xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision e86cf8ada1ac24f3f3119d63d284dc75cf4b8828)
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 #ifndef APPLEKEXT
48 #include "opt_inet6.h"
49 
50 #include <fs/nfs/nfsport.h>
51 #include <fs/nfsclient/nfs.h>
52 #include <sys/extattr.h>
53 #include <sys/sysctl.h>
54 #include <sys/taskqueue.h>
55 
56 SYSCTL_DECL(_vfs_nfs);
57 
58 static int	nfsignore_eexist = 0;
59 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
60     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
61 
62 static int	nfscl_dssameconn = 0;
63 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
64     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
65 
66 /*
67  * Global variables
68  */
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 #endif	/* !APPLEKEXT */
86 
87 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
88 
89 /*
90  * nfscl_getsameserver() can return one of three values:
91  * NFSDSP_USETHISSESSION - Use this session for the DS.
92  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
93  *     session.
94  * NFSDSP_NOTFOUND - No matching server was found.
95  */
96 enum nfsclds_state {
97 	NFSDSP_USETHISSESSION = 0,
98 	NFSDSP_SEQTHISSESSION = 1,
99 	NFSDSP_NOTFOUND = 2,
100 };
101 
102 /*
103  * Do a write RPC on a DS data file, using this structure for the arguments,
104  * so that this function can be executed by a separate kernel process.
105  */
106 struct nfsclwritedsdorpc {
107 	int			done;
108 	int			inprog;
109 	struct task		tsk;
110 	struct vnode		*vp;
111 	int			iomode;
112 	int			must_commit;
113 	nfsv4stateid_t		*stateidp;
114 	struct nfsclds		*dsp;
115 	uint64_t		off;
116 	int			len;
117 #ifdef notyet
118 	int			advise;
119 #endif
120 	struct nfsfh		*fhp;
121 	struct mbuf		*m;
122 	int			vers;
123 	int			minorvers;
124 	struct ucred		*cred;
125 	NFSPROC_T		*p;
126 	int			err;
127 };
128 
129 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
130     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
131 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
132     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
133 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
134     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
135     void *);
136 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
137     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
138     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
139 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
140     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
141     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
142     int *, void *, int *);
143 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
144     struct nfscllockowner *, u_int64_t, u_int64_t,
145     u_int32_t, struct ucred *, NFSPROC_T *, int);
146 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
147     struct acl *, nfsv4stateid_t *, void *);
148 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
149     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
150     struct ucred *, NFSPROC_T *);
151 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
152     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
153     NFSPROC_T *);
154 static void nfscl_initsessionslots(struct nfsclsession *);
155 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
156     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
157     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
158     NFSPROC_T *);
159 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
160     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
161     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
162     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
163 static struct mbuf *nfsm_copym(struct mbuf *, int, int);
164 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
165     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
166     struct ucred *, NFSPROC_T *);
167 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
168     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
169     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
170 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
171     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
172     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
173 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
174     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
175     struct ucred *, NFSPROC_T *);
176 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
177     struct nfsclds *, struct nfsclds **, uint32_t *);
178 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
179     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
180     NFSPROC_T *);
181 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
182     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
183 #ifdef notyet
184 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
185     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
186     NFSPROC_T *);
187 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
188     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
189 #endif
190 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
191     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
192 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
193     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
194 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
195     NFSPROC_T *);
196 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
197     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
198 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
199     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
200     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
201 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
202     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
203     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
204     struct nfsfh **, int *, int *, void *, int *);
205 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
206     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
207     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
208     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
209 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
210     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
211     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
212     struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
213     int, int, int, int *, struct nfsclflayouthead *, int *);
214 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
215     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
216     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
217 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
218     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
219     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
220 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
221     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
222     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
223 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
224     int, struct nfsvattr *, int *, struct ucred *);
225 
226 int nfs_pnfsio(task_fn_t *, void *);
227 
228 /*
229  * nfs null call from vfs.
230  */
231 APPLESTATIC int
232 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
233 {
234 	int error;
235 	struct nfsrv_descript nfsd, *nd = &nfsd;
236 
237 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
238 	error = nfscl_request(nd, vp, p, cred, NULL);
239 	if (nd->nd_repstat && !error)
240 		error = nd->nd_repstat;
241 	m_freem(nd->nd_mrep);
242 	return (error);
243 }
244 
245 /*
246  * nfs access rpc op.
247  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
248  * modes are changed on the server, accesses might still fail later.
249  */
250 APPLESTATIC int
251 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
252     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
253 {
254 	int error;
255 	u_int32_t mode, rmode;
256 
257 	if (acmode & VREAD)
258 		mode = NFSACCESS_READ;
259 	else
260 		mode = 0;
261 	if (vnode_vtype(vp) == VDIR) {
262 		if (acmode & VWRITE)
263 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
264 				 NFSACCESS_DELETE);
265 		if (acmode & VEXEC)
266 			mode |= NFSACCESS_LOOKUP;
267 	} else {
268 		if (acmode & VWRITE)
269 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
270 		if (acmode & VEXEC)
271 			mode |= NFSACCESS_EXECUTE;
272 	}
273 
274 	/*
275 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
276 	 */
277 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
278 	    NULL);
279 
280 	/*
281 	 * The NFS V3 spec does not clarify whether or not
282 	 * the returned access bits can be a superset of
283 	 * the ones requested, so...
284 	 */
285 	if (!error && (rmode & mode) != mode)
286 		error = EACCES;
287 	return (error);
288 }
289 
290 /*
291  * The actual rpc, separated out for Darwin.
292  */
293 APPLESTATIC int
294 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
295     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
296     void *stuff)
297 {
298 	u_int32_t *tl;
299 	u_int32_t supported, rmode;
300 	int error;
301 	struct nfsrv_descript nfsd, *nd = &nfsd;
302 	nfsattrbit_t attrbits;
303 
304 	*attrflagp = 0;
305 	supported = mode;
306 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
307 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
308 	*tl = txdr_unsigned(mode);
309 	if (nd->nd_flag & ND_NFSV4) {
310 		/*
311 		 * And do a Getattr op.
312 		 */
313 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
314 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
315 		NFSGETATTR_ATTRBIT(&attrbits);
316 		(void) nfsrv_putattrbit(nd, &attrbits);
317 	}
318 	error = nfscl_request(nd, vp, p, cred, stuff);
319 	if (error)
320 		return (error);
321 	if (nd->nd_flag & ND_NFSV3) {
322 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
323 		if (error)
324 			goto nfsmout;
325 	}
326 	if (!nd->nd_repstat) {
327 		if (nd->nd_flag & ND_NFSV4) {
328 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
329 			supported = fxdr_unsigned(u_int32_t, *tl++);
330 		} else {
331 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
332 		}
333 		rmode = fxdr_unsigned(u_int32_t, *tl);
334 		if (nd->nd_flag & ND_NFSV4)
335 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
336 
337 		/*
338 		 * It's not obvious what should be done about
339 		 * unsupported access modes. For now, be paranoid
340 		 * and clear the unsupported ones.
341 		 */
342 		rmode &= supported;
343 		*rmodep = rmode;
344 	} else
345 		error = nd->nd_repstat;
346 nfsmout:
347 	m_freem(nd->nd_mrep);
348 	return (error);
349 }
350 
351 /*
352  * nfs open rpc
353  */
354 APPLESTATIC int
355 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
356 {
357 	struct nfsclopen *op;
358 	struct nfscldeleg *dp;
359 	struct nfsfh *nfhp;
360 	struct nfsnode *np = VTONFS(vp);
361 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
362 	u_int32_t mode, clidrev;
363 	int ret, newone, error, expireret = 0, retrycnt;
364 
365 	/*
366 	 * For NFSv4, Open Ops are only done on Regular Files.
367 	 */
368 	if (vnode_vtype(vp) != VREG)
369 		return (0);
370 	mode = 0;
371 	if (amode & FREAD)
372 		mode |= NFSV4OPEN_ACCESSREAD;
373 	if (amode & FWRITE)
374 		mode |= NFSV4OPEN_ACCESSWRITE;
375 	nfhp = np->n_fhp;
376 
377 	retrycnt = 0;
378 #ifdef notdef
379 { char name[100]; int namel;
380 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
381 bcopy(NFS4NODENAME(np->n_v4), name, namel);
382 name[namel] = '\0';
383 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
384 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
385 else printf(" fhl=0\n");
386 }
387 #endif
388 	do {
389 	    dp = NULL;
390 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
391 		cred, p, NULL, &op, &newone, &ret, 1);
392 	    if (error) {
393 		return (error);
394 	    }
395 	    if (nmp->nm_clp != NULL)
396 		clidrev = nmp->nm_clp->nfsc_clientidrev;
397 	    else
398 		clidrev = 0;
399 	    if (ret == NFSCLOPEN_DOOPEN) {
400 		if (np->n_v4 != NULL) {
401 			/*
402 			 * For the first attempt, try and get a layout, if
403 			 * pNFS is enabled for the mount.
404 			 */
405 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
406 			    nfs_numnfscbd == 0 ||
407 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
408 				error = nfsrpc_openrpc(nmp, vp,
409 				    np->n_v4->n4_data,
410 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
411 				    np->n_fhp->nfh_len, mode, op,
412 				    NFS4NODENAME(np->n_v4),
413 				    np->n_v4->n4_namelen,
414 				    &dp, 0, 0x0, cred, p, 0, 0);
415 			else
416 				error = nfsrpc_getopenlayout(nmp, vp,
417 				    np->n_v4->n4_data,
418 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
419 				    np->n_fhp->nfh_len, mode, op,
420 				    NFS4NODENAME(np->n_v4),
421 				    np->n_v4->n4_namelen, &dp, cred, p);
422 			if (dp != NULL) {
423 #ifdef APPLE
424 				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
425 #else
426 				NFSLOCKNODE(np);
427 				np->n_flag &= ~NDELEGMOD;
428 				/*
429 				 * Invalidate the attribute cache, so that
430 				 * attributes that pre-date the issue of a
431 				 * delegation are not cached, since the
432 				 * cached attributes will remain valid while
433 				 * the delegation is held.
434 				 */
435 				NFSINVALATTRCACHE(np);
436 				NFSUNLOCKNODE(np);
437 #endif
438 				(void) nfscl_deleg(nmp->nm_mountp,
439 				    op->nfso_own->nfsow_clp,
440 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
441 			}
442 		} else {
443 			error = EIO;
444 		}
445 		newnfs_copyincred(cred, &op->nfso_cred);
446 	    } else if (ret == NFSCLOPEN_SETCRED)
447 		/*
448 		 * This is a new local open on a delegation. It needs
449 		 * to have credentials so that an open can be done
450 		 * against the server during recovery.
451 		 */
452 		newnfs_copyincred(cred, &op->nfso_cred);
453 
454 	    /*
455 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
456 	     * been done on this Open successfully and a VOP_CLOSE()
457 	     * is expected for each of these.
458 	     * If error is non-zero, don't increment it, since the Open
459 	     * hasn't succeeded yet.
460 	     */
461 	    if (!error)
462 		op->nfso_opencnt++;
463 	    nfscl_openrelease(nmp, op, error, newone);
464 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
465 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
466 		error == NFSERR_BADSESSION) {
467 		(void) nfs_catnap(PZERO, error, "nfs_open");
468 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
469 		&& clidrev != 0) {
470 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
471 		retrycnt++;
472 	    }
473 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
474 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
475 	    error == NFSERR_BADSESSION ||
476 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
477 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
478 	if (error && retrycnt >= 4)
479 		error = EIO;
480 	return (error);
481 }
482 
483 /*
484  * the actual open rpc
485  */
486 APPLESTATIC int
487 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
488     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
489     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
490     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
491     int syscred, int recursed)
492 {
493 	u_int32_t *tl;
494 	struct nfsrv_descript nfsd, *nd = &nfsd;
495 	struct nfscldeleg *dp, *ndp = NULL;
496 	struct nfsvattr nfsva;
497 	u_int32_t rflags, deleg;
498 	nfsattrbit_t attrbits;
499 	int error, ret, acesize, limitby;
500 	struct nfsclsession *tsep;
501 
502 	dp = *dpp;
503 	*dpp = NULL;
504 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
505 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
506 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
507 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
508 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
509 	tsep = nfsmnt_mdssession(nmp);
510 	*tl++ = tsep->nfsess_clientid.lval[0];
511 	*tl = tsep->nfsess_clientid.lval[1];
512 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
513 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
514 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
515 	if (reclaim) {
516 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
517 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
518 		*tl = txdr_unsigned(delegtype);
519 	} else {
520 		if (dp != NULL) {
521 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
522 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
523 			if (NFSHASNFSV4N(nmp))
524 				*tl++ = 0;
525 			else
526 				*tl++ = dp->nfsdl_stateid.seqid;
527 			*tl++ = dp->nfsdl_stateid.other[0];
528 			*tl++ = dp->nfsdl_stateid.other[1];
529 			*tl = dp->nfsdl_stateid.other[2];
530 		} else {
531 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
532 		}
533 		(void) nfsm_strtom(nd, name, namelen);
534 	}
535 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
536 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
537 	NFSZERO_ATTRBIT(&attrbits);
538 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
539 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
540 	(void) nfsrv_putattrbit(nd, &attrbits);
541 	if (syscred)
542 		nd->nd_flag |= ND_USEGSSNAME;
543 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
544 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
545 	if (error)
546 		return (error);
547 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
548 	if (!nd->nd_repstat) {
549 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
550 		    6 * NFSX_UNSIGNED);
551 		op->nfso_stateid.seqid = *tl++;
552 		op->nfso_stateid.other[0] = *tl++;
553 		op->nfso_stateid.other[1] = *tl++;
554 		op->nfso_stateid.other[2] = *tl;
555 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
556 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
557 		if (error)
558 			goto nfsmout;
559 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
560 		deleg = fxdr_unsigned(u_int32_t, *tl);
561 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
562 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
563 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
564 			      NFSCLFLAGS_FIRSTDELEG))
565 				op->nfso_own->nfsow_clp->nfsc_flags |=
566 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
567 			ndp = malloc(
568 			    sizeof (struct nfscldeleg) + newfhlen,
569 			    M_NFSCLDELEG, M_WAITOK);
570 			LIST_INIT(&ndp->nfsdl_owner);
571 			LIST_INIT(&ndp->nfsdl_lock);
572 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
573 			ndp->nfsdl_fhlen = newfhlen;
574 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
575 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
576 			nfscl_lockinit(&ndp->nfsdl_rwlock);
577 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
578 			    NFSX_UNSIGNED);
579 			ndp->nfsdl_stateid.seqid = *tl++;
580 			ndp->nfsdl_stateid.other[0] = *tl++;
581 			ndp->nfsdl_stateid.other[1] = *tl++;
582 			ndp->nfsdl_stateid.other[2] = *tl++;
583 			ret = fxdr_unsigned(int, *tl);
584 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
585 				ndp->nfsdl_flags = NFSCLDL_WRITE;
586 				/*
587 				 * Indicates how much the file can grow.
588 				 */
589 				NFSM_DISSECT(tl, u_int32_t *,
590 				    3 * NFSX_UNSIGNED);
591 				limitby = fxdr_unsigned(int, *tl++);
592 				switch (limitby) {
593 				case NFSV4OPEN_LIMITSIZE:
594 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
595 					break;
596 				case NFSV4OPEN_LIMITBLOCKS:
597 					ndp->nfsdl_sizelimit =
598 					    fxdr_unsigned(u_int64_t, *tl++);
599 					ndp->nfsdl_sizelimit *=
600 					    fxdr_unsigned(u_int64_t, *tl);
601 					break;
602 				default:
603 					error = NFSERR_BADXDR;
604 					goto nfsmout;
605 				}
606 			} else {
607 				ndp->nfsdl_flags = NFSCLDL_READ;
608 			}
609 			if (ret)
610 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
611 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
612 			    &acesize, p);
613 			if (error)
614 				goto nfsmout;
615 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
616 			error = NFSERR_BADXDR;
617 			goto nfsmout;
618 		}
619 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
620 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
621 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
622 		    NULL, NULL, NULL, p, cred);
623 		if (error)
624 			goto nfsmout;
625 		if (ndp != NULL) {
626 			ndp->nfsdl_change = nfsva.na_filerev;
627 			ndp->nfsdl_modtime = nfsva.na_mtime;
628 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
629 		}
630 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
631 		    do {
632 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
633 			    cred, p);
634 			if (ret == NFSERR_DELAY)
635 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
636 		    } while (ret == NFSERR_DELAY);
637 		    error = ret;
638 		}
639 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
640 		    nfscl_assumeposixlocks)
641 		    op->nfso_posixlock = 1;
642 		else
643 		    op->nfso_posixlock = 0;
644 
645 		/*
646 		 * If the server is handing out delegations, but we didn't
647 		 * get one because an OpenConfirm was required, try the
648 		 * Open again, to get a delegation. This is a harmless no-op,
649 		 * from a server's point of view.
650 		 */
651 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
652 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
653 		    && !error && dp == NULL && ndp == NULL && !recursed) {
654 		    do {
655 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
656 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
657 			    cred, p, syscred, 1);
658 			if (ret == NFSERR_DELAY)
659 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
660 		    } while (ret == NFSERR_DELAY);
661 		    if (ret) {
662 			if (ndp != NULL) {
663 				free(ndp, M_NFSCLDELEG);
664 				ndp = NULL;
665 			}
666 			if (ret == NFSERR_STALECLIENTID ||
667 			    ret == NFSERR_STALEDONTRECOVER ||
668 			    ret == NFSERR_BADSESSION)
669 				error = ret;
670 		    }
671 		}
672 	}
673 	if (nd->nd_repstat != 0 && error == 0)
674 		error = nd->nd_repstat;
675 	if (error == NFSERR_STALECLIENTID)
676 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
677 nfsmout:
678 	if (!error)
679 		*dpp = ndp;
680 	else if (ndp != NULL)
681 		free(ndp, M_NFSCLDELEG);
682 	m_freem(nd->nd_mrep);
683 	return (error);
684 }
685 
686 /*
687  * open downgrade rpc
688  */
689 APPLESTATIC int
690 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
691     struct ucred *cred, NFSPROC_T *p)
692 {
693 	u_int32_t *tl;
694 	struct nfsrv_descript nfsd, *nd = &nfsd;
695 	int error;
696 
697 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
698 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
699 	if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
700 		*tl++ = 0;
701 	else
702 		*tl++ = op->nfso_stateid.seqid;
703 	*tl++ = op->nfso_stateid.other[0];
704 	*tl++ = op->nfso_stateid.other[1];
705 	*tl++ = op->nfso_stateid.other[2];
706 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
707 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
708 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
709 	error = nfscl_request(nd, vp, p, cred, NULL);
710 	if (error)
711 		return (error);
712 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
713 	if (!nd->nd_repstat) {
714 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
715 		op->nfso_stateid.seqid = *tl++;
716 		op->nfso_stateid.other[0] = *tl++;
717 		op->nfso_stateid.other[1] = *tl++;
718 		op->nfso_stateid.other[2] = *tl;
719 	}
720 	if (nd->nd_repstat && error == 0)
721 		error = nd->nd_repstat;
722 	if (error == NFSERR_STALESTATEID)
723 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
724 nfsmout:
725 	m_freem(nd->nd_mrep);
726 	return (error);
727 }
728 
729 /*
730  * V4 Close operation.
731  */
732 APPLESTATIC int
733 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
734 {
735 	struct nfsclclient *clp;
736 	int error;
737 
738 	if (vnode_vtype(vp) != VREG)
739 		return (0);
740 	if (doclose)
741 		error = nfscl_doclose(vp, &clp, p);
742 	else
743 		error = nfscl_getclose(vp, &clp);
744 	if (error)
745 		return (error);
746 
747 	nfscl_clientrelease(clp);
748 	return (0);
749 }
750 
751 /*
752  * Close the open.
753  */
754 APPLESTATIC void
755 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
756 {
757 	struct nfsrv_descript nfsd, *nd = &nfsd;
758 	struct nfscllockowner *lp, *nlp;
759 	struct nfscllock *lop, *nlop;
760 	struct ucred *tcred;
761 	u_int64_t off = 0, len = 0;
762 	u_int32_t type = NFSV4LOCKT_READ;
763 	int error, do_unlock, trycnt;
764 
765 	tcred = newnfs_getcred();
766 	newnfs_copycred(&op->nfso_cred, tcred);
767 	/*
768 	 * (Theoretically this could be done in the same
769 	 *  compound as the close, but having multiple
770 	 *  sequenced Ops in the same compound might be
771 	 *  too scary for some servers.)
772 	 */
773 	if (op->nfso_posixlock) {
774 		off = 0;
775 		len = NFS64BITSSET;
776 		type = NFSV4LOCKT_READ;
777 	}
778 
779 	/*
780 	 * Since this function is only called from VOP_INACTIVE(), no
781 	 * other thread will be manipulating this Open. As such, the
782 	 * lock lists are not being changed by other threads, so it should
783 	 * be safe to do this without locking.
784 	 */
785 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
786 		do_unlock = 1;
787 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
788 			if (op->nfso_posixlock == 0) {
789 				off = lop->nfslo_first;
790 				len = lop->nfslo_end - lop->nfslo_first;
791 				if (lop->nfslo_type == F_WRLCK)
792 					type = NFSV4LOCKT_WRITE;
793 				else
794 					type = NFSV4LOCKT_READ;
795 			}
796 			if (do_unlock) {
797 				trycnt = 0;
798 				do {
799 					error = nfsrpc_locku(nd, nmp, lp, off,
800 					    len, type, tcred, p, 0);
801 					if ((nd->nd_repstat == NFSERR_GRACE ||
802 					    nd->nd_repstat == NFSERR_DELAY) &&
803 					    error == 0)
804 						(void) nfs_catnap(PZERO,
805 						    (int)nd->nd_repstat,
806 						    "nfs_close");
807 				} while ((nd->nd_repstat == NFSERR_GRACE ||
808 				    nd->nd_repstat == NFSERR_DELAY) &&
809 				    error == 0 && trycnt++ < 5);
810 				if (op->nfso_posixlock)
811 					do_unlock = 0;
812 			}
813 			nfscl_freelock(lop, 0);
814 		}
815 		/*
816 		 * Do a ReleaseLockOwner.
817 		 * The lock owner name nfsl_owner may be used by other opens for
818 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
819 		 * puts on the wire has the file handle for this file appended
820 		 * to it, so it can be done now.
821 		 */
822 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
823 		    lp->nfsl_open->nfso_fhlen, tcred, p);
824 	}
825 
826 	/*
827 	 * There could be other Opens for different files on the same
828 	 * OpenOwner, so locking is required.
829 	 */
830 	NFSLOCKCLSTATE();
831 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
832 	NFSUNLOCKCLSTATE();
833 	do {
834 		error = nfscl_tryclose(op, tcred, nmp, p);
835 		if (error == NFSERR_GRACE)
836 			(void) nfs_catnap(PZERO, error, "nfs_close");
837 	} while (error == NFSERR_GRACE);
838 	NFSLOCKCLSTATE();
839 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
840 
841 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
842 		nfscl_freelockowner(lp, 0);
843 	nfscl_freeopen(op, 0);
844 	NFSUNLOCKCLSTATE();
845 	NFSFREECRED(tcred);
846 }
847 
848 /*
849  * The actual Close RPC.
850  */
851 APPLESTATIC int
852 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
853     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
854     int syscred)
855 {
856 	u_int32_t *tl;
857 	int error;
858 
859 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
860 	    op->nfso_fhlen, NULL, NULL, 0, 0);
861 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
862 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
863 	if (NFSHASNFSV4N(nmp))
864 		*tl++ = 0;
865 	else
866 		*tl++ = op->nfso_stateid.seqid;
867 	*tl++ = op->nfso_stateid.other[0];
868 	*tl++ = op->nfso_stateid.other[1];
869 	*tl = op->nfso_stateid.other[2];
870 	if (syscred)
871 		nd->nd_flag |= ND_USEGSSNAME;
872 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
873 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
874 	if (error)
875 		return (error);
876 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
877 	if (nd->nd_repstat == 0)
878 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
879 	error = nd->nd_repstat;
880 	if (error == NFSERR_STALESTATEID)
881 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
882 nfsmout:
883 	m_freem(nd->nd_mrep);
884 	return (error);
885 }
886 
887 /*
888  * V4 Open Confirm RPC.
889  */
890 APPLESTATIC int
891 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
892     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
893 {
894 	u_int32_t *tl;
895 	struct nfsrv_descript nfsd, *nd = &nfsd;
896 	struct nfsmount *nmp;
897 	int error;
898 
899 	nmp = VFSTONFS(vnode_mount(vp));
900 	if (NFSHASNFSV4N(nmp))
901 		return (0);		/* No confirmation for NFSv4.1. */
902 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
903 	    0, 0);
904 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
905 	*tl++ = op->nfso_stateid.seqid;
906 	*tl++ = op->nfso_stateid.other[0];
907 	*tl++ = op->nfso_stateid.other[1];
908 	*tl++ = op->nfso_stateid.other[2];
909 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
910 	error = nfscl_request(nd, vp, p, cred, NULL);
911 	if (error)
912 		return (error);
913 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
914 	if (!nd->nd_repstat) {
915 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
916 		op->nfso_stateid.seqid = *tl++;
917 		op->nfso_stateid.other[0] = *tl++;
918 		op->nfso_stateid.other[1] = *tl++;
919 		op->nfso_stateid.other[2] = *tl;
920 	}
921 	error = nd->nd_repstat;
922 	if (error == NFSERR_STALESTATEID)
923 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
924 nfsmout:
925 	m_freem(nd->nd_mrep);
926 	return (error);
927 }
928 
929 /*
930  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
931  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
932  */
933 APPLESTATIC int
934 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
935     bool *retokp, struct ucred *cred, NFSPROC_T *p)
936 {
937 	u_int32_t *tl;
938 	struct nfsrv_descript nfsd;
939 	struct nfsrv_descript *nd = &nfsd;
940 	nfsattrbit_t attrbits;
941 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
942 	u_short port;
943 	int error, isinet6 = 0, callblen;
944 	nfsquad_t confirm;
945 	u_int32_t lease;
946 	static u_int32_t rev = 0;
947 	struct nfsclds *dsp, *odsp;
948 	struct in6_addr a6;
949 	struct nfsclsession *tsep;
950 
951 	if (nfsboottime.tv_sec == 0)
952 		NFSSETBOOTTIME(nfsboottime);
953 	if (NFSHASNFSV4N(nmp)) {
954 		error = NFSERR_BADSESSION;
955 		odsp = dsp = NULL;
956 		if (retokp != NULL) {
957 			NFSLOCKMNT(nmp);
958 			odsp = TAILQ_FIRST(&nmp->nm_sess);
959 			NFSUNLOCKMNT(nmp);
960 		}
961 		if (odsp != NULL) {
962 			/*
963 			 * When a session already exists, first try a
964 			 * CreateSession with the extant ClientID.
965 			 */
966 			dsp = malloc(sizeof(struct nfsclds) +
967 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
968 			    M_WAITOK | M_ZERO);
969 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
970 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
971 			dsp->nfsclds_sess.nfsess_clientid =
972 			    odsp->nfsclds_sess.nfsess_clientid;
973 			dsp->nfsclds_sess.nfsess_sequenceid =
974 			    odsp->nfsclds_sess.nfsess_sequenceid;
975 			dsp->nfsclds_flags = odsp->nfsclds_flags;
976 			if (dsp->nfsclds_servownlen > 0)
977 				memcpy(dsp->nfsclds_serverown,
978 				    odsp->nfsclds_serverown,
979 				    dsp->nfsclds_servownlen + 1);
980 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
981 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
982 			    NULL, MTX_DEF);
983 			nfscl_initsessionslots(&dsp->nfsclds_sess);
984 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
985 			    &nmp->nm_sockreq, NULL,
986 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
987 			NFSCL_DEBUG(1, "create session for extant "
988 			    "ClientID=%d\n", error);
989 			if (error != 0) {
990 				nfscl_freenfsclds(dsp);
991 				dsp = NULL;
992 				/*
993 				 * If *retokp is true, return any error other
994 				 * than NFSERR_STALECLIENTID,
995 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
996 				 * so that nfscl_recover() will not loop.
997 				 */
998 				if (*retokp)
999 					return (NFSERR_IO);
1000 			} else
1001 				*retokp = true;
1002 		} else if (retokp != NULL && *retokp)
1003 			return (NFSERR_IO);
1004 		if (error != 0) {
1005 			/*
1006 			 * Either there was no previous session or the
1007 			 * CreateSession attempt failed, so...
1008 			 * do an ExchangeID followed by the CreateSession.
1009 			 */
1010 			clp->nfsc_rev = rev++;
1011 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1012 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1013 			    cred, p);
1014 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1015 			if (error == 0)
1016 				error = nfsrpc_createsession(nmp,
1017 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1018 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1019 				    cred, p);
1020 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1021 		}
1022 		if (error == 0) {
1023 			NFSLOCKMNT(nmp);
1024 			/*
1025 			 * The old sessions cannot be safely free'd
1026 			 * here, since they may still be used by
1027 			 * in-progress RPCs.
1028 			 */
1029 			tsep = NULL;
1030 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
1031 				tsep = NFSMNT_MDSSESSION(nmp);
1032 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1033 			    nfsclds_list);
1034 			/*
1035 			 * Wake up RPCs waiting for a slot on the
1036 			 * old session. These will then fail with
1037 			 * NFSERR_BADSESSION and be retried with the
1038 			 * new session by nfsv4_setsequence().
1039 			 * Also wakeup() processes waiting for the
1040 			 * new session.
1041 			 */
1042 			if (tsep != NULL)
1043 				wakeup(&tsep->nfsess_slots);
1044 			wakeup(&nmp->nm_sess);
1045 			NFSUNLOCKMNT(nmp);
1046 		} else if (dsp != NULL)
1047 			nfscl_freenfsclds(dsp);
1048 		if (error == 0 && reclaim == 0) {
1049 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1050 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1051 			if (error == NFSERR_COMPLETEALREADY ||
1052 			    error == NFSERR_NOTSUPP)
1053 				/* Ignore this error. */
1054 				error = 0;
1055 		}
1056 		return (error);
1057 	} else if (retokp != NULL && *retokp)
1058 		return (NFSERR_IO);
1059 	clp->nfsc_rev = rev++;
1060 
1061 	/*
1062 	 * Allocate a single session structure for NFSv4.0, because some of
1063 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1064 	 */
1065 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1066 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1067 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1068 	NFSLOCKMNT(nmp);
1069 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1070 	tsep = NFSMNT_MDSSESSION(nmp);
1071 	NFSUNLOCKMNT(nmp);
1072 
1073 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
1074 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1075 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1076 	*tl = txdr_unsigned(clp->nfsc_rev);
1077 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1078 
1079 	/*
1080 	 * set up the callback address
1081 	 */
1082 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1083 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1084 	callblen = strlen(nfsv4_callbackaddr);
1085 	if (callblen == 0)
1086 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1087 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1088 	    (callblen > 0 || cp != NULL)) {
1089 		port = htons(nfsv4_cbport);
1090 		cp2 = (u_int8_t *)&port;
1091 #ifdef INET6
1092 		if ((callblen > 0 &&
1093 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1094 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1095 
1096 			(void) nfsm_strtom(nd, "tcp6", 4);
1097 			if (callblen == 0) {
1098 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1099 				ip6add = ip6buf;
1100 			} else {
1101 				ip6add = nfsv4_callbackaddr;
1102 			}
1103 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1104 			    ip6add, cp2[0], cp2[1]);
1105 		} else
1106 #endif
1107 		{
1108 			(void) nfsm_strtom(nd, "tcp", 3);
1109 			if (callblen == 0)
1110 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1111 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1112 				    cp[2], cp[3], cp2[0], cp2[1]);
1113 			else
1114 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1115 				    "%s.%d.%d", nfsv4_callbackaddr,
1116 				    cp2[0], cp2[1]);
1117 		}
1118 		(void) nfsm_strtom(nd, addr, strlen(addr));
1119 	} else {
1120 		(void) nfsm_strtom(nd, "tcp", 3);
1121 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1122 	}
1123 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1124 	*tl = txdr_unsigned(clp->nfsc_cbident);
1125 	nd->nd_flag |= ND_USEGSSNAME;
1126 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1127 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1128 	if (error)
1129 		return (error);
1130 	if (nd->nd_repstat == 0) {
1131 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1132 	    tsep->nfsess_clientid.lval[0] = *tl++;
1133 	    tsep->nfsess_clientid.lval[1] = *tl++;
1134 	    confirm.lval[0] = *tl++;
1135 	    confirm.lval[1] = *tl;
1136 	    m_freem(nd->nd_mrep);
1137 	    nd->nd_mrep = NULL;
1138 
1139 	    /*
1140 	     * and confirm it.
1141 	     */
1142 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1143 		NULL, 0, 0);
1144 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1145 	    *tl++ = tsep->nfsess_clientid.lval[0];
1146 	    *tl++ = tsep->nfsess_clientid.lval[1];
1147 	    *tl++ = confirm.lval[0];
1148 	    *tl = confirm.lval[1];
1149 	    nd->nd_flag |= ND_USEGSSNAME;
1150 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1151 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1152 	    if (error)
1153 		return (error);
1154 	    m_freem(nd->nd_mrep);
1155 	    nd->nd_mrep = NULL;
1156 	    if (nd->nd_repstat == 0) {
1157 		nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
1158 		    nmp->nm_fhsize, NULL, NULL, 0, 0);
1159 		NFSZERO_ATTRBIT(&attrbits);
1160 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1161 		(void) nfsrv_putattrbit(nd, &attrbits);
1162 		nd->nd_flag |= ND_USEGSSNAME;
1163 		error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1164 		    cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1165 		if (error)
1166 		    return (error);
1167 		if (nd->nd_repstat == 0) {
1168 		    error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
1169 			NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1170 		    if (error)
1171 			goto nfsmout;
1172 		    clp->nfsc_renew = NFSCL_RENEW(lease);
1173 		    clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1174 		    clp->nfsc_clientidrev++;
1175 		    if (clp->nfsc_clientidrev == 0)
1176 			clp->nfsc_clientidrev++;
1177 		}
1178 	    }
1179 	}
1180 	error = nd->nd_repstat;
1181 nfsmout:
1182 	m_freem(nd->nd_mrep);
1183 	return (error);
1184 }
1185 
1186 /*
1187  * nfs getattr call.
1188  */
1189 APPLESTATIC int
1190 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1191     struct nfsvattr *nap, void *stuff)
1192 {
1193 	struct nfsrv_descript nfsd, *nd = &nfsd;
1194 	int error;
1195 	nfsattrbit_t attrbits;
1196 
1197 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1198 	if (nd->nd_flag & ND_NFSV4) {
1199 		NFSGETATTR_ATTRBIT(&attrbits);
1200 		(void) nfsrv_putattrbit(nd, &attrbits);
1201 	}
1202 	error = nfscl_request(nd, vp, p, cred, stuff);
1203 	if (error)
1204 		return (error);
1205 	if (!nd->nd_repstat)
1206 		error = nfsm_loadattr(nd, nap);
1207 	else
1208 		error = nd->nd_repstat;
1209 	m_freem(nd->nd_mrep);
1210 	return (error);
1211 }
1212 
1213 /*
1214  * nfs getattr call with non-vnode arguemnts.
1215  */
1216 APPLESTATIC int
1217 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1218     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1219     uint32_t *leasep)
1220 {
1221 	struct nfsrv_descript nfsd, *nd = &nfsd;
1222 	int error, vers = NFS_VER2;
1223 	nfsattrbit_t attrbits;
1224 
1225 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1226 	if (nd->nd_flag & ND_NFSV4) {
1227 		vers = NFS_VER4;
1228 		NFSGETATTR_ATTRBIT(&attrbits);
1229 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1230 		(void) nfsrv_putattrbit(nd, &attrbits);
1231 	} else if (nd->nd_flag & ND_NFSV3) {
1232 		vers = NFS_VER3;
1233 	}
1234 	if (syscred)
1235 		nd->nd_flag |= ND_USEGSSNAME;
1236 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1237 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1238 	if (error)
1239 		return (error);
1240 	if (nd->nd_repstat == 0) {
1241 		if ((nd->nd_flag & ND_NFSV4) != 0)
1242 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1243 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1244 			    NULL, NULL);
1245 		else
1246 			error = nfsm_loadattr(nd, nap);
1247 	} else
1248 		error = nd->nd_repstat;
1249 	m_freem(nd->nd_mrep);
1250 	return (error);
1251 }
1252 
1253 /*
1254  * Do an nfs setattr operation.
1255  */
1256 APPLESTATIC int
1257 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1258     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1259     void *stuff)
1260 {
1261 	int error, expireret = 0, openerr, retrycnt;
1262 	u_int32_t clidrev = 0, mode;
1263 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1264 	struct nfsfh *nfhp;
1265 	nfsv4stateid_t stateid;
1266 	void *lckp;
1267 
1268 	if (nmp->nm_clp != NULL)
1269 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1270 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1271 		mode = NFSV4OPEN_ACCESSWRITE;
1272 	else
1273 		mode = NFSV4OPEN_ACCESSREAD;
1274 	retrycnt = 0;
1275 	do {
1276 		lckp = NULL;
1277 		openerr = 1;
1278 		if (NFSHASNFSV4(nmp)) {
1279 			nfhp = VTONFS(vp)->n_fhp;
1280 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1281 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1282 			if (error && vnode_vtype(vp) == VREG &&
1283 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1284 			     nfstest_openallsetattr)) {
1285 				/*
1286 				 * No Open stateid, so try and open the file
1287 				 * now.
1288 				 */
1289 				if (mode == NFSV4OPEN_ACCESSWRITE)
1290 					openerr = nfsrpc_open(vp, FWRITE, cred,
1291 					    p);
1292 				else
1293 					openerr = nfsrpc_open(vp, FREAD, cred,
1294 					    p);
1295 				if (!openerr)
1296 					(void) nfscl_getstateid(vp,
1297 					    nfhp->nfh_fh, nfhp->nfh_len,
1298 					    mode, 0, cred, p, &stateid, &lckp);
1299 			}
1300 		}
1301 		if (vap != NULL)
1302 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1303 			    rnap, attrflagp, stuff);
1304 		else
1305 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1306 			    stuff);
1307 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1308 			NFSLOCKMNT(nmp);
1309 			nmp->nm_state |= NFSSTA_OPENMODE;
1310 			NFSUNLOCKMNT(nmp);
1311 		}
1312 		if (error == NFSERR_STALESTATEID)
1313 			nfscl_initiate_recovery(nmp->nm_clp);
1314 		if (lckp != NULL)
1315 			nfscl_lockderef(lckp);
1316 		if (!openerr)
1317 			(void) nfsrpc_close(vp, 0, p);
1318 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1319 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1320 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1321 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1322 		} else if ((error == NFSERR_EXPIRED ||
1323 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1324 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1325 		}
1326 		retrycnt++;
1327 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1328 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1329 	    error == NFSERR_BADSESSION ||
1330 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1331 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1332 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1333 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1334 	     retrycnt < 4));
1335 	if (error && retrycnt >= 4)
1336 		error = EIO;
1337 	return (error);
1338 }
1339 
1340 static int
1341 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1342     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1343     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1344 {
1345 	u_int32_t *tl;
1346 	struct nfsrv_descript nfsd, *nd = &nfsd;
1347 	int error;
1348 	nfsattrbit_t attrbits;
1349 
1350 	*attrflagp = 0;
1351 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1352 	if (nd->nd_flag & ND_NFSV4)
1353 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1354 	vap->va_type = vnode_vtype(vp);
1355 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1356 	if (nd->nd_flag & ND_NFSV3) {
1357 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1358 		*tl = newnfs_false;
1359 	} else if (nd->nd_flag & ND_NFSV4) {
1360 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1361 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1362 		NFSGETATTR_ATTRBIT(&attrbits);
1363 		(void) nfsrv_putattrbit(nd, &attrbits);
1364 	}
1365 	error = nfscl_request(nd, vp, p, cred, stuff);
1366 	if (error)
1367 		return (error);
1368 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1369 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1370 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1371 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1372 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1373 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1374 	m_freem(nd->nd_mrep);
1375 	if (nd->nd_repstat && !error)
1376 		error = nd->nd_repstat;
1377 	return (error);
1378 }
1379 
1380 /*
1381  * nfs lookup rpc
1382  */
1383 APPLESTATIC int
1384 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1385     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1386     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1387 {
1388 	u_int32_t *tl;
1389 	struct nfsrv_descript nfsd, *nd = &nfsd;
1390 	struct nfsmount *nmp;
1391 	struct nfsnode *np;
1392 	struct nfsfh *nfhp;
1393 	nfsattrbit_t attrbits;
1394 	int error = 0, lookupp = 0;
1395 
1396 	*attrflagp = 0;
1397 	*dattrflagp = 0;
1398 	if (vnode_vtype(dvp) != VDIR)
1399 		return (ENOTDIR);
1400 	nmp = VFSTONFS(vnode_mount(dvp));
1401 	if (len > NFS_MAXNAMLEN)
1402 		return (ENAMETOOLONG);
1403 	if (NFSHASNFSV4(nmp) && len == 1 &&
1404 		name[0] == '.') {
1405 		/*
1406 		 * Just return the current dir's fh.
1407 		 */
1408 		np = VTONFS(dvp);
1409 		nfhp = malloc(sizeof (struct nfsfh) +
1410 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1411 		nfhp->nfh_len = np->n_fhp->nfh_len;
1412 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1413 		*nfhpp = nfhp;
1414 		return (0);
1415 	}
1416 	if (NFSHASNFSV4(nmp) && len == 2 &&
1417 		name[0] == '.' && name[1] == '.') {
1418 		lookupp = 1;
1419 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1420 	} else {
1421 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1422 		(void) nfsm_strtom(nd, name, len);
1423 	}
1424 	if (nd->nd_flag & ND_NFSV4) {
1425 		NFSGETATTR_ATTRBIT(&attrbits);
1426 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1427 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1428 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1429 		(void) nfsrv_putattrbit(nd, &attrbits);
1430 	}
1431 	error = nfscl_request(nd, dvp, p, cred, stuff);
1432 	if (error)
1433 		return (error);
1434 	if (nd->nd_repstat) {
1435 		/*
1436 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1437 		 * the lookup is at the root of an fs, so return this dir.
1438 		 */
1439 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1440 		    np = VTONFS(dvp);
1441 		    nfhp = malloc(sizeof (struct nfsfh) +
1442 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1443 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1444 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1445 		    *nfhpp = nfhp;
1446 		    m_freem(nd->nd_mrep);
1447 		    return (0);
1448 		}
1449 		if (nd->nd_flag & ND_NFSV3)
1450 		    error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1451 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1452 		    ND_NFSV4) {
1453 			/* Load the directory attributes. */
1454 			error = nfsm_loadattr(nd, dnap);
1455 			if (error == 0)
1456 				*dattrflagp = 1;
1457 		}
1458 		goto nfsmout;
1459 	}
1460 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1461 		/* Load the directory attributes. */
1462 		error = nfsm_loadattr(nd, dnap);
1463 		if (error != 0)
1464 			goto nfsmout;
1465 		*dattrflagp = 1;
1466 		/* Skip over the Lookup and GetFH operation status values. */
1467 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1468 	}
1469 	error = nfsm_getfh(nd, nfhpp);
1470 	if (error)
1471 		goto nfsmout;
1472 
1473 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1474 	if ((nd->nd_flag & ND_NFSV3) && !error)
1475 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1476 nfsmout:
1477 	m_freem(nd->nd_mrep);
1478 	if (!error && nd->nd_repstat)
1479 		error = nd->nd_repstat;
1480 	return (error);
1481 }
1482 
1483 /*
1484  * Do a readlink rpc.
1485  */
1486 APPLESTATIC int
1487 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1488     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1489 {
1490 	u_int32_t *tl;
1491 	struct nfsrv_descript nfsd, *nd = &nfsd;
1492 	struct nfsnode *np = VTONFS(vp);
1493 	nfsattrbit_t attrbits;
1494 	int error, len, cangetattr = 1;
1495 
1496 	*attrflagp = 0;
1497 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1498 	if (nd->nd_flag & ND_NFSV4) {
1499 		/*
1500 		 * And do a Getattr op.
1501 		 */
1502 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1503 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1504 		NFSGETATTR_ATTRBIT(&attrbits);
1505 		(void) nfsrv_putattrbit(nd, &attrbits);
1506 	}
1507 	error = nfscl_request(nd, vp, p, cred, stuff);
1508 	if (error)
1509 		return (error);
1510 	if (nd->nd_flag & ND_NFSV3)
1511 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1512 	if (!nd->nd_repstat && !error) {
1513 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1514 		/*
1515 		 * This seems weird to me, but must have been added to
1516 		 * FreeBSD for some reason. The only thing I can think of
1517 		 * is that there was/is some server that replies with
1518 		 * more link data than it should?
1519 		 */
1520 		if (len == NFS_MAXPATHLEN) {
1521 			NFSLOCKNODE(np);
1522 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1523 				len = np->n_size;
1524 				cangetattr = 0;
1525 			}
1526 			NFSUNLOCKNODE(np);
1527 		}
1528 		error = nfsm_mbufuio(nd, uiop, len);
1529 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1530 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1531 	}
1532 	if (nd->nd_repstat && !error)
1533 		error = nd->nd_repstat;
1534 nfsmout:
1535 	m_freem(nd->nd_mrep);
1536 	return (error);
1537 }
1538 
1539 /*
1540  * Read operation.
1541  */
1542 APPLESTATIC int
1543 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1544     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1545 {
1546 	int error, expireret = 0, retrycnt;
1547 	u_int32_t clidrev = 0;
1548 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1549 	struct nfsnode *np = VTONFS(vp);
1550 	struct ucred *newcred;
1551 	struct nfsfh *nfhp = NULL;
1552 	nfsv4stateid_t stateid;
1553 	void *lckp;
1554 
1555 	if (nmp->nm_clp != NULL)
1556 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1557 	newcred = cred;
1558 	if (NFSHASNFSV4(nmp)) {
1559 		nfhp = np->n_fhp;
1560 		newcred = NFSNEWCRED(cred);
1561 	}
1562 	retrycnt = 0;
1563 	do {
1564 		lckp = NULL;
1565 		if (NFSHASNFSV4(nmp))
1566 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1567 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1568 			    &lckp);
1569 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1570 		    attrflagp, stuff);
1571 		if (error == NFSERR_OPENMODE) {
1572 			NFSLOCKMNT(nmp);
1573 			nmp->nm_state |= NFSSTA_OPENMODE;
1574 			NFSUNLOCKMNT(nmp);
1575 		}
1576 		if (error == NFSERR_STALESTATEID)
1577 			nfscl_initiate_recovery(nmp->nm_clp);
1578 		if (lckp != NULL)
1579 			nfscl_lockderef(lckp);
1580 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1581 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1582 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1583 			(void) nfs_catnap(PZERO, error, "nfs_read");
1584 		} else if ((error == NFSERR_EXPIRED ||
1585 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1586 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1587 		}
1588 		retrycnt++;
1589 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1590 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1591 	    error == NFSERR_BADSESSION ||
1592 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1593 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1594 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1595 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1596 	if (error && retrycnt >= 4)
1597 		error = EIO;
1598 	if (NFSHASNFSV4(nmp))
1599 		NFSFREECRED(newcred);
1600 	return (error);
1601 }
1602 
1603 /*
1604  * The actual read RPC.
1605  */
1606 static int
1607 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1608     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1609     int *attrflagp, void *stuff)
1610 {
1611 	u_int32_t *tl;
1612 	int error = 0, len, retlen, tsiz, eof = 0;
1613 	struct nfsrv_descript nfsd;
1614 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1615 	struct nfsrv_descript *nd = &nfsd;
1616 	int rsize;
1617 	off_t tmp_off;
1618 
1619 	*attrflagp = 0;
1620 	tsiz = uiop->uio_resid;
1621 	tmp_off = uiop->uio_offset + tsiz;
1622 	NFSLOCKMNT(nmp);
1623 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1624 		NFSUNLOCKMNT(nmp);
1625 		return (EFBIG);
1626 	}
1627 	rsize = nmp->nm_rsize;
1628 	NFSUNLOCKMNT(nmp);
1629 	nd->nd_mrep = NULL;
1630 	while (tsiz > 0) {
1631 		*attrflagp = 0;
1632 		len = (tsiz > rsize) ? rsize : tsiz;
1633 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1634 		if (nd->nd_flag & ND_NFSV4)
1635 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1636 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1637 		if (nd->nd_flag & ND_NFSV2) {
1638 			*tl++ = txdr_unsigned(uiop->uio_offset);
1639 			*tl++ = txdr_unsigned(len);
1640 			*tl = 0;
1641 		} else {
1642 			txdr_hyper(uiop->uio_offset, tl);
1643 			*(tl + 2) = txdr_unsigned(len);
1644 		}
1645 		/*
1646 		 * Since I can't do a Getattr for NFSv4 for Write, there
1647 		 * doesn't seem any point in doing one here, either.
1648 		 * (See the comment in nfsrpc_writerpc() for more info.)
1649 		 */
1650 		error = nfscl_request(nd, vp, p, cred, stuff);
1651 		if (error)
1652 			return (error);
1653 		if (nd->nd_flag & ND_NFSV3) {
1654 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1655 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1656 			error = nfsm_loadattr(nd, nap);
1657 			if (!error)
1658 				*attrflagp = 1;
1659 		}
1660 		if (nd->nd_repstat || error) {
1661 			if (!error)
1662 				error = nd->nd_repstat;
1663 			goto nfsmout;
1664 		}
1665 		if (nd->nd_flag & ND_NFSV3) {
1666 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1667 			eof = fxdr_unsigned(int, *(tl + 1));
1668 		} else if (nd->nd_flag & ND_NFSV4) {
1669 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1670 			eof = fxdr_unsigned(int, *tl);
1671 		}
1672 		NFSM_STRSIZ(retlen, len);
1673 		error = nfsm_mbufuio(nd, uiop, retlen);
1674 		if (error)
1675 			goto nfsmout;
1676 		m_freem(nd->nd_mrep);
1677 		nd->nd_mrep = NULL;
1678 		tsiz -= retlen;
1679 		if (!(nd->nd_flag & ND_NFSV2)) {
1680 			if (eof || retlen == 0)
1681 				tsiz = 0;
1682 		} else if (retlen < len)
1683 			tsiz = 0;
1684 	}
1685 	return (0);
1686 nfsmout:
1687 	if (nd->nd_mrep != NULL)
1688 		m_freem(nd->nd_mrep);
1689 	return (error);
1690 }
1691 
1692 /*
1693  * nfs write operation
1694  * When called_from_strategy != 0, it should return EIO for an error that
1695  * indicates recovery is in progress, so that the buffer will be left
1696  * dirty and be written back to the server later. If it loops around,
1697  * the recovery thread could get stuck waiting for the buffer and recovery
1698  * will then deadlock.
1699  */
1700 APPLESTATIC int
1701 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1702     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1703     void *stuff, int called_from_strategy)
1704 {
1705 	int error, expireret = 0, retrycnt, nostateid;
1706 	u_int32_t clidrev = 0;
1707 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1708 	struct nfsnode *np = VTONFS(vp);
1709 	struct ucred *newcred;
1710 	struct nfsfh *nfhp = NULL;
1711 	nfsv4stateid_t stateid;
1712 	void *lckp;
1713 
1714 	*must_commit = 0;
1715 	if (nmp->nm_clp != NULL)
1716 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1717 	newcred = cred;
1718 	if (NFSHASNFSV4(nmp)) {
1719 		newcred = NFSNEWCRED(cred);
1720 		nfhp = np->n_fhp;
1721 	}
1722 	retrycnt = 0;
1723 	do {
1724 		lckp = NULL;
1725 		nostateid = 0;
1726 		if (NFSHASNFSV4(nmp)) {
1727 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1728 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1729 			    &lckp);
1730 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1731 			    stateid.other[2] == 0) {
1732 				nostateid = 1;
1733 				NFSCL_DEBUG(1, "stateid0 in write\n");
1734 			}
1735 		}
1736 
1737 		/*
1738 		 * If there is no stateid for NFSv4, it means this is an
1739 		 * extraneous write after close. Basically a poorly
1740 		 * implemented buffer cache. Just don't do the write.
1741 		 */
1742 		if (nostateid)
1743 			error = 0;
1744 		else
1745 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1746 			    newcred, &stateid, p, nap, attrflagp, stuff);
1747 		if (error == NFSERR_STALESTATEID)
1748 			nfscl_initiate_recovery(nmp->nm_clp);
1749 		if (lckp != NULL)
1750 			nfscl_lockderef(lckp);
1751 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1752 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1753 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1754 			(void) nfs_catnap(PZERO, error, "nfs_write");
1755 		} else if ((error == NFSERR_EXPIRED ||
1756 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1757 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1758 		}
1759 		retrycnt++;
1760 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1761 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1762 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1763 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1764 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1765 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1766 	if (error != 0 && (retrycnt >= 4 ||
1767 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1768 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1769 		error = EIO;
1770 	if (NFSHASNFSV4(nmp))
1771 		NFSFREECRED(newcred);
1772 	return (error);
1773 }
1774 
1775 /*
1776  * The actual write RPC.
1777  */
1778 static int
1779 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1780     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1781     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1782 {
1783 	u_int32_t *tl;
1784 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1785 	struct nfsnode *np = VTONFS(vp);
1786 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1787 	int wccflag = 0, wsize;
1788 	int32_t backup;
1789 	struct nfsrv_descript nfsd;
1790 	struct nfsrv_descript *nd = &nfsd;
1791 	nfsattrbit_t attrbits;
1792 	off_t tmp_off;
1793 
1794 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1795 	*attrflagp = 0;
1796 	tsiz = uiop->uio_resid;
1797 	tmp_off = uiop->uio_offset + tsiz;
1798 	NFSLOCKMNT(nmp);
1799 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1800 		NFSUNLOCKMNT(nmp);
1801 		return (EFBIG);
1802 	}
1803 	wsize = nmp->nm_wsize;
1804 	NFSUNLOCKMNT(nmp);
1805 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1806 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1807 	while (tsiz > 0) {
1808 		*attrflagp = 0;
1809 		len = (tsiz > wsize) ? wsize : tsiz;
1810 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1811 		if (nd->nd_flag & ND_NFSV4) {
1812 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1813 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1814 			txdr_hyper(uiop->uio_offset, tl);
1815 			tl += 2;
1816 			*tl++ = txdr_unsigned(*iomode);
1817 			*tl = txdr_unsigned(len);
1818 		} else if (nd->nd_flag & ND_NFSV3) {
1819 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1820 			txdr_hyper(uiop->uio_offset, tl);
1821 			tl += 2;
1822 			*tl++ = txdr_unsigned(len);
1823 			*tl++ = txdr_unsigned(*iomode);
1824 			*tl = txdr_unsigned(len);
1825 		} else {
1826 			u_int32_t x;
1827 
1828 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1829 			/*
1830 			 * Not sure why someone changed this, since the
1831 			 * RFC clearly states that "beginoffset" and
1832 			 * "totalcount" are ignored, but it wouldn't
1833 			 * surprise me if there's a busted server out there.
1834 			 */
1835 			/* Set both "begin" and "current" to non-garbage. */
1836 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1837 			*tl++ = x;      /* "begin offset" */
1838 			*tl++ = x;      /* "current offset" */
1839 			x = txdr_unsigned(len);
1840 			*tl++ = x;      /* total to this offset */
1841 			*tl = x;        /* size of this write */
1842 
1843 		}
1844 		nfsm_uiombuf(nd, uiop, len);
1845 		/*
1846 		 * Although it is tempting to do a normal Getattr Op in the
1847 		 * NFSv4 compound, the result can be a nearly hung client
1848 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1849 		 * It occurs when the client can't map either the Owner or
1850 		 * Owner_group name in the Getattr reply to a uid/gid. When
1851 		 * there is a cache miss, the kernel does an upcall to the
1852 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1853 		 * or /etc/group file. It can then block in getnewbuf(),
1854 		 * waiting for dirty writes to be pushed to the NFS server.
1855 		 * The only reason this doesn't result in a complete
1856 		 * deadlock, is that the upcall times out and allows
1857 		 * the write to complete. However, progress is so slow
1858 		 * that it might just as well be deadlocked.
1859 		 * As such, we get the rest of the attributes, but not
1860 		 * Owner or Owner_group.
1861 		 * nb: nfscl_loadattrcache() needs to be told that these
1862 		 *     partial attributes from a write rpc are being
1863 		 *     passed in, via a argument flag.
1864 		 */
1865 		if (nd->nd_flag & ND_NFSV4) {
1866 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
1867 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1868 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
1869 			(void) nfsrv_putattrbit(nd, &attrbits);
1870 		}
1871 		error = nfscl_request(nd, vp, p, cred, stuff);
1872 		if (error)
1873 			return (error);
1874 		if (nd->nd_repstat) {
1875 			/*
1876 			 * In case the rpc gets retried, roll
1877 			 * the uio fileds changed by nfsm_uiombuf()
1878 			 * back.
1879 			 */
1880 			uiop->uio_offset -= len;
1881 			uiop->uio_resid += len;
1882 			uiop->uio_iov->iov_base =
1883 			    (char *)uiop->uio_iov->iov_base - len;
1884 			uiop->uio_iov->iov_len += len;
1885 		}
1886 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1887 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1888 			    &wccflag, stuff);
1889 			if (error)
1890 				goto nfsmout;
1891 		}
1892 		if (!nd->nd_repstat) {
1893 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1894 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1895 					+ NFSX_VERF);
1896 				rlen = fxdr_unsigned(int, *tl++);
1897 				if (rlen == 0) {
1898 					error = NFSERR_IO;
1899 					goto nfsmout;
1900 				} else if (rlen < len) {
1901 					backup = len - rlen;
1902 					uiop->uio_iov->iov_base =
1903 					    (char *)uiop->uio_iov->iov_base -
1904 					    backup;
1905 					uiop->uio_iov->iov_len += backup;
1906 					uiop->uio_offset -= backup;
1907 					uiop->uio_resid += backup;
1908 					len = rlen;
1909 				}
1910 				commit = fxdr_unsigned(int, *tl++);
1911 
1912 				/*
1913 				 * Return the lowest commitment level
1914 				 * obtained by any of the RPCs.
1915 				 */
1916 				if (committed == NFSWRITE_FILESYNC)
1917 					committed = commit;
1918 				else if (committed == NFSWRITE_DATASYNC &&
1919 					commit == NFSWRITE_UNSTABLE)
1920 					committed = commit;
1921 				NFSLOCKMNT(nmp);
1922 				if (!NFSHASWRITEVERF(nmp)) {
1923 					NFSBCOPY((caddr_t)tl,
1924 					    (caddr_t)&nmp->nm_verf[0],
1925 					    NFSX_VERF);
1926 					NFSSETWRITEVERF(nmp);
1927 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1928 				    NFSX_VERF)) {
1929 					*must_commit = 1;
1930 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1931 				}
1932 				NFSUNLOCKMNT(nmp);
1933 			}
1934 			if (nd->nd_flag & ND_NFSV4)
1935 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1936 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1937 				error = nfsm_loadattr(nd, nap);
1938 				if (!error)
1939 					*attrflagp = NFS_LATTR_NOSHRINK;
1940 			}
1941 		} else {
1942 			error = nd->nd_repstat;
1943 		}
1944 		if (error)
1945 			goto nfsmout;
1946 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1947 		m_freem(nd->nd_mrep);
1948 		nd->nd_mrep = NULL;
1949 		tsiz -= len;
1950 	}
1951 nfsmout:
1952 	if (nd->nd_mrep != NULL)
1953 		m_freem(nd->nd_mrep);
1954 	*iomode = committed;
1955 	if (nd->nd_repstat && !error)
1956 		error = nd->nd_repstat;
1957 	return (error);
1958 }
1959 
1960 /*
1961  * nfs mknod rpc
1962  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1963  * mode set to specify the file type and the size field for rdev.
1964  */
1965 APPLESTATIC int
1966 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1967     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1968     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1969     int *attrflagp, int *dattrflagp, void *dstuff)
1970 {
1971 	u_int32_t *tl;
1972 	int error = 0;
1973 	struct nfsrv_descript nfsd, *nd = &nfsd;
1974 	nfsattrbit_t attrbits;
1975 
1976 	*nfhpp = NULL;
1977 	*attrflagp = 0;
1978 	*dattrflagp = 0;
1979 	if (namelen > NFS_MAXNAMLEN)
1980 		return (ENAMETOOLONG);
1981 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1982 	if (nd->nd_flag & ND_NFSV4) {
1983 		if (vtyp == VBLK || vtyp == VCHR) {
1984 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1985 			*tl++ = vtonfsv34_type(vtyp);
1986 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1987 			*tl = txdr_unsigned(NFSMINOR(rdev));
1988 		} else {
1989 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1990 			*tl = vtonfsv34_type(vtyp);
1991 		}
1992 	}
1993 	(void) nfsm_strtom(nd, name, namelen);
1994 	if (nd->nd_flag & ND_NFSV3) {
1995 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1996 		*tl = vtonfsv34_type(vtyp);
1997 	}
1998 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1999 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2000 	if ((nd->nd_flag & ND_NFSV3) &&
2001 	    (vtyp == VCHR || vtyp == VBLK)) {
2002 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2003 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2004 		*tl = txdr_unsigned(NFSMINOR(rdev));
2005 	}
2006 	if (nd->nd_flag & ND_NFSV4) {
2007 		NFSGETATTR_ATTRBIT(&attrbits);
2008 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2009 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2010 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2011 		(void) nfsrv_putattrbit(nd, &attrbits);
2012 	}
2013 	if (nd->nd_flag & ND_NFSV2)
2014 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2015 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2016 	if (error)
2017 		return (error);
2018 	if (nd->nd_flag & ND_NFSV4)
2019 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2020 	if (!nd->nd_repstat) {
2021 		if (nd->nd_flag & ND_NFSV4) {
2022 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2023 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2024 			if (error)
2025 				goto nfsmout;
2026 		}
2027 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2028 		if (error)
2029 			goto nfsmout;
2030 	}
2031 	if (nd->nd_flag & ND_NFSV3)
2032 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2033 	if (!error && nd->nd_repstat)
2034 		error = nd->nd_repstat;
2035 nfsmout:
2036 	m_freem(nd->nd_mrep);
2037 	return (error);
2038 }
2039 
2040 /*
2041  * nfs file create call
2042  * Mostly just call the approriate routine. (I separated out v4, so that
2043  * error recovery wouldn't be as difficult.)
2044  */
2045 APPLESTATIC int
2046 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2047     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2048     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2049     int *attrflagp, int *dattrflagp, void *dstuff)
2050 {
2051 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2052 	struct nfsclowner *owp;
2053 	struct nfscldeleg *dp;
2054 	struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
2055 	u_int32_t clidrev;
2056 
2057 	if (NFSHASNFSV4(nmp)) {
2058 	    retrycnt = 0;
2059 	    do {
2060 		dp = NULL;
2061 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2062 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2063 		    NULL, 1);
2064 		if (error)
2065 			return (error);
2066 		if (nmp->nm_clp != NULL)
2067 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2068 		else
2069 			clidrev = 0;
2070 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2071 		    nfs_numnfscbd == 0 || retrycnt > 0)
2072 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2073 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2074 			  attrflagp, dattrflagp, dstuff, &unlocked);
2075 		else
2076 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2077 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2078 			  attrflagp, dattrflagp, dstuff, &unlocked);
2079 		/*
2080 		 * There is no need to invalidate cached attributes here,
2081 		 * since new post-delegation issue attributes are always
2082 		 * returned by nfsrpc_createv4() and these will update the
2083 		 * attribute cache.
2084 		 */
2085 		if (dp != NULL)
2086 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2087 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2088 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2089 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2090 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2091 		    error == NFSERR_BADSESSION) {
2092 			(void) nfs_catnap(PZERO, error, "nfs_open");
2093 		} else if ((error == NFSERR_EXPIRED ||
2094 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2095 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2096 			retrycnt++;
2097 		}
2098 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2099 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2100 		error == NFSERR_BADSESSION ||
2101 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2102 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2103 	    if (error && retrycnt >= 4)
2104 		    error = EIO;
2105 	} else {
2106 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2107 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2108 		    dstuff);
2109 	}
2110 	return (error);
2111 }
2112 
2113 /*
2114  * The create rpc for v2 and 3.
2115  */
2116 static int
2117 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2118     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2119     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2120     int *attrflagp, int *dattrflagp, void *dstuff)
2121 {
2122 	u_int32_t *tl;
2123 	int error = 0;
2124 	struct nfsrv_descript nfsd, *nd = &nfsd;
2125 
2126 	*nfhpp = NULL;
2127 	*attrflagp = 0;
2128 	*dattrflagp = 0;
2129 	if (namelen > NFS_MAXNAMLEN)
2130 		return (ENAMETOOLONG);
2131 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2132 	(void) nfsm_strtom(nd, name, namelen);
2133 	if (nd->nd_flag & ND_NFSV3) {
2134 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2135 		if (fmode & O_EXCL) {
2136 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2137 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2138 			*tl++ = cverf.lval[0];
2139 			*tl = cverf.lval[1];
2140 		} else {
2141 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2142 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2143 		}
2144 	} else {
2145 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2146 	}
2147 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2148 	if (error)
2149 		return (error);
2150 	if (nd->nd_repstat == 0) {
2151 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2152 		if (error)
2153 			goto nfsmout;
2154 	}
2155 	if (nd->nd_flag & ND_NFSV3)
2156 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2157 	if (nd->nd_repstat != 0 && error == 0)
2158 		error = nd->nd_repstat;
2159 nfsmout:
2160 	m_freem(nd->nd_mrep);
2161 	return (error);
2162 }
2163 
2164 static int
2165 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2166     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2167     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2168     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2169     int *dattrflagp, void *dstuff, int *unlockedp)
2170 {
2171 	u_int32_t *tl;
2172 	int error = 0, deleg, newone, ret, acesize, limitby;
2173 	struct nfsrv_descript nfsd, *nd = &nfsd;
2174 	struct nfsclopen *op;
2175 	struct nfscldeleg *dp = NULL;
2176 	struct nfsnode *np;
2177 	struct nfsfh *nfhp;
2178 	nfsattrbit_t attrbits;
2179 	nfsv4stateid_t stateid;
2180 	u_int32_t rflags;
2181 	struct nfsmount *nmp;
2182 	struct nfsclsession *tsep;
2183 
2184 	nmp = VFSTONFS(dvp->v_mount);
2185 	np = VTONFS(dvp);
2186 	*unlockedp = 0;
2187 	*nfhpp = NULL;
2188 	*dpp = NULL;
2189 	*attrflagp = 0;
2190 	*dattrflagp = 0;
2191 	if (namelen > NFS_MAXNAMLEN)
2192 		return (ENAMETOOLONG);
2193 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2194 	/*
2195 	 * For V4, this is actually an Open op.
2196 	 */
2197 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2198 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2199 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2200 	    NFSV4OPEN_ACCESSREAD);
2201 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2202 	tsep = nfsmnt_mdssession(nmp);
2203 	*tl++ = tsep->nfsess_clientid.lval[0];
2204 	*tl = tsep->nfsess_clientid.lval[1];
2205 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2206 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2207 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2208 	if (fmode & O_EXCL) {
2209 		if (NFSHASNFSV4N(nmp)) {
2210 			if (NFSHASSESSPERSIST(nmp)) {
2211 				/* Use GUARDED for persistent sessions. */
2212 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2213 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2214 			} else {
2215 				/* Otherwise, use EXCLUSIVE4_1. */
2216 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2217 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2218 				*tl++ = cverf.lval[0];
2219 				*tl = cverf.lval[1];
2220 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2221 			}
2222 		} else {
2223 			/* NFSv4.0 */
2224 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2225 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2226 			*tl++ = cverf.lval[0];
2227 			*tl = cverf.lval[1];
2228 		}
2229 	} else {
2230 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2231 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2232 	}
2233 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2234 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2235 	(void) nfsm_strtom(nd, name, namelen);
2236 	/* Get the new file's handle and attributes. */
2237 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2238 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2239 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2240 	NFSGETATTR_ATTRBIT(&attrbits);
2241 	(void) nfsrv_putattrbit(nd, &attrbits);
2242 	/* Get the directory's post-op attributes. */
2243 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2244 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2245 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2246 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2247 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2248 	(void) nfsrv_putattrbit(nd, &attrbits);
2249 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2250 	if (error)
2251 		return (error);
2252 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2253 	if (nd->nd_repstat == 0) {
2254 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2255 		    6 * NFSX_UNSIGNED);
2256 		stateid.seqid = *tl++;
2257 		stateid.other[0] = *tl++;
2258 		stateid.other[1] = *tl++;
2259 		stateid.other[2] = *tl;
2260 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2261 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2262 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2263 		deleg = fxdr_unsigned(int, *tl);
2264 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2265 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2266 			if (!(owp->nfsow_clp->nfsc_flags &
2267 			      NFSCLFLAGS_FIRSTDELEG))
2268 				owp->nfsow_clp->nfsc_flags |=
2269 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2270 			dp = malloc(
2271 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2272 			    M_NFSCLDELEG, M_WAITOK);
2273 			LIST_INIT(&dp->nfsdl_owner);
2274 			LIST_INIT(&dp->nfsdl_lock);
2275 			dp->nfsdl_clp = owp->nfsow_clp;
2276 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2277 			nfscl_lockinit(&dp->nfsdl_rwlock);
2278 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2279 			    NFSX_UNSIGNED);
2280 			dp->nfsdl_stateid.seqid = *tl++;
2281 			dp->nfsdl_stateid.other[0] = *tl++;
2282 			dp->nfsdl_stateid.other[1] = *tl++;
2283 			dp->nfsdl_stateid.other[2] = *tl++;
2284 			ret = fxdr_unsigned(int, *tl);
2285 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2286 				dp->nfsdl_flags = NFSCLDL_WRITE;
2287 				/*
2288 				 * Indicates how much the file can grow.
2289 				 */
2290 				NFSM_DISSECT(tl, u_int32_t *,
2291 				    3 * NFSX_UNSIGNED);
2292 				limitby = fxdr_unsigned(int, *tl++);
2293 				switch (limitby) {
2294 				case NFSV4OPEN_LIMITSIZE:
2295 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2296 					break;
2297 				case NFSV4OPEN_LIMITBLOCKS:
2298 					dp->nfsdl_sizelimit =
2299 					    fxdr_unsigned(u_int64_t, *tl++);
2300 					dp->nfsdl_sizelimit *=
2301 					    fxdr_unsigned(u_int64_t, *tl);
2302 					break;
2303 				default:
2304 					error = NFSERR_BADXDR;
2305 					goto nfsmout;
2306 				}
2307 			} else {
2308 				dp->nfsdl_flags = NFSCLDL_READ;
2309 			}
2310 			if (ret)
2311 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2312 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2313 			    &acesize, p);
2314 			if (error)
2315 				goto nfsmout;
2316 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2317 			error = NFSERR_BADXDR;
2318 			goto nfsmout;
2319 		}
2320 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2321 		if (error)
2322 			goto nfsmout;
2323 		/* Get rid of the PutFH and Getattr status values. */
2324 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2325 		/* Load the directory attributes. */
2326 		error = nfsm_loadattr(nd, dnap);
2327 		if (error)
2328 			goto nfsmout;
2329 		*dattrflagp = 1;
2330 		if (dp != NULL && *attrflagp) {
2331 			dp->nfsdl_change = nnap->na_filerev;
2332 			dp->nfsdl_modtime = nnap->na_mtime;
2333 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2334 		}
2335 		/*
2336 		 * We can now complete the Open state.
2337 		 */
2338 		nfhp = *nfhpp;
2339 		if (dp != NULL) {
2340 			dp->nfsdl_fhlen = nfhp->nfh_len;
2341 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2342 		}
2343 		/*
2344 		 * Get an Open structure that will be
2345 		 * attached to the OpenOwner, acquired already.
2346 		 */
2347 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2348 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2349 		    cred, p, NULL, &op, &newone, NULL, 0);
2350 		if (error)
2351 			goto nfsmout;
2352 		op->nfso_stateid = stateid;
2353 		newnfs_copyincred(cred, &op->nfso_cred);
2354 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2355 		    do {
2356 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2357 			    nfhp->nfh_len, op, cred, p);
2358 			if (ret == NFSERR_DELAY)
2359 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2360 		    } while (ret == NFSERR_DELAY);
2361 		    error = ret;
2362 		}
2363 
2364 		/*
2365 		 * If the server is handing out delegations, but we didn't
2366 		 * get one because an OpenConfirm was required, try the
2367 		 * Open again, to get a delegation. This is a harmless no-op,
2368 		 * from a server's point of view.
2369 		 */
2370 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2371 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2372 		    !error && dp == NULL) {
2373 		    do {
2374 			ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2375 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2376 			    nfhp->nfh_fh, nfhp->nfh_len,
2377 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2378 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2379 			if (ret == NFSERR_DELAY)
2380 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2381 		    } while (ret == NFSERR_DELAY);
2382 		    if (ret) {
2383 			if (dp != NULL) {
2384 				free(dp, M_NFSCLDELEG);
2385 				dp = NULL;
2386 			}
2387 			if (ret == NFSERR_STALECLIENTID ||
2388 			    ret == NFSERR_STALEDONTRECOVER ||
2389 			    ret == NFSERR_BADSESSION)
2390 				error = ret;
2391 		    }
2392 		}
2393 		nfscl_openrelease(nmp, op, error, newone);
2394 		*unlockedp = 1;
2395 	}
2396 	if (nd->nd_repstat != 0 && error == 0)
2397 		error = nd->nd_repstat;
2398 	if (error == NFSERR_STALECLIENTID)
2399 		nfscl_initiate_recovery(owp->nfsow_clp);
2400 nfsmout:
2401 	if (!error)
2402 		*dpp = dp;
2403 	else if (dp != NULL)
2404 		free(dp, M_NFSCLDELEG);
2405 	m_freem(nd->nd_mrep);
2406 	return (error);
2407 }
2408 
2409 /*
2410  * Nfs remove rpc
2411  */
2412 APPLESTATIC int
2413 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2414     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2415     void *dstuff)
2416 {
2417 	u_int32_t *tl;
2418 	struct nfsrv_descript nfsd, *nd = &nfsd;
2419 	struct nfsnode *np;
2420 	struct nfsmount *nmp;
2421 	nfsv4stateid_t dstateid;
2422 	int error, ret = 0, i;
2423 
2424 	*dattrflagp = 0;
2425 	if (namelen > NFS_MAXNAMLEN)
2426 		return (ENAMETOOLONG);
2427 	nmp = VFSTONFS(vnode_mount(dvp));
2428 tryagain:
2429 	if (NFSHASNFSV4(nmp) && ret == 0) {
2430 		ret = nfscl_removedeleg(vp, p, &dstateid);
2431 		if (ret == 1) {
2432 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2433 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2434 			    NFSX_UNSIGNED);
2435 			if (NFSHASNFSV4N(nmp))
2436 				*tl++ = 0;
2437 			else
2438 				*tl++ = dstateid.seqid;
2439 			*tl++ = dstateid.other[0];
2440 			*tl++ = dstateid.other[1];
2441 			*tl++ = dstateid.other[2];
2442 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2443 			np = VTONFS(dvp);
2444 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2445 			    np->n_fhp->nfh_len, 0);
2446 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2447 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2448 		}
2449 	} else {
2450 		ret = 0;
2451 	}
2452 	if (ret == 0)
2453 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2454 	(void) nfsm_strtom(nd, name, namelen);
2455 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2456 	if (error)
2457 		return (error);
2458 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2459 		/* For NFSv4, parse out any Delereturn replies. */
2460 		if (ret > 0 && nd->nd_repstat != 0 &&
2461 		    (nd->nd_flag & ND_NOMOREDATA)) {
2462 			/*
2463 			 * If the Delegreturn failed, try again without
2464 			 * it. The server will Recall, as required.
2465 			 */
2466 			m_freem(nd->nd_mrep);
2467 			goto tryagain;
2468 		}
2469 		for (i = 0; i < (ret * 2); i++) {
2470 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2471 			    ND_NFSV4) {
2472 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2473 			    if (*(tl + 1))
2474 				nd->nd_flag |= ND_NOMOREDATA;
2475 			}
2476 		}
2477 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2478 	}
2479 	if (nd->nd_repstat && !error)
2480 		error = nd->nd_repstat;
2481 nfsmout:
2482 	m_freem(nd->nd_mrep);
2483 	return (error);
2484 }
2485 
2486 /*
2487  * Do an nfs rename rpc.
2488  */
2489 APPLESTATIC int
2490 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2491     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2492     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2493     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2494 {
2495 	u_int32_t *tl;
2496 	struct nfsrv_descript nfsd, *nd = &nfsd;
2497 	struct nfsmount *nmp;
2498 	struct nfsnode *np;
2499 	nfsattrbit_t attrbits;
2500 	nfsv4stateid_t fdstateid, tdstateid;
2501 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2502 
2503 	*fattrflagp = 0;
2504 	*tattrflagp = 0;
2505 	nmp = VFSTONFS(vnode_mount(fdvp));
2506 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2507 		return (ENAMETOOLONG);
2508 tryagain:
2509 	if (NFSHASNFSV4(nmp) && ret == 0) {
2510 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2511 		    &tdstateid, &gottd, p);
2512 		if (gotfd && gottd) {
2513 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2514 		} else if (gotfd) {
2515 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2516 		} else if (gottd) {
2517 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2518 		}
2519 		if (gotfd) {
2520 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2521 			if (NFSHASNFSV4N(nmp))
2522 				*tl++ = 0;
2523 			else
2524 				*tl++ = fdstateid.seqid;
2525 			*tl++ = fdstateid.other[0];
2526 			*tl++ = fdstateid.other[1];
2527 			*tl = fdstateid.other[2];
2528 			if (gottd) {
2529 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2530 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2531 				np = VTONFS(tvp);
2532 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2533 				    np->n_fhp->nfh_len, 0);
2534 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2535 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2536 			}
2537 		}
2538 		if (gottd) {
2539 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2540 			if (NFSHASNFSV4N(nmp))
2541 				*tl++ = 0;
2542 			else
2543 				*tl++ = tdstateid.seqid;
2544 			*tl++ = tdstateid.other[0];
2545 			*tl++ = tdstateid.other[1];
2546 			*tl = tdstateid.other[2];
2547 		}
2548 		if (ret > 0) {
2549 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2550 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2551 			np = VTONFS(fdvp);
2552 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2553 			    np->n_fhp->nfh_len, 0);
2554 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2555 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2556 		}
2557 	} else {
2558 		ret = 0;
2559 	}
2560 	if (ret == 0)
2561 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2562 	if (nd->nd_flag & ND_NFSV4) {
2563 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2564 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2565 		NFSWCCATTR_ATTRBIT(&attrbits);
2566 		(void) nfsrv_putattrbit(nd, &attrbits);
2567 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2568 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2569 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2570 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2571 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2572 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2573 		(void) nfsrv_putattrbit(nd, &attrbits);
2574 		nd->nd_flag |= ND_V4WCCATTR;
2575 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2576 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2577 	}
2578 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2579 	if (!(nd->nd_flag & ND_NFSV4))
2580 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2581 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2582 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2583 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2584 	if (error)
2585 		return (error);
2586 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2587 		/* For NFSv4, parse out any Delereturn replies. */
2588 		if (ret > 0 && nd->nd_repstat != 0 &&
2589 		    (nd->nd_flag & ND_NOMOREDATA)) {
2590 			/*
2591 			 * If the Delegreturn failed, try again without
2592 			 * it. The server will Recall, as required.
2593 			 */
2594 			m_freem(nd->nd_mrep);
2595 			goto tryagain;
2596 		}
2597 		for (i = 0; i < (ret * 2); i++) {
2598 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2599 			    ND_NFSV4) {
2600 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2601 			    if (*(tl + 1)) {
2602 				if (i == 0 && ret > 1) {
2603 				    /*
2604 				     * If the Delegreturn failed, try again
2605 				     * without it. The server will Recall, as
2606 				     * required.
2607 				     * If ret > 1, the first iteration of this
2608 				     * loop is the second DelegReturn result.
2609 				     */
2610 				    m_freem(nd->nd_mrep);
2611 				    goto tryagain;
2612 				} else {
2613 				    nd->nd_flag |= ND_NOMOREDATA;
2614 				}
2615 			    }
2616 			}
2617 		}
2618 		/* Now, the first wcc attribute reply. */
2619 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2620 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2621 			if (*(tl + 1))
2622 				nd->nd_flag |= ND_NOMOREDATA;
2623 		}
2624 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2625 		    fstuff);
2626 		/* and the second wcc attribute reply. */
2627 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2628 		    !error) {
2629 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2630 			if (*(tl + 1))
2631 				nd->nd_flag |= ND_NOMOREDATA;
2632 		}
2633 		if (!error)
2634 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2635 			    NULL, tstuff);
2636 	}
2637 	if (nd->nd_repstat && !error)
2638 		error = nd->nd_repstat;
2639 nfsmout:
2640 	m_freem(nd->nd_mrep);
2641 	return (error);
2642 }
2643 
2644 /*
2645  * nfs hard link create rpc
2646  */
2647 APPLESTATIC int
2648 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2649     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2650     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2651 {
2652 	u_int32_t *tl;
2653 	struct nfsrv_descript nfsd, *nd = &nfsd;
2654 	nfsattrbit_t attrbits;
2655 	int error = 0;
2656 
2657 	*attrflagp = 0;
2658 	*dattrflagp = 0;
2659 	if (namelen > NFS_MAXNAMLEN)
2660 		return (ENAMETOOLONG);
2661 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2662 	if (nd->nd_flag & ND_NFSV4) {
2663 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2664 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2665 	}
2666 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2667 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2668 	if (nd->nd_flag & ND_NFSV4) {
2669 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2670 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2671 		NFSWCCATTR_ATTRBIT(&attrbits);
2672 		(void) nfsrv_putattrbit(nd, &attrbits);
2673 		nd->nd_flag |= ND_V4WCCATTR;
2674 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2675 		*tl = txdr_unsigned(NFSV4OP_LINK);
2676 	}
2677 	(void) nfsm_strtom(nd, name, namelen);
2678 	error = nfscl_request(nd, vp, p, cred, dstuff);
2679 	if (error)
2680 		return (error);
2681 	if (nd->nd_flag & ND_NFSV3) {
2682 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2683 		if (!error)
2684 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2685 			    NULL, dstuff);
2686 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2687 		/*
2688 		 * First, parse out the PutFH and Getattr result.
2689 		 */
2690 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2691 		if (!(*(tl + 1)))
2692 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2693 		if (*(tl + 1))
2694 			nd->nd_flag |= ND_NOMOREDATA;
2695 		/*
2696 		 * Get the pre-op attributes.
2697 		 */
2698 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2699 	}
2700 	if (nd->nd_repstat && !error)
2701 		error = nd->nd_repstat;
2702 nfsmout:
2703 	m_freem(nd->nd_mrep);
2704 	return (error);
2705 }
2706 
2707 /*
2708  * nfs symbolic link create rpc
2709  */
2710 APPLESTATIC int
2711 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
2712     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2713     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2714     int *dattrflagp, void *dstuff)
2715 {
2716 	u_int32_t *tl;
2717 	struct nfsrv_descript nfsd, *nd = &nfsd;
2718 	struct nfsmount *nmp;
2719 	int slen, error = 0;
2720 
2721 	*nfhpp = NULL;
2722 	*attrflagp = 0;
2723 	*dattrflagp = 0;
2724 	nmp = VFSTONFS(vnode_mount(dvp));
2725 	slen = strlen(target);
2726 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2727 		return (ENAMETOOLONG);
2728 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2729 	if (nd->nd_flag & ND_NFSV4) {
2730 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2731 		*tl = txdr_unsigned(NFLNK);
2732 		(void) nfsm_strtom(nd, target, slen);
2733 	}
2734 	(void) nfsm_strtom(nd, name, namelen);
2735 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2736 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2737 	if (!(nd->nd_flag & ND_NFSV4))
2738 		(void) nfsm_strtom(nd, target, slen);
2739 	if (nd->nd_flag & ND_NFSV2)
2740 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2741 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2742 	if (error)
2743 		return (error);
2744 	if (nd->nd_flag & ND_NFSV4)
2745 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2746 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2747 		if (!nd->nd_repstat)
2748 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2749 		if (!error)
2750 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2751 			    NULL, dstuff);
2752 	}
2753 	if (nd->nd_repstat && !error)
2754 		error = nd->nd_repstat;
2755 	m_freem(nd->nd_mrep);
2756 	/*
2757 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2758 	 * Only do this if vfs.nfs.ignore_eexist is set.
2759 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2760 	 * should guarantee "exactly once" RPC semantics.
2761 	 */
2762 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2763 	    nmp->nm_minorvers == 0))
2764 		error = 0;
2765 	return (error);
2766 }
2767 
2768 /*
2769  * nfs make dir rpc
2770  */
2771 APPLESTATIC int
2772 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2773     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2774     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2775     int *dattrflagp, void *dstuff)
2776 {
2777 	u_int32_t *tl;
2778 	struct nfsrv_descript nfsd, *nd = &nfsd;
2779 	nfsattrbit_t attrbits;
2780 	int error = 0;
2781 	struct nfsfh *fhp;
2782 	struct nfsmount *nmp;
2783 
2784 	*nfhpp = NULL;
2785 	*attrflagp = 0;
2786 	*dattrflagp = 0;
2787 	nmp = VFSTONFS(vnode_mount(dvp));
2788 	fhp = VTONFS(dvp)->n_fhp;
2789 	if (namelen > NFS_MAXNAMLEN)
2790 		return (ENAMETOOLONG);
2791 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2792 	if (nd->nd_flag & ND_NFSV4) {
2793 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2794 		*tl = txdr_unsigned(NFDIR);
2795 	}
2796 	(void) nfsm_strtom(nd, name, namelen);
2797 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2798 	if (nd->nd_flag & ND_NFSV4) {
2799 		NFSGETATTR_ATTRBIT(&attrbits);
2800 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2801 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2802 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2803 		(void) nfsrv_putattrbit(nd, &attrbits);
2804 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2805 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2806 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2807 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2808 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2809 		(void) nfsrv_putattrbit(nd, &attrbits);
2810 	}
2811 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2812 	if (error)
2813 		return (error);
2814 	if (nd->nd_flag & ND_NFSV4)
2815 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2816 	if (!nd->nd_repstat && !error) {
2817 		if (nd->nd_flag & ND_NFSV4) {
2818 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2819 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2820 		}
2821 		if (!error)
2822 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2823 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2824 			/* Get rid of the PutFH and Getattr status values. */
2825 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2826 			/* Load the directory attributes. */
2827 			error = nfsm_loadattr(nd, dnap);
2828 			if (error == 0)
2829 				*dattrflagp = 1;
2830 		}
2831 	}
2832 	if ((nd->nd_flag & ND_NFSV3) && !error)
2833 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2834 	if (nd->nd_repstat && !error)
2835 		error = nd->nd_repstat;
2836 nfsmout:
2837 	m_freem(nd->nd_mrep);
2838 	/*
2839 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2840 	 * Only do this if vfs.nfs.ignore_eexist is set.
2841 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2842 	 * should guarantee "exactly once" RPC semantics.
2843 	 */
2844 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2845 	    nmp->nm_minorvers == 0))
2846 		error = 0;
2847 	return (error);
2848 }
2849 
2850 /*
2851  * nfs remove directory call
2852  */
2853 APPLESTATIC int
2854 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2855     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2856 {
2857 	struct nfsrv_descript nfsd, *nd = &nfsd;
2858 	int error = 0;
2859 
2860 	*dattrflagp = 0;
2861 	if (namelen > NFS_MAXNAMLEN)
2862 		return (ENAMETOOLONG);
2863 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2864 	(void) nfsm_strtom(nd, name, namelen);
2865 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2866 	if (error)
2867 		return (error);
2868 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2869 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2870 	if (nd->nd_repstat && !error)
2871 		error = nd->nd_repstat;
2872 	m_freem(nd->nd_mrep);
2873 	/*
2874 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2875 	 */
2876 	if (error == ENOENT)
2877 		error = 0;
2878 	return (error);
2879 }
2880 
2881 /*
2882  * Readdir rpc.
2883  * Always returns with either uio_resid unchanged, if you are at the
2884  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2885  * filled in.
2886  * I felt this would allow caching of directory blocks more easily
2887  * than returning a pertially filled block.
2888  * Directory offset cookies:
2889  * Oh my, what to do with them...
2890  * I can think of three ways to deal with them:
2891  * 1 - have the layer above these RPCs maintain a map between logical
2892  *     directory byte offsets and the NFS directory offset cookies
2893  * 2 - pass the opaque directory offset cookies up into userland
2894  *     and let the libc functions deal with them, via the system call
2895  * 3 - return them to userland in the "struct dirent", so future versions
2896  *     of libc can use them and do whatever is necessary to make things work
2897  *     above these rpc calls, in the meantime
2898  * For now, I do #3 by "hiding" the directory offset cookies after the
2899  * d_name field in struct dirent. This is space inside d_reclen that
2900  * will be ignored by anything that doesn't know about them.
2901  * The directory offset cookies are filled in as the last 8 bytes of
2902  * each directory entry, after d_name. Someday, the userland libc
2903  * functions may be able to use these. In the meantime, it satisfies
2904  * OpenBSD's requirements for cookies being returned.
2905  * If expects the directory offset cookie for the read to be in uio_offset
2906  * and returns the one for the next entry after this directory block in
2907  * there, as well.
2908  */
2909 APPLESTATIC int
2910 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2911     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2912     int *eofp, void *stuff)
2913 {
2914 	int len, left;
2915 	struct dirent *dp = NULL;
2916 	u_int32_t *tl;
2917 	nfsquad_t cookie, ncookie;
2918 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2919 	struct nfsnode *dnp = VTONFS(vp);
2920 	struct nfsvattr nfsva;
2921 	struct nfsrv_descript nfsd, *nd = &nfsd;
2922 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2923 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2924 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2925 	char *cp;
2926 	nfsattrbit_t attrbits, dattrbits;
2927 	u_int32_t rderr, *tl2 = NULL;
2928 	size_t tresid;
2929 
2930 	KASSERT(uiop->uio_iovcnt == 1 &&
2931 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
2932 	    ("nfs readdirrpc bad uio"));
2933 	ncookie.lval[0] = ncookie.lval[1] = 0;
2934 	/*
2935 	 * There is no point in reading a lot more than uio_resid, however
2936 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2937 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2938 	 * will never make readsize > nm_readdirsize.
2939 	 */
2940 	readsize = nmp->nm_readdirsize;
2941 	if (readsize > uiop->uio_resid)
2942 		readsize = uiop->uio_resid + DIRBLKSIZ;
2943 
2944 	*attrflagp = 0;
2945 	if (eofp)
2946 		*eofp = 0;
2947 	tresid = uiop->uio_resid;
2948 	cookie.lval[0] = cookiep->nfsuquad[0];
2949 	cookie.lval[1] = cookiep->nfsuquad[1];
2950 	nd->nd_mrep = NULL;
2951 
2952 	/*
2953 	 * For NFSv4, first create the "." and ".." entries.
2954 	 */
2955 	if (NFSHASNFSV4(nmp)) {
2956 		reqsize = 6 * NFSX_UNSIGNED;
2957 		NFSGETATTR_ATTRBIT(&dattrbits);
2958 		NFSZERO_ATTRBIT(&attrbits);
2959 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2960 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2961 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2962 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2963 			NFSSETBIT_ATTRBIT(&attrbits,
2964 			    NFSATTRBIT_MOUNTEDONFILEID);
2965 			gotmnton = 1;
2966 		} else {
2967 			/*
2968 			 * Must fake it. Use the fileno, except when the
2969 			 * fsid is != to that of the directory. For that
2970 			 * case, generate a fake fileno that is not the same.
2971 			 */
2972 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2973 			gotmnton = 0;
2974 		}
2975 
2976 		/*
2977 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2978 		 */
2979 		if (uiop->uio_offset == 0) {
2980 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2981 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2982 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2983 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2984 			(void) nfsrv_putattrbit(nd, &attrbits);
2985 			error = nfscl_request(nd, vp, p, cred, stuff);
2986 			if (error)
2987 			    return (error);
2988 			dotfileid = 0;	/* Fake out the compiler. */
2989 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2990 			    error = nfsm_loadattr(nd, &nfsva);
2991 			    if (error != 0)
2992 				goto nfsmout;
2993 			    dotfileid = nfsva.na_fileid;
2994 			}
2995 			if (nd->nd_repstat == 0) {
2996 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2997 			    len = fxdr_unsigned(int, *(tl + 4));
2998 			    if (len > 0 && len <= NFSX_V4FHMAX)
2999 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3000 			    else
3001 				error = EPERM;
3002 			    if (!error) {
3003 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3004 				nfsva.na_mntonfileno = UINT64_MAX;
3005 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3006 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3007 				    NULL, NULL, NULL, p, cred);
3008 				if (error) {
3009 				    dotdotfileid = dotfileid;
3010 				} else if (gotmnton) {
3011 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3012 					dotdotfileid = nfsva.na_mntonfileno;
3013 				    else
3014 					dotdotfileid = nfsva.na_fileid;
3015 				} else if (nfsva.na_filesid[0] ==
3016 				    dnp->n_vattr.na_filesid[0] &&
3017 				    nfsva.na_filesid[1] ==
3018 				    dnp->n_vattr.na_filesid[1]) {
3019 				    dotdotfileid = nfsva.na_fileid;
3020 				} else {
3021 				    do {
3022 					fakefileno--;
3023 				    } while (fakefileno ==
3024 					nfsva.na_fileid);
3025 				    dotdotfileid = fakefileno;
3026 				}
3027 			    }
3028 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3029 			    /*
3030 			     * Lookupp returns NFSERR_NOENT when we are
3031 			     * at the root, so just use the current dir.
3032 			     */
3033 			    nd->nd_repstat = 0;
3034 			    dotdotfileid = dotfileid;
3035 			} else {
3036 			    error = nd->nd_repstat;
3037 			}
3038 			m_freem(nd->nd_mrep);
3039 			if (error)
3040 			    return (error);
3041 			nd->nd_mrep = NULL;
3042 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3043 			dp->d_pad0 = dp->d_pad1 = 0;
3044 			dp->d_off = 0;
3045 			dp->d_type = DT_DIR;
3046 			dp->d_fileno = dotfileid;
3047 			dp->d_namlen = 1;
3048 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3049 			dp->d_name[0] = '.';
3050 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3051 			/*
3052 			 * Just make these offset cookie 0.
3053 			 */
3054 			tl = (u_int32_t *)&dp->d_name[8];
3055 			*tl++ = 0;
3056 			*tl = 0;
3057 			blksiz += dp->d_reclen;
3058 			uiop->uio_resid -= dp->d_reclen;
3059 			uiop->uio_offset += dp->d_reclen;
3060 			uiop->uio_iov->iov_base =
3061 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3062 			uiop->uio_iov->iov_len -= dp->d_reclen;
3063 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3064 			dp->d_pad0 = dp->d_pad1 = 0;
3065 			dp->d_off = 0;
3066 			dp->d_type = DT_DIR;
3067 			dp->d_fileno = dotdotfileid;
3068 			dp->d_namlen = 2;
3069 			*((uint64_t *)dp->d_name) = 0;
3070 			dp->d_name[0] = '.';
3071 			dp->d_name[1] = '.';
3072 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3073 			/*
3074 			 * Just make these offset cookie 0.
3075 			 */
3076 			tl = (u_int32_t *)&dp->d_name[8];
3077 			*tl++ = 0;
3078 			*tl = 0;
3079 			blksiz += dp->d_reclen;
3080 			uiop->uio_resid -= dp->d_reclen;
3081 			uiop->uio_offset += dp->d_reclen;
3082 			uiop->uio_iov->iov_base =
3083 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3084 			uiop->uio_iov->iov_len -= dp->d_reclen;
3085 		}
3086 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3087 	} else {
3088 		reqsize = 5 * NFSX_UNSIGNED;
3089 	}
3090 
3091 
3092 	/*
3093 	 * Loop around doing readdir rpc's of size readsize.
3094 	 * The stopping criteria is EOF or buffer full.
3095 	 */
3096 	while (more_dirs && bigenough) {
3097 		*attrflagp = 0;
3098 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3099 		if (nd->nd_flag & ND_NFSV2) {
3100 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3101 			*tl++ = cookie.lval[1];
3102 			*tl = txdr_unsigned(readsize);
3103 		} else {
3104 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3105 			*tl++ = cookie.lval[0];
3106 			*tl++ = cookie.lval[1];
3107 			if (cookie.qval == 0) {
3108 				*tl++ = 0;
3109 				*tl++ = 0;
3110 			} else {
3111 				NFSLOCKNODE(dnp);
3112 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3113 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3114 				NFSUNLOCKNODE(dnp);
3115 			}
3116 			if (nd->nd_flag & ND_NFSV4) {
3117 				*tl++ = txdr_unsigned(readsize);
3118 				*tl = txdr_unsigned(readsize);
3119 				(void) nfsrv_putattrbit(nd, &attrbits);
3120 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3121 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3122 				(void) nfsrv_putattrbit(nd, &dattrbits);
3123 			} else {
3124 				*tl = txdr_unsigned(readsize);
3125 			}
3126 		}
3127 		error = nfscl_request(nd, vp, p, cred, stuff);
3128 		if (error)
3129 			return (error);
3130 		if (!(nd->nd_flag & ND_NFSV2)) {
3131 			if (nd->nd_flag & ND_NFSV3)
3132 				error = nfscl_postop_attr(nd, nap, attrflagp,
3133 				    stuff);
3134 			if (!nd->nd_repstat && !error) {
3135 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3136 				NFSLOCKNODE(dnp);
3137 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3138 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3139 				NFSUNLOCKNODE(dnp);
3140 			}
3141 		}
3142 		if (nd->nd_repstat || error) {
3143 			if (!error)
3144 				error = nd->nd_repstat;
3145 			goto nfsmout;
3146 		}
3147 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3148 		more_dirs = fxdr_unsigned(int, *tl);
3149 		if (!more_dirs)
3150 			tryformoredirs = 0;
3151 
3152 		/* loop through the dir entries, doctoring them to 4bsd form */
3153 		while (more_dirs && bigenough) {
3154 			if (nd->nd_flag & ND_NFSV4) {
3155 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3156 				ncookie.lval[0] = *tl++;
3157 				ncookie.lval[1] = *tl++;
3158 				len = fxdr_unsigned(int, *tl);
3159 			} else if (nd->nd_flag & ND_NFSV3) {
3160 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3161 				nfsva.na_fileid = fxdr_hyper(tl);
3162 				tl += 2;
3163 				len = fxdr_unsigned(int, *tl);
3164 			} else {
3165 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3166 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3167 				    *tl++);
3168 				len = fxdr_unsigned(int, *tl);
3169 			}
3170 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3171 				error = EBADRPC;
3172 				goto nfsmout;
3173 			}
3174 			tlen = roundup2(len, 8);
3175 			if (tlen == len)
3176 				tlen += 8;  /* To ensure null termination. */
3177 			left = DIRBLKSIZ - blksiz;
3178 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3179 				NFSBZERO(uiop->uio_iov->iov_base, left);
3180 				dp->d_reclen += left;
3181 				uiop->uio_iov->iov_base =
3182 				    (char *)uiop->uio_iov->iov_base + left;
3183 				uiop->uio_iov->iov_len -= left;
3184 				uiop->uio_resid -= left;
3185 				uiop->uio_offset += left;
3186 				blksiz = 0;
3187 			}
3188 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3189 			    uiop->uio_resid)
3190 				bigenough = 0;
3191 			if (bigenough) {
3192 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3193 				dp->d_pad0 = dp->d_pad1 = 0;
3194 				dp->d_off = 0;
3195 				dp->d_namlen = len;
3196 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3197 				    NFSX_HYPER;
3198 				dp->d_type = DT_UNKNOWN;
3199 				blksiz += dp->d_reclen;
3200 				if (blksiz == DIRBLKSIZ)
3201 					blksiz = 0;
3202 				uiop->uio_resid -= DIRHDSIZ;
3203 				uiop->uio_offset += DIRHDSIZ;
3204 				uiop->uio_iov->iov_base =
3205 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3206 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3207 				error = nfsm_mbufuio(nd, uiop, len);
3208 				if (error)
3209 					goto nfsmout;
3210 				cp = uiop->uio_iov->iov_base;
3211 				tlen -= len;
3212 				NFSBZERO(cp, tlen);
3213 				cp += tlen;	/* points to cookie storage */
3214 				tl2 = (u_int32_t *)cp;
3215 				uiop->uio_iov->iov_base =
3216 				    (char *)uiop->uio_iov->iov_base + tlen +
3217 				    NFSX_HYPER;
3218 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3219 				uiop->uio_resid -= tlen + NFSX_HYPER;
3220 				uiop->uio_offset += (tlen + NFSX_HYPER);
3221 			} else {
3222 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3223 				if (error)
3224 					goto nfsmout;
3225 			}
3226 			if (nd->nd_flag & ND_NFSV4) {
3227 				rderr = 0;
3228 				nfsva.na_mntonfileno = UINT64_MAX;
3229 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3230 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3231 				    NULL, NULL, &rderr, p, cred);
3232 				if (error)
3233 					goto nfsmout;
3234 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3235 			} else if (nd->nd_flag & ND_NFSV3) {
3236 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3237 				ncookie.lval[0] = *tl++;
3238 				ncookie.lval[1] = *tl++;
3239 			} else {
3240 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3241 				ncookie.lval[0] = 0;
3242 				ncookie.lval[1] = *tl++;
3243 			}
3244 			if (bigenough) {
3245 			    if (nd->nd_flag & ND_NFSV4) {
3246 				if (rderr) {
3247 				    dp->d_fileno = 0;
3248 				} else {
3249 				    if (gotmnton) {
3250 					if (nfsva.na_mntonfileno != UINT64_MAX)
3251 					    dp->d_fileno = nfsva.na_mntonfileno;
3252 					else
3253 					    dp->d_fileno = nfsva.na_fileid;
3254 				    } else if (nfsva.na_filesid[0] ==
3255 					dnp->n_vattr.na_filesid[0] &&
3256 					nfsva.na_filesid[1] ==
3257 					dnp->n_vattr.na_filesid[1]) {
3258 					dp->d_fileno = nfsva.na_fileid;
3259 				    } else {
3260 					do {
3261 					    fakefileno--;
3262 					} while (fakefileno ==
3263 					    nfsva.na_fileid);
3264 					dp->d_fileno = fakefileno;
3265 				    }
3266 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3267 				}
3268 			    } else {
3269 				dp->d_fileno = nfsva.na_fileid;
3270 			    }
3271 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3272 				ncookie.lval[0];
3273 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3274 				ncookie.lval[1];
3275 			}
3276 			more_dirs = fxdr_unsigned(int, *tl);
3277 		}
3278 		/*
3279 		 * If at end of rpc data, get the eof boolean
3280 		 */
3281 		if (!more_dirs) {
3282 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3283 			eof = fxdr_unsigned(int, *tl);
3284 			if (tryformoredirs)
3285 				more_dirs = !eof;
3286 			if (nd->nd_flag & ND_NFSV4) {
3287 				error = nfscl_postop_attr(nd, nap, attrflagp,
3288 				    stuff);
3289 				if (error)
3290 					goto nfsmout;
3291 			}
3292 		}
3293 		m_freem(nd->nd_mrep);
3294 		nd->nd_mrep = NULL;
3295 	}
3296 	/*
3297 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3298 	 * by increasing d_reclen for the last record.
3299 	 */
3300 	if (blksiz > 0) {
3301 		left = DIRBLKSIZ - blksiz;
3302 		NFSBZERO(uiop->uio_iov->iov_base, left);
3303 		dp->d_reclen += left;
3304 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3305 		    left;
3306 		uiop->uio_iov->iov_len -= left;
3307 		uiop->uio_resid -= left;
3308 		uiop->uio_offset += left;
3309 	}
3310 
3311 	/*
3312 	 * If returning no data, assume end of file.
3313 	 * If not bigenough, return not end of file, since you aren't
3314 	 *    returning all the data
3315 	 * Otherwise, return the eof flag from the server.
3316 	 */
3317 	if (eofp) {
3318 		if (tresid == ((size_t)(uiop->uio_resid)))
3319 			*eofp = 1;
3320 		else if (!bigenough)
3321 			*eofp = 0;
3322 		else
3323 			*eofp = eof;
3324 	}
3325 
3326 	/*
3327 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3328 	 */
3329 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3330 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3331 		NFSBZERO(dp, DIRBLKSIZ);
3332 		dp->d_type = DT_UNKNOWN;
3333 		tl = (u_int32_t *)&dp->d_name[4];
3334 		*tl++ = cookie.lval[0];
3335 		*tl = cookie.lval[1];
3336 		dp->d_reclen = DIRBLKSIZ;
3337 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3338 		    DIRBLKSIZ;
3339 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3340 		uiop->uio_resid -= DIRBLKSIZ;
3341 		uiop->uio_offset += DIRBLKSIZ;
3342 	}
3343 
3344 nfsmout:
3345 	if (nd->nd_mrep != NULL)
3346 		m_freem(nd->nd_mrep);
3347 	return (error);
3348 }
3349 
3350 #ifndef APPLE
3351 /*
3352  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3353  * (Also used for NFS V4 when mount flag set.)
3354  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3355  */
3356 APPLESTATIC int
3357 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3358     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3359     int *eofp, void *stuff)
3360 {
3361 	int len, left;
3362 	struct dirent *dp = NULL;
3363 	u_int32_t *tl;
3364 	vnode_t newvp = NULLVP;
3365 	struct nfsrv_descript nfsd, *nd = &nfsd;
3366 	struct nameidata nami, *ndp = &nami;
3367 	struct componentname *cnp = &ndp->ni_cnd;
3368 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3369 	struct nfsnode *dnp = VTONFS(vp), *np;
3370 	struct nfsvattr nfsva;
3371 	struct nfsfh *nfhp;
3372 	nfsquad_t cookie, ncookie;
3373 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3374 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3375 	int isdotdot = 0, unlocknewvp = 0;
3376 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3377 	u_int64_t fileno = 0;
3378 	char *cp;
3379 	nfsattrbit_t attrbits, dattrbits;
3380 	size_t tresid;
3381 	u_int32_t *tl2 = NULL, rderr;
3382 	struct timespec dctime;
3383 
3384 	KASSERT(uiop->uio_iovcnt == 1 &&
3385 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3386 	    ("nfs readdirplusrpc bad uio"));
3387 	ncookie.lval[0] = ncookie.lval[1] = 0;
3388 	timespecclear(&dctime);
3389 	*attrflagp = 0;
3390 	if (eofp != NULL)
3391 		*eofp = 0;
3392 	ndp->ni_dvp = vp;
3393 	nd->nd_mrep = NULL;
3394 	cookie.lval[0] = cookiep->nfsuquad[0];
3395 	cookie.lval[1] = cookiep->nfsuquad[1];
3396 	tresid = uiop->uio_resid;
3397 
3398 	/*
3399 	 * For NFSv4, first create the "." and ".." entries.
3400 	 */
3401 	if (NFSHASNFSV4(nmp)) {
3402 		NFSGETATTR_ATTRBIT(&dattrbits);
3403 		NFSZERO_ATTRBIT(&attrbits);
3404 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3405 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3406 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3407 			NFSSETBIT_ATTRBIT(&attrbits,
3408 			    NFSATTRBIT_MOUNTEDONFILEID);
3409 			gotmnton = 1;
3410 		} else {
3411 			/*
3412 			 * Must fake it. Use the fileno, except when the
3413 			 * fsid is != to that of the directory. For that
3414 			 * case, generate a fake fileno that is not the same.
3415 			 */
3416 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3417 			gotmnton = 0;
3418 		}
3419 
3420 		/*
3421 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3422 		 */
3423 		if (uiop->uio_offset == 0) {
3424 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3425 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3426 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3427 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3428 			(void) nfsrv_putattrbit(nd, &attrbits);
3429 			error = nfscl_request(nd, vp, p, cred, stuff);
3430 			if (error)
3431 			    return (error);
3432 			dotfileid = 0;	/* Fake out the compiler. */
3433 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3434 			    error = nfsm_loadattr(nd, &nfsva);
3435 			    if (error != 0)
3436 				goto nfsmout;
3437 			    dctime = nfsva.na_ctime;
3438 			    dotfileid = nfsva.na_fileid;
3439 			}
3440 			if (nd->nd_repstat == 0) {
3441 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3442 			    len = fxdr_unsigned(int, *(tl + 4));
3443 			    if (len > 0 && len <= NFSX_V4FHMAX)
3444 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3445 			    else
3446 				error = EPERM;
3447 			    if (!error) {
3448 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3449 				nfsva.na_mntonfileno = UINT64_MAX;
3450 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3451 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3452 				    NULL, NULL, NULL, p, cred);
3453 				if (error) {
3454 				    dotdotfileid = dotfileid;
3455 				} else if (gotmnton) {
3456 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3457 					dotdotfileid = nfsva.na_mntonfileno;
3458 				    else
3459 					dotdotfileid = nfsva.na_fileid;
3460 				} else if (nfsva.na_filesid[0] ==
3461 				    dnp->n_vattr.na_filesid[0] &&
3462 				    nfsva.na_filesid[1] ==
3463 				    dnp->n_vattr.na_filesid[1]) {
3464 				    dotdotfileid = nfsva.na_fileid;
3465 				} else {
3466 				    do {
3467 					fakefileno--;
3468 				    } while (fakefileno ==
3469 					nfsva.na_fileid);
3470 				    dotdotfileid = fakefileno;
3471 				}
3472 			    }
3473 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3474 			    /*
3475 			     * Lookupp returns NFSERR_NOENT when we are
3476 			     * at the root, so just use the current dir.
3477 			     */
3478 			    nd->nd_repstat = 0;
3479 			    dotdotfileid = dotfileid;
3480 			} else {
3481 			    error = nd->nd_repstat;
3482 			}
3483 			m_freem(nd->nd_mrep);
3484 			if (error)
3485 			    return (error);
3486 			nd->nd_mrep = NULL;
3487 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3488 			dp->d_pad0 = dp->d_pad1 = 0;
3489 			dp->d_off = 0;
3490 			dp->d_type = DT_DIR;
3491 			dp->d_fileno = dotfileid;
3492 			dp->d_namlen = 1;
3493 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3494 			dp->d_name[0] = '.';
3495 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3496 			/*
3497 			 * Just make these offset cookie 0.
3498 			 */
3499 			tl = (u_int32_t *)&dp->d_name[8];
3500 			*tl++ = 0;
3501 			*tl = 0;
3502 			blksiz += dp->d_reclen;
3503 			uiop->uio_resid -= dp->d_reclen;
3504 			uiop->uio_offset += dp->d_reclen;
3505 			uiop->uio_iov->iov_base =
3506 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3507 			uiop->uio_iov->iov_len -= dp->d_reclen;
3508 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3509 			dp->d_pad0 = dp->d_pad1 = 0;
3510 			dp->d_off = 0;
3511 			dp->d_type = DT_DIR;
3512 			dp->d_fileno = dotdotfileid;
3513 			dp->d_namlen = 2;
3514 			*((uint64_t *)dp->d_name) = 0;
3515 			dp->d_name[0] = '.';
3516 			dp->d_name[1] = '.';
3517 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3518 			/*
3519 			 * Just make these offset cookie 0.
3520 			 */
3521 			tl = (u_int32_t *)&dp->d_name[8];
3522 			*tl++ = 0;
3523 			*tl = 0;
3524 			blksiz += dp->d_reclen;
3525 			uiop->uio_resid -= dp->d_reclen;
3526 			uiop->uio_offset += dp->d_reclen;
3527 			uiop->uio_iov->iov_base =
3528 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3529 			uiop->uio_iov->iov_len -= dp->d_reclen;
3530 		}
3531 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3532 		if (gotmnton)
3533 			NFSSETBIT_ATTRBIT(&attrbits,
3534 			    NFSATTRBIT_MOUNTEDONFILEID);
3535 	}
3536 
3537 	/*
3538 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3539 	 * The stopping criteria is EOF or buffer full.
3540 	 */
3541 	while (more_dirs && bigenough) {
3542 		*attrflagp = 0;
3543 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3544  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3545 		*tl++ = cookie.lval[0];
3546 		*tl++ = cookie.lval[1];
3547 		if (cookie.qval == 0) {
3548 			*tl++ = 0;
3549 			*tl++ = 0;
3550 		} else {
3551 			NFSLOCKNODE(dnp);
3552 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3553 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3554 			NFSUNLOCKNODE(dnp);
3555 		}
3556 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3557 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3558 		if (nd->nd_flag & ND_NFSV4) {
3559 			(void) nfsrv_putattrbit(nd, &attrbits);
3560 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3561 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3562 			(void) nfsrv_putattrbit(nd, &dattrbits);
3563 		}
3564 		error = nfscl_request(nd, vp, p, cred, stuff);
3565 		if (error)
3566 			return (error);
3567 		if (nd->nd_flag & ND_NFSV3)
3568 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3569 		if (nd->nd_repstat || error) {
3570 			if (!error)
3571 				error = nd->nd_repstat;
3572 			goto nfsmout;
3573 		}
3574 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3575 			dctime = nap->na_ctime;
3576 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3577 		NFSLOCKNODE(dnp);
3578 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3579 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3580 		NFSUNLOCKNODE(dnp);
3581 		more_dirs = fxdr_unsigned(int, *tl);
3582 		if (!more_dirs)
3583 			tryformoredirs = 0;
3584 
3585 		/* loop through the dir entries, doctoring them to 4bsd form */
3586 		while (more_dirs && bigenough) {
3587 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3588 			if (nd->nd_flag & ND_NFSV4) {
3589 				ncookie.lval[0] = *tl++;
3590 				ncookie.lval[1] = *tl++;
3591 			} else {
3592 				fileno = fxdr_hyper(tl);
3593 				tl += 2;
3594 			}
3595 			len = fxdr_unsigned(int, *tl);
3596 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3597 				error = EBADRPC;
3598 				goto nfsmout;
3599 			}
3600 			tlen = roundup2(len, 8);
3601 			if (tlen == len)
3602 				tlen += 8;  /* To ensure null termination. */
3603 			left = DIRBLKSIZ - blksiz;
3604 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3605 				NFSBZERO(uiop->uio_iov->iov_base, left);
3606 				dp->d_reclen += left;
3607 				uiop->uio_iov->iov_base =
3608 				    (char *)uiop->uio_iov->iov_base + left;
3609 				uiop->uio_iov->iov_len -= left;
3610 				uiop->uio_resid -= left;
3611 				uiop->uio_offset += left;
3612 				blksiz = 0;
3613 			}
3614 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3615 			    uiop->uio_resid)
3616 				bigenough = 0;
3617 			if (bigenough) {
3618 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3619 				dp->d_pad0 = dp->d_pad1 = 0;
3620 				dp->d_off = 0;
3621 				dp->d_namlen = len;
3622 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3623 				    NFSX_HYPER;
3624 				dp->d_type = DT_UNKNOWN;
3625 				blksiz += dp->d_reclen;
3626 				if (blksiz == DIRBLKSIZ)
3627 					blksiz = 0;
3628 				uiop->uio_resid -= DIRHDSIZ;
3629 				uiop->uio_offset += DIRHDSIZ;
3630 				uiop->uio_iov->iov_base =
3631 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3632 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3633 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3634 				cnp->cn_namelen = len;
3635 				NFSCNHASHZERO(cnp);
3636 				error = nfsm_mbufuio(nd, uiop, len);
3637 				if (error)
3638 					goto nfsmout;
3639 				cp = uiop->uio_iov->iov_base;
3640 				tlen -= len;
3641 				NFSBZERO(cp, tlen);
3642 				cp += tlen;	/* points to cookie storage */
3643 				tl2 = (u_int32_t *)cp;
3644 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3645 				    cnp->cn_nameptr[1] == '.')
3646 					isdotdot = 1;
3647 				else
3648 					isdotdot = 0;
3649 				uiop->uio_iov->iov_base =
3650 				    (char *)uiop->uio_iov->iov_base + tlen +
3651 				    NFSX_HYPER;
3652 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3653 				uiop->uio_resid -= tlen + NFSX_HYPER;
3654 				uiop->uio_offset += (tlen + NFSX_HYPER);
3655 			} else {
3656 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3657 				if (error)
3658 					goto nfsmout;
3659 			}
3660 			nfhp = NULL;
3661 			if (nd->nd_flag & ND_NFSV3) {
3662 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3663 				ncookie.lval[0] = *tl++;
3664 				ncookie.lval[1] = *tl++;
3665 				attrflag = fxdr_unsigned(int, *tl);
3666 				if (attrflag) {
3667 				  error = nfsm_loadattr(nd, &nfsva);
3668 				  if (error)
3669 					goto nfsmout;
3670 				}
3671 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3672 				if (*tl) {
3673 					error = nfsm_getfh(nd, &nfhp);
3674 					if (error)
3675 					    goto nfsmout;
3676 				}
3677 				if (!attrflag && nfhp != NULL) {
3678 					free(nfhp, M_NFSFH);
3679 					nfhp = NULL;
3680 				}
3681 			} else {
3682 				rderr = 0;
3683 				nfsva.na_mntonfileno = 0xffffffff;
3684 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3685 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3686 				    NULL, NULL, &rderr, p, cred);
3687 				if (error)
3688 					goto nfsmout;
3689 			}
3690 
3691 			if (bigenough) {
3692 			    if (nd->nd_flag & ND_NFSV4) {
3693 				if (rderr) {
3694 				    dp->d_fileno = 0;
3695 				} else if (gotmnton) {
3696 				    if (nfsva.na_mntonfileno != 0xffffffff)
3697 					dp->d_fileno = nfsva.na_mntonfileno;
3698 				    else
3699 					dp->d_fileno = nfsva.na_fileid;
3700 				} else if (nfsva.na_filesid[0] ==
3701 				    dnp->n_vattr.na_filesid[0] &&
3702 				    nfsva.na_filesid[1] ==
3703 				    dnp->n_vattr.na_filesid[1]) {
3704 				    dp->d_fileno = nfsva.na_fileid;
3705 				} else {
3706 				    do {
3707 					fakefileno--;
3708 				    } while (fakefileno ==
3709 					nfsva.na_fileid);
3710 				    dp->d_fileno = fakefileno;
3711 				}
3712 			    } else {
3713 				dp->d_fileno = fileno;
3714 			    }
3715 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3716 				ncookie.lval[0];
3717 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3718 				ncookie.lval[1];
3719 
3720 			    if (nfhp != NULL) {
3721 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3722 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3723 				    VREF(vp);
3724 				    newvp = vp;
3725 				    unlocknewvp = 0;
3726 				    free(nfhp, M_NFSFH);
3727 				    np = dnp;
3728 				} else if (isdotdot != 0) {
3729 				    /*
3730 				     * Skip doing a nfscl_nget() call for "..".
3731 				     * There's a race between acquiring the nfs
3732 				     * node here and lookups that look for the
3733 				     * directory being read (in the parent).
3734 				     * It would try to get a lock on ".." here,
3735 				     * owning the lock on the directory being
3736 				     * read. Lookup will hold the lock on ".."
3737 				     * and try to acquire the lock on the
3738 				     * directory being read.
3739 				     * If the directory is unlocked/relocked,
3740 				     * then there is a LOR with the buflock
3741 				     * vp is relocked.
3742 				     */
3743 				    free(nfhp, M_NFSFH);
3744 				} else {
3745 				    error = nfscl_nget(vnode_mount(vp), vp,
3746 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3747 				    if (!error) {
3748 					newvp = NFSTOV(np);
3749 					unlocknewvp = 1;
3750 				    }
3751 				}
3752 				nfhp = NULL;
3753 				if (newvp != NULLVP) {
3754 				    error = nfscl_loadattrcache(&newvp,
3755 					&nfsva, NULL, NULL, 0, 0);
3756 				    if (error) {
3757 					if (unlocknewvp)
3758 					    vput(newvp);
3759 					else
3760 					    vrele(newvp);
3761 					goto nfsmout;
3762 				    }
3763 				    dp->d_type =
3764 					vtonfs_dtype(np->n_vattr.na_type);
3765 				    ndp->ni_vp = newvp;
3766 				    NFSCNHASH(cnp, HASHINIT);
3767 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3768 					(newvp->v_type != VDIR ||
3769 					 dctime.tv_sec != 0)) {
3770 					cache_enter_time(ndp->ni_dvp,
3771 					    ndp->ni_vp, cnp,
3772 					    &nfsva.na_ctime,
3773 					    newvp->v_type != VDIR ? NULL :
3774 					    &dctime);
3775 				    }
3776 				    if (unlocknewvp)
3777 					vput(newvp);
3778 				    else
3779 					vrele(newvp);
3780 				    newvp = NULLVP;
3781 				}
3782 			    }
3783 			} else if (nfhp != NULL) {
3784 			    free(nfhp, M_NFSFH);
3785 			}
3786 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3787 			more_dirs = fxdr_unsigned(int, *tl);
3788 		}
3789 		/*
3790 		 * If at end of rpc data, get the eof boolean
3791 		 */
3792 		if (!more_dirs) {
3793 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3794 			eof = fxdr_unsigned(int, *tl);
3795 			if (tryformoredirs)
3796 				more_dirs = !eof;
3797 			if (nd->nd_flag & ND_NFSV4) {
3798 				error = nfscl_postop_attr(nd, nap, attrflagp,
3799 				    stuff);
3800 				if (error)
3801 					goto nfsmout;
3802 			}
3803 		}
3804 		m_freem(nd->nd_mrep);
3805 		nd->nd_mrep = NULL;
3806 	}
3807 	/*
3808 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3809 	 * by increasing d_reclen for the last record.
3810 	 */
3811 	if (blksiz > 0) {
3812 		left = DIRBLKSIZ - blksiz;
3813 		NFSBZERO(uiop->uio_iov->iov_base, left);
3814 		dp->d_reclen += left;
3815 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3816 		    left;
3817 		uiop->uio_iov->iov_len -= left;
3818 		uiop->uio_resid -= left;
3819 		uiop->uio_offset += left;
3820 	}
3821 
3822 	/*
3823 	 * If returning no data, assume end of file.
3824 	 * If not bigenough, return not end of file, since you aren't
3825 	 *    returning all the data
3826 	 * Otherwise, return the eof flag from the server.
3827 	 */
3828 	if (eofp != NULL) {
3829 		if (tresid == uiop->uio_resid)
3830 			*eofp = 1;
3831 		else if (!bigenough)
3832 			*eofp = 0;
3833 		else
3834 			*eofp = eof;
3835 	}
3836 
3837 	/*
3838 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3839 	 */
3840 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3841 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3842 		NFSBZERO(dp, DIRBLKSIZ);
3843 		dp->d_type = DT_UNKNOWN;
3844 		tl = (u_int32_t *)&dp->d_name[4];
3845 		*tl++ = cookie.lval[0];
3846 		*tl = cookie.lval[1];
3847 		dp->d_reclen = DIRBLKSIZ;
3848 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3849 		    DIRBLKSIZ;
3850 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3851 		uiop->uio_resid -= DIRBLKSIZ;
3852 		uiop->uio_offset += DIRBLKSIZ;
3853 	}
3854 
3855 nfsmout:
3856 	if (nd->nd_mrep != NULL)
3857 		m_freem(nd->nd_mrep);
3858 	return (error);
3859 }
3860 #endif	/* !APPLE */
3861 
3862 /*
3863  * Nfs commit rpc
3864  */
3865 APPLESTATIC int
3866 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3867     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3868 {
3869 	u_int32_t *tl;
3870 	struct nfsrv_descript nfsd, *nd = &nfsd;
3871 	nfsattrbit_t attrbits;
3872 	int error;
3873 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3874 
3875 	*attrflagp = 0;
3876 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3877 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3878 	txdr_hyper(offset, tl);
3879 	tl += 2;
3880 	*tl = txdr_unsigned(cnt);
3881 	if (nd->nd_flag & ND_NFSV4) {
3882 		/*
3883 		 * And do a Getattr op.
3884 		 */
3885 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3886 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3887 		NFSGETATTR_ATTRBIT(&attrbits);
3888 		(void) nfsrv_putattrbit(nd, &attrbits);
3889 	}
3890 	error = nfscl_request(nd, vp, p, cred, stuff);
3891 	if (error)
3892 		return (error);
3893 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3894 	if (!error && !nd->nd_repstat) {
3895 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3896 		NFSLOCKMNT(nmp);
3897 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3898 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3899 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3900 		}
3901 		NFSUNLOCKMNT(nmp);
3902 		if (nd->nd_flag & ND_NFSV4)
3903 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3904 	}
3905 nfsmout:
3906 	if (!error && nd->nd_repstat)
3907 		error = nd->nd_repstat;
3908 	m_freem(nd->nd_mrep);
3909 	return (error);
3910 }
3911 
3912 /*
3913  * NFS byte range lock rpc.
3914  * (Mostly just calls one of the three lower level RPC routines.)
3915  */
3916 APPLESTATIC int
3917 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3918     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3919 {
3920 	struct nfscllockowner *lp;
3921 	struct nfsclclient *clp;
3922 	struct nfsfh *nfhp;
3923 	struct nfsrv_descript nfsd, *nd = &nfsd;
3924 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3925 	u_int64_t off, len;
3926 	off_t start, end;
3927 	u_int32_t clidrev = 0;
3928 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3929 	int callcnt, dorpc;
3930 
3931 	/*
3932 	 * Convert the flock structure into a start and end and do POSIX
3933 	 * bounds checking.
3934 	 */
3935 	switch (fl->l_whence) {
3936 	case SEEK_SET:
3937 	case SEEK_CUR:
3938 		/*
3939 		 * Caller is responsible for adding any necessary offset
3940 		 * when SEEK_CUR is used.
3941 		 */
3942 		start = fl->l_start;
3943 		off = fl->l_start;
3944 		break;
3945 	case SEEK_END:
3946 		start = size + fl->l_start;
3947 		off = size + fl->l_start;
3948 		break;
3949 	default:
3950 		return (EINVAL);
3951 	}
3952 	if (start < 0)
3953 		return (EINVAL);
3954 	if (fl->l_len != 0) {
3955 		end = start + fl->l_len - 1;
3956 		if (end < start)
3957 			return (EINVAL);
3958 	}
3959 
3960 	len = fl->l_len;
3961 	if (len == 0)
3962 		len = NFS64BITSSET;
3963 	retrycnt = 0;
3964 	do {
3965 	    nd->nd_repstat = 0;
3966 	    if (op == F_GETLK) {
3967 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3968 		if (error)
3969 			return (error);
3970 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3971 		if (!error) {
3972 			clidrev = clp->nfsc_clientidrev;
3973 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3974 			    p, id, flags);
3975 		} else if (error == -1) {
3976 			error = 0;
3977 		}
3978 		nfscl_clientrelease(clp);
3979 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3980 		/*
3981 		 * We must loop around for all lockowner cases.
3982 		 */
3983 		callcnt = 0;
3984 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3985 		if (error)
3986 			return (error);
3987 		do {
3988 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3989 			clp, id, flags, &lp, &dorpc);
3990 		    /*
3991 		     * If it returns a NULL lp, we're done.
3992 		     */
3993 		    if (lp == NULL) {
3994 			if (callcnt == 0)
3995 			    nfscl_clientrelease(clp);
3996 			else
3997 			    nfscl_releasealllocks(clp, vp, p, id, flags);
3998 			return (error);
3999 		    }
4000 		    if (nmp->nm_clp != NULL)
4001 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4002 		    else
4003 			clidrev = 0;
4004 		    /*
4005 		     * If the server doesn't support Posix lock semantics,
4006 		     * only allow locks on the entire file, since it won't
4007 		     * handle overlapping byte ranges.
4008 		     * There might still be a problem when a lock
4009 		     * upgrade/downgrade (read<->write) occurs, since the
4010 		     * server "might" expect an unlock first?
4011 		     */
4012 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4013 			(off == 0 && len == NFS64BITSSET))) {
4014 			/*
4015 			 * Since the lock records will go away, we must
4016 			 * wait for grace and delay here.
4017 			 */
4018 			do {
4019 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4020 				NFSV4LOCKT_READ, cred, p, 0);
4021 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4022 				 nd->nd_repstat == NFSERR_DELAY) &&
4023 				error == 0)
4024 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4025 				    "nfs_advlock");
4026 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4027 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4028 		    }
4029 		    callcnt++;
4030 		} while (error == 0 && nd->nd_repstat == 0);
4031 		nfscl_releasealllocks(clp, vp, p, id, flags);
4032 	    } else if (op == F_SETLK) {
4033 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4034 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4035 		if (error || donelocally) {
4036 			return (error);
4037 		}
4038 		if (nmp->nm_clp != NULL)
4039 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4040 		else
4041 			clidrev = 0;
4042 		nfhp = VTONFS(vp)->n_fhp;
4043 		if (!lp->nfsl_open->nfso_posixlock &&
4044 		    (off != 0 || len != NFS64BITSSET)) {
4045 			error = EINVAL;
4046 		} else {
4047 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4048 			    nfhp->nfh_len, lp, newone, reclaim, off,
4049 			    len, fl->l_type, cred, p, 0);
4050 		}
4051 		if (!error)
4052 			error = nd->nd_repstat;
4053 		nfscl_lockrelease(lp, error, newone);
4054 	    } else {
4055 		error = EINVAL;
4056 	    }
4057 	    if (!error)
4058 	        error = nd->nd_repstat;
4059 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4060 		error == NFSERR_STALEDONTRECOVER ||
4061 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4062 		error == NFSERR_BADSESSION) {
4063 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4064 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4065 		&& clidrev != 0) {
4066 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4067 		retrycnt++;
4068 	    }
4069 	} while (error == NFSERR_GRACE ||
4070 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4071 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4072 	    error == NFSERR_BADSESSION ||
4073 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4074 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4075 	if (error && retrycnt >= 4)
4076 		error = EIO;
4077 	return (error);
4078 }
4079 
4080 /*
4081  * The lower level routine for the LockT case.
4082  */
4083 APPLESTATIC int
4084 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4085     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4086     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4087 {
4088 	u_int32_t *tl;
4089 	int error, type, size;
4090 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4091 	struct nfsnode *np;
4092 	struct nfsmount *nmp;
4093 	struct nfsclsession *tsep;
4094 
4095 	nmp = VFSTONFS(vp->v_mount);
4096 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4097 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4098 	if (fl->l_type == F_RDLCK)
4099 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4100 	else
4101 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4102 	txdr_hyper(off, tl);
4103 	tl += 2;
4104 	txdr_hyper(len, tl);
4105 	tl += 2;
4106 	tsep = nfsmnt_mdssession(nmp);
4107 	*tl++ = tsep->nfsess_clientid.lval[0];
4108 	*tl = tsep->nfsess_clientid.lval[1];
4109 	nfscl_filllockowner(id, own, flags);
4110 	np = VTONFS(vp);
4111 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4112 	    np->n_fhp->nfh_len);
4113 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4114 	error = nfscl_request(nd, vp, p, cred, NULL);
4115 	if (error)
4116 		return (error);
4117 	if (nd->nd_repstat == 0) {
4118 		fl->l_type = F_UNLCK;
4119 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4120 		nd->nd_repstat = 0;
4121 		fl->l_whence = SEEK_SET;
4122 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4123 		fl->l_start = fxdr_hyper(tl);
4124 		tl += 2;
4125 		len = fxdr_hyper(tl);
4126 		tl += 2;
4127 		if (len == NFS64BITSSET)
4128 			fl->l_len = 0;
4129 		else
4130 			fl->l_len = len;
4131 		type = fxdr_unsigned(int, *tl++);
4132 		if (type == NFSV4LOCKT_WRITE)
4133 			fl->l_type = F_WRLCK;
4134 		else
4135 			fl->l_type = F_RDLCK;
4136 		/*
4137 		 * XXX For now, I have no idea what to do with the
4138 		 * conflicting lock_owner, so I'll just set the pid == 0
4139 		 * and skip over the lock_owner.
4140 		 */
4141 		fl->l_pid = (pid_t)0;
4142 		tl += 2;
4143 		size = fxdr_unsigned(int, *tl);
4144 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4145 			error = EBADRPC;
4146 		if (!error)
4147 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4148 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4149 		nfscl_initiate_recovery(clp);
4150 nfsmout:
4151 	m_freem(nd->nd_mrep);
4152 	return (error);
4153 }
4154 
4155 /*
4156  * Lower level function that performs the LockU RPC.
4157  */
4158 static int
4159 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4160     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4161     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4162 {
4163 	u_int32_t *tl;
4164 	int error;
4165 
4166 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4167 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4168 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4169 	*tl++ = txdr_unsigned(type);
4170 	*tl = txdr_unsigned(lp->nfsl_seqid);
4171 	if (nfstest_outofseq &&
4172 	    (arc4random() % nfstest_outofseq) == 0)
4173 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4174 	tl++;
4175 	if (NFSHASNFSV4N(nmp))
4176 		*tl++ = 0;
4177 	else
4178 		*tl++ = lp->nfsl_stateid.seqid;
4179 	*tl++ = lp->nfsl_stateid.other[0];
4180 	*tl++ = lp->nfsl_stateid.other[1];
4181 	*tl++ = lp->nfsl_stateid.other[2];
4182 	txdr_hyper(off, tl);
4183 	tl += 2;
4184 	txdr_hyper(len, tl);
4185 	if (syscred)
4186 		nd->nd_flag |= ND_USEGSSNAME;
4187 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4188 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4189 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4190 	if (error)
4191 		return (error);
4192 	if (nd->nd_repstat == 0) {
4193 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4194 		lp->nfsl_stateid.seqid = *tl++;
4195 		lp->nfsl_stateid.other[0] = *tl++;
4196 		lp->nfsl_stateid.other[1] = *tl++;
4197 		lp->nfsl_stateid.other[2] = *tl;
4198 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4199 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4200 nfsmout:
4201 	m_freem(nd->nd_mrep);
4202 	return (error);
4203 }
4204 
4205 /*
4206  * The actual Lock RPC.
4207  */
4208 APPLESTATIC int
4209 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4210     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4211     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4212     NFSPROC_T *p, int syscred)
4213 {
4214 	u_int32_t *tl;
4215 	int error, size;
4216 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4217 	struct nfsclsession *tsep;
4218 
4219 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4220 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4221 	if (type == F_RDLCK)
4222 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4223 	else
4224 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4225 	*tl++ = txdr_unsigned(reclaim);
4226 	txdr_hyper(off, tl);
4227 	tl += 2;
4228 	txdr_hyper(len, tl);
4229 	tl += 2;
4230 	if (newone) {
4231 	    *tl = newnfs_true;
4232 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4233 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4234 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4235 	    if (NFSHASNFSV4N(nmp))
4236 		*tl++ = 0;
4237 	    else
4238 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4239 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4240 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4241 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4242 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4243 	    tsep = nfsmnt_mdssession(nmp);
4244 	    *tl++ = tsep->nfsess_clientid.lval[0];
4245 	    *tl = tsep->nfsess_clientid.lval[1];
4246 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4247 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4248 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4249 	} else {
4250 	    *tl = newnfs_false;
4251 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4252 	    if (NFSHASNFSV4N(nmp))
4253 		*tl++ = 0;
4254 	    else
4255 		*tl++ = lp->nfsl_stateid.seqid;
4256 	    *tl++ = lp->nfsl_stateid.other[0];
4257 	    *tl++ = lp->nfsl_stateid.other[1];
4258 	    *tl++ = lp->nfsl_stateid.other[2];
4259 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4260 	    if (nfstest_outofseq &&
4261 		(arc4random() % nfstest_outofseq) == 0)
4262 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4263 	}
4264 	if (syscred)
4265 		nd->nd_flag |= ND_USEGSSNAME;
4266 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4267 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4268 	if (error)
4269 		return (error);
4270 	if (newone)
4271 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4272 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4273 	if (nd->nd_repstat == 0) {
4274 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4275 		lp->nfsl_stateid.seqid = *tl++;
4276 		lp->nfsl_stateid.other[0] = *tl++;
4277 		lp->nfsl_stateid.other[1] = *tl++;
4278 		lp->nfsl_stateid.other[2] = *tl;
4279 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4280 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4281 		size = fxdr_unsigned(int, *(tl + 7));
4282 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4283 			error = EBADRPC;
4284 		if (!error)
4285 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4286 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4287 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4288 nfsmout:
4289 	m_freem(nd->nd_mrep);
4290 	return (error);
4291 }
4292 
4293 /*
4294  * nfs statfs rpc
4295  * (always called with the vp for the mount point)
4296  */
4297 APPLESTATIC int
4298 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4299     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4300     void *stuff)
4301 {
4302 	u_int32_t *tl = NULL;
4303 	struct nfsrv_descript nfsd, *nd = &nfsd;
4304 	struct nfsmount *nmp;
4305 	nfsattrbit_t attrbits;
4306 	int error;
4307 
4308 	*attrflagp = 0;
4309 	nmp = VFSTONFS(vnode_mount(vp));
4310 	if (NFSHASNFSV4(nmp)) {
4311 		/*
4312 		 * For V4, you actually do a getattr.
4313 		 */
4314 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4315 		NFSSTATFS_GETATTRBIT(&attrbits);
4316 		(void) nfsrv_putattrbit(nd, &attrbits);
4317 		nd->nd_flag |= ND_USEGSSNAME;
4318 		error = nfscl_request(nd, vp, p, cred, stuff);
4319 		if (error)
4320 			return (error);
4321 		if (nd->nd_repstat == 0) {
4322 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4323 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4324 			    cred);
4325 			if (!error) {
4326 				nmp->nm_fsid[0] = nap->na_filesid[0];
4327 				nmp->nm_fsid[1] = nap->na_filesid[1];
4328 				NFSSETHASSETFSID(nmp);
4329 				*attrflagp = 1;
4330 			}
4331 		} else {
4332 			error = nd->nd_repstat;
4333 		}
4334 		if (error)
4335 			goto nfsmout;
4336 	} else {
4337 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4338 		error = nfscl_request(nd, vp, p, cred, stuff);
4339 		if (error)
4340 			return (error);
4341 		if (nd->nd_flag & ND_NFSV3) {
4342 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4343 			if (error)
4344 				goto nfsmout;
4345 		}
4346 		if (nd->nd_repstat) {
4347 			error = nd->nd_repstat;
4348 			goto nfsmout;
4349 		}
4350 		NFSM_DISSECT(tl, u_int32_t *,
4351 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4352 	}
4353 	if (NFSHASNFSV3(nmp)) {
4354 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4355 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4356 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4357 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4358 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4359 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4360 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4361 	} else if (NFSHASNFSV4(nmp) == 0) {
4362 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4363 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4364 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4365 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4366 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4367 	}
4368 nfsmout:
4369 	m_freem(nd->nd_mrep);
4370 	return (error);
4371 }
4372 
4373 /*
4374  * nfs pathconf rpc
4375  */
4376 APPLESTATIC int
4377 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4378     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4379     void *stuff)
4380 {
4381 	struct nfsrv_descript nfsd, *nd = &nfsd;
4382 	struct nfsmount *nmp;
4383 	u_int32_t *tl;
4384 	nfsattrbit_t attrbits;
4385 	int error;
4386 
4387 	*attrflagp = 0;
4388 	nmp = VFSTONFS(vnode_mount(vp));
4389 	if (NFSHASNFSV4(nmp)) {
4390 		/*
4391 		 * For V4, you actually do a getattr.
4392 		 */
4393 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4394 		NFSPATHCONF_GETATTRBIT(&attrbits);
4395 		(void) nfsrv_putattrbit(nd, &attrbits);
4396 		nd->nd_flag |= ND_USEGSSNAME;
4397 		error = nfscl_request(nd, vp, p, cred, stuff);
4398 		if (error)
4399 			return (error);
4400 		if (nd->nd_repstat == 0) {
4401 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4402 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4403 			    cred);
4404 			if (!error)
4405 				*attrflagp = 1;
4406 		} else {
4407 			error = nd->nd_repstat;
4408 		}
4409 	} else {
4410 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4411 		error = nfscl_request(nd, vp, p, cred, stuff);
4412 		if (error)
4413 			return (error);
4414 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4415 		if (nd->nd_repstat && !error)
4416 			error = nd->nd_repstat;
4417 		if (!error) {
4418 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4419 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4420 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4421 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4422 			pc->pc_chownrestricted =
4423 			    fxdr_unsigned(u_int32_t, *tl++);
4424 			pc->pc_caseinsensitive =
4425 			    fxdr_unsigned(u_int32_t, *tl++);
4426 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4427 		}
4428 	}
4429 nfsmout:
4430 	m_freem(nd->nd_mrep);
4431 	return (error);
4432 }
4433 
4434 /*
4435  * nfs version 3 fsinfo rpc call
4436  */
4437 APPLESTATIC int
4438 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4439     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4440 {
4441 	u_int32_t *tl;
4442 	struct nfsrv_descript nfsd, *nd = &nfsd;
4443 	int error;
4444 
4445 	*attrflagp = 0;
4446 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4447 	error = nfscl_request(nd, vp, p, cred, stuff);
4448 	if (error)
4449 		return (error);
4450 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4451 	if (nd->nd_repstat && !error)
4452 		error = nd->nd_repstat;
4453 	if (!error) {
4454 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4455 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4456 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4457 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4458 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4459 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4460 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4461 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4462 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4463 		tl += 2;
4464 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4465 		tl += 2;
4466 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4467 	}
4468 nfsmout:
4469 	m_freem(nd->nd_mrep);
4470 	return (error);
4471 }
4472 
4473 /*
4474  * This function performs the Renew RPC.
4475  */
4476 APPLESTATIC int
4477 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4478     NFSPROC_T *p)
4479 {
4480 	u_int32_t *tl;
4481 	struct nfsrv_descript nfsd;
4482 	struct nfsrv_descript *nd = &nfsd;
4483 	struct nfsmount *nmp;
4484 	int error;
4485 	struct nfssockreq *nrp;
4486 	struct nfsclsession *tsep;
4487 
4488 	nmp = clp->nfsc_nmp;
4489 	if (nmp == NULL)
4490 		return (0);
4491 	if (dsp == NULL)
4492 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4493 		    0);
4494 	else
4495 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4496 		    &dsp->nfsclds_sess, 0, 0);
4497 	if (!NFSHASNFSV4N(nmp)) {
4498 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4499 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4500 		tsep = nfsmnt_mdssession(nmp);
4501 		*tl++ = tsep->nfsess_clientid.lval[0];
4502 		*tl = tsep->nfsess_clientid.lval[1];
4503 	}
4504 	nrp = NULL;
4505 	if (dsp != NULL)
4506 		nrp = dsp->nfsclds_sockp;
4507 	if (nrp == NULL)
4508 		/* If NULL, use the MDS socket. */
4509 		nrp = &nmp->nm_sockreq;
4510 	nd->nd_flag |= ND_USEGSSNAME;
4511 	if (dsp == NULL)
4512 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4513 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4514 	else {
4515 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4516 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4517 		if (error == ENXIO)
4518 			nfscl_cancelreqs(dsp);
4519 	}
4520 	if (error)
4521 		return (error);
4522 	error = nd->nd_repstat;
4523 	m_freem(nd->nd_mrep);
4524 	return (error);
4525 }
4526 
4527 /*
4528  * This function performs the Releaselockowner RPC.
4529  */
4530 APPLESTATIC int
4531 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4532     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4533 {
4534 	struct nfsrv_descript nfsd, *nd = &nfsd;
4535 	u_int32_t *tl;
4536 	int error;
4537 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4538 	struct nfsclsession *tsep;
4539 
4540 	if (NFSHASNFSV4N(nmp)) {
4541 		/* For NFSv4.1, do a FreeStateID. */
4542 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4543 		    NULL, 0, 0);
4544 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4545 	} else {
4546 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4547 		    NULL, 0, 0);
4548 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4549 		tsep = nfsmnt_mdssession(nmp);
4550 		*tl++ = tsep->nfsess_clientid.lval[0];
4551 		*tl = tsep->nfsess_clientid.lval[1];
4552 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4553 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4554 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4555 	}
4556 	nd->nd_flag |= ND_USEGSSNAME;
4557 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4558 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4559 	if (error)
4560 		return (error);
4561 	error = nd->nd_repstat;
4562 	m_freem(nd->nd_mrep);
4563 	return (error);
4564 }
4565 
4566 /*
4567  * This function performs the Compound to get the mount pt FH.
4568  */
4569 APPLESTATIC int
4570 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4571     NFSPROC_T *p)
4572 {
4573 	u_int32_t *tl;
4574 	struct nfsrv_descript nfsd;
4575 	struct nfsrv_descript *nd = &nfsd;
4576 	u_char *cp, *cp2;
4577 	int error, cnt, len, setnil;
4578 	u_int32_t *opcntp;
4579 
4580 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4581 	    0);
4582 	cp = dirpath;
4583 	cnt = 0;
4584 	do {
4585 		setnil = 0;
4586 		while (*cp == '/')
4587 			cp++;
4588 		cp2 = cp;
4589 		while (*cp2 != '\0' && *cp2 != '/')
4590 			cp2++;
4591 		if (*cp2 == '/') {
4592 			setnil = 1;
4593 			*cp2 = '\0';
4594 		}
4595 		if (cp2 != cp) {
4596 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4597 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4598 			nfsm_strtom(nd, cp, strlen(cp));
4599 			cnt++;
4600 		}
4601 		if (setnil)
4602 			*cp2++ = '/';
4603 		cp = cp2;
4604 	} while (*cp != '\0');
4605 	if (NFSHASNFSV4N(nmp))
4606 		/* Has a Sequence Op done by nfscl_reqstart(). */
4607 		*opcntp = txdr_unsigned(3 + cnt);
4608 	else
4609 		*opcntp = txdr_unsigned(2 + cnt);
4610 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4611 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4612 	nd->nd_flag |= ND_USEGSSNAME;
4613 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4614 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4615 	if (error)
4616 		return (error);
4617 	if (nd->nd_repstat == 0) {
4618 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4619 		tl += (2 + 2 * cnt);
4620 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4621 			len > NFSX_FHMAX) {
4622 			nd->nd_repstat = NFSERR_BADXDR;
4623 		} else {
4624 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4625 			if (nd->nd_repstat == 0)
4626 				nmp->nm_fhsize = len;
4627 		}
4628 	}
4629 	error = nd->nd_repstat;
4630 nfsmout:
4631 	m_freem(nd->nd_mrep);
4632 	return (error);
4633 }
4634 
4635 /*
4636  * This function performs the Delegreturn RPC.
4637  */
4638 APPLESTATIC int
4639 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4640     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4641 {
4642 	u_int32_t *tl;
4643 	struct nfsrv_descript nfsd;
4644 	struct nfsrv_descript *nd = &nfsd;
4645 	int error;
4646 
4647 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4648 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4649 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4650 	if (NFSHASNFSV4N(nmp))
4651 		*tl++ = 0;
4652 	else
4653 		*tl++ = dp->nfsdl_stateid.seqid;
4654 	*tl++ = dp->nfsdl_stateid.other[0];
4655 	*tl++ = dp->nfsdl_stateid.other[1];
4656 	*tl = dp->nfsdl_stateid.other[2];
4657 	if (syscred)
4658 		nd->nd_flag |= ND_USEGSSNAME;
4659 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4660 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4661 	if (error)
4662 		return (error);
4663 	error = nd->nd_repstat;
4664 	m_freem(nd->nd_mrep);
4665 	return (error);
4666 }
4667 
4668 /*
4669  * nfs getacl call.
4670  */
4671 APPLESTATIC int
4672 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4673     struct acl *aclp, void *stuff)
4674 {
4675 	struct nfsrv_descript nfsd, *nd = &nfsd;
4676 	int error;
4677 	nfsattrbit_t attrbits;
4678 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4679 
4680 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4681 		return (EOPNOTSUPP);
4682 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4683 	NFSZERO_ATTRBIT(&attrbits);
4684 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4685 	(void) nfsrv_putattrbit(nd, &attrbits);
4686 	error = nfscl_request(nd, vp, p, cred, stuff);
4687 	if (error)
4688 		return (error);
4689 	if (!nd->nd_repstat)
4690 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4691 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4692 	else
4693 		error = nd->nd_repstat;
4694 	m_freem(nd->nd_mrep);
4695 	return (error);
4696 }
4697 
4698 /*
4699  * nfs setacl call.
4700  */
4701 APPLESTATIC int
4702 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4703     struct acl *aclp, void *stuff)
4704 {
4705 	int error;
4706 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4707 
4708 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4709 		return (EOPNOTSUPP);
4710 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4711 	return (error);
4712 }
4713 
4714 /*
4715  * nfs setacl call.
4716  */
4717 static int
4718 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4719     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4720 {
4721 	struct nfsrv_descript nfsd, *nd = &nfsd;
4722 	int error;
4723 	nfsattrbit_t attrbits;
4724 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4725 
4726 	if (!NFSHASNFSV4(nmp))
4727 		return (EOPNOTSUPP);
4728 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4729 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4730 	NFSZERO_ATTRBIT(&attrbits);
4731 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4732 	(void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4733 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4734 	error = nfscl_request(nd, vp, p, cred, stuff);
4735 	if (error)
4736 		return (error);
4737 	/* Don't care about the pre/postop attributes */
4738 	m_freem(nd->nd_mrep);
4739 	return (nd->nd_repstat);
4740 }
4741 
4742 /*
4743  * Do the NFSv4.1 Exchange ID.
4744  */
4745 int
4746 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4747     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
4748     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
4749 {
4750 	uint32_t *tl, v41flags;
4751 	struct nfsrv_descript nfsd;
4752 	struct nfsrv_descript *nd = &nfsd;
4753 	struct nfsclds *dsp;
4754 	struct timespec verstime;
4755 	int error, len;
4756 
4757 	*dspp = NULL;
4758 	if (minorvers == 0)
4759 		minorvers = nmp->nm_minorvers;
4760 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
4761 	    NFS_VER4, minorvers);
4762 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4763 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4764 	*tl = txdr_unsigned(clp->nfsc_rev);
4765 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4766 
4767 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4768 	*tl++ = txdr_unsigned(exchflags);
4769 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4770 
4771 	/* Set the implementation id4 */
4772 	*tl = txdr_unsigned(1);
4773 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4774 	(void) nfsm_strtom(nd, version, strlen(version));
4775 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4776 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4777 	verstime.tv_nsec = 0;
4778 	txdr_nfsv4time(&verstime, tl);
4779 	nd->nd_flag |= ND_USEGSSNAME;
4780 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4781 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4782 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4783 	    (int)nd->nd_repstat);
4784 	if (error != 0)
4785 		return (error);
4786 	if (nd->nd_repstat == 0) {
4787 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4788 		len = fxdr_unsigned(int, *(tl + 7));
4789 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4790 			error = NFSERR_BADXDR;
4791 			goto nfsmout;
4792 		}
4793 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4794 		    M_WAITOK | M_ZERO);
4795 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4796 		dsp->nfsclds_servownlen = len;
4797 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4798 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4799 		dsp->nfsclds_sess.nfsess_sequenceid =
4800 		    fxdr_unsigned(uint32_t, *tl++);
4801 		v41flags = fxdr_unsigned(uint32_t, *tl);
4802 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4803 		    NFSHASPNFSOPT(nmp)) {
4804 			NFSCL_DEBUG(1, "set PNFS\n");
4805 			NFSLOCKMNT(nmp);
4806 			nmp->nm_state |= NFSSTA_PNFS;
4807 			NFSUNLOCKMNT(nmp);
4808 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4809 		}
4810 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4811 			dsp->nfsclds_flags |= NFSCLDS_DS;
4812 		if (minorvers == NFSV42_MINORVERSION)
4813 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
4814 		if (len > 0)
4815 			nd->nd_repstat = nfsrv_mtostr(nd,
4816 			    dsp->nfsclds_serverown, len);
4817 		if (nd->nd_repstat == 0) {
4818 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4819 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4820 			    NULL, MTX_DEF);
4821 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4822 			*dspp = dsp;
4823 		} else
4824 			free(dsp, M_NFSCLDS);
4825 	}
4826 	error = nd->nd_repstat;
4827 nfsmout:
4828 	m_freem(nd->nd_mrep);
4829 	return (error);
4830 }
4831 
4832 /*
4833  * Do the NFSv4.1 Create Session.
4834  */
4835 int
4836 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4837     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
4838     struct ucred *cred, NFSPROC_T *p)
4839 {
4840 	uint32_t crflags, maxval, *tl;
4841 	struct nfsrv_descript nfsd;
4842 	struct nfsrv_descript *nd = &nfsd;
4843 	int error, irdcnt, minorvers;
4844 
4845 	/* Make sure nm_rsize, nm_wsize is set. */
4846 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4847 		nmp->nm_rsize = NFS_MAXBSIZE;
4848 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4849 		nmp->nm_wsize = NFS_MAXBSIZE;
4850 	if (dsp == NULL)
4851 		minorvers = nmp->nm_minorvers;
4852 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
4853 		minorvers = NFSV42_MINORVERSION;
4854 	else
4855 		minorvers = NFSV41_MINORVERSION;
4856 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
4857 	    NFS_VER4, minorvers);
4858 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4859 	*tl++ = sep->nfsess_clientid.lval[0];
4860 	*tl++ = sep->nfsess_clientid.lval[1];
4861 	*tl++ = txdr_unsigned(sequenceid);
4862 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4863 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4864 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4865 	*tl = txdr_unsigned(crflags);
4866 
4867 	/* Fill in fore channel attributes. */
4868 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4869 	*tl++ = 0;				/* Header pad size */
4870 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
4871 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
4872 		/*
4873 		 * NFSv4.2 Extended Attribute operations may want to do
4874 		 * requests/replies that are larger than nm_rsize/nm_wsize.
4875 		 */
4876 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4877 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4878 	} else {
4879 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
4880 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
4881 	}
4882 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4883 	*tl++ = txdr_unsigned(20);		/* Max operations */
4884 	*tl++ = txdr_unsigned(64);		/* Max slots */
4885 	*tl = 0;				/* No rdma ird */
4886 
4887 	/* Fill in back channel attributes. */
4888 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4889 	*tl++ = 0;				/* Header pad size */
4890 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4891 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4892 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4893 	*tl++ = txdr_unsigned(4);		/* Max operations */
4894 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4895 	*tl = 0;				/* No rdma ird */
4896 
4897 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4898 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4899 
4900 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4901 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4902 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4903 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4904 	*tl++ = 0;				/* Null machine name */
4905 	*tl++ = 0;				/* Uid == 0 */
4906 	*tl++ = 0;				/* Gid == 0 */
4907 	*tl = 0;				/* No additional gids */
4908 	nd->nd_flag |= ND_USEGSSNAME;
4909 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4910 	    NFS_VER4, NULL, 1, NULL, NULL);
4911 	if (error != 0)
4912 		return (error);
4913 	if (nd->nd_repstat == 0) {
4914 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4915 		    2 * NFSX_UNSIGNED);
4916 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4917 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4918 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4919 		crflags = fxdr_unsigned(uint32_t, *tl);
4920 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4921 			NFSLOCKMNT(nmp);
4922 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4923 			NFSUNLOCKMNT(nmp);
4924 		}
4925 
4926 		/* Get the fore channel slot count. */
4927 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4928 		tl++;			/* Skip the header pad size. */
4929 
4930 		/* Make sure nm_wsize is small enough. */
4931 		maxval = fxdr_unsigned(uint32_t, *tl++);
4932 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4933 			if (nmp->nm_wsize > 8096)
4934 				nmp->nm_wsize /= 2;
4935 			else
4936 				break;
4937 		}
4938 		sep->nfsess_maxreq = maxval;
4939 
4940 		/* Make sure nm_rsize is small enough. */
4941 		maxval = fxdr_unsigned(uint32_t, *tl++);
4942 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4943 			if (nmp->nm_rsize > 8096)
4944 				nmp->nm_rsize /= 2;
4945 			else
4946 				break;
4947 		}
4948 		sep->nfsess_maxresp = maxval;
4949 
4950 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4951 		tl++;
4952 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4953 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4954 		irdcnt = fxdr_unsigned(int, *tl);
4955 		if (irdcnt > 0)
4956 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4957 
4958 		/* and the back channel slot count. */
4959 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4960 		tl += 5;
4961 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4962 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4963 	}
4964 	error = nd->nd_repstat;
4965 nfsmout:
4966 	m_freem(nd->nd_mrep);
4967 	return (error);
4968 }
4969 
4970 /*
4971  * Do the NFSv4.1 Destroy Session.
4972  */
4973 int
4974 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4975     struct ucred *cred, NFSPROC_T *p)
4976 {
4977 	uint32_t *tl;
4978 	struct nfsrv_descript nfsd;
4979 	struct nfsrv_descript *nd = &nfsd;
4980 	int error;
4981 	struct nfsclsession *tsep;
4982 
4983 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
4984 	    0);
4985 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4986 	tsep = nfsmnt_mdssession(nmp);
4987 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4988 	nd->nd_flag |= ND_USEGSSNAME;
4989 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4990 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4991 	if (error != 0)
4992 		return (error);
4993 	error = nd->nd_repstat;
4994 	m_freem(nd->nd_mrep);
4995 	return (error);
4996 }
4997 
4998 /*
4999  * Do the NFSv4.1 Destroy Client.
5000  */
5001 int
5002 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5003     struct ucred *cred, NFSPROC_T *p)
5004 {
5005 	uint32_t *tl;
5006 	struct nfsrv_descript nfsd;
5007 	struct nfsrv_descript *nd = &nfsd;
5008 	int error;
5009 	struct nfsclsession *tsep;
5010 
5011 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5012 	    0);
5013 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5014 	tsep = nfsmnt_mdssession(nmp);
5015 	*tl++ = tsep->nfsess_clientid.lval[0];
5016 	*tl = tsep->nfsess_clientid.lval[1];
5017 	nd->nd_flag |= ND_USEGSSNAME;
5018 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5019 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5020 	if (error != 0)
5021 		return (error);
5022 	error = nd->nd_repstat;
5023 	m_freem(nd->nd_mrep);
5024 	return (error);
5025 }
5026 
5027 /*
5028  * Do the NFSv4.1 LayoutGet.
5029  */
5030 static int
5031 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5032     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5033     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5034     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
5035     void *stuff)
5036 {
5037 	struct nfsrv_descript nfsd, *nd = &nfsd;
5038 	int error;
5039 
5040 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5041 	    0);
5042 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5043 	    layouttype, layoutlen, 0);
5044 	nd->nd_flag |= ND_USEGSSNAME;
5045 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5046 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5047 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5048 	if (error != 0)
5049 		return (error);
5050 	if (nd->nd_repstat == 0)
5051 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5052 		    flhp);
5053 	if (error == 0 && nd->nd_repstat != 0)
5054 		error = nd->nd_repstat;
5055 	m_freem(nd->nd_mrep);
5056 	return (error);
5057 }
5058 
5059 /*
5060  * Do the NFSv4.1 Get Device Info.
5061  */
5062 int
5063 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5064     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5065     NFSPROC_T *p)
5066 {
5067 	uint32_t cnt, *tl, vers, minorvers;
5068 	struct nfsrv_descript nfsd;
5069 	struct nfsrv_descript *nd = &nfsd;
5070 	struct sockaddr_in sin, ssin;
5071 	struct sockaddr_in6 sin6, ssin6;
5072 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5073 	struct nfscldevinfo *ndi;
5074 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5075 	int stripecnt;
5076 	uint8_t stripeindex;
5077 	sa_family_t af, safilled;
5078 
5079 	ssin.sin_port = 0;		/* To shut up compiler. */
5080 	ssin.sin_addr.s_addr = 0;	/* ditto */
5081 	*ndip = NULL;
5082 	ndi = NULL;
5083 	gotdspp = NULL;
5084 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5085 	    0);
5086 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5087 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5088 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5089 	*tl++ = txdr_unsigned(layouttype);
5090 	*tl++ = txdr_unsigned(100000);
5091 	if (notifybitsp != NULL && *notifybitsp != 0) {
5092 		*tl = txdr_unsigned(1);		/* One word of bits. */
5093 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5094 		*tl = txdr_unsigned(*notifybitsp);
5095 	} else
5096 		*tl = txdr_unsigned(0);
5097 	nd->nd_flag |= ND_USEGSSNAME;
5098 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5099 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5100 	if (error != 0)
5101 		return (error);
5102 	if (nd->nd_repstat == 0) {
5103 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5104 		if (layouttype != fxdr_unsigned(int, *tl))
5105 			printf("EEK! devinfo layout type not same!\n");
5106 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5107 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5108 			stripecnt = fxdr_unsigned(int, *tl);
5109 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5110 			if (stripecnt < 1 || stripecnt > 4096) {
5111 				printf("pNFS File layout devinfo stripecnt %d:"
5112 				    " out of range\n", stripecnt);
5113 				error = NFSERR_BADXDR;
5114 				goto nfsmout;
5115 			}
5116 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5117 			    NFSX_UNSIGNED);
5118 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5119 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5120 			if (addrcnt < 1 || addrcnt > 128) {
5121 				printf("NFS devinfo addrcnt %d: out of range\n",
5122 				    addrcnt);
5123 				error = NFSERR_BADXDR;
5124 				goto nfsmout;
5125 			}
5126 
5127 			/*
5128 			 * Now we know how many stripe indices and addresses, so
5129 			 * we can allocate the structure the correct size.
5130 			 */
5131 			i = (stripecnt * sizeof(uint8_t)) /
5132 			    sizeof(struct nfsclds *) + 1;
5133 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5134 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5135 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5136 			    M_ZERO);
5137 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5138 			    NFSX_V4DEVICEID);
5139 			ndi->nfsdi_refcnt = 0;
5140 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5141 			ndi->nfsdi_stripecnt = stripecnt;
5142 			ndi->nfsdi_addrcnt = addrcnt;
5143 			/* Fill in the stripe indices. */
5144 			for (i = 0; i < stripecnt; i++) {
5145 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5146 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5147 				if (stripeindex >= addrcnt) {
5148 					printf("pNFS File Layout devinfo"
5149 					    " stripeindex %d: too big\n",
5150 					    (int)stripeindex);
5151 					error = NFSERR_BADXDR;
5152 					goto nfsmout;
5153 				}
5154 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5155 			}
5156 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5157 			/* For Flex File, we only get one address list. */
5158 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5159 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5160 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5161 			    NFSX_V4DEVICEID);
5162 			ndi->nfsdi_refcnt = 0;
5163 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5164 			addrcnt = ndi->nfsdi_addrcnt = 1;
5165 		}
5166 
5167 		/* Now, dissect the server address(es). */
5168 		safilled = AF_UNSPEC;
5169 		for (i = 0; i < addrcnt; i++) {
5170 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5171 			cnt = fxdr_unsigned(uint32_t, *tl);
5172 			if (cnt == 0) {
5173 				printf("NFS devinfo 0 len addrlist\n");
5174 				error = NFSERR_BADXDR;
5175 				goto nfsmout;
5176 			}
5177 			dspp = nfsfldi_addr(ndi, i);
5178 			safilled = AF_UNSPEC;
5179 			for (j = 0; j < cnt; j++) {
5180 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5181 				    &isudp);
5182 				if (error != 0 && error != EPERM) {
5183 					error = NFSERR_BADXDR;
5184 					goto nfsmout;
5185 				}
5186 				if (error == 0 && isudp == 0) {
5187 					/*
5188 					 * The priority is:
5189 					 * - Same address family.
5190 					 * Save the address and dspp, so that
5191 					 * the connection can be done after
5192 					 * parsing is complete.
5193 					 */
5194 					if (safilled == AF_UNSPEC ||
5195 					    (af == nmp->nm_nam->sa_family &&
5196 					     safilled != nmp->nm_nam->sa_family)
5197 					   ) {
5198 						if (af == AF_INET)
5199 							ssin = sin;
5200 						else
5201 							ssin6 = sin6;
5202 						safilled = af;
5203 						gotdspp = dspp;
5204 					}
5205 				}
5206 			}
5207 		}
5208 
5209 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5210 		gotminor = NFSV41_MINORVERSION;
5211 		/* For Flex File, we will take one of the versions to use. */
5212 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5213 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5214 			j = fxdr_unsigned(int, *tl);
5215 			if (j < 1 || j > NFSDEV_MAXVERS) {
5216 				printf("pNFS: too many versions\n");
5217 				error = NFSERR_BADXDR;
5218 				goto nfsmout;
5219 			}
5220 			gotvers = 0;
5221 			gotminor = 0;
5222 			for (i = 0; i < j; i++) {
5223 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5224 				vers = fxdr_unsigned(uint32_t, *tl++);
5225 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5226 				if (vers == NFS_VER3)
5227 					minorvers = 0;
5228 				if ((vers == NFS_VER4 && ((minorvers ==
5229 				    NFSV41_MINORVERSION && gotminor == 0) ||
5230 				    minorvers == NFSV42_MINORVERSION)) ||
5231 				    (vers == NFS_VER3 && gotvers == 0)) {
5232 					gotvers = vers;
5233 					gotminor = minorvers;
5234 					/* We'll take this one. */
5235 					ndi->nfsdi_versindex = i;
5236 					ndi->nfsdi_vers = vers;
5237 					ndi->nfsdi_minorvers = minorvers;
5238 					ndi->nfsdi_rsize = fxdr_unsigned(
5239 					    uint32_t, *tl++);
5240 					ndi->nfsdi_wsize = fxdr_unsigned(
5241 					    uint32_t, *tl++);
5242 					if (*tl == newnfs_true)
5243 						ndi->nfsdi_flags |=
5244 						    NFSDI_TIGHTCOUPLED;
5245 					else
5246 						ndi->nfsdi_flags &=
5247 						    ~NFSDI_TIGHTCOUPLED;
5248 				}
5249 			}
5250 			if (gotvers == 0) {
5251 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5252 				error = NFSERR_BADXDR;
5253 				goto nfsmout;
5254 			}
5255 		}
5256 
5257 		/* And the notify bits. */
5258 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5259 		bitcnt = fxdr_unsigned(int, *tl);
5260 		if (bitcnt > 0) {
5261 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5262 			if (notifybitsp != NULL)
5263 				*notifybitsp =
5264 				    fxdr_unsigned(uint32_t, *tl);
5265 		}
5266 		if (safilled != AF_UNSPEC) {
5267 			KASSERT(ndi != NULL, ("ndi is NULL"));
5268 			*ndip = ndi;
5269 		} else
5270 			error = EPERM;
5271 		if (error == 0) {
5272 			/*
5273 			 * Now we can do a TCP connection for the correct
5274 			 * NFS version and IP address.
5275 			 */
5276 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5277 			    gotvers, gotminor, &dsp, p);
5278 		}
5279 		if (error == 0) {
5280 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5281 			*gotdspp = dsp;
5282 		}
5283 	}
5284 	if (nd->nd_repstat != 0 && error == 0)
5285 		error = nd->nd_repstat;
5286 nfsmout:
5287 	if (error != 0 && ndi != NULL)
5288 		nfscl_freedevinfo(ndi);
5289 	m_freem(nd->nd_mrep);
5290 	return (error);
5291 }
5292 
5293 /*
5294  * Do the NFSv4.1 LayoutCommit.
5295  */
5296 int
5297 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5298     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5299     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5300 {
5301 	uint32_t *tl;
5302 	struct nfsrv_descript nfsd, *nd = &nfsd;
5303 	int error;
5304 
5305 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5306 	    0, 0);
5307 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5308 	    NFSX_STATEID);
5309 	txdr_hyper(off, tl);
5310 	tl += 2;
5311 	txdr_hyper(len, tl);
5312 	tl += 2;
5313 	if (reclaim != 0)
5314 		*tl++ = newnfs_true;
5315 	else
5316 		*tl++ = newnfs_false;
5317 	*tl++ = txdr_unsigned(stateidp->seqid);
5318 	*tl++ = stateidp->other[0];
5319 	*tl++ = stateidp->other[1];
5320 	*tl++ = stateidp->other[2];
5321 	*tl++ = newnfs_true;
5322 	if (lastbyte < off)
5323 		lastbyte = off;
5324 	else if (lastbyte >= (off + len))
5325 		lastbyte = off + len - 1;
5326 	txdr_hyper(lastbyte, tl);
5327 	tl += 2;
5328 	*tl++ = newnfs_false;
5329 	*tl++ = txdr_unsigned(layouttype);
5330 	/* All supported layouts are 0 length. */
5331 	*tl = txdr_unsigned(0);
5332 	nd->nd_flag |= ND_USEGSSNAME;
5333 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5334 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5335 	if (error != 0)
5336 		return (error);
5337 	error = nd->nd_repstat;
5338 	m_freem(nd->nd_mrep);
5339 	return (error);
5340 }
5341 
5342 /*
5343  * Do the NFSv4.1 LayoutReturn.
5344  */
5345 int
5346 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5347     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5348     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5349     uint32_t stat, uint32_t op, char *devid)
5350 {
5351 	uint32_t *tl;
5352 	struct nfsrv_descript nfsd, *nd = &nfsd;
5353 	uint64_t tu64;
5354 	int error;
5355 
5356 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5357 	    0, 0);
5358 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5359 	if (reclaim != 0)
5360 		*tl++ = newnfs_true;
5361 	else
5362 		*tl++ = newnfs_false;
5363 	*tl++ = txdr_unsigned(layouttype);
5364 	*tl++ = txdr_unsigned(iomode);
5365 	*tl = txdr_unsigned(layoutreturn);
5366 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5367 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5368 		    NFSX_UNSIGNED);
5369 		txdr_hyper(offset, tl);
5370 		tl += 2;
5371 		txdr_hyper(len, tl);
5372 		tl += 2;
5373 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5374 		*tl++ = txdr_unsigned(stateidp->seqid);
5375 		*tl++ = stateidp->other[0];
5376 		*tl++ = stateidp->other[1];
5377 		*tl++ = stateidp->other[2];
5378 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5379 			*tl = txdr_unsigned(0);
5380 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5381 			if (stat != 0) {
5382 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5383 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5384 				    NFSX_UNSIGNED);
5385 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5386 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5387 				    NFSX_UNSIGNED);
5388 				*tl++ = txdr_unsigned(1);	/* One error. */
5389 				tu64 = 0;			/* Offset. */
5390 				txdr_hyper(tu64, tl); tl += 2;
5391 				tu64 = UINT64_MAX;		/* Length. */
5392 				txdr_hyper(tu64, tl); tl += 2;
5393 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5394 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5395 				*tl++ = txdr_unsigned(1);	/* One error. */
5396 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5397 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5398 				*tl++ = txdr_unsigned(stat);
5399 				*tl++ = txdr_unsigned(op);
5400 			} else {
5401 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5402 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5403 				/* No ioerrs. */
5404 				*tl++ = 0;
5405 			}
5406 			*tl = 0;	/* No stats yet. */
5407 		}
5408 	}
5409 	nd->nd_flag |= ND_USEGSSNAME;
5410 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5411 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5412 	if (error != 0)
5413 		return (error);
5414 	if (nd->nd_repstat == 0) {
5415 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5416 		if (*tl != 0) {
5417 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5418 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5419 			stateidp->other[0] = *tl++;
5420 			stateidp->other[1] = *tl++;
5421 			stateidp->other[2] = *tl;
5422 		}
5423 	} else
5424 		error = nd->nd_repstat;
5425 nfsmout:
5426 	m_freem(nd->nd_mrep);
5427 	return (error);
5428 }
5429 
5430 /*
5431  * Acquire a layout and devinfo, if possible. The caller must have acquired
5432  * a reference count on the nfsclclient structure before calling this.
5433  * Return the layout in lypp with a reference count on it, if successful.
5434  */
5435 static int
5436 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5437     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5438     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5439 {
5440 	struct nfscllayout *lyp;
5441 	struct nfsclflayout *flp;
5442 	struct nfsclflayouthead flh;
5443 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5444 	nfsv4stateid_t stateid;
5445 	struct nfsclsession *tsep;
5446 
5447 	*lypp = NULL;
5448 	if (NFSHASFLEXFILE(nmp))
5449 		layouttype = NFSLAYOUT_FLEXFILE;
5450 	else
5451 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5452 	/*
5453 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5454 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5455 	 * flp == NULL.
5456 	 */
5457 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5458 	    off, &flp, &recalled);
5459 	islocked = 0;
5460 	if (lyp == NULL || flp == NULL) {
5461 		if (recalled != 0)
5462 			return (EIO);
5463 		LIST_INIT(&flh);
5464 		tsep = nfsmnt_mdssession(nmp);
5465 		layoutlen = tsep->nfsess_maxcache -
5466 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5467 		if (lyp == NULL) {
5468 			stateid.seqid = 0;
5469 			stateid.other[0] = stateidp->other[0];
5470 			stateid.other[1] = stateidp->other[1];
5471 			stateid.other[2] = stateidp->other[2];
5472 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5473 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5474 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5475 			    &retonclose, &flh, cred, p, NULL);
5476 		} else {
5477 			islocked = 1;
5478 			stateid.seqid = lyp->nfsly_stateid.seqid;
5479 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5480 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5481 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5482 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5483 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5484 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5485 			    &retonclose, &flh, cred, p, NULL);
5486 		}
5487 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5488 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5489 		    &flh, layouttype, error, NULL, cred, p);
5490 		if (error == 0)
5491 			*lypp = lyp;
5492 		else if (islocked != 0)
5493 			nfscl_rellayout(lyp, 1);
5494 	} else
5495 		*lypp = lyp;
5496 	return (error);
5497 }
5498 
5499 /*
5500  * Do a TCP connection plus exchange id and create session.
5501  * If successful, a "struct nfsclds" is linked into the list for the
5502  * mount point and a pointer to it is returned.
5503  */
5504 static int
5505 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5506     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5507     struct nfsclds **dspp, NFSPROC_T *p)
5508 {
5509 	struct sockaddr_in *msad, *sad;
5510 	struct sockaddr_in6 *msad6, *sad6;
5511 	struct nfsclclient *clp;
5512 	struct nfssockreq *nrp;
5513 	struct nfsclds *dsp, *tdsp;
5514 	int error, firsttry;
5515 	enum nfsclds_state retv;
5516 	uint32_t sequenceid = 0;
5517 
5518 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5519 	    ("nfsrpc_fillsa: NULL nr_cred"));
5520 	NFSLOCKCLSTATE();
5521 	clp = nmp->nm_clp;
5522 	NFSUNLOCKCLSTATE();
5523 	if (clp == NULL)
5524 		return (EPERM);
5525 	if (af == AF_INET) {
5526 		NFSLOCKMNT(nmp);
5527 		/*
5528 		 * Check to see if we already have a session for this
5529 		 * address that is usable for a DS.
5530 		 * Note that the MDS's address is in a different place
5531 		 * than the sessions already acquired for DS's.
5532 		 */
5533 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5534 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5535 		while (tdsp != NULL) {
5536 			if (msad != NULL && msad->sin_family == AF_INET &&
5537 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5538 			    sin->sin_port == msad->sin_port &&
5539 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5540 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5541 				*dspp = tdsp;
5542 				NFSUNLOCKMNT(nmp);
5543 				NFSCL_DEBUG(4, "fnd same addr\n");
5544 				return (0);
5545 			}
5546 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5547 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5548 				msad = (struct sockaddr_in *)
5549 				    tdsp->nfsclds_sockp->nr_nam;
5550 			else
5551 				msad = NULL;
5552 		}
5553 		NFSUNLOCKMNT(nmp);
5554 
5555 		/* No IP address match, so look for new/trunked one. */
5556 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5557 		sad->sin_len = sizeof(*sad);
5558 		sad->sin_family = AF_INET;
5559 		sad->sin_port = sin->sin_port;
5560 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5561 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5562 		nrp->nr_nam = (struct sockaddr *)sad;
5563 	} else if (af == AF_INET6) {
5564 		NFSLOCKMNT(nmp);
5565 		/*
5566 		 * Check to see if we already have a session for this
5567 		 * address that is usable for a DS.
5568 		 * Note that the MDS's address is in a different place
5569 		 * than the sessions already acquired for DS's.
5570 		 */
5571 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5572 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5573 		while (tdsp != NULL) {
5574 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5575 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5576 			    &msad6->sin6_addr) &&
5577 			    sin6->sin6_port == msad6->sin6_port &&
5578 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5579 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5580 				*dspp = tdsp;
5581 				NFSUNLOCKMNT(nmp);
5582 				return (0);
5583 			}
5584 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5585 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5586 				msad6 = (struct sockaddr_in6 *)
5587 				    tdsp->nfsclds_sockp->nr_nam;
5588 			else
5589 				msad6 = NULL;
5590 		}
5591 		NFSUNLOCKMNT(nmp);
5592 
5593 		/* No IP address match, so look for new/trunked one. */
5594 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5595 		sad6->sin6_len = sizeof(*sad6);
5596 		sad6->sin6_family = AF_INET6;
5597 		sad6->sin6_port = sin6->sin6_port;
5598 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5599 		    sizeof(struct in6_addr));
5600 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5601 		nrp->nr_nam = (struct sockaddr *)sad6;
5602 	} else
5603 		return (EPERM);
5604 
5605 	nrp->nr_sotype = SOCK_STREAM;
5606 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5607 	nrp->nr_prog = NFS_PROG;
5608 	nrp->nr_vers = vers;
5609 
5610 	/*
5611 	 * Use the credentials that were used for the mount, which are
5612 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5613 	 * Ref. counting the credentials with crhold() is probably not
5614 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5615 	 * unmount, but I did it anyhow.
5616 	 */
5617 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5618 	error = newnfs_connect(nmp, nrp, NULL, p, 0);
5619 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5620 
5621 	dsp = NULL;
5622 	/* Now, do the exchangeid and create session. */
5623 	if (error == 0) {
5624 		if (vers == NFS_VER4) {
5625 			firsttry = 0;
5626 			do {
5627 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5628 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5629 				    nrp->nr_cred, p);
5630 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5631 				if (error == NFSERR_MINORVERMISMATCH)
5632 					minorvers = NFSV42_MINORVERSION;
5633 			} while (error == NFSERR_MINORVERMISMATCH &&
5634 			    firsttry++ == 0);
5635 			if (error != 0)
5636 				newnfs_disconnect(nrp);
5637 		} else {
5638 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5639 			    M_WAITOK | M_ZERO);
5640 			dsp->nfsclds_flags |= NFSCLDS_DS;
5641 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5642 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5643 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5644 			    NULL, MTX_DEF);
5645 		}
5646 	}
5647 	if (error == 0) {
5648 		dsp->nfsclds_sockp = nrp;
5649 		if (vers == NFS_VER4) {
5650 			NFSLOCKMNT(nmp);
5651 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5652 			    &sequenceid);
5653 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5654 			if (retv == NFSDSP_USETHISSESSION &&
5655 			    nfscl_dssameconn != 0) {
5656 				NFSLOCKDS(tdsp);
5657 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5658 				NFSUNLOCKDS(tdsp);
5659 				NFSUNLOCKMNT(nmp);
5660 				/*
5661 				 * If there is already a session for this
5662 				 * server, use it.
5663 				 */
5664 				(void)newnfs_disconnect(nrp);
5665 				nfscl_freenfsclds(dsp);
5666 				*dspp = tdsp;
5667 				return (0);
5668 			}
5669 			if (retv == NFSDSP_NOTFOUND)
5670 				sequenceid =
5671 				    dsp->nfsclds_sess.nfsess_sequenceid;
5672 			NFSUNLOCKMNT(nmp);
5673 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5674 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
5675 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5676 		}
5677 	} else {
5678 		NFSFREECRED(nrp->nr_cred);
5679 		NFSFREEMUTEX(&nrp->nr_mtx);
5680 		free(nrp->nr_nam, M_SONAME);
5681 		free(nrp, M_NFSSOCKREQ);
5682 	}
5683 	if (error == 0) {
5684 		NFSCL_DEBUG(3, "add DS session\n");
5685 		/*
5686 		 * Put it at the end of the list. That way the list
5687 		 * is ordered by when the entry was added. This matters
5688 		 * since the one done first is the one that should be
5689 		 * used for sequencid'ing any subsequent create sessions.
5690 		 */
5691 		NFSLOCKMNT(nmp);
5692 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5693 		NFSUNLOCKMNT(nmp);
5694 		*dspp = dsp;
5695 	} else if (dsp != NULL) {
5696 		newnfs_disconnect(nrp);
5697 		nfscl_freenfsclds(dsp);
5698 	}
5699 	return (error);
5700 }
5701 
5702 /*
5703  * Do the NFSv4.1 Reclaim Complete.
5704  */
5705 int
5706 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5707 {
5708 	uint32_t *tl;
5709 	struct nfsrv_descript nfsd;
5710 	struct nfsrv_descript *nd = &nfsd;
5711 	int error;
5712 
5713 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5714 	    0);
5715 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5716 	*tl = newnfs_false;
5717 	nd->nd_flag |= ND_USEGSSNAME;
5718 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5719 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5720 	if (error != 0)
5721 		return (error);
5722 	error = nd->nd_repstat;
5723 	m_freem(nd->nd_mrep);
5724 	return (error);
5725 }
5726 
5727 /*
5728  * Initialize the slot tables for a session.
5729  */
5730 static void
5731 nfscl_initsessionslots(struct nfsclsession *sep)
5732 {
5733 	int i;
5734 
5735 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5736 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5737 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5738 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5739 	}
5740 	for (i = 0; i < 64; i++)
5741 		sep->nfsess_slotseq[i] = 0;
5742 	sep->nfsess_slots = 0;
5743 }
5744 
5745 /*
5746  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5747  */
5748 int
5749 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5750     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5751 {
5752 	struct nfsnode *np = VTONFS(vp);
5753 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5754 	struct nfscllayout *layp;
5755 	struct nfscldevinfo *dip;
5756 	struct nfsclflayout *rflp;
5757 	struct mbuf *m;
5758 	struct nfsclwritedsdorpc *drpc, *tdrpc;
5759 	nfsv4stateid_t stateid;
5760 	struct ucred *newcred;
5761 	uint64_t lastbyte, len, off, oresid, xfer;
5762 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5763 	void *lckp;
5764 	uint8_t *dev;
5765 	void *iovbase = NULL;
5766 	size_t iovlen = 0;
5767 	off_t offs = 0;
5768 	ssize_t resid = 0;
5769 
5770 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5771 	    (np->n_flag & NNOLAYOUT) != 0)
5772 		return (EIO);
5773 	/* Now, get a reference cnt on the clientid for this mount. */
5774 	if (nfscl_getref(nmp) == 0)
5775 		return (EIO);
5776 
5777 	/* Find an appropriate stateid. */
5778 	newcred = NFSNEWCRED(cred);
5779 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5780 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5781 	if (error != 0) {
5782 		NFSFREECRED(newcred);
5783 		nfscl_relref(nmp);
5784 		return (error);
5785 	}
5786 	/* Search for a layout for this file. */
5787 	off = uiop->uio_offset;
5788 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5789 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5790 	if (layp == NULL || rflp == NULL) {
5791 		if (recalled != 0) {
5792 			NFSFREECRED(newcred);
5793 			nfscl_relref(nmp);
5794 			return (EIO);
5795 		}
5796 		if (layp != NULL) {
5797 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5798 			layp = NULL;
5799 		}
5800 		/* Try and get a Layout, if it is supported. */
5801 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5802 		    (np->n_flag & NWRITEOPENED) != 0)
5803 			iolaymode = NFSLAYOUTIOMODE_RW;
5804 		else
5805 			iolaymode = NFSLAYOUTIOMODE_READ;
5806 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5807 		    NULL, &stateid, off, &layp, newcred, p);
5808 		if (error != 0) {
5809 			NFSLOCKNODE(np);
5810 			np->n_flag |= NNOLAYOUT;
5811 			NFSUNLOCKNODE(np);
5812 			if (lckp != NULL)
5813 				nfscl_lockderef(lckp);
5814 			NFSFREECRED(newcred);
5815 			if (layp != NULL)
5816 				nfscl_rellayout(layp, 0);
5817 			nfscl_relref(nmp);
5818 			return (error);
5819 		}
5820 	}
5821 
5822 	/*
5823 	 * Loop around finding a layout that works for the first part of
5824 	 * this I/O operation, and then call the function that actually
5825 	 * does the RPC.
5826 	 */
5827 	eof = 0;
5828 	len = (uint64_t)uiop->uio_resid;
5829 	while (len > 0 && error == 0 && eof == 0) {
5830 		off = uiop->uio_offset;
5831 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5832 		if (error == 0) {
5833 			oresid = xfer = (uint64_t)uiop->uio_resid;
5834 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5835 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5836 			/*
5837 			 * For Flex File layout with mirrored DSs, select one
5838 			 * of them at random for reads. For writes and commits,
5839 			 * do all mirrors.
5840 			 */
5841 			m = NULL;
5842 			tdrpc = drpc = NULL;
5843 			firstmirror = 0;
5844 			mirrorcnt = 1;
5845 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5846 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5847 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5848 					firstmirror = arc4random() % mirrorcnt;
5849 					mirrorcnt = firstmirror + 1;
5850 				} else {
5851 					if (docommit == 0) {
5852 						/*
5853 						 * Save values, so uiop can be
5854 						 * rolled back upon a write
5855 						 * error.
5856 						 */
5857 						offs = uiop->uio_offset;
5858 						resid = uiop->uio_resid;
5859 						iovbase =
5860 						    uiop->uio_iov->iov_base;
5861 						iovlen = uiop->uio_iov->iov_len;
5862 						m = nfsm_uiombuflist(uiop, len,
5863 						    NULL, NULL);
5864 					}
5865 					tdrpc = drpc = malloc(sizeof(*drpc) *
5866 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
5867 					    M_ZERO);
5868 				}
5869 			}
5870 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5871 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
5872 					dev = rflp->nfsfl_ffm[i].dev;
5873 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5874 					    rflp->nfsfl_ffm[i].devp);
5875 				} else {
5876 					dev = rflp->nfsfl_dev;
5877 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5878 					    rflp->nfsfl_devp);
5879 				}
5880 				if (dip != NULL) {
5881 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5882 					    != 0)
5883 						error = nfscl_dofflayoutio(vp,
5884 						    uiop, iomode, must_commit,
5885 						    &eof, &stateid, rwaccess,
5886 						    dip, layp, rflp, off, xfer,
5887 						    i, docommit, m, tdrpc,
5888 						    newcred, p);
5889 					else
5890 						error = nfscl_doflayoutio(vp,
5891 						    uiop, iomode, must_commit,
5892 						    &eof, &stateid, rwaccess,
5893 						    dip, layp, rflp, off, xfer,
5894 						    docommit, newcred, p);
5895 					nfscl_reldevinfo(dip);
5896 				} else
5897 					error = EIO;
5898 				tdrpc++;
5899 			}
5900 			if (m != NULL)
5901 				m_freem(m);
5902 			tdrpc = drpc;
5903 			timo = hz / 50;		/* Wait for 20msec. */
5904 			if (timo < 1)
5905 				timo = 1;
5906 			for (i = firstmirror; i < mirrorcnt - 1 &&
5907 			    tdrpc != NULL; i++, tdrpc++) {
5908 				/*
5909 				 * For the unused drpc entries, both inprog and
5910 				 * err == 0, so this loop won't break.
5911 				 */
5912 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
5913 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5914 					    timo);
5915 				if (error == 0 && tdrpc->err != 0)
5916 					error = tdrpc->err;
5917 			}
5918 			free(drpc, M_TEMP);
5919 			if (error == 0) {
5920 				if (mirrorcnt > 1 && rwaccess ==
5921 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5922 					NFSLOCKCLSTATE();
5923 					layp->nfsly_flags |= NFSLY_WRITTEN;
5924 					NFSUNLOCKCLSTATE();
5925 				}
5926 				lastbyte = off + xfer - 1;
5927 				NFSLOCKCLSTATE();
5928 				if (lastbyte > layp->nfsly_lastbyte)
5929 					layp->nfsly_lastbyte = lastbyte;
5930 				NFSUNLOCKCLSTATE();
5931 			} else if (error == NFSERR_OPENMODE &&
5932 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
5933 				NFSLOCKMNT(nmp);
5934 				nmp->nm_state |= NFSSTA_OPENMODE;
5935 				NFSUNLOCKMNT(nmp);
5936 			} else
5937 				error = EIO;
5938 			if (error == 0)
5939 				len -= (oresid - (uint64_t)uiop->uio_resid);
5940 			else if (mirrorcnt > 1 && rwaccess ==
5941 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5942 				/*
5943 				 * In case the rpc gets retried, roll the
5944 				 * uio fields changed by nfsm_uiombuflist()
5945 				 * back.
5946 				 */
5947 				uiop->uio_offset = offs;
5948 				uiop->uio_resid = resid;
5949 				uiop->uio_iov->iov_base = iovbase;
5950 				uiop->uio_iov->iov_len = iovlen;
5951 			}
5952 		}
5953 	}
5954 	if (lckp != NULL)
5955 		nfscl_lockderef(lckp);
5956 	NFSFREECRED(newcred);
5957 	nfscl_rellayout(layp, 0);
5958 	nfscl_relref(nmp);
5959 	return (error);
5960 }
5961 
5962 /*
5963  * Make a copy of the mbuf chain and add an mbuf for null padding, as required.
5964  */
5965 static struct mbuf *
5966 nfsm_copym(struct mbuf *m, int off, int xfer)
5967 {
5968 	struct mbuf *m2, *m3, *m4;
5969 	uint32_t *tl;
5970 	int rem;
5971 
5972 	m2 = m_copym(m, off, xfer, M_WAITOK);
5973 	rem = NFSM_RNDUP(xfer) - xfer;
5974 	if (rem > 0) {
5975 		/*
5976 		 * The zero padding to a multiple of 4 bytes is required by
5977 		 * the XDR. So that the mbufs copied by reference aren't
5978 		 * modified, add an mbuf with the zero'd bytes to the list.
5979 		 * rem will be a maximum of 3, so one zero'd uint32_t is
5980 		 * sufficient.
5981 		 */
5982 		m3 = m2;
5983 		while (m3->m_next != NULL)
5984 			m3 = m3->m_next;
5985 		NFSMGET(m4);
5986 		tl = mtod(m4, uint32_t *);
5987 		*tl = 0;
5988 		m4->m_len = rem;
5989 		m3->m_next = m4;
5990 	}
5991 	return (m2);
5992 }
5993 
5994 /*
5995  * Find a file layout that will handle the first bytes of the requested
5996  * range and return the information from it needed to the I/O operation.
5997  */
5998 int
5999 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
6000     struct nfsclflayout **retflpp)
6001 {
6002 	struct nfsclflayout *flp, *nflp, *rflp;
6003 	uint32_t rw;
6004 
6005 	rflp = NULL;
6006 	rw = rwaccess;
6007 	/* For reading, do the Read list first and then the Write list. */
6008 	do {
6009 		if (rw == NFSV4OPEN_ACCESSREAD)
6010 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6011 		else
6012 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6013 		while (flp != NULL) {
6014 			nflp = LIST_NEXT(flp, nfsfl_list);
6015 			if (flp->nfsfl_off > off)
6016 				break;
6017 			if (flp->nfsfl_end > off &&
6018 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6019 				rflp = flp;
6020 			flp = nflp;
6021 		}
6022 		if (rw == NFSV4OPEN_ACCESSREAD)
6023 			rw = NFSV4OPEN_ACCESSWRITE;
6024 		else
6025 			rw = 0;
6026 	} while (rw != 0);
6027 	if (rflp != NULL) {
6028 		/* This one covers the most bytes starting at off. */
6029 		*retflpp = rflp;
6030 		return (0);
6031 	}
6032 	return (EIO);
6033 }
6034 
6035 /*
6036  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6037  */
6038 static int
6039 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6040     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6041     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6042     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6043 {
6044 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6045 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6046 	struct nfsnode *np;
6047 	struct nfsfh *fhp;
6048 	struct nfsclds **dspp;
6049 
6050 	np = VTONFS(vp);
6051 	rel_off = off - flp->nfsfl_patoff;
6052 	stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
6053 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6054 	    dp->nfsdi_stripecnt;
6055 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6056 	error = 0;
6057 
6058 	/* Loop around, doing I/O for each stripe unit. */
6059 	while (len > 0 && error == 0) {
6060 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6061 		dspp = nfsfldi_addr(dp, stripe_index);
6062 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6063 			minorvers = NFSV42_MINORVERSION;
6064 		else
6065 			minorvers = NFSV41_MINORVERSION;
6066 		if (len > transfer && docommit == 0)
6067 			xfer = transfer;
6068 		else
6069 			xfer = len;
6070 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6071 			/* Dense layout. */
6072 			if (stripe_pos >= flp->nfsfl_fhcnt)
6073 				return (EIO);
6074 			fhp = flp->nfsfl_fh[stripe_pos];
6075 			io_off = (rel_off / (stripe_unit_size *
6076 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6077 			    rel_off % stripe_unit_size;
6078 		} else {
6079 			/* Sparse layout. */
6080 			if (flp->nfsfl_fhcnt > 1) {
6081 				if (stripe_index >= flp->nfsfl_fhcnt)
6082 					return (EIO);
6083 				fhp = flp->nfsfl_fh[stripe_index];
6084 			} else if (flp->nfsfl_fhcnt == 1)
6085 				fhp = flp->nfsfl_fh[0];
6086 			else
6087 				fhp = np->n_fhp;
6088 			io_off = off;
6089 		}
6090 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6091 			commit_thru_mds = 1;
6092 			if (docommit != 0)
6093 				error = EIO;
6094 		} else {
6095 			commit_thru_mds = 0;
6096 			NFSLOCKNODE(np);
6097 			np->n_flag |= NDSCOMMIT;
6098 			NFSUNLOCKNODE(np);
6099 		}
6100 		if (docommit != 0) {
6101 			if (error == 0)
6102 				error = nfsrpc_commitds(vp, io_off, xfer,
6103 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6104 			if (error == 0) {
6105 				/*
6106 				 * Set both eof and uio_resid = 0 to end any
6107 				 * loops.
6108 				 */
6109 				*eofp = 1;
6110 				uiop->uio_resid = 0;
6111 			} else {
6112 				NFSLOCKNODE(np);
6113 				np->n_flag &= ~NDSCOMMIT;
6114 				NFSUNLOCKNODE(np);
6115 			}
6116 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6117 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6118 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6119 		else {
6120 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6121 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6122 			    0, NFS_VER4, minorvers, cred, p);
6123 			if (error == 0) {
6124 				NFSLOCKCLSTATE();
6125 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6126 				NFSUNLOCKCLSTATE();
6127 			}
6128 		}
6129 		if (error == 0) {
6130 			transfer = stripe_unit_size;
6131 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6132 			len -= xfer;
6133 			off += xfer;
6134 		}
6135 	}
6136 	return (error);
6137 }
6138 
6139 /*
6140  * Do I/O using an NFSv4.1 flex file layout.
6141  */
6142 static int
6143 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6144     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6145     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6146     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6147     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6148 {
6149 	uint64_t transfer, xfer;
6150 	int error, rel_off;
6151 	struct nfsnode *np;
6152 	struct nfsfh *fhp;
6153 	struct nfsclds **dspp;
6154 	struct ucred *tcred;
6155 	struct mbuf *m;
6156 
6157 	np = VTONFS(vp);
6158 	error = 0;
6159 	rel_off = 0;
6160 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6161 	    (uintmax_t)len);
6162 	/* Loop around, doing I/O for each stripe unit. */
6163 	while (len > 0 && error == 0) {
6164 		dspp = nfsfldi_addr(dp, 0);
6165 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6166 		stateidp = &flp->nfsfl_ffm[mirror].st;
6167 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6168 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6169 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6170 			tcred = NFSNEWCRED(cred);
6171 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6172 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6173 			tcred->cr_ngroups = 1;
6174 		} else
6175 			tcred = cred;
6176 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6177 			transfer = dp->nfsdi_rsize;
6178 		else
6179 			transfer = dp->nfsdi_wsize;
6180 		NFSLOCKNODE(np);
6181 		np->n_flag |= NDSCOMMIT;
6182 		NFSUNLOCKNODE(np);
6183 		if (len > transfer && docommit == 0)
6184 			xfer = transfer;
6185 		else
6186 			xfer = len;
6187 		if (docommit != 0) {
6188 			if (error == 0) {
6189 				/*
6190 				 * Do last mirrored DS commit with this thread.
6191 				 */
6192 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6193 					error = nfsio_commitds(vp, off, xfer,
6194 					    *dspp, fhp, dp->nfsdi_vers,
6195 					    dp->nfsdi_minorvers, drpc, tcred,
6196 					    p);
6197 				else
6198 					error = nfsrpc_commitds(vp, off, xfer,
6199 					    *dspp, fhp, dp->nfsdi_vers,
6200 					    dp->nfsdi_minorvers, tcred, p);
6201 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6202 				if (error != 0 && error != EACCES && error !=
6203 				    ESTALE) {
6204 					NFSCL_DEBUG(4,
6205 					    "DS layreterr for commit\n");
6206 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6207 					    lyp, *dspp);
6208 				}
6209 			}
6210 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6211 			if (error == 0) {
6212 				/*
6213 				 * Set both eof and uio_resid = 0 to end any
6214 				 * loops.
6215 				 */
6216 				*eofp = 1;
6217 				uiop->uio_resid = 0;
6218 			} else {
6219 				NFSLOCKNODE(np);
6220 				np->n_flag &= ~NDSCOMMIT;
6221 				NFSUNLOCKNODE(np);
6222 			}
6223 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6224 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6225 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6226 			    dp->nfsdi_minorvers, tcred, p);
6227 			NFSCL_DEBUG(4, "readds=%d\n", error);
6228 			if (error != 0 && error != EACCES && error != ESTALE) {
6229 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6230 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6231 				    *dspp);
6232 			}
6233 		} else {
6234 			if (flp->nfsfl_mirrorcnt == 1) {
6235 				error = nfsrpc_writeds(vp, uiop, iomode,
6236 				    must_commit, stateidp, *dspp, off, xfer,
6237 				    fhp, 0, 1, dp->nfsdi_vers,
6238 				    dp->nfsdi_minorvers, tcred, p);
6239 				if (error == 0) {
6240 					NFSLOCKCLSTATE();
6241 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6242 					NFSUNLOCKCLSTATE();
6243 				}
6244 			} else {
6245 				m = nfsm_copym(mp, rel_off, xfer);
6246 				NFSCL_DEBUG(4, "mcopy reloff=%d xfer=%jd\n",
6247 				    rel_off, (uintmax_t)xfer);
6248 				/*
6249 				 * Do last write to a mirrored DS with this
6250 				 * thread.
6251 				 */
6252 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6253 					error = nfsio_writedsmir(vp, iomode,
6254 					    must_commit, stateidp, *dspp, off,
6255 					    xfer, fhp, m, dp->nfsdi_vers,
6256 					    dp->nfsdi_minorvers, drpc, tcred,
6257 					    p);
6258 				else
6259 					error = nfsrpc_writedsmir(vp, iomode,
6260 					    must_commit, stateidp, *dspp, off,
6261 					    xfer, fhp, m, dp->nfsdi_vers,
6262 					    dp->nfsdi_minorvers, tcred, p);
6263 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6264 				if (error != 0 && error != EACCES && error !=
6265 				    ESTALE) {
6266 					NFSCL_DEBUG(4,
6267 					    "DS layreterr for write\n");
6268 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6269 					    lyp, *dspp);
6270 				}
6271 			}
6272 		}
6273 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6274 		if (error == 0) {
6275 			len -= xfer;
6276 			off += xfer;
6277 			rel_off += xfer;
6278 		}
6279 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6280 			NFSFREECRED(tcred);
6281 	}
6282 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6283 	return (error);
6284 }
6285 
6286 /*
6287  * The actual read RPC done to a DS.
6288  */
6289 static int
6290 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6291     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6292     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6293 {
6294 	uint32_t *tl;
6295 	int attrflag, error, retlen;
6296 	struct nfsrv_descript nfsd;
6297 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6298 	struct nfsrv_descript *nd = &nfsd;
6299 	struct nfssockreq *nrp;
6300 	struct nfsvattr na;
6301 
6302 	nd->nd_mrep = NULL;
6303 	if (vers == 0 || vers == NFS_VER4) {
6304 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6305 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6306 		vers = NFS_VER4;
6307 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6308 		if (flex != 0)
6309 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6310 		else
6311 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6312 	} else {
6313 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6314 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6315 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6316 	}
6317 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6318 	txdr_hyper(io_off, tl);
6319 	*(tl + 2) = txdr_unsigned(len);
6320 	nrp = dsp->nfsclds_sockp;
6321 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6322 	if (nrp == NULL)
6323 		/* If NULL, use the MDS socket. */
6324 		nrp = &nmp->nm_sockreq;
6325 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6326 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6327 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6328 	    error);
6329 	if (error != 0)
6330 		return (error);
6331 	if (vers == NFS_VER3) {
6332 		error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6333 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6334 		if (error != 0)
6335 			goto nfsmout;
6336 	}
6337 	if (nd->nd_repstat != 0) {
6338 		error = nd->nd_repstat;
6339 		goto nfsmout;
6340 	}
6341 	if (vers == NFS_VER3) {
6342 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6343 		*eofp = fxdr_unsigned(int, *(tl + 1));
6344 	} else {
6345 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6346 		*eofp = fxdr_unsigned(int, *tl);
6347 	}
6348 	NFSM_STRSIZ(retlen, len);
6349 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6350 	error = nfsm_mbufuio(nd, uiop, retlen);
6351 nfsmout:
6352 	if (nd->nd_mrep != NULL)
6353 		m_freem(nd->nd_mrep);
6354 	return (error);
6355 }
6356 
6357 /*
6358  * The actual write RPC done to a DS.
6359  */
6360 static int
6361 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6362     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6363     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6364     struct ucred *cred, NFSPROC_T *p)
6365 {
6366 	uint32_t *tl;
6367 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6368 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6369 	int32_t backup;
6370 	struct nfsrv_descript nfsd;
6371 	struct nfsrv_descript *nd = &nfsd;
6372 	struct nfssockreq *nrp;
6373 	struct nfsvattr na;
6374 
6375 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6376 	nd->nd_mrep = NULL;
6377 	if (vers == 0 || vers == NFS_VER4) {
6378 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6379 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6380 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6381 		vers = NFS_VER4;
6382 		if (flex != 0)
6383 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6384 		else
6385 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6386 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6387 	} else {
6388 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6389 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6390 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6391 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6392 	}
6393 	txdr_hyper(io_off, tl);
6394 	tl += 2;
6395 	if (vers == NFS_VER3)
6396 		*tl++ = txdr_unsigned(len);
6397 	*tl++ = txdr_unsigned(*iomode);
6398 	*tl = txdr_unsigned(len);
6399 	nfsm_uiombuf(nd, uiop, len);
6400 	nrp = dsp->nfsclds_sockp;
6401 	if (nrp == NULL)
6402 		/* If NULL, use the MDS socket. */
6403 		nrp = &nmp->nm_sockreq;
6404 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6405 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6406 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6407 	    nd->nd_repstat);
6408 	if (error != 0)
6409 		return (error);
6410 	if (nd->nd_repstat != 0) {
6411 		/*
6412 		 * In case the rpc gets retried, roll
6413 		 * the uio fileds changed by nfsm_uiombuf()
6414 		 * back.
6415 		 */
6416 		uiop->uio_offset -= len;
6417 		uiop->uio_resid += len;
6418 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6419 		uiop->uio_iov->iov_len += len;
6420 		error = nd->nd_repstat;
6421 	} else {
6422 		if (vers == NFS_VER3) {
6423 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6424 			    NULL);
6425 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6426 			if (error != 0)
6427 				goto nfsmout;
6428 		}
6429 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6430 		rlen = fxdr_unsigned(int, *tl++);
6431 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6432 		if (rlen == 0) {
6433 			error = NFSERR_IO;
6434 			goto nfsmout;
6435 		} else if (rlen < len) {
6436 			backup = len - rlen;
6437 			uiop->uio_iov->iov_base =
6438 			    (char *)uiop->uio_iov->iov_base - backup;
6439 			uiop->uio_iov->iov_len += backup;
6440 			uiop->uio_offset -= backup;
6441 			uiop->uio_resid += backup;
6442 			len = rlen;
6443 		}
6444 		commit = fxdr_unsigned(int, *tl++);
6445 
6446 		/*
6447 		 * Return the lowest commitment level
6448 		 * obtained by any of the RPCs.
6449 		 */
6450 		if (committed == NFSWRITE_FILESYNC)
6451 			committed = commit;
6452 		else if (committed == NFSWRITE_DATASYNC &&
6453 		    commit == NFSWRITE_UNSTABLE)
6454 			committed = commit;
6455 		if (commit_thru_mds != 0) {
6456 			NFSLOCKMNT(nmp);
6457 			if (!NFSHASWRITEVERF(nmp)) {
6458 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6459 				NFSSETWRITEVERF(nmp);
6460 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6461 				*must_commit = 1;
6462 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6463 			}
6464 			NFSUNLOCKMNT(nmp);
6465 		} else {
6466 			NFSLOCKDS(dsp);
6467 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6468 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6469 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6470 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6471 				*must_commit = 1;
6472 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6473 			}
6474 			NFSUNLOCKDS(dsp);
6475 		}
6476 	}
6477 nfsmout:
6478 	if (nd->nd_mrep != NULL)
6479 		m_freem(nd->nd_mrep);
6480 	*iomode = committed;
6481 	if (nd->nd_repstat != 0 && error == 0)
6482 		error = nd->nd_repstat;
6483 	return (error);
6484 }
6485 
6486 /*
6487  * The actual write RPC done to a DS.
6488  * This variant is called from a separate kernel process for mirrors.
6489  * Any short write is considered an IO error.
6490  */
6491 static int
6492 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6493     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6494     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6495     struct ucred *cred, NFSPROC_T *p)
6496 {
6497 	uint32_t *tl;
6498 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6499 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6500 	struct nfsrv_descript nfsd;
6501 	struct nfsrv_descript *nd = &nfsd;
6502 	struct nfssockreq *nrp;
6503 	struct nfsvattr na;
6504 
6505 	nd->nd_mrep = NULL;
6506 	if (vers == 0 || vers == NFS_VER4) {
6507 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6508 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6509 		vers = NFS_VER4;
6510 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6511 		    minorvers);
6512 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6513 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6514 	} else {
6515 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6516 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6517 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6518 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6519 	}
6520 	txdr_hyper(io_off, tl);
6521 	tl += 2;
6522 	if (vers == NFS_VER3)
6523 		*tl++ = txdr_unsigned(len);
6524 	*tl++ = txdr_unsigned(*iomode);
6525 	*tl = txdr_unsigned(len);
6526 	if (len > 0) {
6527 		/* Put data in mbuf chain. */
6528 		nd->nd_mb->m_next = m;
6529 		/* Set nd_mb and nd_bpos to end of data. */
6530 		while (m->m_next != NULL)
6531 			m = m->m_next;
6532 		nd->nd_mb = m;
6533 		nd->nd_bpos = mtod(m, char *) + m->m_len;
6534 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: lastmb len=%d\n", m->m_len);
6535 	}
6536 	nrp = dsp->nfsclds_sockp;
6537 	if (nrp == NULL)
6538 		/* If NULL, use the MDS socket. */
6539 		nrp = &nmp->nm_sockreq;
6540 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6541 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6542 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6543 	    nd->nd_repstat);
6544 	if (error != 0)
6545 		return (error);
6546 	if (nd->nd_repstat != 0)
6547 		error = nd->nd_repstat;
6548 	else {
6549 		if (vers == NFS_VER3) {
6550 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6551 			    NULL);
6552 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6553 			    error);
6554 			if (error != 0)
6555 				goto nfsmout;
6556 		}
6557 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6558 		rlen = fxdr_unsigned(int, *tl++);
6559 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6560 		    rlen);
6561 		if (rlen != len) {
6562 			error = NFSERR_IO;
6563 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6564 			    len, rlen);
6565 			goto nfsmout;
6566 		}
6567 		commit = fxdr_unsigned(int, *tl++);
6568 
6569 		/*
6570 		 * Return the lowest commitment level
6571 		 * obtained by any of the RPCs.
6572 		 */
6573 		if (committed == NFSWRITE_FILESYNC)
6574 			committed = commit;
6575 		else if (committed == NFSWRITE_DATASYNC &&
6576 		    commit == NFSWRITE_UNSTABLE)
6577 			committed = commit;
6578 		NFSLOCKDS(dsp);
6579 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6580 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6581 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6582 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6583 			*must_commit = 1;
6584 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6585 		}
6586 		NFSUNLOCKDS(dsp);
6587 	}
6588 nfsmout:
6589 	if (nd->nd_mrep != NULL)
6590 		m_freem(nd->nd_mrep);
6591 	*iomode = committed;
6592 	if (nd->nd_repstat != 0 && error == 0)
6593 		error = nd->nd_repstat;
6594 	return (error);
6595 }
6596 
6597 /*
6598  * Start up the thread that will execute nfsrpc_writedsmir().
6599  */
6600 static void
6601 start_writedsmir(void *arg, int pending)
6602 {
6603 	struct nfsclwritedsdorpc *drpc;
6604 
6605 	drpc = (struct nfsclwritedsdorpc *)arg;
6606 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6607 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6608 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6609 	    drpc->p);
6610 	drpc->done = 1;
6611 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6612 }
6613 
6614 /*
6615  * Set up the write DS mirror call for the pNFS I/O thread.
6616  */
6617 static int
6618 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6619     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6620     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6621     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6622 {
6623 	int error, ret;
6624 
6625 	error = 0;
6626 	drpc->done = 0;
6627 	drpc->vp = vp;
6628 	drpc->iomode = *iomode;
6629 	drpc->must_commit = *must_commit;
6630 	drpc->stateidp = stateidp;
6631 	drpc->dsp = dsp;
6632 	drpc->off = off;
6633 	drpc->len = len;
6634 	drpc->fhp = fhp;
6635 	drpc->m = m;
6636 	drpc->vers = vers;
6637 	drpc->minorvers = minorvers;
6638 	drpc->cred = cred;
6639 	drpc->p = p;
6640 	drpc->inprog = 0;
6641 	ret = EIO;
6642 	if (nfs_pnfsiothreads != 0) {
6643 		ret = nfs_pnfsio(start_writedsmir, drpc);
6644 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6645 	}
6646 	if (ret != 0)
6647 		error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6648 		    dsp, off, len, fhp, m, vers, minorvers, cred, p);
6649 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6650 	return (error);
6651 }
6652 
6653 /*
6654  * Free up the nfsclds structure.
6655  */
6656 void
6657 nfscl_freenfsclds(struct nfsclds *dsp)
6658 {
6659 	int i;
6660 
6661 	if (dsp == NULL)
6662 		return;
6663 	if (dsp->nfsclds_sockp != NULL) {
6664 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6665 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6666 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6667 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6668 	}
6669 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
6670 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6671 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6672 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6673 			m_freem(
6674 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6675 	}
6676 	free(dsp, M_NFSCLDS);
6677 }
6678 
6679 static enum nfsclds_state
6680 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6681     struct nfsclds **retdspp, uint32_t *sequencep)
6682 {
6683 	struct nfsclds *dsp;
6684 	int fndseq;
6685 
6686 	/*
6687 	 * Search the list of nfsclds structures for one with the same
6688 	 * server.
6689 	 */
6690 	fndseq = 0;
6691 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6692 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6693 		    dsp->nfsclds_servownlen != 0 &&
6694 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6695 		    dsp->nfsclds_servownlen) &&
6696 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
6697 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6698 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
6699 			    dsp->nfsclds_flags);
6700 			if (fndseq == 0) {
6701 				/* Get sequenceid# from first entry. */
6702 				*sequencep =
6703 				    dsp->nfsclds_sess.nfsess_sequenceid;
6704 				fndseq = 1;
6705 			}
6706 			/* Server major id matches. */
6707 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6708 				*retdspp = dsp;
6709 				return (NFSDSP_USETHISSESSION);
6710 			}
6711 
6712 		}
6713 	}
6714 	if (fndseq != 0)
6715 		return (NFSDSP_SEQTHISSESSION);
6716 	return (NFSDSP_NOTFOUND);
6717 }
6718 
6719 /*
6720  * NFS commit rpc to a NFSv4.1 DS.
6721  */
6722 static int
6723 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6724     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6725     NFSPROC_T *p)
6726 {
6727 	uint32_t *tl;
6728 	struct nfsrv_descript nfsd, *nd = &nfsd;
6729 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6730 	struct nfssockreq *nrp;
6731 	struct nfsvattr na;
6732 	int attrflag, error;
6733 
6734 	nd->nd_mrep = NULL;
6735 	if (vers == 0 || vers == NFS_VER4) {
6736 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6737 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6738 		vers = NFS_VER4;
6739 	} else
6740 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6741 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6742 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6743 	    minorvers);
6744 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6745 	txdr_hyper(offset, tl);
6746 	tl += 2;
6747 	*tl = txdr_unsigned(cnt);
6748 	nrp = dsp->nfsclds_sockp;
6749 	if (nrp == NULL)
6750 		/* If NULL, use the MDS socket. */
6751 		nrp = &nmp->nm_sockreq;
6752 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6753 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6754 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6755 	    nd->nd_repstat);
6756 	if (error != 0)
6757 		return (error);
6758 	if (nd->nd_repstat == 0) {
6759 		if (vers == NFS_VER3) {
6760 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6761 			    NULL);
6762 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6763 			if (error != 0)
6764 				goto nfsmout;
6765 		}
6766 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6767 		NFSLOCKDS(dsp);
6768 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6769 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6770 			error = NFSERR_STALEWRITEVERF;
6771 		}
6772 		NFSUNLOCKDS(dsp);
6773 	}
6774 nfsmout:
6775 	if (error == 0 && nd->nd_repstat != 0)
6776 		error = nd->nd_repstat;
6777 	m_freem(nd->nd_mrep);
6778 	return (error);
6779 }
6780 
6781 /*
6782  * Start up the thread that will execute nfsrpc_commitds().
6783  */
6784 static void
6785 start_commitds(void *arg, int pending)
6786 {
6787 	struct nfsclwritedsdorpc *drpc;
6788 
6789 	drpc = (struct nfsclwritedsdorpc *)arg;
6790 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6791 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6792 	    drpc->p);
6793 	drpc->done = 1;
6794 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6795 }
6796 
6797 /*
6798  * Set up the commit DS mirror call for the pNFS I/O thread.
6799  */
6800 static int
6801 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6802     struct nfsfh *fhp, int vers, int minorvers,
6803     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6804 {
6805 	int error, ret;
6806 
6807 	error = 0;
6808 	drpc->done = 0;
6809 	drpc->vp = vp;
6810 	drpc->off = offset;
6811 	drpc->len = cnt;
6812 	drpc->dsp = dsp;
6813 	drpc->fhp = fhp;
6814 	drpc->vers = vers;
6815 	drpc->minorvers = minorvers;
6816 	drpc->cred = cred;
6817 	drpc->p = p;
6818 	drpc->inprog = 0;
6819 	ret = EIO;
6820 	if (nfs_pnfsiothreads != 0) {
6821 		ret = nfs_pnfsio(start_commitds, drpc);
6822 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6823 	}
6824 	if (ret != 0)
6825 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6826 		    minorvers, cred, p);
6827 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6828 	return (error);
6829 }
6830 
6831 /*
6832  * NFS Advise rpc
6833  */
6834 APPLESTATIC int
6835 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
6836     struct ucred *cred, NFSPROC_T *p)
6837 {
6838 	u_int32_t *tl;
6839 	struct nfsrv_descript nfsd, *nd = &nfsd;
6840 	nfsattrbit_t hints;
6841 	int error;
6842 
6843 	NFSZERO_ATTRBIT(&hints);
6844 	if (advise == POSIX_FADV_WILLNEED)
6845 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6846 	else if (advise == POSIX_FADV_DONTNEED)
6847 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6848 	else
6849 		return (0);
6850 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
6851 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6852 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
6853 	txdr_hyper(offset, tl);
6854 	tl += 2;
6855 	txdr_hyper(cnt, tl);
6856 	nfsrv_putattrbit(nd, &hints);
6857 	error = nfscl_request(nd, vp, p, cred, NULL);
6858 	if (error != 0)
6859 		return (error);
6860 	if (nd->nd_repstat != 0)
6861 		error = nd->nd_repstat;
6862 	m_freem(nd->nd_mrep);
6863 	return (error);
6864 }
6865 
6866 #ifdef notyet
6867 /*
6868  * NFS advise rpc to a NFSv4.2 DS.
6869  */
6870 static int
6871 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6872     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6873     struct ucred *cred, NFSPROC_T *p)
6874 {
6875 	uint32_t *tl;
6876 	struct nfsrv_descript nfsd, *nd = &nfsd;
6877 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6878 	struct nfssockreq *nrp;
6879 	nfsattrbit_t hints;
6880 	int error;
6881 
6882 	/* For NFS DSs prior to NFSv4.2, just return OK. */
6883 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
6884 		return (0);
6885 	NFSZERO_ATTRBIT(&hints);
6886 	if (advise == POSIX_FADV_WILLNEED)
6887 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6888 	else if (advise == POSIX_FADV_DONTNEED)
6889 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6890 	else
6891 		return (0);
6892 	nd->nd_mrep = NULL;
6893 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
6894 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6895 	vers = NFS_VER4;
6896 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
6897 	    minorvers);
6898 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6899 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6900 	txdr_hyper(offset, tl);
6901 	tl += 2;
6902 	*tl = txdr_unsigned(cnt);
6903 	nfsrv_putattrbit(nd, &hints);
6904 	nrp = dsp->nfsclds_sockp;
6905 	if (nrp == NULL)
6906 		/* If NULL, use the MDS socket. */
6907 		nrp = &nmp->nm_sockreq;
6908 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6909 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6910 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
6911 	    nd->nd_repstat);
6912 	if (error != 0)
6913 		return (error);
6914 	if (nd->nd_repstat != 0)
6915 		error = nd->nd_repstat;
6916 	m_freem(nd->nd_mrep);
6917 	return (error);
6918 }
6919 
6920 /*
6921  * Start up the thread that will execute nfsrpc_commitds().
6922  */
6923 static void
6924 start_adviseds(void *arg, int pending)
6925 {
6926 	struct nfsclwritedsdorpc *drpc;
6927 
6928 	drpc = (struct nfsclwritedsdorpc *)arg;
6929 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
6930 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
6931 	    drpc->cred, drpc->p);
6932 	drpc->done = 1;
6933 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
6934 }
6935 
6936 /*
6937  * Set up the commit DS mirror call for the pNFS I/O thread.
6938  */
6939 static int
6940 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6941     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6942     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6943 {
6944 	int error, ret;
6945 
6946 	error = 0;
6947 	drpc->done = 0;
6948 	drpc->vp = vp;
6949 	drpc->off = offset;
6950 	drpc->len = cnt;
6951 	drpc->advise = advise;
6952 	drpc->dsp = dsp;
6953 	drpc->fhp = fhp;
6954 	drpc->vers = vers;
6955 	drpc->minorvers = minorvers;
6956 	drpc->cred = cred;
6957 	drpc->p = p;
6958 	drpc->inprog = 0;
6959 	ret = EIO;
6960 	if (nfs_pnfsiothreads != 0) {
6961 		ret = nfs_pnfsio(start_adviseds, drpc);
6962 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
6963 	}
6964 	if (ret != 0)
6965 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
6966 		    minorvers, cred, p);
6967 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
6968 	return (error);
6969 }
6970 #endif	/* notyet */
6971 
6972 /*
6973  * Do the Allocate operation, retrying for recovery.
6974  */
6975 APPLESTATIC int
6976 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
6977     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
6978 {
6979 	int error, expireret = 0, retrycnt, nostateid;
6980 	uint32_t clidrev = 0;
6981 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6982 	struct nfsfh *nfhp = NULL;
6983 	nfsv4stateid_t stateid;
6984 	off_t tmp_off;
6985 	void *lckp;
6986 
6987 	if (len < 0)
6988 		return (EINVAL);
6989 	if (len == 0)
6990 		return (0);
6991 	tmp_off = off + len;
6992 	NFSLOCKMNT(nmp);
6993 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
6994 		NFSUNLOCKMNT(nmp);
6995 		return (EFBIG);
6996 	}
6997 	if (nmp->nm_clp != NULL)
6998 		clidrev = nmp->nm_clp->nfsc_clientidrev;
6999 	NFSUNLOCKMNT(nmp);
7000 	nfhp = VTONFS(vp)->n_fhp;
7001 	retrycnt = 0;
7002 	do {
7003 		lckp = NULL;
7004 		nostateid = 0;
7005 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7006 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7007 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7008 		    stateid.other[2] == 0) {
7009 			nostateid = 1;
7010 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7011 		}
7012 
7013 		/*
7014 		 * Not finding a stateid should probably never happen,
7015 		 * but just return an error for this case.
7016 		 */
7017 		if (nostateid != 0)
7018 			error = EIO;
7019 		else
7020 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7021 			    nap, attrflagp, cred, p, stuff);
7022 		if (error == NFSERR_STALESTATEID)
7023 			nfscl_initiate_recovery(nmp->nm_clp);
7024 		if (lckp != NULL)
7025 			nfscl_lockderef(lckp);
7026 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7027 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7028 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7029 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7030 		} else if ((error == NFSERR_EXPIRED ||
7031 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7032 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7033 		}
7034 		retrycnt++;
7035 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7036 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7037 	    error == NFSERR_STALEDONTRECOVER ||
7038 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7039 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7040 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7041 	if (error != 0 && retrycnt >= 4)
7042 		error = EIO;
7043 	return (error);
7044 }
7045 
7046 /*
7047  * The allocate RPC.
7048  */
7049 static int
7050 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7051     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p,
7052     void *stuff)
7053 {
7054 	uint32_t *tl;
7055 	int error;
7056 	struct nfsrv_descript nfsd;
7057 	struct nfsrv_descript *nd = &nfsd;
7058 	nfsattrbit_t attrbits;
7059 
7060 	*attrflagp = 0;
7061 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7062 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7063 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7064 	txdr_hyper(off, tl); tl += 2;
7065 	txdr_hyper(len, tl); tl += 2;
7066 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7067 	NFSGETATTR_ATTRBIT(&attrbits);
7068 	nfsrv_putattrbit(nd, &attrbits);
7069 	error = nfscl_request(nd, vp, p, cred, stuff);
7070 	if (error != 0)
7071 		return (error);
7072 	if (nd->nd_repstat == 0) {
7073 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7074 		error = nfsm_loadattr(nd, nap);
7075 		if (error == 0)
7076 			*attrflagp = NFS_LATTR_NOSHRINK;
7077 	} else
7078 		error = nd->nd_repstat;
7079 nfsmout:
7080 	m_freem(nd->nd_mrep);
7081 	return (error);
7082 }
7083 
7084 /*
7085  * Set up the XDR arguments for the LayoutGet operation.
7086  */
7087 static void
7088 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7089     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7090     int layoutlen, int usecurstateid)
7091 {
7092 	uint32_t *tl;
7093 
7094 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7095 	    NFSX_STATEID);
7096 	*tl++ = newnfs_false;		/* Don't signal availability. */
7097 	*tl++ = txdr_unsigned(layouttype);
7098 	*tl++ = txdr_unsigned(iomode);
7099 	txdr_hyper(offset, tl);
7100 	tl += 2;
7101 	txdr_hyper(len, tl);
7102 	tl += 2;
7103 	txdr_hyper(minlen, tl);
7104 	tl += 2;
7105 	if (usecurstateid != 0) {
7106 		/* Special stateid for Current stateid. */
7107 		*tl++ = txdr_unsigned(1);
7108 		*tl++ = 0;
7109 		*tl++ = 0;
7110 		*tl++ = 0;
7111 	} else {
7112 		*tl++ = txdr_unsigned(stateidp->seqid);
7113 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7114 		*tl++ = stateidp->other[0];
7115 		*tl++ = stateidp->other[1];
7116 		*tl++ = stateidp->other[2];
7117 	}
7118 	*tl = txdr_unsigned(layoutlen);
7119 }
7120 
7121 /*
7122  * Parse the reply for a successful LayoutGet operation.
7123  */
7124 static int
7125 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7126     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7127 {
7128 	uint32_t *tl;
7129 	struct nfsclflayout *flp, *prevflp, *tflp;
7130 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7131 	int m, mirrorcnt;
7132 	uint64_t retlen, off;
7133 	struct nfsfh *nfhp;
7134 	uint8_t *cp;
7135 	uid_t user;
7136 	gid_t grp;
7137 
7138 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7139 	error = 0;
7140 	flp = NULL;
7141 	gotiomode = -1;
7142 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7143 	if (*tl++ != 0)
7144 		*retonclosep = 1;
7145 	else
7146 		*retonclosep = 0;
7147 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7148 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7149 	    (int)stateidp->seqid);
7150 	stateidp->other[0] = *tl++;
7151 	stateidp->other[1] = *tl++;
7152 	stateidp->other[2] = *tl++;
7153 	cnt = fxdr_unsigned(int, *tl);
7154 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7155 	if (cnt <= 0 || cnt > 10000) {
7156 		/* Don't accept more than 10000 layouts in reply. */
7157 		error = NFSERR_BADXDR;
7158 		goto nfsmout;
7159 	}
7160 	for (i = 0; i < cnt; i++) {
7161 		/* Dissect to the layout type. */
7162 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7163 		    3 * NFSX_UNSIGNED);
7164 		off = fxdr_hyper(tl); tl += 2;
7165 		retlen = fxdr_hyper(tl); tl += 2;
7166 		iomode = fxdr_unsigned(int, *tl++);
7167 		laytype = fxdr_unsigned(int, *tl);
7168 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7169 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7170 		/* Ignore length of layout body for now. */
7171 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7172 			/* Parse the File layout up to fhcnt. */
7173 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7174 			    NFSX_HYPER + NFSX_V4DEVICEID);
7175 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7176 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7177 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7178 			if (fhcnt < 0 || fhcnt > 100) {
7179 				/* Don't accept more than 100 file handles. */
7180 				error = NFSERR_BADXDR;
7181 				goto nfsmout;
7182 			}
7183 			if (fhcnt > 0)
7184 				flp = malloc(sizeof(*flp) + fhcnt *
7185 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7186 				    M_WAITOK);
7187 			else
7188 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7189 				    M_WAITOK);
7190 			flp->nfsfl_flags = NFSFL_FILE;
7191 			flp->nfsfl_fhcnt = 0;
7192 			flp->nfsfl_devp = NULL;
7193 			flp->nfsfl_off = off;
7194 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7195 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7196 			else
7197 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7198 			flp->nfsfl_iomode = iomode;
7199 			if (gotiomode == -1)
7200 				gotiomode = flp->nfsfl_iomode;
7201 			/* Ignore layout body length for now. */
7202 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7203 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7204 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7205 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7206 			mtx_lock(&nmp->nm_mtx);
7207 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7208 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7209 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7210 			mtx_unlock(&nmp->nm_mtx);
7211 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7212 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7213 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7214 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7215 			for (j = 0; j < fhcnt; j++) {
7216 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7217 				nfhlen = fxdr_unsigned(int, *tl);
7218 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7219 					error = NFSERR_BADXDR;
7220 					goto nfsmout;
7221 				}
7222 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7223 				    M_NFSFH, M_WAITOK);
7224 				flp->nfsfl_fh[j] = nfhp;
7225 				flp->nfsfl_fhcnt++;
7226 				nfhp->nfh_len = nfhlen;
7227 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7228 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7229 			}
7230 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7231 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7232 			    NFSX_HYPER);
7233 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7234 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7235 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7236 				error = NFSERR_BADXDR;
7237 				goto nfsmout;
7238 			}
7239 			flp = malloc(sizeof(*flp) + mirrorcnt *
7240 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7241 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7242 			flp->nfsfl_mirrorcnt = mirrorcnt;
7243 			for (j = 0; j < mirrorcnt; j++)
7244 				flp->nfsfl_ffm[j].devp = NULL;
7245 			flp->nfsfl_off = off;
7246 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7247 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7248 			else
7249 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7250 			flp->nfsfl_iomode = iomode;
7251 			if (gotiomode == -1)
7252 				gotiomode = flp->nfsfl_iomode;
7253 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7254 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7255 			    (uintmax_t)flp->nfsfl_stripeunit);
7256 			for (j = 0; j < mirrorcnt; j++) {
7257 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7258 				k = fxdr_unsigned(int, *tl);
7259 				if (k < 1 || k > 128) {
7260 					error = NFSERR_BADXDR;
7261 					goto nfsmout;
7262 				}
7263 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7264 				for (l = 0; l < k; l++) {
7265 					NFSM_DISSECT(tl, uint32_t *,
7266 					    NFSX_V4DEVICEID + NFSX_STATEID +
7267 					    2 * NFSX_UNSIGNED);
7268 					if (l == 0) {
7269 						/* Just use the first server. */
7270 						NFSBCOPY(tl,
7271 						    flp->nfsfl_ffm[j].dev,
7272 						    NFSX_V4DEVICEID);
7273 						tl += (NFSX_V4DEVICEID /
7274 						    NFSX_UNSIGNED);
7275 						tl++;
7276 						flp->nfsfl_ffm[j].st.seqid =
7277 						    *tl++;
7278 						flp->nfsfl_ffm[j].st.other[0] =
7279 						    *tl++;
7280 						flp->nfsfl_ffm[j].st.other[1] =
7281 						    *tl++;
7282 						flp->nfsfl_ffm[j].st.other[2] =
7283 						    *tl++;
7284 						NFSCL_DEBUG(4, "st.seqid=%u "
7285 						 "st.o0=0x%x st.o1=0x%x "
7286 						 "st.o2=0x%x\n",
7287 						 flp->nfsfl_ffm[j].st.seqid,
7288 						 flp->nfsfl_ffm[j].st.other[0],
7289 						 flp->nfsfl_ffm[j].st.other[1],
7290 						 flp->nfsfl_ffm[j].st.other[2]);
7291 					} else
7292 						tl += ((NFSX_V4DEVICEID +
7293 						    NFSX_STATEID +
7294 						    NFSX_UNSIGNED) /
7295 						    NFSX_UNSIGNED);
7296 					fhcnt = fxdr_unsigned(int, *tl);
7297 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7298 					if (fhcnt < 1 ||
7299 					    fhcnt > NFSDEV_MAXVERS) {
7300 						error = NFSERR_BADXDR;
7301 						goto nfsmout;
7302 					}
7303 					for (m = 0; m < fhcnt; m++) {
7304 						NFSM_DISSECT(tl, uint32_t *,
7305 						    NFSX_UNSIGNED);
7306 						nfhlen = fxdr_unsigned(int,
7307 						    *tl);
7308 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7309 						    nfhlen);
7310 						if (nfhlen <= 0 || nfhlen >
7311 						    NFSX_V4FHMAX) {
7312 							error = NFSERR_BADXDR;
7313 							goto nfsmout;
7314 						}
7315 						NFSM_DISSECT(cp, uint8_t *,
7316 						    NFSM_RNDUP(nfhlen));
7317 						if (l == 0) {
7318 							flp->nfsfl_ffm[j].fhcnt
7319 							    = fhcnt;
7320 							nfhp = malloc(
7321 							    sizeof(*nfhp) +
7322 							    nfhlen - 1, M_NFSFH,
7323 							    M_WAITOK);
7324 							flp->nfsfl_ffm[j].fh[m]
7325 							    = nfhp;
7326 							nfhp->nfh_len = nfhlen;
7327 							NFSBCOPY(cp,
7328 							    nfhp->nfh_fh,
7329 							    nfhlen);
7330 							NFSCL_DEBUG(4,
7331 							    "got fh\n");
7332 						}
7333 					}
7334 					/* Now, get the ffsd_user/ffds_group. */
7335 					error = nfsrv_parseug(nd, 0, &user,
7336 					    &grp, curthread);
7337 					NFSCL_DEBUG(4, "after parseu=%d\n",
7338 					    error);
7339 					if (error == 0)
7340 						error = nfsrv_parseug(nd, 1,
7341 						    &user, &grp, curthread);
7342 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7343 					    grp);
7344 					if (error != 0)
7345 						goto nfsmout;
7346 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7347 					    user, grp);
7348 					if (l == 0) {
7349 						flp->nfsfl_ffm[j].user = user;
7350 						flp->nfsfl_ffm[j].group = grp;
7351 						NFSCL_DEBUG(4,
7352 						    "usr=%d grp=%d\n", user,
7353 						    grp);
7354 					}
7355 				}
7356 			}
7357 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7358 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7359 #ifdef notnow
7360 			/*
7361 			 * At this time, there is no flag.
7362 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7363 			 * added, or it may never exist?
7364 			 */
7365 			mtx_lock(&nmp->nm_mtx);
7366 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7367 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7368 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7369 			mtx_unlock(&nmp->nm_mtx);
7370 #endif
7371 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7372 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7373 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7374 		} else {
7375 			error = NFSERR_BADXDR;
7376 			goto nfsmout;
7377 		}
7378 		if (flp->nfsfl_iomode == gotiomode) {
7379 			/* Keep the list in increasing offset order. */
7380 			tflp = LIST_FIRST(flhp);
7381 			prevflp = NULL;
7382 			while (tflp != NULL &&
7383 			    tflp->nfsfl_off < flp->nfsfl_off) {
7384 				prevflp = tflp;
7385 				tflp = LIST_NEXT(tflp, nfsfl_list);
7386 			}
7387 			if (prevflp == NULL)
7388 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7389 			else
7390 				LIST_INSERT_AFTER(prevflp, flp,
7391 				    nfsfl_list);
7392 			NFSCL_DEBUG(4, "flp inserted\n");
7393 		} else {
7394 			printf("nfscl_layoutget(): got wrong iomode\n");
7395 			nfscl_freeflayout(flp);
7396 		}
7397 		flp = NULL;
7398 	}
7399 nfsmout:
7400 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7401 	if (error != 0 && flp != NULL)
7402 		nfscl_freeflayout(flp);
7403 	return (error);
7404 }
7405 
7406 /*
7407  * Parse a user/group digit string.
7408  */
7409 static int
7410 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7411     NFSPROC_T *p)
7412 {
7413 	uint32_t *tl;
7414 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7415 	uint32_t len = 0;
7416 	int error = 0;
7417 
7418 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7419 	len = fxdr_unsigned(uint32_t, *tl);
7420 	str = NULL;
7421 	if (len > NFSV4_OPAQUELIMIT) {
7422 		error = NFSERR_BADXDR;
7423 		goto nfsmout;
7424 	}
7425 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7426 	if (len == 0) {
7427 		if (dogrp != 0)
7428 			*gidp = GID_NOGROUP;
7429 		else
7430 			*uidp = UID_NOBODY;
7431 		return (0);
7432 	}
7433 	if (len > NFSV4_SMALLSTR)
7434 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7435 	else
7436 		str = str0;
7437 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7438 	NFSBCOPY(cp, str, len);
7439 	str[len] = '\0';
7440 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7441 	if (dogrp != 0)
7442 		error = nfsv4_strtogid(nd, str, len, gidp);
7443 	else
7444 		error = nfsv4_strtouid(nd, str, len, uidp);
7445 nfsmout:
7446 	if (len > NFSV4_SMALLSTR)
7447 		free(str, M_TEMP);
7448 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7449 	return (error);
7450 }
7451 
7452 /*
7453  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7454  * so that it does both an Open and a Layoutget.
7455  */
7456 static int
7457 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7458     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7459     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7460     struct ucred *cred, NFSPROC_T *p)
7461 {
7462 	struct nfscllayout *lyp;
7463 	struct nfsclflayout *flp;
7464 	struct nfsclflayouthead flh;
7465 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7466 	int layouttype, laystat;
7467 	nfsv4stateid_t stateid;
7468 	struct nfsclsession *tsep;
7469 
7470 	error = 0;
7471 	if (NFSHASFLEXFILE(nmp))
7472 		layouttype = NFSLAYOUT_FLEXFILE;
7473 	else
7474 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7475 	/*
7476 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7477 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7478 	 * flp == NULL.
7479 	 */
7480 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7481 	    &recalled);
7482 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7483 	if (lyp == NULL)
7484 		islocked = 0;
7485 	else if (flp != NULL)
7486 		islocked = 1;
7487 	else
7488 		islocked = 2;
7489 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7490 		LIST_INIT(&flh);
7491 		tsep = nfsmnt_mdssession(nmp);
7492 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7493 		    3 * NFSX_UNSIGNED);
7494 		if (lyp == NULL)
7495 			usecurstateid = 1;
7496 		else {
7497 			usecurstateid = 0;
7498 			stateid.seqid = lyp->nfsly_stateid.seqid;
7499 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7500 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7501 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7502 		}
7503 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7504 		    newfhp, newfhlen, mode, op, name, namelen,
7505 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7506 		    &retonclose, &flh, &laystat, cred, p);
7507 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7508 		    laystat, error);
7509 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7510 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7511 		    &islocked, cred, p);
7512 	} else
7513 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7514 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7515 	if (islocked == 2)
7516 		nfscl_rellayout(lyp, 1);
7517 	else if (islocked == 1)
7518 		nfscl_rellayout(lyp, 0);
7519 	return (error);
7520 }
7521 
7522 /*
7523  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7524  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7525  * handled by nfsrpc_openrpc().
7526  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7527  * can be NULL.
7528  */
7529 static int
7530 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7531     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7532     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7533     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7534     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7535     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7536 {
7537 	uint32_t *tl;
7538 	struct nfsrv_descript nfsd, *nd = &nfsd;
7539 	struct nfscldeleg *ndp = NULL;
7540 	struct nfsvattr nfsva;
7541 	struct nfsclsession *tsep;
7542 	uint32_t rflags, deleg;
7543 	nfsattrbit_t attrbits;
7544 	int error, ret, acesize, limitby, iomode;
7545 
7546 	*dpp = NULL;
7547 	*laystatp = ENXIO;
7548 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7549 	    0, 0);
7550 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7551 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7552 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7553 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7554 	tsep = nfsmnt_mdssession(nmp);
7555 	*tl++ = tsep->nfsess_clientid.lval[0];
7556 	*tl = tsep->nfsess_clientid.lval[1];
7557 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7558 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7559 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7560 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7561 	nfsm_strtom(nd, name, namelen);
7562 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7563 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7564 	NFSZERO_ATTRBIT(&attrbits);
7565 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7566 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7567 	nfsrv_putattrbit(nd, &attrbits);
7568 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7569 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7570 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7571 		iomode = NFSLAYOUTIOMODE_RW;
7572 	else
7573 		iomode = NFSLAYOUTIOMODE_READ;
7574 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7575 	    layouttype, layoutlen, usecurstateid);
7576 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7577 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7578 	if (error != 0)
7579 		return (error);
7580 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7581 	if (nd->nd_repstat != 0)
7582 		*laystatp = nd->nd_repstat;
7583 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7584 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7585 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7586 		    6 * NFSX_UNSIGNED);
7587 		op->nfso_stateid.seqid = *tl++;
7588 		op->nfso_stateid.other[0] = *tl++;
7589 		op->nfso_stateid.other[1] = *tl++;
7590 		op->nfso_stateid.other[2] = *tl;
7591 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7592 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7593 		if (error != 0)
7594 			goto nfsmout;
7595 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7596 		deleg = fxdr_unsigned(u_int32_t, *tl);
7597 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7598 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7599 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7600 			      NFSCLFLAGS_FIRSTDELEG))
7601 				op->nfso_own->nfsow_clp->nfsc_flags |=
7602 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7603 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7604 			    M_NFSCLDELEG, M_WAITOK);
7605 			LIST_INIT(&ndp->nfsdl_owner);
7606 			LIST_INIT(&ndp->nfsdl_lock);
7607 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7608 			ndp->nfsdl_fhlen = newfhlen;
7609 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7610 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
7611 			nfscl_lockinit(&ndp->nfsdl_rwlock);
7612 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7613 			    NFSX_UNSIGNED);
7614 			ndp->nfsdl_stateid.seqid = *tl++;
7615 			ndp->nfsdl_stateid.other[0] = *tl++;
7616 			ndp->nfsdl_stateid.other[1] = *tl++;
7617 			ndp->nfsdl_stateid.other[2] = *tl++;
7618 			ret = fxdr_unsigned(int, *tl);
7619 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7620 				ndp->nfsdl_flags = NFSCLDL_WRITE;
7621 				/*
7622 				 * Indicates how much the file can grow.
7623 				 */
7624 				NFSM_DISSECT(tl, u_int32_t *,
7625 				    3 * NFSX_UNSIGNED);
7626 				limitby = fxdr_unsigned(int, *tl++);
7627 				switch (limitby) {
7628 				case NFSV4OPEN_LIMITSIZE:
7629 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7630 					break;
7631 				case NFSV4OPEN_LIMITBLOCKS:
7632 					ndp->nfsdl_sizelimit =
7633 					    fxdr_unsigned(u_int64_t, *tl++);
7634 					ndp->nfsdl_sizelimit *=
7635 					    fxdr_unsigned(u_int64_t, *tl);
7636 					break;
7637 				default:
7638 					error = NFSERR_BADXDR;
7639 					goto nfsmout;
7640 				};
7641 			} else
7642 				ndp->nfsdl_flags = NFSCLDL_READ;
7643 			if (ret != 0)
7644 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
7645 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7646 			    &acesize, p);
7647 			if (error != 0)
7648 				goto nfsmout;
7649 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7650 			error = NFSERR_BADXDR;
7651 			goto nfsmout;
7652 		}
7653 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7654 		    nfscl_assumeposixlocks)
7655 			op->nfso_posixlock = 1;
7656 		else
7657 			op->nfso_posixlock = 0;
7658 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7659 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
7660 		if (*++tl == 0) {
7661 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7662 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7663 			    NULL, NULL, NULL, p, cred);
7664 			if (error != 0)
7665 				goto nfsmout;
7666 			if (ndp != NULL) {
7667 				ndp->nfsdl_change = nfsva.na_filerev;
7668 				ndp->nfsdl_modtime = nfsva.na_mtime;
7669 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7670 				*dpp = ndp;
7671 				ndp = NULL;
7672 			}
7673 			/*
7674 			 * At this point, the Open has succeeded, so set
7675 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
7676 			 * this function just won't return a layout.
7677 			 */
7678 			if (nd->nd_repstat == 0) {
7679 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7680 				*laystatp = fxdr_unsigned(int, *++tl);
7681 				if (*laystatp == 0) {
7682 					error = nfsrv_parselayoutget(nmp, nd,
7683 					    stateidp, retonclosep, flhp);
7684 					if (error != 0)
7685 						*laystatp = error;
7686 				}
7687 			} else
7688 				nd->nd_repstat = 0;	/* Return 0 for Open. */
7689 		}
7690 	}
7691 	if (nd->nd_repstat != 0 && error == 0)
7692 		error = nd->nd_repstat;
7693 nfsmout:
7694 	free(ndp, M_NFSCLDELEG);
7695 	m_freem(nd->nd_mrep);
7696 	return (error);
7697 }
7698 
7699 /*
7700  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7701  * Used only for mounts with pNFS enabled.
7702  */
7703 static int
7704 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7705     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7706     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7707     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7708     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7709     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7710     struct nfsclflayouthead *flhp, int *laystatp)
7711 {
7712 	uint32_t *tl;
7713 	int error = 0, deleg, newone, ret, acesize, limitby;
7714 	struct nfsrv_descript nfsd, *nd = &nfsd;
7715 	struct nfsclopen *op;
7716 	struct nfscldeleg *dp = NULL;
7717 	struct nfsnode *np;
7718 	struct nfsfh *nfhp;
7719 	struct nfsclsession *tsep;
7720 	nfsattrbit_t attrbits;
7721 	nfsv4stateid_t stateid;
7722 	struct nfsmount *nmp;
7723 
7724 	nmp = VFSTONFS(dvp->v_mount);
7725 	np = VTONFS(dvp);
7726 	*laystatp = ENXIO;
7727 	*unlockedp = 0;
7728 	*nfhpp = NULL;
7729 	*dpp = NULL;
7730 	*attrflagp = 0;
7731 	*dattrflagp = 0;
7732 	if (namelen > NFS_MAXNAMLEN)
7733 		return (ENAMETOOLONG);
7734 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7735 	/*
7736 	 * For V4, this is actually an Open op.
7737 	 */
7738 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7739 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
7740 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7741 	    NFSV4OPEN_ACCESSREAD);
7742 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7743 	tsep = nfsmnt_mdssession(nmp);
7744 	*tl++ = tsep->nfsess_clientid.lval[0];
7745 	*tl = tsep->nfsess_clientid.lval[1];
7746 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7747 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7748 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7749 	if ((fmode & O_EXCL) != 0) {
7750 		if (NFSHASSESSPERSIST(nmp)) {
7751 			/* Use GUARDED for persistent sessions. */
7752 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
7753 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7754 		} else {
7755 			/* Otherwise, use EXCLUSIVE4_1. */
7756 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7757 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7758 			*tl++ = cverf.lval[0];
7759 			*tl = cverf.lval[1];
7760 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7761 		}
7762 	} else {
7763 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7764 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
7765 	}
7766 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7767 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7768 	nfsm_strtom(nd, name, namelen);
7769 	/* Get the new file's handle and attributes, plus save the FH. */
7770 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7771 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7772 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
7773 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7774 	NFSGETATTR_ATTRBIT(&attrbits);
7775 	nfsrv_putattrbit(nd, &attrbits);
7776 	/* Get the directory's post-op attributes. */
7777 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7778 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
7779 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7780 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7781 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7782 	nfsrv_putattrbit(nd, &attrbits);
7783 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7784 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7785 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7786 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7787 	    layouttype, layoutlen, usecurstateid);
7788 	error = nfscl_request(nd, dvp, p, cred, dstuff);
7789 	if (error != 0)
7790 		return (error);
7791 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7792 	    error);
7793 	if (nd->nd_repstat != 0)
7794 		*laystatp = nd->nd_repstat;
7795 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7796 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7797 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7798 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7799 		    6 * NFSX_UNSIGNED);
7800 		stateid.seqid = *tl++;
7801 		stateid.other[0] = *tl++;
7802 		stateid.other[1] = *tl++;
7803 		stateid.other[2] = *tl;
7804 		nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7805 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7806 		deleg = fxdr_unsigned(int, *tl);
7807 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7808 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7809 			if (!(owp->nfsow_clp->nfsc_flags &
7810 			      NFSCLFLAGS_FIRSTDELEG))
7811 				owp->nfsow_clp->nfsc_flags |=
7812 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7813 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7814 			    M_NFSCLDELEG, M_WAITOK);
7815 			LIST_INIT(&dp->nfsdl_owner);
7816 			LIST_INIT(&dp->nfsdl_lock);
7817 			dp->nfsdl_clp = owp->nfsow_clp;
7818 			newnfs_copyincred(cred, &dp->nfsdl_cred);
7819 			nfscl_lockinit(&dp->nfsdl_rwlock);
7820 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7821 			    NFSX_UNSIGNED);
7822 			dp->nfsdl_stateid.seqid = *tl++;
7823 			dp->nfsdl_stateid.other[0] = *tl++;
7824 			dp->nfsdl_stateid.other[1] = *tl++;
7825 			dp->nfsdl_stateid.other[2] = *tl++;
7826 			ret = fxdr_unsigned(int, *tl);
7827 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7828 				dp->nfsdl_flags = NFSCLDL_WRITE;
7829 				/*
7830 				 * Indicates how much the file can grow.
7831 				 */
7832 				NFSM_DISSECT(tl, u_int32_t *,
7833 				    3 * NFSX_UNSIGNED);
7834 				limitby = fxdr_unsigned(int, *tl++);
7835 				switch (limitby) {
7836 				case NFSV4OPEN_LIMITSIZE:
7837 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
7838 					break;
7839 				case NFSV4OPEN_LIMITBLOCKS:
7840 					dp->nfsdl_sizelimit =
7841 					    fxdr_unsigned(u_int64_t, *tl++);
7842 					dp->nfsdl_sizelimit *=
7843 					    fxdr_unsigned(u_int64_t, *tl);
7844 					break;
7845 				default:
7846 					error = NFSERR_BADXDR;
7847 					goto nfsmout;
7848 				};
7849 			} else {
7850 				dp->nfsdl_flags = NFSCLDL_READ;
7851 			}
7852 			if (ret != 0)
7853 				dp->nfsdl_flags |= NFSCLDL_RECALL;
7854 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7855 			    &acesize, p);
7856 			if (error != 0)
7857 				goto nfsmout;
7858 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7859 			error = NFSERR_BADXDR;
7860 			goto nfsmout;
7861 		}
7862 
7863 		/* Now, we should have the status for the SaveFH. */
7864 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7865 		if (*++tl == 0) {
7866 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7867 			/*
7868 			 * Now, process the GetFH and Getattr for the newly
7869 			 * created file. nfscl_mtofh() will set
7870 			 * ND_NOMOREDATA if these weren't successful.
7871 			 */
7872 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7873 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7874 			if (error != 0)
7875 				goto nfsmout;
7876 		} else
7877 			nd->nd_flag |= ND_NOMOREDATA;
7878 		/* Now we have the PutFH and Getattr for the directory. */
7879 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7880 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7881 			if (*++tl != 0)
7882 				nd->nd_flag |= ND_NOMOREDATA;
7883 			else {
7884 				NFSM_DISSECT(tl, uint32_t *, 2 *
7885 				    NFSX_UNSIGNED);
7886 				if (*++tl != 0)
7887 					nd->nd_flag |= ND_NOMOREDATA;
7888 			}
7889 		}
7890 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7891 			/* Load the directory attributes. */
7892 			error = nfsm_loadattr(nd, dnap);
7893 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7894 			if (error != 0)
7895 				goto nfsmout;
7896 			*dattrflagp = 1;
7897 			if (dp != NULL && *attrflagp != 0) {
7898 				dp->nfsdl_change = nnap->na_filerev;
7899 				dp->nfsdl_modtime = nnap->na_mtime;
7900 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7901 			}
7902 			/*
7903 			 * We can now complete the Open state.
7904 			 */
7905 			nfhp = *nfhpp;
7906 			if (dp != NULL) {
7907 				dp->nfsdl_fhlen = nfhp->nfh_len;
7908 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7909 				    nfhp->nfh_len);
7910 			}
7911 			/*
7912 			 * Get an Open structure that will be
7913 			 * attached to the OpenOwner, acquired already.
7914 			 */
7915 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7916 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7917 			    cred, p, NULL, &op, &newone, NULL, 0);
7918 			if (error != 0)
7919 				goto nfsmout;
7920 			op->nfso_stateid = stateid;
7921 			newnfs_copyincred(cred, &op->nfso_cred);
7922 
7923 			nfscl_openrelease(nmp, op, error, newone);
7924 			*unlockedp = 1;
7925 
7926 			/* Now, handle the RestoreFH and LayoutGet. */
7927 			if (nd->nd_repstat == 0) {
7928 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7929 				*laystatp = fxdr_unsigned(int, *(tl + 3));
7930 				if (*laystatp == 0) {
7931 					error = nfsrv_parselayoutget(nmp, nd,
7932 					    stateidp, retonclosep, flhp);
7933 					if (error != 0)
7934 						*laystatp = error;
7935 				}
7936 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7937 				    error);
7938 			} else
7939 				nd->nd_repstat = 0;
7940 		}
7941 	}
7942 	if (nd->nd_repstat != 0 && error == 0)
7943 		error = nd->nd_repstat;
7944 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7945 		nfscl_initiate_recovery(owp->nfsow_clp);
7946 nfsmout:
7947 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7948 	if (error == 0)
7949 		*dpp = dp;
7950 	else
7951 		free(dp, M_NFSCLDELEG);
7952 	m_freem(nd->nd_mrep);
7953 	return (error);
7954 }
7955 
7956 /*
7957  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7958  */
7959 static int
7960 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7961     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7962     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7963     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7964     int *dattrflagp, void *dstuff, int *unlockedp)
7965 {
7966 	struct nfscllayout *lyp;
7967 	struct nfsclflayouthead flh;
7968 	struct nfsfh *nfhp;
7969 	struct nfsclsession *tsep;
7970 	struct nfsmount *nmp;
7971 	nfsv4stateid_t stateid;
7972 	int error, layoutlen, layouttype, retonclose, laystat;
7973 
7974 	error = 0;
7975 	nmp = VFSTONFS(dvp->v_mount);
7976 	if (NFSHASFLEXFILE(nmp))
7977 		layouttype = NFSLAYOUT_FLEXFILE;
7978 	else
7979 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7980 	LIST_INIT(&flh);
7981 	tsep = nfsmnt_mdssession(nmp);
7982 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7983 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7984 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7985 	    dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7986 	    &flh, &laystat);
7987 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7988 	    laystat, error);
7989 	lyp = NULL;
7990 	if (laystat == 0) {
7991 		nfhp = *nfhpp;
7992 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
7993 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
7994 		    layouttype, laystat, NULL, cred, p);
7995 	} else
7996 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
7997 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
7998 		    cred, p);
7999 	if (laystat == 0)
8000 		nfscl_rellayout(lyp, 0);
8001 	return (error);
8002 }
8003 
8004 /*
8005  * Process the results of a layoutget() operation.
8006  */
8007 static int
8008 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8009     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8010     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8011     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8012 {
8013 	struct nfsclflayout *tflp;
8014 	struct nfscldevinfo *dip;
8015 	uint8_t *dev;
8016 	int i, mirrorcnt;
8017 
8018 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8019 		NFSLOCKMNT(nmp);
8020 		if (!NFSHASFLEXFILE(nmp)) {
8021 			/* Switch to using Flex File Layout. */
8022 			nmp->nm_state |= NFSSTA_FLEXFILE;
8023 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8024 			/* Disable pNFS. */
8025 			NFSCL_DEBUG(1, "disable PNFS\n");
8026 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8027 		}
8028 		NFSUNLOCKMNT(nmp);
8029 	}
8030 	if (laystat == 0) {
8031 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8032 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8033 			if (layouttype == NFSLAYOUT_FLEXFILE)
8034 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8035 			else
8036 				mirrorcnt = 1;
8037 			for (i = 0; i < mirrorcnt; i++) {
8038 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8039 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8040 				if (laystat != 0) {
8041 					if (layouttype == NFSLAYOUT_FLEXFILE)
8042 						dev = tflp->nfsfl_ffm[i].dev;
8043 					else
8044 						dev = tflp->nfsfl_dev;
8045 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8046 					    layouttype, notifybit, &dip, cred,
8047 					    p);
8048 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8049 					    laystat);
8050 					if (laystat != 0)
8051 						goto out;
8052 					laystat = nfscl_adddevinfo(nmp, dip, i,
8053 					    tflp);
8054 					if (laystat != 0)
8055 						printf("nfsrpc_layoutgetresout"
8056 						    ": cannot add\n");
8057 				}
8058 			}
8059 		}
8060 	}
8061 out:
8062 	if (laystat == 0) {
8063 		/*
8064 		 * nfscl_layout() always returns with the nfsly_lock
8065 		 * set to a refcnt (shared lock).
8066 		 * Passing in dvp is sufficient, since it is only used to
8067 		 * get the fsid for the file system.
8068 		 */
8069 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8070 		    layouttype, retonclose, flhp, lypp, cred, p);
8071 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8072 		    laystat);
8073 		if (laystat == 0 && islockedp != NULL)
8074 			*islockedp = 1;
8075 	}
8076 	return (laystat);
8077 }
8078 
8079 /*
8080  * nfs copy_file_range operation.
8081  */
8082 APPLESTATIC int
8083 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8084     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8085     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8086     struct ucred *cred, bool consecutive, bool *must_commitp)
8087 {
8088 	int commit, error, expireret = 0, retrycnt;
8089 	u_int32_t clidrev = 0;
8090 	struct nfsmount *nmp = VFSTONFS(vnode_mount(invp));
8091 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8092 	nfsv4stateid_t instateid, outstateid;
8093 	void *inlckp, *outlckp;
8094 
8095 	if (nmp->nm_clp != NULL)
8096 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8097 	innfhp = VTONFS(invp)->n_fhp;
8098 	outnfhp = VTONFS(outvp)->n_fhp;
8099 	retrycnt = 0;
8100 	do {
8101 		/* Get both stateids. */
8102 		inlckp = NULL;
8103 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8104 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8105 		    &inlckp);
8106 		outlckp = NULL;
8107 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8108 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8109 		    &outlckp);
8110 
8111 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8112 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8113 		    outattrflagp, consecutive, &commit, cred, curthread);
8114 		if (error == 0) {
8115 			if (commit != NFSWRITE_FILESYNC)
8116 				*must_commitp = true;
8117 			*inoffp += *lenp;
8118 			*outoffp += *lenp;
8119 		} else if (error == NFSERR_STALESTATEID)
8120 			nfscl_initiate_recovery(nmp->nm_clp);
8121 		if (inlckp != NULL)
8122 			nfscl_lockderef(inlckp);
8123 		if (outlckp != NULL)
8124 			nfscl_lockderef(outlckp);
8125 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8126 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8127 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8128 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8129 		} else if ((error == NFSERR_EXPIRED ||
8130 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8131 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8132 			    curthread);
8133 		}
8134 		retrycnt++;
8135 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8136 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8137 	      error == NFSERR_STALEDONTRECOVER ||
8138 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8139 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8140 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8141 	if (error != 0 && (retrycnt >= 4 ||
8142 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8143 	      error == NFSERR_STALEDONTRECOVER))
8144 		error = EIO;
8145 	return (error);
8146 }
8147 
8148 /*
8149  * The copy RPC.
8150  */
8151 static int
8152 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8153     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8154     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8155     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8156     NFSPROC_T *p)
8157 {
8158 	uint32_t *tl;
8159 	int error;
8160 	struct nfsrv_descript nfsd;
8161 	struct nfsrv_descript *nd = &nfsd;
8162 	struct nfsmount *nmp;
8163 	nfsattrbit_t attrbits;
8164 	uint64_t len;
8165 
8166 	nmp = VFSTONFS(outvp->v_mount);
8167 	*inattrflagp = *outattrflagp = 0;
8168 	*commitp = NFSWRITE_UNSTABLE;
8169 	len = *lenp;
8170 	*lenp = 0;
8171 	if (len > nfs_maxcopyrange)
8172 		len = nfs_maxcopyrange;
8173 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8174 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8175 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8176 	NFSGETATTR_ATTRBIT(&attrbits);
8177 	nfsrv_putattrbit(nd, &attrbits);
8178 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8179 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8180 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8181 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8182 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8183 	*tl = txdr_unsigned(NFSV4OP_COPY);
8184 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8185 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8186 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8187 	txdr_hyper(inoff, tl); tl += 2;
8188 	txdr_hyper(outoff, tl); tl += 2;
8189 	txdr_hyper(len, tl); tl += 2;
8190 	if (consecutive)
8191 		*tl++ = newnfs_true;
8192 	else
8193 		*tl++ = newnfs_false;
8194 	*tl++ = newnfs_true;
8195 	*tl++ = 0;
8196 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8197 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8198 	nfsrv_putattrbit(nd, &attrbits);
8199 	error = nfscl_request(nd, invp, p, cred, NULL);
8200 	if (error != 0)
8201 		return (error);
8202 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8203 		/* Get the input file's attributes. */
8204 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8205 		if (*(tl + 1) == 0) {
8206 			error = nfsm_loadattr(nd, innap);
8207 			if (error != 0)
8208 				goto nfsmout;
8209 			*inattrflagp = 1;
8210 		} else
8211 			nd->nd_flag |= ND_NOMOREDATA;
8212 	}
8213 	/* Skip over return stat for PutFH. */
8214 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8215 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8216 		if (*++tl != 0)
8217 			nd->nd_flag |= ND_NOMOREDATA;
8218 	}
8219 	/* Skip over return stat for Copy. */
8220 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8221 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8222 	if (nd->nd_repstat == 0) {
8223 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8224 		if (*tl != 0) {
8225 			/* There should be no callback ids. */
8226 			error = NFSERR_BADXDR;
8227 			goto nfsmout;
8228 		}
8229 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8230 		    NFSX_VERF);
8231 		len = fxdr_hyper(tl); tl += 2;
8232 		*commitp = fxdr_unsigned(int, *tl++);
8233 		NFSLOCKMNT(nmp);
8234 		if (!NFSHASWRITEVERF(nmp)) {
8235 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8236 			NFSSETWRITEVERF(nmp);
8237 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8238 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8239 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8240 		}
8241 		NFSUNLOCKMNT(nmp);
8242 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8243 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8244 			/* Must be a synchronous copy. */
8245 			nd->nd_repstat = NFSERR_NOTSUPP;
8246 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8247 		error = nfsm_loadattr(nd, outnap);
8248 		if (error == 0)
8249 			*outattrflagp = NFS_LATTR_NOSHRINK;
8250 		if (nd->nd_repstat == 0)
8251 			*lenp = len;
8252 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8253 		/*
8254 		 * For the case where consecutive is not supported, but
8255 		 * synchronous is supported, we can try consecutive == false
8256 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8257 		 * since Copy cannot be done.
8258 		 */
8259 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8260 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8261 			if (!consecutive || *++tl == newnfs_false)
8262 				nd->nd_repstat = NFSERR_NOTSUPP;
8263 		} else
8264 			nd->nd_repstat = NFSERR_BADXDR;
8265 	}
8266 	if (error == 0)
8267 		error = nd->nd_repstat;
8268 nfsmout:
8269 	m_freem(nd->nd_mrep);
8270 	return (error);
8271 }
8272 
8273 /*
8274  * Seek operation.
8275  */
8276 APPLESTATIC int
8277 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8278     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8279 {
8280 	int error, expireret = 0, retrycnt;
8281 	u_int32_t clidrev = 0;
8282 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
8283 	struct nfsnode *np = VTONFS(vp);
8284 	struct nfsfh *nfhp = NULL;
8285 	nfsv4stateid_t stateid;
8286 	void *lckp;
8287 
8288 	if (nmp->nm_clp != NULL)
8289 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8290 	nfhp = np->n_fhp;
8291 	retrycnt = 0;
8292 	do {
8293 		lckp = NULL;
8294 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8295 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8296 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8297 		    nap, attrflagp, cred);
8298 		if (error == NFSERR_STALESTATEID)
8299 			nfscl_initiate_recovery(nmp->nm_clp);
8300 		if (lckp != NULL)
8301 			nfscl_lockderef(lckp);
8302 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8303 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8304 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8305 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8306 		} else if ((error == NFSERR_EXPIRED ||
8307 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8308 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8309 			    curthread);
8310 		}
8311 		retrycnt++;
8312 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8313 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8314 	    error == NFSERR_BADSESSION ||
8315 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8316 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8317 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8318 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8319 	if (error && retrycnt >= 4)
8320 		error = EIO;
8321 	return (error);
8322 }
8323 
8324 /*
8325  * The seek RPC.
8326  */
8327 static int
8328 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8329     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8330 {
8331 	uint32_t *tl;
8332 	int error;
8333 	struct nfsrv_descript nfsd;
8334 	struct nfsrv_descript *nd = &nfsd;
8335 	nfsattrbit_t attrbits;
8336 
8337 	*attrflagp = 0;
8338 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8339 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8340 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8341 	txdr_hyper(*offp, tl); tl += 2;
8342 	*tl++ = txdr_unsigned(content);
8343 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8344 	NFSGETATTR_ATTRBIT(&attrbits);
8345 	nfsrv_putattrbit(nd, &attrbits);
8346 	error = nfscl_request(nd, vp, curthread, cred, NULL);
8347 	if (error != 0)
8348 		return (error);
8349 	if (nd->nd_repstat == 0) {
8350 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8351 		if (*tl++ == newnfs_true)
8352 			*eofp = true;
8353 		else
8354 			*eofp = false;
8355 		*offp = fxdr_hyper(tl);
8356 		/* Just skip over Getattr op status. */
8357 		error = nfsm_loadattr(nd, nap);
8358 		if (error == 0)
8359 			*attrflagp = 1;
8360 	}
8361 	error = nd->nd_repstat;
8362 nfsmout:
8363 	m_freem(nd->nd_mrep);
8364 	return (error);
8365 }
8366 
8367 /*
8368  * The getextattr RPC.
8369  */
8370 APPLESTATIC int
8371 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8372     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8373 {
8374 	uint32_t *tl;
8375 	int error;
8376 	struct nfsrv_descript nfsd;
8377 	struct nfsrv_descript *nd = &nfsd;
8378 	nfsattrbit_t attrbits;
8379 	uint32_t len, len2;
8380 
8381 	*attrflagp = 0;
8382 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8383 	nfsm_strtom(nd, name, strlen(name));
8384 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8385 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8386 	NFSGETATTR_ATTRBIT(&attrbits);
8387 	nfsrv_putattrbit(nd, &attrbits);
8388 	error = nfscl_request(nd, vp, p, cred, NULL);
8389 	if (error != 0)
8390 		return (error);
8391 	if (nd->nd_repstat == 0) {
8392 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8393 		len = fxdr_unsigned(uint32_t, *tl);
8394 		/* Sanity check lengths. */
8395 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8396 		    uiop->uio_resid <= UINT32_MAX) {
8397 			len2 = uiop->uio_resid;
8398 			if (len2 >= len)
8399 				error = nfsm_mbufuio(nd, uiop, len);
8400 			else {
8401 				error = nfsm_mbufuio(nd, uiop, len2);
8402 				if (error == 0) {
8403 					/*
8404 					 * nfsm_mbufuio() advances to a multiple
8405 					 * of 4, so round up len2 as well.  Then
8406 					 * we need to advance over the rest of
8407 					 * the data, rounding up the remaining
8408 					 * length.
8409 					 */
8410 					len2 = NFSM_RNDUP(len2);
8411 					len2 = NFSM_RNDUP(len - len2);
8412 					if (len2 > 0)
8413 						error = nfsm_advance(nd, len2,
8414 						    -1);
8415 				}
8416 			}
8417 		} else if (uiop == NULL && len > 0) {
8418 			/* Just wants the length and not the data. */
8419 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8420 		} else if (len > 0)
8421 			error = ENOATTR;
8422 		if (error != 0)
8423 			goto nfsmout;
8424 		*lenp = len;
8425 		/* Just skip over Getattr op status. */
8426 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8427 		error = nfsm_loadattr(nd, nap);
8428 		if (error == 0)
8429 			*attrflagp = 1;
8430 	}
8431 	if (error == 0)
8432 		error = nd->nd_repstat;
8433 nfsmout:
8434 	m_freem(nd->nd_mrep);
8435 	return (error);
8436 }
8437 
8438 /*
8439  * The setextattr RPC.
8440  */
8441 APPLESTATIC int
8442 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8443     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8444 {
8445 	uint32_t *tl;
8446 	int error;
8447 	struct nfsrv_descript nfsd;
8448 	struct nfsrv_descript *nd = &nfsd;
8449 	nfsattrbit_t attrbits;
8450 
8451 	*attrflagp = 0;
8452 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8453 	if (uiop->uio_resid > nd->nd_maxreq) {
8454 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8455 		m_freem(nd->nd_mreq);
8456 		return (EINVAL);
8457 	}
8458 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8459 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8460 	nfsm_strtom(nd, name, strlen(name));
8461 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8462 	*tl = txdr_unsigned(uiop->uio_resid);
8463 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8464 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8465 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8466 	NFSGETATTR_ATTRBIT(&attrbits);
8467 	nfsrv_putattrbit(nd, &attrbits);
8468 	error = nfscl_request(nd, vp, p, cred, NULL);
8469 	if (error != 0)
8470 		return (error);
8471 	if (nd->nd_repstat == 0) {
8472 		/* Just skip over the reply and Getattr op status. */
8473 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8474 		    NFSX_UNSIGNED);
8475 		error = nfsm_loadattr(nd, nap);
8476 		if (error == 0)
8477 			*attrflagp = 1;
8478 	}
8479 	if (error == 0)
8480 		error = nd->nd_repstat;
8481 nfsmout:
8482 	m_freem(nd->nd_mrep);
8483 	return (error);
8484 }
8485 
8486 /*
8487  * The removeextattr RPC.
8488  */
8489 APPLESTATIC int
8490 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8491     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8492 {
8493 	uint32_t *tl;
8494 	int error;
8495 	struct nfsrv_descript nfsd;
8496 	struct nfsrv_descript *nd = &nfsd;
8497 	nfsattrbit_t attrbits;
8498 
8499 	*attrflagp = 0;
8500 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8501 	nfsm_strtom(nd, name, strlen(name));
8502 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8503 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8504 	NFSGETATTR_ATTRBIT(&attrbits);
8505 	nfsrv_putattrbit(nd, &attrbits);
8506 	error = nfscl_request(nd, vp, p, cred, NULL);
8507 	if (error != 0)
8508 		return (error);
8509 	if (nd->nd_repstat == 0) {
8510 		/* Just skip over the reply and Getattr op status. */
8511 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8512 		    NFSX_UNSIGNED);
8513 		error = nfsm_loadattr(nd, nap);
8514 		if (error == 0)
8515 			*attrflagp = 1;
8516 	}
8517 	if (error == 0)
8518 		error = nd->nd_repstat;
8519 nfsmout:
8520 	m_freem(nd->nd_mrep);
8521 	return (error);
8522 }
8523 
8524 /*
8525  * The listextattr RPC.
8526  */
8527 APPLESTATIC int
8528 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8529     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8530     struct ucred *cred, NFSPROC_T *p)
8531 {
8532 	uint32_t *tl;
8533 	int cnt, error, i, len;
8534 	struct nfsrv_descript nfsd;
8535 	struct nfsrv_descript *nd = &nfsd;
8536 	nfsattrbit_t attrbits;
8537 	u_char c;
8538 
8539 	*attrflagp = 0;
8540 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8541 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8542 	txdr_hyper(*cookiep, tl); tl += 2;
8543 	*tl++ = txdr_unsigned(*lenp);
8544 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8545 	NFSGETATTR_ATTRBIT(&attrbits);
8546 	nfsrv_putattrbit(nd, &attrbits);
8547 	error = nfscl_request(nd, vp, p, cred, NULL);
8548 	if (error != 0)
8549 		return (error);
8550 	*eofp = true;
8551 	*lenp = 0;
8552 	if (nd->nd_repstat == 0) {
8553 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8554 		*cookiep = fxdr_hyper(tl); tl += 2;
8555 		cnt = fxdr_unsigned(int, *tl);
8556 		if (cnt < 0) {
8557 			error = EBADRPC;
8558 			goto nfsmout;
8559 		}
8560 		for (i = 0; i < cnt; i++) {
8561 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8562 			len = fxdr_unsigned(int, *tl);
8563 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8564 				error = EBADRPC;
8565 				goto nfsmout;
8566 			}
8567 			if (uiop == NULL)
8568 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8569 			else if (uiop->uio_resid >= len + 1) {
8570 				c = len;
8571 				error = uiomove(&c, sizeof(c), uiop);
8572 				if (error == 0)
8573 					error = nfsm_mbufuio(nd, uiop, len);
8574 			} else {
8575 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8576 				*eofp = false;
8577 			}
8578 			if (error != 0)
8579 				goto nfsmout;
8580 			*lenp += (len + 1);
8581 		}
8582 		/* Get the eof and skip over the Getattr op status. */
8583 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8584 		/*
8585 		 * *eofp is set false above, because it wasn't able to copy
8586 		 * all of the reply.
8587 		 */
8588 		if (*eofp && *tl == 0)
8589 			*eofp = false;
8590 		error = nfsm_loadattr(nd, nap);
8591 		if (error == 0)
8592 			*attrflagp = 1;
8593 	}
8594 	if (error == 0)
8595 		error = nd->nd_repstat;
8596 nfsmout:
8597 	m_freem(nd->nd_mrep);
8598 	return (error);
8599 }
8600 
8601