xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision e3514747256465c52c3b2aedc9795f52c0d3efe9)
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  * 3. 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(nmp, 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 	struct nfsclsession *tsep;
388 
389 	dp = *dpp;
390 	*dpp = NULL;
391 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL);
392 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
393 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
394 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
395 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
396 	tsep = nfsmnt_mdssession(nmp);
397 	*tl++ = tsep->nfsess_clientid.lval[0];
398 	*tl = tsep->nfsess_clientid.lval[1];
399 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
400 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
401 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
402 	if (reclaim) {
403 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
404 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
405 		*tl = txdr_unsigned(delegtype);
406 	} else {
407 		if (dp != NULL) {
408 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
409 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
410 			if (NFSHASNFSV4N(nmp))
411 				*tl++ = 0;
412 			else
413 				*tl++ = dp->nfsdl_stateid.seqid;
414 			*tl++ = dp->nfsdl_stateid.other[0];
415 			*tl++ = dp->nfsdl_stateid.other[1];
416 			*tl = dp->nfsdl_stateid.other[2];
417 		} else {
418 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
419 		}
420 		(void) nfsm_strtom(nd, name, namelen);
421 	}
422 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
423 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
424 	NFSZERO_ATTRBIT(&attrbits);
425 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
426 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
427 	(void) nfsrv_putattrbit(nd, &attrbits);
428 	if (syscred)
429 		nd->nd_flag |= ND_USEGSSNAME;
430 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
431 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
432 	if (error)
433 		return (error);
434 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
435 	if (!nd->nd_repstat) {
436 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
437 		    6 * NFSX_UNSIGNED);
438 		op->nfso_stateid.seqid = *tl++;
439 		op->nfso_stateid.other[0] = *tl++;
440 		op->nfso_stateid.other[1] = *tl++;
441 		op->nfso_stateid.other[2] = *tl;
442 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
443 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
444 		if (error)
445 			goto nfsmout;
446 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
447 		deleg = fxdr_unsigned(u_int32_t, *tl);
448 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
449 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
450 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
451 			      NFSCLFLAGS_FIRSTDELEG))
452 				op->nfso_own->nfsow_clp->nfsc_flags |=
453 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
454 			MALLOC(ndp, struct nfscldeleg *,
455 			    sizeof (struct nfscldeleg) + newfhlen,
456 			    M_NFSCLDELEG, M_WAITOK);
457 			LIST_INIT(&ndp->nfsdl_owner);
458 			LIST_INIT(&ndp->nfsdl_lock);
459 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
460 			ndp->nfsdl_fhlen = newfhlen;
461 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
462 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
463 			nfscl_lockinit(&ndp->nfsdl_rwlock);
464 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
465 			    NFSX_UNSIGNED);
466 			ndp->nfsdl_stateid.seqid = *tl++;
467 			ndp->nfsdl_stateid.other[0] = *tl++;
468 			ndp->nfsdl_stateid.other[1] = *tl++;
469 			ndp->nfsdl_stateid.other[2] = *tl++;
470 			ret = fxdr_unsigned(int, *tl);
471 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
472 				ndp->nfsdl_flags = NFSCLDL_WRITE;
473 				/*
474 				 * Indicates how much the file can grow.
475 				 */
476 				NFSM_DISSECT(tl, u_int32_t *,
477 				    3 * NFSX_UNSIGNED);
478 				limitby = fxdr_unsigned(int, *tl++);
479 				switch (limitby) {
480 				case NFSV4OPEN_LIMITSIZE:
481 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
482 					break;
483 				case NFSV4OPEN_LIMITBLOCKS:
484 					ndp->nfsdl_sizelimit =
485 					    fxdr_unsigned(u_int64_t, *tl++);
486 					ndp->nfsdl_sizelimit *=
487 					    fxdr_unsigned(u_int64_t, *tl);
488 					break;
489 				default:
490 					error = NFSERR_BADXDR;
491 					goto nfsmout;
492 				}
493 			} else {
494 				ndp->nfsdl_flags = NFSCLDL_READ;
495 			}
496 			if (ret)
497 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
498 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
499 			    &acesize, p);
500 			if (error)
501 				goto nfsmout;
502 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
503 			error = NFSERR_BADXDR;
504 			goto nfsmout;
505 		}
506 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
507 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
508 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
509 		    NULL, NULL, NULL, p, cred);
510 		if (error)
511 			goto nfsmout;
512 		if (ndp != NULL) {
513 			ndp->nfsdl_change = nfsva.na_filerev;
514 			ndp->nfsdl_modtime = nfsva.na_mtime;
515 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
516 		}
517 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
518 		    do {
519 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
520 			    cred, p);
521 			if (ret == NFSERR_DELAY)
522 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
523 		    } while (ret == NFSERR_DELAY);
524 		    error = ret;
525 		}
526 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
527 		    nfscl_assumeposixlocks)
528 		    op->nfso_posixlock = 1;
529 		else
530 		    op->nfso_posixlock = 0;
531 
532 		/*
533 		 * If the server is handing out delegations, but we didn't
534 		 * get one because an OpenConfirm was required, try the
535 		 * Open again, to get a delegation. This is a harmless no-op,
536 		 * from a server's point of view.
537 		 */
538 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
539 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
540 		    && !error && dp == NULL && ndp == NULL && !recursed) {
541 		    do {
542 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
543 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
544 			    cred, p, syscred, 1);
545 			if (ret == NFSERR_DELAY)
546 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
547 		    } while (ret == NFSERR_DELAY);
548 		    if (ret) {
549 			if (ndp != NULL) {
550 				FREE((caddr_t)ndp, M_NFSCLDELEG);
551 				ndp = NULL;
552 			}
553 			if (ret == NFSERR_STALECLIENTID ||
554 			    ret == NFSERR_STALEDONTRECOVER ||
555 			    ret == NFSERR_BADSESSION)
556 				error = ret;
557 		    }
558 		}
559 	}
560 	if (nd->nd_repstat != 0 && error == 0)
561 		error = nd->nd_repstat;
562 	if (error == NFSERR_STALECLIENTID)
563 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
564 nfsmout:
565 	if (!error)
566 		*dpp = ndp;
567 	else if (ndp != NULL)
568 		FREE((caddr_t)ndp, M_NFSCLDELEG);
569 	mbuf_freem(nd->nd_mrep);
570 	return (error);
571 }
572 
573 /*
574  * open downgrade rpc
575  */
576 APPLESTATIC int
577 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
578     struct ucred *cred, NFSPROC_T *p)
579 {
580 	u_int32_t *tl;
581 	struct nfsrv_descript nfsd, *nd = &nfsd;
582 	int error;
583 
584 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
585 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
586 	if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
587 		*tl++ = 0;
588 	else
589 		*tl++ = op->nfso_stateid.seqid;
590 	*tl++ = op->nfso_stateid.other[0];
591 	*tl++ = op->nfso_stateid.other[1];
592 	*tl++ = op->nfso_stateid.other[2];
593 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
594 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
595 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
596 	error = nfscl_request(nd, vp, p, cred, NULL);
597 	if (error)
598 		return (error);
599 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
600 	if (!nd->nd_repstat) {
601 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
602 		op->nfso_stateid.seqid = *tl++;
603 		op->nfso_stateid.other[0] = *tl++;
604 		op->nfso_stateid.other[1] = *tl++;
605 		op->nfso_stateid.other[2] = *tl;
606 	}
607 	if (nd->nd_repstat && error == 0)
608 		error = nd->nd_repstat;
609 	if (error == NFSERR_STALESTATEID)
610 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
611 nfsmout:
612 	mbuf_freem(nd->nd_mrep);
613 	return (error);
614 }
615 
616 /*
617  * V4 Close operation.
618  */
619 APPLESTATIC int
620 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
621 {
622 	struct nfsclclient *clp;
623 	int error;
624 
625 	if (vnode_vtype(vp) != VREG)
626 		return (0);
627 	if (doclose)
628 		error = nfscl_doclose(vp, &clp, p);
629 	else
630 		error = nfscl_getclose(vp, &clp);
631 	if (error)
632 		return (error);
633 
634 	nfscl_clientrelease(clp);
635 	return (0);
636 }
637 
638 /*
639  * Close the open.
640  */
641 APPLESTATIC void
642 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
643 {
644 	struct nfsrv_descript nfsd, *nd = &nfsd;
645 	struct nfscllockowner *lp, *nlp;
646 	struct nfscllock *lop, *nlop;
647 	struct ucred *tcred;
648 	u_int64_t off = 0, len = 0;
649 	u_int32_t type = NFSV4LOCKT_READ;
650 	int error, do_unlock, trycnt;
651 
652 	tcred = newnfs_getcred();
653 	newnfs_copycred(&op->nfso_cred, tcred);
654 	/*
655 	 * (Theoretically this could be done in the same
656 	 *  compound as the close, but having multiple
657 	 *  sequenced Ops in the same compound might be
658 	 *  too scary for some servers.)
659 	 */
660 	if (op->nfso_posixlock) {
661 		off = 0;
662 		len = NFS64BITSSET;
663 		type = NFSV4LOCKT_READ;
664 	}
665 
666 	/*
667 	 * Since this function is only called from VOP_INACTIVE(), no
668 	 * other thread will be manipulating this Open. As such, the
669 	 * lock lists are not being changed by other threads, so it should
670 	 * be safe to do this without locking.
671 	 */
672 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
673 		do_unlock = 1;
674 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
675 			if (op->nfso_posixlock == 0) {
676 				off = lop->nfslo_first;
677 				len = lop->nfslo_end - lop->nfslo_first;
678 				if (lop->nfslo_type == F_WRLCK)
679 					type = NFSV4LOCKT_WRITE;
680 				else
681 					type = NFSV4LOCKT_READ;
682 			}
683 			if (do_unlock) {
684 				trycnt = 0;
685 				do {
686 					error = nfsrpc_locku(nd, nmp, lp, off,
687 					    len, type, tcred, p, 0);
688 					if ((nd->nd_repstat == NFSERR_GRACE ||
689 					    nd->nd_repstat == NFSERR_DELAY) &&
690 					    error == 0)
691 						(void) nfs_catnap(PZERO,
692 						    (int)nd->nd_repstat,
693 						    "nfs_close");
694 				} while ((nd->nd_repstat == NFSERR_GRACE ||
695 				    nd->nd_repstat == NFSERR_DELAY) &&
696 				    error == 0 && trycnt++ < 5);
697 				if (op->nfso_posixlock)
698 					do_unlock = 0;
699 			}
700 			nfscl_freelock(lop, 0);
701 		}
702 		/*
703 		 * Do a ReleaseLockOwner.
704 		 * The lock owner name nfsl_owner may be used by other opens for
705 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
706 		 * puts on the wire has the file handle for this file appended
707 		 * to it, so it can be done now.
708 		 */
709 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
710 		    lp->nfsl_open->nfso_fhlen, tcred, p);
711 	}
712 
713 	/*
714 	 * There could be other Opens for different files on the same
715 	 * OpenOwner, so locking is required.
716 	 */
717 	NFSLOCKCLSTATE();
718 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
719 	NFSUNLOCKCLSTATE();
720 	do {
721 		error = nfscl_tryclose(op, tcred, nmp, p);
722 		if (error == NFSERR_GRACE)
723 			(void) nfs_catnap(PZERO, error, "nfs_close");
724 	} while (error == NFSERR_GRACE);
725 	NFSLOCKCLSTATE();
726 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
727 
728 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
729 		nfscl_freelockowner(lp, 0);
730 	nfscl_freeopen(op, 0);
731 	NFSUNLOCKCLSTATE();
732 	NFSFREECRED(tcred);
733 }
734 
735 /*
736  * The actual Close RPC.
737  */
738 APPLESTATIC int
739 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
740     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
741     int syscred)
742 {
743 	u_int32_t *tl;
744 	int error;
745 
746 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
747 	    op->nfso_fhlen, NULL, NULL);
748 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
749 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
750 	if (NFSHASNFSV4N(nmp))
751 		*tl++ = 0;
752 	else
753 		*tl++ = op->nfso_stateid.seqid;
754 	*tl++ = op->nfso_stateid.other[0];
755 	*tl++ = op->nfso_stateid.other[1];
756 	*tl = op->nfso_stateid.other[2];
757 	if (syscred)
758 		nd->nd_flag |= ND_USEGSSNAME;
759 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
760 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
761 	if (error)
762 		return (error);
763 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
764 	if (nd->nd_repstat == 0)
765 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
766 	error = nd->nd_repstat;
767 	if (error == NFSERR_STALESTATEID)
768 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
769 nfsmout:
770 	mbuf_freem(nd->nd_mrep);
771 	return (error);
772 }
773 
774 /*
775  * V4 Open Confirm RPC.
776  */
777 APPLESTATIC int
778 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
779     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
780 {
781 	u_int32_t *tl;
782 	struct nfsrv_descript nfsd, *nd = &nfsd;
783 	struct nfsmount *nmp;
784 	int error;
785 
786 	nmp = VFSTONFS(vnode_mount(vp));
787 	if (NFSHASNFSV4N(nmp))
788 		return (0);		/* No confirmation for NFSv4.1. */
789 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL);
790 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
791 	*tl++ = op->nfso_stateid.seqid;
792 	*tl++ = op->nfso_stateid.other[0];
793 	*tl++ = op->nfso_stateid.other[1];
794 	*tl++ = op->nfso_stateid.other[2];
795 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
796 	error = nfscl_request(nd, vp, p, cred, NULL);
797 	if (error)
798 		return (error);
799 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
800 	if (!nd->nd_repstat) {
801 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
802 		op->nfso_stateid.seqid = *tl++;
803 		op->nfso_stateid.other[0] = *tl++;
804 		op->nfso_stateid.other[1] = *tl++;
805 		op->nfso_stateid.other[2] = *tl;
806 	}
807 	error = nd->nd_repstat;
808 	if (error == NFSERR_STALESTATEID)
809 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
810 nfsmout:
811 	mbuf_freem(nd->nd_mrep);
812 	return (error);
813 }
814 
815 /*
816  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
817  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
818  */
819 APPLESTATIC int
820 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
821     struct ucred *cred, NFSPROC_T *p)
822 {
823 	u_int32_t *tl;
824 	struct nfsrv_descript nfsd;
825 	struct nfsrv_descript *nd = &nfsd;
826 	nfsattrbit_t attrbits;
827 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
828 	u_short port;
829 	int error, isinet6 = 0, callblen;
830 	nfsquad_t confirm;
831 	u_int32_t lease;
832 	static u_int32_t rev = 0;
833 	struct nfsclds *dsp;
834 	struct in6_addr a6;
835 	struct nfsclsession *tsep;
836 
837 	if (nfsboottime.tv_sec == 0)
838 		NFSSETBOOTTIME(nfsboottime);
839 	clp->nfsc_rev = rev++;
840 	if (NFSHASNFSV4N(nmp)) {
841 		/*
842 		 * Either there was no previous session or the
843 		 * previous session has failed, so...
844 		 * do an ExchangeID followed by the CreateSession.
845 		 */
846 		error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq,
847 		    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp, cred, p);
848 		NFSCL_DEBUG(1, "aft exch=%d\n", error);
849 		if (error == 0)
850 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
851 			    &nmp->nm_sockreq,
852 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
853 		if (error == 0) {
854 			NFSLOCKMNT(nmp);
855 			/*
856 			 * The old sessions cannot be safely free'd
857 			 * here, since they may still be used by
858 			 * in-progress RPCs.
859 			 */
860 			tsep = NULL;
861 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
862 				tsep = NFSMNT_MDSSESSION(nmp);
863 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
864 			    nfsclds_list);
865 			/*
866 			 * Wake up RPCs waiting for a slot on the
867 			 * old session. These will then fail with
868 			 * NFSERR_BADSESSION and be retried with the
869 			 * new session by nfsv4_setsequence().
870 			 * Also wakeup() processes waiting for the
871 			 * new session.
872 			 */
873 			if (tsep != NULL)
874 				wakeup(&tsep->nfsess_slots);
875 			wakeup(&nmp->nm_sess);
876 			NFSUNLOCKMNT(nmp);
877 		} else
878 			nfscl_freenfsclds(dsp);
879 		NFSCL_DEBUG(1, "aft createsess=%d\n", error);
880 		if (error == 0 && reclaim == 0) {
881 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
882 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
883 			if (error == NFSERR_COMPLETEALREADY ||
884 			    error == NFSERR_NOTSUPP)
885 				/* Ignore this error. */
886 				error = 0;
887 		}
888 		return (error);
889 	}
890 
891 	/*
892 	 * Allocate a single session structure for NFSv4.0, because some of
893 	 * the fields are used by NFSv4.0 although it doesn't do a session.
894 	 */
895 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
896 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
897 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
898 	NFSLOCKMNT(nmp);
899 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
900 	tsep = NFSMNT_MDSSESSION(nmp);
901 	NFSUNLOCKMNT(nmp);
902 
903 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL);
904 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
905 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
906 	*tl = txdr_unsigned(clp->nfsc_rev);
907 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
908 
909 	/*
910 	 * set up the callback address
911 	 */
912 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
913 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
914 	callblen = strlen(nfsv4_callbackaddr);
915 	if (callblen == 0)
916 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
917 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
918 	    (callblen > 0 || cp != NULL)) {
919 		port = htons(nfsv4_cbport);
920 		cp2 = (u_int8_t *)&port;
921 #ifdef INET6
922 		if ((callblen > 0 &&
923 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
924 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
925 
926 			(void) nfsm_strtom(nd, "tcp6", 4);
927 			if (callblen == 0) {
928 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
929 				ip6add = ip6buf;
930 			} else {
931 				ip6add = nfsv4_callbackaddr;
932 			}
933 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
934 			    ip6add, cp2[0], cp2[1]);
935 		} else
936 #endif
937 		{
938 			(void) nfsm_strtom(nd, "tcp", 3);
939 			if (callblen == 0)
940 				snprintf(addr, INET6_ADDRSTRLEN + 9,
941 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
942 				    cp[2], cp[3], cp2[0], cp2[1]);
943 			else
944 				snprintf(addr, INET6_ADDRSTRLEN + 9,
945 				    "%s.%d.%d", nfsv4_callbackaddr,
946 				    cp2[0], cp2[1]);
947 		}
948 		(void) nfsm_strtom(nd, addr, strlen(addr));
949 	} else {
950 		(void) nfsm_strtom(nd, "tcp", 3);
951 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
952 	}
953 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
954 	*tl = txdr_unsigned(clp->nfsc_cbident);
955 	nd->nd_flag |= ND_USEGSSNAME;
956 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
957 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
958 	if (error)
959 		return (error);
960 	if (nd->nd_repstat == 0) {
961 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
962 	    tsep->nfsess_clientid.lval[0] = *tl++;
963 	    tsep->nfsess_clientid.lval[1] = *tl++;
964 	    confirm.lval[0] = *tl++;
965 	    confirm.lval[1] = *tl;
966 	    mbuf_freem(nd->nd_mrep);
967 	    nd->nd_mrep = NULL;
968 
969 	    /*
970 	     * and confirm it.
971 	     */
972 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
973 		NULL);
974 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
975 	    *tl++ = tsep->nfsess_clientid.lval[0];
976 	    *tl++ = tsep->nfsess_clientid.lval[1];
977 	    *tl++ = confirm.lval[0];
978 	    *tl = confirm.lval[1];
979 	    nd->nd_flag |= ND_USEGSSNAME;
980 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
981 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
982 	    if (error)
983 		return (error);
984 	    mbuf_freem(nd->nd_mrep);
985 	    nd->nd_mrep = NULL;
986 	    if (nd->nd_repstat == 0) {
987 		nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
988 		    nmp->nm_fhsize, NULL, NULL);
989 		NFSZERO_ATTRBIT(&attrbits);
990 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
991 		(void) nfsrv_putattrbit(nd, &attrbits);
992 		nd->nd_flag |= ND_USEGSSNAME;
993 		error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
994 		    cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
995 		if (error)
996 		    return (error);
997 		if (nd->nd_repstat == 0) {
998 		    error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
999 			NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1000 		    if (error)
1001 			goto nfsmout;
1002 		    clp->nfsc_renew = NFSCL_RENEW(lease);
1003 		    clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1004 		    clp->nfsc_clientidrev++;
1005 		    if (clp->nfsc_clientidrev == 0)
1006 			clp->nfsc_clientidrev++;
1007 		}
1008 	    }
1009 	}
1010 	error = nd->nd_repstat;
1011 nfsmout:
1012 	mbuf_freem(nd->nd_mrep);
1013 	return (error);
1014 }
1015 
1016 /*
1017  * nfs getattr call.
1018  */
1019 APPLESTATIC int
1020 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1021     struct nfsvattr *nap, void *stuff)
1022 {
1023 	struct nfsrv_descript nfsd, *nd = &nfsd;
1024 	int error;
1025 	nfsattrbit_t attrbits;
1026 
1027 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1028 	if (nd->nd_flag & ND_NFSV4) {
1029 		NFSGETATTR_ATTRBIT(&attrbits);
1030 		(void) nfsrv_putattrbit(nd, &attrbits);
1031 	}
1032 	error = nfscl_request(nd, vp, p, cred, stuff);
1033 	if (error)
1034 		return (error);
1035 	if (!nd->nd_repstat)
1036 		error = nfsm_loadattr(nd, nap);
1037 	else
1038 		error = nd->nd_repstat;
1039 	mbuf_freem(nd->nd_mrep);
1040 	return (error);
1041 }
1042 
1043 /*
1044  * nfs getattr call with non-vnode arguemnts.
1045  */
1046 APPLESTATIC int
1047 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1048     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1049     uint32_t *leasep)
1050 {
1051 	struct nfsrv_descript nfsd, *nd = &nfsd;
1052 	int error, vers = NFS_VER2;
1053 	nfsattrbit_t attrbits;
1054 
1055 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL);
1056 	if (nd->nd_flag & ND_NFSV4) {
1057 		vers = NFS_VER4;
1058 		NFSGETATTR_ATTRBIT(&attrbits);
1059 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1060 		(void) nfsrv_putattrbit(nd, &attrbits);
1061 	} else if (nd->nd_flag & ND_NFSV3) {
1062 		vers = NFS_VER3;
1063 	}
1064 	if (syscred)
1065 		nd->nd_flag |= ND_USEGSSNAME;
1066 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1067 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1068 	if (error)
1069 		return (error);
1070 	if (nd->nd_repstat == 0) {
1071 		if ((nd->nd_flag & ND_NFSV4) != 0)
1072 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1073 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1074 			    NULL, NULL);
1075 		else
1076 			error = nfsm_loadattr(nd, nap);
1077 	} else
1078 		error = nd->nd_repstat;
1079 	mbuf_freem(nd->nd_mrep);
1080 	return (error);
1081 }
1082 
1083 /*
1084  * Do an nfs setattr operation.
1085  */
1086 APPLESTATIC int
1087 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1088     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1089     void *stuff)
1090 {
1091 	int error, expireret = 0, openerr, retrycnt;
1092 	u_int32_t clidrev = 0, mode;
1093 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1094 	struct nfsfh *nfhp;
1095 	nfsv4stateid_t stateid;
1096 	void *lckp;
1097 
1098 	if (nmp->nm_clp != NULL)
1099 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1100 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1101 		mode = NFSV4OPEN_ACCESSWRITE;
1102 	else
1103 		mode = NFSV4OPEN_ACCESSREAD;
1104 	retrycnt = 0;
1105 	do {
1106 		lckp = NULL;
1107 		openerr = 1;
1108 		if (NFSHASNFSV4(nmp)) {
1109 			nfhp = VTONFS(vp)->n_fhp;
1110 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1111 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1112 			if (error && vnode_vtype(vp) == VREG &&
1113 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1114 			     nfstest_openallsetattr)) {
1115 				/*
1116 				 * No Open stateid, so try and open the file
1117 				 * now.
1118 				 */
1119 				if (mode == NFSV4OPEN_ACCESSWRITE)
1120 					openerr = nfsrpc_open(vp, FWRITE, cred,
1121 					    p);
1122 				else
1123 					openerr = nfsrpc_open(vp, FREAD, cred,
1124 					    p);
1125 				if (!openerr)
1126 					(void) nfscl_getstateid(vp,
1127 					    nfhp->nfh_fh, nfhp->nfh_len,
1128 					    mode, 0, cred, p, &stateid, &lckp);
1129 			}
1130 		}
1131 		if (vap != NULL)
1132 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1133 			    rnap, attrflagp, stuff);
1134 		else
1135 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1136 			    stuff);
1137 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1138 			NFSLOCKMNT(nmp);
1139 			nmp->nm_state |= NFSSTA_OPENMODE;
1140 			NFSUNLOCKMNT(nmp);
1141 		}
1142 		if (error == NFSERR_STALESTATEID)
1143 			nfscl_initiate_recovery(nmp->nm_clp);
1144 		if (lckp != NULL)
1145 			nfscl_lockderef(lckp);
1146 		if (!openerr)
1147 			(void) nfsrpc_close(vp, 0, p);
1148 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1149 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1150 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1151 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1152 		} else if ((error == NFSERR_EXPIRED ||
1153 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1154 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1155 		}
1156 		retrycnt++;
1157 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1158 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1159 	    error == NFSERR_BADSESSION ||
1160 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1161 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1162 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1163 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1164 	     retrycnt < 4));
1165 	if (error && retrycnt >= 4)
1166 		error = EIO;
1167 	return (error);
1168 }
1169 
1170 static int
1171 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1172     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1173     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1174 {
1175 	u_int32_t *tl;
1176 	struct nfsrv_descript nfsd, *nd = &nfsd;
1177 	int error;
1178 	nfsattrbit_t attrbits;
1179 
1180 	*attrflagp = 0;
1181 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1182 	if (nd->nd_flag & ND_NFSV4)
1183 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1184 	vap->va_type = vnode_vtype(vp);
1185 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1186 	if (nd->nd_flag & ND_NFSV3) {
1187 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1188 		*tl = newnfs_false;
1189 	} else if (nd->nd_flag & ND_NFSV4) {
1190 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1191 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1192 		NFSGETATTR_ATTRBIT(&attrbits);
1193 		(void) nfsrv_putattrbit(nd, &attrbits);
1194 	}
1195 	error = nfscl_request(nd, vp, p, cred, stuff);
1196 	if (error)
1197 		return (error);
1198 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1199 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1200 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1201 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1202 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1203 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1204 	mbuf_freem(nd->nd_mrep);
1205 	if (nd->nd_repstat && !error)
1206 		error = nd->nd_repstat;
1207 	return (error);
1208 }
1209 
1210 /*
1211  * nfs lookup rpc
1212  */
1213 APPLESTATIC int
1214 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1215     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1216     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1217 {
1218 	u_int32_t *tl;
1219 	struct nfsrv_descript nfsd, *nd = &nfsd;
1220 	struct nfsmount *nmp;
1221 	struct nfsnode *np;
1222 	struct nfsfh *nfhp;
1223 	nfsattrbit_t attrbits;
1224 	int error = 0, lookupp = 0;
1225 
1226 	*attrflagp = 0;
1227 	*dattrflagp = 0;
1228 	if (vnode_vtype(dvp) != VDIR)
1229 		return (ENOTDIR);
1230 	nmp = VFSTONFS(vnode_mount(dvp));
1231 	if (len > NFS_MAXNAMLEN)
1232 		return (ENAMETOOLONG);
1233 	if (NFSHASNFSV4(nmp) && len == 1 &&
1234 		name[0] == '.') {
1235 		/*
1236 		 * Just return the current dir's fh.
1237 		 */
1238 		np = VTONFS(dvp);
1239 		MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1240 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1241 		nfhp->nfh_len = np->n_fhp->nfh_len;
1242 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1243 		*nfhpp = nfhp;
1244 		return (0);
1245 	}
1246 	if (NFSHASNFSV4(nmp) && len == 2 &&
1247 		name[0] == '.' && name[1] == '.') {
1248 		lookupp = 1;
1249 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1250 	} else {
1251 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1252 		(void) nfsm_strtom(nd, name, len);
1253 	}
1254 	if (nd->nd_flag & ND_NFSV4) {
1255 		NFSGETATTR_ATTRBIT(&attrbits);
1256 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1257 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1258 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1259 		(void) nfsrv_putattrbit(nd, &attrbits);
1260 	}
1261 	error = nfscl_request(nd, dvp, p, cred, stuff);
1262 	if (error)
1263 		return (error);
1264 	if (nd->nd_repstat) {
1265 		/*
1266 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1267 		 * the lookup is at the root of an fs, so return this dir.
1268 		 */
1269 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1270 		    np = VTONFS(dvp);
1271 		    MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) +
1272 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1273 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1274 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1275 		    *nfhpp = nfhp;
1276 		    mbuf_freem(nd->nd_mrep);
1277 		    return (0);
1278 		}
1279 		if (nd->nd_flag & ND_NFSV3)
1280 		    error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1281 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1282 		    ND_NFSV4) {
1283 			/* Load the directory attributes. */
1284 			error = nfsm_loadattr(nd, dnap);
1285 			if (error == 0)
1286 				*dattrflagp = 1;
1287 		}
1288 		goto nfsmout;
1289 	}
1290 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1291 		/* Load the directory attributes. */
1292 		error = nfsm_loadattr(nd, dnap);
1293 		if (error != 0)
1294 			goto nfsmout;
1295 		*dattrflagp = 1;
1296 		/* Skip over the Lookup and GetFH operation status values. */
1297 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1298 	}
1299 	error = nfsm_getfh(nd, nfhpp);
1300 	if (error)
1301 		goto nfsmout;
1302 
1303 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1304 	if ((nd->nd_flag & ND_NFSV3) && !error)
1305 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1306 nfsmout:
1307 	mbuf_freem(nd->nd_mrep);
1308 	if (!error && nd->nd_repstat)
1309 		error = nd->nd_repstat;
1310 	return (error);
1311 }
1312 
1313 /*
1314  * Do a readlink rpc.
1315  */
1316 APPLESTATIC int
1317 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1318     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1319 {
1320 	u_int32_t *tl;
1321 	struct nfsrv_descript nfsd, *nd = &nfsd;
1322 	struct nfsnode *np = VTONFS(vp);
1323 	nfsattrbit_t attrbits;
1324 	int error, len, cangetattr = 1;
1325 
1326 	*attrflagp = 0;
1327 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1328 	if (nd->nd_flag & ND_NFSV4) {
1329 		/*
1330 		 * And do a Getattr op.
1331 		 */
1332 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1333 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1334 		NFSGETATTR_ATTRBIT(&attrbits);
1335 		(void) nfsrv_putattrbit(nd, &attrbits);
1336 	}
1337 	error = nfscl_request(nd, vp, p, cred, stuff);
1338 	if (error)
1339 		return (error);
1340 	if (nd->nd_flag & ND_NFSV3)
1341 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1342 	if (!nd->nd_repstat && !error) {
1343 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1344 		/*
1345 		 * This seems weird to me, but must have been added to
1346 		 * FreeBSD for some reason. The only thing I can think of
1347 		 * is that there was/is some server that replies with
1348 		 * more link data than it should?
1349 		 */
1350 		if (len == NFS_MAXPATHLEN) {
1351 			NFSLOCKNODE(np);
1352 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1353 				len = np->n_size;
1354 				cangetattr = 0;
1355 			}
1356 			NFSUNLOCKNODE(np);
1357 		}
1358 		error = nfsm_mbufuio(nd, uiop, len);
1359 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1360 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1361 	}
1362 	if (nd->nd_repstat && !error)
1363 		error = nd->nd_repstat;
1364 nfsmout:
1365 	mbuf_freem(nd->nd_mrep);
1366 	return (error);
1367 }
1368 
1369 /*
1370  * Read operation.
1371  */
1372 APPLESTATIC int
1373 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1374     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1375 {
1376 	int error, expireret = 0, retrycnt;
1377 	u_int32_t clidrev = 0;
1378 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1379 	struct nfsnode *np = VTONFS(vp);
1380 	struct ucred *newcred;
1381 	struct nfsfh *nfhp = NULL;
1382 	nfsv4stateid_t stateid;
1383 	void *lckp;
1384 
1385 	if (nmp->nm_clp != NULL)
1386 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1387 	newcred = cred;
1388 	if (NFSHASNFSV4(nmp)) {
1389 		nfhp = np->n_fhp;
1390 		newcred = NFSNEWCRED(cred);
1391 	}
1392 	retrycnt = 0;
1393 	do {
1394 		lckp = NULL;
1395 		if (NFSHASNFSV4(nmp))
1396 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1397 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1398 			    &lckp);
1399 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1400 		    attrflagp, stuff);
1401 		if (error == NFSERR_OPENMODE) {
1402 			NFSLOCKMNT(nmp);
1403 			nmp->nm_state |= NFSSTA_OPENMODE;
1404 			NFSUNLOCKMNT(nmp);
1405 		}
1406 		if (error == NFSERR_STALESTATEID)
1407 			nfscl_initiate_recovery(nmp->nm_clp);
1408 		if (lckp != NULL)
1409 			nfscl_lockderef(lckp);
1410 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1411 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1412 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1413 			(void) nfs_catnap(PZERO, error, "nfs_read");
1414 		} else if ((error == NFSERR_EXPIRED ||
1415 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1416 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1417 		}
1418 		retrycnt++;
1419 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1420 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1421 	    error == NFSERR_BADSESSION ||
1422 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1423 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1424 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1425 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1426 	if (error && retrycnt >= 4)
1427 		error = EIO;
1428 	if (NFSHASNFSV4(nmp))
1429 		NFSFREECRED(newcred);
1430 	return (error);
1431 }
1432 
1433 /*
1434  * The actual read RPC.
1435  */
1436 static int
1437 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1438     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1439     int *attrflagp, void *stuff)
1440 {
1441 	u_int32_t *tl;
1442 	int error = 0, len, retlen, tsiz, eof = 0;
1443 	struct nfsrv_descript nfsd;
1444 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1445 	struct nfsrv_descript *nd = &nfsd;
1446 	int rsize;
1447 	off_t tmp_off;
1448 
1449 	*attrflagp = 0;
1450 	tsiz = uio_uio_resid(uiop);
1451 	tmp_off = uiop->uio_offset + tsiz;
1452 	NFSLOCKMNT(nmp);
1453 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1454 		NFSUNLOCKMNT(nmp);
1455 		return (EFBIG);
1456 	}
1457 	rsize = nmp->nm_rsize;
1458 	NFSUNLOCKMNT(nmp);
1459 	nd->nd_mrep = NULL;
1460 	while (tsiz > 0) {
1461 		*attrflagp = 0;
1462 		len = (tsiz > rsize) ? rsize : tsiz;
1463 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1464 		if (nd->nd_flag & ND_NFSV4)
1465 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1466 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1467 		if (nd->nd_flag & ND_NFSV2) {
1468 			*tl++ = txdr_unsigned(uiop->uio_offset);
1469 			*tl++ = txdr_unsigned(len);
1470 			*tl = 0;
1471 		} else {
1472 			txdr_hyper(uiop->uio_offset, tl);
1473 			*(tl + 2) = txdr_unsigned(len);
1474 		}
1475 		/*
1476 		 * Since I can't do a Getattr for NFSv4 for Write, there
1477 		 * doesn't seem any point in doing one here, either.
1478 		 * (See the comment in nfsrpc_writerpc() for more info.)
1479 		 */
1480 		error = nfscl_request(nd, vp, p, cred, stuff);
1481 		if (error)
1482 			return (error);
1483 		if (nd->nd_flag & ND_NFSV3) {
1484 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1485 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1486 			error = nfsm_loadattr(nd, nap);
1487 			if (!error)
1488 				*attrflagp = 1;
1489 		}
1490 		if (nd->nd_repstat || error) {
1491 			if (!error)
1492 				error = nd->nd_repstat;
1493 			goto nfsmout;
1494 		}
1495 		if (nd->nd_flag & ND_NFSV3) {
1496 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1497 			eof = fxdr_unsigned(int, *(tl + 1));
1498 		} else if (nd->nd_flag & ND_NFSV4) {
1499 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1500 			eof = fxdr_unsigned(int, *tl);
1501 		}
1502 		NFSM_STRSIZ(retlen, len);
1503 		error = nfsm_mbufuio(nd, uiop, retlen);
1504 		if (error)
1505 			goto nfsmout;
1506 		mbuf_freem(nd->nd_mrep);
1507 		nd->nd_mrep = NULL;
1508 		tsiz -= retlen;
1509 		if (!(nd->nd_flag & ND_NFSV2)) {
1510 			if (eof || retlen == 0)
1511 				tsiz = 0;
1512 		} else if (retlen < len)
1513 			tsiz = 0;
1514 	}
1515 	return (0);
1516 nfsmout:
1517 	if (nd->nd_mrep != NULL)
1518 		mbuf_freem(nd->nd_mrep);
1519 	return (error);
1520 }
1521 
1522 /*
1523  * nfs write operation
1524  * When called_from_strategy != 0, it should return EIO for an error that
1525  * indicates recovery is in progress, so that the buffer will be left
1526  * dirty and be written back to the server later. If it loops around,
1527  * the recovery thread could get stuck waiting for the buffer and recovery
1528  * will then deadlock.
1529  */
1530 APPLESTATIC int
1531 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1532     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1533     void *stuff, int called_from_strategy)
1534 {
1535 	int error, expireret = 0, retrycnt, nostateid;
1536 	u_int32_t clidrev = 0;
1537 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1538 	struct nfsnode *np = VTONFS(vp);
1539 	struct ucred *newcred;
1540 	struct nfsfh *nfhp = NULL;
1541 	nfsv4stateid_t stateid;
1542 	void *lckp;
1543 
1544 	*must_commit = 0;
1545 	if (nmp->nm_clp != NULL)
1546 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1547 	newcred = cred;
1548 	if (NFSHASNFSV4(nmp)) {
1549 		newcred = NFSNEWCRED(cred);
1550 		nfhp = np->n_fhp;
1551 	}
1552 	retrycnt = 0;
1553 	do {
1554 		lckp = NULL;
1555 		nostateid = 0;
1556 		if (NFSHASNFSV4(nmp)) {
1557 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1558 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1559 			    &lckp);
1560 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1561 			    stateid.other[2] == 0) {
1562 				nostateid = 1;
1563 				NFSCL_DEBUG(1, "stateid0 in write\n");
1564 			}
1565 		}
1566 
1567 		/*
1568 		 * If there is no stateid for NFSv4, it means this is an
1569 		 * extraneous write after close. Basically a poorly
1570 		 * implemented buffer cache. Just don't do the write.
1571 		 */
1572 		if (nostateid)
1573 			error = 0;
1574 		else
1575 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1576 			    newcred, &stateid, p, nap, attrflagp, stuff);
1577 		if (error == NFSERR_STALESTATEID)
1578 			nfscl_initiate_recovery(nmp->nm_clp);
1579 		if (lckp != NULL)
1580 			nfscl_lockderef(lckp);
1581 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1582 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1583 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1584 			(void) nfs_catnap(PZERO, error, "nfs_write");
1585 		} else if ((error == NFSERR_EXPIRED ||
1586 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1587 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1588 		}
1589 		retrycnt++;
1590 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1591 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1592 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1593 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1594 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1595 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1596 	if (error != 0 && (retrycnt >= 4 ||
1597 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1598 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1599 		error = EIO;
1600 	if (NFSHASNFSV4(nmp))
1601 		NFSFREECRED(newcred);
1602 	return (error);
1603 }
1604 
1605 /*
1606  * The actual write RPC.
1607  */
1608 static int
1609 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1610     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1611     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1612 {
1613 	u_int32_t *tl;
1614 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1615 	struct nfsnode *np = VTONFS(vp);
1616 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1617 	int wccflag = 0, wsize;
1618 	int32_t backup;
1619 	struct nfsrv_descript nfsd;
1620 	struct nfsrv_descript *nd = &nfsd;
1621 	nfsattrbit_t attrbits;
1622 	off_t tmp_off;
1623 
1624 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1625 	*attrflagp = 0;
1626 	tsiz = uio_uio_resid(uiop);
1627 	tmp_off = uiop->uio_offset + tsiz;
1628 	NFSLOCKMNT(nmp);
1629 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1630 		NFSUNLOCKMNT(nmp);
1631 		return (EFBIG);
1632 	}
1633 	wsize = nmp->nm_wsize;
1634 	NFSUNLOCKMNT(nmp);
1635 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1636 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1637 	while (tsiz > 0) {
1638 		*attrflagp = 0;
1639 		len = (tsiz > wsize) ? wsize : tsiz;
1640 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1641 		if (nd->nd_flag & ND_NFSV4) {
1642 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1643 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1644 			txdr_hyper(uiop->uio_offset, tl);
1645 			tl += 2;
1646 			*tl++ = txdr_unsigned(*iomode);
1647 			*tl = txdr_unsigned(len);
1648 		} else if (nd->nd_flag & ND_NFSV3) {
1649 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1650 			txdr_hyper(uiop->uio_offset, tl);
1651 			tl += 2;
1652 			*tl++ = txdr_unsigned(len);
1653 			*tl++ = txdr_unsigned(*iomode);
1654 			*tl = txdr_unsigned(len);
1655 		} else {
1656 			u_int32_t x;
1657 
1658 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1659 			/*
1660 			 * Not sure why someone changed this, since the
1661 			 * RFC clearly states that "beginoffset" and
1662 			 * "totalcount" are ignored, but it wouldn't
1663 			 * surprise me if there's a busted server out there.
1664 			 */
1665 			/* Set both "begin" and "current" to non-garbage. */
1666 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1667 			*tl++ = x;      /* "begin offset" */
1668 			*tl++ = x;      /* "current offset" */
1669 			x = txdr_unsigned(len);
1670 			*tl++ = x;      /* total to this offset */
1671 			*tl = x;        /* size of this write */
1672 
1673 		}
1674 		nfsm_uiombuf(nd, uiop, len);
1675 		/*
1676 		 * Although it is tempting to do a normal Getattr Op in the
1677 		 * NFSv4 compound, the result can be a nearly hung client
1678 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1679 		 * It occurs when the client can't map either the Owner or
1680 		 * Owner_group name in the Getattr reply to a uid/gid. When
1681 		 * there is a cache miss, the kernel does an upcall to the
1682 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1683 		 * or /etc/group file. It can then block in getnewbuf(),
1684 		 * waiting for dirty writes to be pushed to the NFS server.
1685 		 * The only reason this doesn't result in a complete
1686 		 * deadlock, is that the upcall times out and allows
1687 		 * the write to complete. However, progress is so slow
1688 		 * that it might just as well be deadlocked.
1689 		 * As such, we get the rest of the attributes, but not
1690 		 * Owner or Owner_group.
1691 		 * nb: nfscl_loadattrcache() needs to be told that these
1692 		 *     partial attributes from a write rpc are being
1693 		 *     passed in, via a argument flag.
1694 		 */
1695 		if (nd->nd_flag & ND_NFSV4) {
1696 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
1697 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1698 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
1699 			(void) nfsrv_putattrbit(nd, &attrbits);
1700 		}
1701 		error = nfscl_request(nd, vp, p, cred, stuff);
1702 		if (error)
1703 			return (error);
1704 		if (nd->nd_repstat) {
1705 			/*
1706 			 * In case the rpc gets retried, roll
1707 			 * the uio fileds changed by nfsm_uiombuf()
1708 			 * back.
1709 			 */
1710 			uiop->uio_offset -= len;
1711 			uio_uio_resid_add(uiop, len);
1712 			uio_iov_base_add(uiop, -len);
1713 			uio_iov_len_add(uiop, len);
1714 		}
1715 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1716 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1717 			    &wccflag, stuff);
1718 			if (error)
1719 				goto nfsmout;
1720 		}
1721 		if (!nd->nd_repstat) {
1722 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1723 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1724 					+ NFSX_VERF);
1725 				rlen = fxdr_unsigned(int, *tl++);
1726 				if (rlen == 0) {
1727 					error = NFSERR_IO;
1728 					goto nfsmout;
1729 				} else if (rlen < len) {
1730 					backup = len - rlen;
1731 					uio_iov_base_add(uiop, -(backup));
1732 					uio_iov_len_add(uiop, backup);
1733 					uiop->uio_offset -= backup;
1734 					uio_uio_resid_add(uiop, backup);
1735 					len = rlen;
1736 				}
1737 				commit = fxdr_unsigned(int, *tl++);
1738 
1739 				/*
1740 				 * Return the lowest commitment level
1741 				 * obtained by any of the RPCs.
1742 				 */
1743 				if (committed == NFSWRITE_FILESYNC)
1744 					committed = commit;
1745 				else if (committed == NFSWRITE_DATASYNC &&
1746 					commit == NFSWRITE_UNSTABLE)
1747 					committed = commit;
1748 				NFSLOCKMNT(nmp);
1749 				if (!NFSHASWRITEVERF(nmp)) {
1750 					NFSBCOPY((caddr_t)tl,
1751 					    (caddr_t)&nmp->nm_verf[0],
1752 					    NFSX_VERF);
1753 					NFSSETWRITEVERF(nmp);
1754 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1755 				    NFSX_VERF)) {
1756 					*must_commit = 1;
1757 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1758 				}
1759 				NFSUNLOCKMNT(nmp);
1760 			}
1761 			if (nd->nd_flag & ND_NFSV4)
1762 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1763 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1764 				error = nfsm_loadattr(nd, nap);
1765 				if (!error)
1766 					*attrflagp = NFS_LATTR_NOSHRINK;
1767 			}
1768 		} else {
1769 			error = nd->nd_repstat;
1770 		}
1771 		if (error)
1772 			goto nfsmout;
1773 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1774 		mbuf_freem(nd->nd_mrep);
1775 		nd->nd_mrep = NULL;
1776 		tsiz -= len;
1777 	}
1778 nfsmout:
1779 	if (nd->nd_mrep != NULL)
1780 		mbuf_freem(nd->nd_mrep);
1781 	*iomode = committed;
1782 	if (nd->nd_repstat && !error)
1783 		error = nd->nd_repstat;
1784 	return (error);
1785 }
1786 
1787 /*
1788  * nfs mknod rpc
1789  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1790  * mode set to specify the file type and the size field for rdev.
1791  */
1792 APPLESTATIC int
1793 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1794     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1795     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1796     int *attrflagp, int *dattrflagp, void *dstuff)
1797 {
1798 	u_int32_t *tl;
1799 	int error = 0;
1800 	struct nfsrv_descript nfsd, *nd = &nfsd;
1801 	nfsattrbit_t attrbits;
1802 
1803 	*nfhpp = NULL;
1804 	*attrflagp = 0;
1805 	*dattrflagp = 0;
1806 	if (namelen > NFS_MAXNAMLEN)
1807 		return (ENAMETOOLONG);
1808 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1809 	if (nd->nd_flag & ND_NFSV4) {
1810 		if (vtyp == VBLK || vtyp == VCHR) {
1811 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1812 			*tl++ = vtonfsv34_type(vtyp);
1813 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1814 			*tl = txdr_unsigned(NFSMINOR(rdev));
1815 		} else {
1816 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1817 			*tl = vtonfsv34_type(vtyp);
1818 		}
1819 	}
1820 	(void) nfsm_strtom(nd, name, namelen);
1821 	if (nd->nd_flag & ND_NFSV3) {
1822 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1823 		*tl = vtonfsv34_type(vtyp);
1824 	}
1825 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1826 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
1827 	if ((nd->nd_flag & ND_NFSV3) &&
1828 	    (vtyp == VCHR || vtyp == VBLK)) {
1829 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1830 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1831 		*tl = txdr_unsigned(NFSMINOR(rdev));
1832 	}
1833 	if (nd->nd_flag & ND_NFSV4) {
1834 		NFSGETATTR_ATTRBIT(&attrbits);
1835 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1836 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1837 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1838 		(void) nfsrv_putattrbit(nd, &attrbits);
1839 	}
1840 	if (nd->nd_flag & ND_NFSV2)
1841 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
1842 	error = nfscl_request(nd, dvp, p, cred, dstuff);
1843 	if (error)
1844 		return (error);
1845 	if (nd->nd_flag & ND_NFSV4)
1846 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1847 	if (!nd->nd_repstat) {
1848 		if (nd->nd_flag & ND_NFSV4) {
1849 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1850 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1851 			if (error)
1852 				goto nfsmout;
1853 		}
1854 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1855 		if (error)
1856 			goto nfsmout;
1857 	}
1858 	if (nd->nd_flag & ND_NFSV3)
1859 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1860 	if (!error && nd->nd_repstat)
1861 		error = nd->nd_repstat;
1862 nfsmout:
1863 	mbuf_freem(nd->nd_mrep);
1864 	return (error);
1865 }
1866 
1867 /*
1868  * nfs file create call
1869  * Mostly just call the approriate routine. (I separated out v4, so that
1870  * error recovery wouldn't be as difficult.)
1871  */
1872 APPLESTATIC int
1873 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1874     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1875     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1876     int *attrflagp, int *dattrflagp, void *dstuff)
1877 {
1878 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
1879 	struct nfsclowner *owp;
1880 	struct nfscldeleg *dp;
1881 	struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
1882 	u_int32_t clidrev;
1883 
1884 	if (NFSHASNFSV4(nmp)) {
1885 	    retrycnt = 0;
1886 	    do {
1887 		dp = NULL;
1888 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
1889 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
1890 		    NULL, 1);
1891 		if (error)
1892 			return (error);
1893 		if (nmp->nm_clp != NULL)
1894 			clidrev = nmp->nm_clp->nfsc_clientidrev;
1895 		else
1896 			clidrev = 0;
1897 		error = nfsrpc_createv4(dvp, name, namelen, vap, cverf, fmode,
1898 		  owp, &dp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1899 		  dstuff, &unlocked);
1900 		/*
1901 		 * There is no need to invalidate cached attributes here,
1902 		 * since new post-delegation issue attributes are always
1903 		 * returned by nfsrpc_createv4() and these will update the
1904 		 * attribute cache.
1905 		 */
1906 		if (dp != NULL)
1907 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
1908 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
1909 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
1910 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1911 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1912 		    error == NFSERR_BADSESSION) {
1913 			(void) nfs_catnap(PZERO, error, "nfs_open");
1914 		} else if ((error == NFSERR_EXPIRED ||
1915 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1916 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1917 			retrycnt++;
1918 		}
1919 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
1920 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1921 		error == NFSERR_BADSESSION ||
1922 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1923 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
1924 	    if (error && retrycnt >= 4)
1925 		    error = EIO;
1926 	} else {
1927 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
1928 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
1929 		    dstuff);
1930 	}
1931 	return (error);
1932 }
1933 
1934 /*
1935  * The create rpc for v2 and 3.
1936  */
1937 static int
1938 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1939     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
1940     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1941     int *attrflagp, int *dattrflagp, void *dstuff)
1942 {
1943 	u_int32_t *tl;
1944 	int error = 0;
1945 	struct nfsrv_descript nfsd, *nd = &nfsd;
1946 
1947 	*nfhpp = NULL;
1948 	*attrflagp = 0;
1949 	*dattrflagp = 0;
1950 	if (namelen > NFS_MAXNAMLEN)
1951 		return (ENAMETOOLONG);
1952 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
1953 	(void) nfsm_strtom(nd, name, namelen);
1954 	if (nd->nd_flag & ND_NFSV3) {
1955 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1956 		if (fmode & O_EXCL) {
1957 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
1958 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
1959 			*tl++ = cverf.lval[0];
1960 			*tl = cverf.lval[1];
1961 		} else {
1962 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
1963 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
1964 		}
1965 	} else {
1966 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
1967 	}
1968 	error = nfscl_request(nd, dvp, p, cred, dstuff);
1969 	if (error)
1970 		return (error);
1971 	if (nd->nd_repstat == 0) {
1972 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
1973 		if (error)
1974 			goto nfsmout;
1975 	}
1976 	if (nd->nd_flag & ND_NFSV3)
1977 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
1978 	if (nd->nd_repstat != 0 && error == 0)
1979 		error = nd->nd_repstat;
1980 nfsmout:
1981 	mbuf_freem(nd->nd_mrep);
1982 	return (error);
1983 }
1984 
1985 static int
1986 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1987     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
1988     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
1989     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
1990     int *dattrflagp, void *dstuff, int *unlockedp)
1991 {
1992 	u_int32_t *tl;
1993 	int error = 0, deleg, newone, ret, acesize, limitby;
1994 	struct nfsrv_descript nfsd, *nd = &nfsd;
1995 	struct nfsclopen *op;
1996 	struct nfscldeleg *dp = NULL;
1997 	struct nfsnode *np;
1998 	struct nfsfh *nfhp;
1999 	nfsattrbit_t attrbits;
2000 	nfsv4stateid_t stateid;
2001 	u_int32_t rflags;
2002 	struct nfsmount *nmp;
2003 	struct nfsclsession *tsep;
2004 
2005 	nmp = VFSTONFS(dvp->v_mount);
2006 	np = VTONFS(dvp);
2007 	*unlockedp = 0;
2008 	*nfhpp = NULL;
2009 	*dpp = NULL;
2010 	*attrflagp = 0;
2011 	*dattrflagp = 0;
2012 	if (namelen > NFS_MAXNAMLEN)
2013 		return (ENAMETOOLONG);
2014 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2015 	/*
2016 	 * For V4, this is actually an Open op.
2017 	 */
2018 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2019 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2020 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2021 	    NFSV4OPEN_ACCESSREAD);
2022 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2023 	tsep = nfsmnt_mdssession(nmp);
2024 	*tl++ = tsep->nfsess_clientid.lval[0];
2025 	*tl = tsep->nfsess_clientid.lval[1];
2026 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2027 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2028 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2029 	if (fmode & O_EXCL) {
2030 		if (NFSHASNFSV4N(nmp)) {
2031 			if (NFSHASSESSPERSIST(nmp)) {
2032 				/* Use GUARDED for persistent sessions. */
2033 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2034 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2035 			} else {
2036 				/* Otherwise, use EXCLUSIVE4_1. */
2037 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2038 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2039 				*tl++ = cverf.lval[0];
2040 				*tl = cverf.lval[1];
2041 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2042 			}
2043 		} else {
2044 			/* NFSv4.0 */
2045 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2046 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2047 			*tl++ = cverf.lval[0];
2048 			*tl = cverf.lval[1];
2049 		}
2050 	} else {
2051 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2052 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2053 	}
2054 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2055 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2056 	(void) nfsm_strtom(nd, name, namelen);
2057 	/* Get the new file's handle and attributes. */
2058 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2059 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2060 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2061 	NFSGETATTR_ATTRBIT(&attrbits);
2062 	(void) nfsrv_putattrbit(nd, &attrbits);
2063 	/* Get the directory's post-op attributes. */
2064 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2065 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2066 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2067 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2068 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2069 	(void) nfsrv_putattrbit(nd, &attrbits);
2070 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2071 	if (error)
2072 		return (error);
2073 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2074 	if (nd->nd_repstat == 0) {
2075 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2076 		    6 * NFSX_UNSIGNED);
2077 		stateid.seqid = *tl++;
2078 		stateid.other[0] = *tl++;
2079 		stateid.other[1] = *tl++;
2080 		stateid.other[2] = *tl;
2081 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2082 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2083 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2084 		deleg = fxdr_unsigned(int, *tl);
2085 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2086 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2087 			if (!(owp->nfsow_clp->nfsc_flags &
2088 			      NFSCLFLAGS_FIRSTDELEG))
2089 				owp->nfsow_clp->nfsc_flags |=
2090 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2091 			MALLOC(dp, struct nfscldeleg *,
2092 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2093 			    M_NFSCLDELEG, M_WAITOK);
2094 			LIST_INIT(&dp->nfsdl_owner);
2095 			LIST_INIT(&dp->nfsdl_lock);
2096 			dp->nfsdl_clp = owp->nfsow_clp;
2097 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2098 			nfscl_lockinit(&dp->nfsdl_rwlock);
2099 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2100 			    NFSX_UNSIGNED);
2101 			dp->nfsdl_stateid.seqid = *tl++;
2102 			dp->nfsdl_stateid.other[0] = *tl++;
2103 			dp->nfsdl_stateid.other[1] = *tl++;
2104 			dp->nfsdl_stateid.other[2] = *tl++;
2105 			ret = fxdr_unsigned(int, *tl);
2106 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2107 				dp->nfsdl_flags = NFSCLDL_WRITE;
2108 				/*
2109 				 * Indicates how much the file can grow.
2110 				 */
2111 				NFSM_DISSECT(tl, u_int32_t *,
2112 				    3 * NFSX_UNSIGNED);
2113 				limitby = fxdr_unsigned(int, *tl++);
2114 				switch (limitby) {
2115 				case NFSV4OPEN_LIMITSIZE:
2116 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2117 					break;
2118 				case NFSV4OPEN_LIMITBLOCKS:
2119 					dp->nfsdl_sizelimit =
2120 					    fxdr_unsigned(u_int64_t, *tl++);
2121 					dp->nfsdl_sizelimit *=
2122 					    fxdr_unsigned(u_int64_t, *tl);
2123 					break;
2124 				default:
2125 					error = NFSERR_BADXDR;
2126 					goto nfsmout;
2127 				}
2128 			} else {
2129 				dp->nfsdl_flags = NFSCLDL_READ;
2130 			}
2131 			if (ret)
2132 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2133 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2134 			    &acesize, p);
2135 			if (error)
2136 				goto nfsmout;
2137 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2138 			error = NFSERR_BADXDR;
2139 			goto nfsmout;
2140 		}
2141 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2142 		if (error)
2143 			goto nfsmout;
2144 		/* Get rid of the PutFH and Getattr status values. */
2145 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2146 		/* Load the directory attributes. */
2147 		error = nfsm_loadattr(nd, dnap);
2148 		if (error)
2149 			goto nfsmout;
2150 		*dattrflagp = 1;
2151 		if (dp != NULL && *attrflagp) {
2152 			dp->nfsdl_change = nnap->na_filerev;
2153 			dp->nfsdl_modtime = nnap->na_mtime;
2154 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2155 		}
2156 		/*
2157 		 * We can now complete the Open state.
2158 		 */
2159 		nfhp = *nfhpp;
2160 		if (dp != NULL) {
2161 			dp->nfsdl_fhlen = nfhp->nfh_len;
2162 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2163 		}
2164 		/*
2165 		 * Get an Open structure that will be
2166 		 * attached to the OpenOwner, acquired already.
2167 		 */
2168 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2169 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2170 		    cred, p, NULL, &op, &newone, NULL, 0);
2171 		if (error)
2172 			goto nfsmout;
2173 		op->nfso_stateid = stateid;
2174 		newnfs_copyincred(cred, &op->nfso_cred);
2175 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2176 		    do {
2177 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2178 			    nfhp->nfh_len, op, cred, p);
2179 			if (ret == NFSERR_DELAY)
2180 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2181 		    } while (ret == NFSERR_DELAY);
2182 		    error = ret;
2183 		}
2184 
2185 		/*
2186 		 * If the server is handing out delegations, but we didn't
2187 		 * get one because an OpenConfirm was required, try the
2188 		 * Open again, to get a delegation. This is a harmless no-op,
2189 		 * from a server's point of view.
2190 		 */
2191 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2192 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2193 		    !error && dp == NULL) {
2194 		    do {
2195 			ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2196 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2197 			    nfhp->nfh_fh, nfhp->nfh_len,
2198 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2199 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2200 			if (ret == NFSERR_DELAY)
2201 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2202 		    } while (ret == NFSERR_DELAY);
2203 		    if (ret) {
2204 			if (dp != NULL) {
2205 				FREE((caddr_t)dp, M_NFSCLDELEG);
2206 				dp = NULL;
2207 			}
2208 			if (ret == NFSERR_STALECLIENTID ||
2209 			    ret == NFSERR_STALEDONTRECOVER ||
2210 			    ret == NFSERR_BADSESSION)
2211 				error = ret;
2212 		    }
2213 		}
2214 		nfscl_openrelease(nmp, op, error, newone);
2215 		*unlockedp = 1;
2216 	}
2217 	if (nd->nd_repstat != 0 && error == 0)
2218 		error = nd->nd_repstat;
2219 	if (error == NFSERR_STALECLIENTID)
2220 		nfscl_initiate_recovery(owp->nfsow_clp);
2221 nfsmout:
2222 	if (!error)
2223 		*dpp = dp;
2224 	else if (dp != NULL)
2225 		FREE((caddr_t)dp, M_NFSCLDELEG);
2226 	mbuf_freem(nd->nd_mrep);
2227 	return (error);
2228 }
2229 
2230 /*
2231  * Nfs remove rpc
2232  */
2233 APPLESTATIC int
2234 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2235     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2236     void *dstuff)
2237 {
2238 	u_int32_t *tl;
2239 	struct nfsrv_descript nfsd, *nd = &nfsd;
2240 	struct nfsnode *np;
2241 	struct nfsmount *nmp;
2242 	nfsv4stateid_t dstateid;
2243 	int error, ret = 0, i;
2244 
2245 	*dattrflagp = 0;
2246 	if (namelen > NFS_MAXNAMLEN)
2247 		return (ENAMETOOLONG);
2248 	nmp = VFSTONFS(vnode_mount(dvp));
2249 tryagain:
2250 	if (NFSHASNFSV4(nmp) && ret == 0) {
2251 		ret = nfscl_removedeleg(vp, p, &dstateid);
2252 		if (ret == 1) {
2253 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2254 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2255 			    NFSX_UNSIGNED);
2256 			if (NFSHASNFSV4N(nmp))
2257 				*tl++ = 0;
2258 			else
2259 				*tl++ = dstateid.seqid;
2260 			*tl++ = dstateid.other[0];
2261 			*tl++ = dstateid.other[1];
2262 			*tl++ = dstateid.other[2];
2263 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2264 			np = VTONFS(dvp);
2265 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2266 			    np->n_fhp->nfh_len, 0);
2267 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2268 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2269 		}
2270 	} else {
2271 		ret = 0;
2272 	}
2273 	if (ret == 0)
2274 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2275 	(void) nfsm_strtom(nd, name, namelen);
2276 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2277 	if (error)
2278 		return (error);
2279 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2280 		/* For NFSv4, parse out any Delereturn replies. */
2281 		if (ret > 0 && nd->nd_repstat != 0 &&
2282 		    (nd->nd_flag & ND_NOMOREDATA)) {
2283 			/*
2284 			 * If the Delegreturn failed, try again without
2285 			 * it. The server will Recall, as required.
2286 			 */
2287 			mbuf_freem(nd->nd_mrep);
2288 			goto tryagain;
2289 		}
2290 		for (i = 0; i < (ret * 2); i++) {
2291 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2292 			    ND_NFSV4) {
2293 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2294 			    if (*(tl + 1))
2295 				nd->nd_flag |= ND_NOMOREDATA;
2296 			}
2297 		}
2298 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2299 	}
2300 	if (nd->nd_repstat && !error)
2301 		error = nd->nd_repstat;
2302 nfsmout:
2303 	mbuf_freem(nd->nd_mrep);
2304 	return (error);
2305 }
2306 
2307 /*
2308  * Do an nfs rename rpc.
2309  */
2310 APPLESTATIC int
2311 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2312     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2313     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2314     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2315 {
2316 	u_int32_t *tl;
2317 	struct nfsrv_descript nfsd, *nd = &nfsd;
2318 	struct nfsmount *nmp;
2319 	struct nfsnode *np;
2320 	nfsattrbit_t attrbits;
2321 	nfsv4stateid_t fdstateid, tdstateid;
2322 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2323 
2324 	*fattrflagp = 0;
2325 	*tattrflagp = 0;
2326 	nmp = VFSTONFS(vnode_mount(fdvp));
2327 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2328 		return (ENAMETOOLONG);
2329 tryagain:
2330 	if (NFSHASNFSV4(nmp) && ret == 0) {
2331 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2332 		    &tdstateid, &gottd, p);
2333 		if (gotfd && gottd) {
2334 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2335 		} else if (gotfd) {
2336 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2337 		} else if (gottd) {
2338 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2339 		}
2340 		if (gotfd) {
2341 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2342 			if (NFSHASNFSV4N(nmp))
2343 				*tl++ = 0;
2344 			else
2345 				*tl++ = fdstateid.seqid;
2346 			*tl++ = fdstateid.other[0];
2347 			*tl++ = fdstateid.other[1];
2348 			*tl = fdstateid.other[2];
2349 			if (gottd) {
2350 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2351 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2352 				np = VTONFS(tvp);
2353 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2354 				    np->n_fhp->nfh_len, 0);
2355 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2356 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2357 			}
2358 		}
2359 		if (gottd) {
2360 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2361 			if (NFSHASNFSV4N(nmp))
2362 				*tl++ = 0;
2363 			else
2364 				*tl++ = tdstateid.seqid;
2365 			*tl++ = tdstateid.other[0];
2366 			*tl++ = tdstateid.other[1];
2367 			*tl = tdstateid.other[2];
2368 		}
2369 		if (ret > 0) {
2370 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2371 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2372 			np = VTONFS(fdvp);
2373 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2374 			    np->n_fhp->nfh_len, 0);
2375 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2376 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2377 		}
2378 	} else {
2379 		ret = 0;
2380 	}
2381 	if (ret == 0)
2382 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2383 	if (nd->nd_flag & ND_NFSV4) {
2384 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2385 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2386 		NFSWCCATTR_ATTRBIT(&attrbits);
2387 		(void) nfsrv_putattrbit(nd, &attrbits);
2388 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2389 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2390 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2391 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2392 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2393 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2394 		(void) nfsrv_putattrbit(nd, &attrbits);
2395 		nd->nd_flag |= ND_V4WCCATTR;
2396 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2397 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2398 	}
2399 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2400 	if (!(nd->nd_flag & ND_NFSV4))
2401 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2402 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2403 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2404 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2405 	if (error)
2406 		return (error);
2407 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2408 		/* For NFSv4, parse out any Delereturn replies. */
2409 		if (ret > 0 && nd->nd_repstat != 0 &&
2410 		    (nd->nd_flag & ND_NOMOREDATA)) {
2411 			/*
2412 			 * If the Delegreturn failed, try again without
2413 			 * it. The server will Recall, as required.
2414 			 */
2415 			mbuf_freem(nd->nd_mrep);
2416 			goto tryagain;
2417 		}
2418 		for (i = 0; i < (ret * 2); i++) {
2419 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2420 			    ND_NFSV4) {
2421 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2422 			    if (*(tl + 1)) {
2423 				if (i == 0 && ret > 1) {
2424 				    /*
2425 				     * If the Delegreturn failed, try again
2426 				     * without it. The server will Recall, as
2427 				     * required.
2428 				     * If ret > 1, the first iteration of this
2429 				     * loop is the second DelegReturn result.
2430 				     */
2431 				    mbuf_freem(nd->nd_mrep);
2432 				    goto tryagain;
2433 				} else {
2434 				    nd->nd_flag |= ND_NOMOREDATA;
2435 				}
2436 			    }
2437 			}
2438 		}
2439 		/* Now, the first wcc attribute reply. */
2440 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2441 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2442 			if (*(tl + 1))
2443 				nd->nd_flag |= ND_NOMOREDATA;
2444 		}
2445 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2446 		    fstuff);
2447 		/* and the second wcc attribute reply. */
2448 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2449 		    !error) {
2450 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2451 			if (*(tl + 1))
2452 				nd->nd_flag |= ND_NOMOREDATA;
2453 		}
2454 		if (!error)
2455 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2456 			    NULL, tstuff);
2457 	}
2458 	if (nd->nd_repstat && !error)
2459 		error = nd->nd_repstat;
2460 nfsmout:
2461 	mbuf_freem(nd->nd_mrep);
2462 	return (error);
2463 }
2464 
2465 /*
2466  * nfs hard link create rpc
2467  */
2468 APPLESTATIC int
2469 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2470     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2471     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2472 {
2473 	u_int32_t *tl;
2474 	struct nfsrv_descript nfsd, *nd = &nfsd;
2475 	nfsattrbit_t attrbits;
2476 	int error = 0;
2477 
2478 	*attrflagp = 0;
2479 	*dattrflagp = 0;
2480 	if (namelen > NFS_MAXNAMLEN)
2481 		return (ENAMETOOLONG);
2482 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2483 	if (nd->nd_flag & ND_NFSV4) {
2484 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2485 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2486 	}
2487 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2488 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2489 	if (nd->nd_flag & ND_NFSV4) {
2490 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2491 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2492 		NFSWCCATTR_ATTRBIT(&attrbits);
2493 		(void) nfsrv_putattrbit(nd, &attrbits);
2494 		nd->nd_flag |= ND_V4WCCATTR;
2495 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2496 		*tl = txdr_unsigned(NFSV4OP_LINK);
2497 	}
2498 	(void) nfsm_strtom(nd, name, namelen);
2499 	error = nfscl_request(nd, vp, p, cred, dstuff);
2500 	if (error)
2501 		return (error);
2502 	if (nd->nd_flag & ND_NFSV3) {
2503 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2504 		if (!error)
2505 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2506 			    NULL, dstuff);
2507 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2508 		/*
2509 		 * First, parse out the PutFH and Getattr result.
2510 		 */
2511 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2512 		if (!(*(tl + 1)))
2513 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2514 		if (*(tl + 1))
2515 			nd->nd_flag |= ND_NOMOREDATA;
2516 		/*
2517 		 * Get the pre-op attributes.
2518 		 */
2519 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2520 	}
2521 	if (nd->nd_repstat && !error)
2522 		error = nd->nd_repstat;
2523 nfsmout:
2524 	mbuf_freem(nd->nd_mrep);
2525 	return (error);
2526 }
2527 
2528 /*
2529  * nfs symbolic link create rpc
2530  */
2531 APPLESTATIC int
2532 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, char *target,
2533     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2534     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2535     int *dattrflagp, void *dstuff)
2536 {
2537 	u_int32_t *tl;
2538 	struct nfsrv_descript nfsd, *nd = &nfsd;
2539 	struct nfsmount *nmp;
2540 	int slen, error = 0;
2541 
2542 	*nfhpp = NULL;
2543 	*attrflagp = 0;
2544 	*dattrflagp = 0;
2545 	nmp = VFSTONFS(vnode_mount(dvp));
2546 	slen = strlen(target);
2547 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2548 		return (ENAMETOOLONG);
2549 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2550 	if (nd->nd_flag & ND_NFSV4) {
2551 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2552 		*tl = txdr_unsigned(NFLNK);
2553 		(void) nfsm_strtom(nd, target, slen);
2554 	}
2555 	(void) nfsm_strtom(nd, name, namelen);
2556 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2557 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2558 	if (!(nd->nd_flag & ND_NFSV4))
2559 		(void) nfsm_strtom(nd, target, slen);
2560 	if (nd->nd_flag & ND_NFSV2)
2561 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2562 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2563 	if (error)
2564 		return (error);
2565 	if (nd->nd_flag & ND_NFSV4)
2566 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2567 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2568 		if (!nd->nd_repstat)
2569 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2570 		if (!error)
2571 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2572 			    NULL, dstuff);
2573 	}
2574 	if (nd->nd_repstat && !error)
2575 		error = nd->nd_repstat;
2576 	mbuf_freem(nd->nd_mrep);
2577 	/*
2578 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2579 	 * Only do this if vfs.nfs.ignore_eexist is set.
2580 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2581 	 * should guarantee "exactly once" RPC semantics.
2582 	 */
2583 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2584 	    nmp->nm_minorvers == 0))
2585 		error = 0;
2586 	return (error);
2587 }
2588 
2589 /*
2590  * nfs make dir rpc
2591  */
2592 APPLESTATIC int
2593 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2594     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2595     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2596     int *dattrflagp, void *dstuff)
2597 {
2598 	u_int32_t *tl;
2599 	struct nfsrv_descript nfsd, *nd = &nfsd;
2600 	nfsattrbit_t attrbits;
2601 	int error = 0;
2602 	struct nfsfh *fhp;
2603 	struct nfsmount *nmp;
2604 
2605 	*nfhpp = NULL;
2606 	*attrflagp = 0;
2607 	*dattrflagp = 0;
2608 	nmp = VFSTONFS(vnode_mount(dvp));
2609 	fhp = VTONFS(dvp)->n_fhp;
2610 	if (namelen > NFS_MAXNAMLEN)
2611 		return (ENAMETOOLONG);
2612 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2613 	if (nd->nd_flag & ND_NFSV4) {
2614 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2615 		*tl = txdr_unsigned(NFDIR);
2616 	}
2617 	(void) nfsm_strtom(nd, name, namelen);
2618 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2619 	if (nd->nd_flag & ND_NFSV4) {
2620 		NFSGETATTR_ATTRBIT(&attrbits);
2621 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2622 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2623 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2624 		(void) nfsrv_putattrbit(nd, &attrbits);
2625 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2626 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2627 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2628 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2629 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2630 		(void) nfsrv_putattrbit(nd, &attrbits);
2631 	}
2632 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2633 	if (error)
2634 		return (error);
2635 	if (nd->nd_flag & ND_NFSV4)
2636 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2637 	if (!nd->nd_repstat && !error) {
2638 		if (nd->nd_flag & ND_NFSV4) {
2639 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2640 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2641 		}
2642 		if (!error)
2643 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2644 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2645 			/* Get rid of the PutFH and Getattr status values. */
2646 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2647 			/* Load the directory attributes. */
2648 			error = nfsm_loadattr(nd, dnap);
2649 			if (error == 0)
2650 				*dattrflagp = 1;
2651 		}
2652 	}
2653 	if ((nd->nd_flag & ND_NFSV3) && !error)
2654 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2655 	if (nd->nd_repstat && !error)
2656 		error = nd->nd_repstat;
2657 nfsmout:
2658 	mbuf_freem(nd->nd_mrep);
2659 	/*
2660 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2661 	 * Only do this if vfs.nfs.ignore_eexist is set.
2662 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2663 	 * should guarantee "exactly once" RPC semantics.
2664 	 */
2665 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2666 	    nmp->nm_minorvers == 0))
2667 		error = 0;
2668 	return (error);
2669 }
2670 
2671 /*
2672  * nfs remove directory call
2673  */
2674 APPLESTATIC int
2675 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2676     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2677 {
2678 	struct nfsrv_descript nfsd, *nd = &nfsd;
2679 	int error = 0;
2680 
2681 	*dattrflagp = 0;
2682 	if (namelen > NFS_MAXNAMLEN)
2683 		return (ENAMETOOLONG);
2684 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2685 	(void) nfsm_strtom(nd, name, namelen);
2686 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2687 	if (error)
2688 		return (error);
2689 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2690 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2691 	if (nd->nd_repstat && !error)
2692 		error = nd->nd_repstat;
2693 	mbuf_freem(nd->nd_mrep);
2694 	/*
2695 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2696 	 */
2697 	if (error == ENOENT)
2698 		error = 0;
2699 	return (error);
2700 }
2701 
2702 /*
2703  * Readdir rpc.
2704  * Always returns with either uio_resid unchanged, if you are at the
2705  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2706  * filled in.
2707  * I felt this would allow caching of directory blocks more easily
2708  * than returning a pertially filled block.
2709  * Directory offset cookies:
2710  * Oh my, what to do with them...
2711  * I can think of three ways to deal with them:
2712  * 1 - have the layer above these RPCs maintain a map between logical
2713  *     directory byte offsets and the NFS directory offset cookies
2714  * 2 - pass the opaque directory offset cookies up into userland
2715  *     and let the libc functions deal with them, via the system call
2716  * 3 - return them to userland in the "struct dirent", so future versions
2717  *     of libc can use them and do whatever is necessary to make things work
2718  *     above these rpc calls, in the meantime
2719  * For now, I do #3 by "hiding" the directory offset cookies after the
2720  * d_name field in struct dirent. This is space inside d_reclen that
2721  * will be ignored by anything that doesn't know about them.
2722  * The directory offset cookies are filled in as the last 8 bytes of
2723  * each directory entry, after d_name. Someday, the userland libc
2724  * functions may be able to use these. In the meantime, it satisfies
2725  * OpenBSD's requirements for cookies being returned.
2726  * If expects the directory offset cookie for the read to be in uio_offset
2727  * and returns the one for the next entry after this directory block in
2728  * there, as well.
2729  */
2730 APPLESTATIC int
2731 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2732     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2733     int *eofp, void *stuff)
2734 {
2735 	int len, left;
2736 	struct dirent *dp = NULL;
2737 	u_int32_t *tl;
2738 	nfsquad_t cookie, ncookie;
2739 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2740 	struct nfsnode *dnp = VTONFS(vp);
2741 	struct nfsvattr nfsva;
2742 	struct nfsrv_descript nfsd, *nd = &nfsd;
2743 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2744 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2745 	long dotfileid, dotdotfileid = 0;
2746 	u_int32_t fakefileno = 0xffffffff, rderr;
2747 	char *cp;
2748 	nfsattrbit_t attrbits, dattrbits;
2749 	u_int32_t *tl2 = NULL;
2750 	size_t tresid;
2751 
2752 	KASSERT(uiop->uio_iovcnt == 1 &&
2753 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2754 	    ("nfs readdirrpc bad uio"));
2755 
2756 	/*
2757 	 * There is no point in reading a lot more than uio_resid, however
2758 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2759 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2760 	 * will never make readsize > nm_readdirsize.
2761 	 */
2762 	readsize = nmp->nm_readdirsize;
2763 	if (readsize > uio_uio_resid(uiop))
2764 		readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2765 
2766 	*attrflagp = 0;
2767 	if (eofp)
2768 		*eofp = 0;
2769 	tresid = uio_uio_resid(uiop);
2770 	cookie.lval[0] = cookiep->nfsuquad[0];
2771 	cookie.lval[1] = cookiep->nfsuquad[1];
2772 	nd->nd_mrep = NULL;
2773 
2774 	/*
2775 	 * For NFSv4, first create the "." and ".." entries.
2776 	 */
2777 	if (NFSHASNFSV4(nmp)) {
2778 		reqsize = 6 * NFSX_UNSIGNED;
2779 		NFSGETATTR_ATTRBIT(&dattrbits);
2780 		NFSZERO_ATTRBIT(&attrbits);
2781 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2782 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2783 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2784 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2785 			NFSSETBIT_ATTRBIT(&attrbits,
2786 			    NFSATTRBIT_MOUNTEDONFILEID);
2787 			gotmnton = 1;
2788 		} else {
2789 			/*
2790 			 * Must fake it. Use the fileno, except when the
2791 			 * fsid is != to that of the directory. For that
2792 			 * case, generate a fake fileno that is not the same.
2793 			 */
2794 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2795 			gotmnton = 0;
2796 		}
2797 
2798 		/*
2799 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2800 		 */
2801 		if (uiop->uio_offset == 0) {
2802 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2803 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2804 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2805 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2806 			(void) nfsrv_putattrbit(nd, &attrbits);
2807 			error = nfscl_request(nd, vp, p, cred, stuff);
2808 			if (error)
2809 			    return (error);
2810 			dotfileid = 0;	/* Fake out the compiler. */
2811 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2812 			    error = nfsm_loadattr(nd, &nfsva);
2813 			    if (error != 0)
2814 				goto nfsmout;
2815 			    dotfileid = nfsva.na_fileid;
2816 			}
2817 			if (nd->nd_repstat == 0) {
2818 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2819 			    len = fxdr_unsigned(int, *(tl + 4));
2820 			    if (len > 0 && len <= NFSX_V4FHMAX)
2821 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2822 			    else
2823 				error = EPERM;
2824 			    if (!error) {
2825 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2826 				nfsva.na_mntonfileno = 0xffffffff;
2827 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2828 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2829 				    NULL, NULL, NULL, p, cred);
2830 				if (error) {
2831 				    dotdotfileid = dotfileid;
2832 				} else if (gotmnton) {
2833 				    if (nfsva.na_mntonfileno != 0xffffffff)
2834 					dotdotfileid = nfsva.na_mntonfileno;
2835 				    else
2836 					dotdotfileid = nfsva.na_fileid;
2837 				} else if (nfsva.na_filesid[0] ==
2838 				    dnp->n_vattr.na_filesid[0] &&
2839 				    nfsva.na_filesid[1] ==
2840 				    dnp->n_vattr.na_filesid[1]) {
2841 				    dotdotfileid = nfsva.na_fileid;
2842 				} else {
2843 				    do {
2844 					fakefileno--;
2845 				    } while (fakefileno ==
2846 					nfsva.na_fileid);
2847 				    dotdotfileid = fakefileno;
2848 				}
2849 			    }
2850 			} else if (nd->nd_repstat == NFSERR_NOENT) {
2851 			    /*
2852 			     * Lookupp returns NFSERR_NOENT when we are
2853 			     * at the root, so just use the current dir.
2854 			     */
2855 			    nd->nd_repstat = 0;
2856 			    dotdotfileid = dotfileid;
2857 			} else {
2858 			    error = nd->nd_repstat;
2859 			}
2860 			mbuf_freem(nd->nd_mrep);
2861 			if (error)
2862 			    return (error);
2863 			nd->nd_mrep = NULL;
2864 			dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2865 			dp->d_type = DT_DIR;
2866 			dp->d_fileno = dotfileid;
2867 			dp->d_namlen = 1;
2868 			dp->d_name[0] = '.';
2869 			dp->d_name[1] = '\0';
2870 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2871 			/*
2872 			 * Just make these offset cookie 0.
2873 			 */
2874 			tl = (u_int32_t *)&dp->d_name[4];
2875 			*tl++ = 0;
2876 			*tl = 0;
2877 			blksiz += dp->d_reclen;
2878 			uio_uio_resid_add(uiop, -(dp->d_reclen));
2879 			uiop->uio_offset += dp->d_reclen;
2880 			uio_iov_base_add(uiop, dp->d_reclen);
2881 			uio_iov_len_add(uiop, -(dp->d_reclen));
2882 			dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
2883 			dp->d_type = DT_DIR;
2884 			dp->d_fileno = dotdotfileid;
2885 			dp->d_namlen = 2;
2886 			dp->d_name[0] = '.';
2887 			dp->d_name[1] = '.';
2888 			dp->d_name[2] = '\0';
2889 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
2890 			/*
2891 			 * Just make these offset cookie 0.
2892 			 */
2893 			tl = (u_int32_t *)&dp->d_name[4];
2894 			*tl++ = 0;
2895 			*tl = 0;
2896 			blksiz += dp->d_reclen;
2897 			uio_uio_resid_add(uiop, -(dp->d_reclen));
2898 			uiop->uio_offset += dp->d_reclen;
2899 			uio_iov_base_add(uiop, dp->d_reclen);
2900 			uio_iov_len_add(uiop, -(dp->d_reclen));
2901 		}
2902 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
2903 	} else {
2904 		reqsize = 5 * NFSX_UNSIGNED;
2905 	}
2906 
2907 
2908 	/*
2909 	 * Loop around doing readdir rpc's of size readsize.
2910 	 * The stopping criteria is EOF or buffer full.
2911 	 */
2912 	while (more_dirs && bigenough) {
2913 		*attrflagp = 0;
2914 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
2915 		if (nd->nd_flag & ND_NFSV2) {
2916 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2917 			*tl++ = cookie.lval[1];
2918 			*tl = txdr_unsigned(readsize);
2919 		} else {
2920 			NFSM_BUILD(tl, u_int32_t *, reqsize);
2921 			*tl++ = cookie.lval[0];
2922 			*tl++ = cookie.lval[1];
2923 			if (cookie.qval == 0) {
2924 				*tl++ = 0;
2925 				*tl++ = 0;
2926 			} else {
2927 				NFSLOCKNODE(dnp);
2928 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
2929 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
2930 				NFSUNLOCKNODE(dnp);
2931 			}
2932 			if (nd->nd_flag & ND_NFSV4) {
2933 				*tl++ = txdr_unsigned(readsize);
2934 				*tl = txdr_unsigned(readsize);
2935 				(void) nfsrv_putattrbit(nd, &attrbits);
2936 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2937 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
2938 				(void) nfsrv_putattrbit(nd, &dattrbits);
2939 			} else {
2940 				*tl = txdr_unsigned(readsize);
2941 			}
2942 		}
2943 		error = nfscl_request(nd, vp, p, cred, stuff);
2944 		if (error)
2945 			return (error);
2946 		if (!(nd->nd_flag & ND_NFSV2)) {
2947 			if (nd->nd_flag & ND_NFSV3)
2948 				error = nfscl_postop_attr(nd, nap, attrflagp,
2949 				    stuff);
2950 			if (!nd->nd_repstat && !error) {
2951 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
2952 				NFSLOCKNODE(dnp);
2953 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
2954 				dnp->n_cookieverf.nfsuquad[1] = *tl;
2955 				NFSUNLOCKNODE(dnp);
2956 			}
2957 		}
2958 		if (nd->nd_repstat || error) {
2959 			if (!error)
2960 				error = nd->nd_repstat;
2961 			goto nfsmout;
2962 		}
2963 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2964 		more_dirs = fxdr_unsigned(int, *tl);
2965 		if (!more_dirs)
2966 			tryformoredirs = 0;
2967 
2968 		/* loop through the dir entries, doctoring them to 4bsd form */
2969 		while (more_dirs && bigenough) {
2970 			if (nd->nd_flag & ND_NFSV4) {
2971 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2972 				ncookie.lval[0] = *tl++;
2973 				ncookie.lval[1] = *tl++;
2974 				len = fxdr_unsigned(int, *tl);
2975 			} else if (nd->nd_flag & ND_NFSV3) {
2976 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
2977 				nfsva.na_fileid = fxdr_hyper(tl);
2978 				tl += 2;
2979 				len = fxdr_unsigned(int, *tl);
2980 			} else {
2981 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2982 				nfsva.na_fileid =
2983 				    fxdr_unsigned(long, *tl++);
2984 				len = fxdr_unsigned(int, *tl);
2985 			}
2986 			if (len <= 0 || len > NFS_MAXNAMLEN) {
2987 				error = EBADRPC;
2988 				goto nfsmout;
2989 			}
2990 			tlen = NFSM_RNDUP(len);
2991 			if (tlen == len)
2992 				tlen += 4;  /* To ensure null termination */
2993 			left = DIRBLKSIZ - blksiz;
2994 			if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > left) {
2995 				dp->d_reclen += left;
2996 				uio_iov_base_add(uiop, left);
2997 				uio_iov_len_add(uiop, -(left));
2998 				uio_uio_resid_add(uiop, -(left));
2999 				uiop->uio_offset += left;
3000 				blksiz = 0;
3001 			}
3002 			if ((int)(tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
3003 				bigenough = 0;
3004 			if (bigenough) {
3005 				dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
3006 				dp->d_namlen = len;
3007 				dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
3008 				dp->d_type = DT_UNKNOWN;
3009 				blksiz += dp->d_reclen;
3010 				if (blksiz == DIRBLKSIZ)
3011 					blksiz = 0;
3012 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
3013 				uiop->uio_offset += DIRHDSIZ;
3014 				uio_iov_base_add(uiop, DIRHDSIZ);
3015 				uio_iov_len_add(uiop, -(DIRHDSIZ));
3016 				error = nfsm_mbufuio(nd, uiop, len);
3017 				if (error)
3018 					goto nfsmout;
3019 				cp = CAST_DOWN(caddr_t, uio_iov_base(uiop));
3020 				tlen -= len;
3021 				*cp = '\0';	/* null terminate */
3022 				cp += tlen;	/* points to cookie storage */
3023 				tl2 = (u_int32_t *)cp;
3024 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3025 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3026 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3027 				uiop->uio_offset += (tlen + NFSX_HYPER);
3028 			} else {
3029 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3030 				if (error)
3031 					goto nfsmout;
3032 			}
3033 			if (nd->nd_flag & ND_NFSV4) {
3034 				rderr = 0;
3035 				nfsva.na_mntonfileno = 0xffffffff;
3036 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3037 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3038 				    NULL, NULL, &rderr, p, cred);
3039 				if (error)
3040 					goto nfsmout;
3041 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3042 			} else if (nd->nd_flag & ND_NFSV3) {
3043 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3044 				ncookie.lval[0] = *tl++;
3045 				ncookie.lval[1] = *tl++;
3046 			} else {
3047 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3048 				ncookie.lval[0] = 0;
3049 				ncookie.lval[1] = *tl++;
3050 			}
3051 			if (bigenough) {
3052 			    if (nd->nd_flag & ND_NFSV4) {
3053 				if (rderr) {
3054 				    dp->d_fileno = 0;
3055 				} else {
3056 				    if (gotmnton) {
3057 					if (nfsva.na_mntonfileno != 0xffffffff)
3058 					    dp->d_fileno = nfsva.na_mntonfileno;
3059 					else
3060 					    dp->d_fileno = nfsva.na_fileid;
3061 				    } else if (nfsva.na_filesid[0] ==
3062 					dnp->n_vattr.na_filesid[0] &&
3063 					nfsva.na_filesid[1] ==
3064 					dnp->n_vattr.na_filesid[1]) {
3065 					dp->d_fileno = nfsva.na_fileid;
3066 				    } else {
3067 					do {
3068 					    fakefileno--;
3069 					} while (fakefileno ==
3070 					    nfsva.na_fileid);
3071 					dp->d_fileno = fakefileno;
3072 				    }
3073 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3074 				}
3075 			    } else {
3076 				dp->d_fileno = nfsva.na_fileid;
3077 			    }
3078 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3079 				ncookie.lval[0];
3080 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3081 				ncookie.lval[1];
3082 			}
3083 			more_dirs = fxdr_unsigned(int, *tl);
3084 		}
3085 		/*
3086 		 * If at end of rpc data, get the eof boolean
3087 		 */
3088 		if (!more_dirs) {
3089 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3090 			eof = fxdr_unsigned(int, *tl);
3091 			if (tryformoredirs)
3092 				more_dirs = !eof;
3093 			if (nd->nd_flag & ND_NFSV4) {
3094 				error = nfscl_postop_attr(nd, nap, attrflagp,
3095 				    stuff);
3096 				if (error)
3097 					goto nfsmout;
3098 			}
3099 		}
3100 		mbuf_freem(nd->nd_mrep);
3101 		nd->nd_mrep = NULL;
3102 	}
3103 	/*
3104 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3105 	 * by increasing d_reclen for the last record.
3106 	 */
3107 	if (blksiz > 0) {
3108 		left = DIRBLKSIZ - blksiz;
3109 		dp->d_reclen += left;
3110 		uio_iov_base_add(uiop, left);
3111 		uio_iov_len_add(uiop, -(left));
3112 		uio_uio_resid_add(uiop, -(left));
3113 		uiop->uio_offset += left;
3114 	}
3115 
3116 	/*
3117 	 * If returning no data, assume end of file.
3118 	 * If not bigenough, return not end of file, since you aren't
3119 	 *    returning all the data
3120 	 * Otherwise, return the eof flag from the server.
3121 	 */
3122 	if (eofp) {
3123 		if (tresid == ((size_t)(uio_uio_resid(uiop))))
3124 			*eofp = 1;
3125 		else if (!bigenough)
3126 			*eofp = 0;
3127 		else
3128 			*eofp = eof;
3129 	}
3130 
3131 	/*
3132 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3133 	 */
3134 	while (uio_uio_resid(uiop) > 0 && ((size_t)(uio_uio_resid(uiop))) != tresid) {
3135 		dp = (struct dirent *) CAST_DOWN(caddr_t, uio_iov_base(uiop));
3136 		dp->d_type = DT_UNKNOWN;
3137 		dp->d_fileno = 0;
3138 		dp->d_namlen = 0;
3139 		dp->d_name[0] = '\0';
3140 		tl = (u_int32_t *)&dp->d_name[4];
3141 		*tl++ = cookie.lval[0];
3142 		*tl = cookie.lval[1];
3143 		dp->d_reclen = DIRBLKSIZ;
3144 		uio_iov_base_add(uiop, DIRBLKSIZ);
3145 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3146 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3147 		uiop->uio_offset += DIRBLKSIZ;
3148 	}
3149 
3150 nfsmout:
3151 	if (nd->nd_mrep != NULL)
3152 		mbuf_freem(nd->nd_mrep);
3153 	return (error);
3154 }
3155 
3156 #ifndef APPLE
3157 /*
3158  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3159  * (Also used for NFS V4 when mount flag set.)
3160  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3161  */
3162 APPLESTATIC int
3163 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3164     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3165     int *eofp, void *stuff)
3166 {
3167 	int len, left;
3168 	struct dirent *dp = NULL;
3169 	u_int32_t *tl;
3170 	vnode_t newvp = NULLVP;
3171 	struct nfsrv_descript nfsd, *nd = &nfsd;
3172 	struct nameidata nami, *ndp = &nami;
3173 	struct componentname *cnp = &ndp->ni_cnd;
3174 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3175 	struct nfsnode *dnp = VTONFS(vp), *np;
3176 	struct nfsvattr nfsva;
3177 	struct nfsfh *nfhp;
3178 	nfsquad_t cookie, ncookie;
3179 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3180 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3181 	int isdotdot = 0, unlocknewvp = 0;
3182 	long dotfileid, dotdotfileid = 0, fileno = 0;
3183 	char *cp;
3184 	nfsattrbit_t attrbits, dattrbits;
3185 	size_t tresid;
3186 	u_int32_t *tl2 = NULL, fakefileno = 0xffffffff, rderr;
3187 	struct timespec dctime;
3188 
3189 	KASSERT(uiop->uio_iovcnt == 1 &&
3190 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3191 	    ("nfs readdirplusrpc bad uio"));
3192 	timespecclear(&dctime);
3193 	*attrflagp = 0;
3194 	if (eofp != NULL)
3195 		*eofp = 0;
3196 	ndp->ni_dvp = vp;
3197 	nd->nd_mrep = NULL;
3198 	cookie.lval[0] = cookiep->nfsuquad[0];
3199 	cookie.lval[1] = cookiep->nfsuquad[1];
3200 	tresid = uio_uio_resid(uiop);
3201 
3202 	/*
3203 	 * For NFSv4, first create the "." and ".." entries.
3204 	 */
3205 	if (NFSHASNFSV4(nmp)) {
3206 		NFSGETATTR_ATTRBIT(&dattrbits);
3207 		NFSZERO_ATTRBIT(&attrbits);
3208 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3209 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3210 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3211 			NFSSETBIT_ATTRBIT(&attrbits,
3212 			    NFSATTRBIT_MOUNTEDONFILEID);
3213 			gotmnton = 1;
3214 		} else {
3215 			/*
3216 			 * Must fake it. Use the fileno, except when the
3217 			 * fsid is != to that of the directory. For that
3218 			 * case, generate a fake fileno that is not the same.
3219 			 */
3220 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3221 			gotmnton = 0;
3222 		}
3223 
3224 		/*
3225 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3226 		 */
3227 		if (uiop->uio_offset == 0) {
3228 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3229 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3230 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3231 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3232 			(void) nfsrv_putattrbit(nd, &attrbits);
3233 			error = nfscl_request(nd, vp, p, cred, stuff);
3234 			if (error)
3235 			    return (error);
3236 			dotfileid = 0;	/* Fake out the compiler. */
3237 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3238 			    error = nfsm_loadattr(nd, &nfsva);
3239 			    if (error != 0)
3240 				goto nfsmout;
3241 			    dctime = nfsva.na_ctime;
3242 			    dotfileid = nfsva.na_fileid;
3243 			}
3244 			if (nd->nd_repstat == 0) {
3245 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3246 			    len = fxdr_unsigned(int, *(tl + 4));
3247 			    if (len > 0 && len <= NFSX_V4FHMAX)
3248 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3249 			    else
3250 				error = EPERM;
3251 			    if (!error) {
3252 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3253 				nfsva.na_mntonfileno = 0xffffffff;
3254 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3255 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3256 				    NULL, NULL, NULL, p, cred);
3257 				if (error) {
3258 				    dotdotfileid = dotfileid;
3259 				} else if (gotmnton) {
3260 				    if (nfsva.na_mntonfileno != 0xffffffff)
3261 					dotdotfileid = nfsva.na_mntonfileno;
3262 				    else
3263 					dotdotfileid = nfsva.na_fileid;
3264 				} else if (nfsva.na_filesid[0] ==
3265 				    dnp->n_vattr.na_filesid[0] &&
3266 				    nfsva.na_filesid[1] ==
3267 				    dnp->n_vattr.na_filesid[1]) {
3268 				    dotdotfileid = nfsva.na_fileid;
3269 				} else {
3270 				    do {
3271 					fakefileno--;
3272 				    } while (fakefileno ==
3273 					nfsva.na_fileid);
3274 				    dotdotfileid = fakefileno;
3275 				}
3276 			    }
3277 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3278 			    /*
3279 			     * Lookupp returns NFSERR_NOENT when we are
3280 			     * at the root, so just use the current dir.
3281 			     */
3282 			    nd->nd_repstat = 0;
3283 			    dotdotfileid = dotfileid;
3284 			} else {
3285 			    error = nd->nd_repstat;
3286 			}
3287 			mbuf_freem(nd->nd_mrep);
3288 			if (error)
3289 			    return (error);
3290 			nd->nd_mrep = NULL;
3291 			dp = (struct dirent *)uio_iov_base(uiop);
3292 			dp->d_type = DT_DIR;
3293 			dp->d_fileno = dotfileid;
3294 			dp->d_namlen = 1;
3295 			dp->d_name[0] = '.';
3296 			dp->d_name[1] = '\0';
3297 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3298 			/*
3299 			 * Just make these offset cookie 0.
3300 			 */
3301 			tl = (u_int32_t *)&dp->d_name[4];
3302 			*tl++ = 0;
3303 			*tl = 0;
3304 			blksiz += dp->d_reclen;
3305 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3306 			uiop->uio_offset += dp->d_reclen;
3307 			uio_iov_base_add(uiop, dp->d_reclen);
3308 			uio_iov_len_add(uiop, -(dp->d_reclen));
3309 			dp = (struct dirent *)uio_iov_base(uiop);
3310 			dp->d_type = DT_DIR;
3311 			dp->d_fileno = dotdotfileid;
3312 			dp->d_namlen = 2;
3313 			dp->d_name[0] = '.';
3314 			dp->d_name[1] = '.';
3315 			dp->d_name[2] = '\0';
3316 			dp->d_reclen = DIRENT_SIZE(dp) + NFSX_HYPER;
3317 			/*
3318 			 * Just make these offset cookie 0.
3319 			 */
3320 			tl = (u_int32_t *)&dp->d_name[4];
3321 			*tl++ = 0;
3322 			*tl = 0;
3323 			blksiz += dp->d_reclen;
3324 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3325 			uiop->uio_offset += dp->d_reclen;
3326 			uio_iov_base_add(uiop, dp->d_reclen);
3327 			uio_iov_len_add(uiop, -(dp->d_reclen));
3328 		}
3329 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3330 		if (gotmnton)
3331 			NFSSETBIT_ATTRBIT(&attrbits,
3332 			    NFSATTRBIT_MOUNTEDONFILEID);
3333 	}
3334 
3335 	/*
3336 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3337 	 * The stopping criteria is EOF or buffer full.
3338 	 */
3339 	while (more_dirs && bigenough) {
3340 		*attrflagp = 0;
3341 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3342  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3343 		*tl++ = cookie.lval[0];
3344 		*tl++ = cookie.lval[1];
3345 		if (cookie.qval == 0) {
3346 			*tl++ = 0;
3347 			*tl++ = 0;
3348 		} else {
3349 			NFSLOCKNODE(dnp);
3350 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3351 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3352 			NFSUNLOCKNODE(dnp);
3353 		}
3354 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3355 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3356 		if (nd->nd_flag & ND_NFSV4) {
3357 			(void) nfsrv_putattrbit(nd, &attrbits);
3358 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3359 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3360 			(void) nfsrv_putattrbit(nd, &dattrbits);
3361 		}
3362 		error = nfscl_request(nd, vp, p, cred, stuff);
3363 		if (error)
3364 			return (error);
3365 		if (nd->nd_flag & ND_NFSV3)
3366 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3367 		if (nd->nd_repstat || error) {
3368 			if (!error)
3369 				error = nd->nd_repstat;
3370 			goto nfsmout;
3371 		}
3372 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3373 			dctime = nap->na_ctime;
3374 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3375 		NFSLOCKNODE(dnp);
3376 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3377 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3378 		NFSUNLOCKNODE(dnp);
3379 		more_dirs = fxdr_unsigned(int, *tl);
3380 		if (!more_dirs)
3381 			tryformoredirs = 0;
3382 
3383 		/* loop through the dir entries, doctoring them to 4bsd form */
3384 		while (more_dirs && bigenough) {
3385 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3386 			if (nd->nd_flag & ND_NFSV4) {
3387 				ncookie.lval[0] = *tl++;
3388 				ncookie.lval[1] = *tl++;
3389 			} else {
3390 				fileno = fxdr_unsigned(long, *++tl);
3391 				tl++;
3392 			}
3393 			len = fxdr_unsigned(int, *tl);
3394 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3395 				error = EBADRPC;
3396 				goto nfsmout;
3397 			}
3398 			tlen = NFSM_RNDUP(len);
3399 			if (tlen == len)
3400 				tlen += 4;  /* To ensure null termination */
3401 			left = DIRBLKSIZ - blksiz;
3402 			if ((tlen + DIRHDSIZ + NFSX_HYPER) > left) {
3403 				dp->d_reclen += left;
3404 				uio_iov_base_add(uiop, left);
3405 				uio_iov_len_add(uiop, -(left));
3406 				uio_uio_resid_add(uiop, -(left));
3407 				uiop->uio_offset += left;
3408 				blksiz = 0;
3409 			}
3410 			if ((tlen + DIRHDSIZ + NFSX_HYPER) > uio_uio_resid(uiop))
3411 				bigenough = 0;
3412 			if (bigenough) {
3413 				dp = (struct dirent *)uio_iov_base(uiop);
3414 				dp->d_namlen = len;
3415 				dp->d_reclen = tlen + DIRHDSIZ + NFSX_HYPER;
3416 				dp->d_type = DT_UNKNOWN;
3417 				blksiz += dp->d_reclen;
3418 				if (blksiz == DIRBLKSIZ)
3419 					blksiz = 0;
3420 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
3421 				uiop->uio_offset += DIRHDSIZ;
3422 				uio_iov_base_add(uiop, DIRHDSIZ);
3423 				uio_iov_len_add(uiop, -(DIRHDSIZ));
3424 				cnp->cn_nameptr = uio_iov_base(uiop);
3425 				cnp->cn_namelen = len;
3426 				NFSCNHASHZERO(cnp);
3427 				error = nfsm_mbufuio(nd, uiop, len);
3428 				if (error)
3429 					goto nfsmout;
3430 				cp = uio_iov_base(uiop);
3431 				tlen -= len;
3432 				*cp = '\0';
3433 				cp += tlen;	/* points to cookie storage */
3434 				tl2 = (u_int32_t *)cp;
3435 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3436 				    cnp->cn_nameptr[1] == '.')
3437 					isdotdot = 1;
3438 				else
3439 					isdotdot = 0;
3440 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3441 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3442 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3443 				uiop->uio_offset += (tlen + NFSX_HYPER);
3444 			} else {
3445 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3446 				if (error)
3447 					goto nfsmout;
3448 			}
3449 			nfhp = NULL;
3450 			if (nd->nd_flag & ND_NFSV3) {
3451 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3452 				ncookie.lval[0] = *tl++;
3453 				ncookie.lval[1] = *tl++;
3454 				attrflag = fxdr_unsigned(int, *tl);
3455 				if (attrflag) {
3456 				  error = nfsm_loadattr(nd, &nfsva);
3457 				  if (error)
3458 					goto nfsmout;
3459 				}
3460 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3461 				if (*tl) {
3462 					error = nfsm_getfh(nd, &nfhp);
3463 					if (error)
3464 					    goto nfsmout;
3465 				}
3466 				if (!attrflag && nfhp != NULL) {
3467 					FREE((caddr_t)nfhp, M_NFSFH);
3468 					nfhp = NULL;
3469 				}
3470 			} else {
3471 				rderr = 0;
3472 				nfsva.na_mntonfileno = 0xffffffff;
3473 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3474 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3475 				    NULL, NULL, &rderr, p, cred);
3476 				if (error)
3477 					goto nfsmout;
3478 			}
3479 
3480 			if (bigenough) {
3481 			    if (nd->nd_flag & ND_NFSV4) {
3482 				if (rderr) {
3483 				    dp->d_fileno = 0;
3484 				} else if (gotmnton) {
3485 				    if (nfsva.na_mntonfileno != 0xffffffff)
3486 					dp->d_fileno = nfsva.na_mntonfileno;
3487 				    else
3488 					dp->d_fileno = nfsva.na_fileid;
3489 				} else if (nfsva.na_filesid[0] ==
3490 				    dnp->n_vattr.na_filesid[0] &&
3491 				    nfsva.na_filesid[1] ==
3492 				    dnp->n_vattr.na_filesid[1]) {
3493 				    dp->d_fileno = nfsva.na_fileid;
3494 				} else {
3495 				    do {
3496 					fakefileno--;
3497 				    } while (fakefileno ==
3498 					nfsva.na_fileid);
3499 				    dp->d_fileno = fakefileno;
3500 				}
3501 			    } else {
3502 				dp->d_fileno = fileno;
3503 			    }
3504 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3505 				ncookie.lval[0];
3506 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3507 				ncookie.lval[1];
3508 
3509 			    if (nfhp != NULL) {
3510 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3511 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3512 				    VREF(vp);
3513 				    newvp = vp;
3514 				    unlocknewvp = 0;
3515 				    FREE((caddr_t)nfhp, M_NFSFH);
3516 				    np = dnp;
3517 				} else if (isdotdot != 0) {
3518 				    /*
3519 				     * Skip doing a nfscl_nget() call for "..".
3520 				     * There's a race between acquiring the nfs
3521 				     * node here and lookups that look for the
3522 				     * directory being read (in the parent).
3523 				     * It would try to get a lock on ".." here,
3524 				     * owning the lock on the directory being
3525 				     * read. Lookup will hold the lock on ".."
3526 				     * and try to acquire the lock on the
3527 				     * directory being read.
3528 				     * If the directory is unlocked/relocked,
3529 				     * then there is a LOR with the buflock
3530 				     * vp is relocked.
3531 				     */
3532 				    free(nfhp, M_NFSFH);
3533 				} else {
3534 				    error = nfscl_nget(vnode_mount(vp), vp,
3535 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3536 				    if (!error) {
3537 					newvp = NFSTOV(np);
3538 					unlocknewvp = 1;
3539 				    }
3540 				}
3541 				nfhp = NULL;
3542 				if (newvp != NULLVP) {
3543 				    error = nfscl_loadattrcache(&newvp,
3544 					&nfsva, NULL, NULL, 0, 0);
3545 				    if (error) {
3546 					if (unlocknewvp)
3547 					    vput(newvp);
3548 					else
3549 					    vrele(newvp);
3550 					goto nfsmout;
3551 				    }
3552 				    dp->d_type =
3553 					vtonfs_dtype(np->n_vattr.na_type);
3554 				    ndp->ni_vp = newvp;
3555 				    NFSCNHASH(cnp, HASHINIT);
3556 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3557 					(newvp->v_type != VDIR ||
3558 					 dctime.tv_sec != 0)) {
3559 					cache_enter_time(ndp->ni_dvp,
3560 					    ndp->ni_vp, cnp,
3561 					    &nfsva.na_ctime,
3562 					    newvp->v_type != VDIR ? NULL :
3563 					    &dctime);
3564 				    }
3565 				    if (unlocknewvp)
3566 					vput(newvp);
3567 				    else
3568 					vrele(newvp);
3569 				    newvp = NULLVP;
3570 				}
3571 			    }
3572 			} else if (nfhp != NULL) {
3573 			    FREE((caddr_t)nfhp, M_NFSFH);
3574 			}
3575 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3576 			more_dirs = fxdr_unsigned(int, *tl);
3577 		}
3578 		/*
3579 		 * If at end of rpc data, get the eof boolean
3580 		 */
3581 		if (!more_dirs) {
3582 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3583 			eof = fxdr_unsigned(int, *tl);
3584 			if (tryformoredirs)
3585 				more_dirs = !eof;
3586 			if (nd->nd_flag & ND_NFSV4) {
3587 				error = nfscl_postop_attr(nd, nap, attrflagp,
3588 				    stuff);
3589 				if (error)
3590 					goto nfsmout;
3591 			}
3592 		}
3593 		mbuf_freem(nd->nd_mrep);
3594 		nd->nd_mrep = NULL;
3595 	}
3596 	/*
3597 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3598 	 * by increasing d_reclen for the last record.
3599 	 */
3600 	if (blksiz > 0) {
3601 		left = DIRBLKSIZ - blksiz;
3602 		dp->d_reclen += left;
3603 		uio_iov_base_add(uiop, left);
3604 		uio_iov_len_add(uiop, -(left));
3605 		uio_uio_resid_add(uiop, -(left));
3606 		uiop->uio_offset += left;
3607 	}
3608 
3609 	/*
3610 	 * If returning no data, assume end of file.
3611 	 * If not bigenough, return not end of file, since you aren't
3612 	 *    returning all the data
3613 	 * Otherwise, return the eof flag from the server.
3614 	 */
3615 	if (eofp != NULL) {
3616 		if (tresid == uio_uio_resid(uiop))
3617 			*eofp = 1;
3618 		else if (!bigenough)
3619 			*eofp = 0;
3620 		else
3621 			*eofp = eof;
3622 	}
3623 
3624 	/*
3625 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3626 	 */
3627 	while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3628 		dp = (struct dirent *)uio_iov_base(uiop);
3629 		dp->d_type = DT_UNKNOWN;
3630 		dp->d_fileno = 0;
3631 		dp->d_namlen = 0;
3632 		dp->d_name[0] = '\0';
3633 		tl = (u_int32_t *)&dp->d_name[4];
3634 		*tl++ = cookie.lval[0];
3635 		*tl = cookie.lval[1];
3636 		dp->d_reclen = DIRBLKSIZ;
3637 		uio_iov_base_add(uiop, DIRBLKSIZ);
3638 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3639 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3640 		uiop->uio_offset += DIRBLKSIZ;
3641 	}
3642 
3643 nfsmout:
3644 	if (nd->nd_mrep != NULL)
3645 		mbuf_freem(nd->nd_mrep);
3646 	return (error);
3647 }
3648 #endif	/* !APPLE */
3649 
3650 /*
3651  * Nfs commit rpc
3652  */
3653 APPLESTATIC int
3654 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3655     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3656 {
3657 	u_int32_t *tl;
3658 	struct nfsrv_descript nfsd, *nd = &nfsd;
3659 	nfsattrbit_t attrbits;
3660 	int error;
3661 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3662 
3663 	*attrflagp = 0;
3664 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3665 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3666 	txdr_hyper(offset, tl);
3667 	tl += 2;
3668 	*tl = txdr_unsigned(cnt);
3669 	if (nd->nd_flag & ND_NFSV4) {
3670 		/*
3671 		 * And do a Getattr op.
3672 		 */
3673 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3674 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3675 		NFSGETATTR_ATTRBIT(&attrbits);
3676 		(void) nfsrv_putattrbit(nd, &attrbits);
3677 	}
3678 	error = nfscl_request(nd, vp, p, cred, stuff);
3679 	if (error)
3680 		return (error);
3681 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3682 	if (!error && !nd->nd_repstat) {
3683 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3684 		NFSLOCKMNT(nmp);
3685 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3686 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3687 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3688 		}
3689 		NFSUNLOCKMNT(nmp);
3690 		if (nd->nd_flag & ND_NFSV4)
3691 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3692 	}
3693 nfsmout:
3694 	if (!error && nd->nd_repstat)
3695 		error = nd->nd_repstat;
3696 	mbuf_freem(nd->nd_mrep);
3697 	return (error);
3698 }
3699 
3700 /*
3701  * NFS byte range lock rpc.
3702  * (Mostly just calls one of the three lower level RPC routines.)
3703  */
3704 APPLESTATIC int
3705 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3706     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3707 {
3708 	struct nfscllockowner *lp;
3709 	struct nfsclclient *clp;
3710 	struct nfsfh *nfhp;
3711 	struct nfsrv_descript nfsd, *nd = &nfsd;
3712 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3713 	u_int64_t off, len;
3714 	off_t start, end;
3715 	u_int32_t clidrev = 0;
3716 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3717 	int callcnt, dorpc;
3718 
3719 	/*
3720 	 * Convert the flock structure into a start and end and do POSIX
3721 	 * bounds checking.
3722 	 */
3723 	switch (fl->l_whence) {
3724 	case SEEK_SET:
3725 	case SEEK_CUR:
3726 		/*
3727 		 * Caller is responsible for adding any necessary offset
3728 		 * when SEEK_CUR is used.
3729 		 */
3730 		start = fl->l_start;
3731 		off = fl->l_start;
3732 		break;
3733 	case SEEK_END:
3734 		start = size + fl->l_start;
3735 		off = size + fl->l_start;
3736 		break;
3737 	default:
3738 		return (EINVAL);
3739 	}
3740 	if (start < 0)
3741 		return (EINVAL);
3742 	if (fl->l_len != 0) {
3743 		end = start + fl->l_len - 1;
3744 		if (end < start)
3745 			return (EINVAL);
3746 	}
3747 
3748 	len = fl->l_len;
3749 	if (len == 0)
3750 		len = NFS64BITSSET;
3751 	retrycnt = 0;
3752 	do {
3753 	    nd->nd_repstat = 0;
3754 	    if (op == F_GETLK) {
3755 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3756 		if (error)
3757 			return (error);
3758 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3759 		if (!error) {
3760 			clidrev = clp->nfsc_clientidrev;
3761 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3762 			    p, id, flags);
3763 		} else if (error == -1) {
3764 			error = 0;
3765 		}
3766 		nfscl_clientrelease(clp);
3767 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3768 		/*
3769 		 * We must loop around for all lockowner cases.
3770 		 */
3771 		callcnt = 0;
3772 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3773 		if (error)
3774 			return (error);
3775 		do {
3776 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3777 			clp, id, flags, &lp, &dorpc);
3778 		    /*
3779 		     * If it returns a NULL lp, we're done.
3780 		     */
3781 		    if (lp == NULL) {
3782 			if (callcnt == 0)
3783 			    nfscl_clientrelease(clp);
3784 			else
3785 			    nfscl_releasealllocks(clp, vp, p, id, flags);
3786 			return (error);
3787 		    }
3788 		    if (nmp->nm_clp != NULL)
3789 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3790 		    else
3791 			clidrev = 0;
3792 		    /*
3793 		     * If the server doesn't support Posix lock semantics,
3794 		     * only allow locks on the entire file, since it won't
3795 		     * handle overlapping byte ranges.
3796 		     * There might still be a problem when a lock
3797 		     * upgrade/downgrade (read<->write) occurs, since the
3798 		     * server "might" expect an unlock first?
3799 		     */
3800 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3801 			(off == 0 && len == NFS64BITSSET))) {
3802 			/*
3803 			 * Since the lock records will go away, we must
3804 			 * wait for grace and delay here.
3805 			 */
3806 			do {
3807 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
3808 				NFSV4LOCKT_READ, cred, p, 0);
3809 			    if ((nd->nd_repstat == NFSERR_GRACE ||
3810 				 nd->nd_repstat == NFSERR_DELAY) &&
3811 				error == 0)
3812 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
3813 				    "nfs_advlock");
3814 			} while ((nd->nd_repstat == NFSERR_GRACE ||
3815 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
3816 		    }
3817 		    callcnt++;
3818 		} while (error == 0 && nd->nd_repstat == 0);
3819 		nfscl_releasealllocks(clp, vp, p, id, flags);
3820 	    } else if (op == F_SETLK) {
3821 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
3822 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
3823 		if (error || donelocally) {
3824 			return (error);
3825 		}
3826 		if (nmp->nm_clp != NULL)
3827 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3828 		else
3829 			clidrev = 0;
3830 		nfhp = VTONFS(vp)->n_fhp;
3831 		if (!lp->nfsl_open->nfso_posixlock &&
3832 		    (off != 0 || len != NFS64BITSSET)) {
3833 			error = EINVAL;
3834 		} else {
3835 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
3836 			    nfhp->nfh_len, lp, newone, reclaim, off,
3837 			    len, fl->l_type, cred, p, 0);
3838 		}
3839 		if (!error)
3840 			error = nd->nd_repstat;
3841 		nfscl_lockrelease(lp, error, newone);
3842 	    } else {
3843 		error = EINVAL;
3844 	    }
3845 	    if (!error)
3846 	        error = nd->nd_repstat;
3847 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
3848 		error == NFSERR_STALEDONTRECOVER ||
3849 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3850 		error == NFSERR_BADSESSION) {
3851 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
3852 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
3853 		&& clidrev != 0) {
3854 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
3855 		retrycnt++;
3856 	    }
3857 	} while (error == NFSERR_GRACE ||
3858 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
3859 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
3860 	    error == NFSERR_BADSESSION ||
3861 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
3862 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
3863 	if (error && retrycnt >= 4)
3864 		error = EIO;
3865 	return (error);
3866 }
3867 
3868 /*
3869  * The lower level routine for the LockT case.
3870  */
3871 APPLESTATIC int
3872 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
3873     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
3874     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3875 {
3876 	u_int32_t *tl;
3877 	int error, type, size;
3878 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
3879 	struct nfsnode *np;
3880 	struct nfsmount *nmp;
3881 	struct nfsclsession *tsep;
3882 
3883 	nmp = VFSTONFS(vp->v_mount);
3884 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
3885 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
3886 	if (fl->l_type == F_RDLCK)
3887 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
3888 	else
3889 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
3890 	txdr_hyper(off, tl);
3891 	tl += 2;
3892 	txdr_hyper(len, tl);
3893 	tl += 2;
3894 	tsep = nfsmnt_mdssession(nmp);
3895 	*tl++ = tsep->nfsess_clientid.lval[0];
3896 	*tl = tsep->nfsess_clientid.lval[1];
3897 	nfscl_filllockowner(id, own, flags);
3898 	np = VTONFS(vp);
3899 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
3900 	    np->n_fhp->nfh_len);
3901 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
3902 	error = nfscl_request(nd, vp, p, cred, NULL);
3903 	if (error)
3904 		return (error);
3905 	if (nd->nd_repstat == 0) {
3906 		fl->l_type = F_UNLCK;
3907 	} else if (nd->nd_repstat == NFSERR_DENIED) {
3908 		nd->nd_repstat = 0;
3909 		fl->l_whence = SEEK_SET;
3910 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
3911 		fl->l_start = fxdr_hyper(tl);
3912 		tl += 2;
3913 		len = fxdr_hyper(tl);
3914 		tl += 2;
3915 		if (len == NFS64BITSSET)
3916 			fl->l_len = 0;
3917 		else
3918 			fl->l_len = len;
3919 		type = fxdr_unsigned(int, *tl++);
3920 		if (type == NFSV4LOCKT_WRITE)
3921 			fl->l_type = F_WRLCK;
3922 		else
3923 			fl->l_type = F_RDLCK;
3924 		/*
3925 		 * XXX For now, I have no idea what to do with the
3926 		 * conflicting lock_owner, so I'll just set the pid == 0
3927 		 * and skip over the lock_owner.
3928 		 */
3929 		fl->l_pid = (pid_t)0;
3930 		tl += 2;
3931 		size = fxdr_unsigned(int, *tl);
3932 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
3933 			error = EBADRPC;
3934 		if (!error)
3935 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
3936 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
3937 		nfscl_initiate_recovery(clp);
3938 nfsmout:
3939 	mbuf_freem(nd->nd_mrep);
3940 	return (error);
3941 }
3942 
3943 /*
3944  * Lower level function that performs the LockU RPC.
3945  */
3946 static int
3947 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
3948     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
3949     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
3950 {
3951 	u_int32_t *tl;
3952 	int error;
3953 
3954 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
3955 	    lp->nfsl_open->nfso_fhlen, NULL, NULL);
3956 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
3957 	*tl++ = txdr_unsigned(type);
3958 	*tl = txdr_unsigned(lp->nfsl_seqid);
3959 	if (nfstest_outofseq &&
3960 	    (arc4random() % nfstest_outofseq) == 0)
3961 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
3962 	tl++;
3963 	if (NFSHASNFSV4N(nmp))
3964 		*tl++ = 0;
3965 	else
3966 		*tl++ = lp->nfsl_stateid.seqid;
3967 	*tl++ = lp->nfsl_stateid.other[0];
3968 	*tl++ = lp->nfsl_stateid.other[1];
3969 	*tl++ = lp->nfsl_stateid.other[2];
3970 	txdr_hyper(off, tl);
3971 	tl += 2;
3972 	txdr_hyper(len, tl);
3973 	if (syscred)
3974 		nd->nd_flag |= ND_USEGSSNAME;
3975 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
3976 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
3977 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
3978 	if (error)
3979 		return (error);
3980 	if (nd->nd_repstat == 0) {
3981 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
3982 		lp->nfsl_stateid.seqid = *tl++;
3983 		lp->nfsl_stateid.other[0] = *tl++;
3984 		lp->nfsl_stateid.other[1] = *tl++;
3985 		lp->nfsl_stateid.other[2] = *tl;
3986 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
3987 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
3988 nfsmout:
3989 	mbuf_freem(nd->nd_mrep);
3990 	return (error);
3991 }
3992 
3993 /*
3994  * The actual Lock RPC.
3995  */
3996 APPLESTATIC int
3997 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
3998     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
3999     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4000     NFSPROC_T *p, int syscred)
4001 {
4002 	u_int32_t *tl;
4003 	int error, size;
4004 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4005 	struct nfsclsession *tsep;
4006 
4007 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL);
4008 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4009 	if (type == F_RDLCK)
4010 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4011 	else
4012 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4013 	*tl++ = txdr_unsigned(reclaim);
4014 	txdr_hyper(off, tl);
4015 	tl += 2;
4016 	txdr_hyper(len, tl);
4017 	tl += 2;
4018 	if (newone) {
4019 	    *tl = newnfs_true;
4020 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4021 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4022 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4023 	    if (NFSHASNFSV4N(nmp))
4024 		*tl++ = 0;
4025 	    else
4026 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4027 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4028 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4029 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4030 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4031 	    tsep = nfsmnt_mdssession(nmp);
4032 	    *tl++ = tsep->nfsess_clientid.lval[0];
4033 	    *tl = tsep->nfsess_clientid.lval[1];
4034 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4035 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4036 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4037 	} else {
4038 	    *tl = newnfs_false;
4039 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4040 	    if (NFSHASNFSV4N(nmp))
4041 		*tl++ = 0;
4042 	    else
4043 		*tl++ = lp->nfsl_stateid.seqid;
4044 	    *tl++ = lp->nfsl_stateid.other[0];
4045 	    *tl++ = lp->nfsl_stateid.other[1];
4046 	    *tl++ = lp->nfsl_stateid.other[2];
4047 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4048 	    if (nfstest_outofseq &&
4049 		(arc4random() % nfstest_outofseq) == 0)
4050 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4051 	}
4052 	if (syscred)
4053 		nd->nd_flag |= ND_USEGSSNAME;
4054 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4055 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4056 	if (error)
4057 		return (error);
4058 	if (newone)
4059 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4060 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4061 	if (nd->nd_repstat == 0) {
4062 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4063 		lp->nfsl_stateid.seqid = *tl++;
4064 		lp->nfsl_stateid.other[0] = *tl++;
4065 		lp->nfsl_stateid.other[1] = *tl++;
4066 		lp->nfsl_stateid.other[2] = *tl;
4067 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4068 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4069 		size = fxdr_unsigned(int, *(tl + 7));
4070 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4071 			error = EBADRPC;
4072 		if (!error)
4073 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4074 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4075 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4076 nfsmout:
4077 	mbuf_freem(nd->nd_mrep);
4078 	return (error);
4079 }
4080 
4081 /*
4082  * nfs statfs rpc
4083  * (always called with the vp for the mount point)
4084  */
4085 APPLESTATIC int
4086 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4087     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4088     void *stuff)
4089 {
4090 	u_int32_t *tl = NULL;
4091 	struct nfsrv_descript nfsd, *nd = &nfsd;
4092 	struct nfsmount *nmp;
4093 	nfsattrbit_t attrbits;
4094 	int error;
4095 
4096 	*attrflagp = 0;
4097 	nmp = VFSTONFS(vnode_mount(vp));
4098 	if (NFSHASNFSV4(nmp)) {
4099 		/*
4100 		 * For V4, you actually do a getattr.
4101 		 */
4102 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4103 		NFSSTATFS_GETATTRBIT(&attrbits);
4104 		(void) nfsrv_putattrbit(nd, &attrbits);
4105 		nd->nd_flag |= ND_USEGSSNAME;
4106 		error = nfscl_request(nd, vp, p, cred, stuff);
4107 		if (error)
4108 			return (error);
4109 		if (nd->nd_repstat == 0) {
4110 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4111 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4112 			    cred);
4113 			if (!error) {
4114 				nmp->nm_fsid[0] = nap->na_filesid[0];
4115 				nmp->nm_fsid[1] = nap->na_filesid[1];
4116 				NFSSETHASSETFSID(nmp);
4117 				*attrflagp = 1;
4118 			}
4119 		} else {
4120 			error = nd->nd_repstat;
4121 		}
4122 		if (error)
4123 			goto nfsmout;
4124 	} else {
4125 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4126 		error = nfscl_request(nd, vp, p, cred, stuff);
4127 		if (error)
4128 			return (error);
4129 		if (nd->nd_flag & ND_NFSV3) {
4130 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4131 			if (error)
4132 				goto nfsmout;
4133 		}
4134 		if (nd->nd_repstat) {
4135 			error = nd->nd_repstat;
4136 			goto nfsmout;
4137 		}
4138 		NFSM_DISSECT(tl, u_int32_t *,
4139 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4140 	}
4141 	if (NFSHASNFSV3(nmp)) {
4142 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4143 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4144 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4145 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4146 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4147 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4148 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4149 	} else if (NFSHASNFSV4(nmp) == 0) {
4150 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4151 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4152 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4153 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4154 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4155 	}
4156 nfsmout:
4157 	mbuf_freem(nd->nd_mrep);
4158 	return (error);
4159 }
4160 
4161 /*
4162  * nfs pathconf rpc
4163  */
4164 APPLESTATIC int
4165 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4166     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4167     void *stuff)
4168 {
4169 	struct nfsrv_descript nfsd, *nd = &nfsd;
4170 	struct nfsmount *nmp;
4171 	u_int32_t *tl;
4172 	nfsattrbit_t attrbits;
4173 	int error;
4174 
4175 	*attrflagp = 0;
4176 	nmp = VFSTONFS(vnode_mount(vp));
4177 	if (NFSHASNFSV4(nmp)) {
4178 		/*
4179 		 * For V4, you actually do a getattr.
4180 		 */
4181 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4182 		NFSPATHCONF_GETATTRBIT(&attrbits);
4183 		(void) nfsrv_putattrbit(nd, &attrbits);
4184 		nd->nd_flag |= ND_USEGSSNAME;
4185 		error = nfscl_request(nd, vp, p, cred, stuff);
4186 		if (error)
4187 			return (error);
4188 		if (nd->nd_repstat == 0) {
4189 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4190 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4191 			    cred);
4192 			if (!error)
4193 				*attrflagp = 1;
4194 		} else {
4195 			error = nd->nd_repstat;
4196 		}
4197 	} else {
4198 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4199 		error = nfscl_request(nd, vp, p, cred, stuff);
4200 		if (error)
4201 			return (error);
4202 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4203 		if (nd->nd_repstat && !error)
4204 			error = nd->nd_repstat;
4205 		if (!error) {
4206 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4207 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4208 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4209 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4210 			pc->pc_chownrestricted =
4211 			    fxdr_unsigned(u_int32_t, *tl++);
4212 			pc->pc_caseinsensitive =
4213 			    fxdr_unsigned(u_int32_t, *tl++);
4214 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4215 		}
4216 	}
4217 nfsmout:
4218 	mbuf_freem(nd->nd_mrep);
4219 	return (error);
4220 }
4221 
4222 /*
4223  * nfs version 3 fsinfo rpc call
4224  */
4225 APPLESTATIC int
4226 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4227     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4228 {
4229 	u_int32_t *tl;
4230 	struct nfsrv_descript nfsd, *nd = &nfsd;
4231 	int error;
4232 
4233 	*attrflagp = 0;
4234 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4235 	error = nfscl_request(nd, vp, p, cred, stuff);
4236 	if (error)
4237 		return (error);
4238 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4239 	if (nd->nd_repstat && !error)
4240 		error = nd->nd_repstat;
4241 	if (!error) {
4242 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4243 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4244 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4245 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4246 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4247 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4248 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4249 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4250 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4251 		tl += 2;
4252 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4253 		tl += 2;
4254 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4255 	}
4256 nfsmout:
4257 	mbuf_freem(nd->nd_mrep);
4258 	return (error);
4259 }
4260 
4261 /*
4262  * This function performs the Renew RPC.
4263  */
4264 APPLESTATIC int
4265 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4266     NFSPROC_T *p)
4267 {
4268 	u_int32_t *tl;
4269 	struct nfsrv_descript nfsd;
4270 	struct nfsrv_descript *nd = &nfsd;
4271 	struct nfsmount *nmp;
4272 	int error;
4273 	struct nfssockreq *nrp;
4274 	struct nfsclsession *tsep;
4275 
4276 	nmp = clp->nfsc_nmp;
4277 	if (nmp == NULL)
4278 		return (0);
4279 	if (dsp == NULL)
4280 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL);
4281 	else
4282 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4283 		    &dsp->nfsclds_sess);
4284 	if (!NFSHASNFSV4N(nmp)) {
4285 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4286 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4287 		tsep = nfsmnt_mdssession(nmp);
4288 		*tl++ = tsep->nfsess_clientid.lval[0];
4289 		*tl = tsep->nfsess_clientid.lval[1];
4290 	}
4291 	nrp = NULL;
4292 	if (dsp != NULL)
4293 		nrp = dsp->nfsclds_sockp;
4294 	if (nrp == NULL)
4295 		/* If NULL, use the MDS socket. */
4296 		nrp = &nmp->nm_sockreq;
4297 	nd->nd_flag |= ND_USEGSSNAME;
4298 	if (dsp == NULL)
4299 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4300 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4301 	else
4302 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4303 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4304 	if (error)
4305 		return (error);
4306 	error = nd->nd_repstat;
4307 	mbuf_freem(nd->nd_mrep);
4308 	return (error);
4309 }
4310 
4311 /*
4312  * This function performs the Releaselockowner RPC.
4313  */
4314 APPLESTATIC int
4315 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4316     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4317 {
4318 	struct nfsrv_descript nfsd, *nd = &nfsd;
4319 	u_int32_t *tl;
4320 	int error;
4321 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4322 	struct nfsclsession *tsep;
4323 
4324 	if (NFSHASNFSV4N(nmp)) {
4325 		/* For NFSv4.1, do a FreeStateID. */
4326 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4327 		    NULL);
4328 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4329 	} else {
4330 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4331 		    NULL);
4332 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4333 		tsep = nfsmnt_mdssession(nmp);
4334 		*tl++ = tsep->nfsess_clientid.lval[0];
4335 		*tl = tsep->nfsess_clientid.lval[1];
4336 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4337 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4338 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4339 	}
4340 	nd->nd_flag |= ND_USEGSSNAME;
4341 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4342 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4343 	if (error)
4344 		return (error);
4345 	error = nd->nd_repstat;
4346 	mbuf_freem(nd->nd_mrep);
4347 	return (error);
4348 }
4349 
4350 /*
4351  * This function performs the Compound to get the mount pt FH.
4352  */
4353 APPLESTATIC int
4354 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4355     NFSPROC_T *p)
4356 {
4357 	u_int32_t *tl;
4358 	struct nfsrv_descript nfsd;
4359 	struct nfsrv_descript *nd = &nfsd;
4360 	u_char *cp, *cp2;
4361 	int error, cnt, len, setnil;
4362 	u_int32_t *opcntp;
4363 
4364 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL);
4365 	cp = dirpath;
4366 	cnt = 0;
4367 	do {
4368 		setnil = 0;
4369 		while (*cp == '/')
4370 			cp++;
4371 		cp2 = cp;
4372 		while (*cp2 != '\0' && *cp2 != '/')
4373 			cp2++;
4374 		if (*cp2 == '/') {
4375 			setnil = 1;
4376 			*cp2 = '\0';
4377 		}
4378 		if (cp2 != cp) {
4379 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4380 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4381 			nfsm_strtom(nd, cp, strlen(cp));
4382 			cnt++;
4383 		}
4384 		if (setnil)
4385 			*cp2++ = '/';
4386 		cp = cp2;
4387 	} while (*cp != '\0');
4388 	if (NFSHASNFSV4N(nmp))
4389 		/* Has a Sequence Op done by nfscl_reqstart(). */
4390 		*opcntp = txdr_unsigned(3 + cnt);
4391 	else
4392 		*opcntp = txdr_unsigned(2 + cnt);
4393 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4394 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4395 	nd->nd_flag |= ND_USEGSSNAME;
4396 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4397 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4398 	if (error)
4399 		return (error);
4400 	if (nd->nd_repstat == 0) {
4401 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4402 		tl += (2 + 2 * cnt);
4403 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4404 			len > NFSX_FHMAX) {
4405 			nd->nd_repstat = NFSERR_BADXDR;
4406 		} else {
4407 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4408 			if (nd->nd_repstat == 0)
4409 				nmp->nm_fhsize = len;
4410 		}
4411 	}
4412 	error = nd->nd_repstat;
4413 nfsmout:
4414 	mbuf_freem(nd->nd_mrep);
4415 	return (error);
4416 }
4417 
4418 /*
4419  * This function performs the Delegreturn RPC.
4420  */
4421 APPLESTATIC int
4422 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4423     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4424 {
4425 	u_int32_t *tl;
4426 	struct nfsrv_descript nfsd;
4427 	struct nfsrv_descript *nd = &nfsd;
4428 	int error;
4429 
4430 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4431 	    dp->nfsdl_fhlen, NULL, NULL);
4432 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4433 	if (NFSHASNFSV4N(nmp))
4434 		*tl++ = 0;
4435 	else
4436 		*tl++ = dp->nfsdl_stateid.seqid;
4437 	*tl++ = dp->nfsdl_stateid.other[0];
4438 	*tl++ = dp->nfsdl_stateid.other[1];
4439 	*tl = dp->nfsdl_stateid.other[2];
4440 	if (syscred)
4441 		nd->nd_flag |= ND_USEGSSNAME;
4442 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4443 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4444 	if (error)
4445 		return (error);
4446 	error = nd->nd_repstat;
4447 	mbuf_freem(nd->nd_mrep);
4448 	return (error);
4449 }
4450 
4451 /*
4452  * nfs getacl call.
4453  */
4454 APPLESTATIC int
4455 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4456     struct acl *aclp, void *stuff)
4457 {
4458 	struct nfsrv_descript nfsd, *nd = &nfsd;
4459 	int error;
4460 	nfsattrbit_t attrbits;
4461 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4462 
4463 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4464 		return (EOPNOTSUPP);
4465 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4466 	NFSZERO_ATTRBIT(&attrbits);
4467 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4468 	(void) nfsrv_putattrbit(nd, &attrbits);
4469 	error = nfscl_request(nd, vp, p, cred, stuff);
4470 	if (error)
4471 		return (error);
4472 	if (!nd->nd_repstat)
4473 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4474 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4475 	else
4476 		error = nd->nd_repstat;
4477 	mbuf_freem(nd->nd_mrep);
4478 	return (error);
4479 }
4480 
4481 /*
4482  * nfs setacl call.
4483  */
4484 APPLESTATIC int
4485 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4486     struct acl *aclp, void *stuff)
4487 {
4488 	int error;
4489 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4490 
4491 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4492 		return (EOPNOTSUPP);
4493 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4494 	return (error);
4495 }
4496 
4497 /*
4498  * nfs setacl call.
4499  */
4500 static int
4501 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4502     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4503 {
4504 	struct nfsrv_descript nfsd, *nd = &nfsd;
4505 	int error;
4506 	nfsattrbit_t attrbits;
4507 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4508 
4509 	if (!NFSHASNFSV4(nmp))
4510 		return (EOPNOTSUPP);
4511 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4512 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4513 	NFSZERO_ATTRBIT(&attrbits);
4514 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4515 	(void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4516 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0);
4517 	error = nfscl_request(nd, vp, p, cred, stuff);
4518 	if (error)
4519 		return (error);
4520 	/* Don't care about the pre/postop attributes */
4521 	mbuf_freem(nd->nd_mrep);
4522 	return (nd->nd_repstat);
4523 }
4524 
4525 /*
4526  * Do the NFSv4.1 Exchange ID.
4527  */
4528 int
4529 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4530     struct nfssockreq *nrp, uint32_t exchflags, struct nfsclds **dspp,
4531     struct ucred *cred, NFSPROC_T *p)
4532 {
4533 	uint32_t *tl, v41flags;
4534 	struct nfsrv_descript nfsd;
4535 	struct nfsrv_descript *nd = &nfsd;
4536 	struct nfsclds *dsp;
4537 	struct timespec verstime;
4538 	int error, len;
4539 
4540 	*dspp = NULL;
4541 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL);
4542 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4543 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4544 	*tl = txdr_unsigned(clp->nfsc_rev);
4545 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4546 
4547 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4548 	*tl++ = txdr_unsigned(exchflags);
4549 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4550 
4551 	/* Set the implementation id4 */
4552 	*tl = txdr_unsigned(1);
4553 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4554 	(void) nfsm_strtom(nd, version, strlen(version));
4555 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4556 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4557 	verstime.tv_nsec = 0;
4558 	txdr_nfsv4time(&verstime, tl);
4559 	nd->nd_flag |= ND_USEGSSNAME;
4560 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4561 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4562 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4563 	    (int)nd->nd_repstat);
4564 	if (error != 0)
4565 		return (error);
4566 	if (nd->nd_repstat == 0) {
4567 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4568 		len = fxdr_unsigned(int, *(tl + 7));
4569 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4570 			error = NFSERR_BADXDR;
4571 			goto nfsmout;
4572 		}
4573 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4574 		    M_WAITOK | M_ZERO);
4575 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4576 		dsp->nfsclds_servownlen = len;
4577 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4578 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4579 		dsp->nfsclds_sess.nfsess_sequenceid =
4580 		    fxdr_unsigned(uint32_t, *tl++);
4581 		v41flags = fxdr_unsigned(uint32_t, *tl);
4582 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4583 		    NFSHASPNFSOPT(nmp)) {
4584 			NFSCL_DEBUG(1, "set PNFS\n");
4585 			NFSLOCKMNT(nmp);
4586 			nmp->nm_state |= NFSSTA_PNFS;
4587 			NFSUNLOCKMNT(nmp);
4588 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4589 		}
4590 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4591 			dsp->nfsclds_flags |= NFSCLDS_DS;
4592 		if (len > 0)
4593 			nd->nd_repstat = nfsrv_mtostr(nd,
4594 			    dsp->nfsclds_serverown, len);
4595 		if (nd->nd_repstat == 0) {
4596 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4597 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4598 			    NULL, MTX_DEF);
4599 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4600 			*dspp = dsp;
4601 		} else
4602 			free(dsp, M_NFSCLDS);
4603 	}
4604 	error = nd->nd_repstat;
4605 nfsmout:
4606 	mbuf_freem(nd->nd_mrep);
4607 	return (error);
4608 }
4609 
4610 /*
4611  * Do the NFSv4.1 Create Session.
4612  */
4613 int
4614 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4615     struct nfssockreq *nrp, uint32_t sequenceid, int mds, struct ucred *cred,
4616     NFSPROC_T *p)
4617 {
4618 	uint32_t crflags, *tl;
4619 	struct nfsrv_descript nfsd;
4620 	struct nfsrv_descript *nd = &nfsd;
4621 	int error, irdcnt;
4622 
4623 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL);
4624 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4625 	*tl++ = sep->nfsess_clientid.lval[0];
4626 	*tl++ = sep->nfsess_clientid.lval[1];
4627 	*tl++ = txdr_unsigned(sequenceid);
4628 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4629 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4630 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4631 	*tl = txdr_unsigned(crflags);
4632 
4633 	/* Fill in fore channel attributes. */
4634 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4635 	*tl++ = 0;				/* Header pad size */
4636 	*tl++ = txdr_unsigned(100000);		/* Max request size */
4637 	*tl++ = txdr_unsigned(100000);		/* Max response size */
4638 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4639 	*tl++ = txdr_unsigned(20);		/* Max operations */
4640 	*tl++ = txdr_unsigned(64);		/* Max slots */
4641 	*tl = 0;				/* No rdma ird */
4642 
4643 	/* Fill in back channel attributes. */
4644 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4645 	*tl++ = 0;				/* Header pad size */
4646 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4647 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4648 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4649 	*tl++ = txdr_unsigned(4);		/* Max operations */
4650 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4651 	*tl = 0;				/* No rdma ird */
4652 
4653 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4654 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4655 
4656 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4657 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4658 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4659 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4660 	*tl++ = 0;				/* Null machine name */
4661 	*tl++ = 0;				/* Uid == 0 */
4662 	*tl++ = 0;				/* Gid == 0 */
4663 	*tl = 0;				/* No additional gids */
4664 	nd->nd_flag |= ND_USEGSSNAME;
4665 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4666 	    NFS_VER4, NULL, 1, NULL, NULL);
4667 	if (error != 0)
4668 		return (error);
4669 	if (nd->nd_repstat == 0) {
4670 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4671 		    2 * NFSX_UNSIGNED);
4672 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4673 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4674 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4675 		crflags = fxdr_unsigned(uint32_t, *tl);
4676 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4677 			NFSLOCKMNT(nmp);
4678 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4679 			NFSUNLOCKMNT(nmp);
4680 		}
4681 
4682 		/* Get the fore channel slot count. */
4683 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4684 		tl += 3;		/* Skip the other counts. */
4685 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4686 		tl++;
4687 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4688 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4689 		irdcnt = fxdr_unsigned(int, *tl);
4690 		if (irdcnt > 0)
4691 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4692 
4693 		/* and the back channel slot count. */
4694 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4695 		tl += 5;
4696 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4697 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4698 	}
4699 	error = nd->nd_repstat;
4700 nfsmout:
4701 	mbuf_freem(nd->nd_mrep);
4702 	return (error);
4703 }
4704 
4705 /*
4706  * Do the NFSv4.1 Destroy Session.
4707  */
4708 int
4709 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4710     struct ucred *cred, NFSPROC_T *p)
4711 {
4712 	uint32_t *tl;
4713 	struct nfsrv_descript nfsd;
4714 	struct nfsrv_descript *nd = &nfsd;
4715 	int error;
4716 	struct nfsclsession *tsep;
4717 
4718 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL);
4719 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4720 	tsep = nfsmnt_mdssession(nmp);
4721 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4722 	nd->nd_flag |= ND_USEGSSNAME;
4723 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4724 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4725 	if (error != 0)
4726 		return (error);
4727 	error = nd->nd_repstat;
4728 	mbuf_freem(nd->nd_mrep);
4729 	return (error);
4730 }
4731 
4732 /*
4733  * Do the NFSv4.1 Destroy Client.
4734  */
4735 int
4736 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4737     struct ucred *cred, NFSPROC_T *p)
4738 {
4739 	uint32_t *tl;
4740 	struct nfsrv_descript nfsd;
4741 	struct nfsrv_descript *nd = &nfsd;
4742 	int error;
4743 	struct nfsclsession *tsep;
4744 
4745 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL);
4746 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4747 	tsep = nfsmnt_mdssession(nmp);
4748 	*tl++ = tsep->nfsess_clientid.lval[0];
4749 	*tl = tsep->nfsess_clientid.lval[1];
4750 	nd->nd_flag |= ND_USEGSSNAME;
4751 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4752 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4753 	if (error != 0)
4754 		return (error);
4755 	error = nd->nd_repstat;
4756 	mbuf_freem(nd->nd_mrep);
4757 	return (error);
4758 }
4759 
4760 /*
4761  * Do the NFSv4.1 LayoutGet.
4762  */
4763 int
4764 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
4765     uint64_t offset, uint64_t len, uint64_t minlen, int layoutlen,
4766     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp,
4767     struct ucred *cred, NFSPROC_T *p, void *stuff)
4768 {
4769 	uint32_t *tl;
4770 	struct nfsrv_descript nfsd, *nd = &nfsd;
4771 	struct nfsfh *nfhp;
4772 	struct nfsclflayout *flp, *prevflp, *tflp;
4773 	int cnt, error, gotiomode, fhcnt, nfhlen, i, j;
4774 	uint8_t *cp;
4775 	uint64_t retlen;
4776 
4777 	flp = NULL;
4778 	gotiomode = -1;
4779 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL);
4780 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
4781 	    NFSX_STATEID);
4782 	*tl++ = newnfs_false;		/* Don't signal availability. */
4783 	*tl++ = txdr_unsigned(NFSLAYOUT_NFSV4_1_FILES);
4784 	*tl++ = txdr_unsigned(iomode);
4785 	txdr_hyper(offset, tl);
4786 	tl += 2;
4787 	txdr_hyper(len, tl);
4788 	tl += 2;
4789 	txdr_hyper(minlen, tl);
4790 	tl += 2;
4791 	*tl++ = txdr_unsigned(stateidp->seqid);
4792 	NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
4793 	*tl++ = stateidp->other[0];
4794 	*tl++ = stateidp->other[1];
4795 	*tl++ = stateidp->other[2];
4796 	*tl = txdr_unsigned(layoutlen);
4797 	nd->nd_flag |= ND_USEGSSNAME;
4798 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4799 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4800 	if (error != 0)
4801 		return (error);
4802 	if (nd->nd_repstat == 0) {
4803 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
4804 		if (*tl++ != 0)
4805 			*retonclosep = 1;
4806 		else
4807 			*retonclosep = 0;
4808 		stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
4809 		NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
4810 		    (int)stateidp->seqid);
4811 		stateidp->other[0] = *tl++;
4812 		stateidp->other[1] = *tl++;
4813 		stateidp->other[2] = *tl++;
4814 		cnt = fxdr_unsigned(int, *tl);
4815 		NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
4816 		if (cnt <= 0 || cnt > 10000) {
4817 			/* Don't accept more than 10000 layouts in reply. */
4818 			error = NFSERR_BADXDR;
4819 			goto nfsmout;
4820 		}
4821 		for (i = 0; i < cnt; i++) {
4822 			/* Dissect all the way to the file handle cnt. */
4823 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_HYPER +
4824 			    6 * NFSX_UNSIGNED + NFSX_V4DEVICEID);
4825 			fhcnt = fxdr_unsigned(int, *(tl + 11 +
4826 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
4827 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
4828 			if (fhcnt < 0 || fhcnt > 100) {
4829 				/* Don't accept more than 100 file handles. */
4830 				error = NFSERR_BADXDR;
4831 				goto nfsmout;
4832 			}
4833 			if (fhcnt > 1)
4834 				flp = malloc(sizeof(*flp) + (fhcnt - 1) *
4835 				    sizeof(struct nfsfh *),
4836 				    M_NFSFLAYOUT, M_WAITOK);
4837 			else
4838 				flp = malloc(sizeof(*flp),
4839 				    M_NFSFLAYOUT, M_WAITOK);
4840 			flp->nfsfl_flags = 0;
4841 			flp->nfsfl_fhcnt = 0;
4842 			flp->nfsfl_devp = NULL;
4843 			flp->nfsfl_off = fxdr_hyper(tl); tl += 2;
4844 			retlen = fxdr_hyper(tl); tl += 2;
4845 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
4846 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
4847 			else
4848 				flp->nfsfl_end = flp->nfsfl_off + retlen;
4849 			flp->nfsfl_iomode = fxdr_unsigned(int, *tl++);
4850 			if (gotiomode == -1)
4851 				gotiomode = flp->nfsfl_iomode;
4852 			NFSCL_DEBUG(4, "layg reqiom=%d retiom=%d\n", iomode,
4853 			    (int)flp->nfsfl_iomode);
4854 			if (fxdr_unsigned(int, *tl++) !=
4855 			    NFSLAYOUT_NFSV4_1_FILES) {
4856 				printf("NFSv4.1: got non-files layout\n");
4857 				error = NFSERR_BADXDR;
4858 				goto nfsmout;
4859 			}
4860 			NFSBCOPY(++tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
4861 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4862 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
4863 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
4864 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
4865 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
4866 			if (fxdr_unsigned(int, *tl) != fhcnt) {
4867 				printf("EEK! bad fhcnt\n");
4868 				error = NFSERR_BADXDR;
4869 				goto nfsmout;
4870 			}
4871 			for (j = 0; j < fhcnt; j++) {
4872 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
4873 				nfhlen = fxdr_unsigned(int, *tl);
4874 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
4875 					error = NFSERR_BADXDR;
4876 					goto nfsmout;
4877 				}
4878 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
4879 				    M_NFSFH, M_WAITOK);
4880 				flp->nfsfl_fh[j] = nfhp;
4881 				flp->nfsfl_fhcnt++;
4882 				nfhp->nfh_len = nfhlen;
4883 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
4884 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
4885 			}
4886 			if (flp->nfsfl_iomode == gotiomode) {
4887 				/* Keep the list in increasing offset order. */
4888 				tflp = LIST_FIRST(flhp);
4889 				prevflp = NULL;
4890 				while (tflp != NULL &&
4891 				    tflp->nfsfl_off < flp->nfsfl_off) {
4892 					prevflp = tflp;
4893 					tflp = LIST_NEXT(tflp, nfsfl_list);
4894 				}
4895 				if (prevflp == NULL)
4896 					LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
4897 				else
4898 					LIST_INSERT_AFTER(prevflp, flp,
4899 					    nfsfl_list);
4900 			} else {
4901 				printf("nfscl_layoutget(): got wrong iomode\n");
4902 				nfscl_freeflayout(flp);
4903 			}
4904 			flp = NULL;
4905 		}
4906 	}
4907 	if (nd->nd_repstat != 0 && error == 0)
4908 		error = nd->nd_repstat;
4909 nfsmout:
4910 	if (error != 0 && flp != NULL)
4911 		nfscl_freeflayout(flp);
4912 	mbuf_freem(nd->nd_mrep);
4913 	return (error);
4914 }
4915 
4916 /*
4917  * Do the NFSv4.1 Get Device Info.
4918  */
4919 int
4920 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
4921     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
4922     NFSPROC_T *p)
4923 {
4924 	uint32_t cnt, *tl;
4925 	struct nfsrv_descript nfsd;
4926 	struct nfsrv_descript *nd = &nfsd;
4927 	struct sockaddr_storage ss;
4928 	struct nfsclds *dsp = NULL, **dspp;
4929 	struct nfscldevinfo *ndi;
4930 	int addrcnt, bitcnt, error, i, isudp, j, pos, safilled, stripecnt;
4931 	uint8_t stripeindex;
4932 
4933 	*ndip = NULL;
4934 	ndi = NULL;
4935 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL);
4936 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
4937 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
4938 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
4939 	*tl++ = txdr_unsigned(layouttype);
4940 	*tl++ = txdr_unsigned(100000);
4941 	if (notifybitsp != NULL && *notifybitsp != 0) {
4942 		*tl = txdr_unsigned(1);		/* One word of bits. */
4943 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
4944 		*tl = txdr_unsigned(*notifybitsp);
4945 	} else
4946 		*tl = txdr_unsigned(0);
4947 	nd->nd_flag |= ND_USEGSSNAME;
4948 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4949 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4950 	if (error != 0)
4951 		return (error);
4952 	if (nd->nd_repstat == 0) {
4953 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4954 		if (layouttype != fxdr_unsigned(int, *tl++))
4955 			printf("EEK! devinfo layout type not same!\n");
4956 		stripecnt = fxdr_unsigned(int, *++tl);
4957 		NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
4958 		if (stripecnt < 1 || stripecnt > 4096) {
4959 			printf("NFS devinfo stripecnt %d: out of range\n",
4960 			    stripecnt);
4961 			error = NFSERR_BADXDR;
4962 			goto nfsmout;
4963 		}
4964 		NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) * NFSX_UNSIGNED);
4965 		addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
4966 		NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
4967 		if (addrcnt < 1 || addrcnt > 128) {
4968 			printf("NFS devinfo addrcnt %d: out of range\n",
4969 			    addrcnt);
4970 			error = NFSERR_BADXDR;
4971 			goto nfsmout;
4972 		}
4973 
4974 		/*
4975 		 * Now we know how many stripe indices and addresses, so
4976 		 * we can allocate the structure the correct size.
4977 		 */
4978 		i = (stripecnt * sizeof(uint8_t)) / sizeof(struct nfsclds *)
4979 		    + 1;
4980 		NFSCL_DEBUG(4, "stripeindices=%d\n", i);
4981 		ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
4982 		    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK | M_ZERO);
4983 		NFSBCOPY(deviceid, ndi->nfsdi_deviceid, NFSX_V4DEVICEID);
4984 		ndi->nfsdi_refcnt = 0;
4985 		ndi->nfsdi_stripecnt = stripecnt;
4986 		ndi->nfsdi_addrcnt = addrcnt;
4987 		/* Fill in the stripe indices. */
4988 		for (i = 0; i < stripecnt; i++) {
4989 			stripeindex = fxdr_unsigned(uint8_t, *tl++);
4990 			NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
4991 			if (stripeindex >= addrcnt) {
4992 				printf("NFS devinfo stripeindex %d: too big\n",
4993 				    (int)stripeindex);
4994 				error = NFSERR_BADXDR;
4995 				goto nfsmout;
4996 			}
4997 			nfsfldi_setstripeindex(ndi, i, stripeindex);
4998 		}
4999 
5000 		/* Now, dissect the server address(es). */
5001 		safilled = 0;
5002 		for (i = 0; i < addrcnt; i++) {
5003 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5004 			cnt = fxdr_unsigned(uint32_t, *tl);
5005 			if (cnt == 0) {
5006 				printf("NFS devinfo 0 len addrlist\n");
5007 				error = NFSERR_BADXDR;
5008 				goto nfsmout;
5009 			}
5010 			dspp = nfsfldi_addr(ndi, i);
5011 			pos = arc4random() % cnt;	/* Choose one. */
5012 			safilled = 0;
5013 			for (j = 0; j < cnt; j++) {
5014 				error = nfsv4_getipaddr(nd, &ss, &isudp);
5015 				if (error != 0 && error != EPERM) {
5016 					error = NFSERR_BADXDR;
5017 					goto nfsmout;
5018 				}
5019 				if (error == 0 && isudp == 0) {
5020 					/*
5021 					 * The algorithm is:
5022 					 * - use "pos" entry if it is of the
5023 					 *   same af_family or none of them
5024 					 *   is of the same af_family
5025 					 * else
5026 					 * - use the first one of the same
5027 					 *   af_family.
5028 					 */
5029 					if ((safilled == 0 && ss.ss_family ==
5030 					     nmp->nm_nam->sa_family) ||
5031 					    (j == pos &&
5032 					     (safilled == 0 || ss.ss_family ==
5033 					      nmp->nm_nam->sa_family)) ||
5034 					    (safilled == 1 && ss.ss_family ==
5035 					     nmp->nm_nam->sa_family)) {
5036 						error = nfsrpc_fillsa(nmp, &ss,
5037 						    &dsp, p);
5038 						if (error == 0) {
5039 							*dspp = dsp;
5040 							if (ss.ss_family ==
5041 							 nmp->nm_nam->sa_family)
5042 								safilled = 2;
5043 							else
5044 								safilled = 1;
5045 						}
5046 					}
5047 				}
5048 			}
5049 			if (safilled == 0)
5050 				break;
5051 		}
5052 
5053 		/* And the notify bits. */
5054 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5055 		if (safilled != 0) {
5056 			bitcnt = fxdr_unsigned(int, *tl);
5057 			if (bitcnt > 0) {
5058 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5059 				if (notifybitsp != NULL)
5060 					*notifybitsp =
5061 					    fxdr_unsigned(uint32_t, *tl);
5062 			}
5063 			*ndip = ndi;
5064 		} else
5065 			error = EPERM;
5066 	}
5067 	if (nd->nd_repstat != 0)
5068 		error = nd->nd_repstat;
5069 nfsmout:
5070 	if (error != 0 && ndi != NULL)
5071 		nfscl_freedevinfo(ndi);
5072 	mbuf_freem(nd->nd_mrep);
5073 	return (error);
5074 }
5075 
5076 /*
5077  * Do the NFSv4.1 LayoutCommit.
5078  */
5079 int
5080 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5081     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5082     int layouttype, int layoutupdatecnt, uint8_t *layp, struct ucred *cred,
5083     NFSPROC_T *p, void *stuff)
5084 {
5085 	uint32_t *tl;
5086 	struct nfsrv_descript nfsd, *nd = &nfsd;
5087 	int error, outcnt, i;
5088 	uint8_t *cp;
5089 
5090 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL);
5091 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5092 	    NFSX_STATEID);
5093 	txdr_hyper(off, tl);
5094 	tl += 2;
5095 	txdr_hyper(len, tl);
5096 	tl += 2;
5097 	if (reclaim != 0)
5098 		*tl++ = newnfs_true;
5099 	else
5100 		*tl++ = newnfs_false;
5101 	*tl++ = txdr_unsigned(stateidp->seqid);
5102 	*tl++ = stateidp->other[0];
5103 	*tl++ = stateidp->other[1];
5104 	*tl++ = stateidp->other[2];
5105 	*tl++ = newnfs_true;
5106 	if (lastbyte < off)
5107 		lastbyte = off;
5108 	else if (lastbyte >= (off + len))
5109 		lastbyte = off + len - 1;
5110 	txdr_hyper(lastbyte, tl);
5111 	tl += 2;
5112 	*tl++ = newnfs_false;
5113 	*tl++ = txdr_unsigned(layouttype);
5114 	*tl = txdr_unsigned(layoutupdatecnt);
5115 	if (layoutupdatecnt > 0) {
5116 		KASSERT(layouttype != NFSLAYOUT_NFSV4_1_FILES,
5117 		    ("Must be nil for Files Layout"));
5118 		outcnt = NFSM_RNDUP(layoutupdatecnt);
5119 		NFSM_BUILD(cp, uint8_t *, outcnt);
5120 		NFSBCOPY(layp, cp, layoutupdatecnt);
5121 		cp += layoutupdatecnt;
5122 		for (i = 0; i < (outcnt - layoutupdatecnt); i++)
5123 			*cp++ = 0x0;
5124 	}
5125 	nd->nd_flag |= ND_USEGSSNAME;
5126 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5127 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5128 	if (error != 0)
5129 		return (error);
5130 	error = nd->nd_repstat;
5131 	mbuf_freem(nd->nd_mrep);
5132 	return (error);
5133 }
5134 
5135 /*
5136  * Do the NFSv4.1 LayoutReturn.
5137  */
5138 int
5139 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5140     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5141     uint64_t len, nfsv4stateid_t *stateidp, int layoutcnt, uint32_t *layp,
5142     struct ucred *cred, NFSPROC_T *p, void *stuff)
5143 {
5144 	uint32_t *tl;
5145 	struct nfsrv_descript nfsd, *nd = &nfsd;
5146 	int error, outcnt, i;
5147 	uint8_t *cp;
5148 
5149 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL);
5150 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5151 	if (reclaim != 0)
5152 		*tl++ = newnfs_true;
5153 	else
5154 		*tl++ = newnfs_false;
5155 	*tl++ = txdr_unsigned(layouttype);
5156 	*tl++ = txdr_unsigned(iomode);
5157 	*tl = txdr_unsigned(layoutreturn);
5158 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5159 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5160 		    NFSX_UNSIGNED);
5161 		txdr_hyper(offset, tl);
5162 		tl += 2;
5163 		txdr_hyper(len, tl);
5164 		tl += 2;
5165 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5166 		*tl++ = txdr_unsigned(stateidp->seqid);
5167 		*tl++ = stateidp->other[0];
5168 		*tl++ = stateidp->other[1];
5169 		*tl++ = stateidp->other[2];
5170 		*tl = txdr_unsigned(layoutcnt);
5171 		if (layoutcnt > 0) {
5172 			outcnt = NFSM_RNDUP(layoutcnt);
5173 			NFSM_BUILD(cp, uint8_t *, outcnt);
5174 			NFSBCOPY(layp, cp, layoutcnt);
5175 			cp += layoutcnt;
5176 			for (i = 0; i < (outcnt - layoutcnt); i++)
5177 				*cp++ = 0x0;
5178 		}
5179 	}
5180 	nd->nd_flag |= ND_USEGSSNAME;
5181 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5182 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5183 	if (error != 0)
5184 		return (error);
5185 	if (nd->nd_repstat == 0) {
5186 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5187 		if (*tl != 0) {
5188 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5189 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5190 			stateidp->other[0] = *tl++;
5191 			stateidp->other[1] = *tl++;
5192 			stateidp->other[2] = *tl;
5193 		}
5194 	} else
5195 		error = nd->nd_repstat;
5196 nfsmout:
5197 	mbuf_freem(nd->nd_mrep);
5198 	return (error);
5199 }
5200 
5201 /*
5202  * Acquire a layout and devinfo, if possible. The caller must have acquired
5203  * a reference count on the nfsclclient structure before calling this.
5204  * Return the layout in lypp with a reference count on it, if successful.
5205  */
5206 static int
5207 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5208     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5209     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5210 {
5211 	struct nfscllayout *lyp;
5212 	struct nfsclflayout *flp, *tflp;
5213 	struct nfscldevinfo *dip;
5214 	struct nfsclflayouthead flh;
5215 	int error = 0, islocked, layoutlen, recalled, retonclose;
5216 	nfsv4stateid_t stateid;
5217 	struct nfsclsession *tsep;
5218 
5219 	*lypp = NULL;
5220 	/*
5221 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5222 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5223 	 * flp == NULL.
5224 	 */
5225 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5226 	    off, &flp, &recalled);
5227 	islocked = 0;
5228 	if (lyp == NULL || flp == NULL) {
5229 		if (recalled != 0)
5230 			return (EIO);
5231 		LIST_INIT(&flh);
5232 		tsep = nfsmnt_mdssession(nmp);
5233 		layoutlen = tsep->nfsess_maxcache -
5234 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5235 		if (lyp == NULL) {
5236 			stateid.seqid = 0;
5237 			stateid.other[0] = stateidp->other[0];
5238 			stateid.other[1] = stateidp->other[1];
5239 			stateid.other[2] = stateidp->other[2];
5240 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5241 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5242 			    (uint64_t)0, layoutlen, &stateid, &retonclose,
5243 			    &flh, cred, p, NULL);
5244 		} else {
5245 			islocked = 1;
5246 			stateid.seqid = lyp->nfsly_stateid.seqid;
5247 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5248 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5249 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5250 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5251 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5252 			    (uint64_t)0, layoutlen, &stateid, &retonclose,
5253 			    &flh, cred, p, NULL);
5254 		}
5255 		if (error == 0)
5256 			LIST_FOREACH(tflp, &flh, nfsfl_list) {
5257 				error = nfscl_adddevinfo(nmp, NULL, tflp);
5258 				if (error != 0) {
5259 					error = nfsrpc_getdeviceinfo(nmp,
5260 					    tflp->nfsfl_dev,
5261 					    NFSLAYOUT_NFSV4_1_FILES,
5262 					    notifybitsp, &dip, cred, p);
5263 					if (error != 0)
5264 						break;
5265 					error = nfscl_adddevinfo(nmp, dip,
5266 					    tflp);
5267 					if (error != 0)
5268 						printf(
5269 						    "getlayout: cannot add\n");
5270 				}
5271 			}
5272 		if (error == 0) {
5273 			/*
5274 			 * nfscl_layout() always returns with the nfsly_lock
5275 			 * set to a refcnt (shared lock).
5276 			 */
5277 			error = nfscl_layout(nmp, vp, nfhp->nfh_fh,
5278 			    nfhp->nfh_len, &stateid, retonclose, &flh, &lyp,
5279 			    cred, p);
5280 			if (error == 0)
5281 				*lypp = lyp;
5282 		} else if (islocked != 0)
5283 			nfsv4_unlock(&lyp->nfsly_lock, 0);
5284 	} else
5285 		*lypp = lyp;
5286 	return (error);
5287 }
5288 
5289 /*
5290  * Do a TCP connection plus exchange id and create session.
5291  * If successful, a "struct nfsclds" is linked into the list for the
5292  * mount point and a pointer to it is returned.
5293  */
5294 static int
5295 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_storage *ssp,
5296     struct nfsclds **dspp, NFSPROC_T *p)
5297 {
5298 	struct sockaddr_in *msad, *sad, *ssd;
5299 	struct sockaddr_in6 *msad6, *sad6, *ssd6;
5300 	struct nfsclclient *clp;
5301 	struct nfssockreq *nrp;
5302 	struct nfsclds *dsp, *tdsp;
5303 	int error;
5304 	enum nfsclds_state retv;
5305 	uint32_t sequenceid;
5306 
5307 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5308 	    ("nfsrpc_fillsa: NULL nr_cred"));
5309 	NFSLOCKCLSTATE();
5310 	clp = nmp->nm_clp;
5311 	NFSUNLOCKCLSTATE();
5312 	if (clp == NULL)
5313 		return (EPERM);
5314 	if (ssp->ss_family == AF_INET) {
5315 		ssd = (struct sockaddr_in *)ssp;
5316 		NFSLOCKMNT(nmp);
5317 
5318 		/*
5319 		 * Check to see if we already have a session for this
5320 		 * address that is usable for a DS.
5321 		 * Note that the MDS's address is in a different place
5322 		 * than the sessions already acquired for DS's.
5323 		 */
5324 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5325 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5326 		while (tdsp != NULL) {
5327 			if (msad != NULL && msad->sin_family == AF_INET &&
5328 			    ssd->sin_addr.s_addr == msad->sin_addr.s_addr &&
5329 			    ssd->sin_port == msad->sin_port &&
5330 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5331 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5332 				*dspp = tdsp;
5333 				NFSUNLOCKMNT(nmp);
5334 				NFSCL_DEBUG(4, "fnd same addr\n");
5335 				return (0);
5336 			}
5337 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5338 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5339 				msad = (struct sockaddr_in *)
5340 				    tdsp->nfsclds_sockp->nr_nam;
5341 			else
5342 				msad = NULL;
5343 		}
5344 		NFSUNLOCKMNT(nmp);
5345 
5346 		/* No IP address match, so look for new/trunked one. */
5347 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5348 		sad->sin_len = sizeof(*sad);
5349 		sad->sin_family = AF_INET;
5350 		sad->sin_port = ssd->sin_port;
5351 		sad->sin_addr.s_addr = ssd->sin_addr.s_addr;
5352 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5353 		nrp->nr_nam = (struct sockaddr *)sad;
5354 	} else if (ssp->ss_family == AF_INET6) {
5355 		ssd6 = (struct sockaddr_in6 *)ssp;
5356 		NFSLOCKMNT(nmp);
5357 
5358 		/*
5359 		 * Check to see if we already have a session for this
5360 		 * address that is usable for a DS.
5361 		 * Note that the MDS's address is in a different place
5362 		 * than the sessions already acquired for DS's.
5363 		 */
5364 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5365 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5366 		while (tdsp != NULL) {
5367 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5368 			    IN6_ARE_ADDR_EQUAL(&ssd6->sin6_addr,
5369 			    &msad6->sin6_addr) &&
5370 			    ssd6->sin6_port == msad6->sin6_port &&
5371 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5372 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5373 				*dspp = tdsp;
5374 				NFSUNLOCKMNT(nmp);
5375 				return (0);
5376 			}
5377 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5378 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5379 				msad6 = (struct sockaddr_in6 *)
5380 				    tdsp->nfsclds_sockp->nr_nam;
5381 			else
5382 				msad6 = NULL;
5383 		}
5384 		NFSUNLOCKMNT(nmp);
5385 
5386 		/* No IP address match, so look for new/trunked one. */
5387 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5388 		sad6->sin6_len = sizeof(*sad6);
5389 		sad6->sin6_family = AF_INET6;
5390 		sad6->sin6_port = ssd6->sin6_port;
5391 		NFSBCOPY(&ssd6->sin6_addr, &sad6->sin6_addr,
5392 		    sizeof(struct in6_addr));
5393 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5394 		nrp->nr_nam = (struct sockaddr *)sad6;
5395 	} else
5396 		return (EPERM);
5397 
5398 	nrp->nr_sotype = SOCK_STREAM;
5399 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5400 	nrp->nr_prog = NFS_PROG;
5401 	nrp->nr_vers = NFS_VER4;
5402 
5403 	/*
5404 	 * Use the credentials that were used for the mount, which are
5405 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5406 	 * Ref. counting the credentials with crhold() is probably not
5407 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5408 	 * unmount, but I did it anyhow.
5409 	 */
5410 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5411 	error = newnfs_connect(nmp, nrp, NULL, p, 0);
5412 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5413 
5414 	/* Now, do the exchangeid and create session. */
5415 	if (error == 0) {
5416 		error = nfsrpc_exchangeid(nmp, clp, nrp, NFSV4EXCH_USEPNFSDS,
5417 		    &dsp, nrp->nr_cred, p);
5418 		NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5419 		if (error != 0)
5420 			newnfs_disconnect(nrp);
5421 	}
5422 	if (error == 0) {
5423 		dsp->nfsclds_sockp = nrp;
5424 		NFSLOCKMNT(nmp);
5425 		retv = nfscl_getsameserver(nmp, dsp, &tdsp);
5426 		NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5427 		if (retv == NFSDSP_USETHISSESSION) {
5428 			NFSUNLOCKMNT(nmp);
5429 			/*
5430 			 * If there is already a session for this server,
5431 			 * use it.
5432 			 */
5433 			(void)newnfs_disconnect(nrp);
5434 			nfscl_freenfsclds(dsp);
5435 			*dspp = tdsp;
5436 			return (0);
5437 		}
5438 		if (retv == NFSDSP_SEQTHISSESSION)
5439 			sequenceid = tdsp->nfsclds_sess.nfsess_sequenceid;
5440 		else
5441 			sequenceid = dsp->nfsclds_sess.nfsess_sequenceid;
5442 		NFSUNLOCKMNT(nmp);
5443 		error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5444 		    nrp, sequenceid, 0, nrp->nr_cred, p);
5445 		NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5446 	} else {
5447 		NFSFREECRED(nrp->nr_cred);
5448 		NFSFREEMUTEX(&nrp->nr_mtx);
5449 		free(nrp->nr_nam, M_SONAME);
5450 		free(nrp, M_NFSSOCKREQ);
5451 	}
5452 	if (error == 0) {
5453 		NFSCL_DEBUG(3, "add DS session\n");
5454 		/*
5455 		 * Put it at the end of the list. That way the list
5456 		 * is ordered by when the entry was added. This matters
5457 		 * since the one done first is the one that should be
5458 		 * used for sequencid'ing any subsequent create sessions.
5459 		 */
5460 		NFSLOCKMNT(nmp);
5461 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5462 		NFSUNLOCKMNT(nmp);
5463 		*dspp = dsp;
5464 	} else if (dsp != NULL) {
5465 		newnfs_disconnect(nrp);
5466 		nfscl_freenfsclds(dsp);
5467 	}
5468 	return (error);
5469 }
5470 
5471 /*
5472  * Do the NFSv4.1 Reclaim Complete.
5473  */
5474 int
5475 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5476 {
5477 	uint32_t *tl;
5478 	struct nfsrv_descript nfsd;
5479 	struct nfsrv_descript *nd = &nfsd;
5480 	int error;
5481 
5482 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL);
5483 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5484 	*tl = newnfs_false;
5485 	nd->nd_flag |= ND_USEGSSNAME;
5486 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5487 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5488 	if (error != 0)
5489 		return (error);
5490 	error = nd->nd_repstat;
5491 	mbuf_freem(nd->nd_mrep);
5492 	return (error);
5493 }
5494 
5495 /*
5496  * Initialize the slot tables for a session.
5497  */
5498 static void
5499 nfscl_initsessionslots(struct nfsclsession *sep)
5500 {
5501 	int i;
5502 
5503 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5504 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5505 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5506 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5507 	}
5508 	for (i = 0; i < 64; i++)
5509 		sep->nfsess_slotseq[i] = 0;
5510 	sep->nfsess_slots = 0;
5511 }
5512 
5513 /*
5514  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5515  */
5516 int
5517 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5518     uint32_t rwaccess, struct ucred *cred, NFSPROC_T *p)
5519 {
5520 	struct nfsnode *np = VTONFS(vp);
5521 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5522 	struct nfscllayout *layp;
5523 	struct nfscldevinfo *dip;
5524 	struct nfsclflayout *rflp;
5525 	nfsv4stateid_t stateid;
5526 	struct ucred *newcred;
5527 	uint64_t lastbyte, len, off, oresid, xfer;
5528 	int eof, error, iolaymode, recalled;
5529 	void *lckp;
5530 
5531 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5532 	    (np->n_flag & NNOLAYOUT) != 0)
5533 		return (EIO);
5534 	/* Now, get a reference cnt on the clientid for this mount. */
5535 	if (nfscl_getref(nmp) == 0)
5536 		return (EIO);
5537 
5538 	/* Find an appropriate stateid. */
5539 	newcred = NFSNEWCRED(cred);
5540 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5541 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5542 	if (error != 0) {
5543 		NFSFREECRED(newcred);
5544 		nfscl_relref(nmp);
5545 		return (error);
5546 	}
5547 	/* Search for a layout for this file. */
5548 	off = uiop->uio_offset;
5549 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5550 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5551 	if (layp == NULL || rflp == NULL) {
5552 		if (recalled != 0) {
5553 			NFSFREECRED(newcred);
5554 			nfscl_relref(nmp);
5555 			return (EIO);
5556 		}
5557 		if (layp != NULL) {
5558 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5559 			layp = NULL;
5560 		}
5561 		/* Try and get a Layout, if it is supported. */
5562 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5563 		    (np->n_flag & NWRITEOPENED) != 0)
5564 			iolaymode = NFSLAYOUTIOMODE_RW;
5565 		else
5566 			iolaymode = NFSLAYOUTIOMODE_READ;
5567 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5568 		    NULL, &stateid, off, &layp, newcred, p);
5569 		if (error != 0) {
5570 			NFSLOCKNODE(np);
5571 			np->n_flag |= NNOLAYOUT;
5572 			NFSUNLOCKNODE(np);
5573 			if (lckp != NULL)
5574 				nfscl_lockderef(lckp);
5575 			NFSFREECRED(newcred);
5576 			if (layp != NULL)
5577 				nfscl_rellayout(layp, 0);
5578 			nfscl_relref(nmp);
5579 			return (error);
5580 		}
5581 	}
5582 
5583 	/*
5584 	 * Loop around finding a layout that works for the first part of
5585 	 * this I/O operation, and then call the function that actually
5586 	 * does the RPC.
5587 	 */
5588 	eof = 0;
5589 	len = (uint64_t)uiop->uio_resid;
5590 	while (len > 0 && error == 0 && eof == 0) {
5591 		off = uiop->uio_offset;
5592 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5593 		if (error == 0) {
5594 			oresid = xfer = (uint64_t)uiop->uio_resid;
5595 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5596 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5597 			dip = nfscl_getdevinfo(nmp->nm_clp, rflp->nfsfl_dev,
5598 			    rflp->nfsfl_devp);
5599 			if (dip != NULL) {
5600 				error = nfscl_doflayoutio(vp, uiop, iomode,
5601 				    must_commit, &eof, &stateid, rwaccess, dip,
5602 				    layp, rflp, off, xfer, newcred, p);
5603 				nfscl_reldevinfo(dip);
5604 				lastbyte = off + xfer - 1;
5605 				if (error == 0) {
5606 					NFSLOCKCLSTATE();
5607 					if (lastbyte > layp->nfsly_lastbyte)
5608 						layp->nfsly_lastbyte = lastbyte;
5609 					NFSUNLOCKCLSTATE();
5610 				} else if (error == NFSERR_OPENMODE &&
5611 				    rwaccess == NFSV4OPEN_ACCESSREAD) {
5612 					NFSLOCKMNT(nmp);
5613 					nmp->nm_state |= NFSSTA_OPENMODE;
5614 					NFSUNLOCKMNT(nmp);
5615 				}
5616 			} else
5617 				error = EIO;
5618 			if (error == 0)
5619 				len -= (oresid - (uint64_t)uiop->uio_resid);
5620 		}
5621 	}
5622 	if (lckp != NULL)
5623 		nfscl_lockderef(lckp);
5624 	NFSFREECRED(newcred);
5625 	nfscl_rellayout(layp, 0);
5626 	nfscl_relref(nmp);
5627 	return (error);
5628 }
5629 
5630 /*
5631  * Find a file layout that will handle the first bytes of the requested
5632  * range and return the information from it needed to to the I/O operation.
5633  */
5634 int
5635 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5636     struct nfsclflayout **retflpp)
5637 {
5638 	struct nfsclflayout *flp, *nflp, *rflp;
5639 	uint32_t rw;
5640 
5641 	rflp = NULL;
5642 	rw = rwaccess;
5643 	/* For reading, do the Read list first and then the Write list. */
5644 	do {
5645 		if (rw == NFSV4OPEN_ACCESSREAD)
5646 			flp = LIST_FIRST(&lyp->nfsly_flayread);
5647 		else
5648 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
5649 		while (flp != NULL) {
5650 			nflp = LIST_NEXT(flp, nfsfl_list);
5651 			if (flp->nfsfl_off > off)
5652 				break;
5653 			if (flp->nfsfl_end > off &&
5654 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5655 				rflp = flp;
5656 			flp = nflp;
5657 		}
5658 		if (rw == NFSV4OPEN_ACCESSREAD)
5659 			rw = NFSV4OPEN_ACCESSWRITE;
5660 		else
5661 			rw = 0;
5662 	} while (rw != 0);
5663 	if (rflp != NULL) {
5664 		/* This one covers the most bytes starting at off. */
5665 		*retflpp = rflp;
5666 		return (0);
5667 	}
5668 	return (EIO);
5669 }
5670 
5671 /*
5672  * Do I/O using an NFSv4.1 file layout.
5673  */
5674 static int
5675 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5676     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
5677     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
5678     uint64_t len, struct ucred *cred, NFSPROC_T *p)
5679 {
5680 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
5681 	int commit_thru_mds, error = 0, stripe_index, stripe_pos;
5682 	struct nfsnode *np;
5683 	struct nfsfh *fhp;
5684 	struct nfsclds **dspp;
5685 
5686 	np = VTONFS(vp);
5687 	rel_off = off - flp->nfsfl_patoff;
5688 	stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
5689 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
5690 	    dp->nfsdi_stripecnt;
5691 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
5692 
5693 	/* Loop around, doing I/O for each stripe unit. */
5694 	while (len > 0 && error == 0) {
5695 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
5696 		dspp = nfsfldi_addr(dp, stripe_index);
5697 		if (len > transfer)
5698 			xfer = transfer;
5699 		else
5700 			xfer = len;
5701 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
5702 			/* Dense layout. */
5703 			if (stripe_pos >= flp->nfsfl_fhcnt)
5704 				return (EIO);
5705 			fhp = flp->nfsfl_fh[stripe_pos];
5706 			io_off = (rel_off / (stripe_unit_size *
5707 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
5708 			    rel_off % stripe_unit_size;
5709 		} else {
5710 			/* Sparse layout. */
5711 			if (flp->nfsfl_fhcnt > 1) {
5712 				if (stripe_index >= flp->nfsfl_fhcnt)
5713 					return (EIO);
5714 				fhp = flp->nfsfl_fh[stripe_index];
5715 			} else if (flp->nfsfl_fhcnt == 1)
5716 				fhp = flp->nfsfl_fh[0];
5717 			else
5718 				fhp = np->n_fhp;
5719 			io_off = off;
5720 		}
5721 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0)
5722 			commit_thru_mds = 1;
5723 		else
5724 			commit_thru_mds = 0;
5725 		if (rwflag == FREAD)
5726 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
5727 			    io_off, xfer, fhp, cred, p);
5728 		else {
5729 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
5730 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
5731 			    cred, p);
5732 			if (error == 0) {
5733 				NFSLOCKCLSTATE();
5734 				lyp->nfsly_flags |= NFSLY_WRITTEN;
5735 				NFSUNLOCKCLSTATE();
5736 			}
5737 		}
5738 		if (error == 0) {
5739 			transfer = stripe_unit_size;
5740 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
5741 			len -= xfer;
5742 			off += xfer;
5743 		}
5744 	}
5745 	return (error);
5746 }
5747 
5748 /*
5749  * The actual read RPC done to a DS.
5750  */
5751 static int
5752 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
5753     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp,
5754     struct ucred *cred, NFSPROC_T *p)
5755 {
5756 	uint32_t *tl;
5757 	int error, retlen;
5758 	struct nfsrv_descript nfsd;
5759 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5760 	struct nfsrv_descript *nd = &nfsd;
5761 	struct nfssockreq *nrp;
5762 
5763 	nd->nd_mrep = NULL;
5764 	nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5765 	    NULL, &dsp->nfsclds_sess);
5766 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5767 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
5768 	txdr_hyper(io_off, tl);
5769 	*(tl + 2) = txdr_unsigned(len);
5770 	nrp = dsp->nfsclds_sockp;
5771 	if (nrp == NULL)
5772 		/* If NULL, use the MDS socket. */
5773 		nrp = &nmp->nm_sockreq;
5774 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5775 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5776 	if (error != 0)
5777 		return (error);
5778 	if (nd->nd_repstat != 0) {
5779 		error = nd->nd_repstat;
5780 		goto nfsmout;
5781 	}
5782 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5783 	*eofp = fxdr_unsigned(int, *tl);
5784 	NFSM_STRSIZ(retlen, len);
5785 	error = nfsm_mbufuio(nd, uiop, retlen);
5786 nfsmout:
5787 	if (nd->nd_mrep != NULL)
5788 		mbuf_freem(nd->nd_mrep);
5789 	return (error);
5790 }
5791 
5792 /*
5793  * The actual write RPC done to a DS.
5794  */
5795 static int
5796 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5797     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
5798     struct nfsfh *fhp, int commit_thru_mds, struct ucred *cred, NFSPROC_T *p)
5799 {
5800 	uint32_t *tl;
5801 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5802 	int error, rlen, commit, committed = NFSWRITE_FILESYNC;
5803 	int32_t backup;
5804 	struct nfsrv_descript nfsd;
5805 	struct nfsrv_descript *nd = &nfsd;
5806 	struct nfssockreq *nrp;
5807 
5808 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
5809 	nd->nd_mrep = NULL;
5810 	nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5811 	    NULL, &dsp->nfsclds_sess);
5812 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
5813 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
5814 	txdr_hyper(io_off, tl);
5815 	tl += 2;
5816 	*tl++ = txdr_unsigned(*iomode);
5817 	*tl = txdr_unsigned(len);
5818 	nfsm_uiombuf(nd, uiop, len);
5819 	nrp = dsp->nfsclds_sockp;
5820 	if (nrp == NULL)
5821 		/* If NULL, use the MDS socket. */
5822 		nrp = &nmp->nm_sockreq;
5823 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5824 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5825 	if (error != 0)
5826 		return (error);
5827 	if (nd->nd_repstat != 0) {
5828 		/*
5829 		 * In case the rpc gets retried, roll
5830 		 * the uio fileds changed by nfsm_uiombuf()
5831 		 * back.
5832 		 */
5833 		uiop->uio_offset -= len;
5834 		uio_uio_resid_add(uiop, len);
5835 		uio_iov_base_add(uiop, -len);
5836 		uio_iov_len_add(uiop, len);
5837 		error = nd->nd_repstat;
5838 	} else {
5839 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
5840 		rlen = fxdr_unsigned(int, *tl++);
5841 		if (rlen == 0) {
5842 			error = NFSERR_IO;
5843 			goto nfsmout;
5844 		} else if (rlen < len) {
5845 			backup = len - rlen;
5846 			uio_iov_base_add(uiop, -(backup));
5847 			uio_iov_len_add(uiop, backup);
5848 			uiop->uio_offset -= backup;
5849 			uio_uio_resid_add(uiop, backup);
5850 			len = rlen;
5851 		}
5852 		commit = fxdr_unsigned(int, *tl++);
5853 
5854 		/*
5855 		 * Return the lowest commitment level
5856 		 * obtained by any of the RPCs.
5857 		 */
5858 		if (committed == NFSWRITE_FILESYNC)
5859 			committed = commit;
5860 		else if (committed == NFSWRITE_DATASYNC &&
5861 		    commit == NFSWRITE_UNSTABLE)
5862 			committed = commit;
5863 		if (commit_thru_mds != 0) {
5864 			NFSLOCKMNT(nmp);
5865 			if (!NFSHASWRITEVERF(nmp)) {
5866 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5867 				NFSSETWRITEVERF(nmp);
5868 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
5869 				*must_commit = 1;
5870 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
5871 			}
5872 			NFSUNLOCKMNT(nmp);
5873 		} else {
5874 			NFSLOCKDS(dsp);
5875 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
5876 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5877 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
5878 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5879 				*must_commit = 1;
5880 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5881 			}
5882 			NFSUNLOCKDS(dsp);
5883 		}
5884 	}
5885 nfsmout:
5886 	if (nd->nd_mrep != NULL)
5887 		mbuf_freem(nd->nd_mrep);
5888 	*iomode = committed;
5889 	if (nd->nd_repstat != 0 && error == 0)
5890 		error = nd->nd_repstat;
5891 	return (error);
5892 }
5893 
5894 /*
5895  * Free up the nfsclds structure.
5896  */
5897 void
5898 nfscl_freenfsclds(struct nfsclds *dsp)
5899 {
5900 	int i;
5901 
5902 	if (dsp == NULL)
5903 		return;
5904 	if (dsp->nfsclds_sockp != NULL) {
5905 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
5906 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
5907 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
5908 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
5909 	}
5910 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
5911 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
5912 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5913 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
5914 			m_freem(
5915 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
5916 	}
5917 	free(dsp, M_NFSCLDS);
5918 }
5919 
5920 static enum nfsclds_state
5921 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
5922     struct nfsclds **retdspp)
5923 {
5924 	struct nfsclds *dsp, *cur_dsp;
5925 
5926 	/*
5927 	 * Search the list of nfsclds structures for one with the same
5928 	 * server.
5929 	 */
5930 	cur_dsp = NULL;
5931 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
5932 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
5933 		    dsp->nfsclds_servownlen != 0 &&
5934 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
5935 		    dsp->nfsclds_servownlen) &&
5936 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
5937 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
5938 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
5939 			    dsp->nfsclds_flags);
5940 			/* Server major id matches. */
5941 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
5942 				*retdspp = dsp;
5943 				return (NFSDSP_USETHISSESSION);
5944 			}
5945 
5946 			/*
5947 			 * Note the first match, so it can be used for
5948 			 * sequence'ing new sessions.
5949 			 */
5950 			if (cur_dsp == NULL)
5951 				cur_dsp = dsp;
5952 		}
5953 	}
5954 	if (cur_dsp != NULL) {
5955 		*retdspp = cur_dsp;
5956 		return (NFSDSP_SEQTHISSESSION);
5957 	}
5958 	return (NFSDSP_NOTFOUND);
5959 }
5960 
5961 #ifdef notyet
5962 /*
5963  * NFS commit rpc to a DS.
5964  */
5965 static int
5966 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
5967     struct nfsfh *fhp, struct ucred *cred, NFSPROC_T *p, void *stuff)
5968 {
5969 	uint32_t *tl;
5970 	struct nfsrv_descript nfsd, *nd = &nfsd;
5971 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5972 	struct nfssockreq *nrp;
5973 	int error;
5974 
5975 	nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh, fhp->nfh_len,
5976 	    NULL, &dsp->nfsclds_sess);
5977 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
5978 	txdr_hyper(offset, tl);
5979 	tl += 2;
5980 	*tl = txdr_unsigned(cnt);
5981 	nrp = dsp->nfsclds_sockp;
5982 	if (nrp == NULL)
5983 		/* If NULL, use the MDS socket. */
5984 		nrp = &nmp->nm_sockreq;
5985 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
5986 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
5987 	if (error)
5988 		return (error);
5989 	if (nd->nd_repstat == 0) {
5990 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
5991 		NFSLOCKDS(dsp);
5992 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
5993 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
5994 			error = NFSERR_STALEWRITEVERF;
5995 		}
5996 		NFSUNLOCKDS(dsp);
5997 	}
5998 nfsmout:
5999 	if (error == 0 && nd->nd_repstat != 0)
6000 		error = nd->nd_repstat;
6001 	mbuf_freem(nd->nd_mrep);
6002 	return (error);
6003 }
6004 #endif
6005 
6006