xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 62ff619dcc3540659a319be71c9a489f1659e14a)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Rpc op calls, generally called from the vnode op calls or through the
41  * buffer cache, for NFS v2, 3 and 4.
42  * These do not normally make any changes to vnode arguments or use
43  * structures that might change between the VFS variants. The returned
44  * arguments are all at the end, after the NFSPROC_T *p one.
45  */
46 
47 #include "opt_inet6.h"
48 
49 #include <fs/nfs/nfsport.h>
50 #include <fs/nfsclient/nfs.h>
51 #include <sys/extattr.h>
52 #include <sys/sysctl.h>
53 #include <sys/taskqueue.h>
54 
55 SYSCTL_DECL(_vfs_nfs);
56 
57 static int	nfsignore_eexist = 0;
58 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
59     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
60 
61 static int	nfscl_dssameconn = 0;
62 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
63     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
64 
65 static uint64_t nfs_maxcopyrange = SSIZE_MAX;
66 SYSCTL_U64(_vfs_nfs, OID_AUTO, maxcopyrange, CTLFLAG_RW,
67     &nfs_maxcopyrange, 0, "Max size of a Copy so RPC times reasonable");
68 
69 /*
70  * Global variables
71  */
72 extern struct nfsstatsv1 nfsstatsv1;
73 extern int nfs_numnfscbd;
74 extern struct timeval nfsboottime;
75 extern u_int32_t newnfs_false, newnfs_true;
76 extern nfstype nfsv34_type[9];
77 extern int nfsrv_useacl;
78 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
79 extern int nfscl_debuglevel;
80 extern int nfs_pnfsiothreads;
81 extern u_long sb_max_adj;
82 NFSCLSTATEMUTEX;
83 int nfstest_outofseq = 0;
84 int nfscl_assumeposixlocks = 1;
85 int nfscl_enablecallb = 0;
86 short nfsv4_cbport = NFSV4_CBPORT;
87 int nfstest_openallsetattr = 0;
88 
89 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
90 
91 /*
92  * nfscl_getsameserver() can return one of three values:
93  * NFSDSP_USETHISSESSION - Use this session for the DS.
94  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
95  *     session.
96  * NFSDSP_NOTFOUND - No matching server was found.
97  */
98 enum nfsclds_state {
99 	NFSDSP_USETHISSESSION = 0,
100 	NFSDSP_SEQTHISSESSION = 1,
101 	NFSDSP_NOTFOUND = 2,
102 };
103 
104 /*
105  * Do a write RPC on a DS data file, using this structure for the arguments,
106  * so that this function can be executed by a separate kernel process.
107  */
108 struct nfsclwritedsdorpc {
109 	int			done;
110 	int			inprog;
111 	struct task		tsk;
112 	struct vnode		*vp;
113 	int			iomode;
114 	int			must_commit;
115 	nfsv4stateid_t		*stateidp;
116 	struct nfsclds		*dsp;
117 	uint64_t		off;
118 	int			len;
119 #ifdef notyet
120 	int			advise;
121 #endif
122 	struct nfsfh		*fhp;
123 	struct mbuf		*m;
124 	int			vers;
125 	int			minorvers;
126 	struct ucred		*cred;
127 	NFSPROC_T		*p;
128 	int			err;
129 };
130 
131 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
132     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *);
133 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
134     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *);
135 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
136     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
137     int);
138 static int nfsrpc_deallocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
139     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
140 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
141     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
142     struct nfsvattr *, struct nfsfh **, int *, int *);
143 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
144     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
145     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
146     int *, int *);
147 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
148     struct nfscllockowner *, u_int64_t, u_int64_t,
149     u_int32_t, struct ucred *, NFSPROC_T *, int);
150 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
151     struct acl *, nfsv4stateid_t *);
152 static int nfsrpc_layouterror(struct nfsmount *, uint8_t *, int, uint64_t,
153     uint64_t, nfsv4stateid_t *, struct ucred *, NFSPROC_T *, uint32_t,
154     uint32_t, char *);
155 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
156     uint32_t, uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
157     struct ucred *, NFSPROC_T *);
158 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
159     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
160     NFSPROC_T *);
161 static void nfscl_initsessionslots(struct nfsclsession *);
162 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
163     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
164     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
165     NFSPROC_T *);
166 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
167     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
168     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
169     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
170 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
171     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
172     struct ucred *, NFSPROC_T *);
173 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
174     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
175     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
176 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
177     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
178     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
179 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
180     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
181     struct ucred *, NFSPROC_T *);
182 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
183     struct nfsclds *, struct nfsclds **, uint32_t *);
184 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
185     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
186     NFSPROC_T *);
187 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
188     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
189 #ifdef notyet
190 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
191     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
192     NFSPROC_T *);
193 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
194     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
195 #endif
196 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
197     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *);
198 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
199     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
200 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
201     NFSPROC_T *);
202 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
203     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
204 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
205     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
206     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
207 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
208     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
209     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
210     struct nfsfh **, int *, int *, int *);
211 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
212     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
213     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
214     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
215 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
216     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
217     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
218     struct nfsfh **, int *, int *, int *, nfsv4stateid_t *,
219     int, int, int, int *, struct nfsclflayouthead *, int *);
220 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
221     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
222     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
223 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
224     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
225     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
226 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
227     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
228     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
229 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
230     int, struct nfsvattr *, int *, struct ucred *);
231 static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
232 
233 int nfs_pnfsio(task_fn_t *, void *);
234 
235 /*
236  * nfs null call from vfs.
237  */
238 int
239 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
240 {
241 	int error;
242 	struct nfsrv_descript nfsd, *nd = &nfsd;
243 
244 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
245 	error = nfscl_request(nd, vp, p, cred);
246 	if (nd->nd_repstat && !error)
247 		error = nd->nd_repstat;
248 	m_freem(nd->nd_mrep);
249 	return (error);
250 }
251 
252 /*
253  * nfs access rpc op.
254  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
255  * modes are changed on the server, accesses might still fail later.
256  */
257 int
258 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
259     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
260 {
261 	int error;
262 	u_int32_t mode, rmode;
263 
264 	if (acmode & VREAD)
265 		mode = NFSACCESS_READ;
266 	else
267 		mode = 0;
268 	if (vnode_vtype(vp) == VDIR) {
269 		if (acmode & VWRITE)
270 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
271 				 NFSACCESS_DELETE);
272 		if (acmode & VEXEC)
273 			mode |= NFSACCESS_LOOKUP;
274 	} else {
275 		if (acmode & VWRITE)
276 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
277 		if (acmode & VEXEC)
278 			mode |= NFSACCESS_EXECUTE;
279 	}
280 
281 	/*
282 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
283 	 */
284 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
285 	    NULL);
286 
287 	/*
288 	 * The NFS V3 spec does not clarify whether or not
289 	 * the returned access bits can be a superset of
290 	 * the ones requested, so...
291 	 */
292 	if (!error && (rmode & mode) != mode)
293 		error = EACCES;
294 	return (error);
295 }
296 
297 /*
298  * The actual rpc, separated out for Darwin.
299  */
300 int
301 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
302     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
303     void *stuff)
304 {
305 	u_int32_t *tl;
306 	u_int32_t supported, rmode;
307 	int error;
308 	struct nfsrv_descript nfsd, *nd = &nfsd;
309 	nfsattrbit_t attrbits;
310 
311 	*attrflagp = 0;
312 	supported = mode;
313 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
314 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
315 	*tl = txdr_unsigned(mode);
316 	if (nd->nd_flag & ND_NFSV4) {
317 		/*
318 		 * And do a Getattr op.
319 		 */
320 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
321 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
322 		NFSGETATTR_ATTRBIT(&attrbits);
323 		(void) nfsrv_putattrbit(nd, &attrbits);
324 	}
325 	error = nfscl_request(nd, vp, p, cred);
326 	if (error)
327 		return (error);
328 	if (nd->nd_flag & ND_NFSV3) {
329 		error = nfscl_postop_attr(nd, nap, attrflagp);
330 		if (error)
331 			goto nfsmout;
332 	}
333 	if (!nd->nd_repstat) {
334 		if (nd->nd_flag & ND_NFSV4) {
335 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
336 			supported = fxdr_unsigned(u_int32_t, *tl++);
337 		} else {
338 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
339 		}
340 		rmode = fxdr_unsigned(u_int32_t, *tl);
341 		if (nd->nd_flag & ND_NFSV4)
342 			error = nfscl_postop_attr(nd, nap, attrflagp);
343 
344 		/*
345 		 * It's not obvious what should be done about
346 		 * unsupported access modes. For now, be paranoid
347 		 * and clear the unsupported ones.
348 		 */
349 		rmode &= supported;
350 		*rmodep = rmode;
351 	} else
352 		error = nd->nd_repstat;
353 nfsmout:
354 	m_freem(nd->nd_mrep);
355 	return (error);
356 }
357 
358 /*
359  * nfs open rpc
360  */
361 int
362 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
363 {
364 	struct nfsclopen *op;
365 	struct nfscldeleg *dp;
366 	struct nfsfh *nfhp;
367 	struct nfsnode *np = VTONFS(vp);
368 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
369 	u_int32_t mode, clidrev;
370 	int ret, newone, error, expireret = 0, retrycnt;
371 
372 	/*
373 	 * For NFSv4, Open Ops are only done on Regular Files.
374 	 */
375 	if (vnode_vtype(vp) != VREG)
376 		return (0);
377 	mode = 0;
378 	if (amode & FREAD)
379 		mode |= NFSV4OPEN_ACCESSREAD;
380 	if (amode & FWRITE)
381 		mode |= NFSV4OPEN_ACCESSWRITE;
382 	nfhp = np->n_fhp;
383 
384 	retrycnt = 0;
385 #ifdef notdef
386 { char name[100]; int namel;
387 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
388 bcopy(NFS4NODENAME(np->n_v4), name, namel);
389 name[namel] = '\0';
390 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
391 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
392 else printf(" fhl=0\n");
393 }
394 #endif
395 	do {
396 	    dp = NULL;
397 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
398 		cred, p, NULL, &op, &newone, &ret, 1, true);
399 	    if (error) {
400 		return (error);
401 	    }
402 	    if (nmp->nm_clp != NULL)
403 		clidrev = nmp->nm_clp->nfsc_clientidrev;
404 	    else
405 		clidrev = 0;
406 	    if (ret == NFSCLOPEN_DOOPEN) {
407 		if (np->n_v4 != NULL) {
408 			/*
409 			 * For the first attempt, try and get a layout, if
410 			 * pNFS is enabled for the mount.
411 			 */
412 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
413 			    nfs_numnfscbd == 0 ||
414 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
415 				error = nfsrpc_openrpc(nmp, vp,
416 				    np->n_v4->n4_data,
417 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
418 				    np->n_fhp->nfh_len, mode, op,
419 				    NFS4NODENAME(np->n_v4),
420 				    np->n_v4->n4_namelen,
421 				    &dp, 0, 0x0, cred, p, 0, 0);
422 			else
423 				error = nfsrpc_getopenlayout(nmp, vp,
424 				    np->n_v4->n4_data,
425 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
426 				    np->n_fhp->nfh_len, mode, op,
427 				    NFS4NODENAME(np->n_v4),
428 				    np->n_v4->n4_namelen, &dp, cred, p);
429 			if (dp != NULL) {
430 #ifdef APPLE
431 				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
432 #else
433 				NFSLOCKNODE(np);
434 				np->n_flag &= ~NDELEGMOD;
435 				/*
436 				 * Invalidate the attribute cache, so that
437 				 * attributes that pre-date the issue of a
438 				 * delegation are not cached, since the
439 				 * cached attributes will remain valid while
440 				 * the delegation is held.
441 				 */
442 				NFSINVALATTRCACHE(np);
443 				NFSUNLOCKNODE(np);
444 #endif
445 				(void) nfscl_deleg(nmp->nm_mountp,
446 				    op->nfso_own->nfsow_clp,
447 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
448 			}
449 		} else {
450 			error = EIO;
451 		}
452 		newnfs_copyincred(cred, &op->nfso_cred);
453 	    } else if (ret == NFSCLOPEN_SETCRED)
454 		/*
455 		 * This is a new local open on a delegation. It needs
456 		 * to have credentials so that an open can be done
457 		 * against the server during recovery.
458 		 */
459 		newnfs_copyincred(cred, &op->nfso_cred);
460 
461 	    /*
462 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
463 	     * been done on this Open successfully and a VOP_CLOSE()
464 	     * is expected for each of these.
465 	     * If error is non-zero, don't increment it, since the Open
466 	     * hasn't succeeded yet.
467 	     */
468 	    if (!error) {
469 		op->nfso_opencnt++;
470 		if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
471 		    NFSLOCKNODE(np);
472 		    np->n_openstateid = op;
473 		    NFSUNLOCKNODE(np);
474 		}
475 	    }
476 	    nfscl_openrelease(nmp, op, error, newone);
477 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
478 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
479 		error == NFSERR_BADSESSION) {
480 		(void) nfs_catnap(PZERO, error, "nfs_open");
481 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
482 		&& clidrev != 0) {
483 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
484 		retrycnt++;
485 	    }
486 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
487 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
488 	    error == NFSERR_BADSESSION ||
489 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
490 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
491 	if (error && retrycnt >= 4)
492 		error = EIO;
493 	return (error);
494 }
495 
496 /*
497  * the actual open rpc
498  */
499 int
500 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
501     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
502     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
503     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
504     int syscred, int recursed)
505 {
506 	u_int32_t *tl;
507 	struct nfsrv_descript nfsd, *nd = &nfsd;
508 	struct nfscldeleg *dp, *ndp = NULL;
509 	struct nfsvattr nfsva;
510 	u_int32_t rflags, deleg;
511 	nfsattrbit_t attrbits;
512 	int error, ret, acesize, limitby;
513 	struct nfsclsession *tsep;
514 
515 	dp = *dpp;
516 	*dpp = NULL;
517 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
518 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
519 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
520 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
521 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
522 	tsep = nfsmnt_mdssession(nmp);
523 	*tl++ = tsep->nfsess_clientid.lval[0];
524 	*tl = tsep->nfsess_clientid.lval[1];
525 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
526 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
527 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
528 	if (reclaim) {
529 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
530 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
531 		*tl = txdr_unsigned(delegtype);
532 	} else {
533 		if (dp != NULL) {
534 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
535 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
536 			if (NFSHASNFSV4N(nmp))
537 				*tl++ = 0;
538 			else
539 				*tl++ = dp->nfsdl_stateid.seqid;
540 			*tl++ = dp->nfsdl_stateid.other[0];
541 			*tl++ = dp->nfsdl_stateid.other[1];
542 			*tl = dp->nfsdl_stateid.other[2];
543 		} else {
544 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
545 		}
546 		(void) nfsm_strtom(nd, name, namelen);
547 	}
548 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
549 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
550 	NFSZERO_ATTRBIT(&attrbits);
551 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
552 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
553 	(void) nfsrv_putattrbit(nd, &attrbits);
554 	if (syscred)
555 		nd->nd_flag |= ND_USEGSSNAME;
556 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
557 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
558 	if (error)
559 		return (error);
560 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
561 	if (!nd->nd_repstat) {
562 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
563 		    6 * NFSX_UNSIGNED);
564 		op->nfso_stateid.seqid = *tl++;
565 		op->nfso_stateid.other[0] = *tl++;
566 		op->nfso_stateid.other[1] = *tl++;
567 		op->nfso_stateid.other[2] = *tl;
568 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
569 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
570 		if (error)
571 			goto nfsmout;
572 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
573 		deleg = fxdr_unsigned(u_int32_t, *tl);
574 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
575 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
576 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
577 			      NFSCLFLAGS_FIRSTDELEG))
578 				op->nfso_own->nfsow_clp->nfsc_flags |=
579 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
580 			ndp = malloc(
581 			    sizeof (struct nfscldeleg) + newfhlen,
582 			    M_NFSCLDELEG, M_WAITOK);
583 			LIST_INIT(&ndp->nfsdl_owner);
584 			LIST_INIT(&ndp->nfsdl_lock);
585 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
586 			ndp->nfsdl_fhlen = newfhlen;
587 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
588 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
589 			nfscl_lockinit(&ndp->nfsdl_rwlock);
590 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
591 			    NFSX_UNSIGNED);
592 			ndp->nfsdl_stateid.seqid = *tl++;
593 			ndp->nfsdl_stateid.other[0] = *tl++;
594 			ndp->nfsdl_stateid.other[1] = *tl++;
595 			ndp->nfsdl_stateid.other[2] = *tl++;
596 			ret = fxdr_unsigned(int, *tl);
597 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
598 				ndp->nfsdl_flags = NFSCLDL_WRITE;
599 				/*
600 				 * Indicates how much the file can grow.
601 				 */
602 				NFSM_DISSECT(tl, u_int32_t *,
603 				    3 * NFSX_UNSIGNED);
604 				limitby = fxdr_unsigned(int, *tl++);
605 				switch (limitby) {
606 				case NFSV4OPEN_LIMITSIZE:
607 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
608 					break;
609 				case NFSV4OPEN_LIMITBLOCKS:
610 					ndp->nfsdl_sizelimit =
611 					    fxdr_unsigned(u_int64_t, *tl++);
612 					ndp->nfsdl_sizelimit *=
613 					    fxdr_unsigned(u_int64_t, *tl);
614 					break;
615 				default:
616 					error = NFSERR_BADXDR;
617 					goto nfsmout;
618 				}
619 			} else {
620 				ndp->nfsdl_flags = NFSCLDL_READ;
621 			}
622 			if (ret)
623 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
624 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
625 			    &ret, &acesize, p);
626 			if (error)
627 				goto nfsmout;
628 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
629 			error = NFSERR_BADXDR;
630 			goto nfsmout;
631 		}
632 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
633 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
634 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
635 		    NULL, NULL, NULL, p, cred);
636 		if (error)
637 			goto nfsmout;
638 		if (ndp != NULL) {
639 			ndp->nfsdl_change = nfsva.na_filerev;
640 			ndp->nfsdl_modtime = nfsva.na_mtime;
641 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
642 		}
643 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
644 		    do {
645 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
646 			    cred, p);
647 			if (ret == NFSERR_DELAY)
648 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
649 		    } while (ret == NFSERR_DELAY);
650 		    error = ret;
651 		}
652 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
653 		    nfscl_assumeposixlocks)
654 		    op->nfso_posixlock = 1;
655 		else
656 		    op->nfso_posixlock = 0;
657 
658 		/*
659 		 * If the server is handing out delegations, but we didn't
660 		 * get one because an OpenConfirm was required, try the
661 		 * Open again, to get a delegation. This is a harmless no-op,
662 		 * from a server's point of view.
663 		 */
664 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
665 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
666 		    && !error && dp == NULL && ndp == NULL && !recursed) {
667 		    do {
668 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
669 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
670 			    cred, p, syscred, 1);
671 			if (ret == NFSERR_DELAY)
672 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
673 		    } while (ret == NFSERR_DELAY);
674 		    if (ret) {
675 			if (ndp != NULL) {
676 				free(ndp, M_NFSCLDELEG);
677 				ndp = NULL;
678 			}
679 			if (ret == NFSERR_STALECLIENTID ||
680 			    ret == NFSERR_STALEDONTRECOVER ||
681 			    ret == NFSERR_BADSESSION)
682 				error = ret;
683 		    }
684 		}
685 	}
686 	if (nd->nd_repstat != 0 && error == 0)
687 		error = nd->nd_repstat;
688 	if (error == NFSERR_STALECLIENTID)
689 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
690 nfsmout:
691 	if (!error)
692 		*dpp = ndp;
693 	else if (ndp != NULL)
694 		free(ndp, M_NFSCLDELEG);
695 	m_freem(nd->nd_mrep);
696 	return (error);
697 }
698 
699 /*
700  * open downgrade rpc
701  */
702 int
703 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
704     struct ucred *cred, NFSPROC_T *p)
705 {
706 	u_int32_t *tl;
707 	struct nfsrv_descript nfsd, *nd = &nfsd;
708 	int error;
709 
710 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
711 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
712 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
713 		*tl++ = 0;
714 	else
715 		*tl++ = op->nfso_stateid.seqid;
716 	*tl++ = op->nfso_stateid.other[0];
717 	*tl++ = op->nfso_stateid.other[1];
718 	*tl++ = op->nfso_stateid.other[2];
719 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
720 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
721 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
722 	error = nfscl_request(nd, vp, p, cred);
723 	if (error)
724 		return (error);
725 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
726 	if (!nd->nd_repstat) {
727 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
728 		op->nfso_stateid.seqid = *tl++;
729 		op->nfso_stateid.other[0] = *tl++;
730 		op->nfso_stateid.other[1] = *tl++;
731 		op->nfso_stateid.other[2] = *tl;
732 	}
733 	if (nd->nd_repstat && error == 0)
734 		error = nd->nd_repstat;
735 	if (error == NFSERR_STALESTATEID)
736 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
737 nfsmout:
738 	m_freem(nd->nd_mrep);
739 	return (error);
740 }
741 
742 /*
743  * V4 Close operation.
744  */
745 int
746 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
747 {
748 	struct nfsclclient *clp;
749 	int error;
750 
751 	if (vnode_vtype(vp) != VREG)
752 		return (0);
753 	if (doclose)
754 		error = nfscl_doclose(vp, &clp, p);
755 	else {
756 		error = nfscl_getclose(vp, &clp);
757 		if (error == 0)
758 			nfscl_clientrelease(clp);
759 	}
760 	return (error);
761 }
762 
763 /*
764  * Close the open.
765  */
766 int
767 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
768     bool loop_on_delayed, bool freeop)
769 {
770 	struct nfsrv_descript nfsd, *nd = &nfsd;
771 	struct nfscllockowner *lp, *nlp;
772 	struct nfscllock *lop, *nlop;
773 	struct ucred *tcred;
774 	u_int64_t off = 0, len = 0;
775 	u_int32_t type = NFSV4LOCKT_READ;
776 	int error, do_unlock, trycnt;
777 
778 	tcred = newnfs_getcred();
779 	newnfs_copycred(&op->nfso_cred, tcred);
780 	/*
781 	 * (Theoretically this could be done in the same
782 	 *  compound as the close, but having multiple
783 	 *  sequenced Ops in the same compound might be
784 	 *  too scary for some servers.)
785 	 */
786 	if (op->nfso_posixlock) {
787 		off = 0;
788 		len = NFS64BITSSET;
789 		type = NFSV4LOCKT_READ;
790 	}
791 
792 	/*
793 	 * Since this function is only called from VOP_INACTIVE(), no
794 	 * other thread will be manipulating this Open. As such, the
795 	 * lock lists are not being changed by other threads, so it should
796 	 * be safe to do this without locking.
797 	 */
798 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
799 		do_unlock = 1;
800 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
801 			if (op->nfso_posixlock == 0) {
802 				off = lop->nfslo_first;
803 				len = lop->nfslo_end - lop->nfslo_first;
804 				if (lop->nfslo_type == F_WRLCK)
805 					type = NFSV4LOCKT_WRITE;
806 				else
807 					type = NFSV4LOCKT_READ;
808 			}
809 			if (do_unlock) {
810 				trycnt = 0;
811 				do {
812 					error = nfsrpc_locku(nd, nmp, lp, off,
813 					    len, type, tcred, p, 0);
814 					if ((nd->nd_repstat == NFSERR_GRACE ||
815 					    nd->nd_repstat == NFSERR_DELAY) &&
816 					    error == 0)
817 						(void) nfs_catnap(PZERO,
818 						    (int)nd->nd_repstat,
819 						    "nfs_close");
820 				} while ((nd->nd_repstat == NFSERR_GRACE ||
821 				    nd->nd_repstat == NFSERR_DELAY) &&
822 				    error == 0 && trycnt++ < 5);
823 				if (op->nfso_posixlock)
824 					do_unlock = 0;
825 			}
826 			nfscl_freelock(lop, 0);
827 		}
828 		/*
829 		 * Do a ReleaseLockOwner.
830 		 * The lock owner name nfsl_owner may be used by other opens for
831 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
832 		 * puts on the wire has the file handle for this file appended
833 		 * to it, so it can be done now.
834 		 */
835 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
836 		    lp->nfsl_open->nfso_fhlen, tcred, p);
837 	}
838 
839 	/*
840 	 * There could be other Opens for different files on the same
841 	 * OpenOwner, so locking is required.
842 	 */
843 	NFSLOCKCLSTATE();
844 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
845 	NFSUNLOCKCLSTATE();
846 	do {
847 		error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
848 		if (error == NFSERR_GRACE)
849 			(void) nfs_catnap(PZERO, error, "nfs_close");
850 	} while (error == NFSERR_GRACE);
851 	NFSLOCKCLSTATE();
852 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
853 
854 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
855 		nfscl_freelockowner(lp, 0);
856 	if (freeop && error != NFSERR_DELAY)
857 		nfscl_freeopen(op, 0, true);
858 	NFSUNLOCKCLSTATE();
859 	NFSFREECRED(tcred);
860 	return (error);
861 }
862 
863 /*
864  * The actual Close RPC.
865  */
866 int
867 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
868     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
869     int syscred)
870 {
871 	u_int32_t *tl;
872 	int error;
873 
874 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
875 	    op->nfso_fhlen, NULL, NULL, 0, 0);
876 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
877 	if (NFSHASNFSV4N(nmp)) {
878 		*tl++ = 0;
879 		*tl++ = 0;
880 	} else {
881 		*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
882 		*tl++ = op->nfso_stateid.seqid;
883 	}
884 	*tl++ = op->nfso_stateid.other[0];
885 	*tl++ = op->nfso_stateid.other[1];
886 	*tl = op->nfso_stateid.other[2];
887 	if (syscred)
888 		nd->nd_flag |= ND_USEGSSNAME;
889 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
890 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
891 	if (error)
892 		return (error);
893 	if (!NFSHASNFSV4N(nmp))
894 		NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
895 	if (nd->nd_repstat == 0)
896 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
897 	error = nd->nd_repstat;
898 	if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
899 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
900 nfsmout:
901 	m_freem(nd->nd_mrep);
902 	return (error);
903 }
904 
905 /*
906  * V4 Open Confirm RPC.
907  */
908 int
909 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
910     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
911 {
912 	u_int32_t *tl;
913 	struct nfsrv_descript nfsd, *nd = &nfsd;
914 	struct nfsmount *nmp;
915 	int error;
916 
917 	nmp = VFSTONFS(vp->v_mount);
918 	if (NFSHASNFSV4N(nmp))
919 		return (0);		/* No confirmation for NFSv4.1. */
920 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
921 	    0, 0);
922 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
923 	*tl++ = op->nfso_stateid.seqid;
924 	*tl++ = op->nfso_stateid.other[0];
925 	*tl++ = op->nfso_stateid.other[1];
926 	*tl++ = op->nfso_stateid.other[2];
927 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
928 	error = nfscl_request(nd, vp, p, cred);
929 	if (error)
930 		return (error);
931 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
932 	if (!nd->nd_repstat) {
933 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
934 		op->nfso_stateid.seqid = *tl++;
935 		op->nfso_stateid.other[0] = *tl++;
936 		op->nfso_stateid.other[1] = *tl++;
937 		op->nfso_stateid.other[2] = *tl;
938 	}
939 	error = nd->nd_repstat;
940 	if (error == NFSERR_STALESTATEID)
941 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
942 nfsmout:
943 	m_freem(nd->nd_mrep);
944 	return (error);
945 }
946 
947 /*
948  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
949  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
950  */
951 int
952 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
953     bool *retokp, struct ucred *cred, NFSPROC_T *p)
954 {
955 	u_int32_t *tl;
956 	struct nfsrv_descript nfsd;
957 	struct nfsrv_descript *nd = &nfsd;
958 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
959 	u_short port;
960 	int error, isinet6 = 0, callblen;
961 	nfsquad_t confirm;
962 	static u_int32_t rev = 0;
963 	struct nfsclds *dsp, *odsp;
964 	struct in6_addr a6;
965 	struct nfsclsession *tsep;
966 	struct rpc_reconupcall recon;
967 	struct nfscl_reconarg *rcp;
968 
969 	if (nfsboottime.tv_sec == 0)
970 		NFSSETBOOTTIME(nfsboottime);
971 	if (NFSHASNFSV4N(nmp)) {
972 		error = NFSERR_BADSESSION;
973 		odsp = dsp = NULL;
974 		if (retokp != NULL) {
975 			NFSLOCKMNT(nmp);
976 			odsp = TAILQ_FIRST(&nmp->nm_sess);
977 			NFSUNLOCKMNT(nmp);
978 		}
979 		if (odsp != NULL) {
980 			/*
981 			 * When a session already exists, first try a
982 			 * CreateSession with the extant ClientID.
983 			 */
984 			dsp = malloc(sizeof(struct nfsclds) +
985 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
986 			    M_WAITOK | M_ZERO);
987 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
988 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
989 			dsp->nfsclds_sess.nfsess_clientid =
990 			    odsp->nfsclds_sess.nfsess_clientid;
991 			dsp->nfsclds_sess.nfsess_sequenceid =
992 			    odsp->nfsclds_sess.nfsess_sequenceid;
993 			dsp->nfsclds_flags = odsp->nfsclds_flags;
994 			if (dsp->nfsclds_servownlen > 0)
995 				memcpy(dsp->nfsclds_serverown,
996 				    odsp->nfsclds_serverown,
997 				    dsp->nfsclds_servownlen + 1);
998 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
999 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
1000 			    NULL, MTX_DEF);
1001 			nfscl_initsessionslots(&dsp->nfsclds_sess);
1002 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
1003 			    &nmp->nm_sockreq, NULL,
1004 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
1005 			NFSCL_DEBUG(1, "create session for extant "
1006 			    "ClientID=%d\n", error);
1007 			if (error != 0) {
1008 				nfscl_freenfsclds(dsp);
1009 				dsp = NULL;
1010 				/*
1011 				 * If *retokp is true, return any error other
1012 				 * than NFSERR_STALECLIENTID,
1013 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
1014 				 * so that nfscl_recover() will not loop.
1015 				 */
1016 				if (*retokp)
1017 					return (NFSERR_IO);
1018 			} else
1019 				*retokp = true;
1020 		} else if (retokp != NULL && *retokp)
1021 			return (NFSERR_IO);
1022 		if (error != 0) {
1023 			/*
1024 			 * Either there was no previous session or the
1025 			 * CreateSession attempt failed, so...
1026 			 * do an ExchangeID followed by the CreateSession.
1027 			 */
1028 			clp->nfsc_rev = rev++;
1029 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1030 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1031 			    cred, p);
1032 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1033 			if (error == 0)
1034 				error = nfsrpc_createsession(nmp,
1035 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1036 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1037 				    cred, p);
1038 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1039 		}
1040 		if (error == 0) {
1041 			/*
1042 			 * If the session supports a backchannel, set up
1043 			 * the BindConnectionToSession call in the krpc
1044 			 * so that it is done on a reconnection.
1045 			 */
1046 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1047 				rcp = mem_alloc(sizeof(*rcp));
1048 				rcp->minorvers = nmp->nm_minorvers;
1049 				memcpy(rcp->sessionid,
1050 				    dsp->nfsclds_sess.nfsess_sessionid,
1051 				    NFSX_V4SESSIONID);
1052 				recon.call = nfsrpc_bindconnsess;
1053 				recon.arg = rcp;
1054 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1055 				    &recon);
1056 			}
1057 
1058 			NFSLOCKMNT(nmp);
1059 			/*
1060 			 * The old sessions cannot be safely free'd
1061 			 * here, since they may still be used by
1062 			 * in-progress RPCs.
1063 			 */
1064 			tsep = NULL;
1065 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL) {
1066 				tsep = NFSMNT_MDSSESSION(nmp);
1067 				if (tsep->nfsess_defunct == 0)
1068 					printf("nfsrpc_setclient: "
1069 					    "nfsess_defunct not set\n");
1070 			}
1071 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1072 			    nfsclds_list);
1073 			/*
1074 			 * Wake up RPCs waiting for a slot on the
1075 			 * old session. These will then fail with
1076 			 * NFSERR_BADSESSION and be retried with the
1077 			 * new session by nfsv4_setsequence().
1078 			 * Also wakeup() processes waiting for the
1079 			 * new session.
1080 			 */
1081 			if (tsep != NULL)
1082 				wakeup(&tsep->nfsess_slots);
1083 			wakeup(&nmp->nm_sess);
1084 			NFSUNLOCKMNT(nmp);
1085 		} else if (dsp != NULL)
1086 			nfscl_freenfsclds(dsp);
1087 		if (error == 0 && reclaim == 0) {
1088 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1089 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1090 			if (error == NFSERR_COMPLETEALREADY ||
1091 			    error == NFSERR_NOTSUPP)
1092 				/* Ignore this error. */
1093 				error = 0;
1094 		}
1095 		return (error);
1096 	} else if (retokp != NULL && *retokp)
1097 		return (NFSERR_IO);
1098 	clp->nfsc_rev = rev++;
1099 
1100 	/*
1101 	 * Allocate a single session structure for NFSv4.0, because some of
1102 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1103 	 */
1104 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1105 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1106 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1107 	NFSLOCKMNT(nmp);
1108 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1109 	tsep = NFSMNT_MDSSESSION(nmp);
1110 	NFSUNLOCKMNT(nmp);
1111 
1112 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
1113 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1114 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1115 	*tl = txdr_unsigned(clp->nfsc_rev);
1116 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1117 
1118 	/*
1119 	 * set up the callback address
1120 	 */
1121 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1122 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1123 	callblen = strlen(nfsv4_callbackaddr);
1124 	if (callblen == 0)
1125 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1126 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1127 	    (callblen > 0 || cp != NULL)) {
1128 		port = htons(nfsv4_cbport);
1129 		cp2 = (u_int8_t *)&port;
1130 #ifdef INET6
1131 		if ((callblen > 0 &&
1132 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1133 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1134 
1135 			(void) nfsm_strtom(nd, "tcp6", 4);
1136 			if (callblen == 0) {
1137 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1138 				ip6add = ip6buf;
1139 			} else {
1140 				ip6add = nfsv4_callbackaddr;
1141 			}
1142 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1143 			    ip6add, cp2[0], cp2[1]);
1144 		} else
1145 #endif
1146 		{
1147 			(void) nfsm_strtom(nd, "tcp", 3);
1148 			if (callblen == 0)
1149 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1150 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1151 				    cp[2], cp[3], cp2[0], cp2[1]);
1152 			else
1153 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1154 				    "%s.%d.%d", nfsv4_callbackaddr,
1155 				    cp2[0], cp2[1]);
1156 		}
1157 		(void) nfsm_strtom(nd, addr, strlen(addr));
1158 	} else {
1159 		(void) nfsm_strtom(nd, "tcp", 3);
1160 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1161 	}
1162 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1163 	*tl = txdr_unsigned(clp->nfsc_cbident);
1164 	nd->nd_flag |= ND_USEGSSNAME;
1165 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1166 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1167 	if (error)
1168 		return (error);
1169 	if (nd->nd_repstat == 0) {
1170 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1171 	    tsep->nfsess_clientid.lval[0] = *tl++;
1172 	    tsep->nfsess_clientid.lval[1] = *tl++;
1173 	    confirm.lval[0] = *tl++;
1174 	    confirm.lval[1] = *tl;
1175 	    m_freem(nd->nd_mrep);
1176 	    nd->nd_mrep = NULL;
1177 
1178 	    /*
1179 	     * and confirm it.
1180 	     */
1181 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1182 		NULL, 0, 0);
1183 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1184 	    *tl++ = tsep->nfsess_clientid.lval[0];
1185 	    *tl++ = tsep->nfsess_clientid.lval[1];
1186 	    *tl++ = confirm.lval[0];
1187 	    *tl = confirm.lval[1];
1188 	    nd->nd_flag |= ND_USEGSSNAME;
1189 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1190 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1191 	    if (error)
1192 		return (error);
1193 	    m_freem(nd->nd_mrep);
1194 	    nd->nd_mrep = NULL;
1195 	}
1196 	error = nd->nd_repstat;
1197 nfsmout:
1198 	m_freem(nd->nd_mrep);
1199 	return (error);
1200 }
1201 
1202 /*
1203  * nfs getattr call.
1204  */
1205 int
1206 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1207     struct nfsvattr *nap, void *stuff)
1208 {
1209 	struct nfsrv_descript nfsd, *nd = &nfsd;
1210 	int error;
1211 	nfsattrbit_t attrbits;
1212 
1213 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1214 	if (nd->nd_flag & ND_NFSV4) {
1215 		NFSGETATTR_ATTRBIT(&attrbits);
1216 		(void) nfsrv_putattrbit(nd, &attrbits);
1217 	}
1218 	error = nfscl_request(nd, vp, p, cred);
1219 	if (error)
1220 		return (error);
1221 	if (!nd->nd_repstat)
1222 		error = nfsm_loadattr(nd, nap);
1223 	else
1224 		error = nd->nd_repstat;
1225 	m_freem(nd->nd_mrep);
1226 	return (error);
1227 }
1228 
1229 /*
1230  * nfs getattr call with non-vnode arguments.
1231  */
1232 int
1233 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1234     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1235     uint32_t *leasep)
1236 {
1237 	struct nfsrv_descript nfsd, *nd = &nfsd;
1238 	int error, vers = NFS_VER2;
1239 	nfsattrbit_t attrbits;
1240 
1241 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1242 	if (nd->nd_flag & ND_NFSV4) {
1243 		vers = NFS_VER4;
1244 		NFSGETATTR_ATTRBIT(&attrbits);
1245 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1246 		(void) nfsrv_putattrbit(nd, &attrbits);
1247 	} else if (nd->nd_flag & ND_NFSV3) {
1248 		vers = NFS_VER3;
1249 	}
1250 	if (syscred)
1251 		nd->nd_flag |= ND_USEGSSNAME;
1252 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1253 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1254 	if (error)
1255 		return (error);
1256 	if (nd->nd_repstat == 0) {
1257 		if ((nd->nd_flag & ND_NFSV4) != 0)
1258 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1259 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1260 			    NULL, NULL);
1261 		else
1262 			error = nfsm_loadattr(nd, nap);
1263 	} else
1264 		error = nd->nd_repstat;
1265 	m_freem(nd->nd_mrep);
1266 	return (error);
1267 }
1268 
1269 /*
1270  * Do an nfs setattr operation.
1271  */
1272 int
1273 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1274     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1275     void *stuff)
1276 {
1277 	int error, expireret = 0, openerr, retrycnt;
1278 	u_int32_t clidrev = 0, mode;
1279 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1280 	struct nfsfh *nfhp;
1281 	nfsv4stateid_t stateid;
1282 	void *lckp;
1283 
1284 	if (nmp->nm_clp != NULL)
1285 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1286 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1287 		mode = NFSV4OPEN_ACCESSWRITE;
1288 	else
1289 		mode = NFSV4OPEN_ACCESSREAD;
1290 	retrycnt = 0;
1291 	do {
1292 		lckp = NULL;
1293 		openerr = 1;
1294 		if (NFSHASNFSV4(nmp)) {
1295 			nfhp = VTONFS(vp)->n_fhp;
1296 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1297 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1298 			if (error && vnode_vtype(vp) == VREG &&
1299 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1300 			     nfstest_openallsetattr)) {
1301 				/*
1302 				 * No Open stateid, so try and open the file
1303 				 * now.
1304 				 */
1305 				if (mode == NFSV4OPEN_ACCESSWRITE)
1306 					openerr = nfsrpc_open(vp, FWRITE, cred,
1307 					    p);
1308 				else
1309 					openerr = nfsrpc_open(vp, FREAD, cred,
1310 					    p);
1311 				if (!openerr)
1312 					(void) nfscl_getstateid(vp,
1313 					    nfhp->nfh_fh, nfhp->nfh_len,
1314 					    mode, 0, cred, p, &stateid, &lckp);
1315 			}
1316 		}
1317 		if (vap != NULL)
1318 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1319 			    rnap, attrflagp);
1320 		else
1321 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid);
1322 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1323 			NFSLOCKMNT(nmp);
1324 			nmp->nm_state |= NFSSTA_OPENMODE;
1325 			NFSUNLOCKMNT(nmp);
1326 		}
1327 		if (error == NFSERR_STALESTATEID)
1328 			nfscl_initiate_recovery(nmp->nm_clp);
1329 		if (lckp != NULL)
1330 			nfscl_lockderef(lckp);
1331 		if (!openerr)
1332 			(void) nfsrpc_close(vp, 0, p);
1333 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1334 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1335 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1336 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1337 		} else if ((error == NFSERR_EXPIRED ||
1338 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1339 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1340 		}
1341 		retrycnt++;
1342 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1343 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1344 	    error == NFSERR_BADSESSION ||
1345 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1346 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1347 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1348 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1349 	     retrycnt < 4));
1350 	if (error && retrycnt >= 4)
1351 		error = EIO;
1352 	return (error);
1353 }
1354 
1355 static int
1356 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1357     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1358     struct nfsvattr *rnap, int *attrflagp)
1359 {
1360 	u_int32_t *tl;
1361 	struct nfsrv_descript nfsd, *nd = &nfsd;
1362 	int error;
1363 	nfsattrbit_t attrbits;
1364 
1365 	*attrflagp = 0;
1366 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1367 	if (nd->nd_flag & ND_NFSV4)
1368 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1369 	vap->va_type = vnode_vtype(vp);
1370 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1371 	if (nd->nd_flag & ND_NFSV3) {
1372 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1373 		*tl = newnfs_false;
1374 	} else if (nd->nd_flag & ND_NFSV4) {
1375 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1376 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1377 		NFSGETATTR_ATTRBIT(&attrbits);
1378 		(void) nfsrv_putattrbit(nd, &attrbits);
1379 	}
1380 	error = nfscl_request(nd, vp, p, cred);
1381 	if (error)
1382 		return (error);
1383 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1384 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, NULL);
1385 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1386 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1387 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1388 		error = nfscl_postop_attr(nd, rnap, attrflagp);
1389 	m_freem(nd->nd_mrep);
1390 	if (nd->nd_repstat && !error)
1391 		error = nd->nd_repstat;
1392 	return (error);
1393 }
1394 
1395 /*
1396  * nfs lookup rpc
1397  */
1398 int
1399 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1400     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1401     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff,
1402     uint32_t openmode)
1403 {
1404 	uint32_t deleg, rflags, *tl;
1405 	struct nfsrv_descript nfsd, *nd = &nfsd;
1406 	struct nfsmount *nmp;
1407 	struct nfsnode *np;
1408 	struct nfsfh *nfhp;
1409 	nfsattrbit_t attrbits;
1410 	int error = 0, lookupp = 0, newone, ret, retop;
1411 	uint8_t own[NFSV4CL_LOCKNAMELEN];
1412 	struct nfsclopen *op;
1413 	struct nfscldeleg *ndp;
1414 	nfsv4stateid_t stateid;
1415 
1416 	*attrflagp = 0;
1417 	*dattrflagp = 0;
1418 	if (vnode_vtype(dvp) != VDIR)
1419 		return (ENOTDIR);
1420 	nmp = VFSTONFS(dvp->v_mount);
1421 	if (len > NFS_MAXNAMLEN)
1422 		return (ENAMETOOLONG);
1423 	if (NFSHASNFSV4(nmp) && len == 1 &&
1424 		name[0] == '.') {
1425 		/*
1426 		 * Just return the current dir's fh.
1427 		 */
1428 		np = VTONFS(dvp);
1429 		nfhp = malloc(sizeof (struct nfsfh) +
1430 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1431 		nfhp->nfh_len = np->n_fhp->nfh_len;
1432 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1433 		*nfhpp = nfhp;
1434 		return (0);
1435 	}
1436 	if (NFSHASNFSV4(nmp) && len == 2 &&
1437 		name[0] == '.' && name[1] == '.') {
1438 		lookupp = 1;
1439 		openmode = 0;
1440 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1441 	} else if (openmode != 0) {
1442 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPOPEN, dvp);
1443 		nfsm_strtom(nd, name, len);
1444 	} else {
1445 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1446 		(void) nfsm_strtom(nd, name, len);
1447 	}
1448 	if (nd->nd_flag & ND_NFSV4) {
1449 		NFSGETATTR_ATTRBIT(&attrbits);
1450 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1451 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1452 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1453 		(void) nfsrv_putattrbit(nd, &attrbits);
1454 		if (openmode != 0) {
1455 			/* Test for a VREG file. */
1456 			NFSZERO_ATTRBIT(&attrbits);
1457 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
1458 			NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
1459 			*tl = txdr_unsigned(NFSV4OP_VERIFY);
1460 			nfsrv_putattrbit(nd, &attrbits);
1461 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1462 			*tl++ = txdr_unsigned(NFSX_UNSIGNED);
1463 			*tl = vtonfsv34_type(VREG);
1464 
1465 			/* Attempt the Open for VREG. */
1466 			nfscl_filllockowner(NULL, own, F_POSIX);
1467 			NFSM_BUILD(tl, uint32_t *, 6 * NFSX_UNSIGNED);
1468 			*tl++ = txdr_unsigned(NFSV4OP_OPEN);
1469 			*tl++ = 0;		/* seqid, ignored. */
1470 			*tl++ = txdr_unsigned(openmode);
1471 			*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1472 			*tl++ = 0;		/* ClientID, ignored. */
1473 			*tl = 0;
1474 			nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN);
1475 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1476 			*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
1477 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
1478 		}
1479 	}
1480 	error = nfscl_request(nd, dvp, p, cred);
1481 	if (error)
1482 		return (error);
1483 	ndp = NULL;
1484 	if (nd->nd_repstat) {
1485 		/*
1486 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1487 		 * the lookup is at the root of an fs, so return this dir.
1488 		 */
1489 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1490 		    np = VTONFS(dvp);
1491 		    nfhp = malloc(sizeof (struct nfsfh) +
1492 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1493 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1494 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1495 		    *nfhpp = nfhp;
1496 		    m_freem(nd->nd_mrep);
1497 		    return (0);
1498 		}
1499 		if (nd->nd_flag & ND_NFSV3)
1500 		    error = nfscl_postop_attr(nd, dnap, dattrflagp);
1501 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1502 		    ND_NFSV4) {
1503 			/* Load the directory attributes. */
1504 			error = nfsm_loadattr(nd, dnap);
1505 			if (error != 0)
1506 				goto nfsmout;
1507 			*dattrflagp = 1;
1508 		}
1509 		/* Check Lookup operation reply status. */
1510 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1511 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1512 			if (*++tl != 0)
1513 				goto nfsmout;
1514 		}
1515 		/* Look for GetFH reply. */
1516 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1517 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1518 			if (*++tl != 0)
1519 				goto nfsmout;
1520 			error = nfsm_getfh(nd, nfhpp);
1521 			if (error)
1522 				goto nfsmout;
1523 		}
1524 		/* Look for Getattr reply. */
1525 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1526 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1527 			if (*++tl != 0)
1528 				goto nfsmout;
1529 			error = nfsm_loadattr(nd, nap);
1530 			if (error == 0) {
1531 				/*
1532 				 * We have now successfully completed the
1533 				 * lookup, so set nd_repstat to 0.
1534 				 */
1535 				nd->nd_repstat = 0;
1536 				*attrflagp = 1;
1537 			}
1538 		}
1539 		goto nfsmout;
1540 	}
1541 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1542 		/* Load the directory attributes. */
1543 		error = nfsm_loadattr(nd, dnap);
1544 		if (error != 0)
1545 			goto nfsmout;
1546 		*dattrflagp = 1;
1547 		/* Skip over the Lookup and GetFH operation status values. */
1548 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1549 	}
1550 	error = nfsm_getfh(nd, nfhpp);
1551 	if (error)
1552 		goto nfsmout;
1553 
1554 	error = nfscl_postop_attr(nd, nap, attrflagp);
1555 	if (openmode != 0 && error == 0) {
1556 		NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID +
1557 		    10 * NFSX_UNSIGNED);
1558 		tl += 4;	/* Skip over Verify+Open status. */
1559 		stateid.seqid = *tl++;
1560 		stateid.other[0] = *tl++;
1561 		stateid.other[1] = *tl++;
1562 		stateid.other[2] = *tl;
1563 		rflags = fxdr_unsigned(uint32_t, *(tl + 6));
1564 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1565 		if (error != 0)
1566 			goto nfsmout;
1567 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1568 		deleg = fxdr_unsigned(uint32_t, *tl);
1569 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
1570 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
1571 			/*
1572 			 * Just need to fill in the fields used by
1573 			 * nfscl_trydelegreturn().
1574 			 * Mark the mount point as acquiring
1575 			 * delegations, so NFSPROC_LOOKUPOPEN will
1576 			 * no longer be done.
1577 			 */
1578 			NFSLOCKMNT(nmp);
1579 			nmp->nm_privflag |= NFSMNTP_DELEGISSUED;
1580 			NFSUNLOCKMNT(nmp);
1581 			ndp = malloc(sizeof(struct nfscldeleg) +
1582 			    (*nfhpp)->nfh_len, M_NFSCLDELEG, M_WAITOK);
1583 			ndp->nfsdl_fhlen = (*nfhpp)->nfh_len;
1584 			NFSBCOPY((*nfhpp)->nfh_fh, ndp->nfsdl_fh,
1585 			    ndp->nfsdl_fhlen);
1586 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
1587 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
1588 			ndp->nfsdl_stateid.seqid = *tl++;
1589 			ndp->nfsdl_stateid.other[0] = *tl++;
1590 			ndp->nfsdl_stateid.other[1] = *tl++;
1591 			ndp->nfsdl_stateid.other[2] = *tl++;
1592 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
1593 			error = NFSERR_BADXDR;
1594 			goto nfsmout;
1595 		}
1596 		ret = nfscl_open(dvp, (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len,
1597 		    openmode, 0, cred, p, NULL, &op, &newone, &retop, 1, true);
1598 		if (ret != 0)
1599 			goto nfsmout;
1600 		if (newone != 0) {
1601 			op->nfso_stateid.seqid = stateid.seqid;
1602 			op->nfso_stateid.other[0] = stateid.other[0];
1603 			op->nfso_stateid.other[1] = stateid.other[1];
1604 			op->nfso_stateid.other[2] = stateid.other[2];
1605 			op->nfso_mode = openmode;
1606 		} else {
1607 			op->nfso_stateid.seqid = stateid.seqid;
1608 			if (retop == NFSCLOPEN_DOOPEN)
1609 				op->nfso_mode |= openmode;
1610 		}
1611 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
1612 		    nfscl_assumeposixlocks)
1613 			op->nfso_posixlock = 1;
1614 		else
1615 			op->nfso_posixlock = 0;
1616 		nfscl_openrelease(nmp, op, 0, 0);
1617 		if (ndp != NULL) {
1618 			/*
1619 			 * Since we do not have the vnode, we
1620 			 * cannot invalidate cached attributes.
1621 			 * Just return the delegation.
1622 			 */
1623 			nfscl_trydelegreturn(ndp, cred, nmp, p);
1624 		}
1625 	}
1626 	if ((nd->nd_flag & ND_NFSV3) && !error)
1627 		error = nfscl_postop_attr(nd, dnap, dattrflagp);
1628 nfsmout:
1629 	m_freem(nd->nd_mrep);
1630 	if (!error && nd->nd_repstat)
1631 		error = nd->nd_repstat;
1632 	free(ndp, M_NFSCLDELEG);
1633 	return (error);
1634 }
1635 
1636 /*
1637  * Do a readlink rpc.
1638  */
1639 int
1640 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1641     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1642 {
1643 	u_int32_t *tl;
1644 	struct nfsrv_descript nfsd, *nd = &nfsd;
1645 	struct nfsnode *np = VTONFS(vp);
1646 	nfsattrbit_t attrbits;
1647 	int error, len, cangetattr = 1;
1648 
1649 	*attrflagp = 0;
1650 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1651 	if (nd->nd_flag & ND_NFSV4) {
1652 		/*
1653 		 * And do a Getattr op.
1654 		 */
1655 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1656 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1657 		NFSGETATTR_ATTRBIT(&attrbits);
1658 		(void) nfsrv_putattrbit(nd, &attrbits);
1659 	}
1660 	error = nfscl_request(nd, vp, p, cred);
1661 	if (error)
1662 		return (error);
1663 	if (nd->nd_flag & ND_NFSV3)
1664 		error = nfscl_postop_attr(nd, nap, attrflagp);
1665 	if (!nd->nd_repstat && !error) {
1666 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1667 		/*
1668 		 * This seems weird to me, but must have been added to
1669 		 * FreeBSD for some reason. The only thing I can think of
1670 		 * is that there was/is some server that replies with
1671 		 * more link data than it should?
1672 		 */
1673 		if (len == NFS_MAXPATHLEN) {
1674 			NFSLOCKNODE(np);
1675 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1676 				len = np->n_size;
1677 				cangetattr = 0;
1678 			}
1679 			NFSUNLOCKNODE(np);
1680 		}
1681 		error = nfsm_mbufuio(nd, uiop, len);
1682 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1683 			error = nfscl_postop_attr(nd, nap, attrflagp);
1684 	}
1685 	if (nd->nd_repstat && !error)
1686 		error = nd->nd_repstat;
1687 nfsmout:
1688 	m_freem(nd->nd_mrep);
1689 	return (error);
1690 }
1691 
1692 /*
1693  * Read operation.
1694  */
1695 int
1696 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1697     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1698 {
1699 	int error, expireret = 0, retrycnt;
1700 	u_int32_t clidrev = 0;
1701 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1702 	struct nfsnode *np = VTONFS(vp);
1703 	struct ucred *newcred;
1704 	struct nfsfh *nfhp = NULL;
1705 	nfsv4stateid_t stateid;
1706 	void *lckp;
1707 
1708 	if (nmp->nm_clp != NULL)
1709 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1710 	newcred = cred;
1711 	if (NFSHASNFSV4(nmp)) {
1712 		nfhp = np->n_fhp;
1713 		newcred = NFSNEWCRED(cred);
1714 	}
1715 	retrycnt = 0;
1716 	do {
1717 		lckp = NULL;
1718 		if (NFSHASNFSV4(nmp))
1719 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1720 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1721 			    &lckp);
1722 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1723 		    attrflagp);
1724 		if (error == NFSERR_OPENMODE) {
1725 			NFSLOCKMNT(nmp);
1726 			nmp->nm_state |= NFSSTA_OPENMODE;
1727 			NFSUNLOCKMNT(nmp);
1728 		}
1729 		if (error == NFSERR_STALESTATEID)
1730 			nfscl_initiate_recovery(nmp->nm_clp);
1731 		if (lckp != NULL)
1732 			nfscl_lockderef(lckp);
1733 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1734 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1735 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1736 			(void) nfs_catnap(PZERO, error, "nfs_read");
1737 		} else if ((error == NFSERR_EXPIRED ||
1738 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1739 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1740 		}
1741 		retrycnt++;
1742 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1743 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1744 	    error == NFSERR_BADSESSION ||
1745 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1746 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1747 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1748 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1749 	if (error && retrycnt >= 4)
1750 		error = EIO;
1751 	if (NFSHASNFSV4(nmp))
1752 		NFSFREECRED(newcred);
1753 	return (error);
1754 }
1755 
1756 /*
1757  * The actual read RPC.
1758  */
1759 static int
1760 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1761     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1762     int *attrflagp)
1763 {
1764 	u_int32_t *tl;
1765 	int error = 0, len, retlen, tsiz, eof = 0;
1766 	struct nfsrv_descript nfsd;
1767 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1768 	struct nfsrv_descript *nd = &nfsd;
1769 	int rsize;
1770 	off_t tmp_off;
1771 
1772 	*attrflagp = 0;
1773 	tsiz = uiop->uio_resid;
1774 	tmp_off = uiop->uio_offset + tsiz;
1775 	NFSLOCKMNT(nmp);
1776 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1777 		NFSUNLOCKMNT(nmp);
1778 		return (EFBIG);
1779 	}
1780 	rsize = nmp->nm_rsize;
1781 	NFSUNLOCKMNT(nmp);
1782 	nd->nd_mrep = NULL;
1783 	while (tsiz > 0) {
1784 		*attrflagp = 0;
1785 		len = (tsiz > rsize) ? rsize : tsiz;
1786 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1787 		if (nd->nd_flag & ND_NFSV4)
1788 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1789 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1790 		if (nd->nd_flag & ND_NFSV2) {
1791 			*tl++ = txdr_unsigned(uiop->uio_offset);
1792 			*tl++ = txdr_unsigned(len);
1793 			*tl = 0;
1794 		} else {
1795 			txdr_hyper(uiop->uio_offset, tl);
1796 			*(tl + 2) = txdr_unsigned(len);
1797 		}
1798 		/*
1799 		 * Since I can't do a Getattr for NFSv4 for Write, there
1800 		 * doesn't seem any point in doing one here, either.
1801 		 * (See the comment in nfsrpc_writerpc() for more info.)
1802 		 */
1803 		error = nfscl_request(nd, vp, p, cred);
1804 		if (error)
1805 			return (error);
1806 		if (nd->nd_flag & ND_NFSV3) {
1807 			error = nfscl_postop_attr(nd, nap, attrflagp);
1808 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1809 			error = nfsm_loadattr(nd, nap);
1810 			if (!error)
1811 				*attrflagp = 1;
1812 		}
1813 		if (nd->nd_repstat || error) {
1814 			if (!error)
1815 				error = nd->nd_repstat;
1816 			goto nfsmout;
1817 		}
1818 		if (nd->nd_flag & ND_NFSV3) {
1819 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1820 			eof = fxdr_unsigned(int, *(tl + 1));
1821 		} else if (nd->nd_flag & ND_NFSV4) {
1822 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1823 			eof = fxdr_unsigned(int, *tl);
1824 		}
1825 		NFSM_STRSIZ(retlen, len);
1826 		error = nfsm_mbufuio(nd, uiop, retlen);
1827 		if (error)
1828 			goto nfsmout;
1829 		m_freem(nd->nd_mrep);
1830 		nd->nd_mrep = NULL;
1831 		tsiz -= retlen;
1832 		if (!(nd->nd_flag & ND_NFSV2)) {
1833 			if (eof || retlen == 0)
1834 				tsiz = 0;
1835 		} else if (retlen < len)
1836 			tsiz = 0;
1837 	}
1838 	return (0);
1839 nfsmout:
1840 	if (nd->nd_mrep != NULL)
1841 		m_freem(nd->nd_mrep);
1842 	return (error);
1843 }
1844 
1845 /*
1846  * nfs write operation
1847  * When called_from_strategy != 0, it should return EIO for an error that
1848  * indicates recovery is in progress, so that the buffer will be left
1849  * dirty and be written back to the server later. If it loops around,
1850  * the recovery thread could get stuck waiting for the buffer and recovery
1851  * will then deadlock.
1852  */
1853 int
1854 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1855     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1856     int called_from_strategy, int ioflag)
1857 {
1858 	int error, expireret = 0, retrycnt, nostateid;
1859 	u_int32_t clidrev = 0;
1860 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1861 	struct nfsnode *np = VTONFS(vp);
1862 	struct ucred *newcred;
1863 	struct nfsfh *nfhp = NULL;
1864 	nfsv4stateid_t stateid;
1865 	void *lckp;
1866 
1867 	KASSERT(*must_commit >= 0 && *must_commit <= 2,
1868 	    ("nfsrpc_write: must_commit out of range=%d", *must_commit));
1869 	if (nmp->nm_clp != NULL)
1870 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1871 	newcred = cred;
1872 	if (NFSHASNFSV4(nmp)) {
1873 		newcred = NFSNEWCRED(cred);
1874 		nfhp = np->n_fhp;
1875 	}
1876 	retrycnt = 0;
1877 	do {
1878 		lckp = NULL;
1879 		nostateid = 0;
1880 		if (NFSHASNFSV4(nmp)) {
1881 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1882 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1883 			    &lckp);
1884 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1885 			    stateid.other[2] == 0) {
1886 				nostateid = 1;
1887 				NFSCL_DEBUG(1, "stateid0 in write\n");
1888 			}
1889 		}
1890 
1891 		/*
1892 		 * If there is no stateid for NFSv4, it means this is an
1893 		 * extraneous write after close. Basically a poorly
1894 		 * implemented buffer cache. Just don't do the write.
1895 		 */
1896 		if (nostateid)
1897 			error = 0;
1898 		else
1899 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1900 			    newcred, &stateid, p, nap, attrflagp, ioflag);
1901 		if (error == NFSERR_STALESTATEID)
1902 			nfscl_initiate_recovery(nmp->nm_clp);
1903 		if (lckp != NULL)
1904 			nfscl_lockderef(lckp);
1905 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1906 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1907 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1908 			(void) nfs_catnap(PZERO, error, "nfs_write");
1909 		} else if ((error == NFSERR_EXPIRED ||
1910 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1911 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1912 		}
1913 		retrycnt++;
1914 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1915 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1916 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1917 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1918 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1919 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1920 	if (error != 0 && (retrycnt >= 4 ||
1921 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1922 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1923 		error = EIO;
1924 	if (NFSHASNFSV4(nmp))
1925 		NFSFREECRED(newcred);
1926 	return (error);
1927 }
1928 
1929 /*
1930  * The actual write RPC.
1931  */
1932 static int
1933 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1934     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1935     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, int ioflag)
1936 {
1937 	u_int32_t *tl;
1938 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1939 	struct nfsnode *np = VTONFS(vp);
1940 	int error = 0, len, rlen, commit, committed = NFSWRITE_FILESYNC;
1941 	int wccflag = 0;
1942 	int32_t backup;
1943 	struct nfsrv_descript *nd;
1944 	nfsattrbit_t attrbits;
1945 	uint64_t tmp_off;
1946 	ssize_t tsiz, wsize;
1947 	bool do_append;
1948 
1949 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1950 	*attrflagp = 0;
1951 	tsiz = uiop->uio_resid;
1952 	tmp_off = uiop->uio_offset + tsiz;
1953 	NFSLOCKMNT(nmp);
1954 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1955 		NFSUNLOCKMNT(nmp);
1956 		return (EFBIG);
1957 	}
1958 	wsize = nmp->nm_wsize;
1959 	do_append = false;
1960 	if ((ioflag & IO_APPEND) != 0 && NFSHASNFSV4(nmp) && !NFSHASPNFS(nmp))
1961 		do_append = true;
1962 	NFSUNLOCKMNT(nmp);
1963 	nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK);
1964 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1965 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1966 	while (tsiz > 0) {
1967 		*attrflagp = 0;
1968 		len = (tsiz > wsize) ? wsize : tsiz;
1969 		if (do_append)
1970 			NFSCL_REQSTART(nd, NFSPROC_APPENDWRITE, vp);
1971 		else
1972 			NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1973 		if (nd->nd_flag & ND_NFSV4) {
1974 			if (do_append) {
1975 				NFSZERO_ATTRBIT(&attrbits);
1976 				NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
1977 				nfsrv_putattrbit(nd, &attrbits);
1978 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED +
1979 				    NFSX_HYPER);
1980 				*tl++ = txdr_unsigned(NFSX_HYPER);
1981 				txdr_hyper(uiop->uio_offset, tl); tl += 2;
1982 				*tl = txdr_unsigned(NFSV4OP_WRITE);
1983 			}
1984 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1985 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1986 			txdr_hyper(uiop->uio_offset, tl);
1987 			tl += 2;
1988 			*tl++ = txdr_unsigned(*iomode);
1989 			*tl = txdr_unsigned(len);
1990 		} else if (nd->nd_flag & ND_NFSV3) {
1991 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1992 			txdr_hyper(uiop->uio_offset, tl);
1993 			tl += 2;
1994 			*tl++ = txdr_unsigned(len);
1995 			*tl++ = txdr_unsigned(*iomode);
1996 			*tl = txdr_unsigned(len);
1997 		} else {
1998 			u_int32_t x;
1999 
2000 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2001 			/*
2002 			 * Not sure why someone changed this, since the
2003 			 * RFC clearly states that "beginoffset" and
2004 			 * "totalcount" are ignored, but it wouldn't
2005 			 * surprise me if there's a busted server out there.
2006 			 */
2007 			/* Set both "begin" and "current" to non-garbage. */
2008 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
2009 			*tl++ = x;      /* "begin offset" */
2010 			*tl++ = x;      /* "current offset" */
2011 			x = txdr_unsigned(len);
2012 			*tl++ = x;      /* total to this offset */
2013 			*tl = x;        /* size of this write */
2014 		}
2015 		nfsm_uiombuf(nd, uiop, len);
2016 		/*
2017 		 * Although it is tempting to do a normal Getattr Op in the
2018 		 * NFSv4 compound, the result can be a nearly hung client
2019 		 * system if the Getattr asks for Owner and/or OwnerGroup.
2020 		 * It occurs when the client can't map either the Owner or
2021 		 * Owner_group name in the Getattr reply to a uid/gid. When
2022 		 * there is a cache miss, the kernel does an upcall to the
2023 		 * nfsuserd. Then, it can try and read the local /etc/passwd
2024 		 * or /etc/group file. It can then block in getnewbuf(),
2025 		 * waiting for dirty writes to be pushed to the NFS server.
2026 		 * The only reason this doesn't result in a complete
2027 		 * deadlock, is that the upcall times out and allows
2028 		 * the write to complete. However, progress is so slow
2029 		 * that it might just as well be deadlocked.
2030 		 * As such, we get the rest of the attributes, but not
2031 		 * Owner or Owner_group.
2032 		 * nb: nfscl_loadattrcache() needs to be told that these
2033 		 *     partial attributes from a write rpc are being
2034 		 *     passed in, via a argument flag.
2035 		 */
2036 		if (nd->nd_flag & ND_NFSV4) {
2037 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
2038 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2039 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2040 			(void) nfsrv_putattrbit(nd, &attrbits);
2041 		}
2042 		error = nfscl_request(nd, vp, p, cred);
2043 		if (error) {
2044 			free(nd, M_TEMP);
2045 			return (error);
2046 		}
2047 		if (nd->nd_repstat) {
2048 			/*
2049 			 * In case the rpc gets retried, roll
2050 			 * the uio fields changed by nfsm_uiombuf()
2051 			 * back.
2052 			 */
2053 			uiop->uio_offset -= len;
2054 			uiop->uio_resid += len;
2055 			uiop->uio_iov->iov_base =
2056 			    (char *)uiop->uio_iov->iov_base - len;
2057 			uiop->uio_iov->iov_len += len;
2058 		}
2059 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2060 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
2061 			    &wccflag, &tmp_off);
2062 			if (error)
2063 				goto nfsmout;
2064 		}
2065 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2066 		    (ND_NFSV4 | ND_NOMOREDATA) &&
2067 		    nd->nd_repstat == NFSERR_NOTSAME && do_append) {
2068 			/*
2069 			 * Verify of the file's size failed, so redo the
2070 			 * write using the file's size as returned in
2071 			 * the wcc attributes.
2072 			 */
2073 			if (tmp_off + tsiz <= nmp->nm_maxfilesize) {
2074 				do_append = false;
2075 				uiop->uio_offset = tmp_off;
2076 				m_freem(nd->nd_mrep);
2077 				nd->nd_mrep = NULL;
2078 				continue;
2079 			} else
2080 				nd->nd_repstat = EFBIG;
2081 		}
2082 		if (!nd->nd_repstat) {
2083 			if (do_append) {
2084 				/* Strip off the Write reply status. */
2085 				do_append = false;
2086 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2087 			}
2088 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2089 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
2090 					+ NFSX_VERF);
2091 				rlen = fxdr_unsigned(int, *tl++);
2092 				if (rlen == 0) {
2093 					error = NFSERR_IO;
2094 					goto nfsmout;
2095 				} else if (rlen < len) {
2096 					backup = len - rlen;
2097 					uiop->uio_iov->iov_base =
2098 					    (char *)uiop->uio_iov->iov_base -
2099 					    backup;
2100 					uiop->uio_iov->iov_len += backup;
2101 					uiop->uio_offset -= backup;
2102 					uiop->uio_resid += backup;
2103 					len = rlen;
2104 				}
2105 				commit = fxdr_unsigned(int, *tl++);
2106 
2107 				/*
2108 				 * Return the lowest commitment level
2109 				 * obtained by any of the RPCs.
2110 				 */
2111 				if (committed == NFSWRITE_FILESYNC)
2112 					committed = commit;
2113 				else if (committed == NFSWRITE_DATASYNC &&
2114 					commit == NFSWRITE_UNSTABLE)
2115 					committed = commit;
2116 				NFSLOCKMNT(nmp);
2117 				if (!NFSHASWRITEVERF(nmp)) {
2118 					NFSBCOPY((caddr_t)tl,
2119 					    (caddr_t)&nmp->nm_verf[0],
2120 					    NFSX_VERF);
2121 					NFSSETWRITEVERF(nmp);
2122 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
2123 				    NFSX_VERF) && *must_commit != 2) {
2124 					*must_commit = 1;
2125 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
2126 				}
2127 				NFSUNLOCKMNT(nmp);
2128 			}
2129 			if (nd->nd_flag & ND_NFSV4)
2130 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2131 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
2132 				error = nfsm_loadattr(nd, nap);
2133 				if (!error)
2134 					*attrflagp = NFS_LATTR_NOSHRINK;
2135 			}
2136 		} else {
2137 			error = nd->nd_repstat;
2138 		}
2139 		if (error)
2140 			goto nfsmout;
2141 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
2142 		m_freem(nd->nd_mrep);
2143 		nd->nd_mrep = NULL;
2144 		tsiz -= len;
2145 	}
2146 nfsmout:
2147 	if (nd->nd_mrep != NULL)
2148 		m_freem(nd->nd_mrep);
2149 	*iomode = committed;
2150 	if (nd->nd_repstat && !error)
2151 		error = nd->nd_repstat;
2152 	free(nd, M_TEMP);
2153 	return (error);
2154 }
2155 
2156 /*
2157  * Do an nfs deallocate operation.
2158  */
2159 int
2160 nfsrpc_deallocate(vnode_t vp, off_t offs, off_t len, struct nfsvattr *nap,
2161     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
2162 {
2163 	int error, expireret = 0, openerr, retrycnt;
2164 	uint32_t clidrev = 0;
2165 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2166 	struct nfsfh *nfhp;
2167 	nfsv4stateid_t stateid;
2168 	void *lckp;
2169 
2170 	if (nmp->nm_clp != NULL)
2171 		clidrev = nmp->nm_clp->nfsc_clientidrev;
2172 	retrycnt = 0;
2173 	do {
2174 		lckp = NULL;
2175 		openerr = 1;
2176 		nfhp = VTONFS(vp)->n_fhp;
2177 		error = nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
2178 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
2179 		if (error != 0) {
2180 			/*
2181 			 * No Open stateid, so try and open the file
2182 			 * now.
2183 			 */
2184 			openerr = nfsrpc_open(vp, FWRITE, cred, p);
2185 			if (openerr == 0)
2186 				nfscl_getstateid(vp, nfhp->nfh_fh,
2187 				    nfhp->nfh_len, NFSV4OPEN_ACCESSWRITE, 0,
2188 				    cred, p, &stateid, &lckp);
2189 		}
2190 		error = nfsrpc_deallocaterpc(vp, offs, len, &stateid, nap,
2191 		    attrflagp, cred, p);
2192 		if (error == NFSERR_STALESTATEID)
2193 			nfscl_initiate_recovery(nmp->nm_clp);
2194 		if (lckp != NULL)
2195 			nfscl_lockderef(lckp);
2196 		if (openerr == 0)
2197 			nfsrpc_close(vp, 0, p);
2198 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2199 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2200 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
2201 			(void) nfs_catnap(PZERO, error, "nfs_deallocate");
2202 		} else if ((error == NFSERR_EXPIRED ||
2203 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2204 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2205 		}
2206 		retrycnt++;
2207 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2208 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2209 	    error == NFSERR_BADSESSION ||
2210 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
2211 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2212 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
2213 	if (error && retrycnt >= 4)
2214 		error = EIO;
2215 	return (error);
2216 }
2217 
2218 /*
2219  * The actual deallocate RPC.
2220  */
2221 static int
2222 nfsrpc_deallocaterpc(vnode_t vp, off_t offs, off_t len,
2223     nfsv4stateid_t *stateidp, struct nfsvattr *nap, int *attrflagp,
2224     struct ucred *cred, NFSPROC_T *p)
2225 {
2226 	uint32_t *tl;
2227 	struct nfsnode *np = VTONFS(vp);
2228 	int error, wccflag;
2229 	struct nfsrv_descript nfsd;
2230 	struct nfsrv_descript *nd = &nfsd;
2231 	nfsattrbit_t attrbits;
2232 
2233 	*attrflagp = 0;
2234 	NFSCL_REQSTART(nd, NFSPROC_DEALLOCATE, vp);
2235 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2236 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
2237 	txdr_hyper(offs, tl);
2238 	tl += 2;
2239 	txdr_hyper(len, tl);
2240 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
2241 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
2242 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2243 	nfsrv_putattrbit(nd, &attrbits);
2244 	error = nfscl_request(nd, vp, p, cred);
2245 	if (error != 0)
2246 		return (error);
2247 	wccflag = 0;
2248 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, &wccflag, NULL);
2249 	if (error != 0)
2250 		goto nfsmout;
2251 	if (nd->nd_repstat == 0) {
2252 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2253 		error = nfsm_loadattr(nd, nap);
2254 		if (error != 0)
2255 			goto nfsmout;
2256 		*attrflagp = NFS_LATTR_NOSHRINK;
2257 	}
2258 	NFSWRITERPC_SETTIME(wccflag, np, nap, 1);
2259 nfsmout:
2260 	m_freem(nd->nd_mrep);
2261 	if (nd->nd_repstat != 0 && error == 0)
2262 		error = nd->nd_repstat;
2263 	return (error);
2264 }
2265 
2266 /*
2267  * nfs mknod rpc
2268  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
2269  * mode set to specify the file type and the size field for rdev.
2270  */
2271 int
2272 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2273     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
2274     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2275     int *attrflagp, int *dattrflagp, void *dstuff)
2276 {
2277 	u_int32_t *tl;
2278 	int error = 0;
2279 	struct nfsrv_descript nfsd, *nd = &nfsd;
2280 	nfsattrbit_t attrbits;
2281 
2282 	*nfhpp = NULL;
2283 	*attrflagp = 0;
2284 	*dattrflagp = 0;
2285 	if (namelen > NFS_MAXNAMLEN)
2286 		return (ENAMETOOLONG);
2287 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
2288 	if (nd->nd_flag & ND_NFSV4) {
2289 		if (vtyp == VBLK || vtyp == VCHR) {
2290 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
2291 			*tl++ = vtonfsv34_type(vtyp);
2292 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2293 			*tl = txdr_unsigned(NFSMINOR(rdev));
2294 		} else {
2295 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2296 			*tl = vtonfsv34_type(vtyp);
2297 		}
2298 	}
2299 	(void) nfsm_strtom(nd, name, namelen);
2300 	if (nd->nd_flag & ND_NFSV3) {
2301 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2302 		*tl = vtonfsv34_type(vtyp);
2303 	}
2304 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2305 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2306 	if ((nd->nd_flag & ND_NFSV3) &&
2307 	    (vtyp == VCHR || vtyp == VBLK)) {
2308 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2309 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2310 		*tl = txdr_unsigned(NFSMINOR(rdev));
2311 	}
2312 	if (nd->nd_flag & ND_NFSV4) {
2313 		NFSGETATTR_ATTRBIT(&attrbits);
2314 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2315 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2316 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2317 		(void) nfsrv_putattrbit(nd, &attrbits);
2318 	}
2319 	if (nd->nd_flag & ND_NFSV2)
2320 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2321 	error = nfscl_request(nd, dvp, p, cred);
2322 	if (error)
2323 		return (error);
2324 	if (nd->nd_flag & ND_NFSV4)
2325 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2326 	if (!nd->nd_repstat) {
2327 		if (nd->nd_flag & ND_NFSV4) {
2328 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2329 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2330 			if (error)
2331 				goto nfsmout;
2332 		}
2333 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2334 		if (error)
2335 			goto nfsmout;
2336 	}
2337 	if (nd->nd_flag & ND_NFSV3)
2338 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2339 	if (!error && nd->nd_repstat)
2340 		error = nd->nd_repstat;
2341 nfsmout:
2342 	m_freem(nd->nd_mrep);
2343 	return (error);
2344 }
2345 
2346 /*
2347  * nfs file create call
2348  * Mostly just call the approriate routine. (I separated out v4, so that
2349  * error recovery wouldn't be as difficult.)
2350  */
2351 int
2352 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2353     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2354     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2355     int *attrflagp, int *dattrflagp, void *dstuff)
2356 {
2357 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2358 	struct nfsclowner *owp;
2359 	struct nfscldeleg *dp;
2360 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2361 	u_int32_t clidrev;
2362 
2363 	if (NFSHASNFSV4(nmp)) {
2364 	    retrycnt = 0;
2365 	    do {
2366 		dp = NULL;
2367 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2368 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2369 		    NULL, 1, true);
2370 		if (error)
2371 			return (error);
2372 		if (nmp->nm_clp != NULL)
2373 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2374 		else
2375 			clidrev = 0;
2376 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2377 		    nfs_numnfscbd == 0 || retrycnt > 0)
2378 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2379 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2380 			  attrflagp, dattrflagp, &unlocked);
2381 		else
2382 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2383 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2384 			  attrflagp, dattrflagp, &unlocked);
2385 		/*
2386 		 * There is no need to invalidate cached attributes here,
2387 		 * since new post-delegation issue attributes are always
2388 		 * returned by nfsrpc_createv4() and these will update the
2389 		 * attribute cache.
2390 		 */
2391 		if (dp != NULL)
2392 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2393 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2394 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2395 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2396 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2397 		    error == NFSERR_BADSESSION) {
2398 			(void) nfs_catnap(PZERO, error, "nfs_open");
2399 		} else if ((error == NFSERR_EXPIRED ||
2400 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2401 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2402 			retrycnt++;
2403 		}
2404 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2405 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2406 		error == NFSERR_BADSESSION ||
2407 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2408 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2409 	    if (error && retrycnt >= 4)
2410 		    error = EIO;
2411 	} else {
2412 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2413 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp);
2414 	}
2415 	return (error);
2416 }
2417 
2418 /*
2419  * The create rpc for v2 and 3.
2420  */
2421 static int
2422 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2423     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2424     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2425     int *attrflagp, int *dattrflagp)
2426 {
2427 	u_int32_t *tl;
2428 	int error = 0;
2429 	struct nfsrv_descript nfsd, *nd = &nfsd;
2430 
2431 	*nfhpp = NULL;
2432 	*attrflagp = 0;
2433 	*dattrflagp = 0;
2434 	if (namelen > NFS_MAXNAMLEN)
2435 		return (ENAMETOOLONG);
2436 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2437 	(void) nfsm_strtom(nd, name, namelen);
2438 	if (nd->nd_flag & ND_NFSV3) {
2439 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2440 		if (fmode & O_EXCL) {
2441 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2442 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2443 			*tl++ = cverf.lval[0];
2444 			*tl = cverf.lval[1];
2445 		} else {
2446 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2447 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2448 		}
2449 	} else {
2450 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2451 	}
2452 	error = nfscl_request(nd, dvp, p, cred);
2453 	if (error)
2454 		return (error);
2455 	if (nd->nd_repstat == 0) {
2456 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2457 		if (error)
2458 			goto nfsmout;
2459 	}
2460 	if (nd->nd_flag & ND_NFSV3)
2461 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2462 	if (nd->nd_repstat != 0 && error == 0)
2463 		error = nd->nd_repstat;
2464 nfsmout:
2465 	m_freem(nd->nd_mrep);
2466 	return (error);
2467 }
2468 
2469 static int
2470 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2471     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2472     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2473     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2474     int *dattrflagp, int *unlockedp)
2475 {
2476 	u_int32_t *tl;
2477 	int error = 0, deleg, newone, ret, acesize, limitby;
2478 	struct nfsrv_descript nfsd, *nd = &nfsd;
2479 	struct nfsclopen *op;
2480 	struct nfscldeleg *dp = NULL;
2481 	struct nfsnode *np;
2482 	struct nfsfh *nfhp;
2483 	nfsattrbit_t attrbits;
2484 	nfsv4stateid_t stateid;
2485 	u_int32_t rflags;
2486 	struct nfsmount *nmp;
2487 	struct nfsclsession *tsep;
2488 
2489 	nmp = VFSTONFS(dvp->v_mount);
2490 	np = VTONFS(dvp);
2491 	*unlockedp = 0;
2492 	*nfhpp = NULL;
2493 	*dpp = NULL;
2494 	*attrflagp = 0;
2495 	*dattrflagp = 0;
2496 	if (namelen > NFS_MAXNAMLEN)
2497 		return (ENAMETOOLONG);
2498 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2499 	/*
2500 	 * For V4, this is actually an Open op.
2501 	 */
2502 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2503 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2504 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2505 	    NFSV4OPEN_ACCESSREAD);
2506 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2507 	tsep = nfsmnt_mdssession(nmp);
2508 	*tl++ = tsep->nfsess_clientid.lval[0];
2509 	*tl = tsep->nfsess_clientid.lval[1];
2510 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2511 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2512 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2513 	if (fmode & O_EXCL) {
2514 		if (NFSHASNFSV4N(nmp)) {
2515 			if (NFSHASSESSPERSIST(nmp)) {
2516 				/* Use GUARDED for persistent sessions. */
2517 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2518 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2519 			} else {
2520 				/* Otherwise, use EXCLUSIVE4_1. */
2521 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2522 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2523 				*tl++ = cverf.lval[0];
2524 				*tl = cverf.lval[1];
2525 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2526 			}
2527 		} else {
2528 			/* NFSv4.0 */
2529 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2530 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2531 			*tl++ = cverf.lval[0];
2532 			*tl = cverf.lval[1];
2533 		}
2534 	} else {
2535 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2536 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2537 	}
2538 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2539 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2540 	(void) nfsm_strtom(nd, name, namelen);
2541 	/* Get the new file's handle and attributes. */
2542 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2543 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2544 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2545 	NFSGETATTR_ATTRBIT(&attrbits);
2546 	(void) nfsrv_putattrbit(nd, &attrbits);
2547 	/* Get the directory's post-op attributes. */
2548 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2549 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2550 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2551 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2552 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2553 	(void) nfsrv_putattrbit(nd, &attrbits);
2554 	error = nfscl_request(nd, dvp, p, cred);
2555 	if (error)
2556 		return (error);
2557 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2558 	if (nd->nd_repstat == 0) {
2559 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2560 		    6 * NFSX_UNSIGNED);
2561 		stateid.seqid = *tl++;
2562 		stateid.other[0] = *tl++;
2563 		stateid.other[1] = *tl++;
2564 		stateid.other[2] = *tl;
2565 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2566 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2567 		if (error)
2568 			goto nfsmout;
2569 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2570 		deleg = fxdr_unsigned(int, *tl);
2571 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2572 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2573 			if (!(owp->nfsow_clp->nfsc_flags &
2574 			      NFSCLFLAGS_FIRSTDELEG))
2575 				owp->nfsow_clp->nfsc_flags |=
2576 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2577 			dp = malloc(
2578 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2579 			    M_NFSCLDELEG, M_WAITOK);
2580 			LIST_INIT(&dp->nfsdl_owner);
2581 			LIST_INIT(&dp->nfsdl_lock);
2582 			dp->nfsdl_clp = owp->nfsow_clp;
2583 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2584 			nfscl_lockinit(&dp->nfsdl_rwlock);
2585 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2586 			    NFSX_UNSIGNED);
2587 			dp->nfsdl_stateid.seqid = *tl++;
2588 			dp->nfsdl_stateid.other[0] = *tl++;
2589 			dp->nfsdl_stateid.other[1] = *tl++;
2590 			dp->nfsdl_stateid.other[2] = *tl++;
2591 			ret = fxdr_unsigned(int, *tl);
2592 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2593 				dp->nfsdl_flags = NFSCLDL_WRITE;
2594 				/*
2595 				 * Indicates how much the file can grow.
2596 				 */
2597 				NFSM_DISSECT(tl, u_int32_t *,
2598 				    3 * NFSX_UNSIGNED);
2599 				limitby = fxdr_unsigned(int, *tl++);
2600 				switch (limitby) {
2601 				case NFSV4OPEN_LIMITSIZE:
2602 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2603 					break;
2604 				case NFSV4OPEN_LIMITBLOCKS:
2605 					dp->nfsdl_sizelimit =
2606 					    fxdr_unsigned(u_int64_t, *tl++);
2607 					dp->nfsdl_sizelimit *=
2608 					    fxdr_unsigned(u_int64_t, *tl);
2609 					break;
2610 				default:
2611 					error = NFSERR_BADXDR;
2612 					goto nfsmout;
2613 				}
2614 			} else {
2615 				dp->nfsdl_flags = NFSCLDL_READ;
2616 			}
2617 			if (ret)
2618 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2619 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
2620 			    &ret, &acesize, p);
2621 			if (error)
2622 				goto nfsmout;
2623 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2624 			error = NFSERR_BADXDR;
2625 			goto nfsmout;
2626 		}
2627 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2628 		if (error)
2629 			goto nfsmout;
2630 		/* Get rid of the PutFH and Getattr status values. */
2631 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2632 		/* Load the directory attributes. */
2633 		error = nfsm_loadattr(nd, dnap);
2634 		if (error)
2635 			goto nfsmout;
2636 		*dattrflagp = 1;
2637 		if (dp != NULL && *attrflagp) {
2638 			dp->nfsdl_change = nnap->na_filerev;
2639 			dp->nfsdl_modtime = nnap->na_mtime;
2640 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2641 		}
2642 		/*
2643 		 * We can now complete the Open state.
2644 		 */
2645 		nfhp = *nfhpp;
2646 		if (dp != NULL) {
2647 			dp->nfsdl_fhlen = nfhp->nfh_len;
2648 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2649 		}
2650 		/*
2651 		 * Get an Open structure that will be
2652 		 * attached to the OpenOwner, acquired already.
2653 		 */
2654 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2655 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2656 		    cred, p, NULL, &op, &newone, NULL, 0, false);
2657 		if (error)
2658 			goto nfsmout;
2659 		op->nfso_stateid = stateid;
2660 		newnfs_copyincred(cred, &op->nfso_cred);
2661 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2662 		    do {
2663 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2664 			    nfhp->nfh_len, op, cred, p);
2665 			if (ret == NFSERR_DELAY)
2666 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2667 		    } while (ret == NFSERR_DELAY);
2668 		    error = ret;
2669 		}
2670 
2671 		/*
2672 		 * If the server is handing out delegations, but we didn't
2673 		 * get one because an OpenConfirm was required, try the
2674 		 * Open again, to get a delegation. This is a harmless no-op,
2675 		 * from a server's point of view.
2676 		 */
2677 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2678 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2679 		    !error && dp == NULL) {
2680 		    do {
2681 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2682 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2683 			    nfhp->nfh_fh, nfhp->nfh_len,
2684 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2685 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2686 			if (ret == NFSERR_DELAY)
2687 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2688 		    } while (ret == NFSERR_DELAY);
2689 		    if (ret) {
2690 			if (dp != NULL) {
2691 				free(dp, M_NFSCLDELEG);
2692 				dp = NULL;
2693 			}
2694 			if (ret == NFSERR_STALECLIENTID ||
2695 			    ret == NFSERR_STALEDONTRECOVER ||
2696 			    ret == NFSERR_BADSESSION)
2697 				error = ret;
2698 		    }
2699 		}
2700 		nfscl_openrelease(nmp, op, error, newone);
2701 		*unlockedp = 1;
2702 	}
2703 	if (nd->nd_repstat != 0 && error == 0)
2704 		error = nd->nd_repstat;
2705 	if (error == NFSERR_STALECLIENTID)
2706 		nfscl_initiate_recovery(owp->nfsow_clp);
2707 nfsmout:
2708 	if (!error)
2709 		*dpp = dp;
2710 	else if (dp != NULL)
2711 		free(dp, M_NFSCLDELEG);
2712 	m_freem(nd->nd_mrep);
2713 	return (error);
2714 }
2715 
2716 /*
2717  * Nfs remove rpc
2718  */
2719 int
2720 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2721     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2722     void *dstuff)
2723 {
2724 	u_int32_t *tl;
2725 	struct nfsrv_descript nfsd, *nd = &nfsd;
2726 	struct nfsnode *np;
2727 	struct nfsmount *nmp;
2728 	nfsv4stateid_t dstateid;
2729 	int error, ret = 0, i;
2730 
2731 	*dattrflagp = 0;
2732 	if (namelen > NFS_MAXNAMLEN)
2733 		return (ENAMETOOLONG);
2734 	nmp = VFSTONFS(dvp->v_mount);
2735 tryagain:
2736 	if (NFSHASNFSV4(nmp) && ret == 0) {
2737 		ret = nfscl_removedeleg(vp, p, &dstateid);
2738 		if (ret == 1) {
2739 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2740 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2741 			    NFSX_UNSIGNED);
2742 			if (NFSHASNFSV4N(nmp))
2743 				*tl++ = 0;
2744 			else
2745 				*tl++ = dstateid.seqid;
2746 			*tl++ = dstateid.other[0];
2747 			*tl++ = dstateid.other[1];
2748 			*tl++ = dstateid.other[2];
2749 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2750 			np = VTONFS(dvp);
2751 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2752 			    np->n_fhp->nfh_len, 0);
2753 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2754 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2755 		}
2756 	} else {
2757 		ret = 0;
2758 	}
2759 	if (ret == 0)
2760 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2761 	(void) nfsm_strtom(nd, name, namelen);
2762 	error = nfscl_request(nd, dvp, p, cred);
2763 	if (error)
2764 		return (error);
2765 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2766 		/* For NFSv4, parse out any Delereturn replies. */
2767 		if (ret > 0 && nd->nd_repstat != 0 &&
2768 		    (nd->nd_flag & ND_NOMOREDATA)) {
2769 			/*
2770 			 * If the Delegreturn failed, try again without
2771 			 * it. The server will Recall, as required.
2772 			 */
2773 			m_freem(nd->nd_mrep);
2774 			goto tryagain;
2775 		}
2776 		for (i = 0; i < (ret * 2); i++) {
2777 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2778 			    ND_NFSV4) {
2779 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2780 			    if (*(tl + 1))
2781 				nd->nd_flag |= ND_NOMOREDATA;
2782 			}
2783 		}
2784 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2785 	}
2786 	if (nd->nd_repstat && !error)
2787 		error = nd->nd_repstat;
2788 nfsmout:
2789 	m_freem(nd->nd_mrep);
2790 	return (error);
2791 }
2792 
2793 /*
2794  * Do an nfs rename rpc.
2795  */
2796 int
2797 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2798     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2799     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2800     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2801 {
2802 	u_int32_t *tl;
2803 	struct nfsrv_descript nfsd, *nd = &nfsd;
2804 	struct nfsmount *nmp;
2805 	struct nfsnode *np;
2806 	nfsattrbit_t attrbits;
2807 	nfsv4stateid_t fdstateid, tdstateid;
2808 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2809 
2810 	*fattrflagp = 0;
2811 	*tattrflagp = 0;
2812 	nmp = VFSTONFS(fdvp->v_mount);
2813 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2814 		return (ENAMETOOLONG);
2815 tryagain:
2816 	if (NFSHASNFSV4(nmp) && ret == 0) {
2817 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2818 		    &tdstateid, &gottd, p);
2819 		if (gotfd && gottd) {
2820 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2821 		} else if (gotfd) {
2822 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2823 		} else if (gottd) {
2824 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2825 		}
2826 		if (gotfd) {
2827 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2828 			if (NFSHASNFSV4N(nmp))
2829 				*tl++ = 0;
2830 			else
2831 				*tl++ = fdstateid.seqid;
2832 			*tl++ = fdstateid.other[0];
2833 			*tl++ = fdstateid.other[1];
2834 			*tl = fdstateid.other[2];
2835 			if (gottd) {
2836 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2837 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2838 				np = VTONFS(tvp);
2839 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2840 				    np->n_fhp->nfh_len, 0);
2841 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2842 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2843 			}
2844 		}
2845 		if (gottd) {
2846 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2847 			if (NFSHASNFSV4N(nmp))
2848 				*tl++ = 0;
2849 			else
2850 				*tl++ = tdstateid.seqid;
2851 			*tl++ = tdstateid.other[0];
2852 			*tl++ = tdstateid.other[1];
2853 			*tl = tdstateid.other[2];
2854 		}
2855 		if (ret > 0) {
2856 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2857 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2858 			np = VTONFS(fdvp);
2859 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2860 			    np->n_fhp->nfh_len, 0);
2861 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2862 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2863 		}
2864 	} else {
2865 		ret = 0;
2866 	}
2867 	if (ret == 0)
2868 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2869 	if (nd->nd_flag & ND_NFSV4) {
2870 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2871 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2872 		NFSWCCATTR_ATTRBIT(&attrbits);
2873 		(void) nfsrv_putattrbit(nd, &attrbits);
2874 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2875 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2876 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2877 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2878 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2879 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2880 		(void) nfsrv_putattrbit(nd, &attrbits);
2881 		nd->nd_flag |= ND_V4WCCATTR;
2882 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2883 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2884 	}
2885 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2886 	if (!(nd->nd_flag & ND_NFSV4))
2887 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2888 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2889 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2890 	error = nfscl_request(nd, fdvp, p, cred);
2891 	if (error)
2892 		return (error);
2893 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2894 		/* For NFSv4, parse out any Delereturn replies. */
2895 		if (ret > 0 && nd->nd_repstat != 0 &&
2896 		    (nd->nd_flag & ND_NOMOREDATA)) {
2897 			/*
2898 			 * If the Delegreturn failed, try again without
2899 			 * it. The server will Recall, as required.
2900 			 */
2901 			m_freem(nd->nd_mrep);
2902 			goto tryagain;
2903 		}
2904 		for (i = 0; i < (ret * 2); i++) {
2905 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2906 			    ND_NFSV4) {
2907 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2908 			    if (*(tl + 1)) {
2909 				if (i == 0 && ret > 1) {
2910 				    /*
2911 				     * If the Delegreturn failed, try again
2912 				     * without it. The server will Recall, as
2913 				     * required.
2914 				     * If ret > 1, the first iteration of this
2915 				     * loop is the second DelegReturn result.
2916 				     */
2917 				    m_freem(nd->nd_mrep);
2918 				    goto tryagain;
2919 				} else {
2920 				    nd->nd_flag |= ND_NOMOREDATA;
2921 				}
2922 			    }
2923 			}
2924 		}
2925 		/* Now, the first wcc attribute reply. */
2926 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2927 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2928 			if (*(tl + 1))
2929 				nd->nd_flag |= ND_NOMOREDATA;
2930 		}
2931 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL, NULL);
2932 		/* and the second wcc attribute reply. */
2933 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2934 		    !error) {
2935 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2936 			if (*(tl + 1))
2937 				nd->nd_flag |= ND_NOMOREDATA;
2938 		}
2939 		if (!error)
2940 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2941 			    NULL, NULL);
2942 	}
2943 	if (nd->nd_repstat && !error)
2944 		error = nd->nd_repstat;
2945 nfsmout:
2946 	m_freem(nd->nd_mrep);
2947 	return (error);
2948 }
2949 
2950 /*
2951  * nfs hard link create rpc
2952  */
2953 int
2954 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2955     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2956     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2957 {
2958 	u_int32_t *tl;
2959 	struct nfsrv_descript nfsd, *nd = &nfsd;
2960 	nfsattrbit_t attrbits;
2961 	int error = 0;
2962 
2963 	*attrflagp = 0;
2964 	*dattrflagp = 0;
2965 	if (namelen > NFS_MAXNAMLEN)
2966 		return (ENAMETOOLONG);
2967 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2968 	if (nd->nd_flag & ND_NFSV4) {
2969 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2970 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2971 	}
2972 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2973 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2974 	if (nd->nd_flag & ND_NFSV4) {
2975 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2976 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2977 		NFSWCCATTR_ATTRBIT(&attrbits);
2978 		(void) nfsrv_putattrbit(nd, &attrbits);
2979 		nd->nd_flag |= ND_V4WCCATTR;
2980 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2981 		*tl = txdr_unsigned(NFSV4OP_LINK);
2982 	}
2983 	(void) nfsm_strtom(nd, name, namelen);
2984 	error = nfscl_request(nd, vp, p, cred);
2985 	if (error)
2986 		return (error);
2987 	if (nd->nd_flag & ND_NFSV3) {
2988 		error = nfscl_postop_attr(nd, nap, attrflagp);
2989 		if (!error)
2990 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2991 			    NULL, NULL);
2992 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2993 		/*
2994 		 * First, parse out the PutFH and Getattr result.
2995 		 */
2996 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2997 		if (!(*(tl + 1)))
2998 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2999 		if (*(tl + 1))
3000 			nd->nd_flag |= ND_NOMOREDATA;
3001 		/*
3002 		 * Get the pre-op attributes.
3003 		 */
3004 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3005 	}
3006 	if (nd->nd_repstat && !error)
3007 		error = nd->nd_repstat;
3008 nfsmout:
3009 	m_freem(nd->nd_mrep);
3010 	return (error);
3011 }
3012 
3013 /*
3014  * nfs symbolic link create rpc
3015  */
3016 int
3017 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
3018     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3019     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3020     int *dattrflagp, void *dstuff)
3021 {
3022 	u_int32_t *tl;
3023 	struct nfsrv_descript nfsd, *nd = &nfsd;
3024 	struct nfsmount *nmp;
3025 	int slen, error = 0;
3026 
3027 	*nfhpp = NULL;
3028 	*attrflagp = 0;
3029 	*dattrflagp = 0;
3030 	nmp = VFSTONFS(dvp->v_mount);
3031 	slen = strlen(target);
3032 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
3033 		return (ENAMETOOLONG);
3034 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
3035 	if (nd->nd_flag & ND_NFSV4) {
3036 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3037 		*tl = txdr_unsigned(NFLNK);
3038 		(void) nfsm_strtom(nd, target, slen);
3039 	}
3040 	(void) nfsm_strtom(nd, name, namelen);
3041 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3042 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
3043 	if (!(nd->nd_flag & ND_NFSV4))
3044 		(void) nfsm_strtom(nd, target, slen);
3045 	if (nd->nd_flag & ND_NFSV2)
3046 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3047 	error = nfscl_request(nd, dvp, p, cred);
3048 	if (error)
3049 		return (error);
3050 	if (nd->nd_flag & ND_NFSV4)
3051 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3052 	if ((nd->nd_flag & ND_NFSV3) && !error) {
3053 		if (!nd->nd_repstat)
3054 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3055 		if (!error)
3056 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3057 			    NULL, NULL);
3058 	}
3059 	if (nd->nd_repstat && !error)
3060 		error = nd->nd_repstat;
3061 	m_freem(nd->nd_mrep);
3062 	/*
3063 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3064 	 * Only do this if vfs.nfs.ignore_eexist is set.
3065 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3066 	 * should guarantee "exactly once" RPC semantics.
3067 	 */
3068 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3069 	    nmp->nm_minorvers == 0))
3070 		error = 0;
3071 	return (error);
3072 }
3073 
3074 /*
3075  * nfs make dir rpc
3076  */
3077 int
3078 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
3079     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3080     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3081     int *dattrflagp, void *dstuff)
3082 {
3083 	u_int32_t *tl;
3084 	struct nfsrv_descript nfsd, *nd = &nfsd;
3085 	nfsattrbit_t attrbits;
3086 	int error = 0;
3087 	struct nfsfh *fhp;
3088 	struct nfsmount *nmp;
3089 
3090 	*nfhpp = NULL;
3091 	*attrflagp = 0;
3092 	*dattrflagp = 0;
3093 	nmp = VFSTONFS(dvp->v_mount);
3094 	fhp = VTONFS(dvp)->n_fhp;
3095 	if (namelen > NFS_MAXNAMLEN)
3096 		return (ENAMETOOLONG);
3097 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
3098 	if (nd->nd_flag & ND_NFSV4) {
3099 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3100 		*tl = txdr_unsigned(NFDIR);
3101 	}
3102 	(void) nfsm_strtom(nd, name, namelen);
3103 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3104 	if (nd->nd_flag & ND_NFSV4) {
3105 		NFSGETATTR_ATTRBIT(&attrbits);
3106 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3107 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3108 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3109 		(void) nfsrv_putattrbit(nd, &attrbits);
3110 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3111 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3112 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
3113 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3114 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3115 		(void) nfsrv_putattrbit(nd, &attrbits);
3116 	}
3117 	error = nfscl_request(nd, dvp, p, cred);
3118 	if (error)
3119 		return (error);
3120 	if (nd->nd_flag & ND_NFSV4)
3121 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3122 	if (!nd->nd_repstat && !error) {
3123 		if (nd->nd_flag & ND_NFSV4) {
3124 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3125 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
3126 		}
3127 		if (!error)
3128 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3129 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
3130 			/* Get rid of the PutFH and Getattr status values. */
3131 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3132 			/* Load the directory attributes. */
3133 			error = nfsm_loadattr(nd, dnap);
3134 			if (error == 0)
3135 				*dattrflagp = 1;
3136 		}
3137 	}
3138 	if ((nd->nd_flag & ND_NFSV3) && !error)
3139 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3140 	if (nd->nd_repstat && !error)
3141 		error = nd->nd_repstat;
3142 nfsmout:
3143 	m_freem(nd->nd_mrep);
3144 	/*
3145 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3146 	 * Only do this if vfs.nfs.ignore_eexist is set.
3147 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3148 	 * should guarantee "exactly once" RPC semantics.
3149 	 */
3150 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3151 	    nmp->nm_minorvers == 0))
3152 		error = 0;
3153 	return (error);
3154 }
3155 
3156 /*
3157  * nfs remove directory call
3158  */
3159 int
3160 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
3161     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
3162 {
3163 	struct nfsrv_descript nfsd, *nd = &nfsd;
3164 	int error = 0;
3165 
3166 	*dattrflagp = 0;
3167 	if (namelen > NFS_MAXNAMLEN)
3168 		return (ENAMETOOLONG);
3169 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
3170 	(void) nfsm_strtom(nd, name, namelen);
3171 	error = nfscl_request(nd, dvp, p, cred);
3172 	if (error)
3173 		return (error);
3174 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3175 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3176 	if (nd->nd_repstat && !error)
3177 		error = nd->nd_repstat;
3178 	m_freem(nd->nd_mrep);
3179 	/*
3180 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
3181 	 */
3182 	if (error == ENOENT)
3183 		error = 0;
3184 	return (error);
3185 }
3186 
3187 /*
3188  * Readdir rpc.
3189  * Always returns with either uio_resid unchanged, if you are at the
3190  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
3191  * filled in.
3192  * I felt this would allow caching of directory blocks more easily
3193  * than returning a pertially filled block.
3194  * Directory offset cookies:
3195  * Oh my, what to do with them...
3196  * I can think of three ways to deal with them:
3197  * 1 - have the layer above these RPCs maintain a map between logical
3198  *     directory byte offsets and the NFS directory offset cookies
3199  * 2 - pass the opaque directory offset cookies up into userland
3200  *     and let the libc functions deal with them, via the system call
3201  * 3 - return them to userland in the "struct dirent", so future versions
3202  *     of libc can use them and do whatever is necessary to make things work
3203  *     above these rpc calls, in the meantime
3204  * For now, I do #3 by "hiding" the directory offset cookies after the
3205  * d_name field in struct dirent. This is space inside d_reclen that
3206  * will be ignored by anything that doesn't know about them.
3207  * The directory offset cookies are filled in as the last 8 bytes of
3208  * each directory entry, after d_name. Someday, the userland libc
3209  * functions may be able to use these. In the meantime, it satisfies
3210  * OpenBSD's requirements for cookies being returned.
3211  * If expects the directory offset cookie for the read to be in uio_offset
3212  * and returns the one for the next entry after this directory block in
3213  * there, as well.
3214  */
3215 int
3216 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3217     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3218     int *eofp, void *stuff)
3219 {
3220 	int len, left;
3221 	struct dirent *dp = NULL;
3222 	u_int32_t *tl;
3223 	nfsquad_t cookie, ncookie;
3224 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3225 	struct nfsnode *dnp = VTONFS(vp);
3226 	struct nfsvattr nfsva;
3227 	struct nfsrv_descript nfsd, *nd = &nfsd;
3228 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3229 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
3230 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3231 	char *cp;
3232 	nfsattrbit_t attrbits, dattrbits;
3233 	u_int32_t rderr, *tl2 = NULL;
3234 	size_t tresid;
3235 
3236 	KASSERT(uiop->uio_iovcnt == 1 &&
3237 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3238 	    ("nfs readdirrpc bad uio"));
3239 	ncookie.lval[0] = ncookie.lval[1] = 0;
3240 	/*
3241 	 * There is no point in reading a lot more than uio_resid, however
3242 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
3243 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
3244 	 * will never make readsize > nm_readdirsize.
3245 	 */
3246 	readsize = nmp->nm_readdirsize;
3247 	if (readsize > uiop->uio_resid)
3248 		readsize = uiop->uio_resid + DIRBLKSIZ;
3249 
3250 	*attrflagp = 0;
3251 	if (eofp)
3252 		*eofp = 0;
3253 	tresid = uiop->uio_resid;
3254 	cookie.lval[0] = cookiep->nfsuquad[0];
3255 	cookie.lval[1] = cookiep->nfsuquad[1];
3256 	nd->nd_mrep = NULL;
3257 
3258 	/*
3259 	 * For NFSv4, first create the "." and ".." entries.
3260 	 */
3261 	if (NFSHASNFSV4(nmp)) {
3262 		reqsize = 6 * NFSX_UNSIGNED;
3263 		NFSGETATTR_ATTRBIT(&dattrbits);
3264 		NFSZERO_ATTRBIT(&attrbits);
3265 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3266 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
3267 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3268 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3269 			NFSSETBIT_ATTRBIT(&attrbits,
3270 			    NFSATTRBIT_MOUNTEDONFILEID);
3271 			gotmnton = 1;
3272 		} else {
3273 			/*
3274 			 * Must fake it. Use the fileno, except when the
3275 			 * fsid is != to that of the directory. For that
3276 			 * case, generate a fake fileno that is not the same.
3277 			 */
3278 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3279 			gotmnton = 0;
3280 		}
3281 
3282 		/*
3283 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3284 		 */
3285 		if (uiop->uio_offset == 0) {
3286 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3287 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3288 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3289 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3290 			(void) nfsrv_putattrbit(nd, &attrbits);
3291 			error = nfscl_request(nd, vp, p, cred);
3292 			if (error)
3293 			    return (error);
3294 			dotfileid = 0;	/* Fake out the compiler. */
3295 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3296 			    error = nfsm_loadattr(nd, &nfsva);
3297 			    if (error != 0)
3298 				goto nfsmout;
3299 			    dotfileid = nfsva.na_fileid;
3300 			}
3301 			if (nd->nd_repstat == 0) {
3302 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3303 			    len = fxdr_unsigned(int, *(tl + 4));
3304 			    if (len > 0 && len <= NFSX_V4FHMAX)
3305 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3306 			    else
3307 				error = EPERM;
3308 			    if (!error) {
3309 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3310 				nfsva.na_mntonfileno = UINT64_MAX;
3311 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3312 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3313 				    NULL, NULL, NULL, p, cred);
3314 				if (error) {
3315 				    dotdotfileid = dotfileid;
3316 				} else if (gotmnton) {
3317 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3318 					dotdotfileid = nfsva.na_mntonfileno;
3319 				    else
3320 					dotdotfileid = nfsva.na_fileid;
3321 				} else if (nfsva.na_filesid[0] ==
3322 				    dnp->n_vattr.na_filesid[0] &&
3323 				    nfsva.na_filesid[1] ==
3324 				    dnp->n_vattr.na_filesid[1]) {
3325 				    dotdotfileid = nfsva.na_fileid;
3326 				} else {
3327 				    do {
3328 					fakefileno--;
3329 				    } while (fakefileno ==
3330 					nfsva.na_fileid);
3331 				    dotdotfileid = fakefileno;
3332 				}
3333 			    }
3334 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3335 			    /*
3336 			     * Lookupp returns NFSERR_NOENT when we are
3337 			     * at the root, so just use the current dir.
3338 			     */
3339 			    nd->nd_repstat = 0;
3340 			    dotdotfileid = dotfileid;
3341 			} else {
3342 			    error = nd->nd_repstat;
3343 			}
3344 			m_freem(nd->nd_mrep);
3345 			if (error)
3346 			    return (error);
3347 			nd->nd_mrep = NULL;
3348 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3349 			dp->d_pad0 = dp->d_pad1 = 0;
3350 			dp->d_off = 0;
3351 			dp->d_type = DT_DIR;
3352 			dp->d_fileno = dotfileid;
3353 			dp->d_namlen = 1;
3354 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3355 			dp->d_name[0] = '.';
3356 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3357 			/*
3358 			 * Just make these offset cookie 0.
3359 			 */
3360 			tl = (u_int32_t *)&dp->d_name[8];
3361 			*tl++ = 0;
3362 			*tl = 0;
3363 			blksiz += dp->d_reclen;
3364 			uiop->uio_resid -= dp->d_reclen;
3365 			uiop->uio_offset += dp->d_reclen;
3366 			uiop->uio_iov->iov_base =
3367 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3368 			uiop->uio_iov->iov_len -= dp->d_reclen;
3369 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3370 			dp->d_pad0 = dp->d_pad1 = 0;
3371 			dp->d_off = 0;
3372 			dp->d_type = DT_DIR;
3373 			dp->d_fileno = dotdotfileid;
3374 			dp->d_namlen = 2;
3375 			*((uint64_t *)dp->d_name) = 0;
3376 			dp->d_name[0] = '.';
3377 			dp->d_name[1] = '.';
3378 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3379 			/*
3380 			 * Just make these offset cookie 0.
3381 			 */
3382 			tl = (u_int32_t *)&dp->d_name[8];
3383 			*tl++ = 0;
3384 			*tl = 0;
3385 			blksiz += dp->d_reclen;
3386 			uiop->uio_resid -= dp->d_reclen;
3387 			uiop->uio_offset += dp->d_reclen;
3388 			uiop->uio_iov->iov_base =
3389 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3390 			uiop->uio_iov->iov_len -= dp->d_reclen;
3391 		}
3392 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3393 	} else {
3394 		reqsize = 5 * NFSX_UNSIGNED;
3395 	}
3396 
3397 	/*
3398 	 * Loop around doing readdir rpc's of size readsize.
3399 	 * The stopping criteria is EOF or buffer full.
3400 	 */
3401 	while (more_dirs && bigenough) {
3402 		*attrflagp = 0;
3403 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3404 		if (nd->nd_flag & ND_NFSV2) {
3405 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3406 			*tl++ = cookie.lval[1];
3407 			*tl = txdr_unsigned(readsize);
3408 		} else {
3409 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3410 			*tl++ = cookie.lval[0];
3411 			*tl++ = cookie.lval[1];
3412 			if (cookie.qval == 0) {
3413 				*tl++ = 0;
3414 				*tl++ = 0;
3415 			} else {
3416 				NFSLOCKNODE(dnp);
3417 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3418 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3419 				NFSUNLOCKNODE(dnp);
3420 			}
3421 			if (nd->nd_flag & ND_NFSV4) {
3422 				*tl++ = txdr_unsigned(readsize);
3423 				*tl = txdr_unsigned(readsize);
3424 				(void) nfsrv_putattrbit(nd, &attrbits);
3425 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3426 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3427 				(void) nfsrv_putattrbit(nd, &dattrbits);
3428 			} else {
3429 				*tl = txdr_unsigned(readsize);
3430 			}
3431 		}
3432 		error = nfscl_request(nd, vp, p, cred);
3433 		if (error)
3434 			return (error);
3435 		if (!(nd->nd_flag & ND_NFSV2)) {
3436 			if (nd->nd_flag & ND_NFSV3)
3437 				error = nfscl_postop_attr(nd, nap, attrflagp);
3438 			if (!nd->nd_repstat && !error) {
3439 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3440 				NFSLOCKNODE(dnp);
3441 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3442 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3443 				NFSUNLOCKNODE(dnp);
3444 			}
3445 		}
3446 		if (nd->nd_repstat || error) {
3447 			if (!error)
3448 				error = nd->nd_repstat;
3449 			goto nfsmout;
3450 		}
3451 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3452 		more_dirs = fxdr_unsigned(int, *tl);
3453 		if (!more_dirs)
3454 			tryformoredirs = 0;
3455 
3456 		/* loop through the dir entries, doctoring them to 4bsd form */
3457 		while (more_dirs && bigenough) {
3458 			if (nd->nd_flag & ND_NFSV4) {
3459 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3460 				ncookie.lval[0] = *tl++;
3461 				ncookie.lval[1] = *tl++;
3462 				len = fxdr_unsigned(int, *tl);
3463 			} else if (nd->nd_flag & ND_NFSV3) {
3464 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3465 				nfsva.na_fileid = fxdr_hyper(tl);
3466 				tl += 2;
3467 				len = fxdr_unsigned(int, *tl);
3468 			} else {
3469 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3470 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3471 				    *tl++);
3472 				len = fxdr_unsigned(int, *tl);
3473 			}
3474 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3475 				error = EBADRPC;
3476 				goto nfsmout;
3477 			}
3478 			tlen = roundup2(len, 8);
3479 			if (tlen == len)
3480 				tlen += 8;  /* To ensure null termination. */
3481 			left = DIRBLKSIZ - blksiz;
3482 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3483 				NFSBZERO(uiop->uio_iov->iov_base, left);
3484 				dp->d_reclen += left;
3485 				uiop->uio_iov->iov_base =
3486 				    (char *)uiop->uio_iov->iov_base + left;
3487 				uiop->uio_iov->iov_len -= left;
3488 				uiop->uio_resid -= left;
3489 				uiop->uio_offset += left;
3490 				blksiz = 0;
3491 			}
3492 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3493 			    uiop->uio_resid)
3494 				bigenough = 0;
3495 			if (bigenough) {
3496 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3497 				dp->d_pad0 = dp->d_pad1 = 0;
3498 				dp->d_off = 0;
3499 				dp->d_namlen = len;
3500 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3501 				    NFSX_HYPER;
3502 				dp->d_type = DT_UNKNOWN;
3503 				blksiz += dp->d_reclen;
3504 				if (blksiz == DIRBLKSIZ)
3505 					blksiz = 0;
3506 				uiop->uio_resid -= DIRHDSIZ;
3507 				uiop->uio_offset += DIRHDSIZ;
3508 				uiop->uio_iov->iov_base =
3509 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3510 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3511 				error = nfsm_mbufuio(nd, uiop, len);
3512 				if (error)
3513 					goto nfsmout;
3514 				cp = uiop->uio_iov->iov_base;
3515 				tlen -= len;
3516 				NFSBZERO(cp, tlen);
3517 				cp += tlen;	/* points to cookie storage */
3518 				tl2 = (u_int32_t *)cp;
3519 				uiop->uio_iov->iov_base =
3520 				    (char *)uiop->uio_iov->iov_base + tlen +
3521 				    NFSX_HYPER;
3522 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3523 				uiop->uio_resid -= tlen + NFSX_HYPER;
3524 				uiop->uio_offset += (tlen + NFSX_HYPER);
3525 			} else {
3526 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3527 				if (error)
3528 					goto nfsmout;
3529 			}
3530 			if (nd->nd_flag & ND_NFSV4) {
3531 				rderr = 0;
3532 				nfsva.na_mntonfileno = UINT64_MAX;
3533 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3534 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3535 				    NULL, NULL, &rderr, p, cred);
3536 				if (error)
3537 					goto nfsmout;
3538 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3539 			} else if (nd->nd_flag & ND_NFSV3) {
3540 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3541 				ncookie.lval[0] = *tl++;
3542 				ncookie.lval[1] = *tl++;
3543 			} else {
3544 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3545 				ncookie.lval[0] = 0;
3546 				ncookie.lval[1] = *tl++;
3547 			}
3548 			if (bigenough) {
3549 			    if (nd->nd_flag & ND_NFSV4) {
3550 				if (rderr) {
3551 				    dp->d_fileno = 0;
3552 				} else {
3553 				    if (gotmnton) {
3554 					if (nfsva.na_mntonfileno != UINT64_MAX)
3555 					    dp->d_fileno = nfsva.na_mntonfileno;
3556 					else
3557 					    dp->d_fileno = nfsva.na_fileid;
3558 				    } else if (nfsva.na_filesid[0] ==
3559 					dnp->n_vattr.na_filesid[0] &&
3560 					nfsva.na_filesid[1] ==
3561 					dnp->n_vattr.na_filesid[1]) {
3562 					dp->d_fileno = nfsva.na_fileid;
3563 				    } else {
3564 					do {
3565 					    fakefileno--;
3566 					} while (fakefileno ==
3567 					    nfsva.na_fileid);
3568 					dp->d_fileno = fakefileno;
3569 				    }
3570 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3571 				}
3572 			    } else {
3573 				dp->d_fileno = nfsva.na_fileid;
3574 			    }
3575 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3576 				ncookie.lval[0];
3577 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3578 				ncookie.lval[1];
3579 			}
3580 			more_dirs = fxdr_unsigned(int, *tl);
3581 		}
3582 		/*
3583 		 * If at end of rpc data, get the eof boolean
3584 		 */
3585 		if (!more_dirs) {
3586 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3587 			eof = fxdr_unsigned(int, *tl);
3588 			if (tryformoredirs)
3589 				more_dirs = !eof;
3590 			if (nd->nd_flag & ND_NFSV4) {
3591 				error = nfscl_postop_attr(nd, nap, attrflagp);
3592 				if (error)
3593 					goto nfsmout;
3594 			}
3595 		}
3596 		m_freem(nd->nd_mrep);
3597 		nd->nd_mrep = NULL;
3598 	}
3599 	/*
3600 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3601 	 * by increasing d_reclen for the last record.
3602 	 */
3603 	if (blksiz > 0) {
3604 		left = DIRBLKSIZ - blksiz;
3605 		NFSBZERO(uiop->uio_iov->iov_base, left);
3606 		dp->d_reclen += left;
3607 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3608 		    left;
3609 		uiop->uio_iov->iov_len -= left;
3610 		uiop->uio_resid -= left;
3611 		uiop->uio_offset += left;
3612 	}
3613 
3614 	/*
3615 	 * If returning no data, assume end of file.
3616 	 * If not bigenough, return not end of file, since you aren't
3617 	 *    returning all the data
3618 	 * Otherwise, return the eof flag from the server.
3619 	 */
3620 	if (eofp) {
3621 		if (tresid == ((size_t)(uiop->uio_resid)))
3622 			*eofp = 1;
3623 		else if (!bigenough)
3624 			*eofp = 0;
3625 		else
3626 			*eofp = eof;
3627 	}
3628 
3629 	/*
3630 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3631 	 */
3632 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3633 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3634 		NFSBZERO(dp, DIRBLKSIZ);
3635 		dp->d_type = DT_UNKNOWN;
3636 		tl = (u_int32_t *)&dp->d_name[4];
3637 		*tl++ = cookie.lval[0];
3638 		*tl = cookie.lval[1];
3639 		dp->d_reclen = DIRBLKSIZ;
3640 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3641 		    DIRBLKSIZ;
3642 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3643 		uiop->uio_resid -= DIRBLKSIZ;
3644 		uiop->uio_offset += DIRBLKSIZ;
3645 	}
3646 
3647 nfsmout:
3648 	if (nd->nd_mrep != NULL)
3649 		m_freem(nd->nd_mrep);
3650 	return (error);
3651 }
3652 
3653 #ifndef APPLE
3654 /*
3655  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3656  * (Also used for NFS V4 when mount flag set.)
3657  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3658  */
3659 int
3660 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3661     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3662     int *eofp, void *stuff)
3663 {
3664 	int len, left;
3665 	struct dirent *dp = NULL;
3666 	u_int32_t *tl;
3667 	vnode_t newvp = NULLVP;
3668 	struct nfsrv_descript nfsd, *nd = &nfsd;
3669 	struct nameidata nami, *ndp = &nami;
3670 	struct componentname *cnp = &ndp->ni_cnd;
3671 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3672 	struct nfsnode *dnp = VTONFS(vp), *np;
3673 	struct nfsvattr nfsva;
3674 	struct nfsfh *nfhp;
3675 	nfsquad_t cookie, ncookie;
3676 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3677 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3678 	int isdotdot = 0, unlocknewvp = 0;
3679 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3680 	u_int64_t fileno = 0;
3681 	char *cp;
3682 	nfsattrbit_t attrbits, dattrbits;
3683 	size_t tresid;
3684 	u_int32_t *tl2 = NULL, rderr;
3685 	struct timespec dctime, ts;
3686 	bool attr_ok;
3687 
3688 	KASSERT(uiop->uio_iovcnt == 1 &&
3689 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3690 	    ("nfs readdirplusrpc bad uio"));
3691 	ncookie.lval[0] = ncookie.lval[1] = 0;
3692 	timespecclear(&dctime);
3693 	*attrflagp = 0;
3694 	if (eofp != NULL)
3695 		*eofp = 0;
3696 	ndp->ni_dvp = vp;
3697 	nd->nd_mrep = NULL;
3698 	cookie.lval[0] = cookiep->nfsuquad[0];
3699 	cookie.lval[1] = cookiep->nfsuquad[1];
3700 	tresid = uiop->uio_resid;
3701 
3702 	/*
3703 	 * For NFSv4, first create the "." and ".." entries.
3704 	 */
3705 	if (NFSHASNFSV4(nmp)) {
3706 		NFSGETATTR_ATTRBIT(&dattrbits);
3707 		NFSZERO_ATTRBIT(&attrbits);
3708 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3709 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3710 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3711 			NFSSETBIT_ATTRBIT(&attrbits,
3712 			    NFSATTRBIT_MOUNTEDONFILEID);
3713 			gotmnton = 1;
3714 		} else {
3715 			/*
3716 			 * Must fake it. Use the fileno, except when the
3717 			 * fsid is != to that of the directory. For that
3718 			 * case, generate a fake fileno that is not the same.
3719 			 */
3720 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3721 			gotmnton = 0;
3722 		}
3723 
3724 		/*
3725 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3726 		 */
3727 		if (uiop->uio_offset == 0) {
3728 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3729 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3730 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3731 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3732 			(void) nfsrv_putattrbit(nd, &attrbits);
3733 			error = nfscl_request(nd, vp, p, cred);
3734 			if (error)
3735 			    return (error);
3736 			dotfileid = 0;	/* Fake out the compiler. */
3737 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3738 			    error = nfsm_loadattr(nd, &nfsva);
3739 			    if (error != 0)
3740 				goto nfsmout;
3741 			    dctime = nfsva.na_ctime;
3742 			    dotfileid = nfsva.na_fileid;
3743 			}
3744 			if (nd->nd_repstat == 0) {
3745 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3746 			    len = fxdr_unsigned(int, *(tl + 4));
3747 			    if (len > 0 && len <= NFSX_V4FHMAX)
3748 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3749 			    else
3750 				error = EPERM;
3751 			    if (!error) {
3752 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3753 				nfsva.na_mntonfileno = UINT64_MAX;
3754 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3755 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3756 				    NULL, NULL, NULL, p, cred);
3757 				if (error) {
3758 				    dotdotfileid = dotfileid;
3759 				} else if (gotmnton) {
3760 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3761 					dotdotfileid = nfsva.na_mntonfileno;
3762 				    else
3763 					dotdotfileid = nfsva.na_fileid;
3764 				} else if (nfsva.na_filesid[0] ==
3765 				    dnp->n_vattr.na_filesid[0] &&
3766 				    nfsva.na_filesid[1] ==
3767 				    dnp->n_vattr.na_filesid[1]) {
3768 				    dotdotfileid = nfsva.na_fileid;
3769 				} else {
3770 				    do {
3771 					fakefileno--;
3772 				    } while (fakefileno ==
3773 					nfsva.na_fileid);
3774 				    dotdotfileid = fakefileno;
3775 				}
3776 			    }
3777 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3778 			    /*
3779 			     * Lookupp returns NFSERR_NOENT when we are
3780 			     * at the root, so just use the current dir.
3781 			     */
3782 			    nd->nd_repstat = 0;
3783 			    dotdotfileid = dotfileid;
3784 			} else {
3785 			    error = nd->nd_repstat;
3786 			}
3787 			m_freem(nd->nd_mrep);
3788 			if (error)
3789 			    return (error);
3790 			nd->nd_mrep = NULL;
3791 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3792 			dp->d_pad0 = dp->d_pad1 = 0;
3793 			dp->d_off = 0;
3794 			dp->d_type = DT_DIR;
3795 			dp->d_fileno = dotfileid;
3796 			dp->d_namlen = 1;
3797 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3798 			dp->d_name[0] = '.';
3799 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3800 			/*
3801 			 * Just make these offset cookie 0.
3802 			 */
3803 			tl = (u_int32_t *)&dp->d_name[8];
3804 			*tl++ = 0;
3805 			*tl = 0;
3806 			blksiz += dp->d_reclen;
3807 			uiop->uio_resid -= dp->d_reclen;
3808 			uiop->uio_offset += dp->d_reclen;
3809 			uiop->uio_iov->iov_base =
3810 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3811 			uiop->uio_iov->iov_len -= dp->d_reclen;
3812 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3813 			dp->d_pad0 = dp->d_pad1 = 0;
3814 			dp->d_off = 0;
3815 			dp->d_type = DT_DIR;
3816 			dp->d_fileno = dotdotfileid;
3817 			dp->d_namlen = 2;
3818 			*((uint64_t *)dp->d_name) = 0;
3819 			dp->d_name[0] = '.';
3820 			dp->d_name[1] = '.';
3821 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3822 			/*
3823 			 * Just make these offset cookie 0.
3824 			 */
3825 			tl = (u_int32_t *)&dp->d_name[8];
3826 			*tl++ = 0;
3827 			*tl = 0;
3828 			blksiz += dp->d_reclen;
3829 			uiop->uio_resid -= dp->d_reclen;
3830 			uiop->uio_offset += dp->d_reclen;
3831 			uiop->uio_iov->iov_base =
3832 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3833 			uiop->uio_iov->iov_len -= dp->d_reclen;
3834 		}
3835 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3836 		if (gotmnton)
3837 			NFSSETBIT_ATTRBIT(&attrbits,
3838 			    NFSATTRBIT_MOUNTEDONFILEID);
3839 		if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3840 		    NFSATTRBIT_TIMECREATE))
3841 			NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
3842 	}
3843 
3844 	/*
3845 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3846 	 * The stopping criteria is EOF or buffer full.
3847 	 */
3848 	while (more_dirs && bigenough) {
3849 		*attrflagp = 0;
3850 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3851  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3852 		*tl++ = cookie.lval[0];
3853 		*tl++ = cookie.lval[1];
3854 		if (cookie.qval == 0) {
3855 			*tl++ = 0;
3856 			*tl++ = 0;
3857 		} else {
3858 			NFSLOCKNODE(dnp);
3859 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3860 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3861 			NFSUNLOCKNODE(dnp);
3862 		}
3863 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3864 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3865 		if (nd->nd_flag & ND_NFSV4) {
3866 			(void) nfsrv_putattrbit(nd, &attrbits);
3867 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3868 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3869 			(void) nfsrv_putattrbit(nd, &dattrbits);
3870 		}
3871 		nanouptime(&ts);
3872 		error = nfscl_request(nd, vp, p, cred);
3873 		if (error)
3874 			return (error);
3875 		if (nd->nd_flag & ND_NFSV3)
3876 			error = nfscl_postop_attr(nd, nap, attrflagp);
3877 		if (nd->nd_repstat || error) {
3878 			if (!error)
3879 				error = nd->nd_repstat;
3880 			goto nfsmout;
3881 		}
3882 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3883 			dctime = nap->na_ctime;
3884 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3885 		NFSLOCKNODE(dnp);
3886 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3887 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3888 		NFSUNLOCKNODE(dnp);
3889 		more_dirs = fxdr_unsigned(int, *tl);
3890 		if (!more_dirs)
3891 			tryformoredirs = 0;
3892 
3893 		/* loop through the dir entries, doctoring them to 4bsd form */
3894 		while (more_dirs && bigenough) {
3895 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3896 			if (nd->nd_flag & ND_NFSV4) {
3897 				ncookie.lval[0] = *tl++;
3898 				ncookie.lval[1] = *tl++;
3899 			} else {
3900 				fileno = fxdr_hyper(tl);
3901 				tl += 2;
3902 			}
3903 			len = fxdr_unsigned(int, *tl);
3904 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3905 				error = EBADRPC;
3906 				goto nfsmout;
3907 			}
3908 			tlen = roundup2(len, 8);
3909 			if (tlen == len)
3910 				tlen += 8;  /* To ensure null termination. */
3911 			left = DIRBLKSIZ - blksiz;
3912 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3913 				NFSBZERO(uiop->uio_iov->iov_base, left);
3914 				dp->d_reclen += left;
3915 				uiop->uio_iov->iov_base =
3916 				    (char *)uiop->uio_iov->iov_base + left;
3917 				uiop->uio_iov->iov_len -= left;
3918 				uiop->uio_resid -= left;
3919 				uiop->uio_offset += left;
3920 				blksiz = 0;
3921 			}
3922 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3923 			    uiop->uio_resid)
3924 				bigenough = 0;
3925 			if (bigenough) {
3926 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3927 				dp->d_pad0 = dp->d_pad1 = 0;
3928 				dp->d_off = 0;
3929 				dp->d_namlen = len;
3930 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3931 				    NFSX_HYPER;
3932 				dp->d_type = DT_UNKNOWN;
3933 				blksiz += dp->d_reclen;
3934 				if (blksiz == DIRBLKSIZ)
3935 					blksiz = 0;
3936 				uiop->uio_resid -= DIRHDSIZ;
3937 				uiop->uio_offset += DIRHDSIZ;
3938 				uiop->uio_iov->iov_base =
3939 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3940 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3941 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3942 				cnp->cn_namelen = len;
3943 				NFSCNHASHZERO(cnp);
3944 				error = nfsm_mbufuio(nd, uiop, len);
3945 				if (error)
3946 					goto nfsmout;
3947 				cp = uiop->uio_iov->iov_base;
3948 				tlen -= len;
3949 				NFSBZERO(cp, tlen);
3950 				cp += tlen;	/* points to cookie storage */
3951 				tl2 = (u_int32_t *)cp;
3952 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3953 				    cnp->cn_nameptr[1] == '.')
3954 					isdotdot = 1;
3955 				else
3956 					isdotdot = 0;
3957 				uiop->uio_iov->iov_base =
3958 				    (char *)uiop->uio_iov->iov_base + tlen +
3959 				    NFSX_HYPER;
3960 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3961 				uiop->uio_resid -= tlen + NFSX_HYPER;
3962 				uiop->uio_offset += (tlen + NFSX_HYPER);
3963 			} else {
3964 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3965 				if (error)
3966 					goto nfsmout;
3967 			}
3968 			nfhp = NULL;
3969 			if (nd->nd_flag & ND_NFSV3) {
3970 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3971 				ncookie.lval[0] = *tl++;
3972 				ncookie.lval[1] = *tl++;
3973 				attrflag = fxdr_unsigned(int, *tl);
3974 				if (attrflag) {
3975 				  error = nfsm_loadattr(nd, &nfsva);
3976 				  if (error)
3977 					goto nfsmout;
3978 				}
3979 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3980 				if (*tl) {
3981 					error = nfsm_getfh(nd, &nfhp);
3982 					if (error)
3983 					    goto nfsmout;
3984 				}
3985 				if (!attrflag && nfhp != NULL) {
3986 					free(nfhp, M_NFSFH);
3987 					nfhp = NULL;
3988 				}
3989 			} else {
3990 				rderr = 0;
3991 				nfsva.na_mntonfileno = 0xffffffff;
3992 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3993 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3994 				    NULL, NULL, &rderr, p, cred);
3995 				if (error)
3996 					goto nfsmout;
3997 			}
3998 
3999 			if (bigenough) {
4000 			    if (nd->nd_flag & ND_NFSV4) {
4001 				if (rderr) {
4002 				    dp->d_fileno = 0;
4003 				} else if (gotmnton) {
4004 				    if (nfsva.na_mntonfileno != 0xffffffff)
4005 					dp->d_fileno = nfsva.na_mntonfileno;
4006 				    else
4007 					dp->d_fileno = nfsva.na_fileid;
4008 				} else if (nfsva.na_filesid[0] ==
4009 				    dnp->n_vattr.na_filesid[0] &&
4010 				    nfsva.na_filesid[1] ==
4011 				    dnp->n_vattr.na_filesid[1]) {
4012 				    dp->d_fileno = nfsva.na_fileid;
4013 				} else {
4014 				    do {
4015 					fakefileno--;
4016 				    } while (fakefileno ==
4017 					nfsva.na_fileid);
4018 				    dp->d_fileno = fakefileno;
4019 				}
4020 			    } else {
4021 				dp->d_fileno = fileno;
4022 			    }
4023 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
4024 				ncookie.lval[0];
4025 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
4026 				ncookie.lval[1];
4027 
4028 			    if (nfhp != NULL) {
4029 				attr_ok = true;
4030 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
4031 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
4032 				    VREF(vp);
4033 				    newvp = vp;
4034 				    unlocknewvp = 0;
4035 				    free(nfhp, M_NFSFH);
4036 				    np = dnp;
4037 				} else if (isdotdot != 0) {
4038 				    /*
4039 				     * Skip doing a nfscl_nget() call for "..".
4040 				     * There's a race between acquiring the nfs
4041 				     * node here and lookups that look for the
4042 				     * directory being read (in the parent).
4043 				     * It would try to get a lock on ".." here,
4044 				     * owning the lock on the directory being
4045 				     * read. Lookup will hold the lock on ".."
4046 				     * and try to acquire the lock on the
4047 				     * directory being read.
4048 				     * If the directory is unlocked/relocked,
4049 				     * then there is a LOR with the buflock
4050 				     * vp is relocked.
4051 				     */
4052 				    free(nfhp, M_NFSFH);
4053 				} else {
4054 				    error = nfscl_nget(vp->v_mount, vp,
4055 				      nfhp, cnp, p, &np, LK_EXCLUSIVE);
4056 				    if (!error) {
4057 					newvp = NFSTOV(np);
4058 					unlocknewvp = 1;
4059 					/*
4060 					 * If n_localmodtime >= time before RPC,
4061 					 * then a file modification operation,
4062 					 * such as VOP_SETATTR() of size, has
4063 					 * occurred while the Lookup RPC and
4064 					 * acquisition of the vnode happened. As
4065 					 * such, the attributes might be stale,
4066 					 * with possibly an incorrect size.
4067 					 */
4068 					NFSLOCKNODE(np);
4069 					if (timespecisset(
4070 					    &np->n_localmodtime) &&
4071 					    timespeccmp(&np->n_localmodtime,
4072 					    &ts, >=)) {
4073 					    NFSCL_DEBUG(4, "nfsrpc_readdirplus:"
4074 						" localmod stale attributes\n");
4075 					    attr_ok = false;
4076 					}
4077 					NFSUNLOCKNODE(np);
4078 				    }
4079 				}
4080 				nfhp = NULL;
4081 				if (newvp != NULLVP) {
4082 				    if (attr_ok)
4083 					error = nfscl_loadattrcache(&newvp,
4084 					    &nfsva, NULL, 0, 0);
4085 				    if (error) {
4086 					if (unlocknewvp)
4087 					    vput(newvp);
4088 					else
4089 					    vrele(newvp);
4090 					goto nfsmout;
4091 				    }
4092 				    dp->d_type =
4093 					vtonfs_dtype(np->n_vattr.na_type);
4094 				    ndp->ni_vp = newvp;
4095 				    NFSCNHASH(cnp, HASHINIT);
4096 				    if (cnp->cn_namelen <= NCHNAMLEN &&
4097 					ndp->ni_dvp != ndp->ni_vp &&
4098 					(newvp->v_type != VDIR ||
4099 					 dctime.tv_sec != 0)) {
4100 					cache_enter_time_flags(ndp->ni_dvp,
4101 					    ndp->ni_vp, cnp,
4102 					    &nfsva.na_ctime,
4103 					    newvp->v_type != VDIR ? NULL :
4104 					    &dctime, VFS_CACHE_DROPOLD);
4105 				    }
4106 				    if (unlocknewvp)
4107 					vput(newvp);
4108 				    else
4109 					vrele(newvp);
4110 				    newvp = NULLVP;
4111 				}
4112 			    }
4113 			} else if (nfhp != NULL) {
4114 			    free(nfhp, M_NFSFH);
4115 			}
4116 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4117 			more_dirs = fxdr_unsigned(int, *tl);
4118 		}
4119 		/*
4120 		 * If at end of rpc data, get the eof boolean
4121 		 */
4122 		if (!more_dirs) {
4123 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4124 			eof = fxdr_unsigned(int, *tl);
4125 			if (tryformoredirs)
4126 				more_dirs = !eof;
4127 			if (nd->nd_flag & ND_NFSV4) {
4128 				error = nfscl_postop_attr(nd, nap, attrflagp);
4129 				if (error)
4130 					goto nfsmout;
4131 			}
4132 		}
4133 		m_freem(nd->nd_mrep);
4134 		nd->nd_mrep = NULL;
4135 	}
4136 	/*
4137 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
4138 	 * by increasing d_reclen for the last record.
4139 	 */
4140 	if (blksiz > 0) {
4141 		left = DIRBLKSIZ - blksiz;
4142 		NFSBZERO(uiop->uio_iov->iov_base, left);
4143 		dp->d_reclen += left;
4144 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4145 		    left;
4146 		uiop->uio_iov->iov_len -= left;
4147 		uiop->uio_resid -= left;
4148 		uiop->uio_offset += left;
4149 	}
4150 
4151 	/*
4152 	 * If returning no data, assume end of file.
4153 	 * If not bigenough, return not end of file, since you aren't
4154 	 *    returning all the data
4155 	 * Otherwise, return the eof flag from the server.
4156 	 */
4157 	if (eofp != NULL) {
4158 		if (tresid == uiop->uio_resid)
4159 			*eofp = 1;
4160 		else if (!bigenough)
4161 			*eofp = 0;
4162 		else
4163 			*eofp = eof;
4164 	}
4165 
4166 	/*
4167 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
4168 	 */
4169 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
4170 		dp = (struct dirent *)uiop->uio_iov->iov_base;
4171 		NFSBZERO(dp, DIRBLKSIZ);
4172 		dp->d_type = DT_UNKNOWN;
4173 		tl = (u_int32_t *)&dp->d_name[4];
4174 		*tl++ = cookie.lval[0];
4175 		*tl = cookie.lval[1];
4176 		dp->d_reclen = DIRBLKSIZ;
4177 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4178 		    DIRBLKSIZ;
4179 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
4180 		uiop->uio_resid -= DIRBLKSIZ;
4181 		uiop->uio_offset += DIRBLKSIZ;
4182 	}
4183 
4184 nfsmout:
4185 	if (nd->nd_mrep != NULL)
4186 		m_freem(nd->nd_mrep);
4187 	return (error);
4188 }
4189 #endif	/* !APPLE */
4190 
4191 /*
4192  * Nfs commit rpc
4193  */
4194 int
4195 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
4196     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4197 {
4198 	u_int32_t *tl;
4199 	struct nfsrv_descript nfsd, *nd = &nfsd;
4200 	nfsattrbit_t attrbits;
4201 	int error;
4202 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4203 
4204 	*attrflagp = 0;
4205 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
4206 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4207 	txdr_hyper(offset, tl);
4208 	tl += 2;
4209 	*tl = txdr_unsigned(cnt);
4210 	if (nd->nd_flag & ND_NFSV4) {
4211 		/*
4212 		 * And do a Getattr op.
4213 		 */
4214 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4215 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
4216 		NFSGETATTR_ATTRBIT(&attrbits);
4217 		(void) nfsrv_putattrbit(nd, &attrbits);
4218 	}
4219 	error = nfscl_request(nd, vp, p, cred);
4220 	if (error)
4221 		return (error);
4222 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, NULL);
4223 	if (!error && !nd->nd_repstat) {
4224 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
4225 		NFSLOCKMNT(nmp);
4226 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
4227 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
4228 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
4229 		}
4230 		NFSUNLOCKMNT(nmp);
4231 		if (nd->nd_flag & ND_NFSV4)
4232 			error = nfscl_postop_attr(nd, nap, attrflagp);
4233 	}
4234 nfsmout:
4235 	if (!error && nd->nd_repstat)
4236 		error = nd->nd_repstat;
4237 	m_freem(nd->nd_mrep);
4238 	return (error);
4239 }
4240 
4241 /*
4242  * NFS byte range lock rpc.
4243  * (Mostly just calls one of the three lower level RPC routines.)
4244  */
4245 int
4246 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
4247     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4248 {
4249 	struct nfscllockowner *lp;
4250 	struct nfsclclient *clp;
4251 	struct nfsfh *nfhp;
4252 	struct nfsrv_descript nfsd, *nd = &nfsd;
4253 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4254 	u_int64_t off, len;
4255 	off_t start, end;
4256 	u_int32_t clidrev = 0;
4257 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
4258 	int callcnt, dorpc;
4259 
4260 	/*
4261 	 * Convert the flock structure into a start and end and do POSIX
4262 	 * bounds checking.
4263 	 */
4264 	switch (fl->l_whence) {
4265 	case SEEK_SET:
4266 	case SEEK_CUR:
4267 		/*
4268 		 * Caller is responsible for adding any necessary offset
4269 		 * when SEEK_CUR is used.
4270 		 */
4271 		start = fl->l_start;
4272 		off = fl->l_start;
4273 		break;
4274 	case SEEK_END:
4275 		start = size + fl->l_start;
4276 		off = size + fl->l_start;
4277 		break;
4278 	default:
4279 		return (EINVAL);
4280 	}
4281 	if (start < 0)
4282 		return (EINVAL);
4283 	if (fl->l_len != 0) {
4284 		end = start + fl->l_len - 1;
4285 		if (end < start)
4286 			return (EINVAL);
4287 	}
4288 
4289 	len = fl->l_len;
4290 	if (len == 0)
4291 		len = NFS64BITSSET;
4292 	retrycnt = 0;
4293 	do {
4294 	    nd->nd_repstat = 0;
4295 	    if (op == F_GETLK) {
4296 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4297 		if (error)
4298 			return (error);
4299 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
4300 		if (!error) {
4301 			clidrev = clp->nfsc_clientidrev;
4302 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
4303 			    p, id, flags);
4304 		} else if (error == -1) {
4305 			error = 0;
4306 		}
4307 		nfscl_clientrelease(clp);
4308 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
4309 		/*
4310 		 * We must loop around for all lockowner cases.
4311 		 */
4312 		callcnt = 0;
4313 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4314 		if (error)
4315 			return (error);
4316 		do {
4317 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
4318 			clp, id, flags, &lp, &dorpc);
4319 		    /*
4320 		     * If it returns a NULL lp, we're done.
4321 		     */
4322 		    if (lp == NULL) {
4323 			if (callcnt == 0)
4324 			    nfscl_clientrelease(clp);
4325 			else
4326 			    nfscl_releasealllocks(clp, vp, p, id, flags);
4327 			return (error);
4328 		    }
4329 		    if (nmp->nm_clp != NULL)
4330 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4331 		    else
4332 			clidrev = 0;
4333 		    /*
4334 		     * If the server doesn't support Posix lock semantics,
4335 		     * only allow locks on the entire file, since it won't
4336 		     * handle overlapping byte ranges.
4337 		     * There might still be a problem when a lock
4338 		     * upgrade/downgrade (read<->write) occurs, since the
4339 		     * server "might" expect an unlock first?
4340 		     */
4341 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4342 			(off == 0 && len == NFS64BITSSET))) {
4343 			/*
4344 			 * Since the lock records will go away, we must
4345 			 * wait for grace and delay here.
4346 			 */
4347 			do {
4348 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4349 				NFSV4LOCKT_READ, cred, p, 0);
4350 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4351 				 nd->nd_repstat == NFSERR_DELAY) &&
4352 				error == 0)
4353 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4354 				    "nfs_advlock");
4355 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4356 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4357 		    }
4358 		    callcnt++;
4359 		} while (error == 0 && nd->nd_repstat == 0);
4360 		nfscl_releasealllocks(clp, vp, p, id, flags);
4361 	    } else if (op == F_SETLK) {
4362 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4363 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4364 		if (error || donelocally) {
4365 			return (error);
4366 		}
4367 		if (nmp->nm_clp != NULL)
4368 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4369 		else
4370 			clidrev = 0;
4371 		nfhp = VTONFS(vp)->n_fhp;
4372 		if (!lp->nfsl_open->nfso_posixlock &&
4373 		    (off != 0 || len != NFS64BITSSET)) {
4374 			error = EINVAL;
4375 		} else {
4376 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4377 			    nfhp->nfh_len, lp, newone, reclaim, off,
4378 			    len, fl->l_type, cred, p, 0);
4379 		}
4380 		if (!error)
4381 			error = nd->nd_repstat;
4382 		nfscl_lockrelease(lp, error, newone);
4383 	    } else {
4384 		error = EINVAL;
4385 	    }
4386 	    if (!error)
4387 	        error = nd->nd_repstat;
4388 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4389 		error == NFSERR_STALEDONTRECOVER ||
4390 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4391 		error == NFSERR_BADSESSION) {
4392 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4393 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4394 		&& clidrev != 0) {
4395 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4396 		retrycnt++;
4397 	    }
4398 	} while (error == NFSERR_GRACE ||
4399 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4400 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4401 	    error == NFSERR_BADSESSION ||
4402 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4403 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4404 	if (error && retrycnt >= 4)
4405 		error = EIO;
4406 	return (error);
4407 }
4408 
4409 /*
4410  * The lower level routine for the LockT case.
4411  */
4412 int
4413 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4414     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4415     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4416 {
4417 	u_int32_t *tl;
4418 	int error, type, size;
4419 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4420 	struct nfsnode *np;
4421 	struct nfsmount *nmp;
4422 	struct nfsclsession *tsep;
4423 
4424 	nmp = VFSTONFS(vp->v_mount);
4425 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4426 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4427 	if (fl->l_type == F_RDLCK)
4428 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4429 	else
4430 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4431 	txdr_hyper(off, tl);
4432 	tl += 2;
4433 	txdr_hyper(len, tl);
4434 	tl += 2;
4435 	tsep = nfsmnt_mdssession(nmp);
4436 	*tl++ = tsep->nfsess_clientid.lval[0];
4437 	*tl = tsep->nfsess_clientid.lval[1];
4438 	nfscl_filllockowner(id, own, flags);
4439 	np = VTONFS(vp);
4440 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4441 	    np->n_fhp->nfh_len);
4442 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4443 	error = nfscl_request(nd, vp, p, cred);
4444 	if (error)
4445 		return (error);
4446 	if (nd->nd_repstat == 0) {
4447 		fl->l_type = F_UNLCK;
4448 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4449 		nd->nd_repstat = 0;
4450 		fl->l_whence = SEEK_SET;
4451 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4452 		fl->l_start = fxdr_hyper(tl);
4453 		tl += 2;
4454 		len = fxdr_hyper(tl);
4455 		tl += 2;
4456 		if (len == NFS64BITSSET)
4457 			fl->l_len = 0;
4458 		else
4459 			fl->l_len = len;
4460 		type = fxdr_unsigned(int, *tl++);
4461 		if (type == NFSV4LOCKT_WRITE)
4462 			fl->l_type = F_WRLCK;
4463 		else
4464 			fl->l_type = F_RDLCK;
4465 		/*
4466 		 * XXX For now, I have no idea what to do with the
4467 		 * conflicting lock_owner, so I'll just set the pid == 0
4468 		 * and skip over the lock_owner.
4469 		 */
4470 		fl->l_pid = (pid_t)0;
4471 		tl += 2;
4472 		size = fxdr_unsigned(int, *tl);
4473 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4474 			error = EBADRPC;
4475 		if (!error)
4476 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4477 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4478 		nfscl_initiate_recovery(clp);
4479 nfsmout:
4480 	m_freem(nd->nd_mrep);
4481 	return (error);
4482 }
4483 
4484 /*
4485  * Lower level function that performs the LockU RPC.
4486  */
4487 static int
4488 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4489     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4490     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4491 {
4492 	u_int32_t *tl;
4493 	int error;
4494 
4495 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4496 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4497 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4498 	*tl++ = txdr_unsigned(type);
4499 	*tl = txdr_unsigned(lp->nfsl_seqid);
4500 	if (nfstest_outofseq &&
4501 	    (arc4random() % nfstest_outofseq) == 0)
4502 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4503 	tl++;
4504 	if (NFSHASNFSV4N(nmp))
4505 		*tl++ = 0;
4506 	else
4507 		*tl++ = lp->nfsl_stateid.seqid;
4508 	*tl++ = lp->nfsl_stateid.other[0];
4509 	*tl++ = lp->nfsl_stateid.other[1];
4510 	*tl++ = lp->nfsl_stateid.other[2];
4511 	txdr_hyper(off, tl);
4512 	tl += 2;
4513 	txdr_hyper(len, tl);
4514 	if (syscred)
4515 		nd->nd_flag |= ND_USEGSSNAME;
4516 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4517 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4518 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4519 	if (error)
4520 		return (error);
4521 	if (nd->nd_repstat == 0) {
4522 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4523 		lp->nfsl_stateid.seqid = *tl++;
4524 		lp->nfsl_stateid.other[0] = *tl++;
4525 		lp->nfsl_stateid.other[1] = *tl++;
4526 		lp->nfsl_stateid.other[2] = *tl;
4527 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4528 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4529 nfsmout:
4530 	m_freem(nd->nd_mrep);
4531 	return (error);
4532 }
4533 
4534 /*
4535  * The actual Lock RPC.
4536  */
4537 int
4538 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4539     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4540     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4541     NFSPROC_T *p, int syscred)
4542 {
4543 	u_int32_t *tl;
4544 	int error, size;
4545 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4546 	struct nfsclsession *tsep;
4547 
4548 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4549 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4550 	if (type == F_RDLCK)
4551 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4552 	else
4553 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4554 	*tl++ = txdr_unsigned(reclaim);
4555 	txdr_hyper(off, tl);
4556 	tl += 2;
4557 	txdr_hyper(len, tl);
4558 	tl += 2;
4559 	if (newone) {
4560 	    *tl = newnfs_true;
4561 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4562 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4563 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4564 	    if (NFSHASNFSV4N(nmp))
4565 		*tl++ = 0;
4566 	    else
4567 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4568 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4569 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4570 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4571 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4572 	    tsep = nfsmnt_mdssession(nmp);
4573 	    *tl++ = tsep->nfsess_clientid.lval[0];
4574 	    *tl = tsep->nfsess_clientid.lval[1];
4575 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4576 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4577 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4578 	} else {
4579 	    *tl = newnfs_false;
4580 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4581 	    if (NFSHASNFSV4N(nmp))
4582 		*tl++ = 0;
4583 	    else
4584 		*tl++ = lp->nfsl_stateid.seqid;
4585 	    *tl++ = lp->nfsl_stateid.other[0];
4586 	    *tl++ = lp->nfsl_stateid.other[1];
4587 	    *tl++ = lp->nfsl_stateid.other[2];
4588 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4589 	    if (nfstest_outofseq &&
4590 		(arc4random() % nfstest_outofseq) == 0)
4591 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4592 	}
4593 	if (syscred)
4594 		nd->nd_flag |= ND_USEGSSNAME;
4595 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4596 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4597 	if (error)
4598 		return (error);
4599 	if (newone)
4600 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4601 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4602 	if (nd->nd_repstat == 0) {
4603 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4604 		lp->nfsl_stateid.seqid = *tl++;
4605 		lp->nfsl_stateid.other[0] = *tl++;
4606 		lp->nfsl_stateid.other[1] = *tl++;
4607 		lp->nfsl_stateid.other[2] = *tl;
4608 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4609 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4610 		size = fxdr_unsigned(int, *(tl + 7));
4611 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4612 			error = EBADRPC;
4613 		if (!error)
4614 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4615 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4616 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4617 nfsmout:
4618 	m_freem(nd->nd_mrep);
4619 	return (error);
4620 }
4621 
4622 /*
4623  * nfs statfs rpc
4624  * (always called with the vp for the mount point)
4625  */
4626 int
4627 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4628     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4629     void *stuff)
4630 {
4631 	u_int32_t *tl = NULL;
4632 	struct nfsrv_descript nfsd, *nd = &nfsd;
4633 	struct nfsmount *nmp;
4634 	nfsattrbit_t attrbits;
4635 	int error;
4636 
4637 	*attrflagp = 0;
4638 	nmp = VFSTONFS(vp->v_mount);
4639 	if (NFSHASNFSV4(nmp)) {
4640 		/*
4641 		 * For V4, you actually do a getattr.
4642 		 */
4643 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4644 		NFSSTATFS_GETATTRBIT(&attrbits);
4645 		(void) nfsrv_putattrbit(nd, &attrbits);
4646 		nd->nd_flag |= ND_USEGSSNAME;
4647 		error = nfscl_request(nd, vp, p, cred);
4648 		if (error)
4649 			return (error);
4650 		if (nd->nd_repstat == 0) {
4651 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4652 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4653 			    cred);
4654 			if (!error) {
4655 				nmp->nm_fsid[0] = nap->na_filesid[0];
4656 				nmp->nm_fsid[1] = nap->na_filesid[1];
4657 				NFSSETHASSETFSID(nmp);
4658 				*attrflagp = 1;
4659 			}
4660 		} else {
4661 			error = nd->nd_repstat;
4662 		}
4663 		if (error)
4664 			goto nfsmout;
4665 	} else {
4666 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4667 		error = nfscl_request(nd, vp, p, cred);
4668 		if (error)
4669 			return (error);
4670 		if (nd->nd_flag & ND_NFSV3) {
4671 			error = nfscl_postop_attr(nd, nap, attrflagp);
4672 			if (error)
4673 				goto nfsmout;
4674 		}
4675 		if (nd->nd_repstat) {
4676 			error = nd->nd_repstat;
4677 			goto nfsmout;
4678 		}
4679 		NFSM_DISSECT(tl, u_int32_t *,
4680 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4681 	}
4682 	if (NFSHASNFSV3(nmp)) {
4683 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4684 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4685 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4686 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4687 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4688 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4689 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4690 	} else if (NFSHASNFSV4(nmp) == 0) {
4691 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4692 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4693 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4694 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4695 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4696 	}
4697 nfsmout:
4698 	m_freem(nd->nd_mrep);
4699 	return (error);
4700 }
4701 
4702 /*
4703  * nfs pathconf rpc
4704  */
4705 int
4706 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4707     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4708     void *stuff)
4709 {
4710 	struct nfsrv_descript nfsd, *nd = &nfsd;
4711 	struct nfsmount *nmp;
4712 	u_int32_t *tl;
4713 	nfsattrbit_t attrbits;
4714 	int error;
4715 
4716 	*attrflagp = 0;
4717 	nmp = VFSTONFS(vp->v_mount);
4718 	if (NFSHASNFSV4(nmp)) {
4719 		/*
4720 		 * For V4, you actually do a getattr.
4721 		 */
4722 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4723 		NFSPATHCONF_GETATTRBIT(&attrbits);
4724 		(void) nfsrv_putattrbit(nd, &attrbits);
4725 		nd->nd_flag |= ND_USEGSSNAME;
4726 		error = nfscl_request(nd, vp, p, cred);
4727 		if (error)
4728 			return (error);
4729 		if (nd->nd_repstat == 0) {
4730 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4731 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4732 			    cred);
4733 			if (!error)
4734 				*attrflagp = 1;
4735 		} else {
4736 			error = nd->nd_repstat;
4737 		}
4738 	} else {
4739 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4740 		error = nfscl_request(nd, vp, p, cred);
4741 		if (error)
4742 			return (error);
4743 		error = nfscl_postop_attr(nd, nap, attrflagp);
4744 		if (nd->nd_repstat && !error)
4745 			error = nd->nd_repstat;
4746 		if (!error) {
4747 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4748 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4749 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4750 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4751 			pc->pc_chownrestricted =
4752 			    fxdr_unsigned(u_int32_t, *tl++);
4753 			pc->pc_caseinsensitive =
4754 			    fxdr_unsigned(u_int32_t, *tl++);
4755 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4756 		}
4757 	}
4758 nfsmout:
4759 	m_freem(nd->nd_mrep);
4760 	return (error);
4761 }
4762 
4763 /*
4764  * nfs version 3 fsinfo rpc call
4765  */
4766 int
4767 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4768     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4769 {
4770 	u_int32_t *tl;
4771 	struct nfsrv_descript nfsd, *nd = &nfsd;
4772 	int error;
4773 
4774 	*attrflagp = 0;
4775 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4776 	error = nfscl_request(nd, vp, p, cred);
4777 	if (error)
4778 		return (error);
4779 	error = nfscl_postop_attr(nd, nap, attrflagp);
4780 	if (nd->nd_repstat && !error)
4781 		error = nd->nd_repstat;
4782 	if (!error) {
4783 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4784 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4785 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4786 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4787 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4788 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4789 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4790 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4791 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4792 		tl += 2;
4793 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4794 		tl += 2;
4795 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4796 	}
4797 nfsmout:
4798 	m_freem(nd->nd_mrep);
4799 	return (error);
4800 }
4801 
4802 /*
4803  * This function performs the Renew RPC.
4804  */
4805 int
4806 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4807     NFSPROC_T *p)
4808 {
4809 	u_int32_t *tl;
4810 	struct nfsrv_descript nfsd;
4811 	struct nfsrv_descript *nd = &nfsd;
4812 	struct nfsmount *nmp;
4813 	int error;
4814 	struct nfssockreq *nrp;
4815 	struct nfsclsession *tsep;
4816 
4817 	nmp = clp->nfsc_nmp;
4818 	if (nmp == NULL)
4819 		return (0);
4820 	if (dsp == NULL)
4821 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4822 		    0);
4823 	else
4824 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4825 		    &dsp->nfsclds_sess, 0, 0);
4826 	if (!NFSHASNFSV4N(nmp)) {
4827 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4828 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4829 		tsep = nfsmnt_mdssession(nmp);
4830 		*tl++ = tsep->nfsess_clientid.lval[0];
4831 		*tl = tsep->nfsess_clientid.lval[1];
4832 	}
4833 	nrp = NULL;
4834 	if (dsp != NULL)
4835 		nrp = dsp->nfsclds_sockp;
4836 	if (nrp == NULL)
4837 		/* If NULL, use the MDS socket. */
4838 		nrp = &nmp->nm_sockreq;
4839 	nd->nd_flag |= ND_USEGSSNAME;
4840 	if (dsp == NULL)
4841 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4842 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4843 	else {
4844 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4845 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4846 		if (error == ENXIO)
4847 			nfscl_cancelreqs(dsp);
4848 	}
4849 	if (error)
4850 		return (error);
4851 	error = nd->nd_repstat;
4852 	m_freem(nd->nd_mrep);
4853 	return (error);
4854 }
4855 
4856 /*
4857  * This function performs the Releaselockowner RPC.
4858  */
4859 int
4860 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4861     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4862 {
4863 	struct nfsrv_descript nfsd, *nd = &nfsd;
4864 	u_int32_t *tl;
4865 	int error;
4866 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4867 	struct nfsclsession *tsep;
4868 
4869 	if (NFSHASNFSV4N(nmp)) {
4870 		/* For NFSv4.1, do a FreeStateID. */
4871 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4872 		    NULL, 0, 0);
4873 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4874 	} else {
4875 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4876 		    NULL, 0, 0);
4877 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4878 		tsep = nfsmnt_mdssession(nmp);
4879 		*tl++ = tsep->nfsess_clientid.lval[0];
4880 		*tl = tsep->nfsess_clientid.lval[1];
4881 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4882 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4883 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4884 	}
4885 	nd->nd_flag |= ND_USEGSSNAME;
4886 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4887 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4888 	if (error)
4889 		return (error);
4890 	error = nd->nd_repstat;
4891 	m_freem(nd->nd_mrep);
4892 	return (error);
4893 }
4894 
4895 /*
4896  * This function performs the Compound to get the mount pt FH.
4897  */
4898 int
4899 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4900     NFSPROC_T *p)
4901 {
4902 	u_int32_t *tl;
4903 	struct nfsrv_descript nfsd;
4904 	struct nfsrv_descript *nd = &nfsd;
4905 	u_char *cp, *cp2;
4906 	int error, cnt, len, setnil;
4907 	u_int32_t *opcntp;
4908 
4909 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4910 	    0);
4911 	cp = dirpath;
4912 	cnt = 0;
4913 	do {
4914 		setnil = 0;
4915 		while (*cp == '/')
4916 			cp++;
4917 		cp2 = cp;
4918 		while (*cp2 != '\0' && *cp2 != '/')
4919 			cp2++;
4920 		if (*cp2 == '/') {
4921 			setnil = 1;
4922 			*cp2 = '\0';
4923 		}
4924 		if (cp2 != cp) {
4925 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4926 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4927 			nfsm_strtom(nd, cp, strlen(cp));
4928 			cnt++;
4929 		}
4930 		if (setnil)
4931 			*cp2++ = '/';
4932 		cp = cp2;
4933 	} while (*cp != '\0');
4934 	if (NFSHASNFSV4N(nmp))
4935 		/* Has a Sequence Op done by nfscl_reqstart(). */
4936 		*opcntp = txdr_unsigned(3 + cnt);
4937 	else
4938 		*opcntp = txdr_unsigned(2 + cnt);
4939 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4940 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4941 	nd->nd_flag |= ND_USEGSSNAME;
4942 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4943 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4944 	if (error)
4945 		return (error);
4946 	if (nd->nd_repstat == 0) {
4947 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4948 		tl += (2 + 2 * cnt);
4949 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4950 			len > NFSX_FHMAX) {
4951 			nd->nd_repstat = NFSERR_BADXDR;
4952 		} else {
4953 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4954 			if (nd->nd_repstat == 0)
4955 				nmp->nm_fhsize = len;
4956 		}
4957 	}
4958 	error = nd->nd_repstat;
4959 nfsmout:
4960 	m_freem(nd->nd_mrep);
4961 	return (error);
4962 }
4963 
4964 /*
4965  * This function performs the Delegreturn RPC.
4966  */
4967 int
4968 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4969     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4970 {
4971 	u_int32_t *tl;
4972 	struct nfsrv_descript nfsd;
4973 	struct nfsrv_descript *nd = &nfsd;
4974 	int error;
4975 
4976 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4977 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4978 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4979 	if (NFSHASNFSV4N(nmp))
4980 		*tl++ = 0;
4981 	else
4982 		*tl++ = dp->nfsdl_stateid.seqid;
4983 	*tl++ = dp->nfsdl_stateid.other[0];
4984 	*tl++ = dp->nfsdl_stateid.other[1];
4985 	*tl = dp->nfsdl_stateid.other[2];
4986 	if (syscred)
4987 		nd->nd_flag |= ND_USEGSSNAME;
4988 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4989 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4990 	if (error)
4991 		return (error);
4992 	error = nd->nd_repstat;
4993 	m_freem(nd->nd_mrep);
4994 	return (error);
4995 }
4996 
4997 /*
4998  * nfs getacl call.
4999  */
5000 int
5001 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
5002     struct acl *aclp, void *stuff)
5003 {
5004 	struct nfsrv_descript nfsd, *nd = &nfsd;
5005 	int error;
5006 	nfsattrbit_t attrbits;
5007 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5008 
5009 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5010 		return (EOPNOTSUPP);
5011 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
5012 	NFSZERO_ATTRBIT(&attrbits);
5013 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5014 	(void) nfsrv_putattrbit(nd, &attrbits);
5015 	error = nfscl_request(nd, vp, p, cred);
5016 	if (error)
5017 		return (error);
5018 	if (!nd->nd_repstat)
5019 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
5020 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
5021 	else
5022 		error = nd->nd_repstat;
5023 	m_freem(nd->nd_mrep);
5024 	return (error);
5025 }
5026 
5027 /*
5028  * nfs setacl call.
5029  */
5030 int
5031 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
5032     struct acl *aclp, void *stuff)
5033 {
5034 	int error;
5035 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5036 
5037 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5038 		return (EOPNOTSUPP);
5039 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
5040 	return (error);
5041 }
5042 
5043 /*
5044  * nfs setacl call.
5045  */
5046 static int
5047 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
5048     struct acl *aclp, nfsv4stateid_t *stateidp)
5049 {
5050 	struct nfsrv_descript nfsd, *nd = &nfsd;
5051 	int error;
5052 	nfsattrbit_t attrbits;
5053 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5054 
5055 	if (!NFSHASNFSV4(nmp))
5056 		return (EOPNOTSUPP);
5057 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
5058 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
5059 	NFSZERO_ATTRBIT(&attrbits);
5060 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5061 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
5062 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
5063 	error = nfscl_request(nd, vp, p, cred);
5064 	if (error)
5065 		return (error);
5066 	/* Don't care about the pre/postop attributes */
5067 	m_freem(nd->nd_mrep);
5068 	return (nd->nd_repstat);
5069 }
5070 
5071 /*
5072  * Do the NFSv4.1 Exchange ID.
5073  */
5074 int
5075 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
5076     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
5077     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
5078 {
5079 	uint32_t *tl, v41flags;
5080 	struct nfsrv_descript nfsd;
5081 	struct nfsrv_descript *nd = &nfsd;
5082 	struct nfsclds *dsp;
5083 	struct timespec verstime;
5084 	int error, len;
5085 
5086 	*dspp = NULL;
5087 	if (minorvers == 0)
5088 		minorvers = nmp->nm_minorvers;
5089 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
5090 	    NFS_VER4, minorvers);
5091 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5092 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
5093 	*tl = txdr_unsigned(clp->nfsc_rev);
5094 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
5095 
5096 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
5097 	*tl++ = txdr_unsigned(exchflags);
5098 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
5099 
5100 	/* Set the implementation id4 */
5101 	*tl = txdr_unsigned(1);
5102 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
5103 	(void) nfsm_strtom(nd, version, strlen(version));
5104 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
5105 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
5106 	verstime.tv_nsec = 0;
5107 	txdr_nfsv4time(&verstime, tl);
5108 	nd->nd_flag |= ND_USEGSSNAME;
5109 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5110 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5111 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
5112 	    (int)nd->nd_repstat);
5113 	if (error != 0)
5114 		return (error);
5115 	if (nd->nd_repstat == 0) {
5116 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
5117 		len = fxdr_unsigned(int, *(tl + 7));
5118 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
5119 			error = NFSERR_BADXDR;
5120 			goto nfsmout;
5121 		}
5122 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
5123 		    M_WAITOK | M_ZERO);
5124 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
5125 		dsp->nfsclds_servownlen = len;
5126 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
5127 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
5128 		dsp->nfsclds_sess.nfsess_sequenceid =
5129 		    fxdr_unsigned(uint32_t, *tl++);
5130 		v41flags = fxdr_unsigned(uint32_t, *tl);
5131 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
5132 		    NFSHASPNFSOPT(nmp)) {
5133 			NFSCL_DEBUG(1, "set PNFS\n");
5134 			NFSLOCKMNT(nmp);
5135 			nmp->nm_state |= NFSSTA_PNFS;
5136 			NFSUNLOCKMNT(nmp);
5137 			dsp->nfsclds_flags |= NFSCLDS_MDS;
5138 		}
5139 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
5140 			dsp->nfsclds_flags |= NFSCLDS_DS;
5141 		if (minorvers == NFSV42_MINORVERSION)
5142 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
5143 		if (len > 0)
5144 			nd->nd_repstat = nfsrv_mtostr(nd,
5145 			    dsp->nfsclds_serverown, len);
5146 		if (nd->nd_repstat == 0) {
5147 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5148 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5149 			    NULL, MTX_DEF);
5150 			nfscl_initsessionslots(&dsp->nfsclds_sess);
5151 			*dspp = dsp;
5152 		} else
5153 			free(dsp, M_NFSCLDS);
5154 	}
5155 	error = nd->nd_repstat;
5156 nfsmout:
5157 	m_freem(nd->nd_mrep);
5158 	return (error);
5159 }
5160 
5161 /*
5162  * Do the NFSv4.1 Create Session.
5163  */
5164 int
5165 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
5166     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
5167     struct ucred *cred, NFSPROC_T *p)
5168 {
5169 	uint32_t crflags, maxval, *tl;
5170 	struct nfsrv_descript nfsd;
5171 	struct nfsrv_descript *nd = &nfsd;
5172 	int error, irdcnt, minorvers;
5173 
5174 	/* Make sure nm_rsize, nm_wsize is set. */
5175 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
5176 		nmp->nm_rsize = NFS_MAXBSIZE;
5177 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
5178 		nmp->nm_wsize = NFS_MAXBSIZE;
5179 	if (dsp == NULL)
5180 		minorvers = nmp->nm_minorvers;
5181 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
5182 		minorvers = NFSV42_MINORVERSION;
5183 	else
5184 		minorvers = NFSV41_MINORVERSION;
5185 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
5186 	    NFS_VER4, minorvers);
5187 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5188 	*tl++ = sep->nfsess_clientid.lval[0];
5189 	*tl++ = sep->nfsess_clientid.lval[1];
5190 	*tl++ = txdr_unsigned(sequenceid);
5191 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
5192 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
5193 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
5194 	*tl = txdr_unsigned(crflags);
5195 
5196 	/* Fill in fore channel attributes. */
5197 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5198 	*tl++ = 0;				/* Header pad size */
5199 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
5200 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
5201 		/*
5202 		 * NFSv4.2 Extended Attribute operations may want to do
5203 		 * requests/replies that are larger than nm_rsize/nm_wsize.
5204 		 */
5205 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5206 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5207 	} else {
5208 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
5209 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
5210 	}
5211 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5212 	*tl++ = txdr_unsigned(20);		/* Max operations */
5213 	*tl++ = txdr_unsigned(64);		/* Max slots */
5214 	*tl = 0;				/* No rdma ird */
5215 
5216 	/* Fill in back channel attributes. */
5217 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5218 	*tl++ = 0;				/* Header pad size */
5219 	*tl++ = txdr_unsigned(10000);		/* Max request size */
5220 	*tl++ = txdr_unsigned(10000);		/* Max response size */
5221 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5222 	*tl++ = txdr_unsigned(4);		/* Max operations */
5223 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
5224 	*tl = 0;				/* No rdma ird */
5225 
5226 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
5227 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
5228 
5229 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
5230 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
5231 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
5232 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
5233 	*tl++ = 0;				/* Null machine name */
5234 	*tl++ = 0;				/* Uid == 0 */
5235 	*tl++ = 0;				/* Gid == 0 */
5236 	*tl = 0;				/* No additional gids */
5237 	nd->nd_flag |= ND_USEGSSNAME;
5238 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
5239 	    NFS_VER4, NULL, 1, NULL, NULL);
5240 	if (error != 0)
5241 		return (error);
5242 	if (nd->nd_repstat == 0) {
5243 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
5244 		    2 * NFSX_UNSIGNED);
5245 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
5246 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
5247 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
5248 		crflags = fxdr_unsigned(uint32_t, *tl);
5249 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
5250 			NFSLOCKMNT(nmp);
5251 			nmp->nm_state |= NFSSTA_SESSPERSIST;
5252 			NFSUNLOCKMNT(nmp);
5253 		}
5254 
5255 		/* Get the fore channel slot count. */
5256 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5257 		tl++;			/* Skip the header pad size. */
5258 
5259 		/* Make sure nm_wsize is small enough. */
5260 		maxval = fxdr_unsigned(uint32_t, *tl++);
5261 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
5262 			if (nmp->nm_wsize > 8096)
5263 				nmp->nm_wsize /= 2;
5264 			else
5265 				break;
5266 		}
5267 		sep->nfsess_maxreq = maxval;
5268 
5269 		/* Make sure nm_rsize is small enough. */
5270 		maxval = fxdr_unsigned(uint32_t, *tl++);
5271 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
5272 			if (nmp->nm_rsize > 8096)
5273 				nmp->nm_rsize /= 2;
5274 			else
5275 				break;
5276 		}
5277 		sep->nfsess_maxresp = maxval;
5278 
5279 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
5280 		tl++;
5281 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
5282 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
5283 		irdcnt = fxdr_unsigned(int, *tl);
5284 		if (irdcnt < 0 || irdcnt > 1) {
5285 			error = NFSERR_BADXDR;
5286 			goto nfsmout;
5287 		}
5288 		if (irdcnt > 0)
5289 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
5290 
5291 		/* and the back channel slot count. */
5292 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5293 		tl += 5;
5294 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
5295 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
5296 	}
5297 	error = nd->nd_repstat;
5298 nfsmout:
5299 	m_freem(nd->nd_mrep);
5300 	return (error);
5301 }
5302 
5303 /*
5304  * Do the NFSv4.1 Destroy Session.
5305  */
5306 int
5307 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
5308     struct ucred *cred, NFSPROC_T *p)
5309 {
5310 	uint32_t *tl;
5311 	struct nfsrv_descript nfsd;
5312 	struct nfsrv_descript *nd = &nfsd;
5313 	int error;
5314 	struct nfsclsession *tsep;
5315 
5316 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
5317 	    0);
5318 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
5319 	tsep = nfsmnt_mdssession(nmp);
5320 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
5321 	nd->nd_flag |= ND_USEGSSNAME;
5322 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5323 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5324 	if (error != 0)
5325 		return (error);
5326 	error = nd->nd_repstat;
5327 	m_freem(nd->nd_mrep);
5328 	return (error);
5329 }
5330 
5331 /*
5332  * Do the NFSv4.1 Destroy Client.
5333  */
5334 int
5335 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5336     struct ucred *cred, NFSPROC_T *p)
5337 {
5338 	uint32_t *tl;
5339 	struct nfsrv_descript nfsd;
5340 	struct nfsrv_descript *nd = &nfsd;
5341 	int error;
5342 	struct nfsclsession *tsep;
5343 
5344 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5345 	    0);
5346 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5347 	tsep = nfsmnt_mdssession(nmp);
5348 	*tl++ = tsep->nfsess_clientid.lval[0];
5349 	*tl = tsep->nfsess_clientid.lval[1];
5350 	nd->nd_flag |= ND_USEGSSNAME;
5351 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5352 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5353 	if (error != 0)
5354 		return (error);
5355 	error = nd->nd_repstat;
5356 	m_freem(nd->nd_mrep);
5357 	return (error);
5358 }
5359 
5360 /*
5361  * Do the NFSv4.1 LayoutGet.
5362  */
5363 static int
5364 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5365     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5366     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5367     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
5368     void *stuff)
5369 {
5370 	struct nfsrv_descript nfsd, *nd = &nfsd;
5371 	int error;
5372 
5373 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5374 	    0);
5375 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5376 	    layouttype, layoutlen, 0);
5377 	nd->nd_flag |= ND_USEGSSNAME;
5378 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5379 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5380 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5381 	if (error != 0)
5382 		return (error);
5383 	if (nd->nd_repstat == 0)
5384 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5385 		    flhp);
5386 	if (error == 0 && nd->nd_repstat != 0)
5387 		error = nd->nd_repstat;
5388 	m_freem(nd->nd_mrep);
5389 	return (error);
5390 }
5391 
5392 /*
5393  * Do the NFSv4.1 Get Device Info.
5394  */
5395 int
5396 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5397     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5398     NFSPROC_T *p)
5399 {
5400 	uint32_t cnt, *tl, vers, minorvers;
5401 	struct nfsrv_descript nfsd;
5402 	struct nfsrv_descript *nd = &nfsd;
5403 	struct sockaddr_in sin, ssin;
5404 	struct sockaddr_in6 sin6, ssin6;
5405 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5406 	struct nfscldevinfo *ndi;
5407 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5408 	int stripecnt;
5409 	uint8_t stripeindex;
5410 	sa_family_t af, safilled;
5411 
5412 	ssin.sin_port = 0;		/* To shut up compiler. */
5413 	ssin.sin_addr.s_addr = 0;	/* ditto */
5414 	*ndip = NULL;
5415 	ndi = NULL;
5416 	gotdspp = NULL;
5417 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5418 	    0);
5419 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5420 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5421 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5422 	*tl++ = txdr_unsigned(layouttype);
5423 	*tl++ = txdr_unsigned(100000);
5424 	if (notifybitsp != NULL && *notifybitsp != 0) {
5425 		*tl = txdr_unsigned(1);		/* One word of bits. */
5426 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5427 		*tl = txdr_unsigned(*notifybitsp);
5428 	} else
5429 		*tl = txdr_unsigned(0);
5430 	nd->nd_flag |= ND_USEGSSNAME;
5431 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5432 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5433 	if (error != 0)
5434 		return (error);
5435 	if (nd->nd_repstat == 0) {
5436 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5437 		if (layouttype != fxdr_unsigned(int, *tl))
5438 			printf("EEK! devinfo layout type not same!\n");
5439 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5440 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5441 			stripecnt = fxdr_unsigned(int, *tl);
5442 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5443 			if (stripecnt < 1 || stripecnt > 4096) {
5444 				printf("pNFS File layout devinfo stripecnt %d:"
5445 				    " out of range\n", stripecnt);
5446 				error = NFSERR_BADXDR;
5447 				goto nfsmout;
5448 			}
5449 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5450 			    NFSX_UNSIGNED);
5451 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5452 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5453 			if (addrcnt < 1 || addrcnt > 128) {
5454 				printf("NFS devinfo addrcnt %d: out of range\n",
5455 				    addrcnt);
5456 				error = NFSERR_BADXDR;
5457 				goto nfsmout;
5458 			}
5459 
5460 			/*
5461 			 * Now we know how many stripe indices and addresses, so
5462 			 * we can allocate the structure the correct size.
5463 			 */
5464 			i = (stripecnt * sizeof(uint8_t)) /
5465 			    sizeof(struct nfsclds *) + 1;
5466 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5467 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5468 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5469 			    M_ZERO);
5470 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5471 			    NFSX_V4DEVICEID);
5472 			ndi->nfsdi_refcnt = 0;
5473 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5474 			ndi->nfsdi_stripecnt = stripecnt;
5475 			ndi->nfsdi_addrcnt = addrcnt;
5476 			/* Fill in the stripe indices. */
5477 			for (i = 0; i < stripecnt; i++) {
5478 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5479 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5480 				if (stripeindex >= addrcnt) {
5481 					printf("pNFS File Layout devinfo"
5482 					    " stripeindex %d: too big\n",
5483 					    (int)stripeindex);
5484 					error = NFSERR_BADXDR;
5485 					goto nfsmout;
5486 				}
5487 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5488 			}
5489 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5490 			/* For Flex File, we only get one address list. */
5491 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5492 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5493 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5494 			    NFSX_V4DEVICEID);
5495 			ndi->nfsdi_refcnt = 0;
5496 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5497 			addrcnt = ndi->nfsdi_addrcnt = 1;
5498 		}
5499 
5500 		/* Now, dissect the server address(es). */
5501 		safilled = AF_UNSPEC;
5502 		for (i = 0; i < addrcnt; i++) {
5503 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5504 			cnt = fxdr_unsigned(uint32_t, *tl);
5505 			if (cnt == 0) {
5506 				printf("NFS devinfo 0 len addrlist\n");
5507 				error = NFSERR_BADXDR;
5508 				goto nfsmout;
5509 			}
5510 			dspp = nfsfldi_addr(ndi, i);
5511 			safilled = AF_UNSPEC;
5512 			for (j = 0; j < cnt; j++) {
5513 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5514 				    &isudp);
5515 				if (error != 0 && error != EPERM) {
5516 					error = NFSERR_BADXDR;
5517 					goto nfsmout;
5518 				}
5519 				if (error == 0 && isudp == 0) {
5520 					/*
5521 					 * The priority is:
5522 					 * - Same address family.
5523 					 * Save the address and dspp, so that
5524 					 * the connection can be done after
5525 					 * parsing is complete.
5526 					 */
5527 					if (safilled == AF_UNSPEC ||
5528 					    (af == nmp->nm_nam->sa_family &&
5529 					     safilled != nmp->nm_nam->sa_family)
5530 					   ) {
5531 						if (af == AF_INET)
5532 							ssin = sin;
5533 						else
5534 							ssin6 = sin6;
5535 						safilled = af;
5536 						gotdspp = dspp;
5537 					}
5538 				}
5539 			}
5540 		}
5541 
5542 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5543 		gotminor = NFSV41_MINORVERSION;
5544 		/* For Flex File, we will take one of the versions to use. */
5545 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5546 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5547 			j = fxdr_unsigned(int, *tl);
5548 			if (j < 1 || j > NFSDEV_MAXVERS) {
5549 				printf("pNFS: too many versions\n");
5550 				error = NFSERR_BADXDR;
5551 				goto nfsmout;
5552 			}
5553 			gotvers = 0;
5554 			gotminor = 0;
5555 			for (i = 0; i < j; i++) {
5556 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5557 				vers = fxdr_unsigned(uint32_t, *tl++);
5558 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5559 				if (vers == NFS_VER3)
5560 					minorvers = 0;
5561 				if ((vers == NFS_VER4 && ((minorvers ==
5562 				    NFSV41_MINORVERSION && gotminor == 0) ||
5563 				    minorvers == NFSV42_MINORVERSION)) ||
5564 				    (vers == NFS_VER3 && gotvers == 0)) {
5565 					gotvers = vers;
5566 					gotminor = minorvers;
5567 					/* We'll take this one. */
5568 					ndi->nfsdi_versindex = i;
5569 					ndi->nfsdi_vers = vers;
5570 					ndi->nfsdi_minorvers = minorvers;
5571 					ndi->nfsdi_rsize = fxdr_unsigned(
5572 					    uint32_t, *tl++);
5573 					ndi->nfsdi_wsize = fxdr_unsigned(
5574 					    uint32_t, *tl++);
5575 					if (*tl == newnfs_true)
5576 						ndi->nfsdi_flags |=
5577 						    NFSDI_TIGHTCOUPLED;
5578 					else
5579 						ndi->nfsdi_flags &=
5580 						    ~NFSDI_TIGHTCOUPLED;
5581 				}
5582 			}
5583 			if (gotvers == 0) {
5584 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5585 				error = NFSERR_BADXDR;
5586 				goto nfsmout;
5587 			}
5588 		}
5589 
5590 		/* And the notify bits. */
5591 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5592 		bitcnt = fxdr_unsigned(int, *tl);
5593 		if (bitcnt > 0) {
5594 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5595 			if (notifybitsp != NULL)
5596 				*notifybitsp =
5597 				    fxdr_unsigned(uint32_t, *tl);
5598 		}
5599 		if (safilled != AF_UNSPEC) {
5600 			KASSERT(ndi != NULL, ("ndi is NULL"));
5601 			*ndip = ndi;
5602 		} else
5603 			error = EPERM;
5604 		if (error == 0) {
5605 			/*
5606 			 * Now we can do a TCP connection for the correct
5607 			 * NFS version and IP address.
5608 			 */
5609 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5610 			    gotvers, gotminor, &dsp, p);
5611 		}
5612 		if (error == 0) {
5613 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5614 			*gotdspp = dsp;
5615 		}
5616 	}
5617 	if (nd->nd_repstat != 0 && error == 0)
5618 		error = nd->nd_repstat;
5619 nfsmout:
5620 	if (error != 0 && ndi != NULL)
5621 		nfscl_freedevinfo(ndi);
5622 	m_freem(nd->nd_mrep);
5623 	return (error);
5624 }
5625 
5626 /*
5627  * Do the NFSv4.1 LayoutCommit.
5628  */
5629 int
5630 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5631     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5632     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5633 {
5634 	uint32_t *tl;
5635 	struct nfsrv_descript nfsd, *nd = &nfsd;
5636 	int error;
5637 
5638 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5639 	    0, 0);
5640 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5641 	    NFSX_STATEID);
5642 	txdr_hyper(off, tl);
5643 	tl += 2;
5644 	txdr_hyper(len, tl);
5645 	tl += 2;
5646 	if (reclaim != 0)
5647 		*tl++ = newnfs_true;
5648 	else
5649 		*tl++ = newnfs_false;
5650 	*tl++ = txdr_unsigned(stateidp->seqid);
5651 	*tl++ = stateidp->other[0];
5652 	*tl++ = stateidp->other[1];
5653 	*tl++ = stateidp->other[2];
5654 	*tl++ = newnfs_true;
5655 	if (lastbyte < off)
5656 		lastbyte = off;
5657 	else if (lastbyte >= (off + len))
5658 		lastbyte = off + len - 1;
5659 	txdr_hyper(lastbyte, tl);
5660 	tl += 2;
5661 	*tl++ = newnfs_false;
5662 	*tl++ = txdr_unsigned(layouttype);
5663 	/* All supported layouts are 0 length. */
5664 	*tl = txdr_unsigned(0);
5665 	nd->nd_flag |= ND_USEGSSNAME;
5666 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5667 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5668 	if (error != 0)
5669 		return (error);
5670 	error = nd->nd_repstat;
5671 	m_freem(nd->nd_mrep);
5672 	return (error);
5673 }
5674 
5675 /*
5676  * Do the NFSv4.1 LayoutReturn.
5677  */
5678 int
5679 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5680     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5681     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5682     uint32_t stat, uint32_t op, char *devid)
5683 {
5684 	uint32_t *tl;
5685 	struct nfsrv_descript nfsd, *nd = &nfsd;
5686 	uint64_t tu64;
5687 	int error;
5688 
5689 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5690 	    0, 0);
5691 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5692 	if (reclaim != 0)
5693 		*tl++ = newnfs_true;
5694 	else
5695 		*tl++ = newnfs_false;
5696 	*tl++ = txdr_unsigned(layouttype);
5697 	*tl++ = txdr_unsigned(iomode);
5698 	*tl = txdr_unsigned(layoutreturn);
5699 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5700 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5701 		    NFSX_UNSIGNED);
5702 		txdr_hyper(offset, tl);
5703 		tl += 2;
5704 		txdr_hyper(len, tl);
5705 		tl += 2;
5706 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5707 		*tl++ = txdr_unsigned(stateidp->seqid);
5708 		*tl++ = stateidp->other[0];
5709 		*tl++ = stateidp->other[1];
5710 		*tl++ = stateidp->other[2];
5711 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5712 			*tl = txdr_unsigned(0);
5713 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5714 			if (stat != 0) {
5715 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5716 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5717 				    NFSX_UNSIGNED);
5718 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5719 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5720 				    NFSX_UNSIGNED);
5721 				*tl++ = txdr_unsigned(1);	/* One error. */
5722 				tu64 = 0;			/* Offset. */
5723 				txdr_hyper(tu64, tl); tl += 2;
5724 				tu64 = UINT64_MAX;		/* Length. */
5725 				txdr_hyper(tu64, tl); tl += 2;
5726 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5727 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5728 				*tl++ = txdr_unsigned(1);	/* One error. */
5729 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5730 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5731 				*tl++ = txdr_unsigned(stat);
5732 				*tl++ = txdr_unsigned(op);
5733 			} else {
5734 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5735 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5736 				/* No ioerrs. */
5737 				*tl++ = 0;
5738 			}
5739 			*tl = 0;	/* No stats yet. */
5740 		}
5741 	}
5742 	nd->nd_flag |= ND_USEGSSNAME;
5743 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5744 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5745 	if (error != 0)
5746 		return (error);
5747 	if (nd->nd_repstat == 0) {
5748 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5749 		if (*tl != 0) {
5750 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5751 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5752 			stateidp->other[0] = *tl++;
5753 			stateidp->other[1] = *tl++;
5754 			stateidp->other[2] = *tl;
5755 		}
5756 	} else
5757 		error = nd->nd_repstat;
5758 nfsmout:
5759 	m_freem(nd->nd_mrep);
5760 	return (error);
5761 }
5762 
5763 /*
5764  * Do the NFSv4.2 LayoutError.
5765  */
5766 static int
5767 nfsrpc_layouterror(struct nfsmount *nmp, uint8_t *fh, int fhlen, uint64_t offset,
5768     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5769     uint32_t stat, uint32_t op, char *devid)
5770 {
5771 	uint32_t *tl;
5772 	struct nfsrv_descript nfsd, *nd = &nfsd;
5773 	int error;
5774 
5775 	nfscl_reqstart(nd, NFSPROC_LAYOUTERROR, nmp, fh, fhlen, NULL, NULL,
5776 	    0, 0);
5777 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5778 	    NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5779 	txdr_hyper(offset, tl); tl += 2;
5780 	txdr_hyper(len, tl); tl += 2;
5781 	*tl++ = txdr_unsigned(stateidp->seqid);
5782 	*tl++ = stateidp->other[0];
5783 	*tl++ = stateidp->other[1];
5784 	*tl++ = stateidp->other[2];
5785 	*tl++ = txdr_unsigned(1);
5786 	NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5787 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5788 	*tl++ = txdr_unsigned(stat);
5789 	*tl = txdr_unsigned(op);
5790 	nd->nd_flag |= ND_USEGSSNAME;
5791 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5792 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5793 	if (error != 0)
5794 		return (error);
5795 	if (nd->nd_repstat != 0)
5796 		error = nd->nd_repstat;
5797 	m_freem(nd->nd_mrep);
5798 	return (error);
5799 }
5800 
5801 /*
5802  * Acquire a layout and devinfo, if possible. The caller must have acquired
5803  * a reference count on the nfsclclient structure before calling this.
5804  * Return the layout in lypp with a reference count on it, if successful.
5805  */
5806 static int
5807 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5808     int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
5809     uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5810 {
5811 	struct nfscllayout *lyp;
5812 	struct nfsclflayout *flp;
5813 	struct nfsclflayouthead flh;
5814 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5815 	nfsv4stateid_t stateid;
5816 	struct nfsclsession *tsep;
5817 
5818 	*lypp = NULL;
5819 	if (NFSHASFLEXFILE(nmp))
5820 		layouttype = NFSLAYOUT_FLEXFILE;
5821 	else
5822 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5823 	/*
5824 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5825 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5826 	 * flp == NULL.
5827 	 */
5828 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5829 	    off, rw, &flp, &recalled);
5830 	islocked = 0;
5831 	if (lyp == NULL || flp == NULL) {
5832 		if (recalled != 0)
5833 			return (EIO);
5834 		LIST_INIT(&flh);
5835 		tsep = nfsmnt_mdssession(nmp);
5836 		layoutlen = tsep->nfsess_maxcache -
5837 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5838 		if (lyp == NULL) {
5839 			stateid.seqid = 0;
5840 			stateid.other[0] = stateidp->other[0];
5841 			stateid.other[1] = stateidp->other[1];
5842 			stateid.other[2] = stateidp->other[2];
5843 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5844 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5845 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5846 			    &retonclose, &flh, cred, p, NULL);
5847 		} else {
5848 			islocked = 1;
5849 			stateid.seqid = lyp->nfsly_stateid.seqid;
5850 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5851 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5852 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5853 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5854 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5855 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5856 			    &retonclose, &flh, cred, p, NULL);
5857 		}
5858 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5859 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5860 		    &flh, layouttype, error, NULL, cred, p);
5861 		if (error == 0)
5862 			*lypp = lyp;
5863 		else if (islocked != 0)
5864 			nfscl_rellayout(lyp, 1);
5865 	} else
5866 		*lypp = lyp;
5867 	return (error);
5868 }
5869 
5870 /*
5871  * Do a TCP connection plus exchange id and create session.
5872  * If successful, a "struct nfsclds" is linked into the list for the
5873  * mount point and a pointer to it is returned.
5874  */
5875 static int
5876 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5877     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5878     struct nfsclds **dspp, NFSPROC_T *p)
5879 {
5880 	struct sockaddr_in *msad, *sad;
5881 	struct sockaddr_in6 *msad6, *sad6;
5882 	struct nfsclclient *clp;
5883 	struct nfssockreq *nrp;
5884 	struct nfsclds *dsp, *tdsp;
5885 	int error, firsttry;
5886 	enum nfsclds_state retv;
5887 	uint32_t sequenceid = 0;
5888 
5889 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5890 	    ("nfsrpc_fillsa: NULL nr_cred"));
5891 	NFSLOCKCLSTATE();
5892 	clp = nmp->nm_clp;
5893 	NFSUNLOCKCLSTATE();
5894 	if (clp == NULL)
5895 		return (EPERM);
5896 	if (af == AF_INET) {
5897 		NFSLOCKMNT(nmp);
5898 		/*
5899 		 * Check to see if we already have a session for this
5900 		 * address that is usable for a DS.
5901 		 * Note that the MDS's address is in a different place
5902 		 * than the sessions already acquired for DS's.
5903 		 */
5904 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5905 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5906 		while (tdsp != NULL) {
5907 			if (msad != NULL && msad->sin_family == AF_INET &&
5908 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5909 			    sin->sin_port == msad->sin_port &&
5910 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5911 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5912 				*dspp = tdsp;
5913 				NFSUNLOCKMNT(nmp);
5914 				NFSCL_DEBUG(4, "fnd same addr\n");
5915 				return (0);
5916 			}
5917 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5918 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5919 				msad = (struct sockaddr_in *)
5920 				    tdsp->nfsclds_sockp->nr_nam;
5921 			else
5922 				msad = NULL;
5923 		}
5924 		NFSUNLOCKMNT(nmp);
5925 
5926 		/* No IP address match, so look for new/trunked one. */
5927 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5928 		sad->sin_len = sizeof(*sad);
5929 		sad->sin_family = AF_INET;
5930 		sad->sin_port = sin->sin_port;
5931 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5932 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5933 		nrp->nr_nam = (struct sockaddr *)sad;
5934 	} else if (af == AF_INET6) {
5935 		NFSLOCKMNT(nmp);
5936 		/*
5937 		 * Check to see if we already have a session for this
5938 		 * address that is usable for a DS.
5939 		 * Note that the MDS's address is in a different place
5940 		 * than the sessions already acquired for DS's.
5941 		 */
5942 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5943 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5944 		while (tdsp != NULL) {
5945 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5946 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5947 			    &msad6->sin6_addr) &&
5948 			    sin6->sin6_port == msad6->sin6_port &&
5949 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5950 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5951 				*dspp = tdsp;
5952 				NFSUNLOCKMNT(nmp);
5953 				return (0);
5954 			}
5955 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5956 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5957 				msad6 = (struct sockaddr_in6 *)
5958 				    tdsp->nfsclds_sockp->nr_nam;
5959 			else
5960 				msad6 = NULL;
5961 		}
5962 		NFSUNLOCKMNT(nmp);
5963 
5964 		/* No IP address match, so look for new/trunked one. */
5965 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5966 		sad6->sin6_len = sizeof(*sad6);
5967 		sad6->sin6_family = AF_INET6;
5968 		sad6->sin6_port = sin6->sin6_port;
5969 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5970 		    sizeof(struct in6_addr));
5971 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5972 		nrp->nr_nam = (struct sockaddr *)sad6;
5973 	} else
5974 		return (EPERM);
5975 
5976 	nrp->nr_sotype = SOCK_STREAM;
5977 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5978 	nrp->nr_prog = NFS_PROG;
5979 	nrp->nr_vers = vers;
5980 
5981 	/*
5982 	 * Use the credentials that were used for the mount, which are
5983 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5984 	 * Ref. counting the credentials with crhold() is probably not
5985 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5986 	 * unmount, but I did it anyhow.
5987 	 */
5988 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5989 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false, &nrp->nr_client);
5990 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5991 
5992 	dsp = NULL;
5993 	/* Now, do the exchangeid and create session. */
5994 	if (error == 0) {
5995 		if (vers == NFS_VER4) {
5996 			firsttry = 0;
5997 			do {
5998 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5999 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
6000 				    nrp->nr_cred, p);
6001 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
6002 				if (error == NFSERR_MINORVERMISMATCH)
6003 					minorvers = NFSV42_MINORVERSION;
6004 			} while (error == NFSERR_MINORVERMISMATCH &&
6005 			    firsttry++ == 0);
6006 			if (error != 0)
6007 				newnfs_disconnect(NULL, nrp);
6008 		} else {
6009 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
6010 			    M_WAITOK | M_ZERO);
6011 			dsp->nfsclds_flags |= NFSCLDS_DS;
6012 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
6013 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
6014 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
6015 			    NULL, MTX_DEF);
6016 		}
6017 	}
6018 	if (error == 0) {
6019 		dsp->nfsclds_sockp = nrp;
6020 		if (vers == NFS_VER4) {
6021 			NFSLOCKMNT(nmp);
6022 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
6023 			    &sequenceid);
6024 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
6025 			if (retv == NFSDSP_USETHISSESSION &&
6026 			    nfscl_dssameconn != 0) {
6027 				NFSLOCKDS(tdsp);
6028 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
6029 				NFSUNLOCKDS(tdsp);
6030 				NFSUNLOCKMNT(nmp);
6031 				/*
6032 				 * If there is already a session for this
6033 				 * server, use it.
6034 				 */
6035 				newnfs_disconnect(NULL, nrp);
6036 				nfscl_freenfsclds(dsp);
6037 				*dspp = tdsp;
6038 				return (0);
6039 			}
6040 			if (retv == NFSDSP_NOTFOUND)
6041 				sequenceid =
6042 				    dsp->nfsclds_sess.nfsess_sequenceid;
6043 			NFSUNLOCKMNT(nmp);
6044 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
6045 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
6046 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
6047 		}
6048 	} else {
6049 		NFSFREECRED(nrp->nr_cred);
6050 		NFSFREEMUTEX(&nrp->nr_mtx);
6051 		free(nrp->nr_nam, M_SONAME);
6052 		free(nrp, M_NFSSOCKREQ);
6053 	}
6054 	if (error == 0) {
6055 		NFSCL_DEBUG(3, "add DS session\n");
6056 		/*
6057 		 * Put it at the end of the list. That way the list
6058 		 * is ordered by when the entry was added. This matters
6059 		 * since the one done first is the one that should be
6060 		 * used for sequencid'ing any subsequent create sessions.
6061 		 */
6062 		NFSLOCKMNT(nmp);
6063 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
6064 		NFSUNLOCKMNT(nmp);
6065 		*dspp = dsp;
6066 	} else if (dsp != NULL) {
6067 		newnfs_disconnect(NULL, nrp);
6068 		nfscl_freenfsclds(dsp);
6069 	}
6070 	return (error);
6071 }
6072 
6073 /*
6074  * Do the NFSv4.1 Reclaim Complete.
6075  */
6076 int
6077 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
6078 {
6079 	uint32_t *tl;
6080 	struct nfsrv_descript nfsd;
6081 	struct nfsrv_descript *nd = &nfsd;
6082 	int error;
6083 
6084 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
6085 	    0);
6086 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
6087 	*tl = newnfs_false;
6088 	nd->nd_flag |= ND_USEGSSNAME;
6089 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
6090 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
6091 	if (error != 0)
6092 		return (error);
6093 	error = nd->nd_repstat;
6094 	m_freem(nd->nd_mrep);
6095 	return (error);
6096 }
6097 
6098 /*
6099  * Initialize the slot tables for a session.
6100  */
6101 static void
6102 nfscl_initsessionslots(struct nfsclsession *sep)
6103 {
6104 	int i;
6105 
6106 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6107 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
6108 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
6109 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
6110 	}
6111 	for (i = 0; i < 64; i++)
6112 		sep->nfsess_slotseq[i] = 0;
6113 	sep->nfsess_slots = 0;
6114 }
6115 
6116 /*
6117  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
6118  */
6119 int
6120 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6121     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
6122 {
6123 	struct nfsnode *np = VTONFS(vp);
6124 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6125 	struct nfscllayout *layp;
6126 	struct nfscldevinfo *dip;
6127 	struct nfsclflayout *rflp;
6128 	struct mbuf *m, *m2;
6129 	struct nfsclwritedsdorpc *drpc, *tdrpc;
6130 	nfsv4stateid_t stateid;
6131 	struct ucred *newcred;
6132 	uint64_t lastbyte, len, off, oresid, xfer;
6133 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
6134 	void *lckp;
6135 	uint8_t *dev;
6136 	void *iovbase = NULL;
6137 	size_t iovlen = 0;
6138 	off_t offs = 0;
6139 	ssize_t resid = 0;
6140 	uint32_t op;
6141 
6142 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
6143 	    (np->n_flag & NNOLAYOUT) != 0)
6144 		return (EIO);
6145 	/* Now, get a reference cnt on the clientid for this mount. */
6146 	if (nfscl_getref(nmp) == 0)
6147 		return (EIO);
6148 
6149 	/* Find an appropriate stateid. */
6150 	newcred = NFSNEWCRED(cred);
6151 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
6152 	    rwaccess, 1, newcred, p, &stateid, &lckp);
6153 	if (error != 0) {
6154 		NFSFREECRED(newcred);
6155 		nfscl_relref(nmp);
6156 		return (error);
6157 	}
6158 	/* Search for a layout for this file. */
6159 	off = uiop->uio_offset;
6160 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
6161 	    np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
6162 	if (layp == NULL || rflp == NULL) {
6163 		if (recalled != 0) {
6164 			NFSFREECRED(newcred);
6165 			if (lckp != NULL)
6166 				nfscl_lockderef(lckp);
6167 			nfscl_relref(nmp);
6168 			return (EIO);
6169 		}
6170 		if (layp != NULL) {
6171 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
6172 			layp = NULL;
6173 		}
6174 		/* Try and get a Layout, if it is supported. */
6175 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
6176 		    (np->n_flag & NWRITEOPENED) != 0)
6177 			iolaymode = NFSLAYOUTIOMODE_RW;
6178 		else
6179 			iolaymode = NFSLAYOUTIOMODE_READ;
6180 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
6181 		    rwaccess, NULL, &stateid, off, &layp, newcred, p);
6182 		if (error != 0) {
6183 			NFSLOCKNODE(np);
6184 			np->n_flag |= NNOLAYOUT;
6185 			NFSUNLOCKNODE(np);
6186 			if (lckp != NULL)
6187 				nfscl_lockderef(lckp);
6188 			NFSFREECRED(newcred);
6189 			if (layp != NULL)
6190 				nfscl_rellayout(layp, 0);
6191 			nfscl_relref(nmp);
6192 			return (error);
6193 		}
6194 	}
6195 
6196 	/*
6197 	 * Loop around finding a layout that works for the first part of
6198 	 * this I/O operation, and then call the function that actually
6199 	 * does the RPC.
6200 	 */
6201 	eof = 0;
6202 	len = (uint64_t)uiop->uio_resid;
6203 	while (len > 0 && error == 0 && eof == 0) {
6204 		off = uiop->uio_offset;
6205 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
6206 		if (error == 0) {
6207 			oresid = xfer = (uint64_t)uiop->uio_resid;
6208 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
6209 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
6210 			/*
6211 			 * For Flex File layout with mirrored DSs, select one
6212 			 * of them at random for reads. For writes and commits,
6213 			 * do all mirrors.
6214 			 */
6215 			m = NULL;
6216 			tdrpc = drpc = NULL;
6217 			firstmirror = 0;
6218 			mirrorcnt = 1;
6219 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
6220 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
6221 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
6222 					firstmirror = arc4random() % mirrorcnt;
6223 					mirrorcnt = firstmirror + 1;
6224 				} else {
6225 					if (docommit == 0) {
6226 						/*
6227 						 * Save values, so uiop can be
6228 						 * rolled back upon a write
6229 						 * error.
6230 						 */
6231 						offs = uiop->uio_offset;
6232 						resid = uiop->uio_resid;
6233 						iovbase =
6234 						    uiop->uio_iov->iov_base;
6235 						iovlen = uiop->uio_iov->iov_len;
6236 						m = nfsm_uiombuflist(uiop, len,
6237 						    0);
6238 					}
6239 					tdrpc = drpc = malloc(sizeof(*drpc) *
6240 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
6241 					    M_ZERO);
6242 				}
6243 			}
6244 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
6245 				m2 = NULL;
6246 				if (m != NULL && i < mirrorcnt - 1)
6247 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
6248 				else {
6249 					m2 = m;
6250 					m = NULL;
6251 				}
6252 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
6253 					dev = rflp->nfsfl_ffm[i].dev;
6254 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6255 					    rflp->nfsfl_ffm[i].devp);
6256 				} else {
6257 					dev = rflp->nfsfl_dev;
6258 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6259 					    rflp->nfsfl_devp);
6260 				}
6261 				if (dip != NULL) {
6262 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
6263 					    != 0)
6264 						error = nfscl_dofflayoutio(vp,
6265 						    uiop, iomode, must_commit,
6266 						    &eof, &stateid, rwaccess,
6267 						    dip, layp, rflp, off, xfer,
6268 						    i, docommit, m2, tdrpc,
6269 						    newcred, p);
6270 					else
6271 						error = nfscl_doflayoutio(vp,
6272 						    uiop, iomode, must_commit,
6273 						    &eof, &stateid, rwaccess,
6274 						    dip, layp, rflp, off, xfer,
6275 						    docommit, newcred, p);
6276 					nfscl_reldevinfo(dip);
6277 				} else {
6278 					if (m2 != NULL)
6279 						m_freem(m2);
6280 					error = EIO;
6281 				}
6282 				tdrpc++;
6283 			}
6284 			if (m != NULL)
6285 				m_freem(m);
6286 			tdrpc = drpc;
6287 			timo = hz / 50;		/* Wait for 20msec. */
6288 			if (timo < 1)
6289 				timo = 1;
6290 			for (i = firstmirror; i < mirrorcnt - 1 &&
6291 			    tdrpc != NULL; i++, tdrpc++) {
6292 				/*
6293 				 * For the unused drpc entries, both inprog and
6294 				 * err == 0, so this loop won't break.
6295 				 */
6296 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
6297 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
6298 					    timo);
6299 				if (error == 0 && tdrpc->err != 0)
6300 					error = tdrpc->err;
6301 				if (rwaccess != NFSV4OPEN_ACCESSREAD &&
6302 				    docommit == 0 && *must_commit == 0 &&
6303 				    tdrpc->must_commit == 1)
6304 					*must_commit = 1;
6305 			}
6306 			free(drpc, M_TEMP);
6307 			if (error == 0) {
6308 				if (mirrorcnt > 1 && rwaccess ==
6309 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6310 					NFSLOCKCLSTATE();
6311 					layp->nfsly_flags |= NFSLY_WRITTEN;
6312 					NFSUNLOCKCLSTATE();
6313 				}
6314 				lastbyte = off + xfer - 1;
6315 				NFSLOCKCLSTATE();
6316 				if (lastbyte > layp->nfsly_lastbyte)
6317 					layp->nfsly_lastbyte = lastbyte;
6318 				NFSUNLOCKCLSTATE();
6319 			} else if (error == NFSERR_OPENMODE &&
6320 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
6321 				NFSLOCKMNT(nmp);
6322 				nmp->nm_state |= NFSSTA_OPENMODE;
6323 				NFSUNLOCKMNT(nmp);
6324 			} else if ((error == NFSERR_NOSPC ||
6325 			    error == NFSERR_IO || error == NFSERR_NXIO) &&
6326 			    nmp->nm_minorvers == NFSV42_MINORVERSION) {
6327 				if (docommit != 0)
6328 					op = NFSV4OP_COMMIT;
6329 				else if (rwaccess == NFSV4OPEN_ACCESSREAD)
6330 					op = NFSV4OP_READ;
6331 				else
6332 					op = NFSV4OP_WRITE;
6333 				nfsrpc_layouterror(nmp, np->n_fhp->nfh_fh,
6334 				    np->n_fhp->nfh_len, off, xfer,
6335 				    &layp->nfsly_stateid, newcred, p, error, op,
6336 				    dip->nfsdi_deviceid);
6337 				error = EIO;
6338 			} else
6339 				error = EIO;
6340 			if (error == 0)
6341 				len -= (oresid - (uint64_t)uiop->uio_resid);
6342 			else if (mirrorcnt > 1 && rwaccess ==
6343 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6344 				/*
6345 				 * In case the rpc gets retried, roll the
6346 				 * uio fields changed by nfsm_uiombuflist()
6347 				 * back.
6348 				 */
6349 				uiop->uio_offset = offs;
6350 				uiop->uio_resid = resid;
6351 				uiop->uio_iov->iov_base = iovbase;
6352 				uiop->uio_iov->iov_len = iovlen;
6353 			}
6354 		}
6355 	}
6356 	if (lckp != NULL)
6357 		nfscl_lockderef(lckp);
6358 	NFSFREECRED(newcred);
6359 	nfscl_rellayout(layp, 0);
6360 	nfscl_relref(nmp);
6361 	return (error);
6362 }
6363 
6364 /*
6365  * Find a file layout that will handle the first bytes of the requested
6366  * range and return the information from it needed to the I/O operation.
6367  */
6368 int
6369 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
6370     struct nfsclflayout **retflpp)
6371 {
6372 	struct nfsclflayout *flp, *nflp, *rflp;
6373 	uint32_t rw;
6374 
6375 	rflp = NULL;
6376 	rw = rwaccess;
6377 	/* For reading, do the Read list first and then the Write list. */
6378 	do {
6379 		if (rw == NFSV4OPEN_ACCESSREAD)
6380 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6381 		else
6382 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6383 		while (flp != NULL) {
6384 			nflp = LIST_NEXT(flp, nfsfl_list);
6385 			if (flp->nfsfl_off > off)
6386 				break;
6387 			if (flp->nfsfl_end > off &&
6388 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6389 				rflp = flp;
6390 			flp = nflp;
6391 		}
6392 		if (rw == NFSV4OPEN_ACCESSREAD)
6393 			rw = NFSV4OPEN_ACCESSWRITE;
6394 		else
6395 			rw = 0;
6396 	} while (rw != 0);
6397 	if (rflp != NULL) {
6398 		/* This one covers the most bytes starting at off. */
6399 		*retflpp = rflp;
6400 		return (0);
6401 	}
6402 	return (EIO);
6403 }
6404 
6405 /*
6406  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6407  */
6408 static int
6409 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6410     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6411     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6412     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6413 {
6414 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6415 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6416 	struct nfsnode *np;
6417 	struct nfsfh *fhp;
6418 	struct nfsclds **dspp;
6419 
6420 	np = VTONFS(vp);
6421 	rel_off = off - flp->nfsfl_patoff;
6422 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6423 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6424 	    dp->nfsdi_stripecnt;
6425 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6426 	error = 0;
6427 
6428 	/* Loop around, doing I/O for each stripe unit. */
6429 	while (len > 0 && error == 0) {
6430 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6431 		dspp = nfsfldi_addr(dp, stripe_index);
6432 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6433 			minorvers = NFSV42_MINORVERSION;
6434 		else
6435 			minorvers = NFSV41_MINORVERSION;
6436 		if (len > transfer && docommit == 0)
6437 			xfer = transfer;
6438 		else
6439 			xfer = len;
6440 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6441 			/* Dense layout. */
6442 			if (stripe_pos >= flp->nfsfl_fhcnt)
6443 				return (EIO);
6444 			fhp = flp->nfsfl_fh[stripe_pos];
6445 			io_off = (rel_off / (stripe_unit_size *
6446 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6447 			    rel_off % stripe_unit_size;
6448 		} else {
6449 			/* Sparse layout. */
6450 			if (flp->nfsfl_fhcnt > 1) {
6451 				if (stripe_index >= flp->nfsfl_fhcnt)
6452 					return (EIO);
6453 				fhp = flp->nfsfl_fh[stripe_index];
6454 			} else if (flp->nfsfl_fhcnt == 1)
6455 				fhp = flp->nfsfl_fh[0];
6456 			else
6457 				fhp = np->n_fhp;
6458 			io_off = off;
6459 		}
6460 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6461 			commit_thru_mds = 1;
6462 			if (docommit != 0)
6463 				error = EIO;
6464 		} else {
6465 			commit_thru_mds = 0;
6466 			NFSLOCKNODE(np);
6467 			np->n_flag |= NDSCOMMIT;
6468 			NFSUNLOCKNODE(np);
6469 		}
6470 		if (docommit != 0) {
6471 			if (error == 0)
6472 				error = nfsrpc_commitds(vp, io_off, xfer,
6473 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6474 			if (error == 0) {
6475 				/*
6476 				 * Set both eof and uio_resid = 0 to end any
6477 				 * loops.
6478 				 */
6479 				*eofp = 1;
6480 				uiop->uio_resid = 0;
6481 			} else {
6482 				NFSLOCKNODE(np);
6483 				np->n_flag &= ~NDSCOMMIT;
6484 				NFSUNLOCKNODE(np);
6485 			}
6486 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6487 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6488 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6489 		else {
6490 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6491 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6492 			    0, NFS_VER4, minorvers, cred, p);
6493 			if (error == 0) {
6494 				NFSLOCKCLSTATE();
6495 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6496 				NFSUNLOCKCLSTATE();
6497 			}
6498 		}
6499 		if (error == 0) {
6500 			transfer = stripe_unit_size;
6501 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6502 			len -= xfer;
6503 			off += xfer;
6504 		}
6505 	}
6506 	return (error);
6507 }
6508 
6509 /*
6510  * Do I/O using an NFSv4.1 flex file layout.
6511  */
6512 static int
6513 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6514     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6515     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6516     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6517     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6518 {
6519 	uint64_t xfer;
6520 	int error;
6521 	struct nfsnode *np;
6522 	struct nfsfh *fhp;
6523 	struct nfsclds **dspp;
6524 	struct ucred *tcred;
6525 	struct mbuf *m, *m2;
6526 	uint32_t copylen;
6527 
6528 	np = VTONFS(vp);
6529 	error = 0;
6530 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6531 	    (uintmax_t)len);
6532 	/* Loop around, doing I/O for each stripe unit. */
6533 	while (len > 0 && error == 0) {
6534 		dspp = nfsfldi_addr(dp, 0);
6535 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6536 		stateidp = &flp->nfsfl_ffm[mirror].st;
6537 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6538 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6539 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6540 			tcred = NFSNEWCRED(cred);
6541 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6542 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6543 			tcred->cr_ngroups = 1;
6544 		} else
6545 			tcred = cred;
6546 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6547 			copylen = dp->nfsdi_rsize;
6548 		else {
6549 			copylen = dp->nfsdi_wsize;
6550 			if (len > copylen && mp != NULL) {
6551 				/*
6552 				 * When a mirrored configuration needs to do
6553 				 * multiple writes to each mirror, all writes
6554 				 * except the last one must be a multiple of
6555 				 * 4 bytes.  This is required so that the XDR
6556 				 * does not need padding.
6557 				 * If possible, clip the size to an exact
6558 				 * multiple of the mbuf length, so that the
6559 				 * split will be on an mbuf boundary.
6560 				 */
6561 				copylen &= 0xfffffffc;
6562 				if (copylen > mp->m_len)
6563 					copylen = copylen / mp->m_len *
6564 					    mp->m_len;
6565 			}
6566 		}
6567 		NFSLOCKNODE(np);
6568 		np->n_flag |= NDSCOMMIT;
6569 		NFSUNLOCKNODE(np);
6570 		if (len > copylen && docommit == 0)
6571 			xfer = copylen;
6572 		else
6573 			xfer = len;
6574 		if (docommit != 0) {
6575 			if (error == 0) {
6576 				/*
6577 				 * Do last mirrored DS commit with this thread.
6578 				 */
6579 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6580 					error = nfsio_commitds(vp, off, xfer,
6581 					    *dspp, fhp, dp->nfsdi_vers,
6582 					    dp->nfsdi_minorvers, drpc, tcred,
6583 					    p);
6584 				else
6585 					error = nfsrpc_commitds(vp, off, xfer,
6586 					    *dspp, fhp, dp->nfsdi_vers,
6587 					    dp->nfsdi_minorvers, tcred, p);
6588 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6589 				if (error != 0 && error != EACCES && error !=
6590 				    ESTALE) {
6591 					NFSCL_DEBUG(4,
6592 					    "DS layreterr for commit\n");
6593 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6594 					    lyp, *dspp);
6595 				}
6596 			}
6597 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6598 			if (error == 0) {
6599 				/*
6600 				 * Set both eof and uio_resid = 0 to end any
6601 				 * loops.
6602 				 */
6603 				*eofp = 1;
6604 				uiop->uio_resid = 0;
6605 			} else {
6606 				NFSLOCKNODE(np);
6607 				np->n_flag &= ~NDSCOMMIT;
6608 				NFSUNLOCKNODE(np);
6609 			}
6610 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6611 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6612 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6613 			    dp->nfsdi_minorvers, tcred, p);
6614 			NFSCL_DEBUG(4, "readds=%d\n", error);
6615 			if (error != 0 && error != EACCES && error != ESTALE) {
6616 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6617 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6618 				    *dspp);
6619 			}
6620 		} else {
6621 			if (flp->nfsfl_mirrorcnt == 1) {
6622 				error = nfsrpc_writeds(vp, uiop, iomode,
6623 				    must_commit, stateidp, *dspp, off, xfer,
6624 				    fhp, 0, 1, dp->nfsdi_vers,
6625 				    dp->nfsdi_minorvers, tcred, p);
6626 				if (error == 0) {
6627 					NFSLOCKCLSTATE();
6628 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6629 					NFSUNLOCKCLSTATE();
6630 				}
6631 			} else {
6632 				m = mp;
6633 				if (xfer < len) {
6634 					/* The mbuf list must be split. */
6635 					m2 = nfsm_split(mp, xfer);
6636 					if (m2 != NULL)
6637 						mp = m2;
6638 					else {
6639 						m_freem(mp);
6640 						error = EIO;
6641 					}
6642 				}
6643 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6644 				    (uintmax_t)len, (uintmax_t)xfer);
6645 				/*
6646 				 * Do last write to a mirrored DS with this
6647 				 * thread.
6648 				 */
6649 				if (error == 0) {
6650 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6651 						error = nfsio_writedsmir(vp,
6652 						    iomode, must_commit,
6653 						    stateidp, *dspp, off,
6654 						    xfer, fhp, m,
6655 						    dp->nfsdi_vers,
6656 						    dp->nfsdi_minorvers, drpc,
6657 						    tcred, p);
6658 					else
6659 						error = nfsrpc_writedsmir(vp,
6660 						    iomode, must_commit,
6661 						    stateidp, *dspp, off,
6662 						    xfer, fhp, m,
6663 						    dp->nfsdi_vers,
6664 						    dp->nfsdi_minorvers, tcred,
6665 						    p);
6666 				}
6667 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6668 				if (error != 0 && error != EACCES && error !=
6669 				    ESTALE) {
6670 					NFSCL_DEBUG(4,
6671 					    "DS layreterr for write\n");
6672 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6673 					    lyp, *dspp);
6674 				}
6675 			}
6676 		}
6677 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6678 		if (error == 0) {
6679 			len -= xfer;
6680 			off += xfer;
6681 		}
6682 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6683 			NFSFREECRED(tcred);
6684 	}
6685 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6686 	return (error);
6687 }
6688 
6689 /*
6690  * The actual read RPC done to a DS.
6691  */
6692 static int
6693 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6694     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6695     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6696 {
6697 	uint32_t *tl;
6698 	int attrflag, error, retlen;
6699 	struct nfsrv_descript nfsd;
6700 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6701 	struct nfsrv_descript *nd = &nfsd;
6702 	struct nfssockreq *nrp;
6703 	struct nfsvattr na;
6704 
6705 	nd->nd_mrep = NULL;
6706 	if (vers == 0 || vers == NFS_VER4) {
6707 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6708 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6709 		vers = NFS_VER4;
6710 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6711 		if (flex != 0)
6712 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6713 		else
6714 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6715 	} else {
6716 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6717 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6718 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6719 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6720 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6721 	}
6722 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6723 	txdr_hyper(io_off, tl);
6724 	*(tl + 2) = txdr_unsigned(len);
6725 	nrp = dsp->nfsclds_sockp;
6726 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6727 	if (nrp == NULL)
6728 		/* If NULL, use the MDS socket. */
6729 		nrp = &nmp->nm_sockreq;
6730 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6731 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6732 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6733 	    error);
6734 	if (error != 0)
6735 		return (error);
6736 	if (vers == NFS_VER3) {
6737 		error = nfscl_postop_attr(nd, &na, &attrflag);
6738 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6739 		if (error != 0)
6740 			goto nfsmout;
6741 	}
6742 	if (nd->nd_repstat != 0) {
6743 		error = nd->nd_repstat;
6744 		goto nfsmout;
6745 	}
6746 	if (vers == NFS_VER3) {
6747 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6748 		*eofp = fxdr_unsigned(int, *(tl + 1));
6749 	} else {
6750 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6751 		*eofp = fxdr_unsigned(int, *tl);
6752 	}
6753 	NFSM_STRSIZ(retlen, len);
6754 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6755 	error = nfsm_mbufuio(nd, uiop, retlen);
6756 nfsmout:
6757 	if (nd->nd_mrep != NULL)
6758 		m_freem(nd->nd_mrep);
6759 	return (error);
6760 }
6761 
6762 /*
6763  * The actual write RPC done to a DS.
6764  */
6765 static int
6766 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6767     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6768     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6769     struct ucred *cred, NFSPROC_T *p)
6770 {
6771 	uint32_t *tl;
6772 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6773 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6774 	int32_t backup;
6775 	struct nfsrv_descript nfsd;
6776 	struct nfsrv_descript *nd = &nfsd;
6777 	struct nfssockreq *nrp;
6778 	struct nfsvattr na;
6779 
6780 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6781 	nd->nd_mrep = NULL;
6782 	if (vers == 0 || vers == NFS_VER4) {
6783 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6784 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6785 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6786 		vers = NFS_VER4;
6787 		if (flex != 0)
6788 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6789 		else
6790 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6791 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6792 	} else {
6793 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6794 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6795 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6796 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6797 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6798 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6799 	}
6800 	txdr_hyper(io_off, tl);
6801 	tl += 2;
6802 	if (vers == NFS_VER3)
6803 		*tl++ = txdr_unsigned(len);
6804 	*tl++ = txdr_unsigned(*iomode);
6805 	*tl = txdr_unsigned(len);
6806 	nfsm_uiombuf(nd, uiop, len);
6807 	nrp = dsp->nfsclds_sockp;
6808 	if (nrp == NULL)
6809 		/* If NULL, use the MDS socket. */
6810 		nrp = &nmp->nm_sockreq;
6811 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6812 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6813 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6814 	    nd->nd_repstat);
6815 	if (error != 0)
6816 		return (error);
6817 	if (nd->nd_repstat != 0) {
6818 		/*
6819 		 * In case the rpc gets retried, roll
6820 		 * the uio fields changed by nfsm_uiombuf()
6821 		 * back.
6822 		 */
6823 		uiop->uio_offset -= len;
6824 		uiop->uio_resid += len;
6825 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6826 		uiop->uio_iov->iov_len += len;
6827 		error = nd->nd_repstat;
6828 	} else {
6829 		if (vers == NFS_VER3) {
6830 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6831 			    NULL);
6832 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6833 			if (error != 0)
6834 				goto nfsmout;
6835 		}
6836 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6837 		rlen = fxdr_unsigned(int, *tl++);
6838 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6839 		if (rlen == 0) {
6840 			error = NFSERR_IO;
6841 			goto nfsmout;
6842 		} else if (rlen < len) {
6843 			backup = len - rlen;
6844 			uiop->uio_iov->iov_base =
6845 			    (char *)uiop->uio_iov->iov_base - backup;
6846 			uiop->uio_iov->iov_len += backup;
6847 			uiop->uio_offset -= backup;
6848 			uiop->uio_resid += backup;
6849 			len = rlen;
6850 		}
6851 		commit = fxdr_unsigned(int, *tl++);
6852 
6853 		/*
6854 		 * Return the lowest commitment level
6855 		 * obtained by any of the RPCs.
6856 		 */
6857 		if (committed == NFSWRITE_FILESYNC)
6858 			committed = commit;
6859 		else if (committed == NFSWRITE_DATASYNC &&
6860 		    commit == NFSWRITE_UNSTABLE)
6861 			committed = commit;
6862 		if (commit_thru_mds != 0) {
6863 			NFSLOCKMNT(nmp);
6864 			if (!NFSHASWRITEVERF(nmp)) {
6865 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6866 				NFSSETWRITEVERF(nmp);
6867 			} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF) &&
6868 			    *must_commit != 2) {
6869 				*must_commit = 1;
6870 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6871 			}
6872 			NFSUNLOCKMNT(nmp);
6873 		} else {
6874 			NFSLOCKDS(dsp);
6875 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6876 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6877 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6878 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
6879 			    *must_commit != 2) {
6880 				*must_commit = 1;
6881 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6882 			}
6883 			NFSUNLOCKDS(dsp);
6884 		}
6885 	}
6886 nfsmout:
6887 	if (nd->nd_mrep != NULL)
6888 		m_freem(nd->nd_mrep);
6889 	*iomode = committed;
6890 	if (nd->nd_repstat != 0 && error == 0)
6891 		error = nd->nd_repstat;
6892 	return (error);
6893 }
6894 
6895 /*
6896  * The actual write RPC done to a DS.
6897  * This variant is called from a separate kernel process for mirrors.
6898  * Any short write is considered an IO error.
6899  */
6900 static int
6901 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6902     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6903     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6904     struct ucred *cred, NFSPROC_T *p)
6905 {
6906 	uint32_t *tl;
6907 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6908 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6909 	struct nfsrv_descript nfsd;
6910 	struct nfsrv_descript *nd = &nfsd;
6911 	struct nfssockreq *nrp;
6912 	struct nfsvattr na;
6913 
6914 	nd->nd_mrep = NULL;
6915 	if (vers == 0 || vers == NFS_VER4) {
6916 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6917 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6918 		vers = NFS_VER4;
6919 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6920 		    minorvers);
6921 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6922 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6923 	} else {
6924 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6925 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6926 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6927 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6928 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6929 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6930 	}
6931 	txdr_hyper(io_off, tl);
6932 	tl += 2;
6933 	if (vers == NFS_VER3)
6934 		*tl++ = txdr_unsigned(len);
6935 	*tl++ = txdr_unsigned(*iomode);
6936 	*tl = txdr_unsigned(len);
6937 	if (len > 0) {
6938 		/* Put data in mbuf chain. */
6939 		nd->nd_mb->m_next = m;
6940 	}
6941 	nrp = dsp->nfsclds_sockp;
6942 	if (nrp == NULL)
6943 		/* If NULL, use the MDS socket. */
6944 		nrp = &nmp->nm_sockreq;
6945 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6946 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6947 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6948 	    nd->nd_repstat);
6949 	if (error != 0)
6950 		return (error);
6951 	if (nd->nd_repstat != 0)
6952 		error = nd->nd_repstat;
6953 	else {
6954 		if (vers == NFS_VER3) {
6955 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6956 			    NULL);
6957 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6958 			    error);
6959 			if (error != 0)
6960 				goto nfsmout;
6961 		}
6962 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6963 		rlen = fxdr_unsigned(int, *tl++);
6964 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6965 		    rlen);
6966 		if (rlen != len) {
6967 			error = NFSERR_IO;
6968 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6969 			    len, rlen);
6970 			goto nfsmout;
6971 		}
6972 		commit = fxdr_unsigned(int, *tl++);
6973 
6974 		/*
6975 		 * Return the lowest commitment level
6976 		 * obtained by any of the RPCs.
6977 		 */
6978 		if (committed == NFSWRITE_FILESYNC)
6979 			committed = commit;
6980 		else if (committed == NFSWRITE_DATASYNC &&
6981 		    commit == NFSWRITE_UNSTABLE)
6982 			committed = commit;
6983 		NFSLOCKDS(dsp);
6984 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6985 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6986 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6987 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
6988 		    *must_commit != 2) {
6989 			*must_commit = 1;
6990 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6991 		}
6992 		NFSUNLOCKDS(dsp);
6993 	}
6994 nfsmout:
6995 	if (nd->nd_mrep != NULL)
6996 		m_freem(nd->nd_mrep);
6997 	*iomode = committed;
6998 	if (nd->nd_repstat != 0 && error == 0)
6999 		error = nd->nd_repstat;
7000 	return (error);
7001 }
7002 
7003 /*
7004  * Start up the thread that will execute nfsrpc_writedsmir().
7005  */
7006 static void
7007 start_writedsmir(void *arg, int pending)
7008 {
7009 	struct nfsclwritedsdorpc *drpc;
7010 
7011 	drpc = (struct nfsclwritedsdorpc *)arg;
7012 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
7013 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
7014 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
7015 	    drpc->p);
7016 	drpc->done = 1;
7017 	crfree(drpc->cred);
7018 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
7019 }
7020 
7021 /*
7022  * Set up the write DS mirror call for the pNFS I/O thread.
7023  */
7024 static int
7025 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
7026     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
7027     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
7028     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7029 {
7030 	int error, ret;
7031 
7032 	error = 0;
7033 	drpc->done = 0;
7034 	drpc->vp = vp;
7035 	drpc->iomode = *iomode;
7036 	drpc->must_commit = *must_commit;
7037 	drpc->stateidp = stateidp;
7038 	drpc->dsp = dsp;
7039 	drpc->off = off;
7040 	drpc->len = len;
7041 	drpc->fhp = fhp;
7042 	drpc->m = m;
7043 	drpc->vers = vers;
7044 	drpc->minorvers = minorvers;
7045 	drpc->cred = crhold(cred);
7046 	drpc->p = p;
7047 	drpc->inprog = 0;
7048 	ret = EIO;
7049 	if (nfs_pnfsiothreads != 0) {
7050 		ret = nfs_pnfsio(start_writedsmir, drpc);
7051 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
7052 	}
7053 	if (ret != 0) {
7054 		error = nfsrpc_writedsmir(vp, iomode, &drpc->must_commit,
7055 		    stateidp, dsp, off, len, fhp, m, vers, minorvers, cred, p);
7056 		crfree(drpc->cred);
7057 	}
7058 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
7059 	return (error);
7060 }
7061 
7062 /*
7063  * Free up the nfsclds structure.
7064  */
7065 void
7066 nfscl_freenfsclds(struct nfsclds *dsp)
7067 {
7068 	int i;
7069 
7070 	if (dsp == NULL)
7071 		return;
7072 	if (dsp->nfsclds_sockp != NULL) {
7073 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
7074 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
7075 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
7076 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
7077 	}
7078 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
7079 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
7080 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
7081 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
7082 			m_freem(
7083 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
7084 	}
7085 	free(dsp, M_NFSCLDS);
7086 }
7087 
7088 static enum nfsclds_state
7089 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
7090     struct nfsclds **retdspp, uint32_t *sequencep)
7091 {
7092 	struct nfsclds *dsp;
7093 	int fndseq;
7094 
7095 	/*
7096 	 * Search the list of nfsclds structures for one with the same
7097 	 * server.
7098 	 */
7099 	fndseq = 0;
7100 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
7101 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
7102 		    dsp->nfsclds_servownlen != 0 &&
7103 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
7104 		    dsp->nfsclds_servownlen) &&
7105 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
7106 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
7107 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
7108 			    dsp->nfsclds_flags);
7109 			if (fndseq == 0) {
7110 				/* Get sequenceid# from first entry. */
7111 				*sequencep =
7112 				    dsp->nfsclds_sess.nfsess_sequenceid;
7113 				fndseq = 1;
7114 			}
7115 			/* Server major id matches. */
7116 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
7117 				*retdspp = dsp;
7118 				return (NFSDSP_USETHISSESSION);
7119 			}
7120 		}
7121 	}
7122 	if (fndseq != 0)
7123 		return (NFSDSP_SEQTHISSESSION);
7124 	return (NFSDSP_NOTFOUND);
7125 }
7126 
7127 /*
7128  * NFS commit rpc to a NFSv4.1 DS.
7129  */
7130 static int
7131 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7132     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
7133     NFSPROC_T *p)
7134 {
7135 	uint32_t *tl;
7136 	struct nfsrv_descript nfsd, *nd = &nfsd;
7137 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7138 	struct nfssockreq *nrp;
7139 	struct nfsvattr na;
7140 	int attrflag, error;
7141 
7142 	nd->nd_mrep = NULL;
7143 	if (vers == 0 || vers == NFS_VER4) {
7144 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
7145 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7146 		vers = NFS_VER4;
7147 	} else {
7148 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
7149 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7150 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
7151 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
7152 	}
7153 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
7154 	    minorvers);
7155 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7156 	txdr_hyper(offset, tl);
7157 	tl += 2;
7158 	*tl = txdr_unsigned(cnt);
7159 	nrp = dsp->nfsclds_sockp;
7160 	if (nrp == NULL)
7161 		/* If NULL, use the MDS socket. */
7162 		nrp = &nmp->nm_sockreq;
7163 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7164 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7165 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
7166 	    nd->nd_repstat);
7167 	if (error != 0)
7168 		return (error);
7169 	if (nd->nd_repstat == 0) {
7170 		if (vers == NFS_VER3) {
7171 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7172 			    NULL);
7173 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
7174 			if (error != 0)
7175 				goto nfsmout;
7176 		}
7177 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
7178 		NFSLOCKDS(dsp);
7179 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
7180 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7181 			error = NFSERR_STALEWRITEVERF;
7182 		}
7183 		NFSUNLOCKDS(dsp);
7184 	}
7185 nfsmout:
7186 	if (error == 0 && nd->nd_repstat != 0)
7187 		error = nd->nd_repstat;
7188 	m_freem(nd->nd_mrep);
7189 	return (error);
7190 }
7191 
7192 /*
7193  * Start up the thread that will execute nfsrpc_commitds().
7194  */
7195 static void
7196 start_commitds(void *arg, int pending)
7197 {
7198 	struct nfsclwritedsdorpc *drpc;
7199 
7200 	drpc = (struct nfsclwritedsdorpc *)arg;
7201 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
7202 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
7203 	    drpc->p);
7204 	drpc->done = 1;
7205 	crfree(drpc->cred);
7206 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
7207 }
7208 
7209 /*
7210  * Set up the commit DS mirror call for the pNFS I/O thread.
7211  */
7212 static int
7213 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7214     struct nfsfh *fhp, int vers, int minorvers,
7215     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7216 {
7217 	int error, ret;
7218 
7219 	error = 0;
7220 	drpc->done = 0;
7221 	drpc->vp = vp;
7222 	drpc->off = offset;
7223 	drpc->len = cnt;
7224 	drpc->dsp = dsp;
7225 	drpc->fhp = fhp;
7226 	drpc->vers = vers;
7227 	drpc->minorvers = minorvers;
7228 	drpc->cred = crhold(cred);
7229 	drpc->p = p;
7230 	drpc->inprog = 0;
7231 	ret = EIO;
7232 	if (nfs_pnfsiothreads != 0) {
7233 		ret = nfs_pnfsio(start_commitds, drpc);
7234 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
7235 	}
7236 	if (ret != 0) {
7237 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
7238 		    minorvers, cred, p);
7239 		crfree(drpc->cred);
7240 	}
7241 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
7242 	return (error);
7243 }
7244 
7245 /*
7246  * NFS Advise rpc
7247  */
7248 int
7249 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
7250     struct ucred *cred, NFSPROC_T *p)
7251 {
7252 	u_int32_t *tl;
7253 	struct nfsrv_descript nfsd, *nd = &nfsd;
7254 	nfsattrbit_t hints;
7255 	int error;
7256 
7257 	NFSZERO_ATTRBIT(&hints);
7258 	if (advise == POSIX_FADV_WILLNEED)
7259 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7260 	else if (advise == POSIX_FADV_DONTNEED)
7261 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7262 	else
7263 		return (0);
7264 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
7265 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7266 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
7267 	txdr_hyper(offset, tl);
7268 	tl += 2;
7269 	txdr_hyper(cnt, tl);
7270 	nfsrv_putattrbit(nd, &hints);
7271 	error = nfscl_request(nd, vp, p, cred);
7272 	if (error != 0)
7273 		return (error);
7274 	if (nd->nd_repstat != 0)
7275 		error = nd->nd_repstat;
7276 	m_freem(nd->nd_mrep);
7277 	return (error);
7278 }
7279 
7280 #ifdef notyet
7281 /*
7282  * NFS advise rpc to a NFSv4.2 DS.
7283  */
7284 static int
7285 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7286     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7287     struct ucred *cred, NFSPROC_T *p)
7288 {
7289 	uint32_t *tl;
7290 	struct nfsrv_descript nfsd, *nd = &nfsd;
7291 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7292 	struct nfssockreq *nrp;
7293 	nfsattrbit_t hints;
7294 	int error;
7295 
7296 	/* For NFS DSs prior to NFSv4.2, just return OK. */
7297 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
7298 		return (0);
7299 	NFSZERO_ATTRBIT(&hints);
7300 	if (advise == POSIX_FADV_WILLNEED)
7301 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7302 	else if (advise == POSIX_FADV_DONTNEED)
7303 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7304 	else
7305 		return (0);
7306 	nd->nd_mrep = NULL;
7307 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
7308 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
7309 	vers = NFS_VER4;
7310 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
7311 	    minorvers);
7312 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7313 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7314 	txdr_hyper(offset, tl);
7315 	tl += 2;
7316 	*tl = txdr_unsigned(cnt);
7317 	nfsrv_putattrbit(nd, &hints);
7318 	nrp = dsp->nfsclds_sockp;
7319 	if (nrp == NULL)
7320 		/* If NULL, use the MDS socket. */
7321 		nrp = &nmp->nm_sockreq;
7322 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7323 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7324 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
7325 	    nd->nd_repstat);
7326 	if (error != 0)
7327 		return (error);
7328 	if (nd->nd_repstat != 0)
7329 		error = nd->nd_repstat;
7330 	m_freem(nd->nd_mrep);
7331 	return (error);
7332 }
7333 
7334 /*
7335  * Start up the thread that will execute nfsrpc_commitds().
7336  */
7337 static void
7338 start_adviseds(void *arg, int pending)
7339 {
7340 	struct nfsclwritedsdorpc *drpc;
7341 
7342 	drpc = (struct nfsclwritedsdorpc *)arg;
7343 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
7344 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
7345 	    drpc->cred, drpc->p);
7346 	drpc->done = 1;
7347 	crfree(drpc->cred);
7348 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
7349 }
7350 
7351 /*
7352  * Set up the advise DS mirror call for the pNFS I/O thread.
7353  */
7354 static int
7355 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7356     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7357     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7358 {
7359 	int error, ret;
7360 
7361 	error = 0;
7362 	drpc->done = 0;
7363 	drpc->vp = vp;
7364 	drpc->off = offset;
7365 	drpc->len = cnt;
7366 	drpc->advise = advise;
7367 	drpc->dsp = dsp;
7368 	drpc->fhp = fhp;
7369 	drpc->vers = vers;
7370 	drpc->minorvers = minorvers;
7371 	drpc->cred = crhold(cred);
7372 	drpc->p = p;
7373 	drpc->inprog = 0;
7374 	ret = EIO;
7375 	if (nfs_pnfsiothreads != 0) {
7376 		ret = nfs_pnfsio(start_adviseds, drpc);
7377 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
7378 	}
7379 	if (ret != 0) {
7380 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
7381 		    minorvers, cred, p);
7382 		crfree(drpc->cred);
7383 	}
7384 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
7385 	return (error);
7386 }
7387 #endif	/* notyet */
7388 
7389 /*
7390  * Do the Allocate operation, retrying for recovery.
7391  */
7392 int
7393 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
7394     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
7395 {
7396 	int error, expireret = 0, retrycnt, nostateid;
7397 	uint32_t clidrev = 0;
7398 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7399 	struct nfsfh *nfhp = NULL;
7400 	nfsv4stateid_t stateid;
7401 	off_t tmp_off;
7402 	void *lckp;
7403 
7404 	if (len < 0)
7405 		return (EINVAL);
7406 	if (len == 0)
7407 		return (0);
7408 	tmp_off = off + len;
7409 	NFSLOCKMNT(nmp);
7410 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7411 		NFSUNLOCKMNT(nmp);
7412 		return (EFBIG);
7413 	}
7414 	if (nmp->nm_clp != NULL)
7415 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7416 	NFSUNLOCKMNT(nmp);
7417 	nfhp = VTONFS(vp)->n_fhp;
7418 	retrycnt = 0;
7419 	do {
7420 		lckp = NULL;
7421 		nostateid = 0;
7422 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7423 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7424 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7425 		    stateid.other[2] == 0) {
7426 			nostateid = 1;
7427 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7428 		}
7429 
7430 		/*
7431 		 * Not finding a stateid should probably never happen,
7432 		 * but just return an error for this case.
7433 		 */
7434 		if (nostateid != 0)
7435 			error = EIO;
7436 		else
7437 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7438 			    nap, attrflagp, cred, p);
7439 		if (error == NFSERR_STALESTATEID)
7440 			nfscl_initiate_recovery(nmp->nm_clp);
7441 		if (lckp != NULL)
7442 			nfscl_lockderef(lckp);
7443 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7444 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7445 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7446 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7447 		} else if ((error == NFSERR_EXPIRED ||
7448 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7449 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7450 		}
7451 		retrycnt++;
7452 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7453 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7454 	    error == NFSERR_STALEDONTRECOVER ||
7455 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7456 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7457 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7458 	if (error != 0 && retrycnt >= 4)
7459 		error = EIO;
7460 	return (error);
7461 }
7462 
7463 /*
7464  * The allocate RPC.
7465  */
7466 static int
7467 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7468     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7469 {
7470 	uint32_t *tl;
7471 	int error;
7472 	struct nfsrv_descript nfsd;
7473 	struct nfsrv_descript *nd = &nfsd;
7474 	nfsattrbit_t attrbits;
7475 
7476 	*attrflagp = 0;
7477 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7478 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7479 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7480 	txdr_hyper(off, tl); tl += 2;
7481 	txdr_hyper(len, tl); tl += 2;
7482 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7483 	NFSGETATTR_ATTRBIT(&attrbits);
7484 	nfsrv_putattrbit(nd, &attrbits);
7485 	error = nfscl_request(nd, vp, p, cred);
7486 	if (error != 0)
7487 		return (error);
7488 	if (nd->nd_repstat == 0) {
7489 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7490 		error = nfsm_loadattr(nd, nap);
7491 		if (error == 0)
7492 			*attrflagp = NFS_LATTR_NOSHRINK;
7493 	} else
7494 		error = nd->nd_repstat;
7495 nfsmout:
7496 	m_freem(nd->nd_mrep);
7497 	return (error);
7498 }
7499 
7500 /*
7501  * Set up the XDR arguments for the LayoutGet operation.
7502  */
7503 static void
7504 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7505     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7506     int layoutlen, int usecurstateid)
7507 {
7508 	uint32_t *tl;
7509 
7510 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7511 	    NFSX_STATEID);
7512 	*tl++ = newnfs_false;		/* Don't signal availability. */
7513 	*tl++ = txdr_unsigned(layouttype);
7514 	*tl++ = txdr_unsigned(iomode);
7515 	txdr_hyper(offset, tl);
7516 	tl += 2;
7517 	txdr_hyper(len, tl);
7518 	tl += 2;
7519 	txdr_hyper(minlen, tl);
7520 	tl += 2;
7521 	if (usecurstateid != 0) {
7522 		/* Special stateid for Current stateid. */
7523 		*tl++ = txdr_unsigned(1);
7524 		*tl++ = 0;
7525 		*tl++ = 0;
7526 		*tl++ = 0;
7527 	} else {
7528 		*tl++ = txdr_unsigned(stateidp->seqid);
7529 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7530 		*tl++ = stateidp->other[0];
7531 		*tl++ = stateidp->other[1];
7532 		*tl++ = stateidp->other[2];
7533 	}
7534 	*tl = txdr_unsigned(layoutlen);
7535 }
7536 
7537 /*
7538  * Parse the reply for a successful LayoutGet operation.
7539  */
7540 static int
7541 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7542     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7543 {
7544 	uint32_t *tl;
7545 	struct nfsclflayout *flp, *prevflp, *tflp;
7546 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7547 	int m, mirrorcnt;
7548 	uint64_t retlen, off;
7549 	struct nfsfh *nfhp;
7550 	uint8_t *cp;
7551 	uid_t user;
7552 	gid_t grp;
7553 
7554 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7555 	error = 0;
7556 	flp = NULL;
7557 	gotiomode = -1;
7558 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7559 	if (*tl++ != 0)
7560 		*retonclosep = 1;
7561 	else
7562 		*retonclosep = 0;
7563 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7564 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7565 	    (int)stateidp->seqid);
7566 	stateidp->other[0] = *tl++;
7567 	stateidp->other[1] = *tl++;
7568 	stateidp->other[2] = *tl++;
7569 	cnt = fxdr_unsigned(int, *tl);
7570 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7571 	if (cnt <= 0 || cnt > 10000) {
7572 		/* Don't accept more than 10000 layouts in reply. */
7573 		error = NFSERR_BADXDR;
7574 		goto nfsmout;
7575 	}
7576 	for (i = 0; i < cnt; i++) {
7577 		/* Dissect to the layout type. */
7578 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7579 		    3 * NFSX_UNSIGNED);
7580 		off = fxdr_hyper(tl); tl += 2;
7581 		retlen = fxdr_hyper(tl); tl += 2;
7582 		iomode = fxdr_unsigned(int, *tl++);
7583 		laytype = fxdr_unsigned(int, *tl);
7584 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7585 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7586 		/* Ignore length of layout body for now. */
7587 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7588 			/* Parse the File layout up to fhcnt. */
7589 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7590 			    NFSX_HYPER + NFSX_V4DEVICEID);
7591 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7592 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7593 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7594 			if (fhcnt < 0 || fhcnt > 100) {
7595 				/* Don't accept more than 100 file handles. */
7596 				error = NFSERR_BADXDR;
7597 				goto nfsmout;
7598 			}
7599 			if (fhcnt > 0)
7600 				flp = malloc(sizeof(*flp) + fhcnt *
7601 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7602 				    M_WAITOK);
7603 			else
7604 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7605 				    M_WAITOK);
7606 			flp->nfsfl_flags = NFSFL_FILE;
7607 			flp->nfsfl_fhcnt = 0;
7608 			flp->nfsfl_devp = NULL;
7609 			flp->nfsfl_off = off;
7610 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7611 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7612 			else
7613 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7614 			flp->nfsfl_iomode = iomode;
7615 			if (gotiomode == -1)
7616 				gotiomode = flp->nfsfl_iomode;
7617 			/* Ignore layout body length for now. */
7618 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7619 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7620 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7621 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7622 			mtx_lock(&nmp->nm_mtx);
7623 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7624 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7625 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7626 			mtx_unlock(&nmp->nm_mtx);
7627 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7628 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7629 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7630 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7631 			for (j = 0; j < fhcnt; j++) {
7632 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7633 				nfhlen = fxdr_unsigned(int, *tl);
7634 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7635 					error = NFSERR_BADXDR;
7636 					goto nfsmout;
7637 				}
7638 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7639 				    M_NFSFH, M_WAITOK);
7640 				flp->nfsfl_fh[j] = nfhp;
7641 				flp->nfsfl_fhcnt++;
7642 				nfhp->nfh_len = nfhlen;
7643 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7644 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7645 			}
7646 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7647 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7648 			    NFSX_HYPER);
7649 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7650 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7651 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7652 				error = NFSERR_BADXDR;
7653 				goto nfsmout;
7654 			}
7655 			flp = malloc(sizeof(*flp) + mirrorcnt *
7656 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7657 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7658 			flp->nfsfl_mirrorcnt = mirrorcnt;
7659 			for (j = 0; j < mirrorcnt; j++)
7660 				flp->nfsfl_ffm[j].devp = NULL;
7661 			flp->nfsfl_off = off;
7662 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7663 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7664 			else
7665 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7666 			flp->nfsfl_iomode = iomode;
7667 			if (gotiomode == -1)
7668 				gotiomode = flp->nfsfl_iomode;
7669 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7670 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7671 			    (uintmax_t)flp->nfsfl_stripeunit);
7672 			for (j = 0; j < mirrorcnt; j++) {
7673 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7674 				k = fxdr_unsigned(int, *tl);
7675 				if (k < 1 || k > 128) {
7676 					error = NFSERR_BADXDR;
7677 					goto nfsmout;
7678 				}
7679 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7680 				for (l = 0; l < k; l++) {
7681 					NFSM_DISSECT(tl, uint32_t *,
7682 					    NFSX_V4DEVICEID + NFSX_STATEID +
7683 					    2 * NFSX_UNSIGNED);
7684 					if (l == 0) {
7685 						/* Just use the first server. */
7686 						NFSBCOPY(tl,
7687 						    flp->nfsfl_ffm[j].dev,
7688 						    NFSX_V4DEVICEID);
7689 						tl += (NFSX_V4DEVICEID /
7690 						    NFSX_UNSIGNED);
7691 						tl++;
7692 						flp->nfsfl_ffm[j].st.seqid =
7693 						    *tl++;
7694 						flp->nfsfl_ffm[j].st.other[0] =
7695 						    *tl++;
7696 						flp->nfsfl_ffm[j].st.other[1] =
7697 						    *tl++;
7698 						flp->nfsfl_ffm[j].st.other[2] =
7699 						    *tl++;
7700 						NFSCL_DEBUG(4, "st.seqid=%u "
7701 						 "st.o0=0x%x st.o1=0x%x "
7702 						 "st.o2=0x%x\n",
7703 						 flp->nfsfl_ffm[j].st.seqid,
7704 						 flp->nfsfl_ffm[j].st.other[0],
7705 						 flp->nfsfl_ffm[j].st.other[1],
7706 						 flp->nfsfl_ffm[j].st.other[2]);
7707 					} else
7708 						tl += ((NFSX_V4DEVICEID +
7709 						    NFSX_STATEID +
7710 						    NFSX_UNSIGNED) /
7711 						    NFSX_UNSIGNED);
7712 					fhcnt = fxdr_unsigned(int, *tl);
7713 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7714 					if (fhcnt < 1 ||
7715 					    fhcnt > NFSDEV_MAXVERS) {
7716 						error = NFSERR_BADXDR;
7717 						goto nfsmout;
7718 					}
7719 					for (m = 0; m < fhcnt; m++) {
7720 						NFSM_DISSECT(tl, uint32_t *,
7721 						    NFSX_UNSIGNED);
7722 						nfhlen = fxdr_unsigned(int,
7723 						    *tl);
7724 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7725 						    nfhlen);
7726 						if (nfhlen <= 0 || nfhlen >
7727 						    NFSX_V4FHMAX) {
7728 							error = NFSERR_BADXDR;
7729 							goto nfsmout;
7730 						}
7731 						NFSM_DISSECT(cp, uint8_t *,
7732 						    NFSM_RNDUP(nfhlen));
7733 						if (l == 0) {
7734 							flp->nfsfl_ffm[j].fhcnt
7735 							    = fhcnt;
7736 							nfhp = malloc(
7737 							    sizeof(*nfhp) +
7738 							    nfhlen - 1, M_NFSFH,
7739 							    M_WAITOK);
7740 							flp->nfsfl_ffm[j].fh[m]
7741 							    = nfhp;
7742 							nfhp->nfh_len = nfhlen;
7743 							NFSBCOPY(cp,
7744 							    nfhp->nfh_fh,
7745 							    nfhlen);
7746 							NFSCL_DEBUG(4,
7747 							    "got fh\n");
7748 						}
7749 					}
7750 					/* Now, get the ffsd_user/ffds_group. */
7751 					error = nfsrv_parseug(nd, 0, &user,
7752 					    &grp, curthread);
7753 					NFSCL_DEBUG(4, "after parseu=%d\n",
7754 					    error);
7755 					if (error == 0)
7756 						error = nfsrv_parseug(nd, 1,
7757 						    &user, &grp, curthread);
7758 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7759 					    grp);
7760 					if (error != 0)
7761 						goto nfsmout;
7762 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7763 					    user, grp);
7764 					if (l == 0) {
7765 						flp->nfsfl_ffm[j].user = user;
7766 						flp->nfsfl_ffm[j].group = grp;
7767 						NFSCL_DEBUG(4,
7768 						    "usr=%d grp=%d\n", user,
7769 						    grp);
7770 					}
7771 				}
7772 			}
7773 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7774 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7775 #ifdef notnow
7776 			/*
7777 			 * At this time, there is no flag.
7778 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7779 			 * added, or it may never exist?
7780 			 */
7781 			mtx_lock(&nmp->nm_mtx);
7782 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7783 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7784 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7785 			mtx_unlock(&nmp->nm_mtx);
7786 #endif
7787 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7788 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7789 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7790 		} else {
7791 			error = NFSERR_BADXDR;
7792 			goto nfsmout;
7793 		}
7794 		if (flp->nfsfl_iomode == gotiomode) {
7795 			/* Keep the list in increasing offset order. */
7796 			tflp = LIST_FIRST(flhp);
7797 			prevflp = NULL;
7798 			while (tflp != NULL &&
7799 			    tflp->nfsfl_off < flp->nfsfl_off) {
7800 				prevflp = tflp;
7801 				tflp = LIST_NEXT(tflp, nfsfl_list);
7802 			}
7803 			if (prevflp == NULL)
7804 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7805 			else
7806 				LIST_INSERT_AFTER(prevflp, flp,
7807 				    nfsfl_list);
7808 			NFSCL_DEBUG(4, "flp inserted\n");
7809 		} else {
7810 			printf("nfscl_layoutget(): got wrong iomode\n");
7811 			nfscl_freeflayout(flp);
7812 		}
7813 		flp = NULL;
7814 	}
7815 nfsmout:
7816 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7817 	if (error != 0 && flp != NULL)
7818 		nfscl_freeflayout(flp);
7819 	return (error);
7820 }
7821 
7822 /*
7823  * Parse a user/group digit string.
7824  */
7825 static int
7826 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7827     NFSPROC_T *p)
7828 {
7829 	uint32_t *tl;
7830 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7831 	uint32_t len = 0;
7832 	int error = 0;
7833 
7834 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7835 	len = fxdr_unsigned(uint32_t, *tl);
7836 	str = NULL;
7837 	if (len > NFSV4_OPAQUELIMIT) {
7838 		error = NFSERR_BADXDR;
7839 		goto nfsmout;
7840 	}
7841 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7842 	if (len == 0) {
7843 		if (dogrp != 0)
7844 			*gidp = GID_NOGROUP;
7845 		else
7846 			*uidp = UID_NOBODY;
7847 		return (0);
7848 	}
7849 	if (len > NFSV4_SMALLSTR)
7850 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7851 	else
7852 		str = str0;
7853 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7854 	NFSBCOPY(cp, str, len);
7855 	str[len] = '\0';
7856 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7857 	if (dogrp != 0)
7858 		error = nfsv4_strtogid(nd, str, len, gidp);
7859 	else
7860 		error = nfsv4_strtouid(nd, str, len, uidp);
7861 nfsmout:
7862 	if (len > NFSV4_SMALLSTR)
7863 		free(str, M_TEMP);
7864 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7865 	return (error);
7866 }
7867 
7868 /*
7869  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7870  * so that it does both an Open and a Layoutget.
7871  */
7872 static int
7873 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7874     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7875     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7876     struct ucred *cred, NFSPROC_T *p)
7877 {
7878 	struct nfscllayout *lyp;
7879 	struct nfsclflayout *flp;
7880 	struct nfsclflayouthead flh;
7881 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7882 	int layouttype, laystat;
7883 	nfsv4stateid_t stateid;
7884 	struct nfsclsession *tsep;
7885 
7886 	error = 0;
7887 	if (NFSHASFLEXFILE(nmp))
7888 		layouttype = NFSLAYOUT_FLEXFILE;
7889 	else
7890 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7891 	/*
7892 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7893 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7894 	 * flp == NULL.
7895 	 */
7896 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
7897 	    &recalled);
7898 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7899 	if (lyp == NULL)
7900 		islocked = 0;
7901 	else if (flp != NULL)
7902 		islocked = 1;
7903 	else
7904 		islocked = 2;
7905 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7906 		LIST_INIT(&flh);
7907 		tsep = nfsmnt_mdssession(nmp);
7908 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7909 		    3 * NFSX_UNSIGNED);
7910 		if (lyp == NULL)
7911 			usecurstateid = 1;
7912 		else {
7913 			usecurstateid = 0;
7914 			stateid.seqid = lyp->nfsly_stateid.seqid;
7915 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7916 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7917 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7918 		}
7919 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7920 		    newfhp, newfhlen, mode, op, name, namelen,
7921 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7922 		    &retonclose, &flh, &laystat, cred, p);
7923 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7924 		    laystat, error);
7925 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7926 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7927 		    &islocked, cred, p);
7928 	} else
7929 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7930 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7931 	if (islocked == 2)
7932 		nfscl_rellayout(lyp, 1);
7933 	else if (islocked == 1)
7934 		nfscl_rellayout(lyp, 0);
7935 	return (error);
7936 }
7937 
7938 /*
7939  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7940  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7941  * handled by nfsrpc_openrpc().
7942  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7943  * can be NULL.
7944  */
7945 static int
7946 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7947     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7948     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7949     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7950     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7951     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7952 {
7953 	uint32_t *tl;
7954 	struct nfsrv_descript nfsd, *nd = &nfsd;
7955 	struct nfscldeleg *ndp = NULL;
7956 	struct nfsvattr nfsva;
7957 	struct nfsclsession *tsep;
7958 	uint32_t rflags, deleg;
7959 	nfsattrbit_t attrbits;
7960 	int error, ret, acesize, limitby, iomode;
7961 
7962 	*dpp = NULL;
7963 	*laystatp = ENXIO;
7964 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7965 	    0, 0);
7966 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7967 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7968 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7969 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7970 	tsep = nfsmnt_mdssession(nmp);
7971 	*tl++ = tsep->nfsess_clientid.lval[0];
7972 	*tl = tsep->nfsess_clientid.lval[1];
7973 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7974 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7975 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7976 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7977 	nfsm_strtom(nd, name, namelen);
7978 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7979 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7980 	NFSZERO_ATTRBIT(&attrbits);
7981 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7982 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7983 	nfsrv_putattrbit(nd, &attrbits);
7984 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7985 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7986 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7987 		iomode = NFSLAYOUTIOMODE_RW;
7988 	else
7989 		iomode = NFSLAYOUTIOMODE_READ;
7990 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7991 	    layouttype, layoutlen, usecurstateid);
7992 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7993 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7994 	if (error != 0)
7995 		return (error);
7996 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7997 	if (nd->nd_repstat != 0)
7998 		*laystatp = nd->nd_repstat;
7999 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8000 		/* ND_NOMOREDATA will be set if the Open operation failed. */
8001 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8002 		    6 * NFSX_UNSIGNED);
8003 		op->nfso_stateid.seqid = *tl++;
8004 		op->nfso_stateid.other[0] = *tl++;
8005 		op->nfso_stateid.other[1] = *tl++;
8006 		op->nfso_stateid.other[2] = *tl;
8007 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
8008 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8009 		if (error != 0)
8010 			goto nfsmout;
8011 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8012 		deleg = fxdr_unsigned(u_int32_t, *tl);
8013 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8014 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8015 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
8016 			      NFSCLFLAGS_FIRSTDELEG))
8017 				op->nfso_own->nfsow_clp->nfsc_flags |=
8018 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8019 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
8020 			    M_NFSCLDELEG, M_WAITOK);
8021 			LIST_INIT(&ndp->nfsdl_owner);
8022 			LIST_INIT(&ndp->nfsdl_lock);
8023 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
8024 			ndp->nfsdl_fhlen = newfhlen;
8025 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
8026 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
8027 			nfscl_lockinit(&ndp->nfsdl_rwlock);
8028 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8029 			    NFSX_UNSIGNED);
8030 			ndp->nfsdl_stateid.seqid = *tl++;
8031 			ndp->nfsdl_stateid.other[0] = *tl++;
8032 			ndp->nfsdl_stateid.other[1] = *tl++;
8033 			ndp->nfsdl_stateid.other[2] = *tl++;
8034 			ret = fxdr_unsigned(int, *tl);
8035 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8036 				ndp->nfsdl_flags = NFSCLDL_WRITE;
8037 				/*
8038 				 * Indicates how much the file can grow.
8039 				 */
8040 				NFSM_DISSECT(tl, u_int32_t *,
8041 				    3 * NFSX_UNSIGNED);
8042 				limitby = fxdr_unsigned(int, *tl++);
8043 				switch (limitby) {
8044 				case NFSV4OPEN_LIMITSIZE:
8045 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
8046 					break;
8047 				case NFSV4OPEN_LIMITBLOCKS:
8048 					ndp->nfsdl_sizelimit =
8049 					    fxdr_unsigned(u_int64_t, *tl++);
8050 					ndp->nfsdl_sizelimit *=
8051 					    fxdr_unsigned(u_int64_t, *tl);
8052 					break;
8053 				default:
8054 					error = NFSERR_BADXDR;
8055 					goto nfsmout;
8056 				};
8057 			} else
8058 				ndp->nfsdl_flags = NFSCLDL_READ;
8059 			if (ret != 0)
8060 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
8061 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
8062 			    &ret, &acesize, p);
8063 			if (error != 0)
8064 				goto nfsmout;
8065 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8066 			error = NFSERR_BADXDR;
8067 			goto nfsmout;
8068 		}
8069 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
8070 		    nfscl_assumeposixlocks)
8071 			op->nfso_posixlock = 1;
8072 		else
8073 			op->nfso_posixlock = 0;
8074 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8075 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
8076 		if (*++tl == 0) {
8077 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
8078 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
8079 			    NULL, NULL, NULL, p, cred);
8080 			if (error != 0)
8081 				goto nfsmout;
8082 			if (ndp != NULL) {
8083 				ndp->nfsdl_change = nfsva.na_filerev;
8084 				ndp->nfsdl_modtime = nfsva.na_mtime;
8085 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8086 				*dpp = ndp;
8087 				ndp = NULL;
8088 			}
8089 			/*
8090 			 * At this point, the Open has succeeded, so set
8091 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
8092 			 * this function just won't return a layout.
8093 			 */
8094 			if (nd->nd_repstat == 0) {
8095 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8096 				*laystatp = fxdr_unsigned(int, *++tl);
8097 				if (*laystatp == 0) {
8098 					error = nfsrv_parselayoutget(nmp, nd,
8099 					    stateidp, retonclosep, flhp);
8100 					if (error != 0)
8101 						*laystatp = error;
8102 				}
8103 			} else
8104 				nd->nd_repstat = 0;	/* Return 0 for Open. */
8105 		}
8106 	}
8107 	if (nd->nd_repstat != 0 && error == 0)
8108 		error = nd->nd_repstat;
8109 nfsmout:
8110 	free(ndp, M_NFSCLDELEG);
8111 	m_freem(nd->nd_mrep);
8112 	return (error);
8113 }
8114 
8115 /*
8116  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
8117  * Used only for mounts with pNFS enabled.
8118  */
8119 static int
8120 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8121     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8122     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8123     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8124     int *dattrflagp, int *unlockedp, nfsv4stateid_t *stateidp,
8125     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
8126     struct nfsclflayouthead *flhp, int *laystatp)
8127 {
8128 	uint32_t *tl;
8129 	int error = 0, deleg, newone, ret, acesize, limitby;
8130 	struct nfsrv_descript nfsd, *nd = &nfsd;
8131 	struct nfsclopen *op;
8132 	struct nfscldeleg *dp = NULL;
8133 	struct nfsnode *np;
8134 	struct nfsfh *nfhp;
8135 	struct nfsclsession *tsep;
8136 	nfsattrbit_t attrbits;
8137 	nfsv4stateid_t stateid;
8138 	struct nfsmount *nmp;
8139 
8140 	nmp = VFSTONFS(dvp->v_mount);
8141 	np = VTONFS(dvp);
8142 	*laystatp = ENXIO;
8143 	*unlockedp = 0;
8144 	*nfhpp = NULL;
8145 	*dpp = NULL;
8146 	*attrflagp = 0;
8147 	*dattrflagp = 0;
8148 	if (namelen > NFS_MAXNAMLEN)
8149 		return (ENAMETOOLONG);
8150 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
8151 	/*
8152 	 * For V4, this is actually an Open op.
8153 	 */
8154 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
8155 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
8156 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8157 	    NFSV4OPEN_ACCESSREAD);
8158 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
8159 	tsep = nfsmnt_mdssession(nmp);
8160 	*tl++ = tsep->nfsess_clientid.lval[0];
8161 	*tl = tsep->nfsess_clientid.lval[1];
8162 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8163 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8164 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
8165 	if ((fmode & O_EXCL) != 0) {
8166 		if (NFSHASSESSPERSIST(nmp)) {
8167 			/* Use GUARDED for persistent sessions. */
8168 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
8169 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8170 		} else {
8171 			/* Otherwise, use EXCLUSIVE4_1. */
8172 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
8173 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
8174 			*tl++ = cverf.lval[0];
8175 			*tl = cverf.lval[1];
8176 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8177 		}
8178 	} else {
8179 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
8180 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
8181 	}
8182 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8183 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8184 	nfsm_strtom(nd, name, namelen);
8185 	/* Get the new file's handle and attributes, plus save the FH. */
8186 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
8187 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
8188 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
8189 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8190 	NFSGETATTR_ATTRBIT(&attrbits);
8191 	nfsrv_putattrbit(nd, &attrbits);
8192 	/* Get the directory's post-op attributes. */
8193 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8194 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8195 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
8196 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8197 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8198 	nfsrv_putattrbit(nd, &attrbits);
8199 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8200 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
8201 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8202 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
8203 	    layouttype, layoutlen, usecurstateid);
8204 	error = nfscl_request(nd, dvp, p, cred);
8205 	if (error != 0)
8206 		return (error);
8207 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
8208 	    error);
8209 	if (nd->nd_repstat != 0)
8210 		*laystatp = nd->nd_repstat;
8211 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
8212 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8213 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
8214 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8215 		    6 * NFSX_UNSIGNED);
8216 		stateid.seqid = *tl++;
8217 		stateid.other[0] = *tl++;
8218 		stateid.other[1] = *tl++;
8219 		stateid.other[2] = *tl;
8220 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8221 		if (error != 0)
8222 			goto nfsmout;
8223 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8224 		deleg = fxdr_unsigned(int, *tl);
8225 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8226 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8227 			if (!(owp->nfsow_clp->nfsc_flags &
8228 			      NFSCLFLAGS_FIRSTDELEG))
8229 				owp->nfsow_clp->nfsc_flags |=
8230 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8231 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
8232 			    M_NFSCLDELEG, M_WAITOK);
8233 			LIST_INIT(&dp->nfsdl_owner);
8234 			LIST_INIT(&dp->nfsdl_lock);
8235 			dp->nfsdl_clp = owp->nfsow_clp;
8236 			newnfs_copyincred(cred, &dp->nfsdl_cred);
8237 			nfscl_lockinit(&dp->nfsdl_rwlock);
8238 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8239 			    NFSX_UNSIGNED);
8240 			dp->nfsdl_stateid.seqid = *tl++;
8241 			dp->nfsdl_stateid.other[0] = *tl++;
8242 			dp->nfsdl_stateid.other[1] = *tl++;
8243 			dp->nfsdl_stateid.other[2] = *tl++;
8244 			ret = fxdr_unsigned(int, *tl);
8245 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8246 				dp->nfsdl_flags = NFSCLDL_WRITE;
8247 				/*
8248 				 * Indicates how much the file can grow.
8249 				 */
8250 				NFSM_DISSECT(tl, u_int32_t *,
8251 				    3 * NFSX_UNSIGNED);
8252 				limitby = fxdr_unsigned(int, *tl++);
8253 				switch (limitby) {
8254 				case NFSV4OPEN_LIMITSIZE:
8255 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
8256 					break;
8257 				case NFSV4OPEN_LIMITBLOCKS:
8258 					dp->nfsdl_sizelimit =
8259 					    fxdr_unsigned(u_int64_t, *tl++);
8260 					dp->nfsdl_sizelimit *=
8261 					    fxdr_unsigned(u_int64_t, *tl);
8262 					break;
8263 				default:
8264 					error = NFSERR_BADXDR;
8265 					goto nfsmout;
8266 				};
8267 			} else {
8268 				dp->nfsdl_flags = NFSCLDL_READ;
8269 			}
8270 			if (ret != 0)
8271 				dp->nfsdl_flags |= NFSCLDL_RECALL;
8272 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
8273 			    &ret, &acesize, p);
8274 			if (error != 0)
8275 				goto nfsmout;
8276 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8277 			error = NFSERR_BADXDR;
8278 			goto nfsmout;
8279 		}
8280 
8281 		/* Now, we should have the status for the SaveFH. */
8282 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8283 		if (*++tl == 0) {
8284 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
8285 			/*
8286 			 * Now, process the GetFH and Getattr for the newly
8287 			 * created file. nfscl_mtofh() will set
8288 			 * ND_NOMOREDATA if these weren't successful.
8289 			 */
8290 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
8291 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
8292 			if (error != 0)
8293 				goto nfsmout;
8294 		} else
8295 			nd->nd_flag |= ND_NOMOREDATA;
8296 		/* Now we have the PutFH and Getattr for the directory. */
8297 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8298 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8299 			if (*++tl != 0)
8300 				nd->nd_flag |= ND_NOMOREDATA;
8301 			else {
8302 				NFSM_DISSECT(tl, uint32_t *, 2 *
8303 				    NFSX_UNSIGNED);
8304 				if (*++tl != 0)
8305 					nd->nd_flag |= ND_NOMOREDATA;
8306 			}
8307 		}
8308 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8309 			/* Load the directory attributes. */
8310 			error = nfsm_loadattr(nd, dnap);
8311 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
8312 			if (error != 0)
8313 				goto nfsmout;
8314 			*dattrflagp = 1;
8315 			if (dp != NULL && *attrflagp != 0) {
8316 				dp->nfsdl_change = nnap->na_filerev;
8317 				dp->nfsdl_modtime = nnap->na_mtime;
8318 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8319 			}
8320 			/*
8321 			 * We can now complete the Open state.
8322 			 */
8323 			nfhp = *nfhpp;
8324 			if (dp != NULL) {
8325 				dp->nfsdl_fhlen = nfhp->nfh_len;
8326 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
8327 				    nfhp->nfh_len);
8328 			}
8329 			/*
8330 			 * Get an Open structure that will be
8331 			 * attached to the OpenOwner, acquired already.
8332 			 */
8333 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
8334 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
8335 			    cred, p, NULL, &op, &newone, NULL, 0, false);
8336 			if (error != 0)
8337 				goto nfsmout;
8338 			op->nfso_stateid = stateid;
8339 			newnfs_copyincred(cred, &op->nfso_cred);
8340 
8341 			nfscl_openrelease(nmp, op, error, newone);
8342 			*unlockedp = 1;
8343 
8344 			/* Now, handle the RestoreFH and LayoutGet. */
8345 			if (nd->nd_repstat == 0) {
8346 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
8347 				*laystatp = fxdr_unsigned(int, *(tl + 3));
8348 				if (*laystatp == 0) {
8349 					error = nfsrv_parselayoutget(nmp, nd,
8350 					    stateidp, retonclosep, flhp);
8351 					if (error != 0)
8352 						*laystatp = error;
8353 				}
8354 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
8355 				    error);
8356 			} else
8357 				nd->nd_repstat = 0;
8358 		}
8359 	}
8360 	if (nd->nd_repstat != 0 && error == 0)
8361 		error = nd->nd_repstat;
8362 	if (error == NFSERR_STALECLIENTID)
8363 		nfscl_initiate_recovery(owp->nfsow_clp);
8364 nfsmout:
8365 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
8366 	if (error == 0)
8367 		*dpp = dp;
8368 	else
8369 		free(dp, M_NFSCLDELEG);
8370 	m_freem(nd->nd_mrep);
8371 	return (error);
8372 }
8373 
8374 /*
8375  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
8376  */
8377 static int
8378 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8379     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8380     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8381     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8382     int *dattrflagp, int *unlockedp)
8383 {
8384 	struct nfscllayout *lyp;
8385 	struct nfsclflayouthead flh;
8386 	struct nfsfh *nfhp;
8387 	struct nfsclsession *tsep;
8388 	struct nfsmount *nmp;
8389 	nfsv4stateid_t stateid;
8390 	int error, layoutlen, layouttype, retonclose, laystat;
8391 
8392 	error = 0;
8393 	nmp = VFSTONFS(dvp->v_mount);
8394 	if (NFSHASFLEXFILE(nmp))
8395 		layouttype = NFSLAYOUT_FLEXFILE;
8396 	else
8397 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8398 	LIST_INIT(&flh);
8399 	tsep = nfsmnt_mdssession(nmp);
8400 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
8401 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
8402 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
8403 	    unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
8404 	    &flh, &laystat);
8405 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
8406 	    laystat, error);
8407 	lyp = NULL;
8408 	if (laystat == 0) {
8409 		nfhp = *nfhpp;
8410 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8411 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8412 		    layouttype, laystat, NULL, cred, p);
8413 	} else
8414 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8415 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8416 		    cred, p);
8417 	if (laystat == 0)
8418 		nfscl_rellayout(lyp, 0);
8419 	return (error);
8420 }
8421 
8422 /*
8423  * Process the results of a layoutget() operation.
8424  */
8425 static int
8426 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8427     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8428     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8429     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8430 {
8431 	struct nfsclflayout *tflp;
8432 	struct nfscldevinfo *dip;
8433 	uint8_t *dev;
8434 	int i, mirrorcnt;
8435 
8436 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8437 		NFSLOCKMNT(nmp);
8438 		if (!NFSHASFLEXFILE(nmp)) {
8439 			/* Switch to using Flex File Layout. */
8440 			nmp->nm_state |= NFSSTA_FLEXFILE;
8441 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8442 			/* Disable pNFS. */
8443 			NFSCL_DEBUG(1, "disable PNFS\n");
8444 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8445 		}
8446 		NFSUNLOCKMNT(nmp);
8447 	}
8448 	if (laystat == 0) {
8449 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8450 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8451 			if (layouttype == NFSLAYOUT_FLEXFILE)
8452 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8453 			else
8454 				mirrorcnt = 1;
8455 			for (i = 0; i < mirrorcnt; i++) {
8456 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8457 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8458 				if (laystat != 0) {
8459 					if (layouttype == NFSLAYOUT_FLEXFILE)
8460 						dev = tflp->nfsfl_ffm[i].dev;
8461 					else
8462 						dev = tflp->nfsfl_dev;
8463 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8464 					    layouttype, notifybit, &dip, cred,
8465 					    p);
8466 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8467 					    laystat);
8468 					if (laystat != 0)
8469 						goto out;
8470 					laystat = nfscl_adddevinfo(nmp, dip, i,
8471 					    tflp);
8472 					if (laystat != 0)
8473 						printf("nfsrpc_layoutgetresout"
8474 						    ": cannot add\n");
8475 				}
8476 			}
8477 		}
8478 	}
8479 out:
8480 	if (laystat == 0) {
8481 		/*
8482 		 * nfscl_layout() always returns with the nfsly_lock
8483 		 * set to a refcnt (shared lock).
8484 		 * Passing in dvp is sufficient, since it is only used to
8485 		 * get the fsid for the file system.
8486 		 */
8487 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8488 		    layouttype, retonclose, flhp, lypp, cred, p);
8489 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8490 		    laystat);
8491 		if (laystat == 0 && islockedp != NULL)
8492 			*islockedp = 1;
8493 	}
8494 	return (laystat);
8495 }
8496 
8497 /*
8498  * nfs copy_file_range operation.
8499  */
8500 int
8501 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8502     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8503     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8504     struct ucred *cred, bool consecutive, bool *must_commitp)
8505 {
8506 	int commit, error, expireret = 0, retrycnt;
8507 	u_int32_t clidrev = 0;
8508 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8509 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8510 	nfsv4stateid_t instateid, outstateid;
8511 	void *inlckp, *outlckp;
8512 
8513 	if (nmp->nm_clp != NULL)
8514 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8515 	innfhp = VTONFS(invp)->n_fhp;
8516 	outnfhp = VTONFS(outvp)->n_fhp;
8517 	retrycnt = 0;
8518 	do {
8519 		/* Get both stateids. */
8520 		inlckp = NULL;
8521 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8522 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8523 		    &inlckp);
8524 		outlckp = NULL;
8525 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8526 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8527 		    &outlckp);
8528 
8529 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8530 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8531 		    outattrflagp, consecutive, &commit, cred, curthread);
8532 		if (error == 0) {
8533 			if (commit != NFSWRITE_FILESYNC)
8534 				*must_commitp = true;
8535 			*inoffp += *lenp;
8536 			*outoffp += *lenp;
8537 		} else if (error == NFSERR_STALESTATEID)
8538 			nfscl_initiate_recovery(nmp->nm_clp);
8539 		if (inlckp != NULL)
8540 			nfscl_lockderef(inlckp);
8541 		if (outlckp != NULL)
8542 			nfscl_lockderef(outlckp);
8543 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8544 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8545 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8546 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8547 		} else if ((error == NFSERR_EXPIRED ||
8548 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8549 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8550 			    curthread);
8551 		}
8552 		retrycnt++;
8553 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8554 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8555 	      error == NFSERR_STALEDONTRECOVER ||
8556 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8557 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8558 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8559 	if (error != 0 && (retrycnt >= 4 ||
8560 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8561 	      error == NFSERR_STALEDONTRECOVER))
8562 		error = EIO;
8563 	return (error);
8564 }
8565 
8566 /*
8567  * The copy RPC.
8568  */
8569 static int
8570 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8571     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8572     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8573     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8574     NFSPROC_T *p)
8575 {
8576 	uint32_t *tl;
8577 	int error;
8578 	struct nfsrv_descript nfsd;
8579 	struct nfsrv_descript *nd = &nfsd;
8580 	struct nfsmount *nmp;
8581 	nfsattrbit_t attrbits;
8582 	uint64_t len;
8583 
8584 	nmp = VFSTONFS(outvp->v_mount);
8585 	*inattrflagp = *outattrflagp = 0;
8586 	*commitp = NFSWRITE_UNSTABLE;
8587 	len = *lenp;
8588 	*lenp = 0;
8589 	if (len > nfs_maxcopyrange)
8590 		len = nfs_maxcopyrange;
8591 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8592 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8593 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8594 	NFSGETATTR_ATTRBIT(&attrbits);
8595 	nfsrv_putattrbit(nd, &attrbits);
8596 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8597 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8598 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8599 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8600 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8601 	*tl = txdr_unsigned(NFSV4OP_COPY);
8602 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8603 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8604 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8605 	txdr_hyper(inoff, tl); tl += 2;
8606 	txdr_hyper(outoff, tl); tl += 2;
8607 	txdr_hyper(len, tl); tl += 2;
8608 	if (consecutive)
8609 		*tl++ = newnfs_true;
8610 	else
8611 		*tl++ = newnfs_false;
8612 	*tl++ = newnfs_true;
8613 	*tl++ = 0;
8614 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8615 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8616 	nfsrv_putattrbit(nd, &attrbits);
8617 	error = nfscl_request(nd, invp, p, cred);
8618 	if (error != 0)
8619 		return (error);
8620 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8621 		/* Get the input file's attributes. */
8622 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8623 		if (*(tl + 1) == 0) {
8624 			error = nfsm_loadattr(nd, innap);
8625 			if (error != 0)
8626 				goto nfsmout;
8627 			*inattrflagp = 1;
8628 		} else
8629 			nd->nd_flag |= ND_NOMOREDATA;
8630 	}
8631 	/* Skip over return stat for PutFH. */
8632 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8633 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8634 		if (*++tl != 0)
8635 			nd->nd_flag |= ND_NOMOREDATA;
8636 	}
8637 	/* Skip over return stat for Copy. */
8638 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8639 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8640 	if (nd->nd_repstat == 0) {
8641 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8642 		if (*tl != 0) {
8643 			/* There should be no callback ids. */
8644 			error = NFSERR_BADXDR;
8645 			goto nfsmout;
8646 		}
8647 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8648 		    NFSX_VERF);
8649 		len = fxdr_hyper(tl); tl += 2;
8650 		*commitp = fxdr_unsigned(int, *tl++);
8651 		NFSLOCKMNT(nmp);
8652 		if (!NFSHASWRITEVERF(nmp)) {
8653 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8654 			NFSSETWRITEVERF(nmp);
8655 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8656 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8657 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8658 		}
8659 		NFSUNLOCKMNT(nmp);
8660 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8661 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8662 			/* Must be a synchronous copy. */
8663 			nd->nd_repstat = NFSERR_NOTSUPP;
8664 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8665 		error = nfsm_loadattr(nd, outnap);
8666 		if (error == 0)
8667 			*outattrflagp = NFS_LATTR_NOSHRINK;
8668 		if (nd->nd_repstat == 0)
8669 			*lenp = len;
8670 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8671 		/*
8672 		 * For the case where consecutive is not supported, but
8673 		 * synchronous is supported, we can try consecutive == false
8674 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8675 		 * since Copy cannot be done.
8676 		 */
8677 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8678 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8679 			if (!consecutive || *++tl == newnfs_false)
8680 				nd->nd_repstat = NFSERR_NOTSUPP;
8681 		} else
8682 			nd->nd_repstat = NFSERR_BADXDR;
8683 	}
8684 	if (error == 0)
8685 		error = nd->nd_repstat;
8686 nfsmout:
8687 	m_freem(nd->nd_mrep);
8688 	return (error);
8689 }
8690 
8691 /*
8692  * Seek operation.
8693  */
8694 int
8695 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8696     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8697 {
8698 	int error, expireret = 0, retrycnt;
8699 	u_int32_t clidrev = 0;
8700 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
8701 	struct nfsnode *np = VTONFS(vp);
8702 	struct nfsfh *nfhp = NULL;
8703 	nfsv4stateid_t stateid;
8704 	void *lckp;
8705 
8706 	if (nmp->nm_clp != NULL)
8707 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8708 	nfhp = np->n_fhp;
8709 	retrycnt = 0;
8710 	do {
8711 		lckp = NULL;
8712 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8713 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8714 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8715 		    nap, attrflagp, cred);
8716 		if (error == NFSERR_STALESTATEID)
8717 			nfscl_initiate_recovery(nmp->nm_clp);
8718 		if (lckp != NULL)
8719 			nfscl_lockderef(lckp);
8720 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8721 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8722 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8723 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8724 		} else if ((error == NFSERR_EXPIRED ||
8725 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8726 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8727 			    curthread);
8728 		}
8729 		retrycnt++;
8730 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8731 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8732 	    error == NFSERR_BADSESSION ||
8733 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8734 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8735 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8736 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8737 	if (error && retrycnt >= 4)
8738 		error = EIO;
8739 	return (error);
8740 }
8741 
8742 /*
8743  * The seek RPC.
8744  */
8745 static int
8746 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8747     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8748 {
8749 	uint32_t *tl;
8750 	int error;
8751 	struct nfsrv_descript nfsd;
8752 	struct nfsrv_descript *nd = &nfsd;
8753 	nfsattrbit_t attrbits;
8754 
8755 	*attrflagp = 0;
8756 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8757 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8758 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8759 	txdr_hyper(*offp, tl); tl += 2;
8760 	*tl++ = txdr_unsigned(content);
8761 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8762 	NFSGETATTR_ATTRBIT(&attrbits);
8763 	nfsrv_putattrbit(nd, &attrbits);
8764 	error = nfscl_request(nd, vp, curthread, cred);
8765 	if (error != 0)
8766 		return (error);
8767 	if (nd->nd_repstat == 0) {
8768 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8769 		if (*tl++ == newnfs_true)
8770 			*eofp = true;
8771 		else
8772 			*eofp = false;
8773 		*offp = fxdr_hyper(tl);
8774 		/* Just skip over Getattr op status. */
8775 		error = nfsm_loadattr(nd, nap);
8776 		if (error == 0)
8777 			*attrflagp = 1;
8778 	}
8779 	error = nd->nd_repstat;
8780 nfsmout:
8781 	m_freem(nd->nd_mrep);
8782 	return (error);
8783 }
8784 
8785 /*
8786  * The getextattr RPC.
8787  */
8788 int
8789 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8790     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8791 {
8792 	uint32_t *tl;
8793 	int error;
8794 	struct nfsrv_descript nfsd;
8795 	struct nfsrv_descript *nd = &nfsd;
8796 	nfsattrbit_t attrbits;
8797 	uint32_t len, len2;
8798 
8799 	*attrflagp = 0;
8800 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8801 	nfsm_strtom(nd, name, strlen(name));
8802 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8803 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8804 	NFSGETATTR_ATTRBIT(&attrbits);
8805 	nfsrv_putattrbit(nd, &attrbits);
8806 	error = nfscl_request(nd, vp, p, cred);
8807 	if (error != 0)
8808 		return (error);
8809 	if (nd->nd_repstat == 0) {
8810 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8811 		len = fxdr_unsigned(uint32_t, *tl);
8812 		/* Sanity check lengths. */
8813 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8814 		    uiop->uio_resid <= UINT32_MAX) {
8815 			len2 = uiop->uio_resid;
8816 			if (len2 >= len)
8817 				error = nfsm_mbufuio(nd, uiop, len);
8818 			else {
8819 				error = nfsm_mbufuio(nd, uiop, len2);
8820 				if (error == 0) {
8821 					/*
8822 					 * nfsm_mbufuio() advances to a multiple
8823 					 * of 4, so round up len2 as well.  Then
8824 					 * we need to advance over the rest of
8825 					 * the data, rounding up the remaining
8826 					 * length.
8827 					 */
8828 					len2 = NFSM_RNDUP(len2);
8829 					len2 = NFSM_RNDUP(len - len2);
8830 					if (len2 > 0)
8831 						error = nfsm_advance(nd, len2,
8832 						    -1);
8833 				}
8834 			}
8835 		} else if (uiop == NULL && len > 0) {
8836 			/* Just wants the length and not the data. */
8837 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8838 		} else if (len > 0)
8839 			error = ENOATTR;
8840 		if (error != 0)
8841 			goto nfsmout;
8842 		*lenp = len;
8843 		/* Just skip over Getattr op status. */
8844 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8845 		error = nfsm_loadattr(nd, nap);
8846 		if (error == 0)
8847 			*attrflagp = 1;
8848 	}
8849 	if (error == 0)
8850 		error = nd->nd_repstat;
8851 nfsmout:
8852 	m_freem(nd->nd_mrep);
8853 	return (error);
8854 }
8855 
8856 /*
8857  * The setextattr RPC.
8858  */
8859 int
8860 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8861     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8862 {
8863 	uint32_t *tl;
8864 	int error;
8865 	struct nfsrv_descript nfsd;
8866 	struct nfsrv_descript *nd = &nfsd;
8867 	nfsattrbit_t attrbits;
8868 
8869 	*attrflagp = 0;
8870 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8871 	if (uiop->uio_resid > nd->nd_maxreq) {
8872 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8873 		m_freem(nd->nd_mreq);
8874 		return (EINVAL);
8875 	}
8876 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8877 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8878 	nfsm_strtom(nd, name, strlen(name));
8879 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8880 	*tl = txdr_unsigned(uiop->uio_resid);
8881 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8882 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8883 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8884 	NFSGETATTR_ATTRBIT(&attrbits);
8885 	nfsrv_putattrbit(nd, &attrbits);
8886 	error = nfscl_request(nd, vp, p, cred);
8887 	if (error != 0)
8888 		return (error);
8889 	if (nd->nd_repstat == 0) {
8890 		/* Just skip over the reply and Getattr op status. */
8891 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8892 		    NFSX_UNSIGNED);
8893 		error = nfsm_loadattr(nd, nap);
8894 		if (error == 0)
8895 			*attrflagp = 1;
8896 	}
8897 	if (error == 0)
8898 		error = nd->nd_repstat;
8899 nfsmout:
8900 	m_freem(nd->nd_mrep);
8901 	return (error);
8902 }
8903 
8904 /*
8905  * The removeextattr RPC.
8906  */
8907 int
8908 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8909     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8910 {
8911 	uint32_t *tl;
8912 	int error;
8913 	struct nfsrv_descript nfsd;
8914 	struct nfsrv_descript *nd = &nfsd;
8915 	nfsattrbit_t attrbits;
8916 
8917 	*attrflagp = 0;
8918 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8919 	nfsm_strtom(nd, name, strlen(name));
8920 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8921 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8922 	NFSGETATTR_ATTRBIT(&attrbits);
8923 	nfsrv_putattrbit(nd, &attrbits);
8924 	error = nfscl_request(nd, vp, p, cred);
8925 	if (error != 0)
8926 		return (error);
8927 	if (nd->nd_repstat == 0) {
8928 		/* Just skip over the reply and Getattr op status. */
8929 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8930 		    NFSX_UNSIGNED);
8931 		error = nfsm_loadattr(nd, nap);
8932 		if (error == 0)
8933 			*attrflagp = 1;
8934 	}
8935 	if (error == 0)
8936 		error = nd->nd_repstat;
8937 nfsmout:
8938 	m_freem(nd->nd_mrep);
8939 	return (error);
8940 }
8941 
8942 /*
8943  * The listextattr RPC.
8944  */
8945 int
8946 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8947     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8948     struct ucred *cred, NFSPROC_T *p)
8949 {
8950 	uint32_t *tl;
8951 	int cnt, error, i, len;
8952 	struct nfsrv_descript nfsd;
8953 	struct nfsrv_descript *nd = &nfsd;
8954 	nfsattrbit_t attrbits;
8955 	u_char c;
8956 
8957 	*attrflagp = 0;
8958 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8959 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8960 	txdr_hyper(*cookiep, tl); tl += 2;
8961 	*tl++ = txdr_unsigned(*lenp);
8962 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8963 	NFSGETATTR_ATTRBIT(&attrbits);
8964 	nfsrv_putattrbit(nd, &attrbits);
8965 	error = nfscl_request(nd, vp, p, cred);
8966 	if (error != 0)
8967 		return (error);
8968 	*eofp = true;
8969 	*lenp = 0;
8970 	if (nd->nd_repstat == 0) {
8971 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8972 		*cookiep = fxdr_hyper(tl); tl += 2;
8973 		cnt = fxdr_unsigned(int, *tl);
8974 		if (cnt < 0) {
8975 			error = EBADRPC;
8976 			goto nfsmout;
8977 		}
8978 		for (i = 0; i < cnt; i++) {
8979 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8980 			len = fxdr_unsigned(int, *tl);
8981 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8982 				error = EBADRPC;
8983 				goto nfsmout;
8984 			}
8985 			if (uiop == NULL)
8986 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8987 			else if (uiop->uio_resid >= len + 1) {
8988 				c = len;
8989 				error = uiomove(&c, sizeof(c), uiop);
8990 				if (error == 0)
8991 					error = nfsm_mbufuio(nd, uiop, len);
8992 			} else {
8993 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8994 				*eofp = false;
8995 			}
8996 			if (error != 0)
8997 				goto nfsmout;
8998 			*lenp += (len + 1);
8999 		}
9000 		/* Get the eof and skip over the Getattr op status. */
9001 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
9002 		/*
9003 		 * *eofp is set false above, because it wasn't able to copy
9004 		 * all of the reply.
9005 		 */
9006 		if (*eofp && *tl == 0)
9007 			*eofp = false;
9008 		error = nfsm_loadattr(nd, nap);
9009 		if (error == 0)
9010 			*attrflagp = 1;
9011 	}
9012 	if (error == 0)
9013 		error = nd->nd_repstat;
9014 nfsmout:
9015 	m_freem(nd->nd_mrep);
9016 	return (error);
9017 }
9018 
9019 /*
9020  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
9021  */
9022 static struct mbuf *
9023 nfsm_split(struct mbuf *mp, uint64_t xfer)
9024 {
9025 	struct mbuf *m, *m2;
9026 	vm_page_t pg;
9027 	int i, j, left, pgno, plen, trim;
9028 	char *cp, *cp2;
9029 
9030 	if ((mp->m_flags & M_EXTPG) == 0) {
9031 		m = m_split(mp, xfer, M_WAITOK);
9032 		return (m);
9033 	}
9034 
9035 	/* Find the correct mbuf to split at. */
9036 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
9037 		xfer -= m->m_len;
9038 	if (m == NULL)
9039 		return (NULL);
9040 
9041 	/* If xfer == m->m_len, we can just split the mbuf list. */
9042 	if (xfer == m->m_len) {
9043 		m2 = m->m_next;
9044 		m->m_next = NULL;
9045 		return (m2);
9046 	}
9047 
9048 	/* Find the page to split at. */
9049 	pgno = 0;
9050 	left = xfer;
9051 	do {
9052 		if (pgno == 0)
9053 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
9054 		else
9055 			plen = m_epg_pagelen(m, pgno, 0);
9056 		if (left <= plen)
9057 			break;
9058 		left -= plen;
9059 		pgno++;
9060 	} while (pgno < m->m_epg_npgs);
9061 	if (pgno == m->m_epg_npgs)
9062 		panic("nfsm_split: eroneous ext_pgs mbuf");
9063 
9064 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
9065 	m2->m_epg_flags |= EPG_FLAG_ANON;
9066 
9067 	/*
9068 	 * If left < plen, allocate a new page for the new mbuf
9069 	 * and copy the data after left in the page to this new
9070 	 * page.
9071 	 */
9072 	if (left < plen) {
9073 		pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
9074 		    VM_ALLOC_WIRED);
9075 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
9076 		m2->m_epg_npgs = 1;
9077 
9078 		/* Copy the data after left to the new page. */
9079 		trim = plen - left;
9080 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
9081 		if (pgno == 0)
9082 			cp += m->m_epg_1st_off;
9083 		cp += left;
9084 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
9085 		if (pgno == m->m_epg_npgs - 1)
9086 			m2->m_epg_last_len = trim;
9087 		else {
9088 			cp2 += PAGE_SIZE - trim;
9089 			m2->m_epg_1st_off = PAGE_SIZE - trim;
9090 			m2->m_epg_last_len = m->m_epg_last_len;
9091 		}
9092 		memcpy(cp2, cp, trim);
9093 		m2->m_len = trim;
9094 	} else {
9095 		m2->m_len = 0;
9096 		m2->m_epg_last_len = m->m_epg_last_len;
9097 	}
9098 
9099 	/* Move the pages beyond pgno to the new mbuf. */
9100 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
9101 		m2->m_epg_pa[j] = m->m_epg_pa[i];
9102 		/* Never moves page 0. */
9103 		m2->m_len += m_epg_pagelen(m, i, 0);
9104 	}
9105 	m2->m_epg_npgs = j;
9106 	m->m_epg_npgs = pgno + 1;
9107 	m->m_epg_last_len = left;
9108 	m->m_len = xfer;
9109 
9110 	m2->m_next = m->m_next;
9111 	m->m_next = NULL;
9112 	return (m2);
9113 }
9114 
9115 /*
9116  * Do the NFSv4.1 Bind Connection to Session.
9117  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
9118  */
9119 void
9120 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
9121 {
9122 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
9123 	uint32_t res, *tl;
9124 	struct nfsrv_descript nfsd;
9125 	struct nfsrv_descript *nd = &nfsd;
9126 	struct rpc_callextra ext;
9127 	struct timeval utimeout;
9128 	enum clnt_stat stat;
9129 	int error;
9130 
9131 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
9132 	    NFS_VER4, rcp->minorvers);
9133 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
9134 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
9135 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9136 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
9137 	*tl = newnfs_false;
9138 
9139 	memset(&ext, 0, sizeof(ext));
9140 	utimeout.tv_sec = 30;
9141 	utimeout.tv_usec = 0;
9142 	ext.rc_auth = authunix_create(cr);
9143 	nd->nd_mrep = NULL;
9144 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
9145 	    &nd->nd_mrep, utimeout);
9146 	AUTH_DESTROY(ext.rc_auth);
9147 	if (stat != RPC_SUCCESS) {
9148 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
9149 		return;
9150 	}
9151 	if (nd->nd_mrep == NULL) {
9152 		printf("nfsrpc_bindconnsess: no reply args\n");
9153 		return;
9154 	}
9155 	error = 0;
9156 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
9157 	nd->nd_md = nd->nd_mrep;
9158 	nd->nd_dpos = mtod(nd->nd_md, char *);
9159 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9160 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
9161 	if (nd->nd_repstat == NFSERR_OK) {
9162 		res = fxdr_unsigned(uint32_t, *tl);
9163 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
9164 		    -1)) != 0)
9165 			goto nfsmout;
9166 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
9167 		    4 * NFSX_UNSIGNED);
9168 		tl += 3;
9169 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
9170 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9171 			res = fxdr_unsigned(uint32_t, *tl);
9172 			if (res != NFSCDFS4_BOTH)
9173 				printf("nfsrpc_bindconnsess: did not "
9174 				    "return FS4_BOTH\n");
9175 		} else
9176 			printf("nfsrpc_bindconnsess: not same "
9177 			    "sessionid\n");
9178 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
9179 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
9180 nfsmout:
9181 	if (error != 0)
9182 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
9183 	m_freem(nd->nd_mrep);
9184 }
9185