xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision ac099daf6742ead81ea7ea86351a8ef4e783041b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Rpc op calls, generally called from the vnode op calls or through the
41  * buffer cache, for NFS v2, 3 and 4.
42  * These do not normally make any changes to vnode arguments or use
43  * structures that might change between the VFS variants. The returned
44  * arguments are all at the end, after the NFSPROC_T *p one.
45  */
46 
47 #include "opt_inet6.h"
48 
49 #include <fs/nfs/nfsport.h>
50 #include <fs/nfsclient/nfs.h>
51 #include <sys/extattr.h>
52 #include <sys/sysctl.h>
53 #include <sys/taskqueue.h>
54 
55 SYSCTL_DECL(_vfs_nfs);
56 
57 static int	nfsignore_eexist = 0;
58 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
59     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
60 
61 static int	nfscl_dssameconn = 0;
62 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
63     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
64 
65 /*
66  * Global variables
67  */
68 extern struct nfsstatsv1 nfsstatsv1;
69 extern int nfs_numnfscbd;
70 extern struct timeval nfsboottime;
71 extern u_int32_t newnfs_false, newnfs_true;
72 extern nfstype nfsv34_type[9];
73 extern int nfsrv_useacl;
74 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
75 extern int nfscl_debuglevel;
76 extern int nfs_pnfsiothreads;
77 extern u_long sb_max_adj;
78 extern int nfs_maxcopyrange;
79 NFSCLSTATEMUTEX;
80 int nfstest_outofseq = 0;
81 int nfscl_assumeposixlocks = 1;
82 int nfscl_enablecallb = 0;
83 short nfsv4_cbport = NFSV4_CBPORT;
84 int nfstest_openallsetattr = 0;
85 
86 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
87 
88 /*
89  * nfscl_getsameserver() can return one of three values:
90  * NFSDSP_USETHISSESSION - Use this session for the DS.
91  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
92  *     session.
93  * NFSDSP_NOTFOUND - No matching server was found.
94  */
95 enum nfsclds_state {
96 	NFSDSP_USETHISSESSION = 0,
97 	NFSDSP_SEQTHISSESSION = 1,
98 	NFSDSP_NOTFOUND = 2,
99 };
100 
101 /*
102  * Do a write RPC on a DS data file, using this structure for the arguments,
103  * so that this function can be executed by a separate kernel process.
104  */
105 struct nfsclwritedsdorpc {
106 	int			done;
107 	int			inprog;
108 	struct task		tsk;
109 	struct vnode		*vp;
110 	int			iomode;
111 	int			must_commit;
112 	nfsv4stateid_t		*stateidp;
113 	struct nfsclds		*dsp;
114 	uint64_t		off;
115 	int			len;
116 #ifdef notyet
117 	int			advise;
118 #endif
119 	struct nfsfh		*fhp;
120 	struct mbuf		*m;
121 	int			vers;
122 	int			minorvers;
123 	struct ucred		*cred;
124 	NFSPROC_T		*p;
125 	int			err;
126 };
127 
128 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
129     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
130 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
131     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
132 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
133     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
134     void *);
135 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
136     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
137     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
138 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
139     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
140     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
141     int *, void *, int *);
142 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
143     struct nfscllockowner *, u_int64_t, u_int64_t,
144     u_int32_t, struct ucred *, NFSPROC_T *, int);
145 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
146     struct acl *, nfsv4stateid_t *, void *);
147 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
148     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
149     struct ucred *, NFSPROC_T *);
150 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
151     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
152     NFSPROC_T *);
153 static void nfscl_initsessionslots(struct nfsclsession *);
154 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
155     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
156     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
157     NFSPROC_T *);
158 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
159     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
160     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
161     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
162 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
163     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
164     struct ucred *, NFSPROC_T *);
165 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
166     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
167     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
168 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
169     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
170     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
171 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
172     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
173     struct ucred *, NFSPROC_T *);
174 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
175     struct nfsclds *, struct nfsclds **, uint32_t *);
176 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
177     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
178     NFSPROC_T *);
179 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
180     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
181 #ifdef notyet
182 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
183     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
184     NFSPROC_T *);
185 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
186     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
187 #endif
188 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
189     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
190 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
191     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
192 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
193     NFSPROC_T *);
194 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
195     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
196 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
197     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
198     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
199 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
200     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
201     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
202     struct nfsfh **, int *, int *, void *, int *);
203 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
204     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
205     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
206     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
207 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
208     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
209     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
210     struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
211     int, int, int, int *, struct nfsclflayouthead *, int *);
212 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
213     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
214     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
215 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
216     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
217     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
218 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
219     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
220     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
221 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
222     int, struct nfsvattr *, int *, struct ucred *);
223 static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
224 
225 int nfs_pnfsio(task_fn_t *, void *);
226 
227 /*
228  * nfs null call from vfs.
229  */
230 int
231 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
232 {
233 	int error;
234 	struct nfsrv_descript nfsd, *nd = &nfsd;
235 
236 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
237 	error = nfscl_request(nd, vp, p, cred, NULL);
238 	if (nd->nd_repstat && !error)
239 		error = nd->nd_repstat;
240 	m_freem(nd->nd_mrep);
241 	return (error);
242 }
243 
244 /*
245  * nfs access rpc op.
246  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
247  * modes are changed on the server, accesses might still fail later.
248  */
249 int
250 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
251     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
252 {
253 	int error;
254 	u_int32_t mode, rmode;
255 
256 	if (acmode & VREAD)
257 		mode = NFSACCESS_READ;
258 	else
259 		mode = 0;
260 	if (vnode_vtype(vp) == VDIR) {
261 		if (acmode & VWRITE)
262 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
263 				 NFSACCESS_DELETE);
264 		if (acmode & VEXEC)
265 			mode |= NFSACCESS_LOOKUP;
266 	} else {
267 		if (acmode & VWRITE)
268 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
269 		if (acmode & VEXEC)
270 			mode |= NFSACCESS_EXECUTE;
271 	}
272 
273 	/*
274 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
275 	 */
276 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
277 	    NULL);
278 
279 	/*
280 	 * The NFS V3 spec does not clarify whether or not
281 	 * the returned access bits can be a superset of
282 	 * the ones requested, so...
283 	 */
284 	if (!error && (rmode & mode) != mode)
285 		error = EACCES;
286 	return (error);
287 }
288 
289 /*
290  * The actual rpc, separated out for Darwin.
291  */
292 int
293 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
294     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
295     void *stuff)
296 {
297 	u_int32_t *tl;
298 	u_int32_t supported, rmode;
299 	int error;
300 	struct nfsrv_descript nfsd, *nd = &nfsd;
301 	nfsattrbit_t attrbits;
302 
303 	*attrflagp = 0;
304 	supported = mode;
305 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
306 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
307 	*tl = txdr_unsigned(mode);
308 	if (nd->nd_flag & ND_NFSV4) {
309 		/*
310 		 * And do a Getattr op.
311 		 */
312 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
313 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
314 		NFSGETATTR_ATTRBIT(&attrbits);
315 		(void) nfsrv_putattrbit(nd, &attrbits);
316 	}
317 	error = nfscl_request(nd, vp, p, cred, stuff);
318 	if (error)
319 		return (error);
320 	if (nd->nd_flag & ND_NFSV3) {
321 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
322 		if (error)
323 			goto nfsmout;
324 	}
325 	if (!nd->nd_repstat) {
326 		if (nd->nd_flag & ND_NFSV4) {
327 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
328 			supported = fxdr_unsigned(u_int32_t, *tl++);
329 		} else {
330 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
331 		}
332 		rmode = fxdr_unsigned(u_int32_t, *tl);
333 		if (nd->nd_flag & ND_NFSV4)
334 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
335 
336 		/*
337 		 * It's not obvious what should be done about
338 		 * unsupported access modes. For now, be paranoid
339 		 * and clear the unsupported ones.
340 		 */
341 		rmode &= supported;
342 		*rmodep = rmode;
343 	} else
344 		error = nd->nd_repstat;
345 nfsmout:
346 	m_freem(nd->nd_mrep);
347 	return (error);
348 }
349 
350 /*
351  * nfs open rpc
352  */
353 int
354 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
355 {
356 	struct nfsclopen *op;
357 	struct nfscldeleg *dp;
358 	struct nfsfh *nfhp;
359 	struct nfsnode *np = VTONFS(vp);
360 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
361 	u_int32_t mode, clidrev;
362 	int ret, newone, error, expireret = 0, retrycnt;
363 
364 	/*
365 	 * For NFSv4, Open Ops are only done on Regular Files.
366 	 */
367 	if (vnode_vtype(vp) != VREG)
368 		return (0);
369 	mode = 0;
370 	if (amode & FREAD)
371 		mode |= NFSV4OPEN_ACCESSREAD;
372 	if (amode & FWRITE)
373 		mode |= NFSV4OPEN_ACCESSWRITE;
374 	nfhp = np->n_fhp;
375 
376 	retrycnt = 0;
377 #ifdef notdef
378 { char name[100]; int namel;
379 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
380 bcopy(NFS4NODENAME(np->n_v4), name, namel);
381 name[namel] = '\0';
382 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
383 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
384 else printf(" fhl=0\n");
385 }
386 #endif
387 	do {
388 	    dp = NULL;
389 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
390 		cred, p, NULL, &op, &newone, &ret, 1);
391 	    if (error) {
392 		return (error);
393 	    }
394 	    if (nmp->nm_clp != NULL)
395 		clidrev = nmp->nm_clp->nfsc_clientidrev;
396 	    else
397 		clidrev = 0;
398 	    if (ret == NFSCLOPEN_DOOPEN) {
399 		if (np->n_v4 != NULL) {
400 			/*
401 			 * For the first attempt, try and get a layout, if
402 			 * pNFS is enabled for the mount.
403 			 */
404 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
405 			    nfs_numnfscbd == 0 ||
406 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
407 				error = nfsrpc_openrpc(nmp, vp,
408 				    np->n_v4->n4_data,
409 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
410 				    np->n_fhp->nfh_len, mode, op,
411 				    NFS4NODENAME(np->n_v4),
412 				    np->n_v4->n4_namelen,
413 				    &dp, 0, 0x0, cred, p, 0, 0);
414 			else
415 				error = nfsrpc_getopenlayout(nmp, vp,
416 				    np->n_v4->n4_data,
417 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
418 				    np->n_fhp->nfh_len, mode, op,
419 				    NFS4NODENAME(np->n_v4),
420 				    np->n_v4->n4_namelen, &dp, cred, p);
421 			if (dp != NULL) {
422 #ifdef APPLE
423 				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
424 #else
425 				NFSLOCKNODE(np);
426 				np->n_flag &= ~NDELEGMOD;
427 				/*
428 				 * Invalidate the attribute cache, so that
429 				 * attributes that pre-date the issue of a
430 				 * delegation are not cached, since the
431 				 * cached attributes will remain valid while
432 				 * the delegation is held.
433 				 */
434 				NFSINVALATTRCACHE(np);
435 				NFSUNLOCKNODE(np);
436 #endif
437 				(void) nfscl_deleg(nmp->nm_mountp,
438 				    op->nfso_own->nfsow_clp,
439 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
440 			}
441 		} else {
442 			error = EIO;
443 		}
444 		newnfs_copyincred(cred, &op->nfso_cred);
445 	    } else if (ret == NFSCLOPEN_SETCRED)
446 		/*
447 		 * This is a new local open on a delegation. It needs
448 		 * to have credentials so that an open can be done
449 		 * against the server during recovery.
450 		 */
451 		newnfs_copyincred(cred, &op->nfso_cred);
452 
453 	    /*
454 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
455 	     * been done on this Open successfully and a VOP_CLOSE()
456 	     * is expected for each of these.
457 	     * If error is non-zero, don't increment it, since the Open
458 	     * hasn't succeeded yet.
459 	     */
460 	    if (!error)
461 		op->nfso_opencnt++;
462 	    nfscl_openrelease(nmp, op, error, newone);
463 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
464 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
465 		error == NFSERR_BADSESSION) {
466 		(void) nfs_catnap(PZERO, error, "nfs_open");
467 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
468 		&& clidrev != 0) {
469 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
470 		retrycnt++;
471 	    }
472 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
473 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
474 	    error == NFSERR_BADSESSION ||
475 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
476 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
477 	if (error && retrycnt >= 4)
478 		error = EIO;
479 	return (error);
480 }
481 
482 /*
483  * the actual open rpc
484  */
485 int
486 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
487     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
488     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
489     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
490     int syscred, int recursed)
491 {
492 	u_int32_t *tl;
493 	struct nfsrv_descript nfsd, *nd = &nfsd;
494 	struct nfscldeleg *dp, *ndp = NULL;
495 	struct nfsvattr nfsva;
496 	u_int32_t rflags, deleg;
497 	nfsattrbit_t attrbits;
498 	int error, ret, acesize, limitby;
499 	struct nfsclsession *tsep;
500 
501 	dp = *dpp;
502 	*dpp = NULL;
503 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
504 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
505 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
506 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
507 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
508 	tsep = nfsmnt_mdssession(nmp);
509 	*tl++ = tsep->nfsess_clientid.lval[0];
510 	*tl = tsep->nfsess_clientid.lval[1];
511 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
512 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
513 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
514 	if (reclaim) {
515 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
516 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
517 		*tl = txdr_unsigned(delegtype);
518 	} else {
519 		if (dp != NULL) {
520 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
521 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
522 			if (NFSHASNFSV4N(nmp))
523 				*tl++ = 0;
524 			else
525 				*tl++ = dp->nfsdl_stateid.seqid;
526 			*tl++ = dp->nfsdl_stateid.other[0];
527 			*tl++ = dp->nfsdl_stateid.other[1];
528 			*tl = dp->nfsdl_stateid.other[2];
529 		} else {
530 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
531 		}
532 		(void) nfsm_strtom(nd, name, namelen);
533 	}
534 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
535 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
536 	NFSZERO_ATTRBIT(&attrbits);
537 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
538 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
539 	(void) nfsrv_putattrbit(nd, &attrbits);
540 	if (syscred)
541 		nd->nd_flag |= ND_USEGSSNAME;
542 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
543 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
544 	if (error)
545 		return (error);
546 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
547 	if (!nd->nd_repstat) {
548 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
549 		    6 * NFSX_UNSIGNED);
550 		op->nfso_stateid.seqid = *tl++;
551 		op->nfso_stateid.other[0] = *tl++;
552 		op->nfso_stateid.other[1] = *tl++;
553 		op->nfso_stateid.other[2] = *tl;
554 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
555 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
556 		if (error)
557 			goto nfsmout;
558 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
559 		deleg = fxdr_unsigned(u_int32_t, *tl);
560 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
561 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
562 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
563 			      NFSCLFLAGS_FIRSTDELEG))
564 				op->nfso_own->nfsow_clp->nfsc_flags |=
565 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
566 			ndp = malloc(
567 			    sizeof (struct nfscldeleg) + newfhlen,
568 			    M_NFSCLDELEG, M_WAITOK);
569 			LIST_INIT(&ndp->nfsdl_owner);
570 			LIST_INIT(&ndp->nfsdl_lock);
571 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
572 			ndp->nfsdl_fhlen = newfhlen;
573 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
574 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
575 			nfscl_lockinit(&ndp->nfsdl_rwlock);
576 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
577 			    NFSX_UNSIGNED);
578 			ndp->nfsdl_stateid.seqid = *tl++;
579 			ndp->nfsdl_stateid.other[0] = *tl++;
580 			ndp->nfsdl_stateid.other[1] = *tl++;
581 			ndp->nfsdl_stateid.other[2] = *tl++;
582 			ret = fxdr_unsigned(int, *tl);
583 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
584 				ndp->nfsdl_flags = NFSCLDL_WRITE;
585 				/*
586 				 * Indicates how much the file can grow.
587 				 */
588 				NFSM_DISSECT(tl, u_int32_t *,
589 				    3 * NFSX_UNSIGNED);
590 				limitby = fxdr_unsigned(int, *tl++);
591 				switch (limitby) {
592 				case NFSV4OPEN_LIMITSIZE:
593 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
594 					break;
595 				case NFSV4OPEN_LIMITBLOCKS:
596 					ndp->nfsdl_sizelimit =
597 					    fxdr_unsigned(u_int64_t, *tl++);
598 					ndp->nfsdl_sizelimit *=
599 					    fxdr_unsigned(u_int64_t, *tl);
600 					break;
601 				default:
602 					error = NFSERR_BADXDR;
603 					goto nfsmout;
604 				}
605 			} else {
606 				ndp->nfsdl_flags = NFSCLDL_READ;
607 			}
608 			if (ret)
609 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
610 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
611 			    &acesize, p);
612 			if (error)
613 				goto nfsmout;
614 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
615 			error = NFSERR_BADXDR;
616 			goto nfsmout;
617 		}
618 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
619 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
620 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
621 		    NULL, NULL, NULL, p, cred);
622 		if (error)
623 			goto nfsmout;
624 		if (ndp != NULL) {
625 			ndp->nfsdl_change = nfsva.na_filerev;
626 			ndp->nfsdl_modtime = nfsva.na_mtime;
627 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
628 		}
629 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
630 		    do {
631 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
632 			    cred, p);
633 			if (ret == NFSERR_DELAY)
634 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
635 		    } while (ret == NFSERR_DELAY);
636 		    error = ret;
637 		}
638 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
639 		    nfscl_assumeposixlocks)
640 		    op->nfso_posixlock = 1;
641 		else
642 		    op->nfso_posixlock = 0;
643 
644 		/*
645 		 * If the server is handing out delegations, but we didn't
646 		 * get one because an OpenConfirm was required, try the
647 		 * Open again, to get a delegation. This is a harmless no-op,
648 		 * from a server's point of view.
649 		 */
650 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
651 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
652 		    && !error && dp == NULL && ndp == NULL && !recursed) {
653 		    do {
654 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
655 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
656 			    cred, p, syscred, 1);
657 			if (ret == NFSERR_DELAY)
658 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
659 		    } while (ret == NFSERR_DELAY);
660 		    if (ret) {
661 			if (ndp != NULL) {
662 				free(ndp, M_NFSCLDELEG);
663 				ndp = NULL;
664 			}
665 			if (ret == NFSERR_STALECLIENTID ||
666 			    ret == NFSERR_STALEDONTRECOVER ||
667 			    ret == NFSERR_BADSESSION)
668 				error = ret;
669 		    }
670 		}
671 	}
672 	if (nd->nd_repstat != 0 && error == 0)
673 		error = nd->nd_repstat;
674 	if (error == NFSERR_STALECLIENTID)
675 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
676 nfsmout:
677 	if (!error)
678 		*dpp = ndp;
679 	else if (ndp != NULL)
680 		free(ndp, M_NFSCLDELEG);
681 	m_freem(nd->nd_mrep);
682 	return (error);
683 }
684 
685 /*
686  * open downgrade rpc
687  */
688 int
689 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
690     struct ucred *cred, NFSPROC_T *p)
691 {
692 	u_int32_t *tl;
693 	struct nfsrv_descript nfsd, *nd = &nfsd;
694 	int error;
695 
696 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
697 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
698 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
699 		*tl++ = 0;
700 	else
701 		*tl++ = op->nfso_stateid.seqid;
702 	*tl++ = op->nfso_stateid.other[0];
703 	*tl++ = op->nfso_stateid.other[1];
704 	*tl++ = op->nfso_stateid.other[2];
705 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
706 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
707 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
708 	error = nfscl_request(nd, vp, p, cred, NULL);
709 	if (error)
710 		return (error);
711 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
712 	if (!nd->nd_repstat) {
713 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
714 		op->nfso_stateid.seqid = *tl++;
715 		op->nfso_stateid.other[0] = *tl++;
716 		op->nfso_stateid.other[1] = *tl++;
717 		op->nfso_stateid.other[2] = *tl;
718 	}
719 	if (nd->nd_repstat && error == 0)
720 		error = nd->nd_repstat;
721 	if (error == NFSERR_STALESTATEID)
722 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
723 nfsmout:
724 	m_freem(nd->nd_mrep);
725 	return (error);
726 }
727 
728 /*
729  * V4 Close operation.
730  */
731 int
732 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
733 {
734 	struct nfsclclient *clp;
735 	int error;
736 
737 	if (vnode_vtype(vp) != VREG)
738 		return (0);
739 	if (doclose)
740 		error = nfscl_doclose(vp, &clp, p);
741 	else
742 		error = nfscl_getclose(vp, &clp);
743 	if (error)
744 		return (error);
745 
746 	nfscl_clientrelease(clp);
747 	return (0);
748 }
749 
750 /*
751  * Close the open.
752  */
753 void
754 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
755 {
756 	struct nfsrv_descript nfsd, *nd = &nfsd;
757 	struct nfscllockowner *lp, *nlp;
758 	struct nfscllock *lop, *nlop;
759 	struct ucred *tcred;
760 	u_int64_t off = 0, len = 0;
761 	u_int32_t type = NFSV4LOCKT_READ;
762 	int error, do_unlock, trycnt;
763 
764 	tcred = newnfs_getcred();
765 	newnfs_copycred(&op->nfso_cred, tcred);
766 	/*
767 	 * (Theoretically this could be done in the same
768 	 *  compound as the close, but having multiple
769 	 *  sequenced Ops in the same compound might be
770 	 *  too scary for some servers.)
771 	 */
772 	if (op->nfso_posixlock) {
773 		off = 0;
774 		len = NFS64BITSSET;
775 		type = NFSV4LOCKT_READ;
776 	}
777 
778 	/*
779 	 * Since this function is only called from VOP_INACTIVE(), no
780 	 * other thread will be manipulating this Open. As such, the
781 	 * lock lists are not being changed by other threads, so it should
782 	 * be safe to do this without locking.
783 	 */
784 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
785 		do_unlock = 1;
786 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
787 			if (op->nfso_posixlock == 0) {
788 				off = lop->nfslo_first;
789 				len = lop->nfslo_end - lop->nfslo_first;
790 				if (lop->nfslo_type == F_WRLCK)
791 					type = NFSV4LOCKT_WRITE;
792 				else
793 					type = NFSV4LOCKT_READ;
794 			}
795 			if (do_unlock) {
796 				trycnt = 0;
797 				do {
798 					error = nfsrpc_locku(nd, nmp, lp, off,
799 					    len, type, tcred, p, 0);
800 					if ((nd->nd_repstat == NFSERR_GRACE ||
801 					    nd->nd_repstat == NFSERR_DELAY) &&
802 					    error == 0)
803 						(void) nfs_catnap(PZERO,
804 						    (int)nd->nd_repstat,
805 						    "nfs_close");
806 				} while ((nd->nd_repstat == NFSERR_GRACE ||
807 				    nd->nd_repstat == NFSERR_DELAY) &&
808 				    error == 0 && trycnt++ < 5);
809 				if (op->nfso_posixlock)
810 					do_unlock = 0;
811 			}
812 			nfscl_freelock(lop, 0);
813 		}
814 		/*
815 		 * Do a ReleaseLockOwner.
816 		 * The lock owner name nfsl_owner may be used by other opens for
817 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
818 		 * puts on the wire has the file handle for this file appended
819 		 * to it, so it can be done now.
820 		 */
821 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
822 		    lp->nfsl_open->nfso_fhlen, tcred, p);
823 	}
824 
825 	/*
826 	 * There could be other Opens for different files on the same
827 	 * OpenOwner, so locking is required.
828 	 */
829 	NFSLOCKCLSTATE();
830 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
831 	NFSUNLOCKCLSTATE();
832 	do {
833 		error = nfscl_tryclose(op, tcred, nmp, p);
834 		if (error == NFSERR_GRACE)
835 			(void) nfs_catnap(PZERO, error, "nfs_close");
836 	} while (error == NFSERR_GRACE);
837 	NFSLOCKCLSTATE();
838 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
839 
840 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
841 		nfscl_freelockowner(lp, 0);
842 	nfscl_freeopen(op, 0);
843 	NFSUNLOCKCLSTATE();
844 	NFSFREECRED(tcred);
845 }
846 
847 /*
848  * The actual Close RPC.
849  */
850 int
851 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
852     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
853     int syscred)
854 {
855 	u_int32_t *tl;
856 	int error;
857 
858 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
859 	    op->nfso_fhlen, NULL, NULL, 0, 0);
860 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
861 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
862 	if (NFSHASNFSV4N(nmp))
863 		*tl++ = 0;
864 	else
865 		*tl++ = op->nfso_stateid.seqid;
866 	*tl++ = op->nfso_stateid.other[0];
867 	*tl++ = op->nfso_stateid.other[1];
868 	*tl = op->nfso_stateid.other[2];
869 	if (syscred)
870 		nd->nd_flag |= ND_USEGSSNAME;
871 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
872 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
873 	if (error)
874 		return (error);
875 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
876 	if (nd->nd_repstat == 0)
877 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
878 	error = nd->nd_repstat;
879 	if (error == NFSERR_STALESTATEID)
880 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
881 nfsmout:
882 	m_freem(nd->nd_mrep);
883 	return (error);
884 }
885 
886 /*
887  * V4 Open Confirm RPC.
888  */
889 int
890 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
891     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
892 {
893 	u_int32_t *tl;
894 	struct nfsrv_descript nfsd, *nd = &nfsd;
895 	struct nfsmount *nmp;
896 	int error;
897 
898 	nmp = VFSTONFS(vp->v_mount);
899 	if (NFSHASNFSV4N(nmp))
900 		return (0);		/* No confirmation for NFSv4.1. */
901 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
902 	    0, 0);
903 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
904 	*tl++ = op->nfso_stateid.seqid;
905 	*tl++ = op->nfso_stateid.other[0];
906 	*tl++ = op->nfso_stateid.other[1];
907 	*tl++ = op->nfso_stateid.other[2];
908 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
909 	error = nfscl_request(nd, vp, p, cred, NULL);
910 	if (error)
911 		return (error);
912 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
913 	if (!nd->nd_repstat) {
914 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
915 		op->nfso_stateid.seqid = *tl++;
916 		op->nfso_stateid.other[0] = *tl++;
917 		op->nfso_stateid.other[1] = *tl++;
918 		op->nfso_stateid.other[2] = *tl;
919 	}
920 	error = nd->nd_repstat;
921 	if (error == NFSERR_STALESTATEID)
922 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
923 nfsmout:
924 	m_freem(nd->nd_mrep);
925 	return (error);
926 }
927 
928 /*
929  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
930  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
931  */
932 int
933 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
934     bool *retokp, struct ucred *cred, NFSPROC_T *p)
935 {
936 	u_int32_t *tl;
937 	struct nfsrv_descript nfsd;
938 	struct nfsrv_descript *nd = &nfsd;
939 	nfsattrbit_t attrbits;
940 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
941 	u_short port;
942 	int error, isinet6 = 0, callblen;
943 	nfsquad_t confirm;
944 	u_int32_t lease;
945 	static u_int32_t rev = 0;
946 	struct nfsclds *dsp, *odsp;
947 	struct in6_addr a6;
948 	struct nfsclsession *tsep;
949 	struct rpc_reconupcall recon;
950 	struct nfscl_reconarg *rcp;
951 
952 	if (nfsboottime.tv_sec == 0)
953 		NFSSETBOOTTIME(nfsboottime);
954 	if (NFSHASNFSV4N(nmp)) {
955 		error = NFSERR_BADSESSION;
956 		odsp = dsp = NULL;
957 		if (retokp != NULL) {
958 			NFSLOCKMNT(nmp);
959 			odsp = TAILQ_FIRST(&nmp->nm_sess);
960 			NFSUNLOCKMNT(nmp);
961 		}
962 		if (odsp != NULL) {
963 			/*
964 			 * When a session already exists, first try a
965 			 * CreateSession with the extant ClientID.
966 			 */
967 			dsp = malloc(sizeof(struct nfsclds) +
968 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
969 			    M_WAITOK | M_ZERO);
970 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
971 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
972 			dsp->nfsclds_sess.nfsess_clientid =
973 			    odsp->nfsclds_sess.nfsess_clientid;
974 			dsp->nfsclds_sess.nfsess_sequenceid =
975 			    odsp->nfsclds_sess.nfsess_sequenceid;
976 			dsp->nfsclds_flags = odsp->nfsclds_flags;
977 			if (dsp->nfsclds_servownlen > 0)
978 				memcpy(dsp->nfsclds_serverown,
979 				    odsp->nfsclds_serverown,
980 				    dsp->nfsclds_servownlen + 1);
981 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
982 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
983 			    NULL, MTX_DEF);
984 			nfscl_initsessionslots(&dsp->nfsclds_sess);
985 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
986 			    &nmp->nm_sockreq, NULL,
987 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
988 			NFSCL_DEBUG(1, "create session for extant "
989 			    "ClientID=%d\n", error);
990 			if (error != 0) {
991 				nfscl_freenfsclds(dsp);
992 				dsp = NULL;
993 				/*
994 				 * If *retokp is true, return any error other
995 				 * than NFSERR_STALECLIENTID,
996 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
997 				 * so that nfscl_recover() will not loop.
998 				 */
999 				if (*retokp)
1000 					return (NFSERR_IO);
1001 			} else
1002 				*retokp = true;
1003 		} else if (retokp != NULL && *retokp)
1004 			return (NFSERR_IO);
1005 		if (error != 0) {
1006 			/*
1007 			 * Either there was no previous session or the
1008 			 * CreateSession attempt failed, so...
1009 			 * do an ExchangeID followed by the CreateSession.
1010 			 */
1011 			clp->nfsc_rev = rev++;
1012 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1013 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1014 			    cred, p);
1015 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1016 			if (error == 0)
1017 				error = nfsrpc_createsession(nmp,
1018 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1019 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1020 				    cred, p);
1021 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1022 		}
1023 		if (error == 0) {
1024 			/*
1025 			 * If the session supports a backchannel, set up
1026 			 * the BindConnectionToSession call in the krpc
1027 			 * so that it is done on a reconnection.
1028 			 */
1029 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1030 				rcp = mem_alloc(sizeof(*rcp));
1031 				rcp->minorvers = nmp->nm_minorvers;
1032 				memcpy(rcp->sessionid,
1033 				    dsp->nfsclds_sess.nfsess_sessionid,
1034 				    NFSX_V4SESSIONID);
1035 				recon.call = nfsrpc_bindconnsess;
1036 				recon.arg = rcp;
1037 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1038 				    &recon);
1039 			}
1040 
1041 			NFSLOCKMNT(nmp);
1042 			/*
1043 			 * The old sessions cannot be safely free'd
1044 			 * here, since they may still be used by
1045 			 * in-progress RPCs.
1046 			 */
1047 			tsep = NULL;
1048 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
1049 				tsep = NFSMNT_MDSSESSION(nmp);
1050 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1051 			    nfsclds_list);
1052 			/*
1053 			 * Wake up RPCs waiting for a slot on the
1054 			 * old session. These will then fail with
1055 			 * NFSERR_BADSESSION and be retried with the
1056 			 * new session by nfsv4_setsequence().
1057 			 * Also wakeup() processes waiting for the
1058 			 * new session.
1059 			 */
1060 			if (tsep != NULL)
1061 				wakeup(&tsep->nfsess_slots);
1062 			wakeup(&nmp->nm_sess);
1063 			NFSUNLOCKMNT(nmp);
1064 		} else if (dsp != NULL)
1065 			nfscl_freenfsclds(dsp);
1066 		if (error == 0 && reclaim == 0) {
1067 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1068 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1069 			if (error == NFSERR_COMPLETEALREADY ||
1070 			    error == NFSERR_NOTSUPP)
1071 				/* Ignore this error. */
1072 				error = 0;
1073 		}
1074 		return (error);
1075 	} else if (retokp != NULL && *retokp)
1076 		return (NFSERR_IO);
1077 	clp->nfsc_rev = rev++;
1078 
1079 	/*
1080 	 * Allocate a single session structure for NFSv4.0, because some of
1081 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1082 	 */
1083 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1084 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1085 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1086 	NFSLOCKMNT(nmp);
1087 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1088 	tsep = NFSMNT_MDSSESSION(nmp);
1089 	NFSUNLOCKMNT(nmp);
1090 
1091 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
1092 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1093 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1094 	*tl = txdr_unsigned(clp->nfsc_rev);
1095 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1096 
1097 	/*
1098 	 * set up the callback address
1099 	 */
1100 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1101 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1102 	callblen = strlen(nfsv4_callbackaddr);
1103 	if (callblen == 0)
1104 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1105 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1106 	    (callblen > 0 || cp != NULL)) {
1107 		port = htons(nfsv4_cbport);
1108 		cp2 = (u_int8_t *)&port;
1109 #ifdef INET6
1110 		if ((callblen > 0 &&
1111 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1112 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1113 
1114 			(void) nfsm_strtom(nd, "tcp6", 4);
1115 			if (callblen == 0) {
1116 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1117 				ip6add = ip6buf;
1118 			} else {
1119 				ip6add = nfsv4_callbackaddr;
1120 			}
1121 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1122 			    ip6add, cp2[0], cp2[1]);
1123 		} else
1124 #endif
1125 		{
1126 			(void) nfsm_strtom(nd, "tcp", 3);
1127 			if (callblen == 0)
1128 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1129 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1130 				    cp[2], cp[3], cp2[0], cp2[1]);
1131 			else
1132 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1133 				    "%s.%d.%d", nfsv4_callbackaddr,
1134 				    cp2[0], cp2[1]);
1135 		}
1136 		(void) nfsm_strtom(nd, addr, strlen(addr));
1137 	} else {
1138 		(void) nfsm_strtom(nd, "tcp", 3);
1139 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1140 	}
1141 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1142 	*tl = txdr_unsigned(clp->nfsc_cbident);
1143 	nd->nd_flag |= ND_USEGSSNAME;
1144 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1145 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1146 	if (error)
1147 		return (error);
1148 	if (nd->nd_repstat == 0) {
1149 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1150 	    tsep->nfsess_clientid.lval[0] = *tl++;
1151 	    tsep->nfsess_clientid.lval[1] = *tl++;
1152 	    confirm.lval[0] = *tl++;
1153 	    confirm.lval[1] = *tl;
1154 	    m_freem(nd->nd_mrep);
1155 	    nd->nd_mrep = NULL;
1156 
1157 	    /*
1158 	     * and confirm it.
1159 	     */
1160 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1161 		NULL, 0, 0);
1162 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1163 	    *tl++ = tsep->nfsess_clientid.lval[0];
1164 	    *tl++ = tsep->nfsess_clientid.lval[1];
1165 	    *tl++ = confirm.lval[0];
1166 	    *tl = confirm.lval[1];
1167 	    nd->nd_flag |= ND_USEGSSNAME;
1168 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1169 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1170 	    if (error)
1171 		return (error);
1172 	    m_freem(nd->nd_mrep);
1173 	    nd->nd_mrep = NULL;
1174 	    if (nd->nd_repstat == 0) {
1175 		nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
1176 		    nmp->nm_fhsize, NULL, NULL, 0, 0);
1177 		NFSZERO_ATTRBIT(&attrbits);
1178 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1179 		(void) nfsrv_putattrbit(nd, &attrbits);
1180 		nd->nd_flag |= ND_USEGSSNAME;
1181 		error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1182 		    cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1183 		if (error)
1184 		    return (error);
1185 		if (nd->nd_repstat == 0) {
1186 		    error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
1187 			NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1188 		    if (error)
1189 			goto nfsmout;
1190 		    clp->nfsc_renew = NFSCL_RENEW(lease);
1191 		    clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1192 		    clp->nfsc_clientidrev++;
1193 		    if (clp->nfsc_clientidrev == 0)
1194 			clp->nfsc_clientidrev++;
1195 		}
1196 	    }
1197 	}
1198 	error = nd->nd_repstat;
1199 nfsmout:
1200 	m_freem(nd->nd_mrep);
1201 	return (error);
1202 }
1203 
1204 /*
1205  * nfs getattr call.
1206  */
1207 int
1208 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1209     struct nfsvattr *nap, void *stuff)
1210 {
1211 	struct nfsrv_descript nfsd, *nd = &nfsd;
1212 	int error;
1213 	nfsattrbit_t attrbits;
1214 
1215 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1216 	if (nd->nd_flag & ND_NFSV4) {
1217 		NFSGETATTR_ATTRBIT(&attrbits);
1218 		(void) nfsrv_putattrbit(nd, &attrbits);
1219 	}
1220 	error = nfscl_request(nd, vp, p, cred, stuff);
1221 	if (error)
1222 		return (error);
1223 	if (!nd->nd_repstat)
1224 		error = nfsm_loadattr(nd, nap);
1225 	else
1226 		error = nd->nd_repstat;
1227 	m_freem(nd->nd_mrep);
1228 	return (error);
1229 }
1230 
1231 /*
1232  * nfs getattr call with non-vnode arguemnts.
1233  */
1234 int
1235 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1236     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1237     uint32_t *leasep)
1238 {
1239 	struct nfsrv_descript nfsd, *nd = &nfsd;
1240 	int error, vers = NFS_VER2;
1241 	nfsattrbit_t attrbits;
1242 
1243 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1244 	if (nd->nd_flag & ND_NFSV4) {
1245 		vers = NFS_VER4;
1246 		NFSGETATTR_ATTRBIT(&attrbits);
1247 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1248 		(void) nfsrv_putattrbit(nd, &attrbits);
1249 	} else if (nd->nd_flag & ND_NFSV3) {
1250 		vers = NFS_VER3;
1251 	}
1252 	if (syscred)
1253 		nd->nd_flag |= ND_USEGSSNAME;
1254 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1255 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1256 	if (error)
1257 		return (error);
1258 	if (nd->nd_repstat == 0) {
1259 		if ((nd->nd_flag & ND_NFSV4) != 0)
1260 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1261 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1262 			    NULL, NULL);
1263 		else
1264 			error = nfsm_loadattr(nd, nap);
1265 	} else
1266 		error = nd->nd_repstat;
1267 	m_freem(nd->nd_mrep);
1268 	return (error);
1269 }
1270 
1271 /*
1272  * Do an nfs setattr operation.
1273  */
1274 int
1275 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1276     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1277     void *stuff)
1278 {
1279 	int error, expireret = 0, openerr, retrycnt;
1280 	u_int32_t clidrev = 0, mode;
1281 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1282 	struct nfsfh *nfhp;
1283 	nfsv4stateid_t stateid;
1284 	void *lckp;
1285 
1286 	if (nmp->nm_clp != NULL)
1287 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1288 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1289 		mode = NFSV4OPEN_ACCESSWRITE;
1290 	else
1291 		mode = NFSV4OPEN_ACCESSREAD;
1292 	retrycnt = 0;
1293 	do {
1294 		lckp = NULL;
1295 		openerr = 1;
1296 		if (NFSHASNFSV4(nmp)) {
1297 			nfhp = VTONFS(vp)->n_fhp;
1298 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1299 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1300 			if (error && vnode_vtype(vp) == VREG &&
1301 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1302 			     nfstest_openallsetattr)) {
1303 				/*
1304 				 * No Open stateid, so try and open the file
1305 				 * now.
1306 				 */
1307 				if (mode == NFSV4OPEN_ACCESSWRITE)
1308 					openerr = nfsrpc_open(vp, FWRITE, cred,
1309 					    p);
1310 				else
1311 					openerr = nfsrpc_open(vp, FREAD, cred,
1312 					    p);
1313 				if (!openerr)
1314 					(void) nfscl_getstateid(vp,
1315 					    nfhp->nfh_fh, nfhp->nfh_len,
1316 					    mode, 0, cred, p, &stateid, &lckp);
1317 			}
1318 		}
1319 		if (vap != NULL)
1320 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1321 			    rnap, attrflagp, stuff);
1322 		else
1323 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1324 			    stuff);
1325 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1326 			NFSLOCKMNT(nmp);
1327 			nmp->nm_state |= NFSSTA_OPENMODE;
1328 			NFSUNLOCKMNT(nmp);
1329 		}
1330 		if (error == NFSERR_STALESTATEID)
1331 			nfscl_initiate_recovery(nmp->nm_clp);
1332 		if (lckp != NULL)
1333 			nfscl_lockderef(lckp);
1334 		if (!openerr)
1335 			(void) nfsrpc_close(vp, 0, p);
1336 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1337 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1338 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1339 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1340 		} else if ((error == NFSERR_EXPIRED ||
1341 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1342 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1343 		}
1344 		retrycnt++;
1345 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1346 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1347 	    error == NFSERR_BADSESSION ||
1348 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1349 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1350 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1351 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1352 	     retrycnt < 4));
1353 	if (error && retrycnt >= 4)
1354 		error = EIO;
1355 	return (error);
1356 }
1357 
1358 static int
1359 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1360     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1361     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1362 {
1363 	u_int32_t *tl;
1364 	struct nfsrv_descript nfsd, *nd = &nfsd;
1365 	int error;
1366 	nfsattrbit_t attrbits;
1367 
1368 	*attrflagp = 0;
1369 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1370 	if (nd->nd_flag & ND_NFSV4)
1371 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1372 	vap->va_type = vnode_vtype(vp);
1373 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1374 	if (nd->nd_flag & ND_NFSV3) {
1375 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1376 		*tl = newnfs_false;
1377 	} else if (nd->nd_flag & ND_NFSV4) {
1378 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1379 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1380 		NFSGETATTR_ATTRBIT(&attrbits);
1381 		(void) nfsrv_putattrbit(nd, &attrbits);
1382 	}
1383 	error = nfscl_request(nd, vp, p, cred, stuff);
1384 	if (error)
1385 		return (error);
1386 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1387 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1388 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1389 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1390 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1391 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1392 	m_freem(nd->nd_mrep);
1393 	if (nd->nd_repstat && !error)
1394 		error = nd->nd_repstat;
1395 	return (error);
1396 }
1397 
1398 /*
1399  * nfs lookup rpc
1400  */
1401 int
1402 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1403     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1404     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1405 {
1406 	u_int32_t *tl;
1407 	struct nfsrv_descript nfsd, *nd = &nfsd;
1408 	struct nfsmount *nmp;
1409 	struct nfsnode *np;
1410 	struct nfsfh *nfhp;
1411 	nfsattrbit_t attrbits;
1412 	int error = 0, lookupp = 0;
1413 
1414 	*attrflagp = 0;
1415 	*dattrflagp = 0;
1416 	if (vnode_vtype(dvp) != VDIR)
1417 		return (ENOTDIR);
1418 	nmp = VFSTONFS(dvp->v_mount);
1419 	if (len > NFS_MAXNAMLEN)
1420 		return (ENAMETOOLONG);
1421 	if (NFSHASNFSV4(nmp) && len == 1 &&
1422 		name[0] == '.') {
1423 		/*
1424 		 * Just return the current dir's fh.
1425 		 */
1426 		np = VTONFS(dvp);
1427 		nfhp = malloc(sizeof (struct nfsfh) +
1428 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1429 		nfhp->nfh_len = np->n_fhp->nfh_len;
1430 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1431 		*nfhpp = nfhp;
1432 		return (0);
1433 	}
1434 	if (NFSHASNFSV4(nmp) && len == 2 &&
1435 		name[0] == '.' && name[1] == '.') {
1436 		lookupp = 1;
1437 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1438 	} else {
1439 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1440 		(void) nfsm_strtom(nd, name, len);
1441 	}
1442 	if (nd->nd_flag & ND_NFSV4) {
1443 		NFSGETATTR_ATTRBIT(&attrbits);
1444 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1445 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1446 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1447 		(void) nfsrv_putattrbit(nd, &attrbits);
1448 	}
1449 	error = nfscl_request(nd, dvp, p, cred, stuff);
1450 	if (error)
1451 		return (error);
1452 	if (nd->nd_repstat) {
1453 		/*
1454 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1455 		 * the lookup is at the root of an fs, so return this dir.
1456 		 */
1457 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1458 		    np = VTONFS(dvp);
1459 		    nfhp = malloc(sizeof (struct nfsfh) +
1460 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1461 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1462 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1463 		    *nfhpp = nfhp;
1464 		    m_freem(nd->nd_mrep);
1465 		    return (0);
1466 		}
1467 		if (nd->nd_flag & ND_NFSV3)
1468 		    error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1469 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1470 		    ND_NFSV4) {
1471 			/* Load the directory attributes. */
1472 			error = nfsm_loadattr(nd, dnap);
1473 			if (error == 0)
1474 				*dattrflagp = 1;
1475 		}
1476 		goto nfsmout;
1477 	}
1478 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1479 		/* Load the directory attributes. */
1480 		error = nfsm_loadattr(nd, dnap);
1481 		if (error != 0)
1482 			goto nfsmout;
1483 		*dattrflagp = 1;
1484 		/* Skip over the Lookup and GetFH operation status values. */
1485 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1486 	}
1487 	error = nfsm_getfh(nd, nfhpp);
1488 	if (error)
1489 		goto nfsmout;
1490 
1491 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1492 	if ((nd->nd_flag & ND_NFSV3) && !error)
1493 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1494 nfsmout:
1495 	m_freem(nd->nd_mrep);
1496 	if (!error && nd->nd_repstat)
1497 		error = nd->nd_repstat;
1498 	return (error);
1499 }
1500 
1501 /*
1502  * Do a readlink rpc.
1503  */
1504 int
1505 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1506     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1507 {
1508 	u_int32_t *tl;
1509 	struct nfsrv_descript nfsd, *nd = &nfsd;
1510 	struct nfsnode *np = VTONFS(vp);
1511 	nfsattrbit_t attrbits;
1512 	int error, len, cangetattr = 1;
1513 
1514 	*attrflagp = 0;
1515 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1516 	if (nd->nd_flag & ND_NFSV4) {
1517 		/*
1518 		 * And do a Getattr op.
1519 		 */
1520 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1521 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1522 		NFSGETATTR_ATTRBIT(&attrbits);
1523 		(void) nfsrv_putattrbit(nd, &attrbits);
1524 	}
1525 	error = nfscl_request(nd, vp, p, cred, stuff);
1526 	if (error)
1527 		return (error);
1528 	if (nd->nd_flag & ND_NFSV3)
1529 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1530 	if (!nd->nd_repstat && !error) {
1531 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1532 		/*
1533 		 * This seems weird to me, but must have been added to
1534 		 * FreeBSD for some reason. The only thing I can think of
1535 		 * is that there was/is some server that replies with
1536 		 * more link data than it should?
1537 		 */
1538 		if (len == NFS_MAXPATHLEN) {
1539 			NFSLOCKNODE(np);
1540 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1541 				len = np->n_size;
1542 				cangetattr = 0;
1543 			}
1544 			NFSUNLOCKNODE(np);
1545 		}
1546 		error = nfsm_mbufuio(nd, uiop, len);
1547 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1548 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1549 	}
1550 	if (nd->nd_repstat && !error)
1551 		error = nd->nd_repstat;
1552 nfsmout:
1553 	m_freem(nd->nd_mrep);
1554 	return (error);
1555 }
1556 
1557 /*
1558  * Read operation.
1559  */
1560 int
1561 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1562     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1563 {
1564 	int error, expireret = 0, retrycnt;
1565 	u_int32_t clidrev = 0;
1566 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1567 	struct nfsnode *np = VTONFS(vp);
1568 	struct ucred *newcred;
1569 	struct nfsfh *nfhp = NULL;
1570 	nfsv4stateid_t stateid;
1571 	void *lckp;
1572 
1573 	if (nmp->nm_clp != NULL)
1574 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1575 	newcred = cred;
1576 	if (NFSHASNFSV4(nmp)) {
1577 		nfhp = np->n_fhp;
1578 		newcred = NFSNEWCRED(cred);
1579 	}
1580 	retrycnt = 0;
1581 	do {
1582 		lckp = NULL;
1583 		if (NFSHASNFSV4(nmp))
1584 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1585 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1586 			    &lckp);
1587 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1588 		    attrflagp, stuff);
1589 		if (error == NFSERR_OPENMODE) {
1590 			NFSLOCKMNT(nmp);
1591 			nmp->nm_state |= NFSSTA_OPENMODE;
1592 			NFSUNLOCKMNT(nmp);
1593 		}
1594 		if (error == NFSERR_STALESTATEID)
1595 			nfscl_initiate_recovery(nmp->nm_clp);
1596 		if (lckp != NULL)
1597 			nfscl_lockderef(lckp);
1598 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1599 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1600 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1601 			(void) nfs_catnap(PZERO, error, "nfs_read");
1602 		} else if ((error == NFSERR_EXPIRED ||
1603 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1604 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1605 		}
1606 		retrycnt++;
1607 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1608 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1609 	    error == NFSERR_BADSESSION ||
1610 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1611 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1612 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1613 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1614 	if (error && retrycnt >= 4)
1615 		error = EIO;
1616 	if (NFSHASNFSV4(nmp))
1617 		NFSFREECRED(newcred);
1618 	return (error);
1619 }
1620 
1621 /*
1622  * The actual read RPC.
1623  */
1624 static int
1625 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1626     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1627     int *attrflagp, void *stuff)
1628 {
1629 	u_int32_t *tl;
1630 	int error = 0, len, retlen, tsiz, eof = 0;
1631 	struct nfsrv_descript nfsd;
1632 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1633 	struct nfsrv_descript *nd = &nfsd;
1634 	int rsize;
1635 	off_t tmp_off;
1636 
1637 	*attrflagp = 0;
1638 	tsiz = uiop->uio_resid;
1639 	tmp_off = uiop->uio_offset + tsiz;
1640 	NFSLOCKMNT(nmp);
1641 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1642 		NFSUNLOCKMNT(nmp);
1643 		return (EFBIG);
1644 	}
1645 	rsize = nmp->nm_rsize;
1646 	NFSUNLOCKMNT(nmp);
1647 	nd->nd_mrep = NULL;
1648 	while (tsiz > 0) {
1649 		*attrflagp = 0;
1650 		len = (tsiz > rsize) ? rsize : tsiz;
1651 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1652 		if (nd->nd_flag & ND_NFSV4)
1653 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1654 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1655 		if (nd->nd_flag & ND_NFSV2) {
1656 			*tl++ = txdr_unsigned(uiop->uio_offset);
1657 			*tl++ = txdr_unsigned(len);
1658 			*tl = 0;
1659 		} else {
1660 			txdr_hyper(uiop->uio_offset, tl);
1661 			*(tl + 2) = txdr_unsigned(len);
1662 		}
1663 		/*
1664 		 * Since I can't do a Getattr for NFSv4 for Write, there
1665 		 * doesn't seem any point in doing one here, either.
1666 		 * (See the comment in nfsrpc_writerpc() for more info.)
1667 		 */
1668 		error = nfscl_request(nd, vp, p, cred, stuff);
1669 		if (error)
1670 			return (error);
1671 		if (nd->nd_flag & ND_NFSV3) {
1672 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1673 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1674 			error = nfsm_loadattr(nd, nap);
1675 			if (!error)
1676 				*attrflagp = 1;
1677 		}
1678 		if (nd->nd_repstat || error) {
1679 			if (!error)
1680 				error = nd->nd_repstat;
1681 			goto nfsmout;
1682 		}
1683 		if (nd->nd_flag & ND_NFSV3) {
1684 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1685 			eof = fxdr_unsigned(int, *(tl + 1));
1686 		} else if (nd->nd_flag & ND_NFSV4) {
1687 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1688 			eof = fxdr_unsigned(int, *tl);
1689 		}
1690 		NFSM_STRSIZ(retlen, len);
1691 		error = nfsm_mbufuio(nd, uiop, retlen);
1692 		if (error)
1693 			goto nfsmout;
1694 		m_freem(nd->nd_mrep);
1695 		nd->nd_mrep = NULL;
1696 		tsiz -= retlen;
1697 		if (!(nd->nd_flag & ND_NFSV2)) {
1698 			if (eof || retlen == 0)
1699 				tsiz = 0;
1700 		} else if (retlen < len)
1701 			tsiz = 0;
1702 	}
1703 	return (0);
1704 nfsmout:
1705 	if (nd->nd_mrep != NULL)
1706 		m_freem(nd->nd_mrep);
1707 	return (error);
1708 }
1709 
1710 /*
1711  * nfs write operation
1712  * When called_from_strategy != 0, it should return EIO for an error that
1713  * indicates recovery is in progress, so that the buffer will be left
1714  * dirty and be written back to the server later. If it loops around,
1715  * the recovery thread could get stuck waiting for the buffer and recovery
1716  * will then deadlock.
1717  */
1718 int
1719 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1720     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1721     void *stuff, int called_from_strategy)
1722 {
1723 	int error, expireret = 0, retrycnt, nostateid;
1724 	u_int32_t clidrev = 0;
1725 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1726 	struct nfsnode *np = VTONFS(vp);
1727 	struct ucred *newcred;
1728 	struct nfsfh *nfhp = NULL;
1729 	nfsv4stateid_t stateid;
1730 	void *lckp;
1731 
1732 	*must_commit = 0;
1733 	if (nmp->nm_clp != NULL)
1734 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1735 	newcred = cred;
1736 	if (NFSHASNFSV4(nmp)) {
1737 		newcred = NFSNEWCRED(cred);
1738 		nfhp = np->n_fhp;
1739 	}
1740 	retrycnt = 0;
1741 	do {
1742 		lckp = NULL;
1743 		nostateid = 0;
1744 		if (NFSHASNFSV4(nmp)) {
1745 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1746 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1747 			    &lckp);
1748 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1749 			    stateid.other[2] == 0) {
1750 				nostateid = 1;
1751 				NFSCL_DEBUG(1, "stateid0 in write\n");
1752 			}
1753 		}
1754 
1755 		/*
1756 		 * If there is no stateid for NFSv4, it means this is an
1757 		 * extraneous write after close. Basically a poorly
1758 		 * implemented buffer cache. Just don't do the write.
1759 		 */
1760 		if (nostateid)
1761 			error = 0;
1762 		else
1763 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1764 			    newcred, &stateid, p, nap, attrflagp, stuff);
1765 		if (error == NFSERR_STALESTATEID)
1766 			nfscl_initiate_recovery(nmp->nm_clp);
1767 		if (lckp != NULL)
1768 			nfscl_lockderef(lckp);
1769 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1770 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1771 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1772 			(void) nfs_catnap(PZERO, error, "nfs_write");
1773 		} else if ((error == NFSERR_EXPIRED ||
1774 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1775 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1776 		}
1777 		retrycnt++;
1778 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1779 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1780 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1781 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1782 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1783 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1784 	if (error != 0 && (retrycnt >= 4 ||
1785 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1786 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1787 		error = EIO;
1788 	if (NFSHASNFSV4(nmp))
1789 		NFSFREECRED(newcred);
1790 	return (error);
1791 }
1792 
1793 /*
1794  * The actual write RPC.
1795  */
1796 static int
1797 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1798     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1799     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1800 {
1801 	u_int32_t *tl;
1802 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1803 	struct nfsnode *np = VTONFS(vp);
1804 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1805 	int wccflag = 0, wsize;
1806 	int32_t backup;
1807 	struct nfsrv_descript nfsd;
1808 	struct nfsrv_descript *nd = &nfsd;
1809 	nfsattrbit_t attrbits;
1810 	off_t tmp_off;
1811 
1812 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1813 	*attrflagp = 0;
1814 	tsiz = uiop->uio_resid;
1815 	tmp_off = uiop->uio_offset + tsiz;
1816 	NFSLOCKMNT(nmp);
1817 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1818 		NFSUNLOCKMNT(nmp);
1819 		return (EFBIG);
1820 	}
1821 	wsize = nmp->nm_wsize;
1822 	NFSUNLOCKMNT(nmp);
1823 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1824 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1825 	while (tsiz > 0) {
1826 		*attrflagp = 0;
1827 		len = (tsiz > wsize) ? wsize : tsiz;
1828 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1829 		if (nd->nd_flag & ND_NFSV4) {
1830 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1831 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1832 			txdr_hyper(uiop->uio_offset, tl);
1833 			tl += 2;
1834 			*tl++ = txdr_unsigned(*iomode);
1835 			*tl = txdr_unsigned(len);
1836 		} else if (nd->nd_flag & ND_NFSV3) {
1837 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1838 			txdr_hyper(uiop->uio_offset, tl);
1839 			tl += 2;
1840 			*tl++ = txdr_unsigned(len);
1841 			*tl++ = txdr_unsigned(*iomode);
1842 			*tl = txdr_unsigned(len);
1843 		} else {
1844 			u_int32_t x;
1845 
1846 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1847 			/*
1848 			 * Not sure why someone changed this, since the
1849 			 * RFC clearly states that "beginoffset" and
1850 			 * "totalcount" are ignored, but it wouldn't
1851 			 * surprise me if there's a busted server out there.
1852 			 */
1853 			/* Set both "begin" and "current" to non-garbage. */
1854 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1855 			*tl++ = x;      /* "begin offset" */
1856 			*tl++ = x;      /* "current offset" */
1857 			x = txdr_unsigned(len);
1858 			*tl++ = x;      /* total to this offset */
1859 			*tl = x;        /* size of this write */
1860 		}
1861 		nfsm_uiombuf(nd, uiop, len);
1862 		/*
1863 		 * Although it is tempting to do a normal Getattr Op in the
1864 		 * NFSv4 compound, the result can be a nearly hung client
1865 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1866 		 * It occurs when the client can't map either the Owner or
1867 		 * Owner_group name in the Getattr reply to a uid/gid. When
1868 		 * there is a cache miss, the kernel does an upcall to the
1869 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1870 		 * or /etc/group file. It can then block in getnewbuf(),
1871 		 * waiting for dirty writes to be pushed to the NFS server.
1872 		 * The only reason this doesn't result in a complete
1873 		 * deadlock, is that the upcall times out and allows
1874 		 * the write to complete. However, progress is so slow
1875 		 * that it might just as well be deadlocked.
1876 		 * As such, we get the rest of the attributes, but not
1877 		 * Owner or Owner_group.
1878 		 * nb: nfscl_loadattrcache() needs to be told that these
1879 		 *     partial attributes from a write rpc are being
1880 		 *     passed in, via a argument flag.
1881 		 */
1882 		if (nd->nd_flag & ND_NFSV4) {
1883 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
1884 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1885 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
1886 			(void) nfsrv_putattrbit(nd, &attrbits);
1887 		}
1888 		error = nfscl_request(nd, vp, p, cred, stuff);
1889 		if (error)
1890 			return (error);
1891 		if (nd->nd_repstat) {
1892 			/*
1893 			 * In case the rpc gets retried, roll
1894 			 * the uio fields changed by nfsm_uiombuf()
1895 			 * back.
1896 			 */
1897 			uiop->uio_offset -= len;
1898 			uiop->uio_resid += len;
1899 			uiop->uio_iov->iov_base =
1900 			    (char *)uiop->uio_iov->iov_base - len;
1901 			uiop->uio_iov->iov_len += len;
1902 		}
1903 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1904 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1905 			    &wccflag, stuff);
1906 			if (error)
1907 				goto nfsmout;
1908 		}
1909 		if (!nd->nd_repstat) {
1910 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1911 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1912 					+ NFSX_VERF);
1913 				rlen = fxdr_unsigned(int, *tl++);
1914 				if (rlen == 0) {
1915 					error = NFSERR_IO;
1916 					goto nfsmout;
1917 				} else if (rlen < len) {
1918 					backup = len - rlen;
1919 					uiop->uio_iov->iov_base =
1920 					    (char *)uiop->uio_iov->iov_base -
1921 					    backup;
1922 					uiop->uio_iov->iov_len += backup;
1923 					uiop->uio_offset -= backup;
1924 					uiop->uio_resid += backup;
1925 					len = rlen;
1926 				}
1927 				commit = fxdr_unsigned(int, *tl++);
1928 
1929 				/*
1930 				 * Return the lowest commitment level
1931 				 * obtained by any of the RPCs.
1932 				 */
1933 				if (committed == NFSWRITE_FILESYNC)
1934 					committed = commit;
1935 				else if (committed == NFSWRITE_DATASYNC &&
1936 					commit == NFSWRITE_UNSTABLE)
1937 					committed = commit;
1938 				NFSLOCKMNT(nmp);
1939 				if (!NFSHASWRITEVERF(nmp)) {
1940 					NFSBCOPY((caddr_t)tl,
1941 					    (caddr_t)&nmp->nm_verf[0],
1942 					    NFSX_VERF);
1943 					NFSSETWRITEVERF(nmp);
1944 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1945 				    NFSX_VERF)) {
1946 					*must_commit = 1;
1947 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1948 				}
1949 				NFSUNLOCKMNT(nmp);
1950 			}
1951 			if (nd->nd_flag & ND_NFSV4)
1952 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1953 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1954 				error = nfsm_loadattr(nd, nap);
1955 				if (!error)
1956 					*attrflagp = NFS_LATTR_NOSHRINK;
1957 			}
1958 		} else {
1959 			error = nd->nd_repstat;
1960 		}
1961 		if (error)
1962 			goto nfsmout;
1963 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1964 		m_freem(nd->nd_mrep);
1965 		nd->nd_mrep = NULL;
1966 		tsiz -= len;
1967 	}
1968 nfsmout:
1969 	if (nd->nd_mrep != NULL)
1970 		m_freem(nd->nd_mrep);
1971 	*iomode = committed;
1972 	if (nd->nd_repstat && !error)
1973 		error = nd->nd_repstat;
1974 	return (error);
1975 }
1976 
1977 /*
1978  * nfs mknod rpc
1979  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1980  * mode set to specify the file type and the size field for rdev.
1981  */
1982 int
1983 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1984     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1985     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1986     int *attrflagp, int *dattrflagp, void *dstuff)
1987 {
1988 	u_int32_t *tl;
1989 	int error = 0;
1990 	struct nfsrv_descript nfsd, *nd = &nfsd;
1991 	nfsattrbit_t attrbits;
1992 
1993 	*nfhpp = NULL;
1994 	*attrflagp = 0;
1995 	*dattrflagp = 0;
1996 	if (namelen > NFS_MAXNAMLEN)
1997 		return (ENAMETOOLONG);
1998 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1999 	if (nd->nd_flag & ND_NFSV4) {
2000 		if (vtyp == VBLK || vtyp == VCHR) {
2001 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
2002 			*tl++ = vtonfsv34_type(vtyp);
2003 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2004 			*tl = txdr_unsigned(NFSMINOR(rdev));
2005 		} else {
2006 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2007 			*tl = vtonfsv34_type(vtyp);
2008 		}
2009 	}
2010 	(void) nfsm_strtom(nd, name, namelen);
2011 	if (nd->nd_flag & ND_NFSV3) {
2012 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2013 		*tl = vtonfsv34_type(vtyp);
2014 	}
2015 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2016 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2017 	if ((nd->nd_flag & ND_NFSV3) &&
2018 	    (vtyp == VCHR || vtyp == VBLK)) {
2019 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2020 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2021 		*tl = txdr_unsigned(NFSMINOR(rdev));
2022 	}
2023 	if (nd->nd_flag & ND_NFSV4) {
2024 		NFSGETATTR_ATTRBIT(&attrbits);
2025 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2026 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2027 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2028 		(void) nfsrv_putattrbit(nd, &attrbits);
2029 	}
2030 	if (nd->nd_flag & ND_NFSV2)
2031 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2032 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2033 	if (error)
2034 		return (error);
2035 	if (nd->nd_flag & ND_NFSV4)
2036 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2037 	if (!nd->nd_repstat) {
2038 		if (nd->nd_flag & ND_NFSV4) {
2039 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2040 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2041 			if (error)
2042 				goto nfsmout;
2043 		}
2044 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2045 		if (error)
2046 			goto nfsmout;
2047 	}
2048 	if (nd->nd_flag & ND_NFSV3)
2049 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2050 	if (!error && nd->nd_repstat)
2051 		error = nd->nd_repstat;
2052 nfsmout:
2053 	m_freem(nd->nd_mrep);
2054 	return (error);
2055 }
2056 
2057 /*
2058  * nfs file create call
2059  * Mostly just call the approriate routine. (I separated out v4, so that
2060  * error recovery wouldn't be as difficult.)
2061  */
2062 int
2063 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2064     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2065     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2066     int *attrflagp, int *dattrflagp, void *dstuff)
2067 {
2068 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2069 	struct nfsclowner *owp;
2070 	struct nfscldeleg *dp;
2071 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2072 	u_int32_t clidrev;
2073 
2074 	if (NFSHASNFSV4(nmp)) {
2075 	    retrycnt = 0;
2076 	    do {
2077 		dp = NULL;
2078 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2079 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2080 		    NULL, 1);
2081 		if (error)
2082 			return (error);
2083 		if (nmp->nm_clp != NULL)
2084 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2085 		else
2086 			clidrev = 0;
2087 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2088 		    nfs_numnfscbd == 0 || retrycnt > 0)
2089 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2090 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2091 			  attrflagp, dattrflagp, dstuff, &unlocked);
2092 		else
2093 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2094 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2095 			  attrflagp, dattrflagp, dstuff, &unlocked);
2096 		/*
2097 		 * There is no need to invalidate cached attributes here,
2098 		 * since new post-delegation issue attributes are always
2099 		 * returned by nfsrpc_createv4() and these will update the
2100 		 * attribute cache.
2101 		 */
2102 		if (dp != NULL)
2103 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2104 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2105 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2106 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2107 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2108 		    error == NFSERR_BADSESSION) {
2109 			(void) nfs_catnap(PZERO, error, "nfs_open");
2110 		} else if ((error == NFSERR_EXPIRED ||
2111 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2112 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2113 			retrycnt++;
2114 		}
2115 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2116 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2117 		error == NFSERR_BADSESSION ||
2118 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2119 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2120 	    if (error && retrycnt >= 4)
2121 		    error = EIO;
2122 	} else {
2123 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2124 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2125 		    dstuff);
2126 	}
2127 	return (error);
2128 }
2129 
2130 /*
2131  * The create rpc for v2 and 3.
2132  */
2133 static int
2134 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2135     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2136     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2137     int *attrflagp, int *dattrflagp, void *dstuff)
2138 {
2139 	u_int32_t *tl;
2140 	int error = 0;
2141 	struct nfsrv_descript nfsd, *nd = &nfsd;
2142 
2143 	*nfhpp = NULL;
2144 	*attrflagp = 0;
2145 	*dattrflagp = 0;
2146 	if (namelen > NFS_MAXNAMLEN)
2147 		return (ENAMETOOLONG);
2148 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2149 	(void) nfsm_strtom(nd, name, namelen);
2150 	if (nd->nd_flag & ND_NFSV3) {
2151 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2152 		if (fmode & O_EXCL) {
2153 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2154 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2155 			*tl++ = cverf.lval[0];
2156 			*tl = cverf.lval[1];
2157 		} else {
2158 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2159 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2160 		}
2161 	} else {
2162 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2163 	}
2164 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2165 	if (error)
2166 		return (error);
2167 	if (nd->nd_repstat == 0) {
2168 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2169 		if (error)
2170 			goto nfsmout;
2171 	}
2172 	if (nd->nd_flag & ND_NFSV3)
2173 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2174 	if (nd->nd_repstat != 0 && error == 0)
2175 		error = nd->nd_repstat;
2176 nfsmout:
2177 	m_freem(nd->nd_mrep);
2178 	return (error);
2179 }
2180 
2181 static int
2182 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2183     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2184     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2185     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2186     int *dattrflagp, void *dstuff, int *unlockedp)
2187 {
2188 	u_int32_t *tl;
2189 	int error = 0, deleg, newone, ret, acesize, limitby;
2190 	struct nfsrv_descript nfsd, *nd = &nfsd;
2191 	struct nfsclopen *op;
2192 	struct nfscldeleg *dp = NULL;
2193 	struct nfsnode *np;
2194 	struct nfsfh *nfhp;
2195 	nfsattrbit_t attrbits;
2196 	nfsv4stateid_t stateid;
2197 	u_int32_t rflags;
2198 	struct nfsmount *nmp;
2199 	struct nfsclsession *tsep;
2200 
2201 	nmp = VFSTONFS(dvp->v_mount);
2202 	np = VTONFS(dvp);
2203 	*unlockedp = 0;
2204 	*nfhpp = NULL;
2205 	*dpp = NULL;
2206 	*attrflagp = 0;
2207 	*dattrflagp = 0;
2208 	if (namelen > NFS_MAXNAMLEN)
2209 		return (ENAMETOOLONG);
2210 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2211 	/*
2212 	 * For V4, this is actually an Open op.
2213 	 */
2214 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2215 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2216 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2217 	    NFSV4OPEN_ACCESSREAD);
2218 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2219 	tsep = nfsmnt_mdssession(nmp);
2220 	*tl++ = tsep->nfsess_clientid.lval[0];
2221 	*tl = tsep->nfsess_clientid.lval[1];
2222 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2223 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2224 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2225 	if (fmode & O_EXCL) {
2226 		if (NFSHASNFSV4N(nmp)) {
2227 			if (NFSHASSESSPERSIST(nmp)) {
2228 				/* Use GUARDED for persistent sessions. */
2229 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2230 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2231 			} else {
2232 				/* Otherwise, use EXCLUSIVE4_1. */
2233 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2234 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2235 				*tl++ = cverf.lval[0];
2236 				*tl = cverf.lval[1];
2237 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2238 			}
2239 		} else {
2240 			/* NFSv4.0 */
2241 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2242 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2243 			*tl++ = cverf.lval[0];
2244 			*tl = cverf.lval[1];
2245 		}
2246 	} else {
2247 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2248 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2249 	}
2250 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2251 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2252 	(void) nfsm_strtom(nd, name, namelen);
2253 	/* Get the new file's handle and attributes. */
2254 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2255 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2256 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2257 	NFSGETATTR_ATTRBIT(&attrbits);
2258 	(void) nfsrv_putattrbit(nd, &attrbits);
2259 	/* Get the directory's post-op attributes. */
2260 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2261 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2262 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2263 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2264 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2265 	(void) nfsrv_putattrbit(nd, &attrbits);
2266 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2267 	if (error)
2268 		return (error);
2269 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2270 	if (nd->nd_repstat == 0) {
2271 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2272 		    6 * NFSX_UNSIGNED);
2273 		stateid.seqid = *tl++;
2274 		stateid.other[0] = *tl++;
2275 		stateid.other[1] = *tl++;
2276 		stateid.other[2] = *tl;
2277 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2278 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2279 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2280 		deleg = fxdr_unsigned(int, *tl);
2281 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2282 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2283 			if (!(owp->nfsow_clp->nfsc_flags &
2284 			      NFSCLFLAGS_FIRSTDELEG))
2285 				owp->nfsow_clp->nfsc_flags |=
2286 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2287 			dp = malloc(
2288 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2289 			    M_NFSCLDELEG, M_WAITOK);
2290 			LIST_INIT(&dp->nfsdl_owner);
2291 			LIST_INIT(&dp->nfsdl_lock);
2292 			dp->nfsdl_clp = owp->nfsow_clp;
2293 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2294 			nfscl_lockinit(&dp->nfsdl_rwlock);
2295 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2296 			    NFSX_UNSIGNED);
2297 			dp->nfsdl_stateid.seqid = *tl++;
2298 			dp->nfsdl_stateid.other[0] = *tl++;
2299 			dp->nfsdl_stateid.other[1] = *tl++;
2300 			dp->nfsdl_stateid.other[2] = *tl++;
2301 			ret = fxdr_unsigned(int, *tl);
2302 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2303 				dp->nfsdl_flags = NFSCLDL_WRITE;
2304 				/*
2305 				 * Indicates how much the file can grow.
2306 				 */
2307 				NFSM_DISSECT(tl, u_int32_t *,
2308 				    3 * NFSX_UNSIGNED);
2309 				limitby = fxdr_unsigned(int, *tl++);
2310 				switch (limitby) {
2311 				case NFSV4OPEN_LIMITSIZE:
2312 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2313 					break;
2314 				case NFSV4OPEN_LIMITBLOCKS:
2315 					dp->nfsdl_sizelimit =
2316 					    fxdr_unsigned(u_int64_t, *tl++);
2317 					dp->nfsdl_sizelimit *=
2318 					    fxdr_unsigned(u_int64_t, *tl);
2319 					break;
2320 				default:
2321 					error = NFSERR_BADXDR;
2322 					goto nfsmout;
2323 				}
2324 			} else {
2325 				dp->nfsdl_flags = NFSCLDL_READ;
2326 			}
2327 			if (ret)
2328 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2329 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2330 			    &acesize, p);
2331 			if (error)
2332 				goto nfsmout;
2333 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2334 			error = NFSERR_BADXDR;
2335 			goto nfsmout;
2336 		}
2337 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2338 		if (error)
2339 			goto nfsmout;
2340 		/* Get rid of the PutFH and Getattr status values. */
2341 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2342 		/* Load the directory attributes. */
2343 		error = nfsm_loadattr(nd, dnap);
2344 		if (error)
2345 			goto nfsmout;
2346 		*dattrflagp = 1;
2347 		if (dp != NULL && *attrflagp) {
2348 			dp->nfsdl_change = nnap->na_filerev;
2349 			dp->nfsdl_modtime = nnap->na_mtime;
2350 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2351 		}
2352 		/*
2353 		 * We can now complete the Open state.
2354 		 */
2355 		nfhp = *nfhpp;
2356 		if (dp != NULL) {
2357 			dp->nfsdl_fhlen = nfhp->nfh_len;
2358 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2359 		}
2360 		/*
2361 		 * Get an Open structure that will be
2362 		 * attached to the OpenOwner, acquired already.
2363 		 */
2364 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2365 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2366 		    cred, p, NULL, &op, &newone, NULL, 0);
2367 		if (error)
2368 			goto nfsmout;
2369 		op->nfso_stateid = stateid;
2370 		newnfs_copyincred(cred, &op->nfso_cred);
2371 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2372 		    do {
2373 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2374 			    nfhp->nfh_len, op, cred, p);
2375 			if (ret == NFSERR_DELAY)
2376 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2377 		    } while (ret == NFSERR_DELAY);
2378 		    error = ret;
2379 		}
2380 
2381 		/*
2382 		 * If the server is handing out delegations, but we didn't
2383 		 * get one because an OpenConfirm was required, try the
2384 		 * Open again, to get a delegation. This is a harmless no-op,
2385 		 * from a server's point of view.
2386 		 */
2387 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2388 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2389 		    !error && dp == NULL) {
2390 		    do {
2391 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2392 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2393 			    nfhp->nfh_fh, nfhp->nfh_len,
2394 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2395 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2396 			if (ret == NFSERR_DELAY)
2397 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2398 		    } while (ret == NFSERR_DELAY);
2399 		    if (ret) {
2400 			if (dp != NULL) {
2401 				free(dp, M_NFSCLDELEG);
2402 				dp = NULL;
2403 			}
2404 			if (ret == NFSERR_STALECLIENTID ||
2405 			    ret == NFSERR_STALEDONTRECOVER ||
2406 			    ret == NFSERR_BADSESSION)
2407 				error = ret;
2408 		    }
2409 		}
2410 		nfscl_openrelease(nmp, op, error, newone);
2411 		*unlockedp = 1;
2412 	}
2413 	if (nd->nd_repstat != 0 && error == 0)
2414 		error = nd->nd_repstat;
2415 	if (error == NFSERR_STALECLIENTID)
2416 		nfscl_initiate_recovery(owp->nfsow_clp);
2417 nfsmout:
2418 	if (!error)
2419 		*dpp = dp;
2420 	else if (dp != NULL)
2421 		free(dp, M_NFSCLDELEG);
2422 	m_freem(nd->nd_mrep);
2423 	return (error);
2424 }
2425 
2426 /*
2427  * Nfs remove rpc
2428  */
2429 int
2430 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2431     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2432     void *dstuff)
2433 {
2434 	u_int32_t *tl;
2435 	struct nfsrv_descript nfsd, *nd = &nfsd;
2436 	struct nfsnode *np;
2437 	struct nfsmount *nmp;
2438 	nfsv4stateid_t dstateid;
2439 	int error, ret = 0, i;
2440 
2441 	*dattrflagp = 0;
2442 	if (namelen > NFS_MAXNAMLEN)
2443 		return (ENAMETOOLONG);
2444 	nmp = VFSTONFS(dvp->v_mount);
2445 tryagain:
2446 	if (NFSHASNFSV4(nmp) && ret == 0) {
2447 		ret = nfscl_removedeleg(vp, p, &dstateid);
2448 		if (ret == 1) {
2449 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2450 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2451 			    NFSX_UNSIGNED);
2452 			if (NFSHASNFSV4N(nmp))
2453 				*tl++ = 0;
2454 			else
2455 				*tl++ = dstateid.seqid;
2456 			*tl++ = dstateid.other[0];
2457 			*tl++ = dstateid.other[1];
2458 			*tl++ = dstateid.other[2];
2459 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2460 			np = VTONFS(dvp);
2461 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2462 			    np->n_fhp->nfh_len, 0);
2463 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2464 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2465 		}
2466 	} else {
2467 		ret = 0;
2468 	}
2469 	if (ret == 0)
2470 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2471 	(void) nfsm_strtom(nd, name, namelen);
2472 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2473 	if (error)
2474 		return (error);
2475 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2476 		/* For NFSv4, parse out any Delereturn replies. */
2477 		if (ret > 0 && nd->nd_repstat != 0 &&
2478 		    (nd->nd_flag & ND_NOMOREDATA)) {
2479 			/*
2480 			 * If the Delegreturn failed, try again without
2481 			 * it. The server will Recall, as required.
2482 			 */
2483 			m_freem(nd->nd_mrep);
2484 			goto tryagain;
2485 		}
2486 		for (i = 0; i < (ret * 2); i++) {
2487 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2488 			    ND_NFSV4) {
2489 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2490 			    if (*(tl + 1))
2491 				nd->nd_flag |= ND_NOMOREDATA;
2492 			}
2493 		}
2494 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2495 	}
2496 	if (nd->nd_repstat && !error)
2497 		error = nd->nd_repstat;
2498 nfsmout:
2499 	m_freem(nd->nd_mrep);
2500 	return (error);
2501 }
2502 
2503 /*
2504  * Do an nfs rename rpc.
2505  */
2506 int
2507 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2508     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2509     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2510     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2511 {
2512 	u_int32_t *tl;
2513 	struct nfsrv_descript nfsd, *nd = &nfsd;
2514 	struct nfsmount *nmp;
2515 	struct nfsnode *np;
2516 	nfsattrbit_t attrbits;
2517 	nfsv4stateid_t fdstateid, tdstateid;
2518 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2519 
2520 	*fattrflagp = 0;
2521 	*tattrflagp = 0;
2522 	nmp = VFSTONFS(fdvp->v_mount);
2523 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2524 		return (ENAMETOOLONG);
2525 tryagain:
2526 	if (NFSHASNFSV4(nmp) && ret == 0) {
2527 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2528 		    &tdstateid, &gottd, p);
2529 		if (gotfd && gottd) {
2530 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2531 		} else if (gotfd) {
2532 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2533 		} else if (gottd) {
2534 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2535 		}
2536 		if (gotfd) {
2537 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2538 			if (NFSHASNFSV4N(nmp))
2539 				*tl++ = 0;
2540 			else
2541 				*tl++ = fdstateid.seqid;
2542 			*tl++ = fdstateid.other[0];
2543 			*tl++ = fdstateid.other[1];
2544 			*tl = fdstateid.other[2];
2545 			if (gottd) {
2546 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2547 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2548 				np = VTONFS(tvp);
2549 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2550 				    np->n_fhp->nfh_len, 0);
2551 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2552 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2553 			}
2554 		}
2555 		if (gottd) {
2556 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2557 			if (NFSHASNFSV4N(nmp))
2558 				*tl++ = 0;
2559 			else
2560 				*tl++ = tdstateid.seqid;
2561 			*tl++ = tdstateid.other[0];
2562 			*tl++ = tdstateid.other[1];
2563 			*tl = tdstateid.other[2];
2564 		}
2565 		if (ret > 0) {
2566 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2567 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2568 			np = VTONFS(fdvp);
2569 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2570 			    np->n_fhp->nfh_len, 0);
2571 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2572 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2573 		}
2574 	} else {
2575 		ret = 0;
2576 	}
2577 	if (ret == 0)
2578 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2579 	if (nd->nd_flag & ND_NFSV4) {
2580 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2581 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2582 		NFSWCCATTR_ATTRBIT(&attrbits);
2583 		(void) nfsrv_putattrbit(nd, &attrbits);
2584 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2585 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2586 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2587 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2588 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2589 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2590 		(void) nfsrv_putattrbit(nd, &attrbits);
2591 		nd->nd_flag |= ND_V4WCCATTR;
2592 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2593 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2594 	}
2595 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2596 	if (!(nd->nd_flag & ND_NFSV4))
2597 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2598 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2599 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2600 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2601 	if (error)
2602 		return (error);
2603 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2604 		/* For NFSv4, parse out any Delereturn replies. */
2605 		if (ret > 0 && nd->nd_repstat != 0 &&
2606 		    (nd->nd_flag & ND_NOMOREDATA)) {
2607 			/*
2608 			 * If the Delegreturn failed, try again without
2609 			 * it. The server will Recall, as required.
2610 			 */
2611 			m_freem(nd->nd_mrep);
2612 			goto tryagain;
2613 		}
2614 		for (i = 0; i < (ret * 2); i++) {
2615 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2616 			    ND_NFSV4) {
2617 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2618 			    if (*(tl + 1)) {
2619 				if (i == 0 && ret > 1) {
2620 				    /*
2621 				     * If the Delegreturn failed, try again
2622 				     * without it. The server will Recall, as
2623 				     * required.
2624 				     * If ret > 1, the first iteration of this
2625 				     * loop is the second DelegReturn result.
2626 				     */
2627 				    m_freem(nd->nd_mrep);
2628 				    goto tryagain;
2629 				} else {
2630 				    nd->nd_flag |= ND_NOMOREDATA;
2631 				}
2632 			    }
2633 			}
2634 		}
2635 		/* Now, the first wcc attribute reply. */
2636 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2637 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2638 			if (*(tl + 1))
2639 				nd->nd_flag |= ND_NOMOREDATA;
2640 		}
2641 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2642 		    fstuff);
2643 		/* and the second wcc attribute reply. */
2644 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2645 		    !error) {
2646 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2647 			if (*(tl + 1))
2648 				nd->nd_flag |= ND_NOMOREDATA;
2649 		}
2650 		if (!error)
2651 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2652 			    NULL, tstuff);
2653 	}
2654 	if (nd->nd_repstat && !error)
2655 		error = nd->nd_repstat;
2656 nfsmout:
2657 	m_freem(nd->nd_mrep);
2658 	return (error);
2659 }
2660 
2661 /*
2662  * nfs hard link create rpc
2663  */
2664 int
2665 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2666     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2667     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2668 {
2669 	u_int32_t *tl;
2670 	struct nfsrv_descript nfsd, *nd = &nfsd;
2671 	nfsattrbit_t attrbits;
2672 	int error = 0;
2673 
2674 	*attrflagp = 0;
2675 	*dattrflagp = 0;
2676 	if (namelen > NFS_MAXNAMLEN)
2677 		return (ENAMETOOLONG);
2678 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2679 	if (nd->nd_flag & ND_NFSV4) {
2680 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2681 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2682 	}
2683 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2684 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2685 	if (nd->nd_flag & ND_NFSV4) {
2686 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2687 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2688 		NFSWCCATTR_ATTRBIT(&attrbits);
2689 		(void) nfsrv_putattrbit(nd, &attrbits);
2690 		nd->nd_flag |= ND_V4WCCATTR;
2691 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2692 		*tl = txdr_unsigned(NFSV4OP_LINK);
2693 	}
2694 	(void) nfsm_strtom(nd, name, namelen);
2695 	error = nfscl_request(nd, vp, p, cred, dstuff);
2696 	if (error)
2697 		return (error);
2698 	if (nd->nd_flag & ND_NFSV3) {
2699 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2700 		if (!error)
2701 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2702 			    NULL, dstuff);
2703 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2704 		/*
2705 		 * First, parse out the PutFH and Getattr result.
2706 		 */
2707 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2708 		if (!(*(tl + 1)))
2709 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2710 		if (*(tl + 1))
2711 			nd->nd_flag |= ND_NOMOREDATA;
2712 		/*
2713 		 * Get the pre-op attributes.
2714 		 */
2715 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2716 	}
2717 	if (nd->nd_repstat && !error)
2718 		error = nd->nd_repstat;
2719 nfsmout:
2720 	m_freem(nd->nd_mrep);
2721 	return (error);
2722 }
2723 
2724 /*
2725  * nfs symbolic link create rpc
2726  */
2727 int
2728 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
2729     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2730     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2731     int *dattrflagp, void *dstuff)
2732 {
2733 	u_int32_t *tl;
2734 	struct nfsrv_descript nfsd, *nd = &nfsd;
2735 	struct nfsmount *nmp;
2736 	int slen, error = 0;
2737 
2738 	*nfhpp = NULL;
2739 	*attrflagp = 0;
2740 	*dattrflagp = 0;
2741 	nmp = VFSTONFS(dvp->v_mount);
2742 	slen = strlen(target);
2743 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2744 		return (ENAMETOOLONG);
2745 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2746 	if (nd->nd_flag & ND_NFSV4) {
2747 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2748 		*tl = txdr_unsigned(NFLNK);
2749 		(void) nfsm_strtom(nd, target, slen);
2750 	}
2751 	(void) nfsm_strtom(nd, name, namelen);
2752 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2753 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2754 	if (!(nd->nd_flag & ND_NFSV4))
2755 		(void) nfsm_strtom(nd, target, slen);
2756 	if (nd->nd_flag & ND_NFSV2)
2757 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2758 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2759 	if (error)
2760 		return (error);
2761 	if (nd->nd_flag & ND_NFSV4)
2762 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2763 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2764 		if (!nd->nd_repstat)
2765 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2766 		if (!error)
2767 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2768 			    NULL, dstuff);
2769 	}
2770 	if (nd->nd_repstat && !error)
2771 		error = nd->nd_repstat;
2772 	m_freem(nd->nd_mrep);
2773 	/*
2774 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2775 	 * Only do this if vfs.nfs.ignore_eexist is set.
2776 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2777 	 * should guarantee "exactly once" RPC semantics.
2778 	 */
2779 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2780 	    nmp->nm_minorvers == 0))
2781 		error = 0;
2782 	return (error);
2783 }
2784 
2785 /*
2786  * nfs make dir rpc
2787  */
2788 int
2789 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2790     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2791     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2792     int *dattrflagp, void *dstuff)
2793 {
2794 	u_int32_t *tl;
2795 	struct nfsrv_descript nfsd, *nd = &nfsd;
2796 	nfsattrbit_t attrbits;
2797 	int error = 0;
2798 	struct nfsfh *fhp;
2799 	struct nfsmount *nmp;
2800 
2801 	*nfhpp = NULL;
2802 	*attrflagp = 0;
2803 	*dattrflagp = 0;
2804 	nmp = VFSTONFS(dvp->v_mount);
2805 	fhp = VTONFS(dvp)->n_fhp;
2806 	if (namelen > NFS_MAXNAMLEN)
2807 		return (ENAMETOOLONG);
2808 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2809 	if (nd->nd_flag & ND_NFSV4) {
2810 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2811 		*tl = txdr_unsigned(NFDIR);
2812 	}
2813 	(void) nfsm_strtom(nd, name, namelen);
2814 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2815 	if (nd->nd_flag & ND_NFSV4) {
2816 		NFSGETATTR_ATTRBIT(&attrbits);
2817 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2818 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2819 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2820 		(void) nfsrv_putattrbit(nd, &attrbits);
2821 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2822 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2823 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2824 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2825 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2826 		(void) nfsrv_putattrbit(nd, &attrbits);
2827 	}
2828 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2829 	if (error)
2830 		return (error);
2831 	if (nd->nd_flag & ND_NFSV4)
2832 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2833 	if (!nd->nd_repstat && !error) {
2834 		if (nd->nd_flag & ND_NFSV4) {
2835 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2836 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2837 		}
2838 		if (!error)
2839 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2840 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2841 			/* Get rid of the PutFH and Getattr status values. */
2842 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2843 			/* Load the directory attributes. */
2844 			error = nfsm_loadattr(nd, dnap);
2845 			if (error == 0)
2846 				*dattrflagp = 1;
2847 		}
2848 	}
2849 	if ((nd->nd_flag & ND_NFSV3) && !error)
2850 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2851 	if (nd->nd_repstat && !error)
2852 		error = nd->nd_repstat;
2853 nfsmout:
2854 	m_freem(nd->nd_mrep);
2855 	/*
2856 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2857 	 * Only do this if vfs.nfs.ignore_eexist is set.
2858 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2859 	 * should guarantee "exactly once" RPC semantics.
2860 	 */
2861 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2862 	    nmp->nm_minorvers == 0))
2863 		error = 0;
2864 	return (error);
2865 }
2866 
2867 /*
2868  * nfs remove directory call
2869  */
2870 int
2871 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2872     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2873 {
2874 	struct nfsrv_descript nfsd, *nd = &nfsd;
2875 	int error = 0;
2876 
2877 	*dattrflagp = 0;
2878 	if (namelen > NFS_MAXNAMLEN)
2879 		return (ENAMETOOLONG);
2880 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2881 	(void) nfsm_strtom(nd, name, namelen);
2882 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2883 	if (error)
2884 		return (error);
2885 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2886 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2887 	if (nd->nd_repstat && !error)
2888 		error = nd->nd_repstat;
2889 	m_freem(nd->nd_mrep);
2890 	/*
2891 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2892 	 */
2893 	if (error == ENOENT)
2894 		error = 0;
2895 	return (error);
2896 }
2897 
2898 /*
2899  * Readdir rpc.
2900  * Always returns with either uio_resid unchanged, if you are at the
2901  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2902  * filled in.
2903  * I felt this would allow caching of directory blocks more easily
2904  * than returning a pertially filled block.
2905  * Directory offset cookies:
2906  * Oh my, what to do with them...
2907  * I can think of three ways to deal with them:
2908  * 1 - have the layer above these RPCs maintain a map between logical
2909  *     directory byte offsets and the NFS directory offset cookies
2910  * 2 - pass the opaque directory offset cookies up into userland
2911  *     and let the libc functions deal with them, via the system call
2912  * 3 - return them to userland in the "struct dirent", so future versions
2913  *     of libc can use them and do whatever is necessary to make things work
2914  *     above these rpc calls, in the meantime
2915  * For now, I do #3 by "hiding" the directory offset cookies after the
2916  * d_name field in struct dirent. This is space inside d_reclen that
2917  * will be ignored by anything that doesn't know about them.
2918  * The directory offset cookies are filled in as the last 8 bytes of
2919  * each directory entry, after d_name. Someday, the userland libc
2920  * functions may be able to use these. In the meantime, it satisfies
2921  * OpenBSD's requirements for cookies being returned.
2922  * If expects the directory offset cookie for the read to be in uio_offset
2923  * and returns the one for the next entry after this directory block in
2924  * there, as well.
2925  */
2926 int
2927 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2928     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2929     int *eofp, void *stuff)
2930 {
2931 	int len, left;
2932 	struct dirent *dp = NULL;
2933 	u_int32_t *tl;
2934 	nfsquad_t cookie, ncookie;
2935 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2936 	struct nfsnode *dnp = VTONFS(vp);
2937 	struct nfsvattr nfsva;
2938 	struct nfsrv_descript nfsd, *nd = &nfsd;
2939 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2940 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2941 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2942 	char *cp;
2943 	nfsattrbit_t attrbits, dattrbits;
2944 	u_int32_t rderr, *tl2 = NULL;
2945 	size_t tresid;
2946 
2947 	KASSERT(uiop->uio_iovcnt == 1 &&
2948 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
2949 	    ("nfs readdirrpc bad uio"));
2950 	ncookie.lval[0] = ncookie.lval[1] = 0;
2951 	/*
2952 	 * There is no point in reading a lot more than uio_resid, however
2953 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2954 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2955 	 * will never make readsize > nm_readdirsize.
2956 	 */
2957 	readsize = nmp->nm_readdirsize;
2958 	if (readsize > uiop->uio_resid)
2959 		readsize = uiop->uio_resid + DIRBLKSIZ;
2960 
2961 	*attrflagp = 0;
2962 	if (eofp)
2963 		*eofp = 0;
2964 	tresid = uiop->uio_resid;
2965 	cookie.lval[0] = cookiep->nfsuquad[0];
2966 	cookie.lval[1] = cookiep->nfsuquad[1];
2967 	nd->nd_mrep = NULL;
2968 
2969 	/*
2970 	 * For NFSv4, first create the "." and ".." entries.
2971 	 */
2972 	if (NFSHASNFSV4(nmp)) {
2973 		reqsize = 6 * NFSX_UNSIGNED;
2974 		NFSGETATTR_ATTRBIT(&dattrbits);
2975 		NFSZERO_ATTRBIT(&attrbits);
2976 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2977 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2978 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2979 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2980 			NFSSETBIT_ATTRBIT(&attrbits,
2981 			    NFSATTRBIT_MOUNTEDONFILEID);
2982 			gotmnton = 1;
2983 		} else {
2984 			/*
2985 			 * Must fake it. Use the fileno, except when the
2986 			 * fsid is != to that of the directory. For that
2987 			 * case, generate a fake fileno that is not the same.
2988 			 */
2989 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2990 			gotmnton = 0;
2991 		}
2992 
2993 		/*
2994 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2995 		 */
2996 		if (uiop->uio_offset == 0) {
2997 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2998 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2999 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3000 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3001 			(void) nfsrv_putattrbit(nd, &attrbits);
3002 			error = nfscl_request(nd, vp, p, cred, stuff);
3003 			if (error)
3004 			    return (error);
3005 			dotfileid = 0;	/* Fake out the compiler. */
3006 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3007 			    error = nfsm_loadattr(nd, &nfsva);
3008 			    if (error != 0)
3009 				goto nfsmout;
3010 			    dotfileid = nfsva.na_fileid;
3011 			}
3012 			if (nd->nd_repstat == 0) {
3013 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3014 			    len = fxdr_unsigned(int, *(tl + 4));
3015 			    if (len > 0 && len <= NFSX_V4FHMAX)
3016 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3017 			    else
3018 				error = EPERM;
3019 			    if (!error) {
3020 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3021 				nfsva.na_mntonfileno = UINT64_MAX;
3022 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3023 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3024 				    NULL, NULL, NULL, p, cred);
3025 				if (error) {
3026 				    dotdotfileid = dotfileid;
3027 				} else if (gotmnton) {
3028 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3029 					dotdotfileid = nfsva.na_mntonfileno;
3030 				    else
3031 					dotdotfileid = nfsva.na_fileid;
3032 				} else if (nfsva.na_filesid[0] ==
3033 				    dnp->n_vattr.na_filesid[0] &&
3034 				    nfsva.na_filesid[1] ==
3035 				    dnp->n_vattr.na_filesid[1]) {
3036 				    dotdotfileid = nfsva.na_fileid;
3037 				} else {
3038 				    do {
3039 					fakefileno--;
3040 				    } while (fakefileno ==
3041 					nfsva.na_fileid);
3042 				    dotdotfileid = fakefileno;
3043 				}
3044 			    }
3045 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3046 			    /*
3047 			     * Lookupp returns NFSERR_NOENT when we are
3048 			     * at the root, so just use the current dir.
3049 			     */
3050 			    nd->nd_repstat = 0;
3051 			    dotdotfileid = dotfileid;
3052 			} else {
3053 			    error = nd->nd_repstat;
3054 			}
3055 			m_freem(nd->nd_mrep);
3056 			if (error)
3057 			    return (error);
3058 			nd->nd_mrep = NULL;
3059 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3060 			dp->d_pad0 = dp->d_pad1 = 0;
3061 			dp->d_off = 0;
3062 			dp->d_type = DT_DIR;
3063 			dp->d_fileno = dotfileid;
3064 			dp->d_namlen = 1;
3065 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3066 			dp->d_name[0] = '.';
3067 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3068 			/*
3069 			 * Just make these offset cookie 0.
3070 			 */
3071 			tl = (u_int32_t *)&dp->d_name[8];
3072 			*tl++ = 0;
3073 			*tl = 0;
3074 			blksiz += dp->d_reclen;
3075 			uiop->uio_resid -= dp->d_reclen;
3076 			uiop->uio_offset += dp->d_reclen;
3077 			uiop->uio_iov->iov_base =
3078 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3079 			uiop->uio_iov->iov_len -= dp->d_reclen;
3080 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3081 			dp->d_pad0 = dp->d_pad1 = 0;
3082 			dp->d_off = 0;
3083 			dp->d_type = DT_DIR;
3084 			dp->d_fileno = dotdotfileid;
3085 			dp->d_namlen = 2;
3086 			*((uint64_t *)dp->d_name) = 0;
3087 			dp->d_name[0] = '.';
3088 			dp->d_name[1] = '.';
3089 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3090 			/*
3091 			 * Just make these offset cookie 0.
3092 			 */
3093 			tl = (u_int32_t *)&dp->d_name[8];
3094 			*tl++ = 0;
3095 			*tl = 0;
3096 			blksiz += dp->d_reclen;
3097 			uiop->uio_resid -= dp->d_reclen;
3098 			uiop->uio_offset += dp->d_reclen;
3099 			uiop->uio_iov->iov_base =
3100 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3101 			uiop->uio_iov->iov_len -= dp->d_reclen;
3102 		}
3103 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3104 	} else {
3105 		reqsize = 5 * NFSX_UNSIGNED;
3106 	}
3107 
3108 	/*
3109 	 * Loop around doing readdir rpc's of size readsize.
3110 	 * The stopping criteria is EOF or buffer full.
3111 	 */
3112 	while (more_dirs && bigenough) {
3113 		*attrflagp = 0;
3114 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3115 		if (nd->nd_flag & ND_NFSV2) {
3116 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3117 			*tl++ = cookie.lval[1];
3118 			*tl = txdr_unsigned(readsize);
3119 		} else {
3120 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3121 			*tl++ = cookie.lval[0];
3122 			*tl++ = cookie.lval[1];
3123 			if (cookie.qval == 0) {
3124 				*tl++ = 0;
3125 				*tl++ = 0;
3126 			} else {
3127 				NFSLOCKNODE(dnp);
3128 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3129 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3130 				NFSUNLOCKNODE(dnp);
3131 			}
3132 			if (nd->nd_flag & ND_NFSV4) {
3133 				*tl++ = txdr_unsigned(readsize);
3134 				*tl = txdr_unsigned(readsize);
3135 				(void) nfsrv_putattrbit(nd, &attrbits);
3136 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3137 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3138 				(void) nfsrv_putattrbit(nd, &dattrbits);
3139 			} else {
3140 				*tl = txdr_unsigned(readsize);
3141 			}
3142 		}
3143 		error = nfscl_request(nd, vp, p, cred, stuff);
3144 		if (error)
3145 			return (error);
3146 		if (!(nd->nd_flag & ND_NFSV2)) {
3147 			if (nd->nd_flag & ND_NFSV3)
3148 				error = nfscl_postop_attr(nd, nap, attrflagp,
3149 				    stuff);
3150 			if (!nd->nd_repstat && !error) {
3151 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3152 				NFSLOCKNODE(dnp);
3153 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3154 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3155 				NFSUNLOCKNODE(dnp);
3156 			}
3157 		}
3158 		if (nd->nd_repstat || error) {
3159 			if (!error)
3160 				error = nd->nd_repstat;
3161 			goto nfsmout;
3162 		}
3163 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3164 		more_dirs = fxdr_unsigned(int, *tl);
3165 		if (!more_dirs)
3166 			tryformoredirs = 0;
3167 
3168 		/* loop through the dir entries, doctoring them to 4bsd form */
3169 		while (more_dirs && bigenough) {
3170 			if (nd->nd_flag & ND_NFSV4) {
3171 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3172 				ncookie.lval[0] = *tl++;
3173 				ncookie.lval[1] = *tl++;
3174 				len = fxdr_unsigned(int, *tl);
3175 			} else if (nd->nd_flag & ND_NFSV3) {
3176 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3177 				nfsva.na_fileid = fxdr_hyper(tl);
3178 				tl += 2;
3179 				len = fxdr_unsigned(int, *tl);
3180 			} else {
3181 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3182 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3183 				    *tl++);
3184 				len = fxdr_unsigned(int, *tl);
3185 			}
3186 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3187 				error = EBADRPC;
3188 				goto nfsmout;
3189 			}
3190 			tlen = roundup2(len, 8);
3191 			if (tlen == len)
3192 				tlen += 8;  /* To ensure null termination. */
3193 			left = DIRBLKSIZ - blksiz;
3194 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3195 				NFSBZERO(uiop->uio_iov->iov_base, left);
3196 				dp->d_reclen += left;
3197 				uiop->uio_iov->iov_base =
3198 				    (char *)uiop->uio_iov->iov_base + left;
3199 				uiop->uio_iov->iov_len -= left;
3200 				uiop->uio_resid -= left;
3201 				uiop->uio_offset += left;
3202 				blksiz = 0;
3203 			}
3204 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3205 			    uiop->uio_resid)
3206 				bigenough = 0;
3207 			if (bigenough) {
3208 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3209 				dp->d_pad0 = dp->d_pad1 = 0;
3210 				dp->d_off = 0;
3211 				dp->d_namlen = len;
3212 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3213 				    NFSX_HYPER;
3214 				dp->d_type = DT_UNKNOWN;
3215 				blksiz += dp->d_reclen;
3216 				if (blksiz == DIRBLKSIZ)
3217 					blksiz = 0;
3218 				uiop->uio_resid -= DIRHDSIZ;
3219 				uiop->uio_offset += DIRHDSIZ;
3220 				uiop->uio_iov->iov_base =
3221 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3222 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3223 				error = nfsm_mbufuio(nd, uiop, len);
3224 				if (error)
3225 					goto nfsmout;
3226 				cp = uiop->uio_iov->iov_base;
3227 				tlen -= len;
3228 				NFSBZERO(cp, tlen);
3229 				cp += tlen;	/* points to cookie storage */
3230 				tl2 = (u_int32_t *)cp;
3231 				uiop->uio_iov->iov_base =
3232 				    (char *)uiop->uio_iov->iov_base + tlen +
3233 				    NFSX_HYPER;
3234 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3235 				uiop->uio_resid -= tlen + NFSX_HYPER;
3236 				uiop->uio_offset += (tlen + NFSX_HYPER);
3237 			} else {
3238 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3239 				if (error)
3240 					goto nfsmout;
3241 			}
3242 			if (nd->nd_flag & ND_NFSV4) {
3243 				rderr = 0;
3244 				nfsva.na_mntonfileno = UINT64_MAX;
3245 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3246 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3247 				    NULL, NULL, &rderr, p, cred);
3248 				if (error)
3249 					goto nfsmout;
3250 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3251 			} else if (nd->nd_flag & ND_NFSV3) {
3252 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3253 				ncookie.lval[0] = *tl++;
3254 				ncookie.lval[1] = *tl++;
3255 			} else {
3256 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3257 				ncookie.lval[0] = 0;
3258 				ncookie.lval[1] = *tl++;
3259 			}
3260 			if (bigenough) {
3261 			    if (nd->nd_flag & ND_NFSV4) {
3262 				if (rderr) {
3263 				    dp->d_fileno = 0;
3264 				} else {
3265 				    if (gotmnton) {
3266 					if (nfsva.na_mntonfileno != UINT64_MAX)
3267 					    dp->d_fileno = nfsva.na_mntonfileno;
3268 					else
3269 					    dp->d_fileno = nfsva.na_fileid;
3270 				    } else if (nfsva.na_filesid[0] ==
3271 					dnp->n_vattr.na_filesid[0] &&
3272 					nfsva.na_filesid[1] ==
3273 					dnp->n_vattr.na_filesid[1]) {
3274 					dp->d_fileno = nfsva.na_fileid;
3275 				    } else {
3276 					do {
3277 					    fakefileno--;
3278 					} while (fakefileno ==
3279 					    nfsva.na_fileid);
3280 					dp->d_fileno = fakefileno;
3281 				    }
3282 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3283 				}
3284 			    } else {
3285 				dp->d_fileno = nfsva.na_fileid;
3286 			    }
3287 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3288 				ncookie.lval[0];
3289 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3290 				ncookie.lval[1];
3291 			}
3292 			more_dirs = fxdr_unsigned(int, *tl);
3293 		}
3294 		/*
3295 		 * If at end of rpc data, get the eof boolean
3296 		 */
3297 		if (!more_dirs) {
3298 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3299 			eof = fxdr_unsigned(int, *tl);
3300 			if (tryformoredirs)
3301 				more_dirs = !eof;
3302 			if (nd->nd_flag & ND_NFSV4) {
3303 				error = nfscl_postop_attr(nd, nap, attrflagp,
3304 				    stuff);
3305 				if (error)
3306 					goto nfsmout;
3307 			}
3308 		}
3309 		m_freem(nd->nd_mrep);
3310 		nd->nd_mrep = NULL;
3311 	}
3312 	/*
3313 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3314 	 * by increasing d_reclen for the last record.
3315 	 */
3316 	if (blksiz > 0) {
3317 		left = DIRBLKSIZ - blksiz;
3318 		NFSBZERO(uiop->uio_iov->iov_base, left);
3319 		dp->d_reclen += left;
3320 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3321 		    left;
3322 		uiop->uio_iov->iov_len -= left;
3323 		uiop->uio_resid -= left;
3324 		uiop->uio_offset += left;
3325 	}
3326 
3327 	/*
3328 	 * If returning no data, assume end of file.
3329 	 * If not bigenough, return not end of file, since you aren't
3330 	 *    returning all the data
3331 	 * Otherwise, return the eof flag from the server.
3332 	 */
3333 	if (eofp) {
3334 		if (tresid == ((size_t)(uiop->uio_resid)))
3335 			*eofp = 1;
3336 		else if (!bigenough)
3337 			*eofp = 0;
3338 		else
3339 			*eofp = eof;
3340 	}
3341 
3342 	/*
3343 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3344 	 */
3345 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3346 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3347 		NFSBZERO(dp, DIRBLKSIZ);
3348 		dp->d_type = DT_UNKNOWN;
3349 		tl = (u_int32_t *)&dp->d_name[4];
3350 		*tl++ = cookie.lval[0];
3351 		*tl = cookie.lval[1];
3352 		dp->d_reclen = DIRBLKSIZ;
3353 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3354 		    DIRBLKSIZ;
3355 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3356 		uiop->uio_resid -= DIRBLKSIZ;
3357 		uiop->uio_offset += DIRBLKSIZ;
3358 	}
3359 
3360 nfsmout:
3361 	if (nd->nd_mrep != NULL)
3362 		m_freem(nd->nd_mrep);
3363 	return (error);
3364 }
3365 
3366 #ifndef APPLE
3367 /*
3368  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3369  * (Also used for NFS V4 when mount flag set.)
3370  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3371  */
3372 int
3373 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3374     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3375     int *eofp, void *stuff)
3376 {
3377 	int len, left;
3378 	struct dirent *dp = NULL;
3379 	u_int32_t *tl;
3380 	vnode_t newvp = NULLVP;
3381 	struct nfsrv_descript nfsd, *nd = &nfsd;
3382 	struct nameidata nami, *ndp = &nami;
3383 	struct componentname *cnp = &ndp->ni_cnd;
3384 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3385 	struct nfsnode *dnp = VTONFS(vp), *np;
3386 	struct nfsvattr nfsva;
3387 	struct nfsfh *nfhp;
3388 	nfsquad_t cookie, ncookie;
3389 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3390 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3391 	int isdotdot = 0, unlocknewvp = 0;
3392 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3393 	u_int64_t fileno = 0;
3394 	char *cp;
3395 	nfsattrbit_t attrbits, dattrbits;
3396 	size_t tresid;
3397 	u_int32_t *tl2 = NULL, rderr;
3398 	struct timespec dctime;
3399 
3400 	KASSERT(uiop->uio_iovcnt == 1 &&
3401 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3402 	    ("nfs readdirplusrpc bad uio"));
3403 	ncookie.lval[0] = ncookie.lval[1] = 0;
3404 	timespecclear(&dctime);
3405 	*attrflagp = 0;
3406 	if (eofp != NULL)
3407 		*eofp = 0;
3408 	ndp->ni_dvp = vp;
3409 	nd->nd_mrep = NULL;
3410 	cookie.lval[0] = cookiep->nfsuquad[0];
3411 	cookie.lval[1] = cookiep->nfsuquad[1];
3412 	tresid = uiop->uio_resid;
3413 
3414 	/*
3415 	 * For NFSv4, first create the "." and ".." entries.
3416 	 */
3417 	if (NFSHASNFSV4(nmp)) {
3418 		NFSGETATTR_ATTRBIT(&dattrbits);
3419 		NFSZERO_ATTRBIT(&attrbits);
3420 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3421 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3422 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3423 			NFSSETBIT_ATTRBIT(&attrbits,
3424 			    NFSATTRBIT_MOUNTEDONFILEID);
3425 			gotmnton = 1;
3426 		} else {
3427 			/*
3428 			 * Must fake it. Use the fileno, except when the
3429 			 * fsid is != to that of the directory. For that
3430 			 * case, generate a fake fileno that is not the same.
3431 			 */
3432 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3433 			gotmnton = 0;
3434 		}
3435 
3436 		/*
3437 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3438 		 */
3439 		if (uiop->uio_offset == 0) {
3440 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3441 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3442 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3443 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3444 			(void) nfsrv_putattrbit(nd, &attrbits);
3445 			error = nfscl_request(nd, vp, p, cred, stuff);
3446 			if (error)
3447 			    return (error);
3448 			dotfileid = 0;	/* Fake out the compiler. */
3449 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3450 			    error = nfsm_loadattr(nd, &nfsva);
3451 			    if (error != 0)
3452 				goto nfsmout;
3453 			    dctime = nfsva.na_ctime;
3454 			    dotfileid = nfsva.na_fileid;
3455 			}
3456 			if (nd->nd_repstat == 0) {
3457 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3458 			    len = fxdr_unsigned(int, *(tl + 4));
3459 			    if (len > 0 && len <= NFSX_V4FHMAX)
3460 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3461 			    else
3462 				error = EPERM;
3463 			    if (!error) {
3464 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3465 				nfsva.na_mntonfileno = UINT64_MAX;
3466 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3467 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3468 				    NULL, NULL, NULL, p, cred);
3469 				if (error) {
3470 				    dotdotfileid = dotfileid;
3471 				} else if (gotmnton) {
3472 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3473 					dotdotfileid = nfsva.na_mntonfileno;
3474 				    else
3475 					dotdotfileid = nfsva.na_fileid;
3476 				} else if (nfsva.na_filesid[0] ==
3477 				    dnp->n_vattr.na_filesid[0] &&
3478 				    nfsva.na_filesid[1] ==
3479 				    dnp->n_vattr.na_filesid[1]) {
3480 				    dotdotfileid = nfsva.na_fileid;
3481 				} else {
3482 				    do {
3483 					fakefileno--;
3484 				    } while (fakefileno ==
3485 					nfsva.na_fileid);
3486 				    dotdotfileid = fakefileno;
3487 				}
3488 			    }
3489 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3490 			    /*
3491 			     * Lookupp returns NFSERR_NOENT when we are
3492 			     * at the root, so just use the current dir.
3493 			     */
3494 			    nd->nd_repstat = 0;
3495 			    dotdotfileid = dotfileid;
3496 			} else {
3497 			    error = nd->nd_repstat;
3498 			}
3499 			m_freem(nd->nd_mrep);
3500 			if (error)
3501 			    return (error);
3502 			nd->nd_mrep = NULL;
3503 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3504 			dp->d_pad0 = dp->d_pad1 = 0;
3505 			dp->d_off = 0;
3506 			dp->d_type = DT_DIR;
3507 			dp->d_fileno = dotfileid;
3508 			dp->d_namlen = 1;
3509 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3510 			dp->d_name[0] = '.';
3511 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3512 			/*
3513 			 * Just make these offset cookie 0.
3514 			 */
3515 			tl = (u_int32_t *)&dp->d_name[8];
3516 			*tl++ = 0;
3517 			*tl = 0;
3518 			blksiz += dp->d_reclen;
3519 			uiop->uio_resid -= dp->d_reclen;
3520 			uiop->uio_offset += dp->d_reclen;
3521 			uiop->uio_iov->iov_base =
3522 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3523 			uiop->uio_iov->iov_len -= dp->d_reclen;
3524 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3525 			dp->d_pad0 = dp->d_pad1 = 0;
3526 			dp->d_off = 0;
3527 			dp->d_type = DT_DIR;
3528 			dp->d_fileno = dotdotfileid;
3529 			dp->d_namlen = 2;
3530 			*((uint64_t *)dp->d_name) = 0;
3531 			dp->d_name[0] = '.';
3532 			dp->d_name[1] = '.';
3533 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3534 			/*
3535 			 * Just make these offset cookie 0.
3536 			 */
3537 			tl = (u_int32_t *)&dp->d_name[8];
3538 			*tl++ = 0;
3539 			*tl = 0;
3540 			blksiz += dp->d_reclen;
3541 			uiop->uio_resid -= dp->d_reclen;
3542 			uiop->uio_offset += dp->d_reclen;
3543 			uiop->uio_iov->iov_base =
3544 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3545 			uiop->uio_iov->iov_len -= dp->d_reclen;
3546 		}
3547 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3548 		if (gotmnton)
3549 			NFSSETBIT_ATTRBIT(&attrbits,
3550 			    NFSATTRBIT_MOUNTEDONFILEID);
3551 	}
3552 
3553 	/*
3554 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3555 	 * The stopping criteria is EOF or buffer full.
3556 	 */
3557 	while (more_dirs && bigenough) {
3558 		*attrflagp = 0;
3559 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3560  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3561 		*tl++ = cookie.lval[0];
3562 		*tl++ = cookie.lval[1];
3563 		if (cookie.qval == 0) {
3564 			*tl++ = 0;
3565 			*tl++ = 0;
3566 		} else {
3567 			NFSLOCKNODE(dnp);
3568 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3569 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3570 			NFSUNLOCKNODE(dnp);
3571 		}
3572 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3573 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3574 		if (nd->nd_flag & ND_NFSV4) {
3575 			(void) nfsrv_putattrbit(nd, &attrbits);
3576 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3577 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3578 			(void) nfsrv_putattrbit(nd, &dattrbits);
3579 		}
3580 		error = nfscl_request(nd, vp, p, cred, stuff);
3581 		if (error)
3582 			return (error);
3583 		if (nd->nd_flag & ND_NFSV3)
3584 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3585 		if (nd->nd_repstat || error) {
3586 			if (!error)
3587 				error = nd->nd_repstat;
3588 			goto nfsmout;
3589 		}
3590 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3591 			dctime = nap->na_ctime;
3592 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3593 		NFSLOCKNODE(dnp);
3594 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3595 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3596 		NFSUNLOCKNODE(dnp);
3597 		more_dirs = fxdr_unsigned(int, *tl);
3598 		if (!more_dirs)
3599 			tryformoredirs = 0;
3600 
3601 		/* loop through the dir entries, doctoring them to 4bsd form */
3602 		while (more_dirs && bigenough) {
3603 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3604 			if (nd->nd_flag & ND_NFSV4) {
3605 				ncookie.lval[0] = *tl++;
3606 				ncookie.lval[1] = *tl++;
3607 			} else {
3608 				fileno = fxdr_hyper(tl);
3609 				tl += 2;
3610 			}
3611 			len = fxdr_unsigned(int, *tl);
3612 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3613 				error = EBADRPC;
3614 				goto nfsmout;
3615 			}
3616 			tlen = roundup2(len, 8);
3617 			if (tlen == len)
3618 				tlen += 8;  /* To ensure null termination. */
3619 			left = DIRBLKSIZ - blksiz;
3620 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3621 				NFSBZERO(uiop->uio_iov->iov_base, left);
3622 				dp->d_reclen += left;
3623 				uiop->uio_iov->iov_base =
3624 				    (char *)uiop->uio_iov->iov_base + left;
3625 				uiop->uio_iov->iov_len -= left;
3626 				uiop->uio_resid -= left;
3627 				uiop->uio_offset += left;
3628 				blksiz = 0;
3629 			}
3630 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3631 			    uiop->uio_resid)
3632 				bigenough = 0;
3633 			if (bigenough) {
3634 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3635 				dp->d_pad0 = dp->d_pad1 = 0;
3636 				dp->d_off = 0;
3637 				dp->d_namlen = len;
3638 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3639 				    NFSX_HYPER;
3640 				dp->d_type = DT_UNKNOWN;
3641 				blksiz += dp->d_reclen;
3642 				if (blksiz == DIRBLKSIZ)
3643 					blksiz = 0;
3644 				uiop->uio_resid -= DIRHDSIZ;
3645 				uiop->uio_offset += DIRHDSIZ;
3646 				uiop->uio_iov->iov_base =
3647 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3648 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3649 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3650 				cnp->cn_namelen = len;
3651 				NFSCNHASHZERO(cnp);
3652 				error = nfsm_mbufuio(nd, uiop, len);
3653 				if (error)
3654 					goto nfsmout;
3655 				cp = uiop->uio_iov->iov_base;
3656 				tlen -= len;
3657 				NFSBZERO(cp, tlen);
3658 				cp += tlen;	/* points to cookie storage */
3659 				tl2 = (u_int32_t *)cp;
3660 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3661 				    cnp->cn_nameptr[1] == '.')
3662 					isdotdot = 1;
3663 				else
3664 					isdotdot = 0;
3665 				uiop->uio_iov->iov_base =
3666 				    (char *)uiop->uio_iov->iov_base + tlen +
3667 				    NFSX_HYPER;
3668 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3669 				uiop->uio_resid -= tlen + NFSX_HYPER;
3670 				uiop->uio_offset += (tlen + NFSX_HYPER);
3671 			} else {
3672 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3673 				if (error)
3674 					goto nfsmout;
3675 			}
3676 			nfhp = NULL;
3677 			if (nd->nd_flag & ND_NFSV3) {
3678 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3679 				ncookie.lval[0] = *tl++;
3680 				ncookie.lval[1] = *tl++;
3681 				attrflag = fxdr_unsigned(int, *tl);
3682 				if (attrflag) {
3683 				  error = nfsm_loadattr(nd, &nfsva);
3684 				  if (error)
3685 					goto nfsmout;
3686 				}
3687 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3688 				if (*tl) {
3689 					error = nfsm_getfh(nd, &nfhp);
3690 					if (error)
3691 					    goto nfsmout;
3692 				}
3693 				if (!attrflag && nfhp != NULL) {
3694 					free(nfhp, M_NFSFH);
3695 					nfhp = NULL;
3696 				}
3697 			} else {
3698 				rderr = 0;
3699 				nfsva.na_mntonfileno = 0xffffffff;
3700 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3701 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3702 				    NULL, NULL, &rderr, p, cred);
3703 				if (error)
3704 					goto nfsmout;
3705 			}
3706 
3707 			if (bigenough) {
3708 			    if (nd->nd_flag & ND_NFSV4) {
3709 				if (rderr) {
3710 				    dp->d_fileno = 0;
3711 				} else if (gotmnton) {
3712 				    if (nfsva.na_mntonfileno != 0xffffffff)
3713 					dp->d_fileno = nfsva.na_mntonfileno;
3714 				    else
3715 					dp->d_fileno = nfsva.na_fileid;
3716 				} else if (nfsva.na_filesid[0] ==
3717 				    dnp->n_vattr.na_filesid[0] &&
3718 				    nfsva.na_filesid[1] ==
3719 				    dnp->n_vattr.na_filesid[1]) {
3720 				    dp->d_fileno = nfsva.na_fileid;
3721 				} else {
3722 				    do {
3723 					fakefileno--;
3724 				    } while (fakefileno ==
3725 					nfsva.na_fileid);
3726 				    dp->d_fileno = fakefileno;
3727 				}
3728 			    } else {
3729 				dp->d_fileno = fileno;
3730 			    }
3731 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3732 				ncookie.lval[0];
3733 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3734 				ncookie.lval[1];
3735 
3736 			    if (nfhp != NULL) {
3737 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3738 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3739 				    VREF(vp);
3740 				    newvp = vp;
3741 				    unlocknewvp = 0;
3742 				    free(nfhp, M_NFSFH);
3743 				    np = dnp;
3744 				} else if (isdotdot != 0) {
3745 				    /*
3746 				     * Skip doing a nfscl_nget() call for "..".
3747 				     * There's a race between acquiring the nfs
3748 				     * node here and lookups that look for the
3749 				     * directory being read (in the parent).
3750 				     * It would try to get a lock on ".." here,
3751 				     * owning the lock on the directory being
3752 				     * read. Lookup will hold the lock on ".."
3753 				     * and try to acquire the lock on the
3754 				     * directory being read.
3755 				     * If the directory is unlocked/relocked,
3756 				     * then there is a LOR with the buflock
3757 				     * vp is relocked.
3758 				     */
3759 				    free(nfhp, M_NFSFH);
3760 				} else {
3761 				    error = nfscl_nget(vp->v_mount, vp,
3762 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3763 				    if (!error) {
3764 					newvp = NFSTOV(np);
3765 					unlocknewvp = 1;
3766 				    }
3767 				}
3768 				nfhp = NULL;
3769 				if (newvp != NULLVP) {
3770 				    error = nfscl_loadattrcache(&newvp,
3771 					&nfsva, NULL, NULL, 0, 0);
3772 				    if (error) {
3773 					if (unlocknewvp)
3774 					    vput(newvp);
3775 					else
3776 					    vrele(newvp);
3777 					goto nfsmout;
3778 				    }
3779 				    dp->d_type =
3780 					vtonfs_dtype(np->n_vattr.na_type);
3781 				    ndp->ni_vp = newvp;
3782 				    NFSCNHASH(cnp, HASHINIT);
3783 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3784 					ndp->ni_dvp != ndp->ni_vp &&
3785 					(newvp->v_type != VDIR ||
3786 					 dctime.tv_sec != 0)) {
3787 					cache_enter_time(ndp->ni_dvp,
3788 					    ndp->ni_vp, cnp,
3789 					    &nfsva.na_ctime,
3790 					    newvp->v_type != VDIR ? NULL :
3791 					    &dctime);
3792 				    }
3793 				    if (unlocknewvp)
3794 					vput(newvp);
3795 				    else
3796 					vrele(newvp);
3797 				    newvp = NULLVP;
3798 				}
3799 			    }
3800 			} else if (nfhp != NULL) {
3801 			    free(nfhp, M_NFSFH);
3802 			}
3803 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3804 			more_dirs = fxdr_unsigned(int, *tl);
3805 		}
3806 		/*
3807 		 * If at end of rpc data, get the eof boolean
3808 		 */
3809 		if (!more_dirs) {
3810 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3811 			eof = fxdr_unsigned(int, *tl);
3812 			if (tryformoredirs)
3813 				more_dirs = !eof;
3814 			if (nd->nd_flag & ND_NFSV4) {
3815 				error = nfscl_postop_attr(nd, nap, attrflagp,
3816 				    stuff);
3817 				if (error)
3818 					goto nfsmout;
3819 			}
3820 		}
3821 		m_freem(nd->nd_mrep);
3822 		nd->nd_mrep = NULL;
3823 	}
3824 	/*
3825 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3826 	 * by increasing d_reclen for the last record.
3827 	 */
3828 	if (blksiz > 0) {
3829 		left = DIRBLKSIZ - blksiz;
3830 		NFSBZERO(uiop->uio_iov->iov_base, left);
3831 		dp->d_reclen += left;
3832 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3833 		    left;
3834 		uiop->uio_iov->iov_len -= left;
3835 		uiop->uio_resid -= left;
3836 		uiop->uio_offset += left;
3837 	}
3838 
3839 	/*
3840 	 * If returning no data, assume end of file.
3841 	 * If not bigenough, return not end of file, since you aren't
3842 	 *    returning all the data
3843 	 * Otherwise, return the eof flag from the server.
3844 	 */
3845 	if (eofp != NULL) {
3846 		if (tresid == uiop->uio_resid)
3847 			*eofp = 1;
3848 		else if (!bigenough)
3849 			*eofp = 0;
3850 		else
3851 			*eofp = eof;
3852 	}
3853 
3854 	/*
3855 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3856 	 */
3857 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3858 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3859 		NFSBZERO(dp, DIRBLKSIZ);
3860 		dp->d_type = DT_UNKNOWN;
3861 		tl = (u_int32_t *)&dp->d_name[4];
3862 		*tl++ = cookie.lval[0];
3863 		*tl = cookie.lval[1];
3864 		dp->d_reclen = DIRBLKSIZ;
3865 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3866 		    DIRBLKSIZ;
3867 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3868 		uiop->uio_resid -= DIRBLKSIZ;
3869 		uiop->uio_offset += DIRBLKSIZ;
3870 	}
3871 
3872 nfsmout:
3873 	if (nd->nd_mrep != NULL)
3874 		m_freem(nd->nd_mrep);
3875 	return (error);
3876 }
3877 #endif	/* !APPLE */
3878 
3879 /*
3880  * Nfs commit rpc
3881  */
3882 int
3883 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3884     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3885 {
3886 	u_int32_t *tl;
3887 	struct nfsrv_descript nfsd, *nd = &nfsd;
3888 	nfsattrbit_t attrbits;
3889 	int error;
3890 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3891 
3892 	*attrflagp = 0;
3893 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3894 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3895 	txdr_hyper(offset, tl);
3896 	tl += 2;
3897 	*tl = txdr_unsigned(cnt);
3898 	if (nd->nd_flag & ND_NFSV4) {
3899 		/*
3900 		 * And do a Getattr op.
3901 		 */
3902 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3903 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3904 		NFSGETATTR_ATTRBIT(&attrbits);
3905 		(void) nfsrv_putattrbit(nd, &attrbits);
3906 	}
3907 	error = nfscl_request(nd, vp, p, cred, stuff);
3908 	if (error)
3909 		return (error);
3910 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3911 	if (!error && !nd->nd_repstat) {
3912 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3913 		NFSLOCKMNT(nmp);
3914 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3915 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3916 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3917 		}
3918 		NFSUNLOCKMNT(nmp);
3919 		if (nd->nd_flag & ND_NFSV4)
3920 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3921 	}
3922 nfsmout:
3923 	if (!error && nd->nd_repstat)
3924 		error = nd->nd_repstat;
3925 	m_freem(nd->nd_mrep);
3926 	return (error);
3927 }
3928 
3929 /*
3930  * NFS byte range lock rpc.
3931  * (Mostly just calls one of the three lower level RPC routines.)
3932  */
3933 int
3934 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3935     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3936 {
3937 	struct nfscllockowner *lp;
3938 	struct nfsclclient *clp;
3939 	struct nfsfh *nfhp;
3940 	struct nfsrv_descript nfsd, *nd = &nfsd;
3941 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3942 	u_int64_t off, len;
3943 	off_t start, end;
3944 	u_int32_t clidrev = 0;
3945 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3946 	int callcnt, dorpc;
3947 
3948 	/*
3949 	 * Convert the flock structure into a start and end and do POSIX
3950 	 * bounds checking.
3951 	 */
3952 	switch (fl->l_whence) {
3953 	case SEEK_SET:
3954 	case SEEK_CUR:
3955 		/*
3956 		 * Caller is responsible for adding any necessary offset
3957 		 * when SEEK_CUR is used.
3958 		 */
3959 		start = fl->l_start;
3960 		off = fl->l_start;
3961 		break;
3962 	case SEEK_END:
3963 		start = size + fl->l_start;
3964 		off = size + fl->l_start;
3965 		break;
3966 	default:
3967 		return (EINVAL);
3968 	}
3969 	if (start < 0)
3970 		return (EINVAL);
3971 	if (fl->l_len != 0) {
3972 		end = start + fl->l_len - 1;
3973 		if (end < start)
3974 			return (EINVAL);
3975 	}
3976 
3977 	len = fl->l_len;
3978 	if (len == 0)
3979 		len = NFS64BITSSET;
3980 	retrycnt = 0;
3981 	do {
3982 	    nd->nd_repstat = 0;
3983 	    if (op == F_GETLK) {
3984 		error = nfscl_getcl(vp->v_mount, cred, p, 1, &clp);
3985 		if (error)
3986 			return (error);
3987 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3988 		if (!error) {
3989 			clidrev = clp->nfsc_clientidrev;
3990 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3991 			    p, id, flags);
3992 		} else if (error == -1) {
3993 			error = 0;
3994 		}
3995 		nfscl_clientrelease(clp);
3996 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3997 		/*
3998 		 * We must loop around for all lockowner cases.
3999 		 */
4000 		callcnt = 0;
4001 		error = nfscl_getcl(vp->v_mount, cred, p, 1, &clp);
4002 		if (error)
4003 			return (error);
4004 		do {
4005 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
4006 			clp, id, flags, &lp, &dorpc);
4007 		    /*
4008 		     * If it returns a NULL lp, we're done.
4009 		     */
4010 		    if (lp == NULL) {
4011 			if (callcnt == 0)
4012 			    nfscl_clientrelease(clp);
4013 			else
4014 			    nfscl_releasealllocks(clp, vp, p, id, flags);
4015 			return (error);
4016 		    }
4017 		    if (nmp->nm_clp != NULL)
4018 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4019 		    else
4020 			clidrev = 0;
4021 		    /*
4022 		     * If the server doesn't support Posix lock semantics,
4023 		     * only allow locks on the entire file, since it won't
4024 		     * handle overlapping byte ranges.
4025 		     * There might still be a problem when a lock
4026 		     * upgrade/downgrade (read<->write) occurs, since the
4027 		     * server "might" expect an unlock first?
4028 		     */
4029 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4030 			(off == 0 && len == NFS64BITSSET))) {
4031 			/*
4032 			 * Since the lock records will go away, we must
4033 			 * wait for grace and delay here.
4034 			 */
4035 			do {
4036 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4037 				NFSV4LOCKT_READ, cred, p, 0);
4038 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4039 				 nd->nd_repstat == NFSERR_DELAY) &&
4040 				error == 0)
4041 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4042 				    "nfs_advlock");
4043 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4044 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4045 		    }
4046 		    callcnt++;
4047 		} while (error == 0 && nd->nd_repstat == 0);
4048 		nfscl_releasealllocks(clp, vp, p, id, flags);
4049 	    } else if (op == F_SETLK) {
4050 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4051 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4052 		if (error || donelocally) {
4053 			return (error);
4054 		}
4055 		if (nmp->nm_clp != NULL)
4056 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4057 		else
4058 			clidrev = 0;
4059 		nfhp = VTONFS(vp)->n_fhp;
4060 		if (!lp->nfsl_open->nfso_posixlock &&
4061 		    (off != 0 || len != NFS64BITSSET)) {
4062 			error = EINVAL;
4063 		} else {
4064 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4065 			    nfhp->nfh_len, lp, newone, reclaim, off,
4066 			    len, fl->l_type, cred, p, 0);
4067 		}
4068 		if (!error)
4069 			error = nd->nd_repstat;
4070 		nfscl_lockrelease(lp, error, newone);
4071 	    } else {
4072 		error = EINVAL;
4073 	    }
4074 	    if (!error)
4075 	        error = nd->nd_repstat;
4076 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4077 		error == NFSERR_STALEDONTRECOVER ||
4078 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4079 		error == NFSERR_BADSESSION) {
4080 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4081 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4082 		&& clidrev != 0) {
4083 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4084 		retrycnt++;
4085 	    }
4086 	} while (error == NFSERR_GRACE ||
4087 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4088 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4089 	    error == NFSERR_BADSESSION ||
4090 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4091 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4092 	if (error && retrycnt >= 4)
4093 		error = EIO;
4094 	return (error);
4095 }
4096 
4097 /*
4098  * The lower level routine for the LockT case.
4099  */
4100 int
4101 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4102     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4103     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4104 {
4105 	u_int32_t *tl;
4106 	int error, type, size;
4107 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4108 	struct nfsnode *np;
4109 	struct nfsmount *nmp;
4110 	struct nfsclsession *tsep;
4111 
4112 	nmp = VFSTONFS(vp->v_mount);
4113 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4114 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4115 	if (fl->l_type == F_RDLCK)
4116 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4117 	else
4118 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4119 	txdr_hyper(off, tl);
4120 	tl += 2;
4121 	txdr_hyper(len, tl);
4122 	tl += 2;
4123 	tsep = nfsmnt_mdssession(nmp);
4124 	*tl++ = tsep->nfsess_clientid.lval[0];
4125 	*tl = tsep->nfsess_clientid.lval[1];
4126 	nfscl_filllockowner(id, own, flags);
4127 	np = VTONFS(vp);
4128 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4129 	    np->n_fhp->nfh_len);
4130 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4131 	error = nfscl_request(nd, vp, p, cred, NULL);
4132 	if (error)
4133 		return (error);
4134 	if (nd->nd_repstat == 0) {
4135 		fl->l_type = F_UNLCK;
4136 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4137 		nd->nd_repstat = 0;
4138 		fl->l_whence = SEEK_SET;
4139 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4140 		fl->l_start = fxdr_hyper(tl);
4141 		tl += 2;
4142 		len = fxdr_hyper(tl);
4143 		tl += 2;
4144 		if (len == NFS64BITSSET)
4145 			fl->l_len = 0;
4146 		else
4147 			fl->l_len = len;
4148 		type = fxdr_unsigned(int, *tl++);
4149 		if (type == NFSV4LOCKT_WRITE)
4150 			fl->l_type = F_WRLCK;
4151 		else
4152 			fl->l_type = F_RDLCK;
4153 		/*
4154 		 * XXX For now, I have no idea what to do with the
4155 		 * conflicting lock_owner, so I'll just set the pid == 0
4156 		 * and skip over the lock_owner.
4157 		 */
4158 		fl->l_pid = (pid_t)0;
4159 		tl += 2;
4160 		size = fxdr_unsigned(int, *tl);
4161 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4162 			error = EBADRPC;
4163 		if (!error)
4164 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4165 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4166 		nfscl_initiate_recovery(clp);
4167 nfsmout:
4168 	m_freem(nd->nd_mrep);
4169 	return (error);
4170 }
4171 
4172 /*
4173  * Lower level function that performs the LockU RPC.
4174  */
4175 static int
4176 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4177     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4178     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4179 {
4180 	u_int32_t *tl;
4181 	int error;
4182 
4183 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4184 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4185 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4186 	*tl++ = txdr_unsigned(type);
4187 	*tl = txdr_unsigned(lp->nfsl_seqid);
4188 	if (nfstest_outofseq &&
4189 	    (arc4random() % nfstest_outofseq) == 0)
4190 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4191 	tl++;
4192 	if (NFSHASNFSV4N(nmp))
4193 		*tl++ = 0;
4194 	else
4195 		*tl++ = lp->nfsl_stateid.seqid;
4196 	*tl++ = lp->nfsl_stateid.other[0];
4197 	*tl++ = lp->nfsl_stateid.other[1];
4198 	*tl++ = lp->nfsl_stateid.other[2];
4199 	txdr_hyper(off, tl);
4200 	tl += 2;
4201 	txdr_hyper(len, tl);
4202 	if (syscred)
4203 		nd->nd_flag |= ND_USEGSSNAME;
4204 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4205 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4206 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4207 	if (error)
4208 		return (error);
4209 	if (nd->nd_repstat == 0) {
4210 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4211 		lp->nfsl_stateid.seqid = *tl++;
4212 		lp->nfsl_stateid.other[0] = *tl++;
4213 		lp->nfsl_stateid.other[1] = *tl++;
4214 		lp->nfsl_stateid.other[2] = *tl;
4215 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4216 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4217 nfsmout:
4218 	m_freem(nd->nd_mrep);
4219 	return (error);
4220 }
4221 
4222 /*
4223  * The actual Lock RPC.
4224  */
4225 int
4226 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4227     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4228     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4229     NFSPROC_T *p, int syscred)
4230 {
4231 	u_int32_t *tl;
4232 	int error, size;
4233 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4234 	struct nfsclsession *tsep;
4235 
4236 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4237 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4238 	if (type == F_RDLCK)
4239 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4240 	else
4241 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4242 	*tl++ = txdr_unsigned(reclaim);
4243 	txdr_hyper(off, tl);
4244 	tl += 2;
4245 	txdr_hyper(len, tl);
4246 	tl += 2;
4247 	if (newone) {
4248 	    *tl = newnfs_true;
4249 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4250 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4251 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4252 	    if (NFSHASNFSV4N(nmp))
4253 		*tl++ = 0;
4254 	    else
4255 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4256 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4257 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4258 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4259 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4260 	    tsep = nfsmnt_mdssession(nmp);
4261 	    *tl++ = tsep->nfsess_clientid.lval[0];
4262 	    *tl = tsep->nfsess_clientid.lval[1];
4263 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4264 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4265 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4266 	} else {
4267 	    *tl = newnfs_false;
4268 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4269 	    if (NFSHASNFSV4N(nmp))
4270 		*tl++ = 0;
4271 	    else
4272 		*tl++ = lp->nfsl_stateid.seqid;
4273 	    *tl++ = lp->nfsl_stateid.other[0];
4274 	    *tl++ = lp->nfsl_stateid.other[1];
4275 	    *tl++ = lp->nfsl_stateid.other[2];
4276 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4277 	    if (nfstest_outofseq &&
4278 		(arc4random() % nfstest_outofseq) == 0)
4279 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4280 	}
4281 	if (syscred)
4282 		nd->nd_flag |= ND_USEGSSNAME;
4283 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4284 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4285 	if (error)
4286 		return (error);
4287 	if (newone)
4288 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4289 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4290 	if (nd->nd_repstat == 0) {
4291 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4292 		lp->nfsl_stateid.seqid = *tl++;
4293 		lp->nfsl_stateid.other[0] = *tl++;
4294 		lp->nfsl_stateid.other[1] = *tl++;
4295 		lp->nfsl_stateid.other[2] = *tl;
4296 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4297 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4298 		size = fxdr_unsigned(int, *(tl + 7));
4299 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4300 			error = EBADRPC;
4301 		if (!error)
4302 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4303 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4304 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4305 nfsmout:
4306 	m_freem(nd->nd_mrep);
4307 	return (error);
4308 }
4309 
4310 /*
4311  * nfs statfs rpc
4312  * (always called with the vp for the mount point)
4313  */
4314 int
4315 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4316     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4317     void *stuff)
4318 {
4319 	u_int32_t *tl = NULL;
4320 	struct nfsrv_descript nfsd, *nd = &nfsd;
4321 	struct nfsmount *nmp;
4322 	nfsattrbit_t attrbits;
4323 	int error;
4324 
4325 	*attrflagp = 0;
4326 	nmp = VFSTONFS(vp->v_mount);
4327 	if (NFSHASNFSV4(nmp)) {
4328 		/*
4329 		 * For V4, you actually do a getattr.
4330 		 */
4331 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4332 		NFSSTATFS_GETATTRBIT(&attrbits);
4333 		(void) nfsrv_putattrbit(nd, &attrbits);
4334 		nd->nd_flag |= ND_USEGSSNAME;
4335 		error = nfscl_request(nd, vp, p, cred, stuff);
4336 		if (error)
4337 			return (error);
4338 		if (nd->nd_repstat == 0) {
4339 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4340 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4341 			    cred);
4342 			if (!error) {
4343 				nmp->nm_fsid[0] = nap->na_filesid[0];
4344 				nmp->nm_fsid[1] = nap->na_filesid[1];
4345 				NFSSETHASSETFSID(nmp);
4346 				*attrflagp = 1;
4347 			}
4348 		} else {
4349 			error = nd->nd_repstat;
4350 		}
4351 		if (error)
4352 			goto nfsmout;
4353 	} else {
4354 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4355 		error = nfscl_request(nd, vp, p, cred, stuff);
4356 		if (error)
4357 			return (error);
4358 		if (nd->nd_flag & ND_NFSV3) {
4359 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4360 			if (error)
4361 				goto nfsmout;
4362 		}
4363 		if (nd->nd_repstat) {
4364 			error = nd->nd_repstat;
4365 			goto nfsmout;
4366 		}
4367 		NFSM_DISSECT(tl, u_int32_t *,
4368 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4369 	}
4370 	if (NFSHASNFSV3(nmp)) {
4371 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4372 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4373 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4374 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4375 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4376 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4377 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4378 	} else if (NFSHASNFSV4(nmp) == 0) {
4379 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4380 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4381 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4382 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4383 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4384 	}
4385 nfsmout:
4386 	m_freem(nd->nd_mrep);
4387 	return (error);
4388 }
4389 
4390 /*
4391  * nfs pathconf rpc
4392  */
4393 int
4394 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4395     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4396     void *stuff)
4397 {
4398 	struct nfsrv_descript nfsd, *nd = &nfsd;
4399 	struct nfsmount *nmp;
4400 	u_int32_t *tl;
4401 	nfsattrbit_t attrbits;
4402 	int error;
4403 
4404 	*attrflagp = 0;
4405 	nmp = VFSTONFS(vp->v_mount);
4406 	if (NFSHASNFSV4(nmp)) {
4407 		/*
4408 		 * For V4, you actually do a getattr.
4409 		 */
4410 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4411 		NFSPATHCONF_GETATTRBIT(&attrbits);
4412 		(void) nfsrv_putattrbit(nd, &attrbits);
4413 		nd->nd_flag |= ND_USEGSSNAME;
4414 		error = nfscl_request(nd, vp, p, cred, stuff);
4415 		if (error)
4416 			return (error);
4417 		if (nd->nd_repstat == 0) {
4418 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4419 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4420 			    cred);
4421 			if (!error)
4422 				*attrflagp = 1;
4423 		} else {
4424 			error = nd->nd_repstat;
4425 		}
4426 	} else {
4427 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4428 		error = nfscl_request(nd, vp, p, cred, stuff);
4429 		if (error)
4430 			return (error);
4431 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4432 		if (nd->nd_repstat && !error)
4433 			error = nd->nd_repstat;
4434 		if (!error) {
4435 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4436 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4437 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4438 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4439 			pc->pc_chownrestricted =
4440 			    fxdr_unsigned(u_int32_t, *tl++);
4441 			pc->pc_caseinsensitive =
4442 			    fxdr_unsigned(u_int32_t, *tl++);
4443 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4444 		}
4445 	}
4446 nfsmout:
4447 	m_freem(nd->nd_mrep);
4448 	return (error);
4449 }
4450 
4451 /*
4452  * nfs version 3 fsinfo rpc call
4453  */
4454 int
4455 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4456     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4457 {
4458 	u_int32_t *tl;
4459 	struct nfsrv_descript nfsd, *nd = &nfsd;
4460 	int error;
4461 
4462 	*attrflagp = 0;
4463 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4464 	error = nfscl_request(nd, vp, p, cred, stuff);
4465 	if (error)
4466 		return (error);
4467 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4468 	if (nd->nd_repstat && !error)
4469 		error = nd->nd_repstat;
4470 	if (!error) {
4471 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4472 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4473 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4474 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4475 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4476 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4477 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4478 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4479 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4480 		tl += 2;
4481 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4482 		tl += 2;
4483 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4484 	}
4485 nfsmout:
4486 	m_freem(nd->nd_mrep);
4487 	return (error);
4488 }
4489 
4490 /*
4491  * This function performs the Renew RPC.
4492  */
4493 int
4494 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4495     NFSPROC_T *p)
4496 {
4497 	u_int32_t *tl;
4498 	struct nfsrv_descript nfsd;
4499 	struct nfsrv_descript *nd = &nfsd;
4500 	struct nfsmount *nmp;
4501 	int error;
4502 	struct nfssockreq *nrp;
4503 	struct nfsclsession *tsep;
4504 
4505 	nmp = clp->nfsc_nmp;
4506 	if (nmp == NULL)
4507 		return (0);
4508 	if (dsp == NULL)
4509 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4510 		    0);
4511 	else
4512 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4513 		    &dsp->nfsclds_sess, 0, 0);
4514 	if (!NFSHASNFSV4N(nmp)) {
4515 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4516 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4517 		tsep = nfsmnt_mdssession(nmp);
4518 		*tl++ = tsep->nfsess_clientid.lval[0];
4519 		*tl = tsep->nfsess_clientid.lval[1];
4520 	}
4521 	nrp = NULL;
4522 	if (dsp != NULL)
4523 		nrp = dsp->nfsclds_sockp;
4524 	if (nrp == NULL)
4525 		/* If NULL, use the MDS socket. */
4526 		nrp = &nmp->nm_sockreq;
4527 	nd->nd_flag |= ND_USEGSSNAME;
4528 	if (dsp == NULL)
4529 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4530 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4531 	else {
4532 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4533 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4534 		if (error == ENXIO)
4535 			nfscl_cancelreqs(dsp);
4536 	}
4537 	if (error)
4538 		return (error);
4539 	error = nd->nd_repstat;
4540 	m_freem(nd->nd_mrep);
4541 	return (error);
4542 }
4543 
4544 /*
4545  * This function performs the Releaselockowner RPC.
4546  */
4547 int
4548 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4549     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4550 {
4551 	struct nfsrv_descript nfsd, *nd = &nfsd;
4552 	u_int32_t *tl;
4553 	int error;
4554 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4555 	struct nfsclsession *tsep;
4556 
4557 	if (NFSHASNFSV4N(nmp)) {
4558 		/* For NFSv4.1, do a FreeStateID. */
4559 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4560 		    NULL, 0, 0);
4561 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4562 	} else {
4563 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4564 		    NULL, 0, 0);
4565 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4566 		tsep = nfsmnt_mdssession(nmp);
4567 		*tl++ = tsep->nfsess_clientid.lval[0];
4568 		*tl = tsep->nfsess_clientid.lval[1];
4569 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4570 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4571 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4572 	}
4573 	nd->nd_flag |= ND_USEGSSNAME;
4574 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4575 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4576 	if (error)
4577 		return (error);
4578 	error = nd->nd_repstat;
4579 	m_freem(nd->nd_mrep);
4580 	return (error);
4581 }
4582 
4583 /*
4584  * This function performs the Compound to get the mount pt FH.
4585  */
4586 int
4587 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4588     NFSPROC_T *p)
4589 {
4590 	u_int32_t *tl;
4591 	struct nfsrv_descript nfsd;
4592 	struct nfsrv_descript *nd = &nfsd;
4593 	u_char *cp, *cp2;
4594 	int error, cnt, len, setnil;
4595 	u_int32_t *opcntp;
4596 
4597 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4598 	    0);
4599 	cp = dirpath;
4600 	cnt = 0;
4601 	do {
4602 		setnil = 0;
4603 		while (*cp == '/')
4604 			cp++;
4605 		cp2 = cp;
4606 		while (*cp2 != '\0' && *cp2 != '/')
4607 			cp2++;
4608 		if (*cp2 == '/') {
4609 			setnil = 1;
4610 			*cp2 = '\0';
4611 		}
4612 		if (cp2 != cp) {
4613 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4614 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4615 			nfsm_strtom(nd, cp, strlen(cp));
4616 			cnt++;
4617 		}
4618 		if (setnil)
4619 			*cp2++ = '/';
4620 		cp = cp2;
4621 	} while (*cp != '\0');
4622 	if (NFSHASNFSV4N(nmp))
4623 		/* Has a Sequence Op done by nfscl_reqstart(). */
4624 		*opcntp = txdr_unsigned(3 + cnt);
4625 	else
4626 		*opcntp = txdr_unsigned(2 + cnt);
4627 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4628 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4629 	nd->nd_flag |= ND_USEGSSNAME;
4630 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4631 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4632 	if (error)
4633 		return (error);
4634 	if (nd->nd_repstat == 0) {
4635 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4636 		tl += (2 + 2 * cnt);
4637 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4638 			len > NFSX_FHMAX) {
4639 			nd->nd_repstat = NFSERR_BADXDR;
4640 		} else {
4641 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4642 			if (nd->nd_repstat == 0)
4643 				nmp->nm_fhsize = len;
4644 		}
4645 	}
4646 	error = nd->nd_repstat;
4647 nfsmout:
4648 	m_freem(nd->nd_mrep);
4649 	return (error);
4650 }
4651 
4652 /*
4653  * This function performs the Delegreturn RPC.
4654  */
4655 int
4656 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4657     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4658 {
4659 	u_int32_t *tl;
4660 	struct nfsrv_descript nfsd;
4661 	struct nfsrv_descript *nd = &nfsd;
4662 	int error;
4663 
4664 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4665 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4666 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4667 	if (NFSHASNFSV4N(nmp))
4668 		*tl++ = 0;
4669 	else
4670 		*tl++ = dp->nfsdl_stateid.seqid;
4671 	*tl++ = dp->nfsdl_stateid.other[0];
4672 	*tl++ = dp->nfsdl_stateid.other[1];
4673 	*tl = dp->nfsdl_stateid.other[2];
4674 	if (syscred)
4675 		nd->nd_flag |= ND_USEGSSNAME;
4676 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4677 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4678 	if (error)
4679 		return (error);
4680 	error = nd->nd_repstat;
4681 	m_freem(nd->nd_mrep);
4682 	return (error);
4683 }
4684 
4685 /*
4686  * nfs getacl call.
4687  */
4688 int
4689 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4690     struct acl *aclp, void *stuff)
4691 {
4692 	struct nfsrv_descript nfsd, *nd = &nfsd;
4693 	int error;
4694 	nfsattrbit_t attrbits;
4695 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4696 
4697 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4698 		return (EOPNOTSUPP);
4699 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4700 	NFSZERO_ATTRBIT(&attrbits);
4701 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4702 	(void) nfsrv_putattrbit(nd, &attrbits);
4703 	error = nfscl_request(nd, vp, p, cred, stuff);
4704 	if (error)
4705 		return (error);
4706 	if (!nd->nd_repstat)
4707 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4708 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4709 	else
4710 		error = nd->nd_repstat;
4711 	m_freem(nd->nd_mrep);
4712 	return (error);
4713 }
4714 
4715 /*
4716  * nfs setacl call.
4717  */
4718 int
4719 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4720     struct acl *aclp, void *stuff)
4721 {
4722 	int error;
4723 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4724 
4725 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4726 		return (EOPNOTSUPP);
4727 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4728 	return (error);
4729 }
4730 
4731 /*
4732  * nfs setacl call.
4733  */
4734 static int
4735 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4736     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4737 {
4738 	struct nfsrv_descript nfsd, *nd = &nfsd;
4739 	int error;
4740 	nfsattrbit_t attrbits;
4741 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4742 
4743 	if (!NFSHASNFSV4(nmp))
4744 		return (EOPNOTSUPP);
4745 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4746 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4747 	NFSZERO_ATTRBIT(&attrbits);
4748 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4749 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
4750 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4751 	error = nfscl_request(nd, vp, p, cred, stuff);
4752 	if (error)
4753 		return (error);
4754 	/* Don't care about the pre/postop attributes */
4755 	m_freem(nd->nd_mrep);
4756 	return (nd->nd_repstat);
4757 }
4758 
4759 /*
4760  * Do the NFSv4.1 Exchange ID.
4761  */
4762 int
4763 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4764     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
4765     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
4766 {
4767 	uint32_t *tl, v41flags;
4768 	struct nfsrv_descript nfsd;
4769 	struct nfsrv_descript *nd = &nfsd;
4770 	struct nfsclds *dsp;
4771 	struct timespec verstime;
4772 	int error, len;
4773 
4774 	*dspp = NULL;
4775 	if (minorvers == 0)
4776 		minorvers = nmp->nm_minorvers;
4777 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
4778 	    NFS_VER4, minorvers);
4779 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4780 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4781 	*tl = txdr_unsigned(clp->nfsc_rev);
4782 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4783 
4784 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4785 	*tl++ = txdr_unsigned(exchflags);
4786 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4787 
4788 	/* Set the implementation id4 */
4789 	*tl = txdr_unsigned(1);
4790 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4791 	(void) nfsm_strtom(nd, version, strlen(version));
4792 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4793 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4794 	verstime.tv_nsec = 0;
4795 	txdr_nfsv4time(&verstime, tl);
4796 	nd->nd_flag |= ND_USEGSSNAME;
4797 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4798 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4799 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4800 	    (int)nd->nd_repstat);
4801 	if (error != 0)
4802 		return (error);
4803 	if (nd->nd_repstat == 0) {
4804 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4805 		len = fxdr_unsigned(int, *(tl + 7));
4806 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4807 			error = NFSERR_BADXDR;
4808 			goto nfsmout;
4809 		}
4810 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4811 		    M_WAITOK | M_ZERO);
4812 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4813 		dsp->nfsclds_servownlen = len;
4814 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4815 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4816 		dsp->nfsclds_sess.nfsess_sequenceid =
4817 		    fxdr_unsigned(uint32_t, *tl++);
4818 		v41flags = fxdr_unsigned(uint32_t, *tl);
4819 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4820 		    NFSHASPNFSOPT(nmp)) {
4821 			NFSCL_DEBUG(1, "set PNFS\n");
4822 			NFSLOCKMNT(nmp);
4823 			nmp->nm_state |= NFSSTA_PNFS;
4824 			NFSUNLOCKMNT(nmp);
4825 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4826 		}
4827 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4828 			dsp->nfsclds_flags |= NFSCLDS_DS;
4829 		if (minorvers == NFSV42_MINORVERSION)
4830 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
4831 		if (len > 0)
4832 			nd->nd_repstat = nfsrv_mtostr(nd,
4833 			    dsp->nfsclds_serverown, len);
4834 		if (nd->nd_repstat == 0) {
4835 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4836 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4837 			    NULL, MTX_DEF);
4838 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4839 			*dspp = dsp;
4840 		} else
4841 			free(dsp, M_NFSCLDS);
4842 	}
4843 	error = nd->nd_repstat;
4844 nfsmout:
4845 	m_freem(nd->nd_mrep);
4846 	return (error);
4847 }
4848 
4849 /*
4850  * Do the NFSv4.1 Create Session.
4851  */
4852 int
4853 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4854     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
4855     struct ucred *cred, NFSPROC_T *p)
4856 {
4857 	uint32_t crflags, maxval, *tl;
4858 	struct nfsrv_descript nfsd;
4859 	struct nfsrv_descript *nd = &nfsd;
4860 	int error, irdcnt, minorvers;
4861 
4862 	/* Make sure nm_rsize, nm_wsize is set. */
4863 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4864 		nmp->nm_rsize = NFS_MAXBSIZE;
4865 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4866 		nmp->nm_wsize = NFS_MAXBSIZE;
4867 	if (dsp == NULL)
4868 		minorvers = nmp->nm_minorvers;
4869 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
4870 		minorvers = NFSV42_MINORVERSION;
4871 	else
4872 		minorvers = NFSV41_MINORVERSION;
4873 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
4874 	    NFS_VER4, minorvers);
4875 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4876 	*tl++ = sep->nfsess_clientid.lval[0];
4877 	*tl++ = sep->nfsess_clientid.lval[1];
4878 	*tl++ = txdr_unsigned(sequenceid);
4879 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4880 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4881 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4882 	*tl = txdr_unsigned(crflags);
4883 
4884 	/* Fill in fore channel attributes. */
4885 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4886 	*tl++ = 0;				/* Header pad size */
4887 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
4888 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
4889 		/*
4890 		 * NFSv4.2 Extended Attribute operations may want to do
4891 		 * requests/replies that are larger than nm_rsize/nm_wsize.
4892 		 */
4893 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4894 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4895 	} else {
4896 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
4897 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
4898 	}
4899 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4900 	*tl++ = txdr_unsigned(20);		/* Max operations */
4901 	*tl++ = txdr_unsigned(64);		/* Max slots */
4902 	*tl = 0;				/* No rdma ird */
4903 
4904 	/* Fill in back channel attributes. */
4905 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4906 	*tl++ = 0;				/* Header pad size */
4907 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4908 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4909 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4910 	*tl++ = txdr_unsigned(4);		/* Max operations */
4911 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4912 	*tl = 0;				/* No rdma ird */
4913 
4914 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4915 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4916 
4917 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4918 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4919 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4920 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4921 	*tl++ = 0;				/* Null machine name */
4922 	*tl++ = 0;				/* Uid == 0 */
4923 	*tl++ = 0;				/* Gid == 0 */
4924 	*tl = 0;				/* No additional gids */
4925 	nd->nd_flag |= ND_USEGSSNAME;
4926 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4927 	    NFS_VER4, NULL, 1, NULL, NULL);
4928 	if (error != 0)
4929 		return (error);
4930 	if (nd->nd_repstat == 0) {
4931 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4932 		    2 * NFSX_UNSIGNED);
4933 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4934 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4935 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4936 		crflags = fxdr_unsigned(uint32_t, *tl);
4937 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4938 			NFSLOCKMNT(nmp);
4939 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4940 			NFSUNLOCKMNT(nmp);
4941 		}
4942 
4943 		/* Get the fore channel slot count. */
4944 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4945 		tl++;			/* Skip the header pad size. */
4946 
4947 		/* Make sure nm_wsize is small enough. */
4948 		maxval = fxdr_unsigned(uint32_t, *tl++);
4949 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4950 			if (nmp->nm_wsize > 8096)
4951 				nmp->nm_wsize /= 2;
4952 			else
4953 				break;
4954 		}
4955 		sep->nfsess_maxreq = maxval;
4956 
4957 		/* Make sure nm_rsize is small enough. */
4958 		maxval = fxdr_unsigned(uint32_t, *tl++);
4959 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4960 			if (nmp->nm_rsize > 8096)
4961 				nmp->nm_rsize /= 2;
4962 			else
4963 				break;
4964 		}
4965 		sep->nfsess_maxresp = maxval;
4966 
4967 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4968 		tl++;
4969 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4970 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4971 		irdcnt = fxdr_unsigned(int, *tl);
4972 		if (irdcnt > 0)
4973 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4974 
4975 		/* and the back channel slot count. */
4976 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4977 		tl += 5;
4978 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4979 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4980 	}
4981 	error = nd->nd_repstat;
4982 nfsmout:
4983 	m_freem(nd->nd_mrep);
4984 	return (error);
4985 }
4986 
4987 /*
4988  * Do the NFSv4.1 Destroy Session.
4989  */
4990 int
4991 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4992     struct ucred *cred, NFSPROC_T *p)
4993 {
4994 	uint32_t *tl;
4995 	struct nfsrv_descript nfsd;
4996 	struct nfsrv_descript *nd = &nfsd;
4997 	int error;
4998 	struct nfsclsession *tsep;
4999 
5000 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
5001 	    0);
5002 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
5003 	tsep = nfsmnt_mdssession(nmp);
5004 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
5005 	nd->nd_flag |= ND_USEGSSNAME;
5006 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5007 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5008 	if (error != 0)
5009 		return (error);
5010 	error = nd->nd_repstat;
5011 	m_freem(nd->nd_mrep);
5012 	return (error);
5013 }
5014 
5015 /*
5016  * Do the NFSv4.1 Destroy Client.
5017  */
5018 int
5019 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5020     struct ucred *cred, NFSPROC_T *p)
5021 {
5022 	uint32_t *tl;
5023 	struct nfsrv_descript nfsd;
5024 	struct nfsrv_descript *nd = &nfsd;
5025 	int error;
5026 	struct nfsclsession *tsep;
5027 
5028 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5029 	    0);
5030 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5031 	tsep = nfsmnt_mdssession(nmp);
5032 	*tl++ = tsep->nfsess_clientid.lval[0];
5033 	*tl = tsep->nfsess_clientid.lval[1];
5034 	nd->nd_flag |= ND_USEGSSNAME;
5035 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5036 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5037 	if (error != 0)
5038 		return (error);
5039 	error = nd->nd_repstat;
5040 	m_freem(nd->nd_mrep);
5041 	return (error);
5042 }
5043 
5044 /*
5045  * Do the NFSv4.1 LayoutGet.
5046  */
5047 static int
5048 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5049     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5050     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5051     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
5052     void *stuff)
5053 {
5054 	struct nfsrv_descript nfsd, *nd = &nfsd;
5055 	int error;
5056 
5057 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5058 	    0);
5059 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5060 	    layouttype, layoutlen, 0);
5061 	nd->nd_flag |= ND_USEGSSNAME;
5062 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5063 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5064 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5065 	if (error != 0)
5066 		return (error);
5067 	if (nd->nd_repstat == 0)
5068 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5069 		    flhp);
5070 	if (error == 0 && nd->nd_repstat != 0)
5071 		error = nd->nd_repstat;
5072 	m_freem(nd->nd_mrep);
5073 	return (error);
5074 }
5075 
5076 /*
5077  * Do the NFSv4.1 Get Device Info.
5078  */
5079 int
5080 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5081     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5082     NFSPROC_T *p)
5083 {
5084 	uint32_t cnt, *tl, vers, minorvers;
5085 	struct nfsrv_descript nfsd;
5086 	struct nfsrv_descript *nd = &nfsd;
5087 	struct sockaddr_in sin, ssin;
5088 	struct sockaddr_in6 sin6, ssin6;
5089 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5090 	struct nfscldevinfo *ndi;
5091 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5092 	int stripecnt;
5093 	uint8_t stripeindex;
5094 	sa_family_t af, safilled;
5095 
5096 	ssin.sin_port = 0;		/* To shut up compiler. */
5097 	ssin.sin_addr.s_addr = 0;	/* ditto */
5098 	*ndip = NULL;
5099 	ndi = NULL;
5100 	gotdspp = NULL;
5101 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5102 	    0);
5103 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5104 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5105 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5106 	*tl++ = txdr_unsigned(layouttype);
5107 	*tl++ = txdr_unsigned(100000);
5108 	if (notifybitsp != NULL && *notifybitsp != 0) {
5109 		*tl = txdr_unsigned(1);		/* One word of bits. */
5110 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5111 		*tl = txdr_unsigned(*notifybitsp);
5112 	} else
5113 		*tl = txdr_unsigned(0);
5114 	nd->nd_flag |= ND_USEGSSNAME;
5115 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5116 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5117 	if (error != 0)
5118 		return (error);
5119 	if (nd->nd_repstat == 0) {
5120 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5121 		if (layouttype != fxdr_unsigned(int, *tl))
5122 			printf("EEK! devinfo layout type not same!\n");
5123 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5124 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5125 			stripecnt = fxdr_unsigned(int, *tl);
5126 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5127 			if (stripecnt < 1 || stripecnt > 4096) {
5128 				printf("pNFS File layout devinfo stripecnt %d:"
5129 				    " out of range\n", stripecnt);
5130 				error = NFSERR_BADXDR;
5131 				goto nfsmout;
5132 			}
5133 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5134 			    NFSX_UNSIGNED);
5135 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5136 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5137 			if (addrcnt < 1 || addrcnt > 128) {
5138 				printf("NFS devinfo addrcnt %d: out of range\n",
5139 				    addrcnt);
5140 				error = NFSERR_BADXDR;
5141 				goto nfsmout;
5142 			}
5143 
5144 			/*
5145 			 * Now we know how many stripe indices and addresses, so
5146 			 * we can allocate the structure the correct size.
5147 			 */
5148 			i = (stripecnt * sizeof(uint8_t)) /
5149 			    sizeof(struct nfsclds *) + 1;
5150 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5151 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5152 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5153 			    M_ZERO);
5154 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5155 			    NFSX_V4DEVICEID);
5156 			ndi->nfsdi_refcnt = 0;
5157 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5158 			ndi->nfsdi_stripecnt = stripecnt;
5159 			ndi->nfsdi_addrcnt = addrcnt;
5160 			/* Fill in the stripe indices. */
5161 			for (i = 0; i < stripecnt; i++) {
5162 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5163 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5164 				if (stripeindex >= addrcnt) {
5165 					printf("pNFS File Layout devinfo"
5166 					    " stripeindex %d: too big\n",
5167 					    (int)stripeindex);
5168 					error = NFSERR_BADXDR;
5169 					goto nfsmout;
5170 				}
5171 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5172 			}
5173 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5174 			/* For Flex File, we only get one address list. */
5175 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5176 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5177 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5178 			    NFSX_V4DEVICEID);
5179 			ndi->nfsdi_refcnt = 0;
5180 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5181 			addrcnt = ndi->nfsdi_addrcnt = 1;
5182 		}
5183 
5184 		/* Now, dissect the server address(es). */
5185 		safilled = AF_UNSPEC;
5186 		for (i = 0; i < addrcnt; i++) {
5187 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5188 			cnt = fxdr_unsigned(uint32_t, *tl);
5189 			if (cnt == 0) {
5190 				printf("NFS devinfo 0 len addrlist\n");
5191 				error = NFSERR_BADXDR;
5192 				goto nfsmout;
5193 			}
5194 			dspp = nfsfldi_addr(ndi, i);
5195 			safilled = AF_UNSPEC;
5196 			for (j = 0; j < cnt; j++) {
5197 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5198 				    &isudp);
5199 				if (error != 0 && error != EPERM) {
5200 					error = NFSERR_BADXDR;
5201 					goto nfsmout;
5202 				}
5203 				if (error == 0 && isudp == 0) {
5204 					/*
5205 					 * The priority is:
5206 					 * - Same address family.
5207 					 * Save the address and dspp, so that
5208 					 * the connection can be done after
5209 					 * parsing is complete.
5210 					 */
5211 					if (safilled == AF_UNSPEC ||
5212 					    (af == nmp->nm_nam->sa_family &&
5213 					     safilled != nmp->nm_nam->sa_family)
5214 					   ) {
5215 						if (af == AF_INET)
5216 							ssin = sin;
5217 						else
5218 							ssin6 = sin6;
5219 						safilled = af;
5220 						gotdspp = dspp;
5221 					}
5222 				}
5223 			}
5224 		}
5225 
5226 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5227 		gotminor = NFSV41_MINORVERSION;
5228 		/* For Flex File, we will take one of the versions to use. */
5229 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5230 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5231 			j = fxdr_unsigned(int, *tl);
5232 			if (j < 1 || j > NFSDEV_MAXVERS) {
5233 				printf("pNFS: too many versions\n");
5234 				error = NFSERR_BADXDR;
5235 				goto nfsmout;
5236 			}
5237 			gotvers = 0;
5238 			gotminor = 0;
5239 			for (i = 0; i < j; i++) {
5240 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5241 				vers = fxdr_unsigned(uint32_t, *tl++);
5242 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5243 				if (vers == NFS_VER3)
5244 					minorvers = 0;
5245 				if ((vers == NFS_VER4 && ((minorvers ==
5246 				    NFSV41_MINORVERSION && gotminor == 0) ||
5247 				    minorvers == NFSV42_MINORVERSION)) ||
5248 				    (vers == NFS_VER3 && gotvers == 0)) {
5249 					gotvers = vers;
5250 					gotminor = minorvers;
5251 					/* We'll take this one. */
5252 					ndi->nfsdi_versindex = i;
5253 					ndi->nfsdi_vers = vers;
5254 					ndi->nfsdi_minorvers = minorvers;
5255 					ndi->nfsdi_rsize = fxdr_unsigned(
5256 					    uint32_t, *tl++);
5257 					ndi->nfsdi_wsize = fxdr_unsigned(
5258 					    uint32_t, *tl++);
5259 					if (*tl == newnfs_true)
5260 						ndi->nfsdi_flags |=
5261 						    NFSDI_TIGHTCOUPLED;
5262 					else
5263 						ndi->nfsdi_flags &=
5264 						    ~NFSDI_TIGHTCOUPLED;
5265 				}
5266 			}
5267 			if (gotvers == 0) {
5268 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5269 				error = NFSERR_BADXDR;
5270 				goto nfsmout;
5271 			}
5272 		}
5273 
5274 		/* And the notify bits. */
5275 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5276 		bitcnt = fxdr_unsigned(int, *tl);
5277 		if (bitcnt > 0) {
5278 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5279 			if (notifybitsp != NULL)
5280 				*notifybitsp =
5281 				    fxdr_unsigned(uint32_t, *tl);
5282 		}
5283 		if (safilled != AF_UNSPEC) {
5284 			KASSERT(ndi != NULL, ("ndi is NULL"));
5285 			*ndip = ndi;
5286 		} else
5287 			error = EPERM;
5288 		if (error == 0) {
5289 			/*
5290 			 * Now we can do a TCP connection for the correct
5291 			 * NFS version and IP address.
5292 			 */
5293 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5294 			    gotvers, gotminor, &dsp, p);
5295 		}
5296 		if (error == 0) {
5297 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5298 			*gotdspp = dsp;
5299 		}
5300 	}
5301 	if (nd->nd_repstat != 0 && error == 0)
5302 		error = nd->nd_repstat;
5303 nfsmout:
5304 	if (error != 0 && ndi != NULL)
5305 		nfscl_freedevinfo(ndi);
5306 	m_freem(nd->nd_mrep);
5307 	return (error);
5308 }
5309 
5310 /*
5311  * Do the NFSv4.1 LayoutCommit.
5312  */
5313 int
5314 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5315     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5316     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5317 {
5318 	uint32_t *tl;
5319 	struct nfsrv_descript nfsd, *nd = &nfsd;
5320 	int error;
5321 
5322 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5323 	    0, 0);
5324 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5325 	    NFSX_STATEID);
5326 	txdr_hyper(off, tl);
5327 	tl += 2;
5328 	txdr_hyper(len, tl);
5329 	tl += 2;
5330 	if (reclaim != 0)
5331 		*tl++ = newnfs_true;
5332 	else
5333 		*tl++ = newnfs_false;
5334 	*tl++ = txdr_unsigned(stateidp->seqid);
5335 	*tl++ = stateidp->other[0];
5336 	*tl++ = stateidp->other[1];
5337 	*tl++ = stateidp->other[2];
5338 	*tl++ = newnfs_true;
5339 	if (lastbyte < off)
5340 		lastbyte = off;
5341 	else if (lastbyte >= (off + len))
5342 		lastbyte = off + len - 1;
5343 	txdr_hyper(lastbyte, tl);
5344 	tl += 2;
5345 	*tl++ = newnfs_false;
5346 	*tl++ = txdr_unsigned(layouttype);
5347 	/* All supported layouts are 0 length. */
5348 	*tl = txdr_unsigned(0);
5349 	nd->nd_flag |= ND_USEGSSNAME;
5350 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5351 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5352 	if (error != 0)
5353 		return (error);
5354 	error = nd->nd_repstat;
5355 	m_freem(nd->nd_mrep);
5356 	return (error);
5357 }
5358 
5359 /*
5360  * Do the NFSv4.1 LayoutReturn.
5361  */
5362 int
5363 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5364     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5365     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5366     uint32_t stat, uint32_t op, char *devid)
5367 {
5368 	uint32_t *tl;
5369 	struct nfsrv_descript nfsd, *nd = &nfsd;
5370 	uint64_t tu64;
5371 	int error;
5372 
5373 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5374 	    0, 0);
5375 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5376 	if (reclaim != 0)
5377 		*tl++ = newnfs_true;
5378 	else
5379 		*tl++ = newnfs_false;
5380 	*tl++ = txdr_unsigned(layouttype);
5381 	*tl++ = txdr_unsigned(iomode);
5382 	*tl = txdr_unsigned(layoutreturn);
5383 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5384 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5385 		    NFSX_UNSIGNED);
5386 		txdr_hyper(offset, tl);
5387 		tl += 2;
5388 		txdr_hyper(len, tl);
5389 		tl += 2;
5390 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5391 		*tl++ = txdr_unsigned(stateidp->seqid);
5392 		*tl++ = stateidp->other[0];
5393 		*tl++ = stateidp->other[1];
5394 		*tl++ = stateidp->other[2];
5395 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5396 			*tl = txdr_unsigned(0);
5397 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5398 			if (stat != 0) {
5399 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5400 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5401 				    NFSX_UNSIGNED);
5402 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5403 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5404 				    NFSX_UNSIGNED);
5405 				*tl++ = txdr_unsigned(1);	/* One error. */
5406 				tu64 = 0;			/* Offset. */
5407 				txdr_hyper(tu64, tl); tl += 2;
5408 				tu64 = UINT64_MAX;		/* Length. */
5409 				txdr_hyper(tu64, tl); tl += 2;
5410 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5411 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5412 				*tl++ = txdr_unsigned(1);	/* One error. */
5413 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5414 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5415 				*tl++ = txdr_unsigned(stat);
5416 				*tl++ = txdr_unsigned(op);
5417 			} else {
5418 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5419 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5420 				/* No ioerrs. */
5421 				*tl++ = 0;
5422 			}
5423 			*tl = 0;	/* No stats yet. */
5424 		}
5425 	}
5426 	nd->nd_flag |= ND_USEGSSNAME;
5427 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5428 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5429 	if (error != 0)
5430 		return (error);
5431 	if (nd->nd_repstat == 0) {
5432 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5433 		if (*tl != 0) {
5434 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5435 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5436 			stateidp->other[0] = *tl++;
5437 			stateidp->other[1] = *tl++;
5438 			stateidp->other[2] = *tl;
5439 		}
5440 	} else
5441 		error = nd->nd_repstat;
5442 nfsmout:
5443 	m_freem(nd->nd_mrep);
5444 	return (error);
5445 }
5446 
5447 /*
5448  * Acquire a layout and devinfo, if possible. The caller must have acquired
5449  * a reference count on the nfsclclient structure before calling this.
5450  * Return the layout in lypp with a reference count on it, if successful.
5451  */
5452 static int
5453 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5454     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5455     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5456 {
5457 	struct nfscllayout *lyp;
5458 	struct nfsclflayout *flp;
5459 	struct nfsclflayouthead flh;
5460 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5461 	nfsv4stateid_t stateid;
5462 	struct nfsclsession *tsep;
5463 
5464 	*lypp = NULL;
5465 	if (NFSHASFLEXFILE(nmp))
5466 		layouttype = NFSLAYOUT_FLEXFILE;
5467 	else
5468 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5469 	/*
5470 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5471 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5472 	 * flp == NULL.
5473 	 */
5474 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5475 	    off, &flp, &recalled);
5476 	islocked = 0;
5477 	if (lyp == NULL || flp == NULL) {
5478 		if (recalled != 0)
5479 			return (EIO);
5480 		LIST_INIT(&flh);
5481 		tsep = nfsmnt_mdssession(nmp);
5482 		layoutlen = tsep->nfsess_maxcache -
5483 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5484 		if (lyp == NULL) {
5485 			stateid.seqid = 0;
5486 			stateid.other[0] = stateidp->other[0];
5487 			stateid.other[1] = stateidp->other[1];
5488 			stateid.other[2] = stateidp->other[2];
5489 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5490 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5491 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5492 			    &retonclose, &flh, cred, p, NULL);
5493 		} else {
5494 			islocked = 1;
5495 			stateid.seqid = lyp->nfsly_stateid.seqid;
5496 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5497 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5498 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5499 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5500 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5501 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5502 			    &retonclose, &flh, cred, p, NULL);
5503 		}
5504 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5505 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5506 		    &flh, layouttype, error, NULL, cred, p);
5507 		if (error == 0)
5508 			*lypp = lyp;
5509 		else if (islocked != 0)
5510 			nfscl_rellayout(lyp, 1);
5511 	} else
5512 		*lypp = lyp;
5513 	return (error);
5514 }
5515 
5516 /*
5517  * Do a TCP connection plus exchange id and create session.
5518  * If successful, a "struct nfsclds" is linked into the list for the
5519  * mount point and a pointer to it is returned.
5520  */
5521 static int
5522 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5523     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5524     struct nfsclds **dspp, NFSPROC_T *p)
5525 {
5526 	struct sockaddr_in *msad, *sad;
5527 	struct sockaddr_in6 *msad6, *sad6;
5528 	struct nfsclclient *clp;
5529 	struct nfssockreq *nrp;
5530 	struct nfsclds *dsp, *tdsp;
5531 	int error, firsttry;
5532 	enum nfsclds_state retv;
5533 	uint32_t sequenceid = 0;
5534 
5535 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5536 	    ("nfsrpc_fillsa: NULL nr_cred"));
5537 	NFSLOCKCLSTATE();
5538 	clp = nmp->nm_clp;
5539 	NFSUNLOCKCLSTATE();
5540 	if (clp == NULL)
5541 		return (EPERM);
5542 	if (af == AF_INET) {
5543 		NFSLOCKMNT(nmp);
5544 		/*
5545 		 * Check to see if we already have a session for this
5546 		 * address that is usable for a DS.
5547 		 * Note that the MDS's address is in a different place
5548 		 * than the sessions already acquired for DS's.
5549 		 */
5550 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5551 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5552 		while (tdsp != NULL) {
5553 			if (msad != NULL && msad->sin_family == AF_INET &&
5554 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5555 			    sin->sin_port == msad->sin_port &&
5556 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5557 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5558 				*dspp = tdsp;
5559 				NFSUNLOCKMNT(nmp);
5560 				NFSCL_DEBUG(4, "fnd same addr\n");
5561 				return (0);
5562 			}
5563 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5564 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5565 				msad = (struct sockaddr_in *)
5566 				    tdsp->nfsclds_sockp->nr_nam;
5567 			else
5568 				msad = NULL;
5569 		}
5570 		NFSUNLOCKMNT(nmp);
5571 
5572 		/* No IP address match, so look for new/trunked one. */
5573 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5574 		sad->sin_len = sizeof(*sad);
5575 		sad->sin_family = AF_INET;
5576 		sad->sin_port = sin->sin_port;
5577 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5578 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5579 		nrp->nr_nam = (struct sockaddr *)sad;
5580 	} else if (af == AF_INET6) {
5581 		NFSLOCKMNT(nmp);
5582 		/*
5583 		 * Check to see if we already have a session for this
5584 		 * address that is usable for a DS.
5585 		 * Note that the MDS's address is in a different place
5586 		 * than the sessions already acquired for DS's.
5587 		 */
5588 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5589 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5590 		while (tdsp != NULL) {
5591 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5592 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5593 			    &msad6->sin6_addr) &&
5594 			    sin6->sin6_port == msad6->sin6_port &&
5595 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5596 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5597 				*dspp = tdsp;
5598 				NFSUNLOCKMNT(nmp);
5599 				return (0);
5600 			}
5601 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5602 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5603 				msad6 = (struct sockaddr_in6 *)
5604 				    tdsp->nfsclds_sockp->nr_nam;
5605 			else
5606 				msad6 = NULL;
5607 		}
5608 		NFSUNLOCKMNT(nmp);
5609 
5610 		/* No IP address match, so look for new/trunked one. */
5611 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5612 		sad6->sin6_len = sizeof(*sad6);
5613 		sad6->sin6_family = AF_INET6;
5614 		sad6->sin6_port = sin6->sin6_port;
5615 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5616 		    sizeof(struct in6_addr));
5617 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5618 		nrp->nr_nam = (struct sockaddr *)sad6;
5619 	} else
5620 		return (EPERM);
5621 
5622 	nrp->nr_sotype = SOCK_STREAM;
5623 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5624 	nrp->nr_prog = NFS_PROG;
5625 	nrp->nr_vers = vers;
5626 
5627 	/*
5628 	 * Use the credentials that were used for the mount, which are
5629 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5630 	 * Ref. counting the credentials with crhold() is probably not
5631 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5632 	 * unmount, but I did it anyhow.
5633 	 */
5634 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5635 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false);
5636 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5637 
5638 	dsp = NULL;
5639 	/* Now, do the exchangeid and create session. */
5640 	if (error == 0) {
5641 		if (vers == NFS_VER4) {
5642 			firsttry = 0;
5643 			do {
5644 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5645 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5646 				    nrp->nr_cred, p);
5647 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5648 				if (error == NFSERR_MINORVERMISMATCH)
5649 					minorvers = NFSV42_MINORVERSION;
5650 			} while (error == NFSERR_MINORVERMISMATCH &&
5651 			    firsttry++ == 0);
5652 			if (error != 0)
5653 				newnfs_disconnect(nrp);
5654 		} else {
5655 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5656 			    M_WAITOK | M_ZERO);
5657 			dsp->nfsclds_flags |= NFSCLDS_DS;
5658 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5659 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5660 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5661 			    NULL, MTX_DEF);
5662 		}
5663 	}
5664 	if (error == 0) {
5665 		dsp->nfsclds_sockp = nrp;
5666 		if (vers == NFS_VER4) {
5667 			NFSLOCKMNT(nmp);
5668 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5669 			    &sequenceid);
5670 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5671 			if (retv == NFSDSP_USETHISSESSION &&
5672 			    nfscl_dssameconn != 0) {
5673 				NFSLOCKDS(tdsp);
5674 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5675 				NFSUNLOCKDS(tdsp);
5676 				NFSUNLOCKMNT(nmp);
5677 				/*
5678 				 * If there is already a session for this
5679 				 * server, use it.
5680 				 */
5681 				(void)newnfs_disconnect(nrp);
5682 				nfscl_freenfsclds(dsp);
5683 				*dspp = tdsp;
5684 				return (0);
5685 			}
5686 			if (retv == NFSDSP_NOTFOUND)
5687 				sequenceid =
5688 				    dsp->nfsclds_sess.nfsess_sequenceid;
5689 			NFSUNLOCKMNT(nmp);
5690 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5691 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
5692 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5693 		}
5694 	} else {
5695 		NFSFREECRED(nrp->nr_cred);
5696 		NFSFREEMUTEX(&nrp->nr_mtx);
5697 		free(nrp->nr_nam, M_SONAME);
5698 		free(nrp, M_NFSSOCKREQ);
5699 	}
5700 	if (error == 0) {
5701 		NFSCL_DEBUG(3, "add DS session\n");
5702 		/*
5703 		 * Put it at the end of the list. That way the list
5704 		 * is ordered by when the entry was added. This matters
5705 		 * since the one done first is the one that should be
5706 		 * used for sequencid'ing any subsequent create sessions.
5707 		 */
5708 		NFSLOCKMNT(nmp);
5709 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5710 		NFSUNLOCKMNT(nmp);
5711 		*dspp = dsp;
5712 	} else if (dsp != NULL) {
5713 		newnfs_disconnect(nrp);
5714 		nfscl_freenfsclds(dsp);
5715 	}
5716 	return (error);
5717 }
5718 
5719 /*
5720  * Do the NFSv4.1 Reclaim Complete.
5721  */
5722 int
5723 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5724 {
5725 	uint32_t *tl;
5726 	struct nfsrv_descript nfsd;
5727 	struct nfsrv_descript *nd = &nfsd;
5728 	int error;
5729 
5730 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5731 	    0);
5732 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5733 	*tl = newnfs_false;
5734 	nd->nd_flag |= ND_USEGSSNAME;
5735 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5736 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5737 	if (error != 0)
5738 		return (error);
5739 	error = nd->nd_repstat;
5740 	m_freem(nd->nd_mrep);
5741 	return (error);
5742 }
5743 
5744 /*
5745  * Initialize the slot tables for a session.
5746  */
5747 static void
5748 nfscl_initsessionslots(struct nfsclsession *sep)
5749 {
5750 	int i;
5751 
5752 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5753 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5754 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5755 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5756 	}
5757 	for (i = 0; i < 64; i++)
5758 		sep->nfsess_slotseq[i] = 0;
5759 	sep->nfsess_slots = 0;
5760 }
5761 
5762 /*
5763  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5764  */
5765 int
5766 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5767     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5768 {
5769 	struct nfsnode *np = VTONFS(vp);
5770 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5771 	struct nfscllayout *layp;
5772 	struct nfscldevinfo *dip;
5773 	struct nfsclflayout *rflp;
5774 	struct mbuf *m, *m2;
5775 	struct nfsclwritedsdorpc *drpc, *tdrpc;
5776 	nfsv4stateid_t stateid;
5777 	struct ucred *newcred;
5778 	uint64_t lastbyte, len, off, oresid, xfer;
5779 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5780 	void *lckp;
5781 	uint8_t *dev;
5782 	void *iovbase = NULL;
5783 	size_t iovlen = 0;
5784 	off_t offs = 0;
5785 	ssize_t resid = 0;
5786 
5787 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5788 	    (np->n_flag & NNOLAYOUT) != 0)
5789 		return (EIO);
5790 	/* Now, get a reference cnt on the clientid for this mount. */
5791 	if (nfscl_getref(nmp) == 0)
5792 		return (EIO);
5793 
5794 	/* Find an appropriate stateid. */
5795 	newcred = NFSNEWCRED(cred);
5796 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5797 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5798 	if (error != 0) {
5799 		NFSFREECRED(newcred);
5800 		nfscl_relref(nmp);
5801 		return (error);
5802 	}
5803 	/* Search for a layout for this file. */
5804 	off = uiop->uio_offset;
5805 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5806 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5807 	if (layp == NULL || rflp == NULL) {
5808 		if (recalled != 0) {
5809 			NFSFREECRED(newcred);
5810 			nfscl_relref(nmp);
5811 			return (EIO);
5812 		}
5813 		if (layp != NULL) {
5814 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5815 			layp = NULL;
5816 		}
5817 		/* Try and get a Layout, if it is supported. */
5818 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5819 		    (np->n_flag & NWRITEOPENED) != 0)
5820 			iolaymode = NFSLAYOUTIOMODE_RW;
5821 		else
5822 			iolaymode = NFSLAYOUTIOMODE_READ;
5823 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5824 		    NULL, &stateid, off, &layp, newcred, p);
5825 		if (error != 0) {
5826 			NFSLOCKNODE(np);
5827 			np->n_flag |= NNOLAYOUT;
5828 			NFSUNLOCKNODE(np);
5829 			if (lckp != NULL)
5830 				nfscl_lockderef(lckp);
5831 			NFSFREECRED(newcred);
5832 			if (layp != NULL)
5833 				nfscl_rellayout(layp, 0);
5834 			nfscl_relref(nmp);
5835 			return (error);
5836 		}
5837 	}
5838 
5839 	/*
5840 	 * Loop around finding a layout that works for the first part of
5841 	 * this I/O operation, and then call the function that actually
5842 	 * does the RPC.
5843 	 */
5844 	eof = 0;
5845 	len = (uint64_t)uiop->uio_resid;
5846 	while (len > 0 && error == 0 && eof == 0) {
5847 		off = uiop->uio_offset;
5848 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5849 		if (error == 0) {
5850 			oresid = xfer = (uint64_t)uiop->uio_resid;
5851 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5852 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5853 			/*
5854 			 * For Flex File layout with mirrored DSs, select one
5855 			 * of them at random for reads. For writes and commits,
5856 			 * do all mirrors.
5857 			 */
5858 			m = NULL;
5859 			tdrpc = drpc = NULL;
5860 			firstmirror = 0;
5861 			mirrorcnt = 1;
5862 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5863 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5864 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5865 					firstmirror = arc4random() % mirrorcnt;
5866 					mirrorcnt = firstmirror + 1;
5867 				} else {
5868 					if (docommit == 0) {
5869 						/*
5870 						 * Save values, so uiop can be
5871 						 * rolled back upon a write
5872 						 * error.
5873 						 */
5874 						offs = uiop->uio_offset;
5875 						resid = uiop->uio_resid;
5876 						iovbase =
5877 						    uiop->uio_iov->iov_base;
5878 						iovlen = uiop->uio_iov->iov_len;
5879 						m = nfsm_uiombuflist(uiop, len,
5880 						    0);
5881 					}
5882 					tdrpc = drpc = malloc(sizeof(*drpc) *
5883 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
5884 					    M_ZERO);
5885 				}
5886 			}
5887 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5888 				m2 = NULL;
5889 				if (m != NULL && i < mirrorcnt - 1)
5890 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
5891 				else {
5892 					m2 = m;
5893 					m = NULL;
5894 				}
5895 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
5896 					dev = rflp->nfsfl_ffm[i].dev;
5897 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5898 					    rflp->nfsfl_ffm[i].devp);
5899 				} else {
5900 					dev = rflp->nfsfl_dev;
5901 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5902 					    rflp->nfsfl_devp);
5903 				}
5904 				if (dip != NULL) {
5905 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5906 					    != 0)
5907 						error = nfscl_dofflayoutio(vp,
5908 						    uiop, iomode, must_commit,
5909 						    &eof, &stateid, rwaccess,
5910 						    dip, layp, rflp, off, xfer,
5911 						    i, docommit, m2, tdrpc,
5912 						    newcred, p);
5913 					else
5914 						error = nfscl_doflayoutio(vp,
5915 						    uiop, iomode, must_commit,
5916 						    &eof, &stateid, rwaccess,
5917 						    dip, layp, rflp, off, xfer,
5918 						    docommit, newcred, p);
5919 					nfscl_reldevinfo(dip);
5920 				} else {
5921 					if (m2 != NULL)
5922 						m_freem(m2);
5923 					error = EIO;
5924 				}
5925 				tdrpc++;
5926 			}
5927 			if (m != NULL)
5928 				m_freem(m);
5929 			tdrpc = drpc;
5930 			timo = hz / 50;		/* Wait for 20msec. */
5931 			if (timo < 1)
5932 				timo = 1;
5933 			for (i = firstmirror; i < mirrorcnt - 1 &&
5934 			    tdrpc != NULL; i++, tdrpc++) {
5935 				/*
5936 				 * For the unused drpc entries, both inprog and
5937 				 * err == 0, so this loop won't break.
5938 				 */
5939 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
5940 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5941 					    timo);
5942 				if (error == 0 && tdrpc->err != 0)
5943 					error = tdrpc->err;
5944 			}
5945 			free(drpc, M_TEMP);
5946 			if (error == 0) {
5947 				if (mirrorcnt > 1 && rwaccess ==
5948 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5949 					NFSLOCKCLSTATE();
5950 					layp->nfsly_flags |= NFSLY_WRITTEN;
5951 					NFSUNLOCKCLSTATE();
5952 				}
5953 				lastbyte = off + xfer - 1;
5954 				NFSLOCKCLSTATE();
5955 				if (lastbyte > layp->nfsly_lastbyte)
5956 					layp->nfsly_lastbyte = lastbyte;
5957 				NFSUNLOCKCLSTATE();
5958 			} else if (error == NFSERR_OPENMODE &&
5959 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
5960 				NFSLOCKMNT(nmp);
5961 				nmp->nm_state |= NFSSTA_OPENMODE;
5962 				NFSUNLOCKMNT(nmp);
5963 			} else
5964 				error = EIO;
5965 			if (error == 0)
5966 				len -= (oresid - (uint64_t)uiop->uio_resid);
5967 			else if (mirrorcnt > 1 && rwaccess ==
5968 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5969 				/*
5970 				 * In case the rpc gets retried, roll the
5971 				 * uio fields changed by nfsm_uiombuflist()
5972 				 * back.
5973 				 */
5974 				uiop->uio_offset = offs;
5975 				uiop->uio_resid = resid;
5976 				uiop->uio_iov->iov_base = iovbase;
5977 				uiop->uio_iov->iov_len = iovlen;
5978 			}
5979 		}
5980 	}
5981 	if (lckp != NULL)
5982 		nfscl_lockderef(lckp);
5983 	NFSFREECRED(newcred);
5984 	nfscl_rellayout(layp, 0);
5985 	nfscl_relref(nmp);
5986 	return (error);
5987 }
5988 
5989 /*
5990  * Find a file layout that will handle the first bytes of the requested
5991  * range and return the information from it needed to the I/O operation.
5992  */
5993 int
5994 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5995     struct nfsclflayout **retflpp)
5996 {
5997 	struct nfsclflayout *flp, *nflp, *rflp;
5998 	uint32_t rw;
5999 
6000 	rflp = NULL;
6001 	rw = rwaccess;
6002 	/* For reading, do the Read list first and then the Write list. */
6003 	do {
6004 		if (rw == NFSV4OPEN_ACCESSREAD)
6005 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6006 		else
6007 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6008 		while (flp != NULL) {
6009 			nflp = LIST_NEXT(flp, nfsfl_list);
6010 			if (flp->nfsfl_off > off)
6011 				break;
6012 			if (flp->nfsfl_end > off &&
6013 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6014 				rflp = flp;
6015 			flp = nflp;
6016 		}
6017 		if (rw == NFSV4OPEN_ACCESSREAD)
6018 			rw = NFSV4OPEN_ACCESSWRITE;
6019 		else
6020 			rw = 0;
6021 	} while (rw != 0);
6022 	if (rflp != NULL) {
6023 		/* This one covers the most bytes starting at off. */
6024 		*retflpp = rflp;
6025 		return (0);
6026 	}
6027 	return (EIO);
6028 }
6029 
6030 /*
6031  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6032  */
6033 static int
6034 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6035     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6036     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6037     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6038 {
6039 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6040 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6041 	struct nfsnode *np;
6042 	struct nfsfh *fhp;
6043 	struct nfsclds **dspp;
6044 
6045 	np = VTONFS(vp);
6046 	rel_off = off - flp->nfsfl_patoff;
6047 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6048 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6049 	    dp->nfsdi_stripecnt;
6050 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6051 	error = 0;
6052 
6053 	/* Loop around, doing I/O for each stripe unit. */
6054 	while (len > 0 && error == 0) {
6055 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6056 		dspp = nfsfldi_addr(dp, stripe_index);
6057 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6058 			minorvers = NFSV42_MINORVERSION;
6059 		else
6060 			minorvers = NFSV41_MINORVERSION;
6061 		if (len > transfer && docommit == 0)
6062 			xfer = transfer;
6063 		else
6064 			xfer = len;
6065 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6066 			/* Dense layout. */
6067 			if (stripe_pos >= flp->nfsfl_fhcnt)
6068 				return (EIO);
6069 			fhp = flp->nfsfl_fh[stripe_pos];
6070 			io_off = (rel_off / (stripe_unit_size *
6071 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6072 			    rel_off % stripe_unit_size;
6073 		} else {
6074 			/* Sparse layout. */
6075 			if (flp->nfsfl_fhcnt > 1) {
6076 				if (stripe_index >= flp->nfsfl_fhcnt)
6077 					return (EIO);
6078 				fhp = flp->nfsfl_fh[stripe_index];
6079 			} else if (flp->nfsfl_fhcnt == 1)
6080 				fhp = flp->nfsfl_fh[0];
6081 			else
6082 				fhp = np->n_fhp;
6083 			io_off = off;
6084 		}
6085 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6086 			commit_thru_mds = 1;
6087 			if (docommit != 0)
6088 				error = EIO;
6089 		} else {
6090 			commit_thru_mds = 0;
6091 			NFSLOCKNODE(np);
6092 			np->n_flag |= NDSCOMMIT;
6093 			NFSUNLOCKNODE(np);
6094 		}
6095 		if (docommit != 0) {
6096 			if (error == 0)
6097 				error = nfsrpc_commitds(vp, io_off, xfer,
6098 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6099 			if (error == 0) {
6100 				/*
6101 				 * Set both eof and uio_resid = 0 to end any
6102 				 * loops.
6103 				 */
6104 				*eofp = 1;
6105 				uiop->uio_resid = 0;
6106 			} else {
6107 				NFSLOCKNODE(np);
6108 				np->n_flag &= ~NDSCOMMIT;
6109 				NFSUNLOCKNODE(np);
6110 			}
6111 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6112 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6113 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6114 		else {
6115 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6116 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6117 			    0, NFS_VER4, minorvers, cred, p);
6118 			if (error == 0) {
6119 				NFSLOCKCLSTATE();
6120 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6121 				NFSUNLOCKCLSTATE();
6122 			}
6123 		}
6124 		if (error == 0) {
6125 			transfer = stripe_unit_size;
6126 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6127 			len -= xfer;
6128 			off += xfer;
6129 		}
6130 	}
6131 	return (error);
6132 }
6133 
6134 /*
6135  * Do I/O using an NFSv4.1 flex file layout.
6136  */
6137 static int
6138 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6139     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6140     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6141     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6142     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6143 {
6144 	uint64_t xfer;
6145 	int error;
6146 	struct nfsnode *np;
6147 	struct nfsfh *fhp;
6148 	struct nfsclds **dspp;
6149 	struct ucred *tcred;
6150 	struct mbuf *m, *m2;
6151 	uint32_t copylen;
6152 
6153 	np = VTONFS(vp);
6154 	error = 0;
6155 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6156 	    (uintmax_t)len);
6157 	/* Loop around, doing I/O for each stripe unit. */
6158 	while (len > 0 && error == 0) {
6159 		dspp = nfsfldi_addr(dp, 0);
6160 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6161 		stateidp = &flp->nfsfl_ffm[mirror].st;
6162 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6163 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6164 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6165 			tcred = NFSNEWCRED(cred);
6166 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6167 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6168 			tcred->cr_ngroups = 1;
6169 		} else
6170 			tcred = cred;
6171 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6172 			copylen = dp->nfsdi_rsize;
6173 		else {
6174 			copylen = dp->nfsdi_wsize;
6175 			if (len > copylen && mp != NULL) {
6176 				/*
6177 				 * When a mirrored configuration needs to do
6178 				 * multiple writes to each mirror, all writes
6179 				 * except the last one must be a multiple of
6180 				 * 4 bytes.  This is required so that the XDR
6181 				 * does not need padding.
6182 				 * If possible, clip the size to an exact
6183 				 * multiple of the mbuf length, so that the
6184 				 * split will be on an mbuf boundary.
6185 				 */
6186 				copylen &= 0xfffffffc;
6187 				if (copylen > mp->m_len)
6188 					copylen = copylen / mp->m_len *
6189 					    mp->m_len;
6190 			}
6191 		}
6192 		NFSLOCKNODE(np);
6193 		np->n_flag |= NDSCOMMIT;
6194 		NFSUNLOCKNODE(np);
6195 		if (len > copylen && docommit == 0)
6196 			xfer = copylen;
6197 		else
6198 			xfer = len;
6199 		if (docommit != 0) {
6200 			if (error == 0) {
6201 				/*
6202 				 * Do last mirrored DS commit with this thread.
6203 				 */
6204 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6205 					error = nfsio_commitds(vp, off, xfer,
6206 					    *dspp, fhp, dp->nfsdi_vers,
6207 					    dp->nfsdi_minorvers, drpc, tcred,
6208 					    p);
6209 				else
6210 					error = nfsrpc_commitds(vp, off, xfer,
6211 					    *dspp, fhp, dp->nfsdi_vers,
6212 					    dp->nfsdi_minorvers, tcred, p);
6213 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6214 				if (error != 0 && error != EACCES && error !=
6215 				    ESTALE) {
6216 					NFSCL_DEBUG(4,
6217 					    "DS layreterr for commit\n");
6218 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6219 					    lyp, *dspp);
6220 				}
6221 			}
6222 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6223 			if (error == 0) {
6224 				/*
6225 				 * Set both eof and uio_resid = 0 to end any
6226 				 * loops.
6227 				 */
6228 				*eofp = 1;
6229 				uiop->uio_resid = 0;
6230 			} else {
6231 				NFSLOCKNODE(np);
6232 				np->n_flag &= ~NDSCOMMIT;
6233 				NFSUNLOCKNODE(np);
6234 			}
6235 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6236 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6237 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6238 			    dp->nfsdi_minorvers, tcred, p);
6239 			NFSCL_DEBUG(4, "readds=%d\n", error);
6240 			if (error != 0 && error != EACCES && error != ESTALE) {
6241 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6242 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6243 				    *dspp);
6244 			}
6245 		} else {
6246 			if (flp->nfsfl_mirrorcnt == 1) {
6247 				error = nfsrpc_writeds(vp, uiop, iomode,
6248 				    must_commit, stateidp, *dspp, off, xfer,
6249 				    fhp, 0, 1, dp->nfsdi_vers,
6250 				    dp->nfsdi_minorvers, tcred, p);
6251 				if (error == 0) {
6252 					NFSLOCKCLSTATE();
6253 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6254 					NFSUNLOCKCLSTATE();
6255 				}
6256 			} else {
6257 				m = mp;
6258 				if (xfer < len) {
6259 					/* The mbuf list must be split. */
6260 					m2 = nfsm_split(mp, xfer);
6261 					if (m2 != NULL)
6262 						mp = m2;
6263 					else {
6264 						m_freem(mp);
6265 						error = EIO;
6266 					}
6267 				}
6268 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6269 				    (uintmax_t)len, (uintmax_t)xfer);
6270 				/*
6271 				 * Do last write to a mirrored DS with this
6272 				 * thread.
6273 				 */
6274 				if (error == 0) {
6275 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6276 						error = nfsio_writedsmir(vp,
6277 						    iomode, must_commit,
6278 						    stateidp, *dspp, off,
6279 						    xfer, fhp, m,
6280 						    dp->nfsdi_vers,
6281 						    dp->nfsdi_minorvers, drpc,
6282 						    tcred, p);
6283 					else
6284 						error = nfsrpc_writedsmir(vp,
6285 						    iomode, must_commit,
6286 						    stateidp, *dspp, off,
6287 						    xfer, fhp, m,
6288 						    dp->nfsdi_vers,
6289 						    dp->nfsdi_minorvers, tcred,
6290 						    p);
6291 				}
6292 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6293 				if (error != 0 && error != EACCES && error !=
6294 				    ESTALE) {
6295 					NFSCL_DEBUG(4,
6296 					    "DS layreterr for write\n");
6297 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6298 					    lyp, *dspp);
6299 				}
6300 			}
6301 		}
6302 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6303 		if (error == 0) {
6304 			len -= xfer;
6305 			off += xfer;
6306 		}
6307 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6308 			NFSFREECRED(tcred);
6309 	}
6310 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6311 	return (error);
6312 }
6313 
6314 /*
6315  * The actual read RPC done to a DS.
6316  */
6317 static int
6318 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6319     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6320     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6321 {
6322 	uint32_t *tl;
6323 	int attrflag, error, retlen;
6324 	struct nfsrv_descript nfsd;
6325 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6326 	struct nfsrv_descript *nd = &nfsd;
6327 	struct nfssockreq *nrp;
6328 	struct nfsvattr na;
6329 
6330 	nd->nd_mrep = NULL;
6331 	if (vers == 0 || vers == NFS_VER4) {
6332 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6333 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6334 		vers = NFS_VER4;
6335 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6336 		if (flex != 0)
6337 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6338 		else
6339 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6340 	} else {
6341 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6342 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6343 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6344 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6345 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6346 	}
6347 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6348 	txdr_hyper(io_off, tl);
6349 	*(tl + 2) = txdr_unsigned(len);
6350 	nrp = dsp->nfsclds_sockp;
6351 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6352 	if (nrp == NULL)
6353 		/* If NULL, use the MDS socket. */
6354 		nrp = &nmp->nm_sockreq;
6355 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6356 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6357 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6358 	    error);
6359 	if (error != 0)
6360 		return (error);
6361 	if (vers == NFS_VER3) {
6362 		error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6363 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6364 		if (error != 0)
6365 			goto nfsmout;
6366 	}
6367 	if (nd->nd_repstat != 0) {
6368 		error = nd->nd_repstat;
6369 		goto nfsmout;
6370 	}
6371 	if (vers == NFS_VER3) {
6372 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6373 		*eofp = fxdr_unsigned(int, *(tl + 1));
6374 	} else {
6375 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6376 		*eofp = fxdr_unsigned(int, *tl);
6377 	}
6378 	NFSM_STRSIZ(retlen, len);
6379 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6380 	error = nfsm_mbufuio(nd, uiop, retlen);
6381 nfsmout:
6382 	if (nd->nd_mrep != NULL)
6383 		m_freem(nd->nd_mrep);
6384 	return (error);
6385 }
6386 
6387 /*
6388  * The actual write RPC done to a DS.
6389  */
6390 static int
6391 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6392     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6393     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6394     struct ucred *cred, NFSPROC_T *p)
6395 {
6396 	uint32_t *tl;
6397 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6398 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6399 	int32_t backup;
6400 	struct nfsrv_descript nfsd;
6401 	struct nfsrv_descript *nd = &nfsd;
6402 	struct nfssockreq *nrp;
6403 	struct nfsvattr na;
6404 
6405 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6406 	nd->nd_mrep = NULL;
6407 	if (vers == 0 || vers == NFS_VER4) {
6408 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6409 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6410 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6411 		vers = NFS_VER4;
6412 		if (flex != 0)
6413 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6414 		else
6415 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6416 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6417 	} else {
6418 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6419 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6420 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6421 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6422 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6423 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6424 	}
6425 	txdr_hyper(io_off, tl);
6426 	tl += 2;
6427 	if (vers == NFS_VER3)
6428 		*tl++ = txdr_unsigned(len);
6429 	*tl++ = txdr_unsigned(*iomode);
6430 	*tl = txdr_unsigned(len);
6431 	nfsm_uiombuf(nd, uiop, len);
6432 	nrp = dsp->nfsclds_sockp;
6433 	if (nrp == NULL)
6434 		/* If NULL, use the MDS socket. */
6435 		nrp = &nmp->nm_sockreq;
6436 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6437 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6438 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6439 	    nd->nd_repstat);
6440 	if (error != 0)
6441 		return (error);
6442 	if (nd->nd_repstat != 0) {
6443 		/*
6444 		 * In case the rpc gets retried, roll
6445 		 * the uio fields changed by nfsm_uiombuf()
6446 		 * back.
6447 		 */
6448 		uiop->uio_offset -= len;
6449 		uiop->uio_resid += len;
6450 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6451 		uiop->uio_iov->iov_len += len;
6452 		error = nd->nd_repstat;
6453 	} else {
6454 		if (vers == NFS_VER3) {
6455 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6456 			    NULL);
6457 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6458 			if (error != 0)
6459 				goto nfsmout;
6460 		}
6461 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6462 		rlen = fxdr_unsigned(int, *tl++);
6463 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6464 		if (rlen == 0) {
6465 			error = NFSERR_IO;
6466 			goto nfsmout;
6467 		} else if (rlen < len) {
6468 			backup = len - rlen;
6469 			uiop->uio_iov->iov_base =
6470 			    (char *)uiop->uio_iov->iov_base - backup;
6471 			uiop->uio_iov->iov_len += backup;
6472 			uiop->uio_offset -= backup;
6473 			uiop->uio_resid += backup;
6474 			len = rlen;
6475 		}
6476 		commit = fxdr_unsigned(int, *tl++);
6477 
6478 		/*
6479 		 * Return the lowest commitment level
6480 		 * obtained by any of the RPCs.
6481 		 */
6482 		if (committed == NFSWRITE_FILESYNC)
6483 			committed = commit;
6484 		else if (committed == NFSWRITE_DATASYNC &&
6485 		    commit == NFSWRITE_UNSTABLE)
6486 			committed = commit;
6487 		if (commit_thru_mds != 0) {
6488 			NFSLOCKMNT(nmp);
6489 			if (!NFSHASWRITEVERF(nmp)) {
6490 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6491 				NFSSETWRITEVERF(nmp);
6492 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6493 				*must_commit = 1;
6494 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6495 			}
6496 			NFSUNLOCKMNT(nmp);
6497 		} else {
6498 			NFSLOCKDS(dsp);
6499 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6500 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6501 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6502 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6503 				*must_commit = 1;
6504 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6505 			}
6506 			NFSUNLOCKDS(dsp);
6507 		}
6508 	}
6509 nfsmout:
6510 	if (nd->nd_mrep != NULL)
6511 		m_freem(nd->nd_mrep);
6512 	*iomode = committed;
6513 	if (nd->nd_repstat != 0 && error == 0)
6514 		error = nd->nd_repstat;
6515 	return (error);
6516 }
6517 
6518 /*
6519  * The actual write RPC done to a DS.
6520  * This variant is called from a separate kernel process for mirrors.
6521  * Any short write is considered an IO error.
6522  */
6523 static int
6524 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6525     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6526     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6527     struct ucred *cred, NFSPROC_T *p)
6528 {
6529 	uint32_t *tl;
6530 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6531 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6532 	struct nfsrv_descript nfsd;
6533 	struct nfsrv_descript *nd = &nfsd;
6534 	struct nfssockreq *nrp;
6535 	struct nfsvattr na;
6536 
6537 	nd->nd_mrep = NULL;
6538 	if (vers == 0 || vers == NFS_VER4) {
6539 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6540 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6541 		vers = NFS_VER4;
6542 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6543 		    minorvers);
6544 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6545 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6546 	} else {
6547 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6548 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6549 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6550 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6551 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6552 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6553 	}
6554 	txdr_hyper(io_off, tl);
6555 	tl += 2;
6556 	if (vers == NFS_VER3)
6557 		*tl++ = txdr_unsigned(len);
6558 	*tl++ = txdr_unsigned(*iomode);
6559 	*tl = txdr_unsigned(len);
6560 	if (len > 0) {
6561 		/* Put data in mbuf chain. */
6562 		nd->nd_mb->m_next = m;
6563 	}
6564 	nrp = dsp->nfsclds_sockp;
6565 	if (nrp == NULL)
6566 		/* If NULL, use the MDS socket. */
6567 		nrp = &nmp->nm_sockreq;
6568 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6569 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6570 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6571 	    nd->nd_repstat);
6572 	if (error != 0)
6573 		return (error);
6574 	if (nd->nd_repstat != 0)
6575 		error = nd->nd_repstat;
6576 	else {
6577 		if (vers == NFS_VER3) {
6578 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6579 			    NULL);
6580 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6581 			    error);
6582 			if (error != 0)
6583 				goto nfsmout;
6584 		}
6585 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6586 		rlen = fxdr_unsigned(int, *tl++);
6587 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6588 		    rlen);
6589 		if (rlen != len) {
6590 			error = NFSERR_IO;
6591 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6592 			    len, rlen);
6593 			goto nfsmout;
6594 		}
6595 		commit = fxdr_unsigned(int, *tl++);
6596 
6597 		/*
6598 		 * Return the lowest commitment level
6599 		 * obtained by any of the RPCs.
6600 		 */
6601 		if (committed == NFSWRITE_FILESYNC)
6602 			committed = commit;
6603 		else if (committed == NFSWRITE_DATASYNC &&
6604 		    commit == NFSWRITE_UNSTABLE)
6605 			committed = commit;
6606 		NFSLOCKDS(dsp);
6607 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6608 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6609 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6610 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6611 			*must_commit = 1;
6612 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6613 		}
6614 		NFSUNLOCKDS(dsp);
6615 	}
6616 nfsmout:
6617 	if (nd->nd_mrep != NULL)
6618 		m_freem(nd->nd_mrep);
6619 	*iomode = committed;
6620 	if (nd->nd_repstat != 0 && error == 0)
6621 		error = nd->nd_repstat;
6622 	return (error);
6623 }
6624 
6625 /*
6626  * Start up the thread that will execute nfsrpc_writedsmir().
6627  */
6628 static void
6629 start_writedsmir(void *arg, int pending)
6630 {
6631 	struct nfsclwritedsdorpc *drpc;
6632 
6633 	drpc = (struct nfsclwritedsdorpc *)arg;
6634 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6635 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6636 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6637 	    drpc->p);
6638 	drpc->done = 1;
6639 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6640 }
6641 
6642 /*
6643  * Set up the write DS mirror call for the pNFS I/O thread.
6644  */
6645 static int
6646 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6647     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6648     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6649     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6650 {
6651 	int error, ret;
6652 
6653 	error = 0;
6654 	drpc->done = 0;
6655 	drpc->vp = vp;
6656 	drpc->iomode = *iomode;
6657 	drpc->must_commit = *must_commit;
6658 	drpc->stateidp = stateidp;
6659 	drpc->dsp = dsp;
6660 	drpc->off = off;
6661 	drpc->len = len;
6662 	drpc->fhp = fhp;
6663 	drpc->m = m;
6664 	drpc->vers = vers;
6665 	drpc->minorvers = minorvers;
6666 	drpc->cred = cred;
6667 	drpc->p = p;
6668 	drpc->inprog = 0;
6669 	ret = EIO;
6670 	if (nfs_pnfsiothreads != 0) {
6671 		ret = nfs_pnfsio(start_writedsmir, drpc);
6672 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6673 	}
6674 	if (ret != 0)
6675 		error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6676 		    dsp, off, len, fhp, m, vers, minorvers, cred, p);
6677 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6678 	return (error);
6679 }
6680 
6681 /*
6682  * Free up the nfsclds structure.
6683  */
6684 void
6685 nfscl_freenfsclds(struct nfsclds *dsp)
6686 {
6687 	int i;
6688 
6689 	if (dsp == NULL)
6690 		return;
6691 	if (dsp->nfsclds_sockp != NULL) {
6692 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6693 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6694 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6695 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6696 	}
6697 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
6698 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6699 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6700 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6701 			m_freem(
6702 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6703 	}
6704 	free(dsp, M_NFSCLDS);
6705 }
6706 
6707 static enum nfsclds_state
6708 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6709     struct nfsclds **retdspp, uint32_t *sequencep)
6710 {
6711 	struct nfsclds *dsp;
6712 	int fndseq;
6713 
6714 	/*
6715 	 * Search the list of nfsclds structures for one with the same
6716 	 * server.
6717 	 */
6718 	fndseq = 0;
6719 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6720 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6721 		    dsp->nfsclds_servownlen != 0 &&
6722 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6723 		    dsp->nfsclds_servownlen) &&
6724 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
6725 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6726 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
6727 			    dsp->nfsclds_flags);
6728 			if (fndseq == 0) {
6729 				/* Get sequenceid# from first entry. */
6730 				*sequencep =
6731 				    dsp->nfsclds_sess.nfsess_sequenceid;
6732 				fndseq = 1;
6733 			}
6734 			/* Server major id matches. */
6735 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6736 				*retdspp = dsp;
6737 				return (NFSDSP_USETHISSESSION);
6738 			}
6739 		}
6740 	}
6741 	if (fndseq != 0)
6742 		return (NFSDSP_SEQTHISSESSION);
6743 	return (NFSDSP_NOTFOUND);
6744 }
6745 
6746 /*
6747  * NFS commit rpc to a NFSv4.1 DS.
6748  */
6749 static int
6750 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6751     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6752     NFSPROC_T *p)
6753 {
6754 	uint32_t *tl;
6755 	struct nfsrv_descript nfsd, *nd = &nfsd;
6756 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6757 	struct nfssockreq *nrp;
6758 	struct nfsvattr na;
6759 	int attrflag, error;
6760 
6761 	nd->nd_mrep = NULL;
6762 	if (vers == 0 || vers == NFS_VER4) {
6763 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6764 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6765 		vers = NFS_VER4;
6766 	} else {
6767 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6768 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6769 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
6770 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
6771 	}
6772 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6773 	    minorvers);
6774 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6775 	txdr_hyper(offset, tl);
6776 	tl += 2;
6777 	*tl = txdr_unsigned(cnt);
6778 	nrp = dsp->nfsclds_sockp;
6779 	if (nrp == NULL)
6780 		/* If NULL, use the MDS socket. */
6781 		nrp = &nmp->nm_sockreq;
6782 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6783 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6784 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6785 	    nd->nd_repstat);
6786 	if (error != 0)
6787 		return (error);
6788 	if (nd->nd_repstat == 0) {
6789 		if (vers == NFS_VER3) {
6790 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6791 			    NULL);
6792 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6793 			if (error != 0)
6794 				goto nfsmout;
6795 		}
6796 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6797 		NFSLOCKDS(dsp);
6798 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6799 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6800 			error = NFSERR_STALEWRITEVERF;
6801 		}
6802 		NFSUNLOCKDS(dsp);
6803 	}
6804 nfsmout:
6805 	if (error == 0 && nd->nd_repstat != 0)
6806 		error = nd->nd_repstat;
6807 	m_freem(nd->nd_mrep);
6808 	return (error);
6809 }
6810 
6811 /*
6812  * Start up the thread that will execute nfsrpc_commitds().
6813  */
6814 static void
6815 start_commitds(void *arg, int pending)
6816 {
6817 	struct nfsclwritedsdorpc *drpc;
6818 
6819 	drpc = (struct nfsclwritedsdorpc *)arg;
6820 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6821 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6822 	    drpc->p);
6823 	drpc->done = 1;
6824 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6825 }
6826 
6827 /*
6828  * Set up the commit DS mirror call for the pNFS I/O thread.
6829  */
6830 static int
6831 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6832     struct nfsfh *fhp, int vers, int minorvers,
6833     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6834 {
6835 	int error, ret;
6836 
6837 	error = 0;
6838 	drpc->done = 0;
6839 	drpc->vp = vp;
6840 	drpc->off = offset;
6841 	drpc->len = cnt;
6842 	drpc->dsp = dsp;
6843 	drpc->fhp = fhp;
6844 	drpc->vers = vers;
6845 	drpc->minorvers = minorvers;
6846 	drpc->cred = cred;
6847 	drpc->p = p;
6848 	drpc->inprog = 0;
6849 	ret = EIO;
6850 	if (nfs_pnfsiothreads != 0) {
6851 		ret = nfs_pnfsio(start_commitds, drpc);
6852 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6853 	}
6854 	if (ret != 0)
6855 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6856 		    minorvers, cred, p);
6857 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6858 	return (error);
6859 }
6860 
6861 /*
6862  * NFS Advise rpc
6863  */
6864 int
6865 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
6866     struct ucred *cred, NFSPROC_T *p)
6867 {
6868 	u_int32_t *tl;
6869 	struct nfsrv_descript nfsd, *nd = &nfsd;
6870 	nfsattrbit_t hints;
6871 	int error;
6872 
6873 	NFSZERO_ATTRBIT(&hints);
6874 	if (advise == POSIX_FADV_WILLNEED)
6875 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6876 	else if (advise == POSIX_FADV_DONTNEED)
6877 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6878 	else
6879 		return (0);
6880 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
6881 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6882 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
6883 	txdr_hyper(offset, tl);
6884 	tl += 2;
6885 	txdr_hyper(cnt, tl);
6886 	nfsrv_putattrbit(nd, &hints);
6887 	error = nfscl_request(nd, vp, p, cred, NULL);
6888 	if (error != 0)
6889 		return (error);
6890 	if (nd->nd_repstat != 0)
6891 		error = nd->nd_repstat;
6892 	m_freem(nd->nd_mrep);
6893 	return (error);
6894 }
6895 
6896 #ifdef notyet
6897 /*
6898  * NFS advise rpc to a NFSv4.2 DS.
6899  */
6900 static int
6901 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6902     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6903     struct ucred *cred, NFSPROC_T *p)
6904 {
6905 	uint32_t *tl;
6906 	struct nfsrv_descript nfsd, *nd = &nfsd;
6907 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6908 	struct nfssockreq *nrp;
6909 	nfsattrbit_t hints;
6910 	int error;
6911 
6912 	/* For NFS DSs prior to NFSv4.2, just return OK. */
6913 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
6914 		return (0);
6915 	NFSZERO_ATTRBIT(&hints);
6916 	if (advise == POSIX_FADV_WILLNEED)
6917 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6918 	else if (advise == POSIX_FADV_DONTNEED)
6919 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6920 	else
6921 		return (0);
6922 	nd->nd_mrep = NULL;
6923 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
6924 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6925 	vers = NFS_VER4;
6926 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
6927 	    minorvers);
6928 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6929 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6930 	txdr_hyper(offset, tl);
6931 	tl += 2;
6932 	*tl = txdr_unsigned(cnt);
6933 	nfsrv_putattrbit(nd, &hints);
6934 	nrp = dsp->nfsclds_sockp;
6935 	if (nrp == NULL)
6936 		/* If NULL, use the MDS socket. */
6937 		nrp = &nmp->nm_sockreq;
6938 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6939 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6940 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
6941 	    nd->nd_repstat);
6942 	if (error != 0)
6943 		return (error);
6944 	if (nd->nd_repstat != 0)
6945 		error = nd->nd_repstat;
6946 	m_freem(nd->nd_mrep);
6947 	return (error);
6948 }
6949 
6950 /*
6951  * Start up the thread that will execute nfsrpc_commitds().
6952  */
6953 static void
6954 start_adviseds(void *arg, int pending)
6955 {
6956 	struct nfsclwritedsdorpc *drpc;
6957 
6958 	drpc = (struct nfsclwritedsdorpc *)arg;
6959 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
6960 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
6961 	    drpc->cred, drpc->p);
6962 	drpc->done = 1;
6963 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
6964 }
6965 
6966 /*
6967  * Set up the commit DS mirror call for the pNFS I/O thread.
6968  */
6969 static int
6970 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6971     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6972     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6973 {
6974 	int error, ret;
6975 
6976 	error = 0;
6977 	drpc->done = 0;
6978 	drpc->vp = vp;
6979 	drpc->off = offset;
6980 	drpc->len = cnt;
6981 	drpc->advise = advise;
6982 	drpc->dsp = dsp;
6983 	drpc->fhp = fhp;
6984 	drpc->vers = vers;
6985 	drpc->minorvers = minorvers;
6986 	drpc->cred = cred;
6987 	drpc->p = p;
6988 	drpc->inprog = 0;
6989 	ret = EIO;
6990 	if (nfs_pnfsiothreads != 0) {
6991 		ret = nfs_pnfsio(start_adviseds, drpc);
6992 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
6993 	}
6994 	if (ret != 0)
6995 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
6996 		    minorvers, cred, p);
6997 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
6998 	return (error);
6999 }
7000 #endif	/* notyet */
7001 
7002 /*
7003  * Do the Allocate operation, retrying for recovery.
7004  */
7005 int
7006 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
7007     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
7008 {
7009 	int error, expireret = 0, retrycnt, nostateid;
7010 	uint32_t clidrev = 0;
7011 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7012 	struct nfsfh *nfhp = NULL;
7013 	nfsv4stateid_t stateid;
7014 	off_t tmp_off;
7015 	void *lckp;
7016 
7017 	if (len < 0)
7018 		return (EINVAL);
7019 	if (len == 0)
7020 		return (0);
7021 	tmp_off = off + len;
7022 	NFSLOCKMNT(nmp);
7023 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7024 		NFSUNLOCKMNT(nmp);
7025 		return (EFBIG);
7026 	}
7027 	if (nmp->nm_clp != NULL)
7028 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7029 	NFSUNLOCKMNT(nmp);
7030 	nfhp = VTONFS(vp)->n_fhp;
7031 	retrycnt = 0;
7032 	do {
7033 		lckp = NULL;
7034 		nostateid = 0;
7035 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7036 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7037 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7038 		    stateid.other[2] == 0) {
7039 			nostateid = 1;
7040 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7041 		}
7042 
7043 		/*
7044 		 * Not finding a stateid should probably never happen,
7045 		 * but just return an error for this case.
7046 		 */
7047 		if (nostateid != 0)
7048 			error = EIO;
7049 		else
7050 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7051 			    nap, attrflagp, cred, p, stuff);
7052 		if (error == NFSERR_STALESTATEID)
7053 			nfscl_initiate_recovery(nmp->nm_clp);
7054 		if (lckp != NULL)
7055 			nfscl_lockderef(lckp);
7056 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7057 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7058 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7059 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7060 		} else if ((error == NFSERR_EXPIRED ||
7061 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7062 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7063 		}
7064 		retrycnt++;
7065 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7066 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7067 	    error == NFSERR_STALEDONTRECOVER ||
7068 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7069 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7070 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7071 	if (error != 0 && retrycnt >= 4)
7072 		error = EIO;
7073 	return (error);
7074 }
7075 
7076 /*
7077  * The allocate RPC.
7078  */
7079 static int
7080 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7081     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p,
7082     void *stuff)
7083 {
7084 	uint32_t *tl;
7085 	int error;
7086 	struct nfsrv_descript nfsd;
7087 	struct nfsrv_descript *nd = &nfsd;
7088 	nfsattrbit_t attrbits;
7089 
7090 	*attrflagp = 0;
7091 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7092 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7093 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7094 	txdr_hyper(off, tl); tl += 2;
7095 	txdr_hyper(len, tl); tl += 2;
7096 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7097 	NFSGETATTR_ATTRBIT(&attrbits);
7098 	nfsrv_putattrbit(nd, &attrbits);
7099 	error = nfscl_request(nd, vp, p, cred, stuff);
7100 	if (error != 0)
7101 		return (error);
7102 	if (nd->nd_repstat == 0) {
7103 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7104 		error = nfsm_loadattr(nd, nap);
7105 		if (error == 0)
7106 			*attrflagp = NFS_LATTR_NOSHRINK;
7107 	} else
7108 		error = nd->nd_repstat;
7109 nfsmout:
7110 	m_freem(nd->nd_mrep);
7111 	return (error);
7112 }
7113 
7114 /*
7115  * Set up the XDR arguments for the LayoutGet operation.
7116  */
7117 static void
7118 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7119     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7120     int layoutlen, int usecurstateid)
7121 {
7122 	uint32_t *tl;
7123 
7124 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7125 	    NFSX_STATEID);
7126 	*tl++ = newnfs_false;		/* Don't signal availability. */
7127 	*tl++ = txdr_unsigned(layouttype);
7128 	*tl++ = txdr_unsigned(iomode);
7129 	txdr_hyper(offset, tl);
7130 	tl += 2;
7131 	txdr_hyper(len, tl);
7132 	tl += 2;
7133 	txdr_hyper(minlen, tl);
7134 	tl += 2;
7135 	if (usecurstateid != 0) {
7136 		/* Special stateid for Current stateid. */
7137 		*tl++ = txdr_unsigned(1);
7138 		*tl++ = 0;
7139 		*tl++ = 0;
7140 		*tl++ = 0;
7141 	} else {
7142 		*tl++ = txdr_unsigned(stateidp->seqid);
7143 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7144 		*tl++ = stateidp->other[0];
7145 		*tl++ = stateidp->other[1];
7146 		*tl++ = stateidp->other[2];
7147 	}
7148 	*tl = txdr_unsigned(layoutlen);
7149 }
7150 
7151 /*
7152  * Parse the reply for a successful LayoutGet operation.
7153  */
7154 static int
7155 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7156     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7157 {
7158 	uint32_t *tl;
7159 	struct nfsclflayout *flp, *prevflp, *tflp;
7160 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7161 	int m, mirrorcnt;
7162 	uint64_t retlen, off;
7163 	struct nfsfh *nfhp;
7164 	uint8_t *cp;
7165 	uid_t user;
7166 	gid_t grp;
7167 
7168 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7169 	error = 0;
7170 	flp = NULL;
7171 	gotiomode = -1;
7172 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7173 	if (*tl++ != 0)
7174 		*retonclosep = 1;
7175 	else
7176 		*retonclosep = 0;
7177 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7178 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7179 	    (int)stateidp->seqid);
7180 	stateidp->other[0] = *tl++;
7181 	stateidp->other[1] = *tl++;
7182 	stateidp->other[2] = *tl++;
7183 	cnt = fxdr_unsigned(int, *tl);
7184 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7185 	if (cnt <= 0 || cnt > 10000) {
7186 		/* Don't accept more than 10000 layouts in reply. */
7187 		error = NFSERR_BADXDR;
7188 		goto nfsmout;
7189 	}
7190 	for (i = 0; i < cnt; i++) {
7191 		/* Dissect to the layout type. */
7192 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7193 		    3 * NFSX_UNSIGNED);
7194 		off = fxdr_hyper(tl); tl += 2;
7195 		retlen = fxdr_hyper(tl); tl += 2;
7196 		iomode = fxdr_unsigned(int, *tl++);
7197 		laytype = fxdr_unsigned(int, *tl);
7198 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7199 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7200 		/* Ignore length of layout body for now. */
7201 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7202 			/* Parse the File layout up to fhcnt. */
7203 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7204 			    NFSX_HYPER + NFSX_V4DEVICEID);
7205 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7206 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7207 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7208 			if (fhcnt < 0 || fhcnt > 100) {
7209 				/* Don't accept more than 100 file handles. */
7210 				error = NFSERR_BADXDR;
7211 				goto nfsmout;
7212 			}
7213 			if (fhcnt > 0)
7214 				flp = malloc(sizeof(*flp) + fhcnt *
7215 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7216 				    M_WAITOK);
7217 			else
7218 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7219 				    M_WAITOK);
7220 			flp->nfsfl_flags = NFSFL_FILE;
7221 			flp->nfsfl_fhcnt = 0;
7222 			flp->nfsfl_devp = NULL;
7223 			flp->nfsfl_off = off;
7224 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7225 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7226 			else
7227 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7228 			flp->nfsfl_iomode = iomode;
7229 			if (gotiomode == -1)
7230 				gotiomode = flp->nfsfl_iomode;
7231 			/* Ignore layout body length for now. */
7232 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7233 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7234 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7235 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7236 			mtx_lock(&nmp->nm_mtx);
7237 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7238 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7239 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7240 			mtx_unlock(&nmp->nm_mtx);
7241 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7242 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7243 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7244 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7245 			for (j = 0; j < fhcnt; j++) {
7246 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7247 				nfhlen = fxdr_unsigned(int, *tl);
7248 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7249 					error = NFSERR_BADXDR;
7250 					goto nfsmout;
7251 				}
7252 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7253 				    M_NFSFH, M_WAITOK);
7254 				flp->nfsfl_fh[j] = nfhp;
7255 				flp->nfsfl_fhcnt++;
7256 				nfhp->nfh_len = nfhlen;
7257 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7258 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7259 			}
7260 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7261 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7262 			    NFSX_HYPER);
7263 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7264 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7265 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7266 				error = NFSERR_BADXDR;
7267 				goto nfsmout;
7268 			}
7269 			flp = malloc(sizeof(*flp) + mirrorcnt *
7270 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7271 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7272 			flp->nfsfl_mirrorcnt = mirrorcnt;
7273 			for (j = 0; j < mirrorcnt; j++)
7274 				flp->nfsfl_ffm[j].devp = NULL;
7275 			flp->nfsfl_off = off;
7276 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7277 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7278 			else
7279 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7280 			flp->nfsfl_iomode = iomode;
7281 			if (gotiomode == -1)
7282 				gotiomode = flp->nfsfl_iomode;
7283 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7284 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7285 			    (uintmax_t)flp->nfsfl_stripeunit);
7286 			for (j = 0; j < mirrorcnt; j++) {
7287 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7288 				k = fxdr_unsigned(int, *tl);
7289 				if (k < 1 || k > 128) {
7290 					error = NFSERR_BADXDR;
7291 					goto nfsmout;
7292 				}
7293 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7294 				for (l = 0; l < k; l++) {
7295 					NFSM_DISSECT(tl, uint32_t *,
7296 					    NFSX_V4DEVICEID + NFSX_STATEID +
7297 					    2 * NFSX_UNSIGNED);
7298 					if (l == 0) {
7299 						/* Just use the first server. */
7300 						NFSBCOPY(tl,
7301 						    flp->nfsfl_ffm[j].dev,
7302 						    NFSX_V4DEVICEID);
7303 						tl += (NFSX_V4DEVICEID /
7304 						    NFSX_UNSIGNED);
7305 						tl++;
7306 						flp->nfsfl_ffm[j].st.seqid =
7307 						    *tl++;
7308 						flp->nfsfl_ffm[j].st.other[0] =
7309 						    *tl++;
7310 						flp->nfsfl_ffm[j].st.other[1] =
7311 						    *tl++;
7312 						flp->nfsfl_ffm[j].st.other[2] =
7313 						    *tl++;
7314 						NFSCL_DEBUG(4, "st.seqid=%u "
7315 						 "st.o0=0x%x st.o1=0x%x "
7316 						 "st.o2=0x%x\n",
7317 						 flp->nfsfl_ffm[j].st.seqid,
7318 						 flp->nfsfl_ffm[j].st.other[0],
7319 						 flp->nfsfl_ffm[j].st.other[1],
7320 						 flp->nfsfl_ffm[j].st.other[2]);
7321 					} else
7322 						tl += ((NFSX_V4DEVICEID +
7323 						    NFSX_STATEID +
7324 						    NFSX_UNSIGNED) /
7325 						    NFSX_UNSIGNED);
7326 					fhcnt = fxdr_unsigned(int, *tl);
7327 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7328 					if (fhcnt < 1 ||
7329 					    fhcnt > NFSDEV_MAXVERS) {
7330 						error = NFSERR_BADXDR;
7331 						goto nfsmout;
7332 					}
7333 					for (m = 0; m < fhcnt; m++) {
7334 						NFSM_DISSECT(tl, uint32_t *,
7335 						    NFSX_UNSIGNED);
7336 						nfhlen = fxdr_unsigned(int,
7337 						    *tl);
7338 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7339 						    nfhlen);
7340 						if (nfhlen <= 0 || nfhlen >
7341 						    NFSX_V4FHMAX) {
7342 							error = NFSERR_BADXDR;
7343 							goto nfsmout;
7344 						}
7345 						NFSM_DISSECT(cp, uint8_t *,
7346 						    NFSM_RNDUP(nfhlen));
7347 						if (l == 0) {
7348 							flp->nfsfl_ffm[j].fhcnt
7349 							    = fhcnt;
7350 							nfhp = malloc(
7351 							    sizeof(*nfhp) +
7352 							    nfhlen - 1, M_NFSFH,
7353 							    M_WAITOK);
7354 							flp->nfsfl_ffm[j].fh[m]
7355 							    = nfhp;
7356 							nfhp->nfh_len = nfhlen;
7357 							NFSBCOPY(cp,
7358 							    nfhp->nfh_fh,
7359 							    nfhlen);
7360 							NFSCL_DEBUG(4,
7361 							    "got fh\n");
7362 						}
7363 					}
7364 					/* Now, get the ffsd_user/ffds_group. */
7365 					error = nfsrv_parseug(nd, 0, &user,
7366 					    &grp, curthread);
7367 					NFSCL_DEBUG(4, "after parseu=%d\n",
7368 					    error);
7369 					if (error == 0)
7370 						error = nfsrv_parseug(nd, 1,
7371 						    &user, &grp, curthread);
7372 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7373 					    grp);
7374 					if (error != 0)
7375 						goto nfsmout;
7376 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7377 					    user, grp);
7378 					if (l == 0) {
7379 						flp->nfsfl_ffm[j].user = user;
7380 						flp->nfsfl_ffm[j].group = grp;
7381 						NFSCL_DEBUG(4,
7382 						    "usr=%d grp=%d\n", user,
7383 						    grp);
7384 					}
7385 				}
7386 			}
7387 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7388 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7389 #ifdef notnow
7390 			/*
7391 			 * At this time, there is no flag.
7392 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7393 			 * added, or it may never exist?
7394 			 */
7395 			mtx_lock(&nmp->nm_mtx);
7396 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7397 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7398 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7399 			mtx_unlock(&nmp->nm_mtx);
7400 #endif
7401 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7402 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7403 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7404 		} else {
7405 			error = NFSERR_BADXDR;
7406 			goto nfsmout;
7407 		}
7408 		if (flp->nfsfl_iomode == gotiomode) {
7409 			/* Keep the list in increasing offset order. */
7410 			tflp = LIST_FIRST(flhp);
7411 			prevflp = NULL;
7412 			while (tflp != NULL &&
7413 			    tflp->nfsfl_off < flp->nfsfl_off) {
7414 				prevflp = tflp;
7415 				tflp = LIST_NEXT(tflp, nfsfl_list);
7416 			}
7417 			if (prevflp == NULL)
7418 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7419 			else
7420 				LIST_INSERT_AFTER(prevflp, flp,
7421 				    nfsfl_list);
7422 			NFSCL_DEBUG(4, "flp inserted\n");
7423 		} else {
7424 			printf("nfscl_layoutget(): got wrong iomode\n");
7425 			nfscl_freeflayout(flp);
7426 		}
7427 		flp = NULL;
7428 	}
7429 nfsmout:
7430 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7431 	if (error != 0 && flp != NULL)
7432 		nfscl_freeflayout(flp);
7433 	return (error);
7434 }
7435 
7436 /*
7437  * Parse a user/group digit string.
7438  */
7439 static int
7440 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7441     NFSPROC_T *p)
7442 {
7443 	uint32_t *tl;
7444 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7445 	uint32_t len = 0;
7446 	int error = 0;
7447 
7448 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7449 	len = fxdr_unsigned(uint32_t, *tl);
7450 	str = NULL;
7451 	if (len > NFSV4_OPAQUELIMIT) {
7452 		error = NFSERR_BADXDR;
7453 		goto nfsmout;
7454 	}
7455 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7456 	if (len == 0) {
7457 		if (dogrp != 0)
7458 			*gidp = GID_NOGROUP;
7459 		else
7460 			*uidp = UID_NOBODY;
7461 		return (0);
7462 	}
7463 	if (len > NFSV4_SMALLSTR)
7464 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7465 	else
7466 		str = str0;
7467 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7468 	NFSBCOPY(cp, str, len);
7469 	str[len] = '\0';
7470 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7471 	if (dogrp != 0)
7472 		error = nfsv4_strtogid(nd, str, len, gidp);
7473 	else
7474 		error = nfsv4_strtouid(nd, str, len, uidp);
7475 nfsmout:
7476 	if (len > NFSV4_SMALLSTR)
7477 		free(str, M_TEMP);
7478 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7479 	return (error);
7480 }
7481 
7482 /*
7483  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7484  * so that it does both an Open and a Layoutget.
7485  */
7486 static int
7487 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7488     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7489     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7490     struct ucred *cred, NFSPROC_T *p)
7491 {
7492 	struct nfscllayout *lyp;
7493 	struct nfsclflayout *flp;
7494 	struct nfsclflayouthead flh;
7495 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7496 	int layouttype, laystat;
7497 	nfsv4stateid_t stateid;
7498 	struct nfsclsession *tsep;
7499 
7500 	error = 0;
7501 	if (NFSHASFLEXFILE(nmp))
7502 		layouttype = NFSLAYOUT_FLEXFILE;
7503 	else
7504 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7505 	/*
7506 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7507 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7508 	 * flp == NULL.
7509 	 */
7510 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7511 	    &recalled);
7512 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7513 	if (lyp == NULL)
7514 		islocked = 0;
7515 	else if (flp != NULL)
7516 		islocked = 1;
7517 	else
7518 		islocked = 2;
7519 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7520 		LIST_INIT(&flh);
7521 		tsep = nfsmnt_mdssession(nmp);
7522 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7523 		    3 * NFSX_UNSIGNED);
7524 		if (lyp == NULL)
7525 			usecurstateid = 1;
7526 		else {
7527 			usecurstateid = 0;
7528 			stateid.seqid = lyp->nfsly_stateid.seqid;
7529 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7530 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7531 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7532 		}
7533 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7534 		    newfhp, newfhlen, mode, op, name, namelen,
7535 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7536 		    &retonclose, &flh, &laystat, cred, p);
7537 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7538 		    laystat, error);
7539 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7540 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7541 		    &islocked, cred, p);
7542 	} else
7543 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7544 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7545 	if (islocked == 2)
7546 		nfscl_rellayout(lyp, 1);
7547 	else if (islocked == 1)
7548 		nfscl_rellayout(lyp, 0);
7549 	return (error);
7550 }
7551 
7552 /*
7553  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7554  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7555  * handled by nfsrpc_openrpc().
7556  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7557  * can be NULL.
7558  */
7559 static int
7560 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7561     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7562     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7563     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7564     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7565     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7566 {
7567 	uint32_t *tl;
7568 	struct nfsrv_descript nfsd, *nd = &nfsd;
7569 	struct nfscldeleg *ndp = NULL;
7570 	struct nfsvattr nfsva;
7571 	struct nfsclsession *tsep;
7572 	uint32_t rflags, deleg;
7573 	nfsattrbit_t attrbits;
7574 	int error, ret, acesize, limitby, iomode;
7575 
7576 	*dpp = NULL;
7577 	*laystatp = ENXIO;
7578 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7579 	    0, 0);
7580 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7581 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7582 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7583 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7584 	tsep = nfsmnt_mdssession(nmp);
7585 	*tl++ = tsep->nfsess_clientid.lval[0];
7586 	*tl = tsep->nfsess_clientid.lval[1];
7587 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7588 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7589 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7590 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7591 	nfsm_strtom(nd, name, namelen);
7592 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7593 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7594 	NFSZERO_ATTRBIT(&attrbits);
7595 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7596 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7597 	nfsrv_putattrbit(nd, &attrbits);
7598 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7599 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7600 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7601 		iomode = NFSLAYOUTIOMODE_RW;
7602 	else
7603 		iomode = NFSLAYOUTIOMODE_READ;
7604 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7605 	    layouttype, layoutlen, usecurstateid);
7606 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7607 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7608 	if (error != 0)
7609 		return (error);
7610 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7611 	if (nd->nd_repstat != 0)
7612 		*laystatp = nd->nd_repstat;
7613 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7614 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7615 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7616 		    6 * NFSX_UNSIGNED);
7617 		op->nfso_stateid.seqid = *tl++;
7618 		op->nfso_stateid.other[0] = *tl++;
7619 		op->nfso_stateid.other[1] = *tl++;
7620 		op->nfso_stateid.other[2] = *tl;
7621 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7622 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7623 		if (error != 0)
7624 			goto nfsmout;
7625 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7626 		deleg = fxdr_unsigned(u_int32_t, *tl);
7627 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7628 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7629 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7630 			      NFSCLFLAGS_FIRSTDELEG))
7631 				op->nfso_own->nfsow_clp->nfsc_flags |=
7632 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7633 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7634 			    M_NFSCLDELEG, M_WAITOK);
7635 			LIST_INIT(&ndp->nfsdl_owner);
7636 			LIST_INIT(&ndp->nfsdl_lock);
7637 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7638 			ndp->nfsdl_fhlen = newfhlen;
7639 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7640 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
7641 			nfscl_lockinit(&ndp->nfsdl_rwlock);
7642 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7643 			    NFSX_UNSIGNED);
7644 			ndp->nfsdl_stateid.seqid = *tl++;
7645 			ndp->nfsdl_stateid.other[0] = *tl++;
7646 			ndp->nfsdl_stateid.other[1] = *tl++;
7647 			ndp->nfsdl_stateid.other[2] = *tl++;
7648 			ret = fxdr_unsigned(int, *tl);
7649 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7650 				ndp->nfsdl_flags = NFSCLDL_WRITE;
7651 				/*
7652 				 * Indicates how much the file can grow.
7653 				 */
7654 				NFSM_DISSECT(tl, u_int32_t *,
7655 				    3 * NFSX_UNSIGNED);
7656 				limitby = fxdr_unsigned(int, *tl++);
7657 				switch (limitby) {
7658 				case NFSV4OPEN_LIMITSIZE:
7659 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7660 					break;
7661 				case NFSV4OPEN_LIMITBLOCKS:
7662 					ndp->nfsdl_sizelimit =
7663 					    fxdr_unsigned(u_int64_t, *tl++);
7664 					ndp->nfsdl_sizelimit *=
7665 					    fxdr_unsigned(u_int64_t, *tl);
7666 					break;
7667 				default:
7668 					error = NFSERR_BADXDR;
7669 					goto nfsmout;
7670 				};
7671 			} else
7672 				ndp->nfsdl_flags = NFSCLDL_READ;
7673 			if (ret != 0)
7674 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
7675 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7676 			    &acesize, p);
7677 			if (error != 0)
7678 				goto nfsmout;
7679 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7680 			error = NFSERR_BADXDR;
7681 			goto nfsmout;
7682 		}
7683 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7684 		    nfscl_assumeposixlocks)
7685 			op->nfso_posixlock = 1;
7686 		else
7687 			op->nfso_posixlock = 0;
7688 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7689 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
7690 		if (*++tl == 0) {
7691 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7692 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7693 			    NULL, NULL, NULL, p, cred);
7694 			if (error != 0)
7695 				goto nfsmout;
7696 			if (ndp != NULL) {
7697 				ndp->nfsdl_change = nfsva.na_filerev;
7698 				ndp->nfsdl_modtime = nfsva.na_mtime;
7699 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7700 				*dpp = ndp;
7701 				ndp = NULL;
7702 			}
7703 			/*
7704 			 * At this point, the Open has succeeded, so set
7705 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
7706 			 * this function just won't return a layout.
7707 			 */
7708 			if (nd->nd_repstat == 0) {
7709 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7710 				*laystatp = fxdr_unsigned(int, *++tl);
7711 				if (*laystatp == 0) {
7712 					error = nfsrv_parselayoutget(nmp, nd,
7713 					    stateidp, retonclosep, flhp);
7714 					if (error != 0)
7715 						*laystatp = error;
7716 				}
7717 			} else
7718 				nd->nd_repstat = 0;	/* Return 0 for Open. */
7719 		}
7720 	}
7721 	if (nd->nd_repstat != 0 && error == 0)
7722 		error = nd->nd_repstat;
7723 nfsmout:
7724 	free(ndp, M_NFSCLDELEG);
7725 	m_freem(nd->nd_mrep);
7726 	return (error);
7727 }
7728 
7729 /*
7730  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7731  * Used only for mounts with pNFS enabled.
7732  */
7733 static int
7734 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7735     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7736     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7737     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7738     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7739     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7740     struct nfsclflayouthead *flhp, int *laystatp)
7741 {
7742 	uint32_t *tl;
7743 	int error = 0, deleg, newone, ret, acesize, limitby;
7744 	struct nfsrv_descript nfsd, *nd = &nfsd;
7745 	struct nfsclopen *op;
7746 	struct nfscldeleg *dp = NULL;
7747 	struct nfsnode *np;
7748 	struct nfsfh *nfhp;
7749 	struct nfsclsession *tsep;
7750 	nfsattrbit_t attrbits;
7751 	nfsv4stateid_t stateid;
7752 	struct nfsmount *nmp;
7753 
7754 	nmp = VFSTONFS(dvp->v_mount);
7755 	np = VTONFS(dvp);
7756 	*laystatp = ENXIO;
7757 	*unlockedp = 0;
7758 	*nfhpp = NULL;
7759 	*dpp = NULL;
7760 	*attrflagp = 0;
7761 	*dattrflagp = 0;
7762 	if (namelen > NFS_MAXNAMLEN)
7763 		return (ENAMETOOLONG);
7764 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7765 	/*
7766 	 * For V4, this is actually an Open op.
7767 	 */
7768 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7769 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
7770 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7771 	    NFSV4OPEN_ACCESSREAD);
7772 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7773 	tsep = nfsmnt_mdssession(nmp);
7774 	*tl++ = tsep->nfsess_clientid.lval[0];
7775 	*tl = tsep->nfsess_clientid.lval[1];
7776 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7777 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7778 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7779 	if ((fmode & O_EXCL) != 0) {
7780 		if (NFSHASSESSPERSIST(nmp)) {
7781 			/* Use GUARDED for persistent sessions. */
7782 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
7783 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7784 		} else {
7785 			/* Otherwise, use EXCLUSIVE4_1. */
7786 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7787 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7788 			*tl++ = cverf.lval[0];
7789 			*tl = cverf.lval[1];
7790 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7791 		}
7792 	} else {
7793 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7794 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
7795 	}
7796 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7797 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7798 	nfsm_strtom(nd, name, namelen);
7799 	/* Get the new file's handle and attributes, plus save the FH. */
7800 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7801 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7802 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
7803 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7804 	NFSGETATTR_ATTRBIT(&attrbits);
7805 	nfsrv_putattrbit(nd, &attrbits);
7806 	/* Get the directory's post-op attributes. */
7807 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7808 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
7809 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7810 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7811 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7812 	nfsrv_putattrbit(nd, &attrbits);
7813 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7814 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7815 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7816 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7817 	    layouttype, layoutlen, usecurstateid);
7818 	error = nfscl_request(nd, dvp, p, cred, dstuff);
7819 	if (error != 0)
7820 		return (error);
7821 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7822 	    error);
7823 	if (nd->nd_repstat != 0)
7824 		*laystatp = nd->nd_repstat;
7825 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7826 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7827 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7828 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7829 		    6 * NFSX_UNSIGNED);
7830 		stateid.seqid = *tl++;
7831 		stateid.other[0] = *tl++;
7832 		stateid.other[1] = *tl++;
7833 		stateid.other[2] = *tl;
7834 		nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7835 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7836 		deleg = fxdr_unsigned(int, *tl);
7837 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7838 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7839 			if (!(owp->nfsow_clp->nfsc_flags &
7840 			      NFSCLFLAGS_FIRSTDELEG))
7841 				owp->nfsow_clp->nfsc_flags |=
7842 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7843 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7844 			    M_NFSCLDELEG, M_WAITOK);
7845 			LIST_INIT(&dp->nfsdl_owner);
7846 			LIST_INIT(&dp->nfsdl_lock);
7847 			dp->nfsdl_clp = owp->nfsow_clp;
7848 			newnfs_copyincred(cred, &dp->nfsdl_cred);
7849 			nfscl_lockinit(&dp->nfsdl_rwlock);
7850 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7851 			    NFSX_UNSIGNED);
7852 			dp->nfsdl_stateid.seqid = *tl++;
7853 			dp->nfsdl_stateid.other[0] = *tl++;
7854 			dp->nfsdl_stateid.other[1] = *tl++;
7855 			dp->nfsdl_stateid.other[2] = *tl++;
7856 			ret = fxdr_unsigned(int, *tl);
7857 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7858 				dp->nfsdl_flags = NFSCLDL_WRITE;
7859 				/*
7860 				 * Indicates how much the file can grow.
7861 				 */
7862 				NFSM_DISSECT(tl, u_int32_t *,
7863 				    3 * NFSX_UNSIGNED);
7864 				limitby = fxdr_unsigned(int, *tl++);
7865 				switch (limitby) {
7866 				case NFSV4OPEN_LIMITSIZE:
7867 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
7868 					break;
7869 				case NFSV4OPEN_LIMITBLOCKS:
7870 					dp->nfsdl_sizelimit =
7871 					    fxdr_unsigned(u_int64_t, *tl++);
7872 					dp->nfsdl_sizelimit *=
7873 					    fxdr_unsigned(u_int64_t, *tl);
7874 					break;
7875 				default:
7876 					error = NFSERR_BADXDR;
7877 					goto nfsmout;
7878 				};
7879 			} else {
7880 				dp->nfsdl_flags = NFSCLDL_READ;
7881 			}
7882 			if (ret != 0)
7883 				dp->nfsdl_flags |= NFSCLDL_RECALL;
7884 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7885 			    &acesize, p);
7886 			if (error != 0)
7887 				goto nfsmout;
7888 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7889 			error = NFSERR_BADXDR;
7890 			goto nfsmout;
7891 		}
7892 
7893 		/* Now, we should have the status for the SaveFH. */
7894 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7895 		if (*++tl == 0) {
7896 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7897 			/*
7898 			 * Now, process the GetFH and Getattr for the newly
7899 			 * created file. nfscl_mtofh() will set
7900 			 * ND_NOMOREDATA if these weren't successful.
7901 			 */
7902 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7903 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7904 			if (error != 0)
7905 				goto nfsmout;
7906 		} else
7907 			nd->nd_flag |= ND_NOMOREDATA;
7908 		/* Now we have the PutFH and Getattr for the directory. */
7909 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7910 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7911 			if (*++tl != 0)
7912 				nd->nd_flag |= ND_NOMOREDATA;
7913 			else {
7914 				NFSM_DISSECT(tl, uint32_t *, 2 *
7915 				    NFSX_UNSIGNED);
7916 				if (*++tl != 0)
7917 					nd->nd_flag |= ND_NOMOREDATA;
7918 			}
7919 		}
7920 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7921 			/* Load the directory attributes. */
7922 			error = nfsm_loadattr(nd, dnap);
7923 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7924 			if (error != 0)
7925 				goto nfsmout;
7926 			*dattrflagp = 1;
7927 			if (dp != NULL && *attrflagp != 0) {
7928 				dp->nfsdl_change = nnap->na_filerev;
7929 				dp->nfsdl_modtime = nnap->na_mtime;
7930 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7931 			}
7932 			/*
7933 			 * We can now complete the Open state.
7934 			 */
7935 			nfhp = *nfhpp;
7936 			if (dp != NULL) {
7937 				dp->nfsdl_fhlen = nfhp->nfh_len;
7938 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7939 				    nfhp->nfh_len);
7940 			}
7941 			/*
7942 			 * Get an Open structure that will be
7943 			 * attached to the OpenOwner, acquired already.
7944 			 */
7945 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7946 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7947 			    cred, p, NULL, &op, &newone, NULL, 0);
7948 			if (error != 0)
7949 				goto nfsmout;
7950 			op->nfso_stateid = stateid;
7951 			newnfs_copyincred(cred, &op->nfso_cred);
7952 
7953 			nfscl_openrelease(nmp, op, error, newone);
7954 			*unlockedp = 1;
7955 
7956 			/* Now, handle the RestoreFH and LayoutGet. */
7957 			if (nd->nd_repstat == 0) {
7958 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7959 				*laystatp = fxdr_unsigned(int, *(tl + 3));
7960 				if (*laystatp == 0) {
7961 					error = nfsrv_parselayoutget(nmp, nd,
7962 					    stateidp, retonclosep, flhp);
7963 					if (error != 0)
7964 						*laystatp = error;
7965 				}
7966 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7967 				    error);
7968 			} else
7969 				nd->nd_repstat = 0;
7970 		}
7971 	}
7972 	if (nd->nd_repstat != 0 && error == 0)
7973 		error = nd->nd_repstat;
7974 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7975 		nfscl_initiate_recovery(owp->nfsow_clp);
7976 nfsmout:
7977 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7978 	if (error == 0)
7979 		*dpp = dp;
7980 	else
7981 		free(dp, M_NFSCLDELEG);
7982 	m_freem(nd->nd_mrep);
7983 	return (error);
7984 }
7985 
7986 /*
7987  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7988  */
7989 static int
7990 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7991     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7992     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7993     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7994     int *dattrflagp, void *dstuff, int *unlockedp)
7995 {
7996 	struct nfscllayout *lyp;
7997 	struct nfsclflayouthead flh;
7998 	struct nfsfh *nfhp;
7999 	struct nfsclsession *tsep;
8000 	struct nfsmount *nmp;
8001 	nfsv4stateid_t stateid;
8002 	int error, layoutlen, layouttype, retonclose, laystat;
8003 
8004 	error = 0;
8005 	nmp = VFSTONFS(dvp->v_mount);
8006 	if (NFSHASFLEXFILE(nmp))
8007 		layouttype = NFSLAYOUT_FLEXFILE;
8008 	else
8009 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8010 	LIST_INIT(&flh);
8011 	tsep = nfsmnt_mdssession(nmp);
8012 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
8013 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
8014 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
8015 	    dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
8016 	    &flh, &laystat);
8017 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
8018 	    laystat, error);
8019 	lyp = NULL;
8020 	if (laystat == 0) {
8021 		nfhp = *nfhpp;
8022 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8023 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8024 		    layouttype, laystat, NULL, cred, p);
8025 	} else
8026 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8027 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8028 		    cred, p);
8029 	if (laystat == 0)
8030 		nfscl_rellayout(lyp, 0);
8031 	return (error);
8032 }
8033 
8034 /*
8035  * Process the results of a layoutget() operation.
8036  */
8037 static int
8038 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8039     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8040     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8041     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8042 {
8043 	struct nfsclflayout *tflp;
8044 	struct nfscldevinfo *dip;
8045 	uint8_t *dev;
8046 	int i, mirrorcnt;
8047 
8048 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8049 		NFSLOCKMNT(nmp);
8050 		if (!NFSHASFLEXFILE(nmp)) {
8051 			/* Switch to using Flex File Layout. */
8052 			nmp->nm_state |= NFSSTA_FLEXFILE;
8053 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8054 			/* Disable pNFS. */
8055 			NFSCL_DEBUG(1, "disable PNFS\n");
8056 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8057 		}
8058 		NFSUNLOCKMNT(nmp);
8059 	}
8060 	if (laystat == 0) {
8061 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8062 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8063 			if (layouttype == NFSLAYOUT_FLEXFILE)
8064 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8065 			else
8066 				mirrorcnt = 1;
8067 			for (i = 0; i < mirrorcnt; i++) {
8068 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8069 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8070 				if (laystat != 0) {
8071 					if (layouttype == NFSLAYOUT_FLEXFILE)
8072 						dev = tflp->nfsfl_ffm[i].dev;
8073 					else
8074 						dev = tflp->nfsfl_dev;
8075 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8076 					    layouttype, notifybit, &dip, cred,
8077 					    p);
8078 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8079 					    laystat);
8080 					if (laystat != 0)
8081 						goto out;
8082 					laystat = nfscl_adddevinfo(nmp, dip, i,
8083 					    tflp);
8084 					if (laystat != 0)
8085 						printf("nfsrpc_layoutgetresout"
8086 						    ": cannot add\n");
8087 				}
8088 			}
8089 		}
8090 	}
8091 out:
8092 	if (laystat == 0) {
8093 		/*
8094 		 * nfscl_layout() always returns with the nfsly_lock
8095 		 * set to a refcnt (shared lock).
8096 		 * Passing in dvp is sufficient, since it is only used to
8097 		 * get the fsid for the file system.
8098 		 */
8099 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8100 		    layouttype, retonclose, flhp, lypp, cred, p);
8101 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8102 		    laystat);
8103 		if (laystat == 0 && islockedp != NULL)
8104 			*islockedp = 1;
8105 	}
8106 	return (laystat);
8107 }
8108 
8109 /*
8110  * nfs copy_file_range operation.
8111  */
8112 int
8113 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8114     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8115     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8116     struct ucred *cred, bool consecutive, bool *must_commitp)
8117 {
8118 	int commit, error, expireret = 0, retrycnt;
8119 	u_int32_t clidrev = 0;
8120 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8121 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8122 	nfsv4stateid_t instateid, outstateid;
8123 	void *inlckp, *outlckp;
8124 
8125 	if (nmp->nm_clp != NULL)
8126 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8127 	innfhp = VTONFS(invp)->n_fhp;
8128 	outnfhp = VTONFS(outvp)->n_fhp;
8129 	retrycnt = 0;
8130 	do {
8131 		/* Get both stateids. */
8132 		inlckp = NULL;
8133 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8134 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8135 		    &inlckp);
8136 		outlckp = NULL;
8137 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8138 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8139 		    &outlckp);
8140 
8141 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8142 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8143 		    outattrflagp, consecutive, &commit, cred, curthread);
8144 		if (error == 0) {
8145 			if (commit != NFSWRITE_FILESYNC)
8146 				*must_commitp = true;
8147 			*inoffp += *lenp;
8148 			*outoffp += *lenp;
8149 		} else if (error == NFSERR_STALESTATEID)
8150 			nfscl_initiate_recovery(nmp->nm_clp);
8151 		if (inlckp != NULL)
8152 			nfscl_lockderef(inlckp);
8153 		if (outlckp != NULL)
8154 			nfscl_lockderef(outlckp);
8155 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8156 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8157 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8158 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8159 		} else if ((error == NFSERR_EXPIRED ||
8160 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8161 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8162 			    curthread);
8163 		}
8164 		retrycnt++;
8165 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8166 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8167 	      error == NFSERR_STALEDONTRECOVER ||
8168 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8169 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8170 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8171 	if (error != 0 && (retrycnt >= 4 ||
8172 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8173 	      error == NFSERR_STALEDONTRECOVER))
8174 		error = EIO;
8175 	return (error);
8176 }
8177 
8178 /*
8179  * The copy RPC.
8180  */
8181 static int
8182 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8183     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8184     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8185     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8186     NFSPROC_T *p)
8187 {
8188 	uint32_t *tl;
8189 	int error;
8190 	struct nfsrv_descript nfsd;
8191 	struct nfsrv_descript *nd = &nfsd;
8192 	struct nfsmount *nmp;
8193 	nfsattrbit_t attrbits;
8194 	uint64_t len;
8195 
8196 	nmp = VFSTONFS(outvp->v_mount);
8197 	*inattrflagp = *outattrflagp = 0;
8198 	*commitp = NFSWRITE_UNSTABLE;
8199 	len = *lenp;
8200 	*lenp = 0;
8201 	if (len > nfs_maxcopyrange)
8202 		len = nfs_maxcopyrange;
8203 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8204 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8205 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8206 	NFSGETATTR_ATTRBIT(&attrbits);
8207 	nfsrv_putattrbit(nd, &attrbits);
8208 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8209 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8210 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8211 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8212 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8213 	*tl = txdr_unsigned(NFSV4OP_COPY);
8214 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8215 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8216 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8217 	txdr_hyper(inoff, tl); tl += 2;
8218 	txdr_hyper(outoff, tl); tl += 2;
8219 	txdr_hyper(len, tl); tl += 2;
8220 	if (consecutive)
8221 		*tl++ = newnfs_true;
8222 	else
8223 		*tl++ = newnfs_false;
8224 	*tl++ = newnfs_true;
8225 	*tl++ = 0;
8226 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8227 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8228 	nfsrv_putattrbit(nd, &attrbits);
8229 	error = nfscl_request(nd, invp, p, cred, NULL);
8230 	if (error != 0)
8231 		return (error);
8232 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8233 		/* Get the input file's attributes. */
8234 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8235 		if (*(tl + 1) == 0) {
8236 			error = nfsm_loadattr(nd, innap);
8237 			if (error != 0)
8238 				goto nfsmout;
8239 			*inattrflagp = 1;
8240 		} else
8241 			nd->nd_flag |= ND_NOMOREDATA;
8242 	}
8243 	/* Skip over return stat for PutFH. */
8244 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8245 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8246 		if (*++tl != 0)
8247 			nd->nd_flag |= ND_NOMOREDATA;
8248 	}
8249 	/* Skip over return stat for Copy. */
8250 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8251 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8252 	if (nd->nd_repstat == 0) {
8253 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8254 		if (*tl != 0) {
8255 			/* There should be no callback ids. */
8256 			error = NFSERR_BADXDR;
8257 			goto nfsmout;
8258 		}
8259 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8260 		    NFSX_VERF);
8261 		len = fxdr_hyper(tl); tl += 2;
8262 		*commitp = fxdr_unsigned(int, *tl++);
8263 		NFSLOCKMNT(nmp);
8264 		if (!NFSHASWRITEVERF(nmp)) {
8265 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8266 			NFSSETWRITEVERF(nmp);
8267 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8268 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8269 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8270 		}
8271 		NFSUNLOCKMNT(nmp);
8272 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8273 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8274 			/* Must be a synchronous copy. */
8275 			nd->nd_repstat = NFSERR_NOTSUPP;
8276 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8277 		error = nfsm_loadattr(nd, outnap);
8278 		if (error == 0)
8279 			*outattrflagp = NFS_LATTR_NOSHRINK;
8280 		if (nd->nd_repstat == 0)
8281 			*lenp = len;
8282 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8283 		/*
8284 		 * For the case where consecutive is not supported, but
8285 		 * synchronous is supported, we can try consecutive == false
8286 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8287 		 * since Copy cannot be done.
8288 		 */
8289 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8290 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8291 			if (!consecutive || *++tl == newnfs_false)
8292 				nd->nd_repstat = NFSERR_NOTSUPP;
8293 		} else
8294 			nd->nd_repstat = NFSERR_BADXDR;
8295 	}
8296 	if (error == 0)
8297 		error = nd->nd_repstat;
8298 nfsmout:
8299 	m_freem(nd->nd_mrep);
8300 	return (error);
8301 }
8302 
8303 /*
8304  * Seek operation.
8305  */
8306 int
8307 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8308     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8309 {
8310 	int error, expireret = 0, retrycnt;
8311 	u_int32_t clidrev = 0;
8312 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
8313 	struct nfsnode *np = VTONFS(vp);
8314 	struct nfsfh *nfhp = NULL;
8315 	nfsv4stateid_t stateid;
8316 	void *lckp;
8317 
8318 	if (nmp->nm_clp != NULL)
8319 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8320 	nfhp = np->n_fhp;
8321 	retrycnt = 0;
8322 	do {
8323 		lckp = NULL;
8324 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8325 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8326 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8327 		    nap, attrflagp, cred);
8328 		if (error == NFSERR_STALESTATEID)
8329 			nfscl_initiate_recovery(nmp->nm_clp);
8330 		if (lckp != NULL)
8331 			nfscl_lockderef(lckp);
8332 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8333 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8334 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8335 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8336 		} else if ((error == NFSERR_EXPIRED ||
8337 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8338 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8339 			    curthread);
8340 		}
8341 		retrycnt++;
8342 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8343 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8344 	    error == NFSERR_BADSESSION ||
8345 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8346 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8347 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8348 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8349 	if (error && retrycnt >= 4)
8350 		error = EIO;
8351 	return (error);
8352 }
8353 
8354 /*
8355  * The seek RPC.
8356  */
8357 static int
8358 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8359     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8360 {
8361 	uint32_t *tl;
8362 	int error;
8363 	struct nfsrv_descript nfsd;
8364 	struct nfsrv_descript *nd = &nfsd;
8365 	nfsattrbit_t attrbits;
8366 
8367 	*attrflagp = 0;
8368 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8369 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8370 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8371 	txdr_hyper(*offp, tl); tl += 2;
8372 	*tl++ = txdr_unsigned(content);
8373 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8374 	NFSGETATTR_ATTRBIT(&attrbits);
8375 	nfsrv_putattrbit(nd, &attrbits);
8376 	error = nfscl_request(nd, vp, curthread, cred, NULL);
8377 	if (error != 0)
8378 		return (error);
8379 	if (nd->nd_repstat == 0) {
8380 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8381 		if (*tl++ == newnfs_true)
8382 			*eofp = true;
8383 		else
8384 			*eofp = false;
8385 		*offp = fxdr_hyper(tl);
8386 		/* Just skip over Getattr op status. */
8387 		error = nfsm_loadattr(nd, nap);
8388 		if (error == 0)
8389 			*attrflagp = 1;
8390 	}
8391 	error = nd->nd_repstat;
8392 nfsmout:
8393 	m_freem(nd->nd_mrep);
8394 	return (error);
8395 }
8396 
8397 /*
8398  * The getextattr RPC.
8399  */
8400 int
8401 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8402     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8403 {
8404 	uint32_t *tl;
8405 	int error;
8406 	struct nfsrv_descript nfsd;
8407 	struct nfsrv_descript *nd = &nfsd;
8408 	nfsattrbit_t attrbits;
8409 	uint32_t len, len2;
8410 
8411 	*attrflagp = 0;
8412 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8413 	nfsm_strtom(nd, name, strlen(name));
8414 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8415 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8416 	NFSGETATTR_ATTRBIT(&attrbits);
8417 	nfsrv_putattrbit(nd, &attrbits);
8418 	error = nfscl_request(nd, vp, p, cred, NULL);
8419 	if (error != 0)
8420 		return (error);
8421 	if (nd->nd_repstat == 0) {
8422 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8423 		len = fxdr_unsigned(uint32_t, *tl);
8424 		/* Sanity check lengths. */
8425 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8426 		    uiop->uio_resid <= UINT32_MAX) {
8427 			len2 = uiop->uio_resid;
8428 			if (len2 >= len)
8429 				error = nfsm_mbufuio(nd, uiop, len);
8430 			else {
8431 				error = nfsm_mbufuio(nd, uiop, len2);
8432 				if (error == 0) {
8433 					/*
8434 					 * nfsm_mbufuio() advances to a multiple
8435 					 * of 4, so round up len2 as well.  Then
8436 					 * we need to advance over the rest of
8437 					 * the data, rounding up the remaining
8438 					 * length.
8439 					 */
8440 					len2 = NFSM_RNDUP(len2);
8441 					len2 = NFSM_RNDUP(len - len2);
8442 					if (len2 > 0)
8443 						error = nfsm_advance(nd, len2,
8444 						    -1);
8445 				}
8446 			}
8447 		} else if (uiop == NULL && len > 0) {
8448 			/* Just wants the length and not the data. */
8449 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8450 		} else if (len > 0)
8451 			error = ENOATTR;
8452 		if (error != 0)
8453 			goto nfsmout;
8454 		*lenp = len;
8455 		/* Just skip over Getattr op status. */
8456 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8457 		error = nfsm_loadattr(nd, nap);
8458 		if (error == 0)
8459 			*attrflagp = 1;
8460 	}
8461 	if (error == 0)
8462 		error = nd->nd_repstat;
8463 nfsmout:
8464 	m_freem(nd->nd_mrep);
8465 	return (error);
8466 }
8467 
8468 /*
8469  * The setextattr RPC.
8470  */
8471 int
8472 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8473     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8474 {
8475 	uint32_t *tl;
8476 	int error;
8477 	struct nfsrv_descript nfsd;
8478 	struct nfsrv_descript *nd = &nfsd;
8479 	nfsattrbit_t attrbits;
8480 
8481 	*attrflagp = 0;
8482 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8483 	if (uiop->uio_resid > nd->nd_maxreq) {
8484 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8485 		m_freem(nd->nd_mreq);
8486 		return (EINVAL);
8487 	}
8488 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8489 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8490 	nfsm_strtom(nd, name, strlen(name));
8491 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8492 	*tl = txdr_unsigned(uiop->uio_resid);
8493 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8494 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8495 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8496 	NFSGETATTR_ATTRBIT(&attrbits);
8497 	nfsrv_putattrbit(nd, &attrbits);
8498 	error = nfscl_request(nd, vp, p, cred, NULL);
8499 	if (error != 0)
8500 		return (error);
8501 	if (nd->nd_repstat == 0) {
8502 		/* Just skip over the reply and Getattr op status. */
8503 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8504 		    NFSX_UNSIGNED);
8505 		error = nfsm_loadattr(nd, nap);
8506 		if (error == 0)
8507 			*attrflagp = 1;
8508 	}
8509 	if (error == 0)
8510 		error = nd->nd_repstat;
8511 nfsmout:
8512 	m_freem(nd->nd_mrep);
8513 	return (error);
8514 }
8515 
8516 /*
8517  * The removeextattr RPC.
8518  */
8519 int
8520 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8521     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8522 {
8523 	uint32_t *tl;
8524 	int error;
8525 	struct nfsrv_descript nfsd;
8526 	struct nfsrv_descript *nd = &nfsd;
8527 	nfsattrbit_t attrbits;
8528 
8529 	*attrflagp = 0;
8530 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8531 	nfsm_strtom(nd, name, strlen(name));
8532 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8533 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8534 	NFSGETATTR_ATTRBIT(&attrbits);
8535 	nfsrv_putattrbit(nd, &attrbits);
8536 	error = nfscl_request(nd, vp, p, cred, NULL);
8537 	if (error != 0)
8538 		return (error);
8539 	if (nd->nd_repstat == 0) {
8540 		/* Just skip over the reply and Getattr op status. */
8541 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8542 		    NFSX_UNSIGNED);
8543 		error = nfsm_loadattr(nd, nap);
8544 		if (error == 0)
8545 			*attrflagp = 1;
8546 	}
8547 	if (error == 0)
8548 		error = nd->nd_repstat;
8549 nfsmout:
8550 	m_freem(nd->nd_mrep);
8551 	return (error);
8552 }
8553 
8554 /*
8555  * The listextattr RPC.
8556  */
8557 int
8558 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8559     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8560     struct ucred *cred, NFSPROC_T *p)
8561 {
8562 	uint32_t *tl;
8563 	int cnt, error, i, len;
8564 	struct nfsrv_descript nfsd;
8565 	struct nfsrv_descript *nd = &nfsd;
8566 	nfsattrbit_t attrbits;
8567 	u_char c;
8568 
8569 	*attrflagp = 0;
8570 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8571 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8572 	txdr_hyper(*cookiep, tl); tl += 2;
8573 	*tl++ = txdr_unsigned(*lenp);
8574 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8575 	NFSGETATTR_ATTRBIT(&attrbits);
8576 	nfsrv_putattrbit(nd, &attrbits);
8577 	error = nfscl_request(nd, vp, p, cred, NULL);
8578 	if (error != 0)
8579 		return (error);
8580 	*eofp = true;
8581 	*lenp = 0;
8582 	if (nd->nd_repstat == 0) {
8583 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8584 		*cookiep = fxdr_hyper(tl); tl += 2;
8585 		cnt = fxdr_unsigned(int, *tl);
8586 		if (cnt < 0) {
8587 			error = EBADRPC;
8588 			goto nfsmout;
8589 		}
8590 		for (i = 0; i < cnt; i++) {
8591 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8592 			len = fxdr_unsigned(int, *tl);
8593 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8594 				error = EBADRPC;
8595 				goto nfsmout;
8596 			}
8597 			if (uiop == NULL)
8598 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8599 			else if (uiop->uio_resid >= len + 1) {
8600 				c = len;
8601 				error = uiomove(&c, sizeof(c), uiop);
8602 				if (error == 0)
8603 					error = nfsm_mbufuio(nd, uiop, len);
8604 			} else {
8605 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8606 				*eofp = false;
8607 			}
8608 			if (error != 0)
8609 				goto nfsmout;
8610 			*lenp += (len + 1);
8611 		}
8612 		/* Get the eof and skip over the Getattr op status. */
8613 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8614 		/*
8615 		 * *eofp is set false above, because it wasn't able to copy
8616 		 * all of the reply.
8617 		 */
8618 		if (*eofp && *tl == 0)
8619 			*eofp = false;
8620 		error = nfsm_loadattr(nd, nap);
8621 		if (error == 0)
8622 			*attrflagp = 1;
8623 	}
8624 	if (error == 0)
8625 		error = nd->nd_repstat;
8626 nfsmout:
8627 	m_freem(nd->nd_mrep);
8628 	return (error);
8629 }
8630 
8631 /*
8632  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
8633  */
8634 static struct mbuf *
8635 nfsm_split(struct mbuf *mp, uint64_t xfer)
8636 {
8637 	struct mbuf *m, *m2;
8638 	vm_page_t pg;
8639 	int i, j, left, pgno, plen, trim;
8640 	char *cp, *cp2;
8641 
8642 	if ((mp->m_flags & M_EXTPG) == 0) {
8643 		m = m_split(mp, xfer, M_WAITOK);
8644 		return (m);
8645 	}
8646 
8647 	/* Find the correct mbuf to split at. */
8648 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
8649 		xfer -= m->m_len;
8650 	if (m == NULL)
8651 		return (NULL);
8652 
8653 	/* If xfer == m->m_len, we can just split the mbuf list. */
8654 	if (xfer == m->m_len) {
8655 		m2 = m->m_next;
8656 		m->m_next = NULL;
8657 		return (m2);
8658 	}
8659 
8660 	/* Find the page to split at. */
8661 	pgno = 0;
8662 	left = xfer;
8663 	do {
8664 		if (pgno == 0)
8665 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
8666 		else
8667 			plen = m_epg_pagelen(m, pgno, 0);
8668 		if (left <= plen)
8669 			break;
8670 		left -= plen;
8671 		pgno++;
8672 	} while (pgno < m->m_epg_npgs);
8673 	if (pgno == m->m_epg_npgs)
8674 		panic("nfsm_split: eroneous ext_pgs mbuf");
8675 
8676 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
8677 	m2->m_epg_flags |= EPG_FLAG_ANON;
8678 
8679 	/*
8680 	 * If left < plen, allocate a new page for the new mbuf
8681 	 * and copy the data after left in the page to this new
8682 	 * page.
8683 	 */
8684 	if (left < plen) {
8685 		do {
8686 			pg = vm_page_alloc(NULL, 0, VM_ALLOC_NORMAL |
8687 			    VM_ALLOC_NOOBJ | VM_ALLOC_NODUMP |
8688 			    VM_ALLOC_WIRED);
8689 			if (pg == NULL)
8690 				vm_wait(NULL);
8691 		} while (pg == NULL);
8692 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
8693 		m2->m_epg_npgs = 1;
8694 
8695 		/* Copy the data after left to the new page. */
8696 		trim = plen - left;
8697 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
8698 		if (pgno == 0)
8699 			cp += m->m_epg_1st_off;
8700 		cp += left;
8701 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
8702 		if (pgno == m->m_epg_npgs - 1)
8703 			m2->m_epg_last_len = trim;
8704 		else {
8705 			cp2 += PAGE_SIZE - trim;
8706 			m2->m_epg_1st_off = PAGE_SIZE - trim;
8707 			m2->m_epg_last_len = m->m_epg_last_len;
8708 		}
8709 		memcpy(cp2, cp, trim);
8710 		m2->m_len = trim;
8711 	} else {
8712 		m2->m_len = 0;
8713 		m2->m_epg_last_len = m->m_epg_last_len;
8714 	}
8715 
8716 	/* Move the pages beyond pgno to the new mbuf. */
8717 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
8718 		m2->m_epg_pa[j] = m->m_epg_pa[i];
8719 		/* Never moves page 0. */
8720 		m2->m_len += m_epg_pagelen(m, i, 0);
8721 	}
8722 	m2->m_epg_npgs = j;
8723 	m->m_epg_npgs = pgno + 1;
8724 	m->m_epg_last_len = left;
8725 	m->m_len = xfer;
8726 
8727 	m2->m_next = m->m_next;
8728 	m->m_next = NULL;
8729 	return (m2);
8730 }
8731 
8732 /*
8733  * Do the NFSv4.1 Bind Connection to Session.
8734  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
8735  */
8736 void
8737 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
8738 {
8739 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
8740 	uint32_t res, *tl;
8741 	struct nfsrv_descript nfsd;
8742 	struct nfsrv_descript *nd = &nfsd;
8743 	struct rpc_callextra ext;
8744 	struct timeval utimeout;
8745 	enum clnt_stat stat;
8746 	int error;
8747 
8748 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
8749 	    NFS_VER4, rcp->minorvers);
8750 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
8751 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
8752 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
8753 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
8754 	*tl = newnfs_false;
8755 
8756 	memset(&ext, 0, sizeof(ext));
8757 	utimeout.tv_sec = 30;
8758 	utimeout.tv_usec = 0;
8759 	ext.rc_auth = authunix_create(cr);
8760 	nd->nd_mrep = NULL;
8761 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
8762 	    &nd->nd_mrep, utimeout);
8763 	AUTH_DESTROY(ext.rc_auth);
8764 	if (stat != RPC_SUCCESS) {
8765 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
8766 		return;
8767 	}
8768 	if (nd->nd_mrep == NULL) {
8769 		printf("nfsrpc_bindconnsess: no reply args\n");
8770 		return;
8771 	}
8772 	error = 0;
8773 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
8774 	nd->nd_md = nd->nd_mrep;
8775 	nd->nd_dpos = mtod(nd->nd_md, char *);
8776 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8777 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
8778 	if (nd->nd_repstat == NFSERR_OK) {
8779 		res = fxdr_unsigned(uint32_t, *tl);
8780 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
8781 		    -1)) != 0)
8782 			goto nfsmout;
8783 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
8784 		    4 * NFSX_UNSIGNED);
8785 		tl += 3;
8786 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
8787 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
8788 			res = fxdr_unsigned(uint32_t, *tl);
8789 			if (res != NFSCDFS4_BOTH)
8790 				printf("nfsrpc_bindconnsess: did not "
8791 				    "return FS4_BOTH\n");
8792 		} else
8793 			printf("nfsrpc_bindconnsess: not same "
8794 			    "sessionid\n");
8795 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
8796 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
8797 nfsmout:
8798 	if (error != 0)
8799 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
8800 	m_freem(nd->nd_mrep);
8801 }
8802