xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision e4e041744071ffade1a1c5e08714ed1d5db8e6a6)
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 = uio_uio_resid(uiop);
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 = uio_uio_resid(uiop);
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 			uio_uio_resid_add(uiop, len);
1882 			uio_iov_base_add(uiop, -len);
1883 			uio_iov_len_add(uiop, len);
1884 		}
1885 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1886 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1887 			    &wccflag, stuff);
1888 			if (error)
1889 				goto nfsmout;
1890 		}
1891 		if (!nd->nd_repstat) {
1892 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1893 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1894 					+ NFSX_VERF);
1895 				rlen = fxdr_unsigned(int, *tl++);
1896 				if (rlen == 0) {
1897 					error = NFSERR_IO;
1898 					goto nfsmout;
1899 				} else if (rlen < len) {
1900 					backup = len - rlen;
1901 					uio_iov_base_add(uiop, -(backup));
1902 					uio_iov_len_add(uiop, backup);
1903 					uiop->uio_offset -= backup;
1904 					uio_uio_resid_add(uiop, backup);
1905 					len = rlen;
1906 				}
1907 				commit = fxdr_unsigned(int, *tl++);
1908 
1909 				/*
1910 				 * Return the lowest commitment level
1911 				 * obtained by any of the RPCs.
1912 				 */
1913 				if (committed == NFSWRITE_FILESYNC)
1914 					committed = commit;
1915 				else if (committed == NFSWRITE_DATASYNC &&
1916 					commit == NFSWRITE_UNSTABLE)
1917 					committed = commit;
1918 				NFSLOCKMNT(nmp);
1919 				if (!NFSHASWRITEVERF(nmp)) {
1920 					NFSBCOPY((caddr_t)tl,
1921 					    (caddr_t)&nmp->nm_verf[0],
1922 					    NFSX_VERF);
1923 					NFSSETWRITEVERF(nmp);
1924 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1925 				    NFSX_VERF)) {
1926 					*must_commit = 1;
1927 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1928 				}
1929 				NFSUNLOCKMNT(nmp);
1930 			}
1931 			if (nd->nd_flag & ND_NFSV4)
1932 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1933 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1934 				error = nfsm_loadattr(nd, nap);
1935 				if (!error)
1936 					*attrflagp = NFS_LATTR_NOSHRINK;
1937 			}
1938 		} else {
1939 			error = nd->nd_repstat;
1940 		}
1941 		if (error)
1942 			goto nfsmout;
1943 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1944 		m_freem(nd->nd_mrep);
1945 		nd->nd_mrep = NULL;
1946 		tsiz -= len;
1947 	}
1948 nfsmout:
1949 	if (nd->nd_mrep != NULL)
1950 		m_freem(nd->nd_mrep);
1951 	*iomode = committed;
1952 	if (nd->nd_repstat && !error)
1953 		error = nd->nd_repstat;
1954 	return (error);
1955 }
1956 
1957 /*
1958  * nfs mknod rpc
1959  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1960  * mode set to specify the file type and the size field for rdev.
1961  */
1962 APPLESTATIC int
1963 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1964     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1965     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1966     int *attrflagp, int *dattrflagp, void *dstuff)
1967 {
1968 	u_int32_t *tl;
1969 	int error = 0;
1970 	struct nfsrv_descript nfsd, *nd = &nfsd;
1971 	nfsattrbit_t attrbits;
1972 
1973 	*nfhpp = NULL;
1974 	*attrflagp = 0;
1975 	*dattrflagp = 0;
1976 	if (namelen > NFS_MAXNAMLEN)
1977 		return (ENAMETOOLONG);
1978 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1979 	if (nd->nd_flag & ND_NFSV4) {
1980 		if (vtyp == VBLK || vtyp == VCHR) {
1981 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1982 			*tl++ = vtonfsv34_type(vtyp);
1983 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1984 			*tl = txdr_unsigned(NFSMINOR(rdev));
1985 		} else {
1986 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1987 			*tl = vtonfsv34_type(vtyp);
1988 		}
1989 	}
1990 	(void) nfsm_strtom(nd, name, namelen);
1991 	if (nd->nd_flag & ND_NFSV3) {
1992 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1993 		*tl = vtonfsv34_type(vtyp);
1994 	}
1995 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1996 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
1997 	if ((nd->nd_flag & ND_NFSV3) &&
1998 	    (vtyp == VCHR || vtyp == VBLK)) {
1999 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2000 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2001 		*tl = txdr_unsigned(NFSMINOR(rdev));
2002 	}
2003 	if (nd->nd_flag & ND_NFSV4) {
2004 		NFSGETATTR_ATTRBIT(&attrbits);
2005 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2006 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2007 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2008 		(void) nfsrv_putattrbit(nd, &attrbits);
2009 	}
2010 	if (nd->nd_flag & ND_NFSV2)
2011 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2012 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2013 	if (error)
2014 		return (error);
2015 	if (nd->nd_flag & ND_NFSV4)
2016 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2017 	if (!nd->nd_repstat) {
2018 		if (nd->nd_flag & ND_NFSV4) {
2019 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2020 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2021 			if (error)
2022 				goto nfsmout;
2023 		}
2024 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2025 		if (error)
2026 			goto nfsmout;
2027 	}
2028 	if (nd->nd_flag & ND_NFSV3)
2029 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2030 	if (!error && nd->nd_repstat)
2031 		error = nd->nd_repstat;
2032 nfsmout:
2033 	m_freem(nd->nd_mrep);
2034 	return (error);
2035 }
2036 
2037 /*
2038  * nfs file create call
2039  * Mostly just call the approriate routine. (I separated out v4, so that
2040  * error recovery wouldn't be as difficult.)
2041  */
2042 APPLESTATIC int
2043 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2044     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2045     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2046     int *attrflagp, int *dattrflagp, void *dstuff)
2047 {
2048 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2049 	struct nfsclowner *owp;
2050 	struct nfscldeleg *dp;
2051 	struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
2052 	u_int32_t clidrev;
2053 
2054 	if (NFSHASNFSV4(nmp)) {
2055 	    retrycnt = 0;
2056 	    do {
2057 		dp = NULL;
2058 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2059 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2060 		    NULL, 1);
2061 		if (error)
2062 			return (error);
2063 		if (nmp->nm_clp != NULL)
2064 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2065 		else
2066 			clidrev = 0;
2067 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2068 		    nfs_numnfscbd == 0 || retrycnt > 0)
2069 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2070 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2071 			  attrflagp, dattrflagp, dstuff, &unlocked);
2072 		else
2073 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2074 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2075 			  attrflagp, dattrflagp, dstuff, &unlocked);
2076 		/*
2077 		 * There is no need to invalidate cached attributes here,
2078 		 * since new post-delegation issue attributes are always
2079 		 * returned by nfsrpc_createv4() and these will update the
2080 		 * attribute cache.
2081 		 */
2082 		if (dp != NULL)
2083 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2084 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2085 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2086 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2087 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2088 		    error == NFSERR_BADSESSION) {
2089 			(void) nfs_catnap(PZERO, error, "nfs_open");
2090 		} else if ((error == NFSERR_EXPIRED ||
2091 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2092 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2093 			retrycnt++;
2094 		}
2095 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2096 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2097 		error == NFSERR_BADSESSION ||
2098 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2099 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2100 	    if (error && retrycnt >= 4)
2101 		    error = EIO;
2102 	} else {
2103 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2104 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2105 		    dstuff);
2106 	}
2107 	return (error);
2108 }
2109 
2110 /*
2111  * The create rpc for v2 and 3.
2112  */
2113 static int
2114 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2115     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2116     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2117     int *attrflagp, int *dattrflagp, void *dstuff)
2118 {
2119 	u_int32_t *tl;
2120 	int error = 0;
2121 	struct nfsrv_descript nfsd, *nd = &nfsd;
2122 
2123 	*nfhpp = NULL;
2124 	*attrflagp = 0;
2125 	*dattrflagp = 0;
2126 	if (namelen > NFS_MAXNAMLEN)
2127 		return (ENAMETOOLONG);
2128 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2129 	(void) nfsm_strtom(nd, name, namelen);
2130 	if (nd->nd_flag & ND_NFSV3) {
2131 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2132 		if (fmode & O_EXCL) {
2133 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2134 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2135 			*tl++ = cverf.lval[0];
2136 			*tl = cverf.lval[1];
2137 		} else {
2138 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2139 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2140 		}
2141 	} else {
2142 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2143 	}
2144 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2145 	if (error)
2146 		return (error);
2147 	if (nd->nd_repstat == 0) {
2148 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2149 		if (error)
2150 			goto nfsmout;
2151 	}
2152 	if (nd->nd_flag & ND_NFSV3)
2153 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2154 	if (nd->nd_repstat != 0 && error == 0)
2155 		error = nd->nd_repstat;
2156 nfsmout:
2157 	m_freem(nd->nd_mrep);
2158 	return (error);
2159 }
2160 
2161 static int
2162 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2163     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2164     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2165     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2166     int *dattrflagp, void *dstuff, int *unlockedp)
2167 {
2168 	u_int32_t *tl;
2169 	int error = 0, deleg, newone, ret, acesize, limitby;
2170 	struct nfsrv_descript nfsd, *nd = &nfsd;
2171 	struct nfsclopen *op;
2172 	struct nfscldeleg *dp = NULL;
2173 	struct nfsnode *np;
2174 	struct nfsfh *nfhp;
2175 	nfsattrbit_t attrbits;
2176 	nfsv4stateid_t stateid;
2177 	u_int32_t rflags;
2178 	struct nfsmount *nmp;
2179 	struct nfsclsession *tsep;
2180 
2181 	nmp = VFSTONFS(dvp->v_mount);
2182 	np = VTONFS(dvp);
2183 	*unlockedp = 0;
2184 	*nfhpp = NULL;
2185 	*dpp = NULL;
2186 	*attrflagp = 0;
2187 	*dattrflagp = 0;
2188 	if (namelen > NFS_MAXNAMLEN)
2189 		return (ENAMETOOLONG);
2190 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2191 	/*
2192 	 * For V4, this is actually an Open op.
2193 	 */
2194 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2195 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2196 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2197 	    NFSV4OPEN_ACCESSREAD);
2198 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2199 	tsep = nfsmnt_mdssession(nmp);
2200 	*tl++ = tsep->nfsess_clientid.lval[0];
2201 	*tl = tsep->nfsess_clientid.lval[1];
2202 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2203 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2204 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2205 	if (fmode & O_EXCL) {
2206 		if (NFSHASNFSV4N(nmp)) {
2207 			if (NFSHASSESSPERSIST(nmp)) {
2208 				/* Use GUARDED for persistent sessions. */
2209 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2210 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2211 			} else {
2212 				/* Otherwise, use EXCLUSIVE4_1. */
2213 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2214 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2215 				*tl++ = cverf.lval[0];
2216 				*tl = cverf.lval[1];
2217 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2218 			}
2219 		} else {
2220 			/* NFSv4.0 */
2221 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2222 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2223 			*tl++ = cverf.lval[0];
2224 			*tl = cverf.lval[1];
2225 		}
2226 	} else {
2227 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2228 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2229 	}
2230 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2231 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2232 	(void) nfsm_strtom(nd, name, namelen);
2233 	/* Get the new file's handle and attributes. */
2234 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2235 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2236 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2237 	NFSGETATTR_ATTRBIT(&attrbits);
2238 	(void) nfsrv_putattrbit(nd, &attrbits);
2239 	/* Get the directory's post-op attributes. */
2240 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2241 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2242 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2243 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2244 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2245 	(void) nfsrv_putattrbit(nd, &attrbits);
2246 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2247 	if (error)
2248 		return (error);
2249 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2250 	if (nd->nd_repstat == 0) {
2251 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2252 		    6 * NFSX_UNSIGNED);
2253 		stateid.seqid = *tl++;
2254 		stateid.other[0] = *tl++;
2255 		stateid.other[1] = *tl++;
2256 		stateid.other[2] = *tl;
2257 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2258 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2259 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2260 		deleg = fxdr_unsigned(int, *tl);
2261 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2262 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2263 			if (!(owp->nfsow_clp->nfsc_flags &
2264 			      NFSCLFLAGS_FIRSTDELEG))
2265 				owp->nfsow_clp->nfsc_flags |=
2266 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2267 			dp = malloc(
2268 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2269 			    M_NFSCLDELEG, M_WAITOK);
2270 			LIST_INIT(&dp->nfsdl_owner);
2271 			LIST_INIT(&dp->nfsdl_lock);
2272 			dp->nfsdl_clp = owp->nfsow_clp;
2273 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2274 			nfscl_lockinit(&dp->nfsdl_rwlock);
2275 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2276 			    NFSX_UNSIGNED);
2277 			dp->nfsdl_stateid.seqid = *tl++;
2278 			dp->nfsdl_stateid.other[0] = *tl++;
2279 			dp->nfsdl_stateid.other[1] = *tl++;
2280 			dp->nfsdl_stateid.other[2] = *tl++;
2281 			ret = fxdr_unsigned(int, *tl);
2282 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2283 				dp->nfsdl_flags = NFSCLDL_WRITE;
2284 				/*
2285 				 * Indicates how much the file can grow.
2286 				 */
2287 				NFSM_DISSECT(tl, u_int32_t *,
2288 				    3 * NFSX_UNSIGNED);
2289 				limitby = fxdr_unsigned(int, *tl++);
2290 				switch (limitby) {
2291 				case NFSV4OPEN_LIMITSIZE:
2292 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2293 					break;
2294 				case NFSV4OPEN_LIMITBLOCKS:
2295 					dp->nfsdl_sizelimit =
2296 					    fxdr_unsigned(u_int64_t, *tl++);
2297 					dp->nfsdl_sizelimit *=
2298 					    fxdr_unsigned(u_int64_t, *tl);
2299 					break;
2300 				default:
2301 					error = NFSERR_BADXDR;
2302 					goto nfsmout;
2303 				}
2304 			} else {
2305 				dp->nfsdl_flags = NFSCLDL_READ;
2306 			}
2307 			if (ret)
2308 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2309 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2310 			    &acesize, p);
2311 			if (error)
2312 				goto nfsmout;
2313 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2314 			error = NFSERR_BADXDR;
2315 			goto nfsmout;
2316 		}
2317 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2318 		if (error)
2319 			goto nfsmout;
2320 		/* Get rid of the PutFH and Getattr status values. */
2321 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2322 		/* Load the directory attributes. */
2323 		error = nfsm_loadattr(nd, dnap);
2324 		if (error)
2325 			goto nfsmout;
2326 		*dattrflagp = 1;
2327 		if (dp != NULL && *attrflagp) {
2328 			dp->nfsdl_change = nnap->na_filerev;
2329 			dp->nfsdl_modtime = nnap->na_mtime;
2330 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2331 		}
2332 		/*
2333 		 * We can now complete the Open state.
2334 		 */
2335 		nfhp = *nfhpp;
2336 		if (dp != NULL) {
2337 			dp->nfsdl_fhlen = nfhp->nfh_len;
2338 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2339 		}
2340 		/*
2341 		 * Get an Open structure that will be
2342 		 * attached to the OpenOwner, acquired already.
2343 		 */
2344 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2345 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2346 		    cred, p, NULL, &op, &newone, NULL, 0);
2347 		if (error)
2348 			goto nfsmout;
2349 		op->nfso_stateid = stateid;
2350 		newnfs_copyincred(cred, &op->nfso_cred);
2351 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2352 		    do {
2353 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2354 			    nfhp->nfh_len, op, cred, p);
2355 			if (ret == NFSERR_DELAY)
2356 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2357 		    } while (ret == NFSERR_DELAY);
2358 		    error = ret;
2359 		}
2360 
2361 		/*
2362 		 * If the server is handing out delegations, but we didn't
2363 		 * get one because an OpenConfirm was required, try the
2364 		 * Open again, to get a delegation. This is a harmless no-op,
2365 		 * from a server's point of view.
2366 		 */
2367 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2368 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2369 		    !error && dp == NULL) {
2370 		    do {
2371 			ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2372 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2373 			    nfhp->nfh_fh, nfhp->nfh_len,
2374 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2375 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2376 			if (ret == NFSERR_DELAY)
2377 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2378 		    } while (ret == NFSERR_DELAY);
2379 		    if (ret) {
2380 			if (dp != NULL) {
2381 				free(dp, M_NFSCLDELEG);
2382 				dp = NULL;
2383 			}
2384 			if (ret == NFSERR_STALECLIENTID ||
2385 			    ret == NFSERR_STALEDONTRECOVER ||
2386 			    ret == NFSERR_BADSESSION)
2387 				error = ret;
2388 		    }
2389 		}
2390 		nfscl_openrelease(nmp, op, error, newone);
2391 		*unlockedp = 1;
2392 	}
2393 	if (nd->nd_repstat != 0 && error == 0)
2394 		error = nd->nd_repstat;
2395 	if (error == NFSERR_STALECLIENTID)
2396 		nfscl_initiate_recovery(owp->nfsow_clp);
2397 nfsmout:
2398 	if (!error)
2399 		*dpp = dp;
2400 	else if (dp != NULL)
2401 		free(dp, M_NFSCLDELEG);
2402 	m_freem(nd->nd_mrep);
2403 	return (error);
2404 }
2405 
2406 /*
2407  * Nfs remove rpc
2408  */
2409 APPLESTATIC int
2410 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2411     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2412     void *dstuff)
2413 {
2414 	u_int32_t *tl;
2415 	struct nfsrv_descript nfsd, *nd = &nfsd;
2416 	struct nfsnode *np;
2417 	struct nfsmount *nmp;
2418 	nfsv4stateid_t dstateid;
2419 	int error, ret = 0, i;
2420 
2421 	*dattrflagp = 0;
2422 	if (namelen > NFS_MAXNAMLEN)
2423 		return (ENAMETOOLONG);
2424 	nmp = VFSTONFS(vnode_mount(dvp));
2425 tryagain:
2426 	if (NFSHASNFSV4(nmp) && ret == 0) {
2427 		ret = nfscl_removedeleg(vp, p, &dstateid);
2428 		if (ret == 1) {
2429 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2430 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2431 			    NFSX_UNSIGNED);
2432 			if (NFSHASNFSV4N(nmp))
2433 				*tl++ = 0;
2434 			else
2435 				*tl++ = dstateid.seqid;
2436 			*tl++ = dstateid.other[0];
2437 			*tl++ = dstateid.other[1];
2438 			*tl++ = dstateid.other[2];
2439 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2440 			np = VTONFS(dvp);
2441 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2442 			    np->n_fhp->nfh_len, 0);
2443 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2444 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2445 		}
2446 	} else {
2447 		ret = 0;
2448 	}
2449 	if (ret == 0)
2450 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2451 	(void) nfsm_strtom(nd, name, namelen);
2452 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2453 	if (error)
2454 		return (error);
2455 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2456 		/* For NFSv4, parse out any Delereturn replies. */
2457 		if (ret > 0 && nd->nd_repstat != 0 &&
2458 		    (nd->nd_flag & ND_NOMOREDATA)) {
2459 			/*
2460 			 * If the Delegreturn failed, try again without
2461 			 * it. The server will Recall, as required.
2462 			 */
2463 			m_freem(nd->nd_mrep);
2464 			goto tryagain;
2465 		}
2466 		for (i = 0; i < (ret * 2); i++) {
2467 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2468 			    ND_NFSV4) {
2469 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2470 			    if (*(tl + 1))
2471 				nd->nd_flag |= ND_NOMOREDATA;
2472 			}
2473 		}
2474 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2475 	}
2476 	if (nd->nd_repstat && !error)
2477 		error = nd->nd_repstat;
2478 nfsmout:
2479 	m_freem(nd->nd_mrep);
2480 	return (error);
2481 }
2482 
2483 /*
2484  * Do an nfs rename rpc.
2485  */
2486 APPLESTATIC int
2487 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2488     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2489     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2490     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2491 {
2492 	u_int32_t *tl;
2493 	struct nfsrv_descript nfsd, *nd = &nfsd;
2494 	struct nfsmount *nmp;
2495 	struct nfsnode *np;
2496 	nfsattrbit_t attrbits;
2497 	nfsv4stateid_t fdstateid, tdstateid;
2498 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2499 
2500 	*fattrflagp = 0;
2501 	*tattrflagp = 0;
2502 	nmp = VFSTONFS(vnode_mount(fdvp));
2503 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2504 		return (ENAMETOOLONG);
2505 tryagain:
2506 	if (NFSHASNFSV4(nmp) && ret == 0) {
2507 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2508 		    &tdstateid, &gottd, p);
2509 		if (gotfd && gottd) {
2510 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2511 		} else if (gotfd) {
2512 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2513 		} else if (gottd) {
2514 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2515 		}
2516 		if (gotfd) {
2517 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2518 			if (NFSHASNFSV4N(nmp))
2519 				*tl++ = 0;
2520 			else
2521 				*tl++ = fdstateid.seqid;
2522 			*tl++ = fdstateid.other[0];
2523 			*tl++ = fdstateid.other[1];
2524 			*tl = fdstateid.other[2];
2525 			if (gottd) {
2526 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2527 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2528 				np = VTONFS(tvp);
2529 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2530 				    np->n_fhp->nfh_len, 0);
2531 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2532 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2533 			}
2534 		}
2535 		if (gottd) {
2536 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2537 			if (NFSHASNFSV4N(nmp))
2538 				*tl++ = 0;
2539 			else
2540 				*tl++ = tdstateid.seqid;
2541 			*tl++ = tdstateid.other[0];
2542 			*tl++ = tdstateid.other[1];
2543 			*tl = tdstateid.other[2];
2544 		}
2545 		if (ret > 0) {
2546 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2547 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2548 			np = VTONFS(fdvp);
2549 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2550 			    np->n_fhp->nfh_len, 0);
2551 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2552 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2553 		}
2554 	} else {
2555 		ret = 0;
2556 	}
2557 	if (ret == 0)
2558 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2559 	if (nd->nd_flag & ND_NFSV4) {
2560 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2561 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2562 		NFSWCCATTR_ATTRBIT(&attrbits);
2563 		(void) nfsrv_putattrbit(nd, &attrbits);
2564 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2565 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2566 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2567 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2568 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2569 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2570 		(void) nfsrv_putattrbit(nd, &attrbits);
2571 		nd->nd_flag |= ND_V4WCCATTR;
2572 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2573 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2574 	}
2575 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2576 	if (!(nd->nd_flag & ND_NFSV4))
2577 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2578 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2579 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2580 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2581 	if (error)
2582 		return (error);
2583 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2584 		/* For NFSv4, parse out any Delereturn replies. */
2585 		if (ret > 0 && nd->nd_repstat != 0 &&
2586 		    (nd->nd_flag & ND_NOMOREDATA)) {
2587 			/*
2588 			 * If the Delegreturn failed, try again without
2589 			 * it. The server will Recall, as required.
2590 			 */
2591 			m_freem(nd->nd_mrep);
2592 			goto tryagain;
2593 		}
2594 		for (i = 0; i < (ret * 2); i++) {
2595 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2596 			    ND_NFSV4) {
2597 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2598 			    if (*(tl + 1)) {
2599 				if (i == 0 && ret > 1) {
2600 				    /*
2601 				     * If the Delegreturn failed, try again
2602 				     * without it. The server will Recall, as
2603 				     * required.
2604 				     * If ret > 1, the first iteration of this
2605 				     * loop is the second DelegReturn result.
2606 				     */
2607 				    m_freem(nd->nd_mrep);
2608 				    goto tryagain;
2609 				} else {
2610 				    nd->nd_flag |= ND_NOMOREDATA;
2611 				}
2612 			    }
2613 			}
2614 		}
2615 		/* Now, the first wcc attribute reply. */
2616 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2617 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2618 			if (*(tl + 1))
2619 				nd->nd_flag |= ND_NOMOREDATA;
2620 		}
2621 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2622 		    fstuff);
2623 		/* and the second wcc attribute reply. */
2624 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2625 		    !error) {
2626 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2627 			if (*(tl + 1))
2628 				nd->nd_flag |= ND_NOMOREDATA;
2629 		}
2630 		if (!error)
2631 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2632 			    NULL, tstuff);
2633 	}
2634 	if (nd->nd_repstat && !error)
2635 		error = nd->nd_repstat;
2636 nfsmout:
2637 	m_freem(nd->nd_mrep);
2638 	return (error);
2639 }
2640 
2641 /*
2642  * nfs hard link create rpc
2643  */
2644 APPLESTATIC int
2645 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2646     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2647     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2648 {
2649 	u_int32_t *tl;
2650 	struct nfsrv_descript nfsd, *nd = &nfsd;
2651 	nfsattrbit_t attrbits;
2652 	int error = 0;
2653 
2654 	*attrflagp = 0;
2655 	*dattrflagp = 0;
2656 	if (namelen > NFS_MAXNAMLEN)
2657 		return (ENAMETOOLONG);
2658 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2659 	if (nd->nd_flag & ND_NFSV4) {
2660 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2661 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2662 	}
2663 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2664 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2665 	if (nd->nd_flag & ND_NFSV4) {
2666 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2667 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2668 		NFSWCCATTR_ATTRBIT(&attrbits);
2669 		(void) nfsrv_putattrbit(nd, &attrbits);
2670 		nd->nd_flag |= ND_V4WCCATTR;
2671 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2672 		*tl = txdr_unsigned(NFSV4OP_LINK);
2673 	}
2674 	(void) nfsm_strtom(nd, name, namelen);
2675 	error = nfscl_request(nd, vp, p, cred, dstuff);
2676 	if (error)
2677 		return (error);
2678 	if (nd->nd_flag & ND_NFSV3) {
2679 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2680 		if (!error)
2681 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2682 			    NULL, dstuff);
2683 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2684 		/*
2685 		 * First, parse out the PutFH and Getattr result.
2686 		 */
2687 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2688 		if (!(*(tl + 1)))
2689 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2690 		if (*(tl + 1))
2691 			nd->nd_flag |= ND_NOMOREDATA;
2692 		/*
2693 		 * Get the pre-op attributes.
2694 		 */
2695 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2696 	}
2697 	if (nd->nd_repstat && !error)
2698 		error = nd->nd_repstat;
2699 nfsmout:
2700 	m_freem(nd->nd_mrep);
2701 	return (error);
2702 }
2703 
2704 /*
2705  * nfs symbolic link create rpc
2706  */
2707 APPLESTATIC int
2708 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
2709     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2710     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2711     int *dattrflagp, void *dstuff)
2712 {
2713 	u_int32_t *tl;
2714 	struct nfsrv_descript nfsd, *nd = &nfsd;
2715 	struct nfsmount *nmp;
2716 	int slen, error = 0;
2717 
2718 	*nfhpp = NULL;
2719 	*attrflagp = 0;
2720 	*dattrflagp = 0;
2721 	nmp = VFSTONFS(vnode_mount(dvp));
2722 	slen = strlen(target);
2723 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2724 		return (ENAMETOOLONG);
2725 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2726 	if (nd->nd_flag & ND_NFSV4) {
2727 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2728 		*tl = txdr_unsigned(NFLNK);
2729 		(void) nfsm_strtom(nd, target, slen);
2730 	}
2731 	(void) nfsm_strtom(nd, name, namelen);
2732 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2733 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2734 	if (!(nd->nd_flag & ND_NFSV4))
2735 		(void) nfsm_strtom(nd, target, slen);
2736 	if (nd->nd_flag & ND_NFSV2)
2737 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2738 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2739 	if (error)
2740 		return (error);
2741 	if (nd->nd_flag & ND_NFSV4)
2742 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2743 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2744 		if (!nd->nd_repstat)
2745 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2746 		if (!error)
2747 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2748 			    NULL, dstuff);
2749 	}
2750 	if (nd->nd_repstat && !error)
2751 		error = nd->nd_repstat;
2752 	m_freem(nd->nd_mrep);
2753 	/*
2754 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2755 	 * Only do this if vfs.nfs.ignore_eexist is set.
2756 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2757 	 * should guarantee "exactly once" RPC semantics.
2758 	 */
2759 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2760 	    nmp->nm_minorvers == 0))
2761 		error = 0;
2762 	return (error);
2763 }
2764 
2765 /*
2766  * nfs make dir rpc
2767  */
2768 APPLESTATIC int
2769 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2770     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2771     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2772     int *dattrflagp, void *dstuff)
2773 {
2774 	u_int32_t *tl;
2775 	struct nfsrv_descript nfsd, *nd = &nfsd;
2776 	nfsattrbit_t attrbits;
2777 	int error = 0;
2778 	struct nfsfh *fhp;
2779 	struct nfsmount *nmp;
2780 
2781 	*nfhpp = NULL;
2782 	*attrflagp = 0;
2783 	*dattrflagp = 0;
2784 	nmp = VFSTONFS(vnode_mount(dvp));
2785 	fhp = VTONFS(dvp)->n_fhp;
2786 	if (namelen > NFS_MAXNAMLEN)
2787 		return (ENAMETOOLONG);
2788 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2789 	if (nd->nd_flag & ND_NFSV4) {
2790 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2791 		*tl = txdr_unsigned(NFDIR);
2792 	}
2793 	(void) nfsm_strtom(nd, name, namelen);
2794 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2795 	if (nd->nd_flag & ND_NFSV4) {
2796 		NFSGETATTR_ATTRBIT(&attrbits);
2797 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2798 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2799 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2800 		(void) nfsrv_putattrbit(nd, &attrbits);
2801 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2802 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2803 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2804 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2805 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2806 		(void) nfsrv_putattrbit(nd, &attrbits);
2807 	}
2808 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2809 	if (error)
2810 		return (error);
2811 	if (nd->nd_flag & ND_NFSV4)
2812 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2813 	if (!nd->nd_repstat && !error) {
2814 		if (nd->nd_flag & ND_NFSV4) {
2815 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2816 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2817 		}
2818 		if (!error)
2819 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2820 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2821 			/* Get rid of the PutFH and Getattr status values. */
2822 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2823 			/* Load the directory attributes. */
2824 			error = nfsm_loadattr(nd, dnap);
2825 			if (error == 0)
2826 				*dattrflagp = 1;
2827 		}
2828 	}
2829 	if ((nd->nd_flag & ND_NFSV3) && !error)
2830 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2831 	if (nd->nd_repstat && !error)
2832 		error = nd->nd_repstat;
2833 nfsmout:
2834 	m_freem(nd->nd_mrep);
2835 	/*
2836 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2837 	 * Only do this if vfs.nfs.ignore_eexist is set.
2838 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2839 	 * should guarantee "exactly once" RPC semantics.
2840 	 */
2841 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2842 	    nmp->nm_minorvers == 0))
2843 		error = 0;
2844 	return (error);
2845 }
2846 
2847 /*
2848  * nfs remove directory call
2849  */
2850 APPLESTATIC int
2851 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2852     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2853 {
2854 	struct nfsrv_descript nfsd, *nd = &nfsd;
2855 	int error = 0;
2856 
2857 	*dattrflagp = 0;
2858 	if (namelen > NFS_MAXNAMLEN)
2859 		return (ENAMETOOLONG);
2860 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2861 	(void) nfsm_strtom(nd, name, namelen);
2862 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2863 	if (error)
2864 		return (error);
2865 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2866 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2867 	if (nd->nd_repstat && !error)
2868 		error = nd->nd_repstat;
2869 	m_freem(nd->nd_mrep);
2870 	/*
2871 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2872 	 */
2873 	if (error == ENOENT)
2874 		error = 0;
2875 	return (error);
2876 }
2877 
2878 /*
2879  * Readdir rpc.
2880  * Always returns with either uio_resid unchanged, if you are at the
2881  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2882  * filled in.
2883  * I felt this would allow caching of directory blocks more easily
2884  * than returning a pertially filled block.
2885  * Directory offset cookies:
2886  * Oh my, what to do with them...
2887  * I can think of three ways to deal with them:
2888  * 1 - have the layer above these RPCs maintain a map between logical
2889  *     directory byte offsets and the NFS directory offset cookies
2890  * 2 - pass the opaque directory offset cookies up into userland
2891  *     and let the libc functions deal with them, via the system call
2892  * 3 - return them to userland in the "struct dirent", so future versions
2893  *     of libc can use them and do whatever is necessary to make things work
2894  *     above these rpc calls, in the meantime
2895  * For now, I do #3 by "hiding" the directory offset cookies after the
2896  * d_name field in struct dirent. This is space inside d_reclen that
2897  * will be ignored by anything that doesn't know about them.
2898  * The directory offset cookies are filled in as the last 8 bytes of
2899  * each directory entry, after d_name. Someday, the userland libc
2900  * functions may be able to use these. In the meantime, it satisfies
2901  * OpenBSD's requirements for cookies being returned.
2902  * If expects the directory offset cookie for the read to be in uio_offset
2903  * and returns the one for the next entry after this directory block in
2904  * there, as well.
2905  */
2906 APPLESTATIC int
2907 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2908     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2909     int *eofp, void *stuff)
2910 {
2911 	int len, left;
2912 	struct dirent *dp = NULL;
2913 	u_int32_t *tl;
2914 	nfsquad_t cookie, ncookie;
2915 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2916 	struct nfsnode *dnp = VTONFS(vp);
2917 	struct nfsvattr nfsva;
2918 	struct nfsrv_descript nfsd, *nd = &nfsd;
2919 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2920 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2921 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2922 	char *cp;
2923 	nfsattrbit_t attrbits, dattrbits;
2924 	u_int32_t rderr, *tl2 = NULL;
2925 	size_t tresid;
2926 
2927 	KASSERT(uiop->uio_iovcnt == 1 &&
2928 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
2929 	    ("nfs readdirrpc bad uio"));
2930 	ncookie.lval[0] = ncookie.lval[1] = 0;
2931 	/*
2932 	 * There is no point in reading a lot more than uio_resid, however
2933 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2934 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2935 	 * will never make readsize > nm_readdirsize.
2936 	 */
2937 	readsize = nmp->nm_readdirsize;
2938 	if (readsize > uio_uio_resid(uiop))
2939 		readsize = uio_uio_resid(uiop) + DIRBLKSIZ;
2940 
2941 	*attrflagp = 0;
2942 	if (eofp)
2943 		*eofp = 0;
2944 	tresid = uio_uio_resid(uiop);
2945 	cookie.lval[0] = cookiep->nfsuquad[0];
2946 	cookie.lval[1] = cookiep->nfsuquad[1];
2947 	nd->nd_mrep = NULL;
2948 
2949 	/*
2950 	 * For NFSv4, first create the "." and ".." entries.
2951 	 */
2952 	if (NFSHASNFSV4(nmp)) {
2953 		reqsize = 6 * NFSX_UNSIGNED;
2954 		NFSGETATTR_ATTRBIT(&dattrbits);
2955 		NFSZERO_ATTRBIT(&attrbits);
2956 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2957 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2958 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2959 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2960 			NFSSETBIT_ATTRBIT(&attrbits,
2961 			    NFSATTRBIT_MOUNTEDONFILEID);
2962 			gotmnton = 1;
2963 		} else {
2964 			/*
2965 			 * Must fake it. Use the fileno, except when the
2966 			 * fsid is != to that of the directory. For that
2967 			 * case, generate a fake fileno that is not the same.
2968 			 */
2969 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2970 			gotmnton = 0;
2971 		}
2972 
2973 		/*
2974 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2975 		 */
2976 		if (uiop->uio_offset == 0) {
2977 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2978 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2979 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2980 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2981 			(void) nfsrv_putattrbit(nd, &attrbits);
2982 			error = nfscl_request(nd, vp, p, cred, stuff);
2983 			if (error)
2984 			    return (error);
2985 			dotfileid = 0;	/* Fake out the compiler. */
2986 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2987 			    error = nfsm_loadattr(nd, &nfsva);
2988 			    if (error != 0)
2989 				goto nfsmout;
2990 			    dotfileid = nfsva.na_fileid;
2991 			}
2992 			if (nd->nd_repstat == 0) {
2993 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2994 			    len = fxdr_unsigned(int, *(tl + 4));
2995 			    if (len > 0 && len <= NFSX_V4FHMAX)
2996 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2997 			    else
2998 				error = EPERM;
2999 			    if (!error) {
3000 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3001 				nfsva.na_mntonfileno = UINT64_MAX;
3002 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3003 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3004 				    NULL, NULL, NULL, p, cred);
3005 				if (error) {
3006 				    dotdotfileid = dotfileid;
3007 				} else if (gotmnton) {
3008 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3009 					dotdotfileid = nfsva.na_mntonfileno;
3010 				    else
3011 					dotdotfileid = nfsva.na_fileid;
3012 				} else if (nfsva.na_filesid[0] ==
3013 				    dnp->n_vattr.na_filesid[0] &&
3014 				    nfsva.na_filesid[1] ==
3015 				    dnp->n_vattr.na_filesid[1]) {
3016 				    dotdotfileid = nfsva.na_fileid;
3017 				} else {
3018 				    do {
3019 					fakefileno--;
3020 				    } while (fakefileno ==
3021 					nfsva.na_fileid);
3022 				    dotdotfileid = fakefileno;
3023 				}
3024 			    }
3025 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3026 			    /*
3027 			     * Lookupp returns NFSERR_NOENT when we are
3028 			     * at the root, so just use the current dir.
3029 			     */
3030 			    nd->nd_repstat = 0;
3031 			    dotdotfileid = dotfileid;
3032 			} else {
3033 			    error = nd->nd_repstat;
3034 			}
3035 			m_freem(nd->nd_mrep);
3036 			if (error)
3037 			    return (error);
3038 			nd->nd_mrep = NULL;
3039 			dp = (struct dirent *)uio_iov_base(uiop);
3040 			dp->d_pad0 = dp->d_pad1 = 0;
3041 			dp->d_off = 0;
3042 			dp->d_type = DT_DIR;
3043 			dp->d_fileno = dotfileid;
3044 			dp->d_namlen = 1;
3045 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3046 			dp->d_name[0] = '.';
3047 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3048 			/*
3049 			 * Just make these offset cookie 0.
3050 			 */
3051 			tl = (u_int32_t *)&dp->d_name[8];
3052 			*tl++ = 0;
3053 			*tl = 0;
3054 			blksiz += dp->d_reclen;
3055 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3056 			uiop->uio_offset += dp->d_reclen;
3057 			uio_iov_base_add(uiop, dp->d_reclen);
3058 			uio_iov_len_add(uiop, -(dp->d_reclen));
3059 			dp = (struct dirent *)uio_iov_base(uiop);
3060 			dp->d_pad0 = dp->d_pad1 = 0;
3061 			dp->d_off = 0;
3062 			dp->d_type = DT_DIR;
3063 			dp->d_fileno = dotdotfileid;
3064 			dp->d_namlen = 2;
3065 			*((uint64_t *)dp->d_name) = 0;
3066 			dp->d_name[0] = '.';
3067 			dp->d_name[1] = '.';
3068 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3069 			/*
3070 			 * Just make these offset cookie 0.
3071 			 */
3072 			tl = (u_int32_t *)&dp->d_name[8];
3073 			*tl++ = 0;
3074 			*tl = 0;
3075 			blksiz += dp->d_reclen;
3076 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3077 			uiop->uio_offset += dp->d_reclen;
3078 			uio_iov_base_add(uiop, dp->d_reclen);
3079 			uio_iov_len_add(uiop, -(dp->d_reclen));
3080 		}
3081 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3082 	} else {
3083 		reqsize = 5 * NFSX_UNSIGNED;
3084 	}
3085 
3086 
3087 	/*
3088 	 * Loop around doing readdir rpc's of size readsize.
3089 	 * The stopping criteria is EOF or buffer full.
3090 	 */
3091 	while (more_dirs && bigenough) {
3092 		*attrflagp = 0;
3093 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3094 		if (nd->nd_flag & ND_NFSV2) {
3095 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3096 			*tl++ = cookie.lval[1];
3097 			*tl = txdr_unsigned(readsize);
3098 		} else {
3099 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3100 			*tl++ = cookie.lval[0];
3101 			*tl++ = cookie.lval[1];
3102 			if (cookie.qval == 0) {
3103 				*tl++ = 0;
3104 				*tl++ = 0;
3105 			} else {
3106 				NFSLOCKNODE(dnp);
3107 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3108 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3109 				NFSUNLOCKNODE(dnp);
3110 			}
3111 			if (nd->nd_flag & ND_NFSV4) {
3112 				*tl++ = txdr_unsigned(readsize);
3113 				*tl = txdr_unsigned(readsize);
3114 				(void) nfsrv_putattrbit(nd, &attrbits);
3115 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3116 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3117 				(void) nfsrv_putattrbit(nd, &dattrbits);
3118 			} else {
3119 				*tl = txdr_unsigned(readsize);
3120 			}
3121 		}
3122 		error = nfscl_request(nd, vp, p, cred, stuff);
3123 		if (error)
3124 			return (error);
3125 		if (!(nd->nd_flag & ND_NFSV2)) {
3126 			if (nd->nd_flag & ND_NFSV3)
3127 				error = nfscl_postop_attr(nd, nap, attrflagp,
3128 				    stuff);
3129 			if (!nd->nd_repstat && !error) {
3130 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3131 				NFSLOCKNODE(dnp);
3132 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3133 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3134 				NFSUNLOCKNODE(dnp);
3135 			}
3136 		}
3137 		if (nd->nd_repstat || error) {
3138 			if (!error)
3139 				error = nd->nd_repstat;
3140 			goto nfsmout;
3141 		}
3142 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3143 		more_dirs = fxdr_unsigned(int, *tl);
3144 		if (!more_dirs)
3145 			tryformoredirs = 0;
3146 
3147 		/* loop through the dir entries, doctoring them to 4bsd form */
3148 		while (more_dirs && bigenough) {
3149 			if (nd->nd_flag & ND_NFSV4) {
3150 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3151 				ncookie.lval[0] = *tl++;
3152 				ncookie.lval[1] = *tl++;
3153 				len = fxdr_unsigned(int, *tl);
3154 			} else if (nd->nd_flag & ND_NFSV3) {
3155 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3156 				nfsva.na_fileid = fxdr_hyper(tl);
3157 				tl += 2;
3158 				len = fxdr_unsigned(int, *tl);
3159 			} else {
3160 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3161 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3162 				    *tl++);
3163 				len = fxdr_unsigned(int, *tl);
3164 			}
3165 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3166 				error = EBADRPC;
3167 				goto nfsmout;
3168 			}
3169 			tlen = roundup2(len, 8);
3170 			if (tlen == len)
3171 				tlen += 8;  /* To ensure null termination. */
3172 			left = DIRBLKSIZ - blksiz;
3173 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3174 				NFSBZERO(uio_iov_base(uiop), left);
3175 				dp->d_reclen += left;
3176 				uio_iov_base_add(uiop, left);
3177 				uio_iov_len_add(uiop, -(left));
3178 				uio_uio_resid_add(uiop, -(left));
3179 				uiop->uio_offset += left;
3180 				blksiz = 0;
3181 			}
3182 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3183 			    uio_uio_resid(uiop))
3184 				bigenough = 0;
3185 			if (bigenough) {
3186 				dp = (struct dirent *)uio_iov_base(uiop);
3187 				dp->d_pad0 = dp->d_pad1 = 0;
3188 				dp->d_off = 0;
3189 				dp->d_namlen = len;
3190 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3191 				    NFSX_HYPER;
3192 				dp->d_type = DT_UNKNOWN;
3193 				blksiz += dp->d_reclen;
3194 				if (blksiz == DIRBLKSIZ)
3195 					blksiz = 0;
3196 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
3197 				uiop->uio_offset += DIRHDSIZ;
3198 				uio_iov_base_add(uiop, DIRHDSIZ);
3199 				uio_iov_len_add(uiop, -(DIRHDSIZ));
3200 				error = nfsm_mbufuio(nd, uiop, len);
3201 				if (error)
3202 					goto nfsmout;
3203 				cp = uio_iov_base(uiop);
3204 				tlen -= len;
3205 				NFSBZERO(cp, tlen);
3206 				cp += tlen;	/* points to cookie storage */
3207 				tl2 = (u_int32_t *)cp;
3208 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3209 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3210 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3211 				uiop->uio_offset += (tlen + NFSX_HYPER);
3212 			} else {
3213 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3214 				if (error)
3215 					goto nfsmout;
3216 			}
3217 			if (nd->nd_flag & ND_NFSV4) {
3218 				rderr = 0;
3219 				nfsva.na_mntonfileno = UINT64_MAX;
3220 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3221 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3222 				    NULL, NULL, &rderr, p, cred);
3223 				if (error)
3224 					goto nfsmout;
3225 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3226 			} else if (nd->nd_flag & ND_NFSV3) {
3227 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3228 				ncookie.lval[0] = *tl++;
3229 				ncookie.lval[1] = *tl++;
3230 			} else {
3231 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3232 				ncookie.lval[0] = 0;
3233 				ncookie.lval[1] = *tl++;
3234 			}
3235 			if (bigenough) {
3236 			    if (nd->nd_flag & ND_NFSV4) {
3237 				if (rderr) {
3238 				    dp->d_fileno = 0;
3239 				} else {
3240 				    if (gotmnton) {
3241 					if (nfsva.na_mntonfileno != UINT64_MAX)
3242 					    dp->d_fileno = nfsva.na_mntonfileno;
3243 					else
3244 					    dp->d_fileno = nfsva.na_fileid;
3245 				    } else if (nfsva.na_filesid[0] ==
3246 					dnp->n_vattr.na_filesid[0] &&
3247 					nfsva.na_filesid[1] ==
3248 					dnp->n_vattr.na_filesid[1]) {
3249 					dp->d_fileno = nfsva.na_fileid;
3250 				    } else {
3251 					do {
3252 					    fakefileno--;
3253 					} while (fakefileno ==
3254 					    nfsva.na_fileid);
3255 					dp->d_fileno = fakefileno;
3256 				    }
3257 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3258 				}
3259 			    } else {
3260 				dp->d_fileno = nfsva.na_fileid;
3261 			    }
3262 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3263 				ncookie.lval[0];
3264 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3265 				ncookie.lval[1];
3266 			}
3267 			more_dirs = fxdr_unsigned(int, *tl);
3268 		}
3269 		/*
3270 		 * If at end of rpc data, get the eof boolean
3271 		 */
3272 		if (!more_dirs) {
3273 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3274 			eof = fxdr_unsigned(int, *tl);
3275 			if (tryformoredirs)
3276 				more_dirs = !eof;
3277 			if (nd->nd_flag & ND_NFSV4) {
3278 				error = nfscl_postop_attr(nd, nap, attrflagp,
3279 				    stuff);
3280 				if (error)
3281 					goto nfsmout;
3282 			}
3283 		}
3284 		m_freem(nd->nd_mrep);
3285 		nd->nd_mrep = NULL;
3286 	}
3287 	/*
3288 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3289 	 * by increasing d_reclen for the last record.
3290 	 */
3291 	if (blksiz > 0) {
3292 		left = DIRBLKSIZ - blksiz;
3293 		NFSBZERO(uio_iov_base(uiop), left);
3294 		dp->d_reclen += left;
3295 		uio_iov_base_add(uiop, left);
3296 		uio_iov_len_add(uiop, -(left));
3297 		uio_uio_resid_add(uiop, -(left));
3298 		uiop->uio_offset += left;
3299 	}
3300 
3301 	/*
3302 	 * If returning no data, assume end of file.
3303 	 * If not bigenough, return not end of file, since you aren't
3304 	 *    returning all the data
3305 	 * Otherwise, return the eof flag from the server.
3306 	 */
3307 	if (eofp) {
3308 		if (tresid == ((size_t)(uio_uio_resid(uiop))))
3309 			*eofp = 1;
3310 		else if (!bigenough)
3311 			*eofp = 0;
3312 		else
3313 			*eofp = eof;
3314 	}
3315 
3316 	/*
3317 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3318 	 */
3319 	while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3320 		dp = (struct dirent *)uio_iov_base(uiop);
3321 		NFSBZERO(dp, DIRBLKSIZ);
3322 		dp->d_type = DT_UNKNOWN;
3323 		tl = (u_int32_t *)&dp->d_name[4];
3324 		*tl++ = cookie.lval[0];
3325 		*tl = cookie.lval[1];
3326 		dp->d_reclen = DIRBLKSIZ;
3327 		uio_iov_base_add(uiop, DIRBLKSIZ);
3328 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3329 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3330 		uiop->uio_offset += DIRBLKSIZ;
3331 	}
3332 
3333 nfsmout:
3334 	if (nd->nd_mrep != NULL)
3335 		m_freem(nd->nd_mrep);
3336 	return (error);
3337 }
3338 
3339 #ifndef APPLE
3340 /*
3341  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3342  * (Also used for NFS V4 when mount flag set.)
3343  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3344  */
3345 APPLESTATIC int
3346 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3347     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3348     int *eofp, void *stuff)
3349 {
3350 	int len, left;
3351 	struct dirent *dp = NULL;
3352 	u_int32_t *tl;
3353 	vnode_t newvp = NULLVP;
3354 	struct nfsrv_descript nfsd, *nd = &nfsd;
3355 	struct nameidata nami, *ndp = &nami;
3356 	struct componentname *cnp = &ndp->ni_cnd;
3357 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3358 	struct nfsnode *dnp = VTONFS(vp), *np;
3359 	struct nfsvattr nfsva;
3360 	struct nfsfh *nfhp;
3361 	nfsquad_t cookie, ncookie;
3362 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3363 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3364 	int isdotdot = 0, unlocknewvp = 0;
3365 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3366 	u_int64_t fileno = 0;
3367 	char *cp;
3368 	nfsattrbit_t attrbits, dattrbits;
3369 	size_t tresid;
3370 	u_int32_t *tl2 = NULL, rderr;
3371 	struct timespec dctime;
3372 
3373 	KASSERT(uiop->uio_iovcnt == 1 &&
3374 	    (uio_uio_resid(uiop) & (DIRBLKSIZ - 1)) == 0,
3375 	    ("nfs readdirplusrpc bad uio"));
3376 	ncookie.lval[0] = ncookie.lval[1] = 0;
3377 	timespecclear(&dctime);
3378 	*attrflagp = 0;
3379 	if (eofp != NULL)
3380 		*eofp = 0;
3381 	ndp->ni_dvp = vp;
3382 	nd->nd_mrep = NULL;
3383 	cookie.lval[0] = cookiep->nfsuquad[0];
3384 	cookie.lval[1] = cookiep->nfsuquad[1];
3385 	tresid = uio_uio_resid(uiop);
3386 
3387 	/*
3388 	 * For NFSv4, first create the "." and ".." entries.
3389 	 */
3390 	if (NFSHASNFSV4(nmp)) {
3391 		NFSGETATTR_ATTRBIT(&dattrbits);
3392 		NFSZERO_ATTRBIT(&attrbits);
3393 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3394 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3395 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3396 			NFSSETBIT_ATTRBIT(&attrbits,
3397 			    NFSATTRBIT_MOUNTEDONFILEID);
3398 			gotmnton = 1;
3399 		} else {
3400 			/*
3401 			 * Must fake it. Use the fileno, except when the
3402 			 * fsid is != to that of the directory. For that
3403 			 * case, generate a fake fileno that is not the same.
3404 			 */
3405 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3406 			gotmnton = 0;
3407 		}
3408 
3409 		/*
3410 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3411 		 */
3412 		if (uiop->uio_offset == 0) {
3413 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3414 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3415 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3416 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3417 			(void) nfsrv_putattrbit(nd, &attrbits);
3418 			error = nfscl_request(nd, vp, p, cred, stuff);
3419 			if (error)
3420 			    return (error);
3421 			dotfileid = 0;	/* Fake out the compiler. */
3422 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3423 			    error = nfsm_loadattr(nd, &nfsva);
3424 			    if (error != 0)
3425 				goto nfsmout;
3426 			    dctime = nfsva.na_ctime;
3427 			    dotfileid = nfsva.na_fileid;
3428 			}
3429 			if (nd->nd_repstat == 0) {
3430 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3431 			    len = fxdr_unsigned(int, *(tl + 4));
3432 			    if (len > 0 && len <= NFSX_V4FHMAX)
3433 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3434 			    else
3435 				error = EPERM;
3436 			    if (!error) {
3437 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3438 				nfsva.na_mntonfileno = UINT64_MAX;
3439 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3440 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3441 				    NULL, NULL, NULL, p, cred);
3442 				if (error) {
3443 				    dotdotfileid = dotfileid;
3444 				} else if (gotmnton) {
3445 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3446 					dotdotfileid = nfsva.na_mntonfileno;
3447 				    else
3448 					dotdotfileid = nfsva.na_fileid;
3449 				} else if (nfsva.na_filesid[0] ==
3450 				    dnp->n_vattr.na_filesid[0] &&
3451 				    nfsva.na_filesid[1] ==
3452 				    dnp->n_vattr.na_filesid[1]) {
3453 				    dotdotfileid = nfsva.na_fileid;
3454 				} else {
3455 				    do {
3456 					fakefileno--;
3457 				    } while (fakefileno ==
3458 					nfsva.na_fileid);
3459 				    dotdotfileid = fakefileno;
3460 				}
3461 			    }
3462 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3463 			    /*
3464 			     * Lookupp returns NFSERR_NOENT when we are
3465 			     * at the root, so just use the current dir.
3466 			     */
3467 			    nd->nd_repstat = 0;
3468 			    dotdotfileid = dotfileid;
3469 			} else {
3470 			    error = nd->nd_repstat;
3471 			}
3472 			m_freem(nd->nd_mrep);
3473 			if (error)
3474 			    return (error);
3475 			nd->nd_mrep = NULL;
3476 			dp = (struct dirent *)uio_iov_base(uiop);
3477 			dp->d_pad0 = dp->d_pad1 = 0;
3478 			dp->d_off = 0;
3479 			dp->d_type = DT_DIR;
3480 			dp->d_fileno = dotfileid;
3481 			dp->d_namlen = 1;
3482 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3483 			dp->d_name[0] = '.';
3484 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3485 			/*
3486 			 * Just make these offset cookie 0.
3487 			 */
3488 			tl = (u_int32_t *)&dp->d_name[8];
3489 			*tl++ = 0;
3490 			*tl = 0;
3491 			blksiz += dp->d_reclen;
3492 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3493 			uiop->uio_offset += dp->d_reclen;
3494 			uio_iov_base_add(uiop, dp->d_reclen);
3495 			uio_iov_len_add(uiop, -(dp->d_reclen));
3496 			dp = (struct dirent *)uio_iov_base(uiop);
3497 			dp->d_pad0 = dp->d_pad1 = 0;
3498 			dp->d_off = 0;
3499 			dp->d_type = DT_DIR;
3500 			dp->d_fileno = dotdotfileid;
3501 			dp->d_namlen = 2;
3502 			*((uint64_t *)dp->d_name) = 0;
3503 			dp->d_name[0] = '.';
3504 			dp->d_name[1] = '.';
3505 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3506 			/*
3507 			 * Just make these offset cookie 0.
3508 			 */
3509 			tl = (u_int32_t *)&dp->d_name[8];
3510 			*tl++ = 0;
3511 			*tl = 0;
3512 			blksiz += dp->d_reclen;
3513 			uio_uio_resid_add(uiop, -(dp->d_reclen));
3514 			uiop->uio_offset += dp->d_reclen;
3515 			uio_iov_base_add(uiop, dp->d_reclen);
3516 			uio_iov_len_add(uiop, -(dp->d_reclen));
3517 		}
3518 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3519 		if (gotmnton)
3520 			NFSSETBIT_ATTRBIT(&attrbits,
3521 			    NFSATTRBIT_MOUNTEDONFILEID);
3522 	}
3523 
3524 	/*
3525 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3526 	 * The stopping criteria is EOF or buffer full.
3527 	 */
3528 	while (more_dirs && bigenough) {
3529 		*attrflagp = 0;
3530 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3531  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3532 		*tl++ = cookie.lval[0];
3533 		*tl++ = cookie.lval[1];
3534 		if (cookie.qval == 0) {
3535 			*tl++ = 0;
3536 			*tl++ = 0;
3537 		} else {
3538 			NFSLOCKNODE(dnp);
3539 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3540 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3541 			NFSUNLOCKNODE(dnp);
3542 		}
3543 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3544 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3545 		if (nd->nd_flag & ND_NFSV4) {
3546 			(void) nfsrv_putattrbit(nd, &attrbits);
3547 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3548 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3549 			(void) nfsrv_putattrbit(nd, &dattrbits);
3550 		}
3551 		error = nfscl_request(nd, vp, p, cred, stuff);
3552 		if (error)
3553 			return (error);
3554 		if (nd->nd_flag & ND_NFSV3)
3555 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3556 		if (nd->nd_repstat || error) {
3557 			if (!error)
3558 				error = nd->nd_repstat;
3559 			goto nfsmout;
3560 		}
3561 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3562 			dctime = nap->na_ctime;
3563 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3564 		NFSLOCKNODE(dnp);
3565 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3566 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3567 		NFSUNLOCKNODE(dnp);
3568 		more_dirs = fxdr_unsigned(int, *tl);
3569 		if (!more_dirs)
3570 			tryformoredirs = 0;
3571 
3572 		/* loop through the dir entries, doctoring them to 4bsd form */
3573 		while (more_dirs && bigenough) {
3574 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3575 			if (nd->nd_flag & ND_NFSV4) {
3576 				ncookie.lval[0] = *tl++;
3577 				ncookie.lval[1] = *tl++;
3578 			} else {
3579 				fileno = fxdr_hyper(tl);
3580 				tl += 2;
3581 			}
3582 			len = fxdr_unsigned(int, *tl);
3583 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3584 				error = EBADRPC;
3585 				goto nfsmout;
3586 			}
3587 			tlen = roundup2(len, 8);
3588 			if (tlen == len)
3589 				tlen += 8;  /* To ensure null termination. */
3590 			left = DIRBLKSIZ - blksiz;
3591 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3592 				NFSBZERO(uio_iov_base(uiop), left);
3593 				dp->d_reclen += left;
3594 				uio_iov_base_add(uiop, left);
3595 				uio_iov_len_add(uiop, -(left));
3596 				uio_uio_resid_add(uiop, -(left));
3597 				uiop->uio_offset += left;
3598 				blksiz = 0;
3599 			}
3600 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3601 			    uio_uio_resid(uiop))
3602 				bigenough = 0;
3603 			if (bigenough) {
3604 				dp = (struct dirent *)uio_iov_base(uiop);
3605 				dp->d_pad0 = dp->d_pad1 = 0;
3606 				dp->d_off = 0;
3607 				dp->d_namlen = len;
3608 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3609 				    NFSX_HYPER;
3610 				dp->d_type = DT_UNKNOWN;
3611 				blksiz += dp->d_reclen;
3612 				if (blksiz == DIRBLKSIZ)
3613 					blksiz = 0;
3614 				uio_uio_resid_add(uiop, -(DIRHDSIZ));
3615 				uiop->uio_offset += DIRHDSIZ;
3616 				uio_iov_base_add(uiop, DIRHDSIZ);
3617 				uio_iov_len_add(uiop, -(DIRHDSIZ));
3618 				cnp->cn_nameptr = uio_iov_base(uiop);
3619 				cnp->cn_namelen = len;
3620 				NFSCNHASHZERO(cnp);
3621 				error = nfsm_mbufuio(nd, uiop, len);
3622 				if (error)
3623 					goto nfsmout;
3624 				cp = uio_iov_base(uiop);
3625 				tlen -= len;
3626 				NFSBZERO(cp, tlen);
3627 				cp += tlen;	/* points to cookie storage */
3628 				tl2 = (u_int32_t *)cp;
3629 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3630 				    cnp->cn_nameptr[1] == '.')
3631 					isdotdot = 1;
3632 				else
3633 					isdotdot = 0;
3634 				uio_iov_base_add(uiop, (tlen + NFSX_HYPER));
3635 				uio_iov_len_add(uiop, -(tlen + NFSX_HYPER));
3636 				uio_uio_resid_add(uiop, -(tlen + NFSX_HYPER));
3637 				uiop->uio_offset += (tlen + NFSX_HYPER);
3638 			} else {
3639 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3640 				if (error)
3641 					goto nfsmout;
3642 			}
3643 			nfhp = NULL;
3644 			if (nd->nd_flag & ND_NFSV3) {
3645 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3646 				ncookie.lval[0] = *tl++;
3647 				ncookie.lval[1] = *tl++;
3648 				attrflag = fxdr_unsigned(int, *tl);
3649 				if (attrflag) {
3650 				  error = nfsm_loadattr(nd, &nfsva);
3651 				  if (error)
3652 					goto nfsmout;
3653 				}
3654 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3655 				if (*tl) {
3656 					error = nfsm_getfh(nd, &nfhp);
3657 					if (error)
3658 					    goto nfsmout;
3659 				}
3660 				if (!attrflag && nfhp != NULL) {
3661 					free(nfhp, M_NFSFH);
3662 					nfhp = NULL;
3663 				}
3664 			} else {
3665 				rderr = 0;
3666 				nfsva.na_mntonfileno = 0xffffffff;
3667 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3668 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3669 				    NULL, NULL, &rderr, p, cred);
3670 				if (error)
3671 					goto nfsmout;
3672 			}
3673 
3674 			if (bigenough) {
3675 			    if (nd->nd_flag & ND_NFSV4) {
3676 				if (rderr) {
3677 				    dp->d_fileno = 0;
3678 				} else if (gotmnton) {
3679 				    if (nfsva.na_mntonfileno != 0xffffffff)
3680 					dp->d_fileno = nfsva.na_mntonfileno;
3681 				    else
3682 					dp->d_fileno = nfsva.na_fileid;
3683 				} else if (nfsva.na_filesid[0] ==
3684 				    dnp->n_vattr.na_filesid[0] &&
3685 				    nfsva.na_filesid[1] ==
3686 				    dnp->n_vattr.na_filesid[1]) {
3687 				    dp->d_fileno = nfsva.na_fileid;
3688 				} else {
3689 				    do {
3690 					fakefileno--;
3691 				    } while (fakefileno ==
3692 					nfsva.na_fileid);
3693 				    dp->d_fileno = fakefileno;
3694 				}
3695 			    } else {
3696 				dp->d_fileno = fileno;
3697 			    }
3698 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3699 				ncookie.lval[0];
3700 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3701 				ncookie.lval[1];
3702 
3703 			    if (nfhp != NULL) {
3704 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3705 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3706 				    VREF(vp);
3707 				    newvp = vp;
3708 				    unlocknewvp = 0;
3709 				    free(nfhp, M_NFSFH);
3710 				    np = dnp;
3711 				} else if (isdotdot != 0) {
3712 				    /*
3713 				     * Skip doing a nfscl_nget() call for "..".
3714 				     * There's a race between acquiring the nfs
3715 				     * node here and lookups that look for the
3716 				     * directory being read (in the parent).
3717 				     * It would try to get a lock on ".." here,
3718 				     * owning the lock on the directory being
3719 				     * read. Lookup will hold the lock on ".."
3720 				     * and try to acquire the lock on the
3721 				     * directory being read.
3722 				     * If the directory is unlocked/relocked,
3723 				     * then there is a LOR with the buflock
3724 				     * vp is relocked.
3725 				     */
3726 				    free(nfhp, M_NFSFH);
3727 				} else {
3728 				    error = nfscl_nget(vnode_mount(vp), vp,
3729 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3730 				    if (!error) {
3731 					newvp = NFSTOV(np);
3732 					unlocknewvp = 1;
3733 				    }
3734 				}
3735 				nfhp = NULL;
3736 				if (newvp != NULLVP) {
3737 				    error = nfscl_loadattrcache(&newvp,
3738 					&nfsva, NULL, NULL, 0, 0);
3739 				    if (error) {
3740 					if (unlocknewvp)
3741 					    vput(newvp);
3742 					else
3743 					    vrele(newvp);
3744 					goto nfsmout;
3745 				    }
3746 				    dp->d_type =
3747 					vtonfs_dtype(np->n_vattr.na_type);
3748 				    ndp->ni_vp = newvp;
3749 				    NFSCNHASH(cnp, HASHINIT);
3750 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3751 					(newvp->v_type != VDIR ||
3752 					 dctime.tv_sec != 0)) {
3753 					cache_enter_time(ndp->ni_dvp,
3754 					    ndp->ni_vp, cnp,
3755 					    &nfsva.na_ctime,
3756 					    newvp->v_type != VDIR ? NULL :
3757 					    &dctime);
3758 				    }
3759 				    if (unlocknewvp)
3760 					vput(newvp);
3761 				    else
3762 					vrele(newvp);
3763 				    newvp = NULLVP;
3764 				}
3765 			    }
3766 			} else if (nfhp != NULL) {
3767 			    free(nfhp, M_NFSFH);
3768 			}
3769 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3770 			more_dirs = fxdr_unsigned(int, *tl);
3771 		}
3772 		/*
3773 		 * If at end of rpc data, get the eof boolean
3774 		 */
3775 		if (!more_dirs) {
3776 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3777 			eof = fxdr_unsigned(int, *tl);
3778 			if (tryformoredirs)
3779 				more_dirs = !eof;
3780 			if (nd->nd_flag & ND_NFSV4) {
3781 				error = nfscl_postop_attr(nd, nap, attrflagp,
3782 				    stuff);
3783 				if (error)
3784 					goto nfsmout;
3785 			}
3786 		}
3787 		m_freem(nd->nd_mrep);
3788 		nd->nd_mrep = NULL;
3789 	}
3790 	/*
3791 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3792 	 * by increasing d_reclen for the last record.
3793 	 */
3794 	if (blksiz > 0) {
3795 		left = DIRBLKSIZ - blksiz;
3796 		NFSBZERO(uio_iov_base(uiop), left);
3797 		dp->d_reclen += left;
3798 		uio_iov_base_add(uiop, left);
3799 		uio_iov_len_add(uiop, -(left));
3800 		uio_uio_resid_add(uiop, -(left));
3801 		uiop->uio_offset += left;
3802 	}
3803 
3804 	/*
3805 	 * If returning no data, assume end of file.
3806 	 * If not bigenough, return not end of file, since you aren't
3807 	 *    returning all the data
3808 	 * Otherwise, return the eof flag from the server.
3809 	 */
3810 	if (eofp != NULL) {
3811 		if (tresid == uio_uio_resid(uiop))
3812 			*eofp = 1;
3813 		else if (!bigenough)
3814 			*eofp = 0;
3815 		else
3816 			*eofp = eof;
3817 	}
3818 
3819 	/*
3820 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3821 	 */
3822 	while (uio_uio_resid(uiop) > 0 && uio_uio_resid(uiop) != tresid) {
3823 		dp = (struct dirent *)uio_iov_base(uiop);
3824 		NFSBZERO(dp, DIRBLKSIZ);
3825 		dp->d_type = DT_UNKNOWN;
3826 		tl = (u_int32_t *)&dp->d_name[4];
3827 		*tl++ = cookie.lval[0];
3828 		*tl = cookie.lval[1];
3829 		dp->d_reclen = DIRBLKSIZ;
3830 		uio_iov_base_add(uiop, DIRBLKSIZ);
3831 		uio_iov_len_add(uiop, -(DIRBLKSIZ));
3832 		uio_uio_resid_add(uiop, -(DIRBLKSIZ));
3833 		uiop->uio_offset += DIRBLKSIZ;
3834 	}
3835 
3836 nfsmout:
3837 	if (nd->nd_mrep != NULL)
3838 		m_freem(nd->nd_mrep);
3839 	return (error);
3840 }
3841 #endif	/* !APPLE */
3842 
3843 /*
3844  * Nfs commit rpc
3845  */
3846 APPLESTATIC int
3847 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3848     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3849 {
3850 	u_int32_t *tl;
3851 	struct nfsrv_descript nfsd, *nd = &nfsd;
3852 	nfsattrbit_t attrbits;
3853 	int error;
3854 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3855 
3856 	*attrflagp = 0;
3857 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3858 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3859 	txdr_hyper(offset, tl);
3860 	tl += 2;
3861 	*tl = txdr_unsigned(cnt);
3862 	if (nd->nd_flag & ND_NFSV4) {
3863 		/*
3864 		 * And do a Getattr op.
3865 		 */
3866 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3867 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3868 		NFSGETATTR_ATTRBIT(&attrbits);
3869 		(void) nfsrv_putattrbit(nd, &attrbits);
3870 	}
3871 	error = nfscl_request(nd, vp, p, cred, stuff);
3872 	if (error)
3873 		return (error);
3874 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3875 	if (!error && !nd->nd_repstat) {
3876 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3877 		NFSLOCKMNT(nmp);
3878 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3879 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3880 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3881 		}
3882 		NFSUNLOCKMNT(nmp);
3883 		if (nd->nd_flag & ND_NFSV4)
3884 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3885 	}
3886 nfsmout:
3887 	if (!error && nd->nd_repstat)
3888 		error = nd->nd_repstat;
3889 	m_freem(nd->nd_mrep);
3890 	return (error);
3891 }
3892 
3893 /*
3894  * NFS byte range lock rpc.
3895  * (Mostly just calls one of the three lower level RPC routines.)
3896  */
3897 APPLESTATIC int
3898 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3899     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3900 {
3901 	struct nfscllockowner *lp;
3902 	struct nfsclclient *clp;
3903 	struct nfsfh *nfhp;
3904 	struct nfsrv_descript nfsd, *nd = &nfsd;
3905 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3906 	u_int64_t off, len;
3907 	off_t start, end;
3908 	u_int32_t clidrev = 0;
3909 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3910 	int callcnt, dorpc;
3911 
3912 	/*
3913 	 * Convert the flock structure into a start and end and do POSIX
3914 	 * bounds checking.
3915 	 */
3916 	switch (fl->l_whence) {
3917 	case SEEK_SET:
3918 	case SEEK_CUR:
3919 		/*
3920 		 * Caller is responsible for adding any necessary offset
3921 		 * when SEEK_CUR is used.
3922 		 */
3923 		start = fl->l_start;
3924 		off = fl->l_start;
3925 		break;
3926 	case SEEK_END:
3927 		start = size + fl->l_start;
3928 		off = size + fl->l_start;
3929 		break;
3930 	default:
3931 		return (EINVAL);
3932 	}
3933 	if (start < 0)
3934 		return (EINVAL);
3935 	if (fl->l_len != 0) {
3936 		end = start + fl->l_len - 1;
3937 		if (end < start)
3938 			return (EINVAL);
3939 	}
3940 
3941 	len = fl->l_len;
3942 	if (len == 0)
3943 		len = NFS64BITSSET;
3944 	retrycnt = 0;
3945 	do {
3946 	    nd->nd_repstat = 0;
3947 	    if (op == F_GETLK) {
3948 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3949 		if (error)
3950 			return (error);
3951 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3952 		if (!error) {
3953 			clidrev = clp->nfsc_clientidrev;
3954 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3955 			    p, id, flags);
3956 		} else if (error == -1) {
3957 			error = 0;
3958 		}
3959 		nfscl_clientrelease(clp);
3960 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3961 		/*
3962 		 * We must loop around for all lockowner cases.
3963 		 */
3964 		callcnt = 0;
3965 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3966 		if (error)
3967 			return (error);
3968 		do {
3969 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3970 			clp, id, flags, &lp, &dorpc);
3971 		    /*
3972 		     * If it returns a NULL lp, we're done.
3973 		     */
3974 		    if (lp == NULL) {
3975 			if (callcnt == 0)
3976 			    nfscl_clientrelease(clp);
3977 			else
3978 			    nfscl_releasealllocks(clp, vp, p, id, flags);
3979 			return (error);
3980 		    }
3981 		    if (nmp->nm_clp != NULL)
3982 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3983 		    else
3984 			clidrev = 0;
3985 		    /*
3986 		     * If the server doesn't support Posix lock semantics,
3987 		     * only allow locks on the entire file, since it won't
3988 		     * handle overlapping byte ranges.
3989 		     * There might still be a problem when a lock
3990 		     * upgrade/downgrade (read<->write) occurs, since the
3991 		     * server "might" expect an unlock first?
3992 		     */
3993 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
3994 			(off == 0 && len == NFS64BITSSET))) {
3995 			/*
3996 			 * Since the lock records will go away, we must
3997 			 * wait for grace and delay here.
3998 			 */
3999 			do {
4000 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4001 				NFSV4LOCKT_READ, cred, p, 0);
4002 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4003 				 nd->nd_repstat == NFSERR_DELAY) &&
4004 				error == 0)
4005 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4006 				    "nfs_advlock");
4007 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4008 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4009 		    }
4010 		    callcnt++;
4011 		} while (error == 0 && nd->nd_repstat == 0);
4012 		nfscl_releasealllocks(clp, vp, p, id, flags);
4013 	    } else if (op == F_SETLK) {
4014 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4015 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4016 		if (error || donelocally) {
4017 			return (error);
4018 		}
4019 		if (nmp->nm_clp != NULL)
4020 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4021 		else
4022 			clidrev = 0;
4023 		nfhp = VTONFS(vp)->n_fhp;
4024 		if (!lp->nfsl_open->nfso_posixlock &&
4025 		    (off != 0 || len != NFS64BITSSET)) {
4026 			error = EINVAL;
4027 		} else {
4028 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4029 			    nfhp->nfh_len, lp, newone, reclaim, off,
4030 			    len, fl->l_type, cred, p, 0);
4031 		}
4032 		if (!error)
4033 			error = nd->nd_repstat;
4034 		nfscl_lockrelease(lp, error, newone);
4035 	    } else {
4036 		error = EINVAL;
4037 	    }
4038 	    if (!error)
4039 	        error = nd->nd_repstat;
4040 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4041 		error == NFSERR_STALEDONTRECOVER ||
4042 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4043 		error == NFSERR_BADSESSION) {
4044 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4045 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4046 		&& clidrev != 0) {
4047 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4048 		retrycnt++;
4049 	    }
4050 	} while (error == NFSERR_GRACE ||
4051 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4052 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4053 	    error == NFSERR_BADSESSION ||
4054 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4055 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4056 	if (error && retrycnt >= 4)
4057 		error = EIO;
4058 	return (error);
4059 }
4060 
4061 /*
4062  * The lower level routine for the LockT case.
4063  */
4064 APPLESTATIC int
4065 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4066     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4067     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4068 {
4069 	u_int32_t *tl;
4070 	int error, type, size;
4071 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4072 	struct nfsnode *np;
4073 	struct nfsmount *nmp;
4074 	struct nfsclsession *tsep;
4075 
4076 	nmp = VFSTONFS(vp->v_mount);
4077 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4078 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4079 	if (fl->l_type == F_RDLCK)
4080 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4081 	else
4082 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4083 	txdr_hyper(off, tl);
4084 	tl += 2;
4085 	txdr_hyper(len, tl);
4086 	tl += 2;
4087 	tsep = nfsmnt_mdssession(nmp);
4088 	*tl++ = tsep->nfsess_clientid.lval[0];
4089 	*tl = tsep->nfsess_clientid.lval[1];
4090 	nfscl_filllockowner(id, own, flags);
4091 	np = VTONFS(vp);
4092 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4093 	    np->n_fhp->nfh_len);
4094 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4095 	error = nfscl_request(nd, vp, p, cred, NULL);
4096 	if (error)
4097 		return (error);
4098 	if (nd->nd_repstat == 0) {
4099 		fl->l_type = F_UNLCK;
4100 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4101 		nd->nd_repstat = 0;
4102 		fl->l_whence = SEEK_SET;
4103 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4104 		fl->l_start = fxdr_hyper(tl);
4105 		tl += 2;
4106 		len = fxdr_hyper(tl);
4107 		tl += 2;
4108 		if (len == NFS64BITSSET)
4109 			fl->l_len = 0;
4110 		else
4111 			fl->l_len = len;
4112 		type = fxdr_unsigned(int, *tl++);
4113 		if (type == NFSV4LOCKT_WRITE)
4114 			fl->l_type = F_WRLCK;
4115 		else
4116 			fl->l_type = F_RDLCK;
4117 		/*
4118 		 * XXX For now, I have no idea what to do with the
4119 		 * conflicting lock_owner, so I'll just set the pid == 0
4120 		 * and skip over the lock_owner.
4121 		 */
4122 		fl->l_pid = (pid_t)0;
4123 		tl += 2;
4124 		size = fxdr_unsigned(int, *tl);
4125 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4126 			error = EBADRPC;
4127 		if (!error)
4128 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4129 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4130 		nfscl_initiate_recovery(clp);
4131 nfsmout:
4132 	m_freem(nd->nd_mrep);
4133 	return (error);
4134 }
4135 
4136 /*
4137  * Lower level function that performs the LockU RPC.
4138  */
4139 static int
4140 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4141     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4142     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4143 {
4144 	u_int32_t *tl;
4145 	int error;
4146 
4147 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4148 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4149 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4150 	*tl++ = txdr_unsigned(type);
4151 	*tl = txdr_unsigned(lp->nfsl_seqid);
4152 	if (nfstest_outofseq &&
4153 	    (arc4random() % nfstest_outofseq) == 0)
4154 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4155 	tl++;
4156 	if (NFSHASNFSV4N(nmp))
4157 		*tl++ = 0;
4158 	else
4159 		*tl++ = lp->nfsl_stateid.seqid;
4160 	*tl++ = lp->nfsl_stateid.other[0];
4161 	*tl++ = lp->nfsl_stateid.other[1];
4162 	*tl++ = lp->nfsl_stateid.other[2];
4163 	txdr_hyper(off, tl);
4164 	tl += 2;
4165 	txdr_hyper(len, tl);
4166 	if (syscred)
4167 		nd->nd_flag |= ND_USEGSSNAME;
4168 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4169 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4170 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4171 	if (error)
4172 		return (error);
4173 	if (nd->nd_repstat == 0) {
4174 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4175 		lp->nfsl_stateid.seqid = *tl++;
4176 		lp->nfsl_stateid.other[0] = *tl++;
4177 		lp->nfsl_stateid.other[1] = *tl++;
4178 		lp->nfsl_stateid.other[2] = *tl;
4179 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4180 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4181 nfsmout:
4182 	m_freem(nd->nd_mrep);
4183 	return (error);
4184 }
4185 
4186 /*
4187  * The actual Lock RPC.
4188  */
4189 APPLESTATIC int
4190 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4191     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4192     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4193     NFSPROC_T *p, int syscred)
4194 {
4195 	u_int32_t *tl;
4196 	int error, size;
4197 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4198 	struct nfsclsession *tsep;
4199 
4200 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4201 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4202 	if (type == F_RDLCK)
4203 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4204 	else
4205 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4206 	*tl++ = txdr_unsigned(reclaim);
4207 	txdr_hyper(off, tl);
4208 	tl += 2;
4209 	txdr_hyper(len, tl);
4210 	tl += 2;
4211 	if (newone) {
4212 	    *tl = newnfs_true;
4213 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4214 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4215 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4216 	    if (NFSHASNFSV4N(nmp))
4217 		*tl++ = 0;
4218 	    else
4219 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4220 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4221 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4222 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4223 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4224 	    tsep = nfsmnt_mdssession(nmp);
4225 	    *tl++ = tsep->nfsess_clientid.lval[0];
4226 	    *tl = tsep->nfsess_clientid.lval[1];
4227 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4228 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4229 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4230 	} else {
4231 	    *tl = newnfs_false;
4232 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4233 	    if (NFSHASNFSV4N(nmp))
4234 		*tl++ = 0;
4235 	    else
4236 		*tl++ = lp->nfsl_stateid.seqid;
4237 	    *tl++ = lp->nfsl_stateid.other[0];
4238 	    *tl++ = lp->nfsl_stateid.other[1];
4239 	    *tl++ = lp->nfsl_stateid.other[2];
4240 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4241 	    if (nfstest_outofseq &&
4242 		(arc4random() % nfstest_outofseq) == 0)
4243 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4244 	}
4245 	if (syscred)
4246 		nd->nd_flag |= ND_USEGSSNAME;
4247 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4248 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4249 	if (error)
4250 		return (error);
4251 	if (newone)
4252 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4253 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4254 	if (nd->nd_repstat == 0) {
4255 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4256 		lp->nfsl_stateid.seqid = *tl++;
4257 		lp->nfsl_stateid.other[0] = *tl++;
4258 		lp->nfsl_stateid.other[1] = *tl++;
4259 		lp->nfsl_stateid.other[2] = *tl;
4260 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4261 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4262 		size = fxdr_unsigned(int, *(tl + 7));
4263 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4264 			error = EBADRPC;
4265 		if (!error)
4266 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4267 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4268 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4269 nfsmout:
4270 	m_freem(nd->nd_mrep);
4271 	return (error);
4272 }
4273 
4274 /*
4275  * nfs statfs rpc
4276  * (always called with the vp for the mount point)
4277  */
4278 APPLESTATIC int
4279 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4280     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4281     void *stuff)
4282 {
4283 	u_int32_t *tl = NULL;
4284 	struct nfsrv_descript nfsd, *nd = &nfsd;
4285 	struct nfsmount *nmp;
4286 	nfsattrbit_t attrbits;
4287 	int error;
4288 
4289 	*attrflagp = 0;
4290 	nmp = VFSTONFS(vnode_mount(vp));
4291 	if (NFSHASNFSV4(nmp)) {
4292 		/*
4293 		 * For V4, you actually do a getattr.
4294 		 */
4295 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4296 		NFSSTATFS_GETATTRBIT(&attrbits);
4297 		(void) nfsrv_putattrbit(nd, &attrbits);
4298 		nd->nd_flag |= ND_USEGSSNAME;
4299 		error = nfscl_request(nd, vp, p, cred, stuff);
4300 		if (error)
4301 			return (error);
4302 		if (nd->nd_repstat == 0) {
4303 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4304 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4305 			    cred);
4306 			if (!error) {
4307 				nmp->nm_fsid[0] = nap->na_filesid[0];
4308 				nmp->nm_fsid[1] = nap->na_filesid[1];
4309 				NFSSETHASSETFSID(nmp);
4310 				*attrflagp = 1;
4311 			}
4312 		} else {
4313 			error = nd->nd_repstat;
4314 		}
4315 		if (error)
4316 			goto nfsmout;
4317 	} else {
4318 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4319 		error = nfscl_request(nd, vp, p, cred, stuff);
4320 		if (error)
4321 			return (error);
4322 		if (nd->nd_flag & ND_NFSV3) {
4323 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4324 			if (error)
4325 				goto nfsmout;
4326 		}
4327 		if (nd->nd_repstat) {
4328 			error = nd->nd_repstat;
4329 			goto nfsmout;
4330 		}
4331 		NFSM_DISSECT(tl, u_int32_t *,
4332 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4333 	}
4334 	if (NFSHASNFSV3(nmp)) {
4335 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4336 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4337 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4338 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4339 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4340 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4341 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4342 	} else if (NFSHASNFSV4(nmp) == 0) {
4343 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4344 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4345 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4346 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4347 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4348 	}
4349 nfsmout:
4350 	m_freem(nd->nd_mrep);
4351 	return (error);
4352 }
4353 
4354 /*
4355  * nfs pathconf rpc
4356  */
4357 APPLESTATIC int
4358 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4359     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4360     void *stuff)
4361 {
4362 	struct nfsrv_descript nfsd, *nd = &nfsd;
4363 	struct nfsmount *nmp;
4364 	u_int32_t *tl;
4365 	nfsattrbit_t attrbits;
4366 	int error;
4367 
4368 	*attrflagp = 0;
4369 	nmp = VFSTONFS(vnode_mount(vp));
4370 	if (NFSHASNFSV4(nmp)) {
4371 		/*
4372 		 * For V4, you actually do a getattr.
4373 		 */
4374 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4375 		NFSPATHCONF_GETATTRBIT(&attrbits);
4376 		(void) nfsrv_putattrbit(nd, &attrbits);
4377 		nd->nd_flag |= ND_USEGSSNAME;
4378 		error = nfscl_request(nd, vp, p, cred, stuff);
4379 		if (error)
4380 			return (error);
4381 		if (nd->nd_repstat == 0) {
4382 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4383 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4384 			    cred);
4385 			if (!error)
4386 				*attrflagp = 1;
4387 		} else {
4388 			error = nd->nd_repstat;
4389 		}
4390 	} else {
4391 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4392 		error = nfscl_request(nd, vp, p, cred, stuff);
4393 		if (error)
4394 			return (error);
4395 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4396 		if (nd->nd_repstat && !error)
4397 			error = nd->nd_repstat;
4398 		if (!error) {
4399 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4400 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4401 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4402 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4403 			pc->pc_chownrestricted =
4404 			    fxdr_unsigned(u_int32_t, *tl++);
4405 			pc->pc_caseinsensitive =
4406 			    fxdr_unsigned(u_int32_t, *tl++);
4407 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4408 		}
4409 	}
4410 nfsmout:
4411 	m_freem(nd->nd_mrep);
4412 	return (error);
4413 }
4414 
4415 /*
4416  * nfs version 3 fsinfo rpc call
4417  */
4418 APPLESTATIC int
4419 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4420     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4421 {
4422 	u_int32_t *tl;
4423 	struct nfsrv_descript nfsd, *nd = &nfsd;
4424 	int error;
4425 
4426 	*attrflagp = 0;
4427 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4428 	error = nfscl_request(nd, vp, p, cred, stuff);
4429 	if (error)
4430 		return (error);
4431 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4432 	if (nd->nd_repstat && !error)
4433 		error = nd->nd_repstat;
4434 	if (!error) {
4435 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4436 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4437 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4438 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4439 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4440 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4441 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4442 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4443 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4444 		tl += 2;
4445 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4446 		tl += 2;
4447 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4448 	}
4449 nfsmout:
4450 	m_freem(nd->nd_mrep);
4451 	return (error);
4452 }
4453 
4454 /*
4455  * This function performs the Renew RPC.
4456  */
4457 APPLESTATIC int
4458 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4459     NFSPROC_T *p)
4460 {
4461 	u_int32_t *tl;
4462 	struct nfsrv_descript nfsd;
4463 	struct nfsrv_descript *nd = &nfsd;
4464 	struct nfsmount *nmp;
4465 	int error;
4466 	struct nfssockreq *nrp;
4467 	struct nfsclsession *tsep;
4468 
4469 	nmp = clp->nfsc_nmp;
4470 	if (nmp == NULL)
4471 		return (0);
4472 	if (dsp == NULL)
4473 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4474 		    0);
4475 	else
4476 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4477 		    &dsp->nfsclds_sess, 0, 0);
4478 	if (!NFSHASNFSV4N(nmp)) {
4479 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4480 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4481 		tsep = nfsmnt_mdssession(nmp);
4482 		*tl++ = tsep->nfsess_clientid.lval[0];
4483 		*tl = tsep->nfsess_clientid.lval[1];
4484 	}
4485 	nrp = NULL;
4486 	if (dsp != NULL)
4487 		nrp = dsp->nfsclds_sockp;
4488 	if (nrp == NULL)
4489 		/* If NULL, use the MDS socket. */
4490 		nrp = &nmp->nm_sockreq;
4491 	nd->nd_flag |= ND_USEGSSNAME;
4492 	if (dsp == NULL)
4493 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4494 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4495 	else {
4496 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4497 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4498 		if (error == ENXIO)
4499 			nfscl_cancelreqs(dsp);
4500 	}
4501 	if (error)
4502 		return (error);
4503 	error = nd->nd_repstat;
4504 	m_freem(nd->nd_mrep);
4505 	return (error);
4506 }
4507 
4508 /*
4509  * This function performs the Releaselockowner RPC.
4510  */
4511 APPLESTATIC int
4512 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4513     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4514 {
4515 	struct nfsrv_descript nfsd, *nd = &nfsd;
4516 	u_int32_t *tl;
4517 	int error;
4518 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4519 	struct nfsclsession *tsep;
4520 
4521 	if (NFSHASNFSV4N(nmp)) {
4522 		/* For NFSv4.1, do a FreeStateID. */
4523 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4524 		    NULL, 0, 0);
4525 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4526 	} else {
4527 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4528 		    NULL, 0, 0);
4529 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4530 		tsep = nfsmnt_mdssession(nmp);
4531 		*tl++ = tsep->nfsess_clientid.lval[0];
4532 		*tl = tsep->nfsess_clientid.lval[1];
4533 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4534 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4535 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4536 	}
4537 	nd->nd_flag |= ND_USEGSSNAME;
4538 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4539 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4540 	if (error)
4541 		return (error);
4542 	error = nd->nd_repstat;
4543 	m_freem(nd->nd_mrep);
4544 	return (error);
4545 }
4546 
4547 /*
4548  * This function performs the Compound to get the mount pt FH.
4549  */
4550 APPLESTATIC int
4551 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4552     NFSPROC_T *p)
4553 {
4554 	u_int32_t *tl;
4555 	struct nfsrv_descript nfsd;
4556 	struct nfsrv_descript *nd = &nfsd;
4557 	u_char *cp, *cp2;
4558 	int error, cnt, len, setnil;
4559 	u_int32_t *opcntp;
4560 
4561 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4562 	    0);
4563 	cp = dirpath;
4564 	cnt = 0;
4565 	do {
4566 		setnil = 0;
4567 		while (*cp == '/')
4568 			cp++;
4569 		cp2 = cp;
4570 		while (*cp2 != '\0' && *cp2 != '/')
4571 			cp2++;
4572 		if (*cp2 == '/') {
4573 			setnil = 1;
4574 			*cp2 = '\0';
4575 		}
4576 		if (cp2 != cp) {
4577 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4578 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4579 			nfsm_strtom(nd, cp, strlen(cp));
4580 			cnt++;
4581 		}
4582 		if (setnil)
4583 			*cp2++ = '/';
4584 		cp = cp2;
4585 	} while (*cp != '\0');
4586 	if (NFSHASNFSV4N(nmp))
4587 		/* Has a Sequence Op done by nfscl_reqstart(). */
4588 		*opcntp = txdr_unsigned(3 + cnt);
4589 	else
4590 		*opcntp = txdr_unsigned(2 + cnt);
4591 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4592 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4593 	nd->nd_flag |= ND_USEGSSNAME;
4594 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4595 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4596 	if (error)
4597 		return (error);
4598 	if (nd->nd_repstat == 0) {
4599 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4600 		tl += (2 + 2 * cnt);
4601 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4602 			len > NFSX_FHMAX) {
4603 			nd->nd_repstat = NFSERR_BADXDR;
4604 		} else {
4605 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4606 			if (nd->nd_repstat == 0)
4607 				nmp->nm_fhsize = len;
4608 		}
4609 	}
4610 	error = nd->nd_repstat;
4611 nfsmout:
4612 	m_freem(nd->nd_mrep);
4613 	return (error);
4614 }
4615 
4616 /*
4617  * This function performs the Delegreturn RPC.
4618  */
4619 APPLESTATIC int
4620 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4621     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4622 {
4623 	u_int32_t *tl;
4624 	struct nfsrv_descript nfsd;
4625 	struct nfsrv_descript *nd = &nfsd;
4626 	int error;
4627 
4628 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4629 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4630 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4631 	if (NFSHASNFSV4N(nmp))
4632 		*tl++ = 0;
4633 	else
4634 		*tl++ = dp->nfsdl_stateid.seqid;
4635 	*tl++ = dp->nfsdl_stateid.other[0];
4636 	*tl++ = dp->nfsdl_stateid.other[1];
4637 	*tl = dp->nfsdl_stateid.other[2];
4638 	if (syscred)
4639 		nd->nd_flag |= ND_USEGSSNAME;
4640 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4641 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4642 	if (error)
4643 		return (error);
4644 	error = nd->nd_repstat;
4645 	m_freem(nd->nd_mrep);
4646 	return (error);
4647 }
4648 
4649 /*
4650  * nfs getacl call.
4651  */
4652 APPLESTATIC int
4653 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4654     struct acl *aclp, void *stuff)
4655 {
4656 	struct nfsrv_descript nfsd, *nd = &nfsd;
4657 	int error;
4658 	nfsattrbit_t attrbits;
4659 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4660 
4661 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4662 		return (EOPNOTSUPP);
4663 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4664 	NFSZERO_ATTRBIT(&attrbits);
4665 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4666 	(void) nfsrv_putattrbit(nd, &attrbits);
4667 	error = nfscl_request(nd, vp, p, cred, stuff);
4668 	if (error)
4669 		return (error);
4670 	if (!nd->nd_repstat)
4671 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4672 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4673 	else
4674 		error = nd->nd_repstat;
4675 	m_freem(nd->nd_mrep);
4676 	return (error);
4677 }
4678 
4679 /*
4680  * nfs setacl call.
4681  */
4682 APPLESTATIC int
4683 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4684     struct acl *aclp, void *stuff)
4685 {
4686 	int error;
4687 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4688 
4689 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4690 		return (EOPNOTSUPP);
4691 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4692 	return (error);
4693 }
4694 
4695 /*
4696  * nfs setacl call.
4697  */
4698 static int
4699 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4700     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4701 {
4702 	struct nfsrv_descript nfsd, *nd = &nfsd;
4703 	int error;
4704 	nfsattrbit_t attrbits;
4705 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4706 
4707 	if (!NFSHASNFSV4(nmp))
4708 		return (EOPNOTSUPP);
4709 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4710 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4711 	NFSZERO_ATTRBIT(&attrbits);
4712 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4713 	(void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4714 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4715 	error = nfscl_request(nd, vp, p, cred, stuff);
4716 	if (error)
4717 		return (error);
4718 	/* Don't care about the pre/postop attributes */
4719 	m_freem(nd->nd_mrep);
4720 	return (nd->nd_repstat);
4721 }
4722 
4723 /*
4724  * Do the NFSv4.1 Exchange ID.
4725  */
4726 int
4727 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4728     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
4729     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
4730 {
4731 	uint32_t *tl, v41flags;
4732 	struct nfsrv_descript nfsd;
4733 	struct nfsrv_descript *nd = &nfsd;
4734 	struct nfsclds *dsp;
4735 	struct timespec verstime;
4736 	int error, len;
4737 
4738 	*dspp = NULL;
4739 	if (minorvers == 0)
4740 		minorvers = nmp->nm_minorvers;
4741 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
4742 	    NFS_VER4, minorvers);
4743 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4744 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4745 	*tl = txdr_unsigned(clp->nfsc_rev);
4746 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4747 
4748 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4749 	*tl++ = txdr_unsigned(exchflags);
4750 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4751 
4752 	/* Set the implementation id4 */
4753 	*tl = txdr_unsigned(1);
4754 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4755 	(void) nfsm_strtom(nd, version, strlen(version));
4756 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4757 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4758 	verstime.tv_nsec = 0;
4759 	txdr_nfsv4time(&verstime, tl);
4760 	nd->nd_flag |= ND_USEGSSNAME;
4761 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4762 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4763 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4764 	    (int)nd->nd_repstat);
4765 	if (error != 0)
4766 		return (error);
4767 	if (nd->nd_repstat == 0) {
4768 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4769 		len = fxdr_unsigned(int, *(tl + 7));
4770 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4771 			error = NFSERR_BADXDR;
4772 			goto nfsmout;
4773 		}
4774 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4775 		    M_WAITOK | M_ZERO);
4776 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4777 		dsp->nfsclds_servownlen = len;
4778 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4779 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4780 		dsp->nfsclds_sess.nfsess_sequenceid =
4781 		    fxdr_unsigned(uint32_t, *tl++);
4782 		v41flags = fxdr_unsigned(uint32_t, *tl);
4783 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4784 		    NFSHASPNFSOPT(nmp)) {
4785 			NFSCL_DEBUG(1, "set PNFS\n");
4786 			NFSLOCKMNT(nmp);
4787 			nmp->nm_state |= NFSSTA_PNFS;
4788 			NFSUNLOCKMNT(nmp);
4789 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4790 		}
4791 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4792 			dsp->nfsclds_flags |= NFSCLDS_DS;
4793 		if (minorvers == NFSV42_MINORVERSION)
4794 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
4795 		if (len > 0)
4796 			nd->nd_repstat = nfsrv_mtostr(nd,
4797 			    dsp->nfsclds_serverown, len);
4798 		if (nd->nd_repstat == 0) {
4799 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4800 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4801 			    NULL, MTX_DEF);
4802 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4803 			*dspp = dsp;
4804 		} else
4805 			free(dsp, M_NFSCLDS);
4806 	}
4807 	error = nd->nd_repstat;
4808 nfsmout:
4809 	m_freem(nd->nd_mrep);
4810 	return (error);
4811 }
4812 
4813 /*
4814  * Do the NFSv4.1 Create Session.
4815  */
4816 int
4817 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4818     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
4819     struct ucred *cred, NFSPROC_T *p)
4820 {
4821 	uint32_t crflags, maxval, *tl;
4822 	struct nfsrv_descript nfsd;
4823 	struct nfsrv_descript *nd = &nfsd;
4824 	int error, irdcnt, minorvers;
4825 
4826 	/* Make sure nm_rsize, nm_wsize is set. */
4827 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4828 		nmp->nm_rsize = NFS_MAXBSIZE;
4829 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4830 		nmp->nm_wsize = NFS_MAXBSIZE;
4831 	if (dsp == NULL)
4832 		minorvers = nmp->nm_minorvers;
4833 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
4834 		minorvers = NFSV42_MINORVERSION;
4835 	else
4836 		minorvers = NFSV41_MINORVERSION;
4837 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
4838 	    NFS_VER4, minorvers);
4839 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4840 	*tl++ = sep->nfsess_clientid.lval[0];
4841 	*tl++ = sep->nfsess_clientid.lval[1];
4842 	*tl++ = txdr_unsigned(sequenceid);
4843 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4844 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4845 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4846 	*tl = txdr_unsigned(crflags);
4847 
4848 	/* Fill in fore channel attributes. */
4849 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4850 	*tl++ = 0;				/* Header pad size */
4851 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
4852 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
4853 		/*
4854 		 * NFSv4.2 Extended Attribute operations may want to do
4855 		 * requests/replies that are larger than nm_rsize/nm_wsize.
4856 		 */
4857 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4858 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4859 	} else {
4860 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
4861 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
4862 	}
4863 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4864 	*tl++ = txdr_unsigned(20);		/* Max operations */
4865 	*tl++ = txdr_unsigned(64);		/* Max slots */
4866 	*tl = 0;				/* No rdma ird */
4867 
4868 	/* Fill in back channel attributes. */
4869 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4870 	*tl++ = 0;				/* Header pad size */
4871 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4872 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4873 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4874 	*tl++ = txdr_unsigned(4);		/* Max operations */
4875 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4876 	*tl = 0;				/* No rdma ird */
4877 
4878 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4879 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4880 
4881 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4882 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4883 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4884 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4885 	*tl++ = 0;				/* Null machine name */
4886 	*tl++ = 0;				/* Uid == 0 */
4887 	*tl++ = 0;				/* Gid == 0 */
4888 	*tl = 0;				/* No additional gids */
4889 	nd->nd_flag |= ND_USEGSSNAME;
4890 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4891 	    NFS_VER4, NULL, 1, NULL, NULL);
4892 	if (error != 0)
4893 		return (error);
4894 	if (nd->nd_repstat == 0) {
4895 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4896 		    2 * NFSX_UNSIGNED);
4897 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4898 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4899 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4900 		crflags = fxdr_unsigned(uint32_t, *tl);
4901 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4902 			NFSLOCKMNT(nmp);
4903 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4904 			NFSUNLOCKMNT(nmp);
4905 		}
4906 
4907 		/* Get the fore channel slot count. */
4908 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4909 		tl++;			/* Skip the header pad size. */
4910 
4911 		/* Make sure nm_wsize is small enough. */
4912 		maxval = fxdr_unsigned(uint32_t, *tl++);
4913 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4914 			if (nmp->nm_wsize > 8096)
4915 				nmp->nm_wsize /= 2;
4916 			else
4917 				break;
4918 		}
4919 		sep->nfsess_maxreq = maxval;
4920 
4921 		/* Make sure nm_rsize is small enough. */
4922 		maxval = fxdr_unsigned(uint32_t, *tl++);
4923 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4924 			if (nmp->nm_rsize > 8096)
4925 				nmp->nm_rsize /= 2;
4926 			else
4927 				break;
4928 		}
4929 		sep->nfsess_maxresp = maxval;
4930 
4931 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4932 		tl++;
4933 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4934 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4935 		irdcnt = fxdr_unsigned(int, *tl);
4936 		if (irdcnt > 0)
4937 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4938 
4939 		/* and the back channel slot count. */
4940 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4941 		tl += 5;
4942 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4943 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4944 	}
4945 	error = nd->nd_repstat;
4946 nfsmout:
4947 	m_freem(nd->nd_mrep);
4948 	return (error);
4949 }
4950 
4951 /*
4952  * Do the NFSv4.1 Destroy Session.
4953  */
4954 int
4955 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4956     struct ucred *cred, NFSPROC_T *p)
4957 {
4958 	uint32_t *tl;
4959 	struct nfsrv_descript nfsd;
4960 	struct nfsrv_descript *nd = &nfsd;
4961 	int error;
4962 	struct nfsclsession *tsep;
4963 
4964 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
4965 	    0);
4966 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4967 	tsep = nfsmnt_mdssession(nmp);
4968 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4969 	nd->nd_flag |= ND_USEGSSNAME;
4970 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4971 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4972 	if (error != 0)
4973 		return (error);
4974 	error = nd->nd_repstat;
4975 	m_freem(nd->nd_mrep);
4976 	return (error);
4977 }
4978 
4979 /*
4980  * Do the NFSv4.1 Destroy Client.
4981  */
4982 int
4983 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4984     struct ucred *cred, NFSPROC_T *p)
4985 {
4986 	uint32_t *tl;
4987 	struct nfsrv_descript nfsd;
4988 	struct nfsrv_descript *nd = &nfsd;
4989 	int error;
4990 	struct nfsclsession *tsep;
4991 
4992 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
4993 	    0);
4994 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4995 	tsep = nfsmnt_mdssession(nmp);
4996 	*tl++ = tsep->nfsess_clientid.lval[0];
4997 	*tl = tsep->nfsess_clientid.lval[1];
4998 	nd->nd_flag |= ND_USEGSSNAME;
4999 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5000 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5001 	if (error != 0)
5002 		return (error);
5003 	error = nd->nd_repstat;
5004 	m_freem(nd->nd_mrep);
5005 	return (error);
5006 }
5007 
5008 /*
5009  * Do the NFSv4.1 LayoutGet.
5010  */
5011 static int
5012 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5013     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5014     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5015     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
5016     void *stuff)
5017 {
5018 	struct nfsrv_descript nfsd, *nd = &nfsd;
5019 	int error;
5020 
5021 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5022 	    0);
5023 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5024 	    layouttype, layoutlen, 0);
5025 	nd->nd_flag |= ND_USEGSSNAME;
5026 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5027 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5028 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5029 	if (error != 0)
5030 		return (error);
5031 	if (nd->nd_repstat == 0)
5032 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5033 		    flhp);
5034 	if (error == 0 && nd->nd_repstat != 0)
5035 		error = nd->nd_repstat;
5036 	m_freem(nd->nd_mrep);
5037 	return (error);
5038 }
5039 
5040 /*
5041  * Do the NFSv4.1 Get Device Info.
5042  */
5043 int
5044 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5045     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5046     NFSPROC_T *p)
5047 {
5048 	uint32_t cnt, *tl, vers, minorvers;
5049 	struct nfsrv_descript nfsd;
5050 	struct nfsrv_descript *nd = &nfsd;
5051 	struct sockaddr_in sin, ssin;
5052 	struct sockaddr_in6 sin6, ssin6;
5053 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5054 	struct nfscldevinfo *ndi;
5055 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5056 	int stripecnt;
5057 	uint8_t stripeindex;
5058 	sa_family_t af, safilled;
5059 
5060 	ssin.sin_port = 0;		/* To shut up compiler. */
5061 	ssin.sin_addr.s_addr = 0;	/* ditto */
5062 	*ndip = NULL;
5063 	ndi = NULL;
5064 	gotdspp = NULL;
5065 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5066 	    0);
5067 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5068 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5069 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5070 	*tl++ = txdr_unsigned(layouttype);
5071 	*tl++ = txdr_unsigned(100000);
5072 	if (notifybitsp != NULL && *notifybitsp != 0) {
5073 		*tl = txdr_unsigned(1);		/* One word of bits. */
5074 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5075 		*tl = txdr_unsigned(*notifybitsp);
5076 	} else
5077 		*tl = txdr_unsigned(0);
5078 	nd->nd_flag |= ND_USEGSSNAME;
5079 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5080 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5081 	if (error != 0)
5082 		return (error);
5083 	if (nd->nd_repstat == 0) {
5084 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5085 		if (layouttype != fxdr_unsigned(int, *tl))
5086 			printf("EEK! devinfo layout type not same!\n");
5087 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5088 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5089 			stripecnt = fxdr_unsigned(int, *tl);
5090 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5091 			if (stripecnt < 1 || stripecnt > 4096) {
5092 				printf("pNFS File layout devinfo stripecnt %d:"
5093 				    " out of range\n", stripecnt);
5094 				error = NFSERR_BADXDR;
5095 				goto nfsmout;
5096 			}
5097 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5098 			    NFSX_UNSIGNED);
5099 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5100 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5101 			if (addrcnt < 1 || addrcnt > 128) {
5102 				printf("NFS devinfo addrcnt %d: out of range\n",
5103 				    addrcnt);
5104 				error = NFSERR_BADXDR;
5105 				goto nfsmout;
5106 			}
5107 
5108 			/*
5109 			 * Now we know how many stripe indices and addresses, so
5110 			 * we can allocate the structure the correct size.
5111 			 */
5112 			i = (stripecnt * sizeof(uint8_t)) /
5113 			    sizeof(struct nfsclds *) + 1;
5114 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5115 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5116 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5117 			    M_ZERO);
5118 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5119 			    NFSX_V4DEVICEID);
5120 			ndi->nfsdi_refcnt = 0;
5121 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5122 			ndi->nfsdi_stripecnt = stripecnt;
5123 			ndi->nfsdi_addrcnt = addrcnt;
5124 			/* Fill in the stripe indices. */
5125 			for (i = 0; i < stripecnt; i++) {
5126 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5127 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5128 				if (stripeindex >= addrcnt) {
5129 					printf("pNFS File Layout devinfo"
5130 					    " stripeindex %d: too big\n",
5131 					    (int)stripeindex);
5132 					error = NFSERR_BADXDR;
5133 					goto nfsmout;
5134 				}
5135 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5136 			}
5137 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5138 			/* For Flex File, we only get one address list. */
5139 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5140 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5141 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5142 			    NFSX_V4DEVICEID);
5143 			ndi->nfsdi_refcnt = 0;
5144 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5145 			addrcnt = ndi->nfsdi_addrcnt = 1;
5146 		}
5147 
5148 		/* Now, dissect the server address(es). */
5149 		safilled = AF_UNSPEC;
5150 		for (i = 0; i < addrcnt; i++) {
5151 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5152 			cnt = fxdr_unsigned(uint32_t, *tl);
5153 			if (cnt == 0) {
5154 				printf("NFS devinfo 0 len addrlist\n");
5155 				error = NFSERR_BADXDR;
5156 				goto nfsmout;
5157 			}
5158 			dspp = nfsfldi_addr(ndi, i);
5159 			safilled = AF_UNSPEC;
5160 			for (j = 0; j < cnt; j++) {
5161 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5162 				    &isudp);
5163 				if (error != 0 && error != EPERM) {
5164 					error = NFSERR_BADXDR;
5165 					goto nfsmout;
5166 				}
5167 				if (error == 0 && isudp == 0) {
5168 					/*
5169 					 * The priority is:
5170 					 * - Same address family.
5171 					 * Save the address and dspp, so that
5172 					 * the connection can be done after
5173 					 * parsing is complete.
5174 					 */
5175 					if (safilled == AF_UNSPEC ||
5176 					    (af == nmp->nm_nam->sa_family &&
5177 					     safilled != nmp->nm_nam->sa_family)
5178 					   ) {
5179 						if (af == AF_INET)
5180 							ssin = sin;
5181 						else
5182 							ssin6 = sin6;
5183 						safilled = af;
5184 						gotdspp = dspp;
5185 					}
5186 				}
5187 			}
5188 		}
5189 
5190 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5191 		gotminor = NFSV41_MINORVERSION;
5192 		/* For Flex File, we will take one of the versions to use. */
5193 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5194 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5195 			j = fxdr_unsigned(int, *tl);
5196 			if (j < 1 || j > NFSDEV_MAXVERS) {
5197 				printf("pNFS: too many versions\n");
5198 				error = NFSERR_BADXDR;
5199 				goto nfsmout;
5200 			}
5201 			gotvers = 0;
5202 			gotminor = 0;
5203 			for (i = 0; i < j; i++) {
5204 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5205 				vers = fxdr_unsigned(uint32_t, *tl++);
5206 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5207 				if (vers == NFS_VER3)
5208 					minorvers = 0;
5209 				if ((vers == NFS_VER4 && ((minorvers ==
5210 				    NFSV41_MINORVERSION && gotminor == 0) ||
5211 				    minorvers == NFSV42_MINORVERSION)) ||
5212 				    (vers == NFS_VER3 && gotvers == 0)) {
5213 					gotvers = vers;
5214 					gotminor = minorvers;
5215 					/* We'll take this one. */
5216 					ndi->nfsdi_versindex = i;
5217 					ndi->nfsdi_vers = vers;
5218 					ndi->nfsdi_minorvers = minorvers;
5219 					ndi->nfsdi_rsize = fxdr_unsigned(
5220 					    uint32_t, *tl++);
5221 					ndi->nfsdi_wsize = fxdr_unsigned(
5222 					    uint32_t, *tl++);
5223 					if (*tl == newnfs_true)
5224 						ndi->nfsdi_flags |=
5225 						    NFSDI_TIGHTCOUPLED;
5226 					else
5227 						ndi->nfsdi_flags &=
5228 						    ~NFSDI_TIGHTCOUPLED;
5229 				}
5230 			}
5231 			if (gotvers == 0) {
5232 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5233 				error = NFSERR_BADXDR;
5234 				goto nfsmout;
5235 			}
5236 		}
5237 
5238 		/* And the notify bits. */
5239 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5240 		bitcnt = fxdr_unsigned(int, *tl);
5241 		if (bitcnt > 0) {
5242 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5243 			if (notifybitsp != NULL)
5244 				*notifybitsp =
5245 				    fxdr_unsigned(uint32_t, *tl);
5246 		}
5247 		if (safilled != AF_UNSPEC) {
5248 			KASSERT(ndi != NULL, ("ndi is NULL"));
5249 			*ndip = ndi;
5250 		} else
5251 			error = EPERM;
5252 		if (error == 0) {
5253 			/*
5254 			 * Now we can do a TCP connection for the correct
5255 			 * NFS version and IP address.
5256 			 */
5257 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5258 			    gotvers, gotminor, &dsp, p);
5259 		}
5260 		if (error == 0) {
5261 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5262 			*gotdspp = dsp;
5263 		}
5264 	}
5265 	if (nd->nd_repstat != 0 && error == 0)
5266 		error = nd->nd_repstat;
5267 nfsmout:
5268 	if (error != 0 && ndi != NULL)
5269 		nfscl_freedevinfo(ndi);
5270 	m_freem(nd->nd_mrep);
5271 	return (error);
5272 }
5273 
5274 /*
5275  * Do the NFSv4.1 LayoutCommit.
5276  */
5277 int
5278 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5279     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5280     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5281 {
5282 	uint32_t *tl;
5283 	struct nfsrv_descript nfsd, *nd = &nfsd;
5284 	int error;
5285 
5286 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5287 	    0, 0);
5288 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5289 	    NFSX_STATEID);
5290 	txdr_hyper(off, tl);
5291 	tl += 2;
5292 	txdr_hyper(len, tl);
5293 	tl += 2;
5294 	if (reclaim != 0)
5295 		*tl++ = newnfs_true;
5296 	else
5297 		*tl++ = newnfs_false;
5298 	*tl++ = txdr_unsigned(stateidp->seqid);
5299 	*tl++ = stateidp->other[0];
5300 	*tl++ = stateidp->other[1];
5301 	*tl++ = stateidp->other[2];
5302 	*tl++ = newnfs_true;
5303 	if (lastbyte < off)
5304 		lastbyte = off;
5305 	else if (lastbyte >= (off + len))
5306 		lastbyte = off + len - 1;
5307 	txdr_hyper(lastbyte, tl);
5308 	tl += 2;
5309 	*tl++ = newnfs_false;
5310 	*tl++ = txdr_unsigned(layouttype);
5311 	/* All supported layouts are 0 length. */
5312 	*tl = txdr_unsigned(0);
5313 	nd->nd_flag |= ND_USEGSSNAME;
5314 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5315 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5316 	if (error != 0)
5317 		return (error);
5318 	error = nd->nd_repstat;
5319 	m_freem(nd->nd_mrep);
5320 	return (error);
5321 }
5322 
5323 /*
5324  * Do the NFSv4.1 LayoutReturn.
5325  */
5326 int
5327 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5328     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5329     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5330     uint32_t stat, uint32_t op, char *devid)
5331 {
5332 	uint32_t *tl;
5333 	struct nfsrv_descript nfsd, *nd = &nfsd;
5334 	uint64_t tu64;
5335 	int error;
5336 
5337 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5338 	    0, 0);
5339 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5340 	if (reclaim != 0)
5341 		*tl++ = newnfs_true;
5342 	else
5343 		*tl++ = newnfs_false;
5344 	*tl++ = txdr_unsigned(layouttype);
5345 	*tl++ = txdr_unsigned(iomode);
5346 	*tl = txdr_unsigned(layoutreturn);
5347 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5348 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5349 		    NFSX_UNSIGNED);
5350 		txdr_hyper(offset, tl);
5351 		tl += 2;
5352 		txdr_hyper(len, tl);
5353 		tl += 2;
5354 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5355 		*tl++ = txdr_unsigned(stateidp->seqid);
5356 		*tl++ = stateidp->other[0];
5357 		*tl++ = stateidp->other[1];
5358 		*tl++ = stateidp->other[2];
5359 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5360 			*tl = txdr_unsigned(0);
5361 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5362 			if (stat != 0) {
5363 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5364 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5365 				    NFSX_UNSIGNED);
5366 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5367 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5368 				    NFSX_UNSIGNED);
5369 				*tl++ = txdr_unsigned(1);	/* One error. */
5370 				tu64 = 0;			/* Offset. */
5371 				txdr_hyper(tu64, tl); tl += 2;
5372 				tu64 = UINT64_MAX;		/* Length. */
5373 				txdr_hyper(tu64, tl); tl += 2;
5374 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5375 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5376 				*tl++ = txdr_unsigned(1);	/* One error. */
5377 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5378 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5379 				*tl++ = txdr_unsigned(stat);
5380 				*tl++ = txdr_unsigned(op);
5381 			} else {
5382 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5383 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5384 				/* No ioerrs. */
5385 				*tl++ = 0;
5386 			}
5387 			*tl = 0;	/* No stats yet. */
5388 		}
5389 	}
5390 	nd->nd_flag |= ND_USEGSSNAME;
5391 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5392 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5393 	if (error != 0)
5394 		return (error);
5395 	if (nd->nd_repstat == 0) {
5396 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5397 		if (*tl != 0) {
5398 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5399 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5400 			stateidp->other[0] = *tl++;
5401 			stateidp->other[1] = *tl++;
5402 			stateidp->other[2] = *tl;
5403 		}
5404 	} else
5405 		error = nd->nd_repstat;
5406 nfsmout:
5407 	m_freem(nd->nd_mrep);
5408 	return (error);
5409 }
5410 
5411 /*
5412  * Acquire a layout and devinfo, if possible. The caller must have acquired
5413  * a reference count on the nfsclclient structure before calling this.
5414  * Return the layout in lypp with a reference count on it, if successful.
5415  */
5416 static int
5417 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5418     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5419     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5420 {
5421 	struct nfscllayout *lyp;
5422 	struct nfsclflayout *flp;
5423 	struct nfsclflayouthead flh;
5424 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5425 	nfsv4stateid_t stateid;
5426 	struct nfsclsession *tsep;
5427 
5428 	*lypp = NULL;
5429 	if (NFSHASFLEXFILE(nmp))
5430 		layouttype = NFSLAYOUT_FLEXFILE;
5431 	else
5432 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5433 	/*
5434 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5435 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5436 	 * flp == NULL.
5437 	 */
5438 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5439 	    off, &flp, &recalled);
5440 	islocked = 0;
5441 	if (lyp == NULL || flp == NULL) {
5442 		if (recalled != 0)
5443 			return (EIO);
5444 		LIST_INIT(&flh);
5445 		tsep = nfsmnt_mdssession(nmp);
5446 		layoutlen = tsep->nfsess_maxcache -
5447 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5448 		if (lyp == NULL) {
5449 			stateid.seqid = 0;
5450 			stateid.other[0] = stateidp->other[0];
5451 			stateid.other[1] = stateidp->other[1];
5452 			stateid.other[2] = stateidp->other[2];
5453 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5454 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5455 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5456 			    &retonclose, &flh, cred, p, NULL);
5457 		} else {
5458 			islocked = 1;
5459 			stateid.seqid = lyp->nfsly_stateid.seqid;
5460 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5461 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5462 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5463 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5464 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5465 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5466 			    &retonclose, &flh, cred, p, NULL);
5467 		}
5468 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5469 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5470 		    &flh, layouttype, error, NULL, cred, p);
5471 		if (error == 0)
5472 			*lypp = lyp;
5473 		else if (islocked != 0)
5474 			nfscl_rellayout(lyp, 1);
5475 	} else
5476 		*lypp = lyp;
5477 	return (error);
5478 }
5479 
5480 /*
5481  * Do a TCP connection plus exchange id and create session.
5482  * If successful, a "struct nfsclds" is linked into the list for the
5483  * mount point and a pointer to it is returned.
5484  */
5485 static int
5486 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5487     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5488     struct nfsclds **dspp, NFSPROC_T *p)
5489 {
5490 	struct sockaddr_in *msad, *sad;
5491 	struct sockaddr_in6 *msad6, *sad6;
5492 	struct nfsclclient *clp;
5493 	struct nfssockreq *nrp;
5494 	struct nfsclds *dsp, *tdsp;
5495 	int error, firsttry;
5496 	enum nfsclds_state retv;
5497 	uint32_t sequenceid = 0;
5498 
5499 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5500 	    ("nfsrpc_fillsa: NULL nr_cred"));
5501 	NFSLOCKCLSTATE();
5502 	clp = nmp->nm_clp;
5503 	NFSUNLOCKCLSTATE();
5504 	if (clp == NULL)
5505 		return (EPERM);
5506 	if (af == AF_INET) {
5507 		NFSLOCKMNT(nmp);
5508 		/*
5509 		 * Check to see if we already have a session for this
5510 		 * address that is usable for a DS.
5511 		 * Note that the MDS's address is in a different place
5512 		 * than the sessions already acquired for DS's.
5513 		 */
5514 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5515 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5516 		while (tdsp != NULL) {
5517 			if (msad != NULL && msad->sin_family == AF_INET &&
5518 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5519 			    sin->sin_port == msad->sin_port &&
5520 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5521 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5522 				*dspp = tdsp;
5523 				NFSUNLOCKMNT(nmp);
5524 				NFSCL_DEBUG(4, "fnd same addr\n");
5525 				return (0);
5526 			}
5527 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5528 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5529 				msad = (struct sockaddr_in *)
5530 				    tdsp->nfsclds_sockp->nr_nam;
5531 			else
5532 				msad = NULL;
5533 		}
5534 		NFSUNLOCKMNT(nmp);
5535 
5536 		/* No IP address match, so look for new/trunked one. */
5537 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5538 		sad->sin_len = sizeof(*sad);
5539 		sad->sin_family = AF_INET;
5540 		sad->sin_port = sin->sin_port;
5541 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5542 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5543 		nrp->nr_nam = (struct sockaddr *)sad;
5544 	} else if (af == AF_INET6) {
5545 		NFSLOCKMNT(nmp);
5546 		/*
5547 		 * Check to see if we already have a session for this
5548 		 * address that is usable for a DS.
5549 		 * Note that the MDS's address is in a different place
5550 		 * than the sessions already acquired for DS's.
5551 		 */
5552 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5553 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5554 		while (tdsp != NULL) {
5555 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5556 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5557 			    &msad6->sin6_addr) &&
5558 			    sin6->sin6_port == msad6->sin6_port &&
5559 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5560 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5561 				*dspp = tdsp;
5562 				NFSUNLOCKMNT(nmp);
5563 				return (0);
5564 			}
5565 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5566 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5567 				msad6 = (struct sockaddr_in6 *)
5568 				    tdsp->nfsclds_sockp->nr_nam;
5569 			else
5570 				msad6 = NULL;
5571 		}
5572 		NFSUNLOCKMNT(nmp);
5573 
5574 		/* No IP address match, so look for new/trunked one. */
5575 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5576 		sad6->sin6_len = sizeof(*sad6);
5577 		sad6->sin6_family = AF_INET6;
5578 		sad6->sin6_port = sin6->sin6_port;
5579 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5580 		    sizeof(struct in6_addr));
5581 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5582 		nrp->nr_nam = (struct sockaddr *)sad6;
5583 	} else
5584 		return (EPERM);
5585 
5586 	nrp->nr_sotype = SOCK_STREAM;
5587 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5588 	nrp->nr_prog = NFS_PROG;
5589 	nrp->nr_vers = vers;
5590 
5591 	/*
5592 	 * Use the credentials that were used for the mount, which are
5593 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5594 	 * Ref. counting the credentials with crhold() is probably not
5595 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5596 	 * unmount, but I did it anyhow.
5597 	 */
5598 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5599 	error = newnfs_connect(nmp, nrp, NULL, p, 0);
5600 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5601 
5602 	dsp = NULL;
5603 	/* Now, do the exchangeid and create session. */
5604 	if (error == 0) {
5605 		if (vers == NFS_VER4) {
5606 			firsttry = 0;
5607 			do {
5608 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5609 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5610 				    nrp->nr_cred, p);
5611 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5612 				if (error == NFSERR_MINORVERMISMATCH)
5613 					minorvers = NFSV42_MINORVERSION;
5614 			} while (error == NFSERR_MINORVERMISMATCH &&
5615 			    firsttry++ == 0);
5616 			if (error != 0)
5617 				newnfs_disconnect(nrp);
5618 		} else {
5619 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5620 			    M_WAITOK | M_ZERO);
5621 			dsp->nfsclds_flags |= NFSCLDS_DS;
5622 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5623 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5624 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5625 			    NULL, MTX_DEF);
5626 		}
5627 	}
5628 	if (error == 0) {
5629 		dsp->nfsclds_sockp = nrp;
5630 		if (vers == NFS_VER4) {
5631 			NFSLOCKMNT(nmp);
5632 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5633 			    &sequenceid);
5634 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5635 			if (retv == NFSDSP_USETHISSESSION &&
5636 			    nfscl_dssameconn != 0) {
5637 				NFSLOCKDS(tdsp);
5638 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5639 				NFSUNLOCKDS(tdsp);
5640 				NFSUNLOCKMNT(nmp);
5641 				/*
5642 				 * If there is already a session for this
5643 				 * server, use it.
5644 				 */
5645 				(void)newnfs_disconnect(nrp);
5646 				nfscl_freenfsclds(dsp);
5647 				*dspp = tdsp;
5648 				return (0);
5649 			}
5650 			if (retv == NFSDSP_NOTFOUND)
5651 				sequenceid =
5652 				    dsp->nfsclds_sess.nfsess_sequenceid;
5653 			NFSUNLOCKMNT(nmp);
5654 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5655 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
5656 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5657 		}
5658 	} else {
5659 		NFSFREECRED(nrp->nr_cred);
5660 		NFSFREEMUTEX(&nrp->nr_mtx);
5661 		free(nrp->nr_nam, M_SONAME);
5662 		free(nrp, M_NFSSOCKREQ);
5663 	}
5664 	if (error == 0) {
5665 		NFSCL_DEBUG(3, "add DS session\n");
5666 		/*
5667 		 * Put it at the end of the list. That way the list
5668 		 * is ordered by when the entry was added. This matters
5669 		 * since the one done first is the one that should be
5670 		 * used for sequencid'ing any subsequent create sessions.
5671 		 */
5672 		NFSLOCKMNT(nmp);
5673 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5674 		NFSUNLOCKMNT(nmp);
5675 		*dspp = dsp;
5676 	} else if (dsp != NULL) {
5677 		newnfs_disconnect(nrp);
5678 		nfscl_freenfsclds(dsp);
5679 	}
5680 	return (error);
5681 }
5682 
5683 /*
5684  * Do the NFSv4.1 Reclaim Complete.
5685  */
5686 int
5687 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5688 {
5689 	uint32_t *tl;
5690 	struct nfsrv_descript nfsd;
5691 	struct nfsrv_descript *nd = &nfsd;
5692 	int error;
5693 
5694 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5695 	    0);
5696 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5697 	*tl = newnfs_false;
5698 	nd->nd_flag |= ND_USEGSSNAME;
5699 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5700 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5701 	if (error != 0)
5702 		return (error);
5703 	error = nd->nd_repstat;
5704 	m_freem(nd->nd_mrep);
5705 	return (error);
5706 }
5707 
5708 /*
5709  * Initialize the slot tables for a session.
5710  */
5711 static void
5712 nfscl_initsessionslots(struct nfsclsession *sep)
5713 {
5714 	int i;
5715 
5716 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5717 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5718 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5719 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5720 	}
5721 	for (i = 0; i < 64; i++)
5722 		sep->nfsess_slotseq[i] = 0;
5723 	sep->nfsess_slots = 0;
5724 }
5725 
5726 /*
5727  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5728  */
5729 int
5730 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5731     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5732 {
5733 	struct nfsnode *np = VTONFS(vp);
5734 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5735 	struct nfscllayout *layp;
5736 	struct nfscldevinfo *dip;
5737 	struct nfsclflayout *rflp;
5738 	struct mbuf *m;
5739 	struct nfsclwritedsdorpc *drpc, *tdrpc;
5740 	nfsv4stateid_t stateid;
5741 	struct ucred *newcred;
5742 	uint64_t lastbyte, len, off, oresid, xfer;
5743 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5744 	void *lckp;
5745 	uint8_t *dev;
5746 	void *iovbase = NULL;
5747 	size_t iovlen = 0;
5748 	off_t offs = 0;
5749 	ssize_t resid = 0;
5750 
5751 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5752 	    (np->n_flag & NNOLAYOUT) != 0)
5753 		return (EIO);
5754 	/* Now, get a reference cnt on the clientid for this mount. */
5755 	if (nfscl_getref(nmp) == 0)
5756 		return (EIO);
5757 
5758 	/* Find an appropriate stateid. */
5759 	newcred = NFSNEWCRED(cred);
5760 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5761 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5762 	if (error != 0) {
5763 		NFSFREECRED(newcred);
5764 		nfscl_relref(nmp);
5765 		return (error);
5766 	}
5767 	/* Search for a layout for this file. */
5768 	off = uiop->uio_offset;
5769 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5770 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5771 	if (layp == NULL || rflp == NULL) {
5772 		if (recalled != 0) {
5773 			NFSFREECRED(newcred);
5774 			nfscl_relref(nmp);
5775 			return (EIO);
5776 		}
5777 		if (layp != NULL) {
5778 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5779 			layp = NULL;
5780 		}
5781 		/* Try and get a Layout, if it is supported. */
5782 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5783 		    (np->n_flag & NWRITEOPENED) != 0)
5784 			iolaymode = NFSLAYOUTIOMODE_RW;
5785 		else
5786 			iolaymode = NFSLAYOUTIOMODE_READ;
5787 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5788 		    NULL, &stateid, off, &layp, newcred, p);
5789 		if (error != 0) {
5790 			NFSLOCKNODE(np);
5791 			np->n_flag |= NNOLAYOUT;
5792 			NFSUNLOCKNODE(np);
5793 			if (lckp != NULL)
5794 				nfscl_lockderef(lckp);
5795 			NFSFREECRED(newcred);
5796 			if (layp != NULL)
5797 				nfscl_rellayout(layp, 0);
5798 			nfscl_relref(nmp);
5799 			return (error);
5800 		}
5801 	}
5802 
5803 	/*
5804 	 * Loop around finding a layout that works for the first part of
5805 	 * this I/O operation, and then call the function that actually
5806 	 * does the RPC.
5807 	 */
5808 	eof = 0;
5809 	len = (uint64_t)uiop->uio_resid;
5810 	while (len > 0 && error == 0 && eof == 0) {
5811 		off = uiop->uio_offset;
5812 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5813 		if (error == 0) {
5814 			oresid = xfer = (uint64_t)uiop->uio_resid;
5815 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5816 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5817 			/*
5818 			 * For Flex File layout with mirrored DSs, select one
5819 			 * of them at random for reads. For writes and commits,
5820 			 * do all mirrors.
5821 			 */
5822 			m = NULL;
5823 			tdrpc = drpc = NULL;
5824 			firstmirror = 0;
5825 			mirrorcnt = 1;
5826 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5827 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5828 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5829 					firstmirror = arc4random() % mirrorcnt;
5830 					mirrorcnt = firstmirror + 1;
5831 				} else {
5832 					if (docommit == 0) {
5833 						/*
5834 						 * Save values, so uiop can be
5835 						 * rolled back upon a write
5836 						 * error.
5837 						 */
5838 						offs = uiop->uio_offset;
5839 						resid = uiop->uio_resid;
5840 						iovbase =
5841 						    uiop->uio_iov->iov_base;
5842 						iovlen = uiop->uio_iov->iov_len;
5843 						m = nfsm_uiombuflist(uiop, len,
5844 						    NULL, NULL);
5845 					}
5846 					tdrpc = drpc = malloc(sizeof(*drpc) *
5847 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
5848 					    M_ZERO);
5849 				}
5850 			}
5851 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5852 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
5853 					dev = rflp->nfsfl_ffm[i].dev;
5854 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5855 					    rflp->nfsfl_ffm[i].devp);
5856 				} else {
5857 					dev = rflp->nfsfl_dev;
5858 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5859 					    rflp->nfsfl_devp);
5860 				}
5861 				if (dip != NULL) {
5862 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5863 					    != 0)
5864 						error = nfscl_dofflayoutio(vp,
5865 						    uiop, iomode, must_commit,
5866 						    &eof, &stateid, rwaccess,
5867 						    dip, layp, rflp, off, xfer,
5868 						    i, docommit, m, tdrpc,
5869 						    newcred, p);
5870 					else
5871 						error = nfscl_doflayoutio(vp,
5872 						    uiop, iomode, must_commit,
5873 						    &eof, &stateid, rwaccess,
5874 						    dip, layp, rflp, off, xfer,
5875 						    docommit, newcred, p);
5876 					nfscl_reldevinfo(dip);
5877 				} else
5878 					error = EIO;
5879 				tdrpc++;
5880 			}
5881 			if (m != NULL)
5882 				m_freem(m);
5883 			tdrpc = drpc;
5884 			timo = hz / 50;		/* Wait for 20msec. */
5885 			if (timo < 1)
5886 				timo = 1;
5887 			for (i = firstmirror; i < mirrorcnt - 1 &&
5888 			    tdrpc != NULL; i++, tdrpc++) {
5889 				/*
5890 				 * For the unused drpc entries, both inprog and
5891 				 * err == 0, so this loop won't break.
5892 				 */
5893 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
5894 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5895 					    timo);
5896 				if (error == 0 && tdrpc->err != 0)
5897 					error = tdrpc->err;
5898 			}
5899 			free(drpc, M_TEMP);
5900 			if (error == 0) {
5901 				if (mirrorcnt > 1 && rwaccess ==
5902 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5903 					NFSLOCKCLSTATE();
5904 					layp->nfsly_flags |= NFSLY_WRITTEN;
5905 					NFSUNLOCKCLSTATE();
5906 				}
5907 				lastbyte = off + xfer - 1;
5908 				NFSLOCKCLSTATE();
5909 				if (lastbyte > layp->nfsly_lastbyte)
5910 					layp->nfsly_lastbyte = lastbyte;
5911 				NFSUNLOCKCLSTATE();
5912 			} else if (error == NFSERR_OPENMODE &&
5913 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
5914 				NFSLOCKMNT(nmp);
5915 				nmp->nm_state |= NFSSTA_OPENMODE;
5916 				NFSUNLOCKMNT(nmp);
5917 			} else
5918 				error = EIO;
5919 			if (error == 0)
5920 				len -= (oresid - (uint64_t)uiop->uio_resid);
5921 			else if (mirrorcnt > 1 && rwaccess ==
5922 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5923 				/*
5924 				 * In case the rpc gets retried, roll the
5925 				 * uio fields changed by nfsm_uiombuflist()
5926 				 * back.
5927 				 */
5928 				uiop->uio_offset = offs;
5929 				uiop->uio_resid = resid;
5930 				uiop->uio_iov->iov_base = iovbase;
5931 				uiop->uio_iov->iov_len = iovlen;
5932 			}
5933 		}
5934 	}
5935 	if (lckp != NULL)
5936 		nfscl_lockderef(lckp);
5937 	NFSFREECRED(newcred);
5938 	nfscl_rellayout(layp, 0);
5939 	nfscl_relref(nmp);
5940 	return (error);
5941 }
5942 
5943 /*
5944  * Make a copy of the mbuf chain and add an mbuf for null padding, as required.
5945  */
5946 static struct mbuf *
5947 nfsm_copym(struct mbuf *m, int off, int xfer)
5948 {
5949 	struct mbuf *m2, *m3, *m4;
5950 	uint32_t *tl;
5951 	int rem;
5952 
5953 	m2 = m_copym(m, off, xfer, M_WAITOK);
5954 	rem = NFSM_RNDUP(xfer) - xfer;
5955 	if (rem > 0) {
5956 		/*
5957 		 * The zero padding to a multiple of 4 bytes is required by
5958 		 * the XDR. So that the mbufs copied by reference aren't
5959 		 * modified, add an mbuf with the zero'd bytes to the list.
5960 		 * rem will be a maximum of 3, so one zero'd uint32_t is
5961 		 * sufficient.
5962 		 */
5963 		m3 = m2;
5964 		while (m3->m_next != NULL)
5965 			m3 = m3->m_next;
5966 		NFSMGET(m4);
5967 		tl = mtod(m4, uint32_t *);
5968 		*tl = 0;
5969 		m4->m_len = rem;
5970 		m3->m_next = m4;
5971 	}
5972 	return (m2);
5973 }
5974 
5975 /*
5976  * Find a file layout that will handle the first bytes of the requested
5977  * range and return the information from it needed to the I/O operation.
5978  */
5979 int
5980 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5981     struct nfsclflayout **retflpp)
5982 {
5983 	struct nfsclflayout *flp, *nflp, *rflp;
5984 	uint32_t rw;
5985 
5986 	rflp = NULL;
5987 	rw = rwaccess;
5988 	/* For reading, do the Read list first and then the Write list. */
5989 	do {
5990 		if (rw == NFSV4OPEN_ACCESSREAD)
5991 			flp = LIST_FIRST(&lyp->nfsly_flayread);
5992 		else
5993 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
5994 		while (flp != NULL) {
5995 			nflp = LIST_NEXT(flp, nfsfl_list);
5996 			if (flp->nfsfl_off > off)
5997 				break;
5998 			if (flp->nfsfl_end > off &&
5999 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6000 				rflp = flp;
6001 			flp = nflp;
6002 		}
6003 		if (rw == NFSV4OPEN_ACCESSREAD)
6004 			rw = NFSV4OPEN_ACCESSWRITE;
6005 		else
6006 			rw = 0;
6007 	} while (rw != 0);
6008 	if (rflp != NULL) {
6009 		/* This one covers the most bytes starting at off. */
6010 		*retflpp = rflp;
6011 		return (0);
6012 	}
6013 	return (EIO);
6014 }
6015 
6016 /*
6017  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6018  */
6019 static int
6020 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6021     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6022     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6023     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6024 {
6025 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6026 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6027 	struct nfsnode *np;
6028 	struct nfsfh *fhp;
6029 	struct nfsclds **dspp;
6030 
6031 	np = VTONFS(vp);
6032 	rel_off = off - flp->nfsfl_patoff;
6033 	stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
6034 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6035 	    dp->nfsdi_stripecnt;
6036 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6037 	error = 0;
6038 
6039 	/* Loop around, doing I/O for each stripe unit. */
6040 	while (len > 0 && error == 0) {
6041 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6042 		dspp = nfsfldi_addr(dp, stripe_index);
6043 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6044 			minorvers = NFSV42_MINORVERSION;
6045 		else
6046 			minorvers = NFSV41_MINORVERSION;
6047 		if (len > transfer && docommit == 0)
6048 			xfer = transfer;
6049 		else
6050 			xfer = len;
6051 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6052 			/* Dense layout. */
6053 			if (stripe_pos >= flp->nfsfl_fhcnt)
6054 				return (EIO);
6055 			fhp = flp->nfsfl_fh[stripe_pos];
6056 			io_off = (rel_off / (stripe_unit_size *
6057 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6058 			    rel_off % stripe_unit_size;
6059 		} else {
6060 			/* Sparse layout. */
6061 			if (flp->nfsfl_fhcnt > 1) {
6062 				if (stripe_index >= flp->nfsfl_fhcnt)
6063 					return (EIO);
6064 				fhp = flp->nfsfl_fh[stripe_index];
6065 			} else if (flp->nfsfl_fhcnt == 1)
6066 				fhp = flp->nfsfl_fh[0];
6067 			else
6068 				fhp = np->n_fhp;
6069 			io_off = off;
6070 		}
6071 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6072 			commit_thru_mds = 1;
6073 			if (docommit != 0)
6074 				error = EIO;
6075 		} else {
6076 			commit_thru_mds = 0;
6077 			NFSLOCKNODE(np);
6078 			np->n_flag |= NDSCOMMIT;
6079 			NFSUNLOCKNODE(np);
6080 		}
6081 		if (docommit != 0) {
6082 			if (error == 0)
6083 				error = nfsrpc_commitds(vp, io_off, xfer,
6084 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6085 			if (error == 0) {
6086 				/*
6087 				 * Set both eof and uio_resid = 0 to end any
6088 				 * loops.
6089 				 */
6090 				*eofp = 1;
6091 				uiop->uio_resid = 0;
6092 			} else {
6093 				NFSLOCKNODE(np);
6094 				np->n_flag &= ~NDSCOMMIT;
6095 				NFSUNLOCKNODE(np);
6096 			}
6097 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6098 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6099 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6100 		else {
6101 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6102 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6103 			    0, NFS_VER4, minorvers, cred, p);
6104 			if (error == 0) {
6105 				NFSLOCKCLSTATE();
6106 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6107 				NFSUNLOCKCLSTATE();
6108 			}
6109 		}
6110 		if (error == 0) {
6111 			transfer = stripe_unit_size;
6112 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6113 			len -= xfer;
6114 			off += xfer;
6115 		}
6116 	}
6117 	return (error);
6118 }
6119 
6120 /*
6121  * Do I/O using an NFSv4.1 flex file layout.
6122  */
6123 static int
6124 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6125     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6126     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6127     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6128     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6129 {
6130 	uint64_t transfer, xfer;
6131 	int error, rel_off;
6132 	struct nfsnode *np;
6133 	struct nfsfh *fhp;
6134 	struct nfsclds **dspp;
6135 	struct ucred *tcred;
6136 	struct mbuf *m;
6137 
6138 	np = VTONFS(vp);
6139 	error = 0;
6140 	rel_off = 0;
6141 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6142 	    (uintmax_t)len);
6143 	/* Loop around, doing I/O for each stripe unit. */
6144 	while (len > 0 && error == 0) {
6145 		dspp = nfsfldi_addr(dp, 0);
6146 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6147 		stateidp = &flp->nfsfl_ffm[mirror].st;
6148 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6149 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6150 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6151 			tcred = NFSNEWCRED(cred);
6152 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6153 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6154 			tcred->cr_ngroups = 1;
6155 		} else
6156 			tcred = cred;
6157 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6158 			transfer = dp->nfsdi_rsize;
6159 		else
6160 			transfer = dp->nfsdi_wsize;
6161 		NFSLOCKNODE(np);
6162 		np->n_flag |= NDSCOMMIT;
6163 		NFSUNLOCKNODE(np);
6164 		if (len > transfer && docommit == 0)
6165 			xfer = transfer;
6166 		else
6167 			xfer = len;
6168 		if (docommit != 0) {
6169 			if (error == 0) {
6170 				/*
6171 				 * Do last mirrored DS commit with this thread.
6172 				 */
6173 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6174 					error = nfsio_commitds(vp, off, xfer,
6175 					    *dspp, fhp, dp->nfsdi_vers,
6176 					    dp->nfsdi_minorvers, drpc, tcred,
6177 					    p);
6178 				else
6179 					error = nfsrpc_commitds(vp, off, xfer,
6180 					    *dspp, fhp, dp->nfsdi_vers,
6181 					    dp->nfsdi_minorvers, tcred, p);
6182 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6183 				if (error != 0 && error != EACCES && error !=
6184 				    ESTALE) {
6185 					NFSCL_DEBUG(4,
6186 					    "DS layreterr for commit\n");
6187 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6188 					    lyp, *dspp);
6189 				}
6190 			}
6191 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6192 			if (error == 0) {
6193 				/*
6194 				 * Set both eof and uio_resid = 0 to end any
6195 				 * loops.
6196 				 */
6197 				*eofp = 1;
6198 				uiop->uio_resid = 0;
6199 			} else {
6200 				NFSLOCKNODE(np);
6201 				np->n_flag &= ~NDSCOMMIT;
6202 				NFSUNLOCKNODE(np);
6203 			}
6204 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6205 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6206 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6207 			    dp->nfsdi_minorvers, tcred, p);
6208 			NFSCL_DEBUG(4, "readds=%d\n", error);
6209 			if (error != 0 && error != EACCES && error != ESTALE) {
6210 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6211 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6212 				    *dspp);
6213 			}
6214 		} else {
6215 			if (flp->nfsfl_mirrorcnt == 1) {
6216 				error = nfsrpc_writeds(vp, uiop, iomode,
6217 				    must_commit, stateidp, *dspp, off, xfer,
6218 				    fhp, 0, 1, dp->nfsdi_vers,
6219 				    dp->nfsdi_minorvers, tcred, p);
6220 				if (error == 0) {
6221 					NFSLOCKCLSTATE();
6222 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6223 					NFSUNLOCKCLSTATE();
6224 				}
6225 			} else {
6226 				m = nfsm_copym(mp, rel_off, xfer);
6227 				NFSCL_DEBUG(4, "mcopy reloff=%d xfer=%jd\n",
6228 				    rel_off, (uintmax_t)xfer);
6229 				/*
6230 				 * Do last write to a mirrored DS with this
6231 				 * thread.
6232 				 */
6233 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6234 					error = nfsio_writedsmir(vp, iomode,
6235 					    must_commit, stateidp, *dspp, off,
6236 					    xfer, fhp, m, dp->nfsdi_vers,
6237 					    dp->nfsdi_minorvers, drpc, tcred,
6238 					    p);
6239 				else
6240 					error = nfsrpc_writedsmir(vp, iomode,
6241 					    must_commit, stateidp, *dspp, off,
6242 					    xfer, fhp, m, dp->nfsdi_vers,
6243 					    dp->nfsdi_minorvers, tcred, p);
6244 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6245 				if (error != 0 && error != EACCES && error !=
6246 				    ESTALE) {
6247 					NFSCL_DEBUG(4,
6248 					    "DS layreterr for write\n");
6249 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6250 					    lyp, *dspp);
6251 				}
6252 			}
6253 		}
6254 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6255 		if (error == 0) {
6256 			len -= xfer;
6257 			off += xfer;
6258 			rel_off += xfer;
6259 		}
6260 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6261 			NFSFREECRED(tcred);
6262 	}
6263 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6264 	return (error);
6265 }
6266 
6267 /*
6268  * The actual read RPC done to a DS.
6269  */
6270 static int
6271 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6272     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6273     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6274 {
6275 	uint32_t *tl;
6276 	int attrflag, error, retlen;
6277 	struct nfsrv_descript nfsd;
6278 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6279 	struct nfsrv_descript *nd = &nfsd;
6280 	struct nfssockreq *nrp;
6281 	struct nfsvattr na;
6282 
6283 	nd->nd_mrep = NULL;
6284 	if (vers == 0 || vers == NFS_VER4) {
6285 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6286 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6287 		vers = NFS_VER4;
6288 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6289 		if (flex != 0)
6290 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6291 		else
6292 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6293 	} else {
6294 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6295 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6296 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6297 	}
6298 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6299 	txdr_hyper(io_off, tl);
6300 	*(tl + 2) = txdr_unsigned(len);
6301 	nrp = dsp->nfsclds_sockp;
6302 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6303 	if (nrp == NULL)
6304 		/* If NULL, use the MDS socket. */
6305 		nrp = &nmp->nm_sockreq;
6306 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6307 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6308 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6309 	    error);
6310 	if (error != 0)
6311 		return (error);
6312 	if (vers == NFS_VER3) {
6313 		error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6314 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6315 		if (error != 0)
6316 			goto nfsmout;
6317 	}
6318 	if (nd->nd_repstat != 0) {
6319 		error = nd->nd_repstat;
6320 		goto nfsmout;
6321 	}
6322 	if (vers == NFS_VER3) {
6323 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6324 		*eofp = fxdr_unsigned(int, *(tl + 1));
6325 	} else {
6326 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6327 		*eofp = fxdr_unsigned(int, *tl);
6328 	}
6329 	NFSM_STRSIZ(retlen, len);
6330 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6331 	error = nfsm_mbufuio(nd, uiop, retlen);
6332 nfsmout:
6333 	if (nd->nd_mrep != NULL)
6334 		m_freem(nd->nd_mrep);
6335 	return (error);
6336 }
6337 
6338 /*
6339  * The actual write RPC done to a DS.
6340  */
6341 static int
6342 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6343     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6344     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6345     struct ucred *cred, NFSPROC_T *p)
6346 {
6347 	uint32_t *tl;
6348 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6349 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6350 	int32_t backup;
6351 	struct nfsrv_descript nfsd;
6352 	struct nfsrv_descript *nd = &nfsd;
6353 	struct nfssockreq *nrp;
6354 	struct nfsvattr na;
6355 
6356 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6357 	nd->nd_mrep = NULL;
6358 	if (vers == 0 || vers == NFS_VER4) {
6359 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6360 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6361 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6362 		vers = NFS_VER4;
6363 		if (flex != 0)
6364 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6365 		else
6366 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6367 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6368 	} else {
6369 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6370 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6371 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6372 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6373 	}
6374 	txdr_hyper(io_off, tl);
6375 	tl += 2;
6376 	if (vers == NFS_VER3)
6377 		*tl++ = txdr_unsigned(len);
6378 	*tl++ = txdr_unsigned(*iomode);
6379 	*tl = txdr_unsigned(len);
6380 	nfsm_uiombuf(nd, uiop, len);
6381 	nrp = dsp->nfsclds_sockp;
6382 	if (nrp == NULL)
6383 		/* If NULL, use the MDS socket. */
6384 		nrp = &nmp->nm_sockreq;
6385 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6386 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6387 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6388 	    nd->nd_repstat);
6389 	if (error != 0)
6390 		return (error);
6391 	if (nd->nd_repstat != 0) {
6392 		/*
6393 		 * In case the rpc gets retried, roll
6394 		 * the uio fileds changed by nfsm_uiombuf()
6395 		 * back.
6396 		 */
6397 		uiop->uio_offset -= len;
6398 		uio_uio_resid_add(uiop, len);
6399 		uio_iov_base_add(uiop, -len);
6400 		uio_iov_len_add(uiop, len);
6401 		error = nd->nd_repstat;
6402 	} else {
6403 		if (vers == NFS_VER3) {
6404 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6405 			    NULL);
6406 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6407 			if (error != 0)
6408 				goto nfsmout;
6409 		}
6410 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6411 		rlen = fxdr_unsigned(int, *tl++);
6412 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6413 		if (rlen == 0) {
6414 			error = NFSERR_IO;
6415 			goto nfsmout;
6416 		} else if (rlen < len) {
6417 			backup = len - rlen;
6418 			uio_iov_base_add(uiop, -(backup));
6419 			uio_iov_len_add(uiop, backup);
6420 			uiop->uio_offset -= backup;
6421 			uio_uio_resid_add(uiop, backup);
6422 			len = rlen;
6423 		}
6424 		commit = fxdr_unsigned(int, *tl++);
6425 
6426 		/*
6427 		 * Return the lowest commitment level
6428 		 * obtained by any of the RPCs.
6429 		 */
6430 		if (committed == NFSWRITE_FILESYNC)
6431 			committed = commit;
6432 		else if (committed == NFSWRITE_DATASYNC &&
6433 		    commit == NFSWRITE_UNSTABLE)
6434 			committed = commit;
6435 		if (commit_thru_mds != 0) {
6436 			NFSLOCKMNT(nmp);
6437 			if (!NFSHASWRITEVERF(nmp)) {
6438 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6439 				NFSSETWRITEVERF(nmp);
6440 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6441 				*must_commit = 1;
6442 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6443 			}
6444 			NFSUNLOCKMNT(nmp);
6445 		} else {
6446 			NFSLOCKDS(dsp);
6447 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6448 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6449 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6450 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6451 				*must_commit = 1;
6452 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6453 			}
6454 			NFSUNLOCKDS(dsp);
6455 		}
6456 	}
6457 nfsmout:
6458 	if (nd->nd_mrep != NULL)
6459 		m_freem(nd->nd_mrep);
6460 	*iomode = committed;
6461 	if (nd->nd_repstat != 0 && error == 0)
6462 		error = nd->nd_repstat;
6463 	return (error);
6464 }
6465 
6466 /*
6467  * The actual write RPC done to a DS.
6468  * This variant is called from a separate kernel process for mirrors.
6469  * Any short write is considered an IO error.
6470  */
6471 static int
6472 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6473     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6474     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6475     struct ucred *cred, NFSPROC_T *p)
6476 {
6477 	uint32_t *tl;
6478 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6479 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6480 	struct nfsrv_descript nfsd;
6481 	struct nfsrv_descript *nd = &nfsd;
6482 	struct nfssockreq *nrp;
6483 	struct nfsvattr na;
6484 
6485 	nd->nd_mrep = NULL;
6486 	if (vers == 0 || vers == NFS_VER4) {
6487 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6488 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6489 		vers = NFS_VER4;
6490 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6491 		    minorvers);
6492 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6493 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6494 	} else {
6495 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6496 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6497 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6498 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6499 	}
6500 	txdr_hyper(io_off, tl);
6501 	tl += 2;
6502 	if (vers == NFS_VER3)
6503 		*tl++ = txdr_unsigned(len);
6504 	*tl++ = txdr_unsigned(*iomode);
6505 	*tl = txdr_unsigned(len);
6506 	if (len > 0) {
6507 		/* Put data in mbuf chain. */
6508 		nd->nd_mb->m_next = m;
6509 		/* Set nd_mb and nd_bpos to end of data. */
6510 		while (m->m_next != NULL)
6511 			m = m->m_next;
6512 		nd->nd_mb = m;
6513 		nd->nd_bpos = mtod(m, char *) + m->m_len;
6514 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: lastmb len=%d\n", m->m_len);
6515 	}
6516 	nrp = dsp->nfsclds_sockp;
6517 	if (nrp == NULL)
6518 		/* If NULL, use the MDS socket. */
6519 		nrp = &nmp->nm_sockreq;
6520 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6521 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6522 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6523 	    nd->nd_repstat);
6524 	if (error != 0)
6525 		return (error);
6526 	if (nd->nd_repstat != 0)
6527 		error = nd->nd_repstat;
6528 	else {
6529 		if (vers == NFS_VER3) {
6530 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6531 			    NULL);
6532 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6533 			    error);
6534 			if (error != 0)
6535 				goto nfsmout;
6536 		}
6537 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6538 		rlen = fxdr_unsigned(int, *tl++);
6539 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6540 		    rlen);
6541 		if (rlen != len) {
6542 			error = NFSERR_IO;
6543 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6544 			    len, rlen);
6545 			goto nfsmout;
6546 		}
6547 		commit = fxdr_unsigned(int, *tl++);
6548 
6549 		/*
6550 		 * Return the lowest commitment level
6551 		 * obtained by any of the RPCs.
6552 		 */
6553 		if (committed == NFSWRITE_FILESYNC)
6554 			committed = commit;
6555 		else if (committed == NFSWRITE_DATASYNC &&
6556 		    commit == NFSWRITE_UNSTABLE)
6557 			committed = commit;
6558 		NFSLOCKDS(dsp);
6559 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6560 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6561 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6562 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6563 			*must_commit = 1;
6564 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6565 		}
6566 		NFSUNLOCKDS(dsp);
6567 	}
6568 nfsmout:
6569 	if (nd->nd_mrep != NULL)
6570 		m_freem(nd->nd_mrep);
6571 	*iomode = committed;
6572 	if (nd->nd_repstat != 0 && error == 0)
6573 		error = nd->nd_repstat;
6574 	return (error);
6575 }
6576 
6577 /*
6578  * Start up the thread that will execute nfsrpc_writedsmir().
6579  */
6580 static void
6581 start_writedsmir(void *arg, int pending)
6582 {
6583 	struct nfsclwritedsdorpc *drpc;
6584 
6585 	drpc = (struct nfsclwritedsdorpc *)arg;
6586 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6587 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6588 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6589 	    drpc->p);
6590 	drpc->done = 1;
6591 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6592 }
6593 
6594 /*
6595  * Set up the write DS mirror call for the pNFS I/O thread.
6596  */
6597 static int
6598 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6599     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6600     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6601     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6602 {
6603 	int error, ret;
6604 
6605 	error = 0;
6606 	drpc->done = 0;
6607 	drpc->vp = vp;
6608 	drpc->iomode = *iomode;
6609 	drpc->must_commit = *must_commit;
6610 	drpc->stateidp = stateidp;
6611 	drpc->dsp = dsp;
6612 	drpc->off = off;
6613 	drpc->len = len;
6614 	drpc->fhp = fhp;
6615 	drpc->m = m;
6616 	drpc->vers = vers;
6617 	drpc->minorvers = minorvers;
6618 	drpc->cred = cred;
6619 	drpc->p = p;
6620 	drpc->inprog = 0;
6621 	ret = EIO;
6622 	if (nfs_pnfsiothreads != 0) {
6623 		ret = nfs_pnfsio(start_writedsmir, drpc);
6624 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6625 	}
6626 	if (ret != 0)
6627 		error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6628 		    dsp, off, len, fhp, m, vers, minorvers, cred, p);
6629 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6630 	return (error);
6631 }
6632 
6633 /*
6634  * Free up the nfsclds structure.
6635  */
6636 void
6637 nfscl_freenfsclds(struct nfsclds *dsp)
6638 {
6639 	int i;
6640 
6641 	if (dsp == NULL)
6642 		return;
6643 	if (dsp->nfsclds_sockp != NULL) {
6644 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6645 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6646 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6647 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6648 	}
6649 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
6650 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6651 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6652 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6653 			m_freem(
6654 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6655 	}
6656 	free(dsp, M_NFSCLDS);
6657 }
6658 
6659 static enum nfsclds_state
6660 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6661     struct nfsclds **retdspp, uint32_t *sequencep)
6662 {
6663 	struct nfsclds *dsp;
6664 	int fndseq;
6665 
6666 	/*
6667 	 * Search the list of nfsclds structures for one with the same
6668 	 * server.
6669 	 */
6670 	fndseq = 0;
6671 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6672 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6673 		    dsp->nfsclds_servownlen != 0 &&
6674 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6675 		    dsp->nfsclds_servownlen) &&
6676 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
6677 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6678 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
6679 			    dsp->nfsclds_flags);
6680 			if (fndseq == 0) {
6681 				/* Get sequenceid# from first entry. */
6682 				*sequencep =
6683 				    dsp->nfsclds_sess.nfsess_sequenceid;
6684 				fndseq = 1;
6685 			}
6686 			/* Server major id matches. */
6687 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6688 				*retdspp = dsp;
6689 				return (NFSDSP_USETHISSESSION);
6690 			}
6691 
6692 		}
6693 	}
6694 	if (fndseq != 0)
6695 		return (NFSDSP_SEQTHISSESSION);
6696 	return (NFSDSP_NOTFOUND);
6697 }
6698 
6699 /*
6700  * NFS commit rpc to a NFSv4.1 DS.
6701  */
6702 static int
6703 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6704     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6705     NFSPROC_T *p)
6706 {
6707 	uint32_t *tl;
6708 	struct nfsrv_descript nfsd, *nd = &nfsd;
6709 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6710 	struct nfssockreq *nrp;
6711 	struct nfsvattr na;
6712 	int attrflag, error;
6713 
6714 	nd->nd_mrep = NULL;
6715 	if (vers == 0 || vers == NFS_VER4) {
6716 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6717 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6718 		vers = NFS_VER4;
6719 	} else
6720 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6721 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6722 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6723 	    minorvers);
6724 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6725 	txdr_hyper(offset, tl);
6726 	tl += 2;
6727 	*tl = txdr_unsigned(cnt);
6728 	nrp = dsp->nfsclds_sockp;
6729 	if (nrp == NULL)
6730 		/* If NULL, use the MDS socket. */
6731 		nrp = &nmp->nm_sockreq;
6732 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6733 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6734 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6735 	    nd->nd_repstat);
6736 	if (error != 0)
6737 		return (error);
6738 	if (nd->nd_repstat == 0) {
6739 		if (vers == NFS_VER3) {
6740 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6741 			    NULL);
6742 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6743 			if (error != 0)
6744 				goto nfsmout;
6745 		}
6746 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6747 		NFSLOCKDS(dsp);
6748 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6749 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6750 			error = NFSERR_STALEWRITEVERF;
6751 		}
6752 		NFSUNLOCKDS(dsp);
6753 	}
6754 nfsmout:
6755 	if (error == 0 && nd->nd_repstat != 0)
6756 		error = nd->nd_repstat;
6757 	m_freem(nd->nd_mrep);
6758 	return (error);
6759 }
6760 
6761 /*
6762  * Start up the thread that will execute nfsrpc_commitds().
6763  */
6764 static void
6765 start_commitds(void *arg, int pending)
6766 {
6767 	struct nfsclwritedsdorpc *drpc;
6768 
6769 	drpc = (struct nfsclwritedsdorpc *)arg;
6770 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6771 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6772 	    drpc->p);
6773 	drpc->done = 1;
6774 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6775 }
6776 
6777 /*
6778  * Set up the commit DS mirror call for the pNFS I/O thread.
6779  */
6780 static int
6781 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6782     struct nfsfh *fhp, int vers, int minorvers,
6783     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6784 {
6785 	int error, ret;
6786 
6787 	error = 0;
6788 	drpc->done = 0;
6789 	drpc->vp = vp;
6790 	drpc->off = offset;
6791 	drpc->len = cnt;
6792 	drpc->dsp = dsp;
6793 	drpc->fhp = fhp;
6794 	drpc->vers = vers;
6795 	drpc->minorvers = minorvers;
6796 	drpc->cred = cred;
6797 	drpc->p = p;
6798 	drpc->inprog = 0;
6799 	ret = EIO;
6800 	if (nfs_pnfsiothreads != 0) {
6801 		ret = nfs_pnfsio(start_commitds, drpc);
6802 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6803 	}
6804 	if (ret != 0)
6805 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6806 		    minorvers, cred, p);
6807 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6808 	return (error);
6809 }
6810 
6811 /*
6812  * NFS Advise rpc
6813  */
6814 APPLESTATIC int
6815 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
6816     struct ucred *cred, NFSPROC_T *p)
6817 {
6818 	u_int32_t *tl;
6819 	struct nfsrv_descript nfsd, *nd = &nfsd;
6820 	nfsattrbit_t hints;
6821 	int error;
6822 
6823 	NFSZERO_ATTRBIT(&hints);
6824 	if (advise == POSIX_FADV_WILLNEED)
6825 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6826 	else if (advise == POSIX_FADV_DONTNEED)
6827 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6828 	else
6829 		return (0);
6830 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
6831 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6832 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
6833 	txdr_hyper(offset, tl);
6834 	tl += 2;
6835 	txdr_hyper(cnt, tl);
6836 	nfsrv_putattrbit(nd, &hints);
6837 	error = nfscl_request(nd, vp, p, cred, NULL);
6838 	if (error != 0)
6839 		return (error);
6840 	if (nd->nd_repstat != 0)
6841 		error = nd->nd_repstat;
6842 	m_freem(nd->nd_mrep);
6843 	return (error);
6844 }
6845 
6846 #ifdef notyet
6847 /*
6848  * NFS advise rpc to a NFSv4.2 DS.
6849  */
6850 static int
6851 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6852     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6853     struct ucred *cred, NFSPROC_T *p)
6854 {
6855 	uint32_t *tl;
6856 	struct nfsrv_descript nfsd, *nd = &nfsd;
6857 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6858 	struct nfssockreq *nrp;
6859 	nfsattrbit_t hints;
6860 	int error;
6861 
6862 	/* For NFS DSs prior to NFSv4.2, just return OK. */
6863 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
6864 		return (0);
6865 	NFSZERO_ATTRBIT(&hints);
6866 	if (advise == POSIX_FADV_WILLNEED)
6867 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6868 	else if (advise == POSIX_FADV_DONTNEED)
6869 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6870 	else
6871 		return (0);
6872 	nd->nd_mrep = NULL;
6873 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
6874 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6875 	vers = NFS_VER4;
6876 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
6877 	    minorvers);
6878 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6879 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6880 	txdr_hyper(offset, tl);
6881 	tl += 2;
6882 	*tl = txdr_unsigned(cnt);
6883 	nfsrv_putattrbit(nd, &hints);
6884 	nrp = dsp->nfsclds_sockp;
6885 	if (nrp == NULL)
6886 		/* If NULL, use the MDS socket. */
6887 		nrp = &nmp->nm_sockreq;
6888 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6889 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6890 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
6891 	    nd->nd_repstat);
6892 	if (error != 0)
6893 		return (error);
6894 	if (nd->nd_repstat != 0)
6895 		error = nd->nd_repstat;
6896 	m_freem(nd->nd_mrep);
6897 	return (error);
6898 }
6899 
6900 /*
6901  * Start up the thread that will execute nfsrpc_commitds().
6902  */
6903 static void
6904 start_adviseds(void *arg, int pending)
6905 {
6906 	struct nfsclwritedsdorpc *drpc;
6907 
6908 	drpc = (struct nfsclwritedsdorpc *)arg;
6909 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
6910 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
6911 	    drpc->cred, drpc->p);
6912 	drpc->done = 1;
6913 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
6914 }
6915 
6916 /*
6917  * Set up the commit DS mirror call for the pNFS I/O thread.
6918  */
6919 static int
6920 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6921     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6922     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6923 {
6924 	int error, ret;
6925 
6926 	error = 0;
6927 	drpc->done = 0;
6928 	drpc->vp = vp;
6929 	drpc->off = offset;
6930 	drpc->len = cnt;
6931 	drpc->advise = advise;
6932 	drpc->dsp = dsp;
6933 	drpc->fhp = fhp;
6934 	drpc->vers = vers;
6935 	drpc->minorvers = minorvers;
6936 	drpc->cred = cred;
6937 	drpc->p = p;
6938 	drpc->inprog = 0;
6939 	ret = EIO;
6940 	if (nfs_pnfsiothreads != 0) {
6941 		ret = nfs_pnfsio(start_adviseds, drpc);
6942 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
6943 	}
6944 	if (ret != 0)
6945 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
6946 		    minorvers, cred, p);
6947 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
6948 	return (error);
6949 }
6950 #endif	/* notyet */
6951 
6952 /*
6953  * Do the Allocate operation, retrying for recovery.
6954  */
6955 APPLESTATIC int
6956 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
6957     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
6958 {
6959 	int error, expireret = 0, retrycnt, nostateid;
6960 	uint32_t clidrev = 0;
6961 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6962 	struct nfsfh *nfhp = NULL;
6963 	nfsv4stateid_t stateid;
6964 	off_t tmp_off;
6965 	void *lckp;
6966 
6967 	if (len < 0)
6968 		return (EINVAL);
6969 	if (len == 0)
6970 		return (0);
6971 	tmp_off = off + len;
6972 	NFSLOCKMNT(nmp);
6973 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
6974 		NFSUNLOCKMNT(nmp);
6975 		return (EFBIG);
6976 	}
6977 	if (nmp->nm_clp != NULL)
6978 		clidrev = nmp->nm_clp->nfsc_clientidrev;
6979 	NFSUNLOCKMNT(nmp);
6980 	nfhp = VTONFS(vp)->n_fhp;
6981 	retrycnt = 0;
6982 	do {
6983 		lckp = NULL;
6984 		nostateid = 0;
6985 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
6986 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
6987 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
6988 		    stateid.other[2] == 0) {
6989 			nostateid = 1;
6990 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
6991 		}
6992 
6993 		/*
6994 		 * Not finding a stateid should probably never happen,
6995 		 * but just return an error for this case.
6996 		 */
6997 		if (nostateid != 0)
6998 			error = EIO;
6999 		else
7000 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7001 			    nap, attrflagp, cred, p, stuff);
7002 		if (error == NFSERR_STALESTATEID)
7003 			nfscl_initiate_recovery(nmp->nm_clp);
7004 		if (lckp != NULL)
7005 			nfscl_lockderef(lckp);
7006 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7007 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7008 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7009 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7010 		} else if ((error == NFSERR_EXPIRED ||
7011 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7012 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7013 		}
7014 		retrycnt++;
7015 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7016 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7017 	    error == NFSERR_STALEDONTRECOVER ||
7018 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7019 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7020 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7021 	if (error != 0 && retrycnt >= 4)
7022 		error = EIO;
7023 	return (error);
7024 }
7025 
7026 /*
7027  * The allocate RPC.
7028  */
7029 static int
7030 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7031     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p,
7032     void *stuff)
7033 {
7034 	uint32_t *tl;
7035 	int error;
7036 	struct nfsrv_descript nfsd;
7037 	struct nfsrv_descript *nd = &nfsd;
7038 	nfsattrbit_t attrbits;
7039 
7040 	*attrflagp = 0;
7041 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7042 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7043 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7044 	txdr_hyper(off, tl); tl += 2;
7045 	txdr_hyper(len, tl); tl += 2;
7046 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7047 	NFSGETATTR_ATTRBIT(&attrbits);
7048 	nfsrv_putattrbit(nd, &attrbits);
7049 	error = nfscl_request(nd, vp, p, cred, stuff);
7050 	if (error != 0)
7051 		return (error);
7052 	if (nd->nd_repstat == 0) {
7053 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7054 		error = nfsm_loadattr(nd, nap);
7055 		if (error == 0)
7056 			*attrflagp = NFS_LATTR_NOSHRINK;
7057 	} else
7058 		error = nd->nd_repstat;
7059 nfsmout:
7060 	m_freem(nd->nd_mrep);
7061 	return (error);
7062 }
7063 
7064 /*
7065  * Set up the XDR arguments for the LayoutGet operation.
7066  */
7067 static void
7068 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7069     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7070     int layoutlen, int usecurstateid)
7071 {
7072 	uint32_t *tl;
7073 
7074 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7075 	    NFSX_STATEID);
7076 	*tl++ = newnfs_false;		/* Don't signal availability. */
7077 	*tl++ = txdr_unsigned(layouttype);
7078 	*tl++ = txdr_unsigned(iomode);
7079 	txdr_hyper(offset, tl);
7080 	tl += 2;
7081 	txdr_hyper(len, tl);
7082 	tl += 2;
7083 	txdr_hyper(minlen, tl);
7084 	tl += 2;
7085 	if (usecurstateid != 0) {
7086 		/* Special stateid for Current stateid. */
7087 		*tl++ = txdr_unsigned(1);
7088 		*tl++ = 0;
7089 		*tl++ = 0;
7090 		*tl++ = 0;
7091 	} else {
7092 		*tl++ = txdr_unsigned(stateidp->seqid);
7093 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7094 		*tl++ = stateidp->other[0];
7095 		*tl++ = stateidp->other[1];
7096 		*tl++ = stateidp->other[2];
7097 	}
7098 	*tl = txdr_unsigned(layoutlen);
7099 }
7100 
7101 /*
7102  * Parse the reply for a successful LayoutGet operation.
7103  */
7104 static int
7105 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7106     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7107 {
7108 	uint32_t *tl;
7109 	struct nfsclflayout *flp, *prevflp, *tflp;
7110 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7111 	int m, mirrorcnt;
7112 	uint64_t retlen, off;
7113 	struct nfsfh *nfhp;
7114 	uint8_t *cp;
7115 	uid_t user;
7116 	gid_t grp;
7117 
7118 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7119 	error = 0;
7120 	flp = NULL;
7121 	gotiomode = -1;
7122 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7123 	if (*tl++ != 0)
7124 		*retonclosep = 1;
7125 	else
7126 		*retonclosep = 0;
7127 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7128 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7129 	    (int)stateidp->seqid);
7130 	stateidp->other[0] = *tl++;
7131 	stateidp->other[1] = *tl++;
7132 	stateidp->other[2] = *tl++;
7133 	cnt = fxdr_unsigned(int, *tl);
7134 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7135 	if (cnt <= 0 || cnt > 10000) {
7136 		/* Don't accept more than 10000 layouts in reply. */
7137 		error = NFSERR_BADXDR;
7138 		goto nfsmout;
7139 	}
7140 	for (i = 0; i < cnt; i++) {
7141 		/* Dissect to the layout type. */
7142 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7143 		    3 * NFSX_UNSIGNED);
7144 		off = fxdr_hyper(tl); tl += 2;
7145 		retlen = fxdr_hyper(tl); tl += 2;
7146 		iomode = fxdr_unsigned(int, *tl++);
7147 		laytype = fxdr_unsigned(int, *tl);
7148 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7149 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7150 		/* Ignore length of layout body for now. */
7151 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7152 			/* Parse the File layout up to fhcnt. */
7153 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7154 			    NFSX_HYPER + NFSX_V4DEVICEID);
7155 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7156 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7157 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7158 			if (fhcnt < 0 || fhcnt > 100) {
7159 				/* Don't accept more than 100 file handles. */
7160 				error = NFSERR_BADXDR;
7161 				goto nfsmout;
7162 			}
7163 			if (fhcnt > 0)
7164 				flp = malloc(sizeof(*flp) + fhcnt *
7165 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7166 				    M_WAITOK);
7167 			else
7168 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7169 				    M_WAITOK);
7170 			flp->nfsfl_flags = NFSFL_FILE;
7171 			flp->nfsfl_fhcnt = 0;
7172 			flp->nfsfl_devp = NULL;
7173 			flp->nfsfl_off = off;
7174 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7175 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7176 			else
7177 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7178 			flp->nfsfl_iomode = iomode;
7179 			if (gotiomode == -1)
7180 				gotiomode = flp->nfsfl_iomode;
7181 			/* Ignore layout body length for now. */
7182 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7183 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7184 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7185 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7186 			mtx_lock(&nmp->nm_mtx);
7187 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7188 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7189 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7190 			mtx_unlock(&nmp->nm_mtx);
7191 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7192 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7193 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7194 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7195 			for (j = 0; j < fhcnt; j++) {
7196 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7197 				nfhlen = fxdr_unsigned(int, *tl);
7198 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7199 					error = NFSERR_BADXDR;
7200 					goto nfsmout;
7201 				}
7202 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7203 				    M_NFSFH, M_WAITOK);
7204 				flp->nfsfl_fh[j] = nfhp;
7205 				flp->nfsfl_fhcnt++;
7206 				nfhp->nfh_len = nfhlen;
7207 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7208 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7209 			}
7210 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7211 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7212 			    NFSX_HYPER);
7213 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7214 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7215 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7216 				error = NFSERR_BADXDR;
7217 				goto nfsmout;
7218 			}
7219 			flp = malloc(sizeof(*flp) + mirrorcnt *
7220 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7221 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7222 			flp->nfsfl_mirrorcnt = mirrorcnt;
7223 			for (j = 0; j < mirrorcnt; j++)
7224 				flp->nfsfl_ffm[j].devp = NULL;
7225 			flp->nfsfl_off = off;
7226 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7227 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7228 			else
7229 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7230 			flp->nfsfl_iomode = iomode;
7231 			if (gotiomode == -1)
7232 				gotiomode = flp->nfsfl_iomode;
7233 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7234 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7235 			    (uintmax_t)flp->nfsfl_stripeunit);
7236 			for (j = 0; j < mirrorcnt; j++) {
7237 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7238 				k = fxdr_unsigned(int, *tl);
7239 				if (k < 1 || k > 128) {
7240 					error = NFSERR_BADXDR;
7241 					goto nfsmout;
7242 				}
7243 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7244 				for (l = 0; l < k; l++) {
7245 					NFSM_DISSECT(tl, uint32_t *,
7246 					    NFSX_V4DEVICEID + NFSX_STATEID +
7247 					    2 * NFSX_UNSIGNED);
7248 					if (l == 0) {
7249 						/* Just use the first server. */
7250 						NFSBCOPY(tl,
7251 						    flp->nfsfl_ffm[j].dev,
7252 						    NFSX_V4DEVICEID);
7253 						tl += (NFSX_V4DEVICEID /
7254 						    NFSX_UNSIGNED);
7255 						tl++;
7256 						flp->nfsfl_ffm[j].st.seqid =
7257 						    *tl++;
7258 						flp->nfsfl_ffm[j].st.other[0] =
7259 						    *tl++;
7260 						flp->nfsfl_ffm[j].st.other[1] =
7261 						    *tl++;
7262 						flp->nfsfl_ffm[j].st.other[2] =
7263 						    *tl++;
7264 						NFSCL_DEBUG(4, "st.seqid=%u "
7265 						 "st.o0=0x%x st.o1=0x%x "
7266 						 "st.o2=0x%x\n",
7267 						 flp->nfsfl_ffm[j].st.seqid,
7268 						 flp->nfsfl_ffm[j].st.other[0],
7269 						 flp->nfsfl_ffm[j].st.other[1],
7270 						 flp->nfsfl_ffm[j].st.other[2]);
7271 					} else
7272 						tl += ((NFSX_V4DEVICEID +
7273 						    NFSX_STATEID +
7274 						    NFSX_UNSIGNED) /
7275 						    NFSX_UNSIGNED);
7276 					fhcnt = fxdr_unsigned(int, *tl);
7277 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7278 					if (fhcnt < 1 ||
7279 					    fhcnt > NFSDEV_MAXVERS) {
7280 						error = NFSERR_BADXDR;
7281 						goto nfsmout;
7282 					}
7283 					for (m = 0; m < fhcnt; m++) {
7284 						NFSM_DISSECT(tl, uint32_t *,
7285 						    NFSX_UNSIGNED);
7286 						nfhlen = fxdr_unsigned(int,
7287 						    *tl);
7288 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7289 						    nfhlen);
7290 						if (nfhlen <= 0 || nfhlen >
7291 						    NFSX_V4FHMAX) {
7292 							error = NFSERR_BADXDR;
7293 							goto nfsmout;
7294 						}
7295 						NFSM_DISSECT(cp, uint8_t *,
7296 						    NFSM_RNDUP(nfhlen));
7297 						if (l == 0) {
7298 							flp->nfsfl_ffm[j].fhcnt
7299 							    = fhcnt;
7300 							nfhp = malloc(
7301 							    sizeof(*nfhp) +
7302 							    nfhlen - 1, M_NFSFH,
7303 							    M_WAITOK);
7304 							flp->nfsfl_ffm[j].fh[m]
7305 							    = nfhp;
7306 							nfhp->nfh_len = nfhlen;
7307 							NFSBCOPY(cp,
7308 							    nfhp->nfh_fh,
7309 							    nfhlen);
7310 							NFSCL_DEBUG(4,
7311 							    "got fh\n");
7312 						}
7313 					}
7314 					/* Now, get the ffsd_user/ffds_group. */
7315 					error = nfsrv_parseug(nd, 0, &user,
7316 					    &grp, curthread);
7317 					NFSCL_DEBUG(4, "after parseu=%d\n",
7318 					    error);
7319 					if (error == 0)
7320 						error = nfsrv_parseug(nd, 1,
7321 						    &user, &grp, curthread);
7322 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7323 					    grp);
7324 					if (error != 0)
7325 						goto nfsmout;
7326 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7327 					    user, grp);
7328 					if (l == 0) {
7329 						flp->nfsfl_ffm[j].user = user;
7330 						flp->nfsfl_ffm[j].group = grp;
7331 						NFSCL_DEBUG(4,
7332 						    "usr=%d grp=%d\n", user,
7333 						    grp);
7334 					}
7335 				}
7336 			}
7337 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7338 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7339 #ifdef notnow
7340 			/*
7341 			 * At this time, there is no flag.
7342 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7343 			 * added, or it may never exist?
7344 			 */
7345 			mtx_lock(&nmp->nm_mtx);
7346 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7347 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7348 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7349 			mtx_unlock(&nmp->nm_mtx);
7350 #endif
7351 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7352 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7353 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7354 		} else {
7355 			error = NFSERR_BADXDR;
7356 			goto nfsmout;
7357 		}
7358 		if (flp->nfsfl_iomode == gotiomode) {
7359 			/* Keep the list in increasing offset order. */
7360 			tflp = LIST_FIRST(flhp);
7361 			prevflp = NULL;
7362 			while (tflp != NULL &&
7363 			    tflp->nfsfl_off < flp->nfsfl_off) {
7364 				prevflp = tflp;
7365 				tflp = LIST_NEXT(tflp, nfsfl_list);
7366 			}
7367 			if (prevflp == NULL)
7368 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7369 			else
7370 				LIST_INSERT_AFTER(prevflp, flp,
7371 				    nfsfl_list);
7372 			NFSCL_DEBUG(4, "flp inserted\n");
7373 		} else {
7374 			printf("nfscl_layoutget(): got wrong iomode\n");
7375 			nfscl_freeflayout(flp);
7376 		}
7377 		flp = NULL;
7378 	}
7379 nfsmout:
7380 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7381 	if (error != 0 && flp != NULL)
7382 		nfscl_freeflayout(flp);
7383 	return (error);
7384 }
7385 
7386 /*
7387  * Parse a user/group digit string.
7388  */
7389 static int
7390 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7391     NFSPROC_T *p)
7392 {
7393 	uint32_t *tl;
7394 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7395 	uint32_t len = 0;
7396 	int error = 0;
7397 
7398 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7399 	len = fxdr_unsigned(uint32_t, *tl);
7400 	str = NULL;
7401 	if (len > NFSV4_OPAQUELIMIT) {
7402 		error = NFSERR_BADXDR;
7403 		goto nfsmout;
7404 	}
7405 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7406 	if (len == 0) {
7407 		if (dogrp != 0)
7408 			*gidp = GID_NOGROUP;
7409 		else
7410 			*uidp = UID_NOBODY;
7411 		return (0);
7412 	}
7413 	if (len > NFSV4_SMALLSTR)
7414 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7415 	else
7416 		str = str0;
7417 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7418 	NFSBCOPY(cp, str, len);
7419 	str[len] = '\0';
7420 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7421 	if (dogrp != 0)
7422 		error = nfsv4_strtogid(nd, str, len, gidp);
7423 	else
7424 		error = nfsv4_strtouid(nd, str, len, uidp);
7425 nfsmout:
7426 	if (len > NFSV4_SMALLSTR)
7427 		free(str, M_TEMP);
7428 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7429 	return (error);
7430 }
7431 
7432 /*
7433  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7434  * so that it does both an Open and a Layoutget.
7435  */
7436 static int
7437 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7438     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7439     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7440     struct ucred *cred, NFSPROC_T *p)
7441 {
7442 	struct nfscllayout *lyp;
7443 	struct nfsclflayout *flp;
7444 	struct nfsclflayouthead flh;
7445 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7446 	int layouttype, laystat;
7447 	nfsv4stateid_t stateid;
7448 	struct nfsclsession *tsep;
7449 
7450 	error = 0;
7451 	if (NFSHASFLEXFILE(nmp))
7452 		layouttype = NFSLAYOUT_FLEXFILE;
7453 	else
7454 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7455 	/*
7456 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7457 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7458 	 * flp == NULL.
7459 	 */
7460 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7461 	    &recalled);
7462 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7463 	if (lyp == NULL)
7464 		islocked = 0;
7465 	else if (flp != NULL)
7466 		islocked = 1;
7467 	else
7468 		islocked = 2;
7469 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7470 		LIST_INIT(&flh);
7471 		tsep = nfsmnt_mdssession(nmp);
7472 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7473 		    3 * NFSX_UNSIGNED);
7474 		if (lyp == NULL)
7475 			usecurstateid = 1;
7476 		else {
7477 			usecurstateid = 0;
7478 			stateid.seqid = lyp->nfsly_stateid.seqid;
7479 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7480 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7481 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7482 		}
7483 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7484 		    newfhp, newfhlen, mode, op, name, namelen,
7485 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7486 		    &retonclose, &flh, &laystat, cred, p);
7487 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7488 		    laystat, error);
7489 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7490 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7491 		    &islocked, cred, p);
7492 	} else
7493 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7494 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7495 	if (islocked == 2)
7496 		nfscl_rellayout(lyp, 1);
7497 	else if (islocked == 1)
7498 		nfscl_rellayout(lyp, 0);
7499 	return (error);
7500 }
7501 
7502 /*
7503  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7504  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7505  * handled by nfsrpc_openrpc().
7506  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7507  * can be NULL.
7508  */
7509 static int
7510 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7511     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7512     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7513     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7514     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7515     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7516 {
7517 	uint32_t *tl;
7518 	struct nfsrv_descript nfsd, *nd = &nfsd;
7519 	struct nfscldeleg *ndp = NULL;
7520 	struct nfsvattr nfsva;
7521 	struct nfsclsession *tsep;
7522 	uint32_t rflags, deleg;
7523 	nfsattrbit_t attrbits;
7524 	int error, ret, acesize, limitby, iomode;
7525 
7526 	*dpp = NULL;
7527 	*laystatp = ENXIO;
7528 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7529 	    0, 0);
7530 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7531 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7532 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7533 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7534 	tsep = nfsmnt_mdssession(nmp);
7535 	*tl++ = tsep->nfsess_clientid.lval[0];
7536 	*tl = tsep->nfsess_clientid.lval[1];
7537 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7538 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7539 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7540 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7541 	nfsm_strtom(nd, name, namelen);
7542 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7543 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7544 	NFSZERO_ATTRBIT(&attrbits);
7545 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7546 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7547 	nfsrv_putattrbit(nd, &attrbits);
7548 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7549 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7550 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7551 		iomode = NFSLAYOUTIOMODE_RW;
7552 	else
7553 		iomode = NFSLAYOUTIOMODE_READ;
7554 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7555 	    layouttype, layoutlen, usecurstateid);
7556 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7557 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7558 	if (error != 0)
7559 		return (error);
7560 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7561 	if (nd->nd_repstat != 0)
7562 		*laystatp = nd->nd_repstat;
7563 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7564 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7565 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7566 		    6 * NFSX_UNSIGNED);
7567 		op->nfso_stateid.seqid = *tl++;
7568 		op->nfso_stateid.other[0] = *tl++;
7569 		op->nfso_stateid.other[1] = *tl++;
7570 		op->nfso_stateid.other[2] = *tl;
7571 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7572 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7573 		if (error != 0)
7574 			goto nfsmout;
7575 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7576 		deleg = fxdr_unsigned(u_int32_t, *tl);
7577 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7578 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7579 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7580 			      NFSCLFLAGS_FIRSTDELEG))
7581 				op->nfso_own->nfsow_clp->nfsc_flags |=
7582 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7583 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7584 			    M_NFSCLDELEG, M_WAITOK);
7585 			LIST_INIT(&ndp->nfsdl_owner);
7586 			LIST_INIT(&ndp->nfsdl_lock);
7587 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7588 			ndp->nfsdl_fhlen = newfhlen;
7589 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7590 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
7591 			nfscl_lockinit(&ndp->nfsdl_rwlock);
7592 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7593 			    NFSX_UNSIGNED);
7594 			ndp->nfsdl_stateid.seqid = *tl++;
7595 			ndp->nfsdl_stateid.other[0] = *tl++;
7596 			ndp->nfsdl_stateid.other[1] = *tl++;
7597 			ndp->nfsdl_stateid.other[2] = *tl++;
7598 			ret = fxdr_unsigned(int, *tl);
7599 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7600 				ndp->nfsdl_flags = NFSCLDL_WRITE;
7601 				/*
7602 				 * Indicates how much the file can grow.
7603 				 */
7604 				NFSM_DISSECT(tl, u_int32_t *,
7605 				    3 * NFSX_UNSIGNED);
7606 				limitby = fxdr_unsigned(int, *tl++);
7607 				switch (limitby) {
7608 				case NFSV4OPEN_LIMITSIZE:
7609 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7610 					break;
7611 				case NFSV4OPEN_LIMITBLOCKS:
7612 					ndp->nfsdl_sizelimit =
7613 					    fxdr_unsigned(u_int64_t, *tl++);
7614 					ndp->nfsdl_sizelimit *=
7615 					    fxdr_unsigned(u_int64_t, *tl);
7616 					break;
7617 				default:
7618 					error = NFSERR_BADXDR;
7619 					goto nfsmout;
7620 				};
7621 			} else
7622 				ndp->nfsdl_flags = NFSCLDL_READ;
7623 			if (ret != 0)
7624 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
7625 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7626 			    &acesize, p);
7627 			if (error != 0)
7628 				goto nfsmout;
7629 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7630 			error = NFSERR_BADXDR;
7631 			goto nfsmout;
7632 		}
7633 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7634 		    nfscl_assumeposixlocks)
7635 			op->nfso_posixlock = 1;
7636 		else
7637 			op->nfso_posixlock = 0;
7638 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7639 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
7640 		if (*++tl == 0) {
7641 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7642 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7643 			    NULL, NULL, NULL, p, cred);
7644 			if (error != 0)
7645 				goto nfsmout;
7646 			if (ndp != NULL) {
7647 				ndp->nfsdl_change = nfsva.na_filerev;
7648 				ndp->nfsdl_modtime = nfsva.na_mtime;
7649 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7650 				*dpp = ndp;
7651 				ndp = NULL;
7652 			}
7653 			/*
7654 			 * At this point, the Open has succeeded, so set
7655 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
7656 			 * this function just won't return a layout.
7657 			 */
7658 			if (nd->nd_repstat == 0) {
7659 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7660 				*laystatp = fxdr_unsigned(int, *++tl);
7661 				if (*laystatp == 0) {
7662 					error = nfsrv_parselayoutget(nmp, nd,
7663 					    stateidp, retonclosep, flhp);
7664 					if (error != 0)
7665 						*laystatp = error;
7666 				}
7667 			} else
7668 				nd->nd_repstat = 0;	/* Return 0 for Open. */
7669 		}
7670 	}
7671 	if (nd->nd_repstat != 0 && error == 0)
7672 		error = nd->nd_repstat;
7673 nfsmout:
7674 	free(ndp, M_NFSCLDELEG);
7675 	m_freem(nd->nd_mrep);
7676 	return (error);
7677 }
7678 
7679 /*
7680  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7681  * Used only for mounts with pNFS enabled.
7682  */
7683 static int
7684 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7685     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7686     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7687     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7688     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7689     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7690     struct nfsclflayouthead *flhp, int *laystatp)
7691 {
7692 	uint32_t *tl;
7693 	int error = 0, deleg, newone, ret, acesize, limitby;
7694 	struct nfsrv_descript nfsd, *nd = &nfsd;
7695 	struct nfsclopen *op;
7696 	struct nfscldeleg *dp = NULL;
7697 	struct nfsnode *np;
7698 	struct nfsfh *nfhp;
7699 	struct nfsclsession *tsep;
7700 	nfsattrbit_t attrbits;
7701 	nfsv4stateid_t stateid;
7702 	struct nfsmount *nmp;
7703 
7704 	nmp = VFSTONFS(dvp->v_mount);
7705 	np = VTONFS(dvp);
7706 	*laystatp = ENXIO;
7707 	*unlockedp = 0;
7708 	*nfhpp = NULL;
7709 	*dpp = NULL;
7710 	*attrflagp = 0;
7711 	*dattrflagp = 0;
7712 	if (namelen > NFS_MAXNAMLEN)
7713 		return (ENAMETOOLONG);
7714 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7715 	/*
7716 	 * For V4, this is actually an Open op.
7717 	 */
7718 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7719 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
7720 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7721 	    NFSV4OPEN_ACCESSREAD);
7722 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7723 	tsep = nfsmnt_mdssession(nmp);
7724 	*tl++ = tsep->nfsess_clientid.lval[0];
7725 	*tl = tsep->nfsess_clientid.lval[1];
7726 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7727 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7728 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7729 	if ((fmode & O_EXCL) != 0) {
7730 		if (NFSHASSESSPERSIST(nmp)) {
7731 			/* Use GUARDED for persistent sessions. */
7732 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
7733 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7734 		} else {
7735 			/* Otherwise, use EXCLUSIVE4_1. */
7736 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7737 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7738 			*tl++ = cverf.lval[0];
7739 			*tl = cverf.lval[1];
7740 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7741 		}
7742 	} else {
7743 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7744 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
7745 	}
7746 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7747 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7748 	nfsm_strtom(nd, name, namelen);
7749 	/* Get the new file's handle and attributes, plus save the FH. */
7750 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7751 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7752 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
7753 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7754 	NFSGETATTR_ATTRBIT(&attrbits);
7755 	nfsrv_putattrbit(nd, &attrbits);
7756 	/* Get the directory's post-op attributes. */
7757 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7758 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
7759 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7760 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7761 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7762 	nfsrv_putattrbit(nd, &attrbits);
7763 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7764 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7765 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7766 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7767 	    layouttype, layoutlen, usecurstateid);
7768 	error = nfscl_request(nd, dvp, p, cred, dstuff);
7769 	if (error != 0)
7770 		return (error);
7771 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7772 	    error);
7773 	if (nd->nd_repstat != 0)
7774 		*laystatp = nd->nd_repstat;
7775 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7776 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7777 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7778 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7779 		    6 * NFSX_UNSIGNED);
7780 		stateid.seqid = *tl++;
7781 		stateid.other[0] = *tl++;
7782 		stateid.other[1] = *tl++;
7783 		stateid.other[2] = *tl;
7784 		nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7785 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7786 		deleg = fxdr_unsigned(int, *tl);
7787 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7788 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7789 			if (!(owp->nfsow_clp->nfsc_flags &
7790 			      NFSCLFLAGS_FIRSTDELEG))
7791 				owp->nfsow_clp->nfsc_flags |=
7792 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7793 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7794 			    M_NFSCLDELEG, M_WAITOK);
7795 			LIST_INIT(&dp->nfsdl_owner);
7796 			LIST_INIT(&dp->nfsdl_lock);
7797 			dp->nfsdl_clp = owp->nfsow_clp;
7798 			newnfs_copyincred(cred, &dp->nfsdl_cred);
7799 			nfscl_lockinit(&dp->nfsdl_rwlock);
7800 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7801 			    NFSX_UNSIGNED);
7802 			dp->nfsdl_stateid.seqid = *tl++;
7803 			dp->nfsdl_stateid.other[0] = *tl++;
7804 			dp->nfsdl_stateid.other[1] = *tl++;
7805 			dp->nfsdl_stateid.other[2] = *tl++;
7806 			ret = fxdr_unsigned(int, *tl);
7807 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7808 				dp->nfsdl_flags = NFSCLDL_WRITE;
7809 				/*
7810 				 * Indicates how much the file can grow.
7811 				 */
7812 				NFSM_DISSECT(tl, u_int32_t *,
7813 				    3 * NFSX_UNSIGNED);
7814 				limitby = fxdr_unsigned(int, *tl++);
7815 				switch (limitby) {
7816 				case NFSV4OPEN_LIMITSIZE:
7817 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
7818 					break;
7819 				case NFSV4OPEN_LIMITBLOCKS:
7820 					dp->nfsdl_sizelimit =
7821 					    fxdr_unsigned(u_int64_t, *tl++);
7822 					dp->nfsdl_sizelimit *=
7823 					    fxdr_unsigned(u_int64_t, *tl);
7824 					break;
7825 				default:
7826 					error = NFSERR_BADXDR;
7827 					goto nfsmout;
7828 				};
7829 			} else {
7830 				dp->nfsdl_flags = NFSCLDL_READ;
7831 			}
7832 			if (ret != 0)
7833 				dp->nfsdl_flags |= NFSCLDL_RECALL;
7834 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7835 			    &acesize, p);
7836 			if (error != 0)
7837 				goto nfsmout;
7838 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7839 			error = NFSERR_BADXDR;
7840 			goto nfsmout;
7841 		}
7842 
7843 		/* Now, we should have the status for the SaveFH. */
7844 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7845 		if (*++tl == 0) {
7846 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7847 			/*
7848 			 * Now, process the GetFH and Getattr for the newly
7849 			 * created file. nfscl_mtofh() will set
7850 			 * ND_NOMOREDATA if these weren't successful.
7851 			 */
7852 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7853 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7854 			if (error != 0)
7855 				goto nfsmout;
7856 		} else
7857 			nd->nd_flag |= ND_NOMOREDATA;
7858 		/* Now we have the PutFH and Getattr for the directory. */
7859 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7860 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7861 			if (*++tl != 0)
7862 				nd->nd_flag |= ND_NOMOREDATA;
7863 			else {
7864 				NFSM_DISSECT(tl, uint32_t *, 2 *
7865 				    NFSX_UNSIGNED);
7866 				if (*++tl != 0)
7867 					nd->nd_flag |= ND_NOMOREDATA;
7868 			}
7869 		}
7870 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7871 			/* Load the directory attributes. */
7872 			error = nfsm_loadattr(nd, dnap);
7873 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7874 			if (error != 0)
7875 				goto nfsmout;
7876 			*dattrflagp = 1;
7877 			if (dp != NULL && *attrflagp != 0) {
7878 				dp->nfsdl_change = nnap->na_filerev;
7879 				dp->nfsdl_modtime = nnap->na_mtime;
7880 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7881 			}
7882 			/*
7883 			 * We can now complete the Open state.
7884 			 */
7885 			nfhp = *nfhpp;
7886 			if (dp != NULL) {
7887 				dp->nfsdl_fhlen = nfhp->nfh_len;
7888 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7889 				    nfhp->nfh_len);
7890 			}
7891 			/*
7892 			 * Get an Open structure that will be
7893 			 * attached to the OpenOwner, acquired already.
7894 			 */
7895 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7896 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7897 			    cred, p, NULL, &op, &newone, NULL, 0);
7898 			if (error != 0)
7899 				goto nfsmout;
7900 			op->nfso_stateid = stateid;
7901 			newnfs_copyincred(cred, &op->nfso_cred);
7902 
7903 			nfscl_openrelease(nmp, op, error, newone);
7904 			*unlockedp = 1;
7905 
7906 			/* Now, handle the RestoreFH and LayoutGet. */
7907 			if (nd->nd_repstat == 0) {
7908 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7909 				*laystatp = fxdr_unsigned(int, *(tl + 3));
7910 				if (*laystatp == 0) {
7911 					error = nfsrv_parselayoutget(nmp, nd,
7912 					    stateidp, retonclosep, flhp);
7913 					if (error != 0)
7914 						*laystatp = error;
7915 				}
7916 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7917 				    error);
7918 			} else
7919 				nd->nd_repstat = 0;
7920 		}
7921 	}
7922 	if (nd->nd_repstat != 0 && error == 0)
7923 		error = nd->nd_repstat;
7924 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7925 		nfscl_initiate_recovery(owp->nfsow_clp);
7926 nfsmout:
7927 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7928 	if (error == 0)
7929 		*dpp = dp;
7930 	else
7931 		free(dp, M_NFSCLDELEG);
7932 	m_freem(nd->nd_mrep);
7933 	return (error);
7934 }
7935 
7936 /*
7937  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7938  */
7939 static int
7940 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7941     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7942     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7943     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7944     int *dattrflagp, void *dstuff, int *unlockedp)
7945 {
7946 	struct nfscllayout *lyp;
7947 	struct nfsclflayouthead flh;
7948 	struct nfsfh *nfhp;
7949 	struct nfsclsession *tsep;
7950 	struct nfsmount *nmp;
7951 	nfsv4stateid_t stateid;
7952 	int error, layoutlen, layouttype, retonclose, laystat;
7953 
7954 	error = 0;
7955 	nmp = VFSTONFS(dvp->v_mount);
7956 	if (NFSHASFLEXFILE(nmp))
7957 		layouttype = NFSLAYOUT_FLEXFILE;
7958 	else
7959 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7960 	LIST_INIT(&flh);
7961 	tsep = nfsmnt_mdssession(nmp);
7962 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7963 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7964 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7965 	    dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7966 	    &flh, &laystat);
7967 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7968 	    laystat, error);
7969 	lyp = NULL;
7970 	if (laystat == 0) {
7971 		nfhp = *nfhpp;
7972 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
7973 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
7974 		    layouttype, laystat, NULL, cred, p);
7975 	} else
7976 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
7977 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
7978 		    cred, p);
7979 	if (laystat == 0)
7980 		nfscl_rellayout(lyp, 0);
7981 	return (error);
7982 }
7983 
7984 /*
7985  * Process the results of a layoutget() operation.
7986  */
7987 static int
7988 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
7989     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
7990     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
7991     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
7992 {
7993 	struct nfsclflayout *tflp;
7994 	struct nfscldevinfo *dip;
7995 	uint8_t *dev;
7996 	int i, mirrorcnt;
7997 
7998 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
7999 		NFSLOCKMNT(nmp);
8000 		if (!NFSHASFLEXFILE(nmp)) {
8001 			/* Switch to using Flex File Layout. */
8002 			nmp->nm_state |= NFSSTA_FLEXFILE;
8003 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8004 			/* Disable pNFS. */
8005 			NFSCL_DEBUG(1, "disable PNFS\n");
8006 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8007 		}
8008 		NFSUNLOCKMNT(nmp);
8009 	}
8010 	if (laystat == 0) {
8011 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8012 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8013 			if (layouttype == NFSLAYOUT_FLEXFILE)
8014 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8015 			else
8016 				mirrorcnt = 1;
8017 			for (i = 0; i < mirrorcnt; i++) {
8018 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8019 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8020 				if (laystat != 0) {
8021 					if (layouttype == NFSLAYOUT_FLEXFILE)
8022 						dev = tflp->nfsfl_ffm[i].dev;
8023 					else
8024 						dev = tflp->nfsfl_dev;
8025 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8026 					    layouttype, notifybit, &dip, cred,
8027 					    p);
8028 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8029 					    laystat);
8030 					if (laystat != 0)
8031 						goto out;
8032 					laystat = nfscl_adddevinfo(nmp, dip, i,
8033 					    tflp);
8034 					if (laystat != 0)
8035 						printf("nfsrpc_layoutgetresout"
8036 						    ": cannot add\n");
8037 				}
8038 			}
8039 		}
8040 	}
8041 out:
8042 	if (laystat == 0) {
8043 		/*
8044 		 * nfscl_layout() always returns with the nfsly_lock
8045 		 * set to a refcnt (shared lock).
8046 		 * Passing in dvp is sufficient, since it is only used to
8047 		 * get the fsid for the file system.
8048 		 */
8049 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8050 		    layouttype, retonclose, flhp, lypp, cred, p);
8051 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8052 		    laystat);
8053 		if (laystat == 0 && islockedp != NULL)
8054 			*islockedp = 1;
8055 	}
8056 	return (laystat);
8057 }
8058 
8059 /*
8060  * nfs copy_file_range operation.
8061  */
8062 APPLESTATIC int
8063 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8064     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8065     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8066     struct ucred *cred, bool consecutive, bool *must_commitp)
8067 {
8068 	int commit, error, expireret = 0, retrycnt;
8069 	u_int32_t clidrev = 0;
8070 	struct nfsmount *nmp = VFSTONFS(vnode_mount(invp));
8071 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8072 	nfsv4stateid_t instateid, outstateid;
8073 	void *inlckp, *outlckp;
8074 
8075 	if (nmp->nm_clp != NULL)
8076 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8077 	innfhp = VTONFS(invp)->n_fhp;
8078 	outnfhp = VTONFS(outvp)->n_fhp;
8079 	retrycnt = 0;
8080 	do {
8081 		/* Get both stateids. */
8082 		inlckp = NULL;
8083 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8084 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8085 		    &inlckp);
8086 		outlckp = NULL;
8087 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8088 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8089 		    &outlckp);
8090 
8091 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8092 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8093 		    outattrflagp, consecutive, &commit, cred, curthread);
8094 		if (error == 0) {
8095 			if (commit != NFSWRITE_FILESYNC)
8096 				*must_commitp = true;
8097 			*inoffp += *lenp;
8098 			*outoffp += *lenp;
8099 		} else if (error == NFSERR_STALESTATEID)
8100 			nfscl_initiate_recovery(nmp->nm_clp);
8101 		if (inlckp != NULL)
8102 			nfscl_lockderef(inlckp);
8103 		if (outlckp != NULL)
8104 			nfscl_lockderef(outlckp);
8105 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8106 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8107 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8108 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8109 		} else if ((error == NFSERR_EXPIRED ||
8110 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8111 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8112 			    curthread);
8113 		}
8114 		retrycnt++;
8115 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8116 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8117 	      error == NFSERR_STALEDONTRECOVER ||
8118 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8119 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8120 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8121 	if (error != 0 && (retrycnt >= 4 ||
8122 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8123 	      error == NFSERR_STALEDONTRECOVER))
8124 		error = EIO;
8125 	return (error);
8126 }
8127 
8128 /*
8129  * The copy RPC.
8130  */
8131 static int
8132 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8133     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8134     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8135     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8136     NFSPROC_T *p)
8137 {
8138 	uint32_t *tl;
8139 	int error;
8140 	struct nfsrv_descript nfsd;
8141 	struct nfsrv_descript *nd = &nfsd;
8142 	struct nfsmount *nmp;
8143 	nfsattrbit_t attrbits;
8144 	uint64_t len;
8145 
8146 	nmp = VFSTONFS(outvp->v_mount);
8147 	*inattrflagp = *outattrflagp = 0;
8148 	*commitp = NFSWRITE_UNSTABLE;
8149 	len = *lenp;
8150 	*lenp = 0;
8151 	if (len > nfs_maxcopyrange)
8152 		len = nfs_maxcopyrange;
8153 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8154 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8155 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8156 	NFSGETATTR_ATTRBIT(&attrbits);
8157 	nfsrv_putattrbit(nd, &attrbits);
8158 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8159 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8160 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8161 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8162 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8163 	*tl = txdr_unsigned(NFSV4OP_COPY);
8164 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8165 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8166 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8167 	txdr_hyper(inoff, tl); tl += 2;
8168 	txdr_hyper(outoff, tl); tl += 2;
8169 	txdr_hyper(len, tl); tl += 2;
8170 	if (consecutive)
8171 		*tl++ = newnfs_true;
8172 	else
8173 		*tl++ = newnfs_false;
8174 	*tl++ = newnfs_true;
8175 	*tl++ = 0;
8176 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8177 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8178 	nfsrv_putattrbit(nd, &attrbits);
8179 	error = nfscl_request(nd, invp, p, cred, NULL);
8180 	if (error != 0)
8181 		return (error);
8182 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8183 		/* Get the input file's attributes. */
8184 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8185 		if (*(tl + 1) == 0) {
8186 			error = nfsm_loadattr(nd, innap);
8187 			if (error != 0)
8188 				goto nfsmout;
8189 			*inattrflagp = 1;
8190 		} else
8191 			nd->nd_flag |= ND_NOMOREDATA;
8192 	}
8193 	/* Skip over return stat for PutFH. */
8194 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8195 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8196 		if (*++tl != 0)
8197 			nd->nd_flag |= ND_NOMOREDATA;
8198 	}
8199 	/* Skip over return stat for Copy. */
8200 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8201 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8202 	if (nd->nd_repstat == 0) {
8203 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8204 		if (*tl != 0) {
8205 			/* There should be no callback ids. */
8206 			error = NFSERR_BADXDR;
8207 			goto nfsmout;
8208 		}
8209 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8210 		    NFSX_VERF);
8211 		len = fxdr_hyper(tl); tl += 2;
8212 		*commitp = fxdr_unsigned(int, *tl++);
8213 		NFSLOCKMNT(nmp);
8214 		if (!NFSHASWRITEVERF(nmp)) {
8215 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8216 			NFSSETWRITEVERF(nmp);
8217 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8218 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8219 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8220 		}
8221 		NFSUNLOCKMNT(nmp);
8222 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8223 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8224 			/* Must be a synchronous copy. */
8225 			nd->nd_repstat = NFSERR_NOTSUPP;
8226 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8227 		error = nfsm_loadattr(nd, outnap);
8228 		if (error == 0)
8229 			*outattrflagp = NFS_LATTR_NOSHRINK;
8230 		if (nd->nd_repstat == 0)
8231 			*lenp = len;
8232 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8233 		/*
8234 		 * For the case where consecutive is not supported, but
8235 		 * synchronous is supported, we can try consecutive == false
8236 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8237 		 * since Copy cannot be done.
8238 		 */
8239 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8240 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8241 			if (!consecutive || *++tl == newnfs_false)
8242 				nd->nd_repstat = NFSERR_NOTSUPP;
8243 		} else
8244 			nd->nd_repstat = NFSERR_BADXDR;
8245 	}
8246 	if (error == 0)
8247 		error = nd->nd_repstat;
8248 nfsmout:
8249 	m_freem(nd->nd_mrep);
8250 	return (error);
8251 }
8252 
8253 /*
8254  * Seek operation.
8255  */
8256 APPLESTATIC int
8257 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8258     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8259 {
8260 	int error, expireret = 0, retrycnt;
8261 	u_int32_t clidrev = 0;
8262 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
8263 	struct nfsnode *np = VTONFS(vp);
8264 	struct nfsfh *nfhp = NULL;
8265 	nfsv4stateid_t stateid;
8266 	void *lckp;
8267 
8268 	if (nmp->nm_clp != NULL)
8269 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8270 	nfhp = np->n_fhp;
8271 	retrycnt = 0;
8272 	do {
8273 		lckp = NULL;
8274 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8275 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8276 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8277 		    nap, attrflagp, cred);
8278 		if (error == NFSERR_STALESTATEID)
8279 			nfscl_initiate_recovery(nmp->nm_clp);
8280 		if (lckp != NULL)
8281 			nfscl_lockderef(lckp);
8282 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8283 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8284 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8285 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8286 		} else if ((error == NFSERR_EXPIRED ||
8287 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8288 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8289 			    curthread);
8290 		}
8291 		retrycnt++;
8292 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8293 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8294 	    error == NFSERR_BADSESSION ||
8295 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8296 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8297 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8298 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8299 	if (error && retrycnt >= 4)
8300 		error = EIO;
8301 	return (error);
8302 }
8303 
8304 /*
8305  * The seek RPC.
8306  */
8307 static int
8308 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8309     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8310 {
8311 	uint32_t *tl;
8312 	int error;
8313 	struct nfsrv_descript nfsd;
8314 	struct nfsrv_descript *nd = &nfsd;
8315 	nfsattrbit_t attrbits;
8316 
8317 	*attrflagp = 0;
8318 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8319 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8320 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8321 	txdr_hyper(*offp, tl); tl += 2;
8322 	*tl++ = txdr_unsigned(content);
8323 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8324 	NFSGETATTR_ATTRBIT(&attrbits);
8325 	nfsrv_putattrbit(nd, &attrbits);
8326 	error = nfscl_request(nd, vp, curthread, cred, NULL);
8327 	if (error != 0)
8328 		return (error);
8329 	if (nd->nd_repstat == 0) {
8330 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8331 		if (*tl++ == newnfs_true)
8332 			*eofp = true;
8333 		else
8334 			*eofp = false;
8335 		*offp = fxdr_hyper(tl);
8336 		/* Just skip over Getattr op status. */
8337 		error = nfsm_loadattr(nd, nap);
8338 		if (error == 0)
8339 			*attrflagp = 1;
8340 	}
8341 	error = nd->nd_repstat;
8342 nfsmout:
8343 	m_freem(nd->nd_mrep);
8344 	return (error);
8345 }
8346 
8347 /*
8348  * The getextattr RPC.
8349  */
8350 APPLESTATIC int
8351 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8352     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8353 {
8354 	uint32_t *tl;
8355 	int error;
8356 	struct nfsrv_descript nfsd;
8357 	struct nfsrv_descript *nd = &nfsd;
8358 	nfsattrbit_t attrbits;
8359 	uint32_t len, len2;
8360 
8361 	*attrflagp = 0;
8362 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8363 	nfsm_strtom(nd, name, strlen(name));
8364 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8365 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8366 	NFSGETATTR_ATTRBIT(&attrbits);
8367 	nfsrv_putattrbit(nd, &attrbits);
8368 	error = nfscl_request(nd, vp, p, cred, NULL);
8369 	if (error != 0)
8370 		return (error);
8371 	if (nd->nd_repstat == 0) {
8372 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8373 		len = fxdr_unsigned(uint32_t, *tl);
8374 		/* Sanity check lengths. */
8375 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8376 		    uiop->uio_resid <= UINT32_MAX) {
8377 			len2 = uiop->uio_resid;
8378 			if (len2 >= len)
8379 				error = nfsm_mbufuio(nd, uiop, len);
8380 			else {
8381 				error = nfsm_mbufuio(nd, uiop, len2);
8382 				if (error == 0) {
8383 					/*
8384 					 * nfsm_mbufuio() advances to a multiple
8385 					 * of 4, so round up len2 as well.  Then
8386 					 * we need to advance over the rest of
8387 					 * the data, rounding up the remaining
8388 					 * length.
8389 					 */
8390 					len2 = NFSM_RNDUP(len2);
8391 					len2 = NFSM_RNDUP(len - len2);
8392 					if (len2 > 0)
8393 						error = nfsm_advance(nd, len2,
8394 						    -1);
8395 				}
8396 			}
8397 		} else if (uiop == NULL && len > 0) {
8398 			/* Just wants the length and not the data. */
8399 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8400 		} else if (len > 0)
8401 			error = ENOATTR;
8402 		if (error != 0)
8403 			goto nfsmout;
8404 		*lenp = len;
8405 		/* Just skip over Getattr op status. */
8406 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8407 		error = nfsm_loadattr(nd, nap);
8408 		if (error == 0)
8409 			*attrflagp = 1;
8410 	}
8411 	if (error == 0)
8412 		error = nd->nd_repstat;
8413 nfsmout:
8414 	m_freem(nd->nd_mrep);
8415 	return (error);
8416 }
8417 
8418 /*
8419  * The setextattr RPC.
8420  */
8421 APPLESTATIC int
8422 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8423     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8424 {
8425 	uint32_t *tl;
8426 	int error;
8427 	struct nfsrv_descript nfsd;
8428 	struct nfsrv_descript *nd = &nfsd;
8429 	nfsattrbit_t attrbits;
8430 
8431 	*attrflagp = 0;
8432 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8433 	if (uiop->uio_resid > nd->nd_maxreq) {
8434 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8435 		m_freem(nd->nd_mreq);
8436 		return (EINVAL);
8437 	}
8438 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8439 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8440 	nfsm_strtom(nd, name, strlen(name));
8441 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8442 	*tl = txdr_unsigned(uiop->uio_resid);
8443 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8444 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8445 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8446 	NFSGETATTR_ATTRBIT(&attrbits);
8447 	nfsrv_putattrbit(nd, &attrbits);
8448 	error = nfscl_request(nd, vp, p, cred, NULL);
8449 	if (error != 0)
8450 		return (error);
8451 	if (nd->nd_repstat == 0) {
8452 		/* Just skip over the reply and Getattr op status. */
8453 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8454 		    NFSX_UNSIGNED);
8455 		error = nfsm_loadattr(nd, nap);
8456 		if (error == 0)
8457 			*attrflagp = 1;
8458 	}
8459 	if (error == 0)
8460 		error = nd->nd_repstat;
8461 nfsmout:
8462 	m_freem(nd->nd_mrep);
8463 	return (error);
8464 }
8465 
8466 /*
8467  * The removeextattr RPC.
8468  */
8469 APPLESTATIC int
8470 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8471     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8472 {
8473 	uint32_t *tl;
8474 	int error;
8475 	struct nfsrv_descript nfsd;
8476 	struct nfsrv_descript *nd = &nfsd;
8477 	nfsattrbit_t attrbits;
8478 
8479 	*attrflagp = 0;
8480 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8481 	nfsm_strtom(nd, name, strlen(name));
8482 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8483 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8484 	NFSGETATTR_ATTRBIT(&attrbits);
8485 	nfsrv_putattrbit(nd, &attrbits);
8486 	error = nfscl_request(nd, vp, p, cred, NULL);
8487 	if (error != 0)
8488 		return (error);
8489 	if (nd->nd_repstat == 0) {
8490 		/* Just skip over the reply and Getattr op status. */
8491 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8492 		    NFSX_UNSIGNED);
8493 		error = nfsm_loadattr(nd, nap);
8494 		if (error == 0)
8495 			*attrflagp = 1;
8496 	}
8497 	if (error == 0)
8498 		error = nd->nd_repstat;
8499 nfsmout:
8500 	m_freem(nd->nd_mrep);
8501 	return (error);
8502 }
8503 
8504 /*
8505  * The listextattr RPC.
8506  */
8507 APPLESTATIC int
8508 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8509     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8510     struct ucred *cred, NFSPROC_T *p)
8511 {
8512 	uint32_t *tl;
8513 	int cnt, error, i, len;
8514 	struct nfsrv_descript nfsd;
8515 	struct nfsrv_descript *nd = &nfsd;
8516 	nfsattrbit_t attrbits;
8517 	u_char c;
8518 
8519 	*attrflagp = 0;
8520 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8521 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8522 	txdr_hyper(*cookiep, tl); tl += 2;
8523 	*tl++ = txdr_unsigned(*lenp);
8524 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8525 	NFSGETATTR_ATTRBIT(&attrbits);
8526 	nfsrv_putattrbit(nd, &attrbits);
8527 	error = nfscl_request(nd, vp, p, cred, NULL);
8528 	if (error != 0)
8529 		return (error);
8530 	*eofp = true;
8531 	*lenp = 0;
8532 	if (nd->nd_repstat == 0) {
8533 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8534 		*cookiep = fxdr_hyper(tl); tl += 2;
8535 		cnt = fxdr_unsigned(int, *tl);
8536 		if (cnt < 0) {
8537 			error = EBADRPC;
8538 			goto nfsmout;
8539 		}
8540 		for (i = 0; i < cnt; i++) {
8541 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8542 			len = fxdr_unsigned(int, *tl);
8543 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8544 				error = EBADRPC;
8545 				goto nfsmout;
8546 			}
8547 			if (uiop == NULL)
8548 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8549 			else if (uiop->uio_resid >= len + 1) {
8550 				c = len;
8551 				error = uiomove(&c, sizeof(c), uiop);
8552 				if (error == 0)
8553 					error = nfsm_mbufuio(nd, uiop, len);
8554 			} else {
8555 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8556 				*eofp = false;
8557 			}
8558 			if (error != 0)
8559 				goto nfsmout;
8560 			*lenp += (len + 1);
8561 		}
8562 		/* Get the eof and skip over the Getattr op status. */
8563 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8564 		/*
8565 		 * *eofp is set false above, because it wasn't able to copy
8566 		 * all of the reply.
8567 		 */
8568 		if (*eofp && *tl == 0)
8569 			*eofp = false;
8570 		error = nfsm_loadattr(nd, nap);
8571 		if (error == 0)
8572 			*attrflagp = 1;
8573 	}
8574 	if (error == 0)
8575 		error = nd->nd_repstat;
8576 nfsmout:
8577 	m_freem(nd->nd_mrep);
8578 	return (error);
8579 }
8580 
8581