xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 81ad626541db97eb356e2c1d4a20eb2a26a766ab)
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 *);
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, NULL);
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 (vp->v_type == 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 
286 	/*
287 	 * The NFS V3 spec does not clarify whether or not
288 	 * the returned access bits can be a superset of
289 	 * the ones requested, so...
290 	 */
291 	if (!error && (rmode & mode) != mode)
292 		error = EACCES;
293 	return (error);
294 }
295 
296 /*
297  * The actual rpc, separated out for Darwin.
298  */
299 int
300 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
301     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep)
302 {
303 	u_int32_t *tl;
304 	u_int32_t supported, rmode;
305 	int error;
306 	struct nfsrv_descript nfsd, *nd = &nfsd;
307 	nfsattrbit_t attrbits;
308 
309 	*attrflagp = 0;
310 	supported = mode;
311 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp, cred);
312 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
313 	*tl = txdr_unsigned(mode);
314 	if (nd->nd_flag & ND_NFSV4) {
315 		/*
316 		 * And do a Getattr op.
317 		 */
318 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
319 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
320 		NFSGETATTR_ATTRBIT(&attrbits);
321 		(void) nfsrv_putattrbit(nd, &attrbits);
322 	}
323 	error = nfscl_request(nd, vp, p, cred);
324 	if (error)
325 		return (error);
326 	if (nd->nd_flag & ND_NFSV3) {
327 		error = nfscl_postop_attr(nd, nap, attrflagp);
328 		if (error)
329 			goto nfsmout;
330 	}
331 	if (!nd->nd_repstat) {
332 		if (nd->nd_flag & ND_NFSV4) {
333 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
334 			supported = fxdr_unsigned(u_int32_t, *tl++);
335 		} else {
336 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
337 		}
338 		rmode = fxdr_unsigned(u_int32_t, *tl);
339 		if (nd->nd_flag & ND_NFSV4)
340 			error = nfscl_postop_attr(nd, nap, attrflagp);
341 
342 		/*
343 		 * It's not obvious what should be done about
344 		 * unsupported access modes. For now, be paranoid
345 		 * and clear the unsupported ones.
346 		 */
347 		rmode &= supported;
348 		*rmodep = rmode;
349 	} else
350 		error = nd->nd_repstat;
351 nfsmout:
352 	m_freem(nd->nd_mrep);
353 	return (error);
354 }
355 
356 /*
357  * nfs open rpc
358  */
359 int
360 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
361 {
362 	struct nfsclopen *op;
363 	struct nfscldeleg *dp;
364 	struct nfsfh *nfhp;
365 	struct nfsnode *np = VTONFS(vp);
366 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
367 	u_int32_t mode, clidrev;
368 	int ret, newone, error, expireret = 0, retrycnt;
369 
370 	/*
371 	 * For NFSv4, Open Ops are only done on Regular Files.
372 	 */
373 	if (vp->v_type != VREG)
374 		return (0);
375 	mode = 0;
376 	if (amode & FREAD)
377 		mode |= NFSV4OPEN_ACCESSREAD;
378 	if (amode & FWRITE)
379 		mode |= NFSV4OPEN_ACCESSWRITE;
380 	nfhp = np->n_fhp;
381 
382 	retrycnt = 0;
383 #ifdef notdef
384 { char name[100]; int namel;
385 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
386 bcopy(NFS4NODENAME(np->n_v4), name, namel);
387 name[namel] = '\0';
388 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
389 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
390 else printf(" fhl=0\n");
391 }
392 #endif
393 	do {
394 	    dp = NULL;
395 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
396 		cred, p, NULL, &op, &newone, &ret, 1, true);
397 	    if (error) {
398 		return (error);
399 	    }
400 	    if (nmp->nm_clp != NULL)
401 		clidrev = nmp->nm_clp->nfsc_clientidrev;
402 	    else
403 		clidrev = 0;
404 	    if (ret == NFSCLOPEN_DOOPEN) {
405 		if (np->n_v4 != NULL) {
406 			/*
407 			 * For the first attempt, try and get a layout, if
408 			 * pNFS is enabled for the mount.
409 			 */
410 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
411 			    nfs_numnfscbd == 0 ||
412 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
413 				error = nfsrpc_openrpc(nmp, vp,
414 				    np->n_v4->n4_data,
415 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
416 				    np->n_fhp->nfh_len, mode, op,
417 				    NFS4NODENAME(np->n_v4),
418 				    np->n_v4->n4_namelen,
419 				    &dp, 0, 0x0, cred, p, 0, 0);
420 			else
421 				error = nfsrpc_getopenlayout(nmp, vp,
422 				    np->n_v4->n4_data,
423 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
424 				    np->n_fhp->nfh_len, mode, op,
425 				    NFS4NODENAME(np->n_v4),
426 				    np->n_v4->n4_namelen, &dp, cred, p);
427 			if (dp != NULL) {
428 				NFSLOCKNODE(np);
429 				np->n_flag &= ~NDELEGMOD;
430 				/*
431 				 * Invalidate the attribute cache, so that
432 				 * attributes that pre-date the issue of a
433 				 * delegation are not cached, since the
434 				 * cached attributes will remain valid while
435 				 * the delegation is held.
436 				 */
437 				NFSINVALATTRCACHE(np);
438 				NFSUNLOCKNODE(np);
439 				(void) nfscl_deleg(nmp->nm_mountp,
440 				    op->nfso_own->nfsow_clp,
441 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
442 			}
443 		} else {
444 			error = EIO;
445 		}
446 		newnfs_copyincred(cred, &op->nfso_cred);
447 	    } else if (ret == NFSCLOPEN_SETCRED)
448 		/*
449 		 * This is a new local open on a delegation. It needs
450 		 * to have credentials so that an open can be done
451 		 * against the server during recovery.
452 		 */
453 		newnfs_copyincred(cred, &op->nfso_cred);
454 
455 	    /*
456 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
457 	     * been done on this Open successfully and a VOP_CLOSE()
458 	     * is expected for each of these.
459 	     * If error is non-zero, don't increment it, since the Open
460 	     * hasn't succeeded yet.
461 	     */
462 	    if (!error) {
463 		op->nfso_opencnt++;
464 		if (NFSHASNFSV4N(nmp) && NFSHASONEOPENOWN(nmp)) {
465 		    NFSLOCKNODE(np);
466 		    np->n_openstateid = op;
467 		    NFSUNLOCKNODE(np);
468 		}
469 	    }
470 	    nfscl_openrelease(nmp, op, error, newone);
471 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
472 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
473 		error == NFSERR_BADSESSION) {
474 		(void) nfs_catnap(PZERO, error, "nfs_open");
475 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
476 		&& clidrev != 0) {
477 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
478 		retrycnt++;
479 	    }
480 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
481 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
482 	    error == NFSERR_BADSESSION ||
483 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
484 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
485 	if (error && retrycnt >= 4)
486 		error = EIO;
487 	return (error);
488 }
489 
490 /*
491  * the actual open rpc
492  */
493 int
494 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
495     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
496     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
497     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
498     int syscred, int recursed)
499 {
500 	u_int32_t *tl;
501 	struct nfsrv_descript nfsd, *nd = &nfsd;
502 	struct nfscldeleg *dp, *ndp = NULL;
503 	struct nfsvattr nfsva;
504 	u_int32_t rflags, deleg;
505 	nfsattrbit_t attrbits;
506 	int error, ret, acesize, limitby;
507 	struct nfsclsession *tsep;
508 
509 	dp = *dpp;
510 	*dpp = NULL;
511 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0,
512 	    cred);
513 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
514 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
515 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
516 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
517 	tsep = nfsmnt_mdssession(nmp);
518 	*tl++ = tsep->nfsess_clientid.lval[0];
519 	*tl = tsep->nfsess_clientid.lval[1];
520 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
521 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
522 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
523 	if (reclaim) {
524 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
525 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
526 		*tl = txdr_unsigned(delegtype);
527 	} else {
528 		if (dp != NULL) {
529 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
530 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
531 			if (NFSHASNFSV4N(nmp))
532 				*tl++ = 0;
533 			else
534 				*tl++ = dp->nfsdl_stateid.seqid;
535 			*tl++ = dp->nfsdl_stateid.other[0];
536 			*tl++ = dp->nfsdl_stateid.other[1];
537 			*tl = dp->nfsdl_stateid.other[2];
538 		} else {
539 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
540 		}
541 		(void) nfsm_strtom(nd, name, namelen);
542 	}
543 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
544 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
545 	NFSZERO_ATTRBIT(&attrbits);
546 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
547 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
548 	(void) nfsrv_putattrbit(nd, &attrbits);
549 	if (syscred)
550 		nd->nd_flag |= ND_USEGSSNAME;
551 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
552 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
553 	if (error)
554 		return (error);
555 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
556 	if (!nd->nd_repstat) {
557 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
558 		    6 * NFSX_UNSIGNED);
559 		op->nfso_stateid.seqid = *tl++;
560 		op->nfso_stateid.other[0] = *tl++;
561 		op->nfso_stateid.other[1] = *tl++;
562 		op->nfso_stateid.other[2] = *tl;
563 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
564 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
565 		if (error)
566 			goto nfsmout;
567 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
568 		deleg = fxdr_unsigned(u_int32_t, *tl);
569 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
570 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
571 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
572 			      NFSCLFLAGS_FIRSTDELEG))
573 				op->nfso_own->nfsow_clp->nfsc_flags |=
574 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
575 			ndp = malloc(
576 			    sizeof (struct nfscldeleg) + newfhlen,
577 			    M_NFSCLDELEG, M_WAITOK);
578 			LIST_INIT(&ndp->nfsdl_owner);
579 			LIST_INIT(&ndp->nfsdl_lock);
580 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
581 			ndp->nfsdl_fhlen = newfhlen;
582 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
583 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
584 			nfscl_lockinit(&ndp->nfsdl_rwlock);
585 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
586 			    NFSX_UNSIGNED);
587 			ndp->nfsdl_stateid.seqid = *tl++;
588 			ndp->nfsdl_stateid.other[0] = *tl++;
589 			ndp->nfsdl_stateid.other[1] = *tl++;
590 			ndp->nfsdl_stateid.other[2] = *tl++;
591 			ret = fxdr_unsigned(int, *tl);
592 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
593 				ndp->nfsdl_flags = NFSCLDL_WRITE;
594 				/*
595 				 * Indicates how much the file can grow.
596 				 */
597 				NFSM_DISSECT(tl, u_int32_t *,
598 				    3 * NFSX_UNSIGNED);
599 				limitby = fxdr_unsigned(int, *tl++);
600 				switch (limitby) {
601 				case NFSV4OPEN_LIMITSIZE:
602 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
603 					break;
604 				case NFSV4OPEN_LIMITBLOCKS:
605 					ndp->nfsdl_sizelimit =
606 					    fxdr_unsigned(u_int64_t, *tl++);
607 					ndp->nfsdl_sizelimit *=
608 					    fxdr_unsigned(u_int64_t, *tl);
609 					break;
610 				default:
611 					error = NFSERR_BADXDR;
612 					goto nfsmout;
613 				}
614 			} else {
615 				ndp->nfsdl_flags = NFSCLDL_READ;
616 			}
617 			if (ret)
618 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
619 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
620 			    &ret, &acesize, p);
621 			if (error)
622 				goto nfsmout;
623 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
624 			error = NFSERR_BADXDR;
625 			goto nfsmout;
626 		}
627 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
628 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
629 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
630 		    NULL, NULL, NULL, p, cred);
631 		if (error)
632 			goto nfsmout;
633 		if (ndp != NULL) {
634 			ndp->nfsdl_change = nfsva.na_filerev;
635 			ndp->nfsdl_modtime = nfsva.na_mtime;
636 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
637 		}
638 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
639 		    do {
640 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
641 			    cred, p);
642 			if (ret == NFSERR_DELAY)
643 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
644 		    } while (ret == NFSERR_DELAY);
645 		    error = ret;
646 		}
647 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
648 		    nfscl_assumeposixlocks)
649 		    op->nfso_posixlock = 1;
650 		else
651 		    op->nfso_posixlock = 0;
652 
653 		/*
654 		 * If the server is handing out delegations, but we didn't
655 		 * get one because an OpenConfirm was required, try the
656 		 * Open again, to get a delegation. This is a harmless no-op,
657 		 * from a server's point of view.
658 		 */
659 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
660 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
661 		    && !error && dp == NULL && ndp == NULL && !recursed) {
662 		    do {
663 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
664 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
665 			    cred, p, syscred, 1);
666 			if (ret == NFSERR_DELAY)
667 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
668 		    } while (ret == NFSERR_DELAY);
669 		    if (ret) {
670 			if (ndp != NULL) {
671 				free(ndp, M_NFSCLDELEG);
672 				ndp = NULL;
673 			}
674 			if (ret == NFSERR_STALECLIENTID ||
675 			    ret == NFSERR_STALEDONTRECOVER ||
676 			    ret == NFSERR_BADSESSION)
677 				error = ret;
678 		    }
679 		}
680 	}
681 	if (nd->nd_repstat != 0 && error == 0)
682 		error = nd->nd_repstat;
683 	if (error == NFSERR_STALECLIENTID)
684 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
685 nfsmout:
686 	if (!error)
687 		*dpp = ndp;
688 	else if (ndp != NULL)
689 		free(ndp, M_NFSCLDELEG);
690 	m_freem(nd->nd_mrep);
691 	return (error);
692 }
693 
694 /*
695  * open downgrade rpc
696  */
697 int
698 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
699     struct ucred *cred, NFSPROC_T *p)
700 {
701 	u_int32_t *tl;
702 	struct nfsrv_descript nfsd, *nd = &nfsd;
703 	int error;
704 
705 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp, cred);
706 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
707 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
708 		*tl++ = 0;
709 	else
710 		*tl++ = op->nfso_stateid.seqid;
711 	*tl++ = op->nfso_stateid.other[0];
712 	*tl++ = op->nfso_stateid.other[1];
713 	*tl++ = op->nfso_stateid.other[2];
714 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
715 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
716 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
717 	error = nfscl_request(nd, vp, p, cred);
718 	if (error)
719 		return (error);
720 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
721 	if (!nd->nd_repstat) {
722 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
723 		op->nfso_stateid.seqid = *tl++;
724 		op->nfso_stateid.other[0] = *tl++;
725 		op->nfso_stateid.other[1] = *tl++;
726 		op->nfso_stateid.other[2] = *tl;
727 	}
728 	if (nd->nd_repstat && error == 0)
729 		error = nd->nd_repstat;
730 	if (error == NFSERR_STALESTATEID)
731 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
732 nfsmout:
733 	m_freem(nd->nd_mrep);
734 	return (error);
735 }
736 
737 /*
738  * V4 Close operation.
739  */
740 int
741 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
742 {
743 	struct nfsclclient *clp;
744 	int error;
745 
746 	if (vp->v_type != VREG)
747 		return (0);
748 	if (doclose)
749 		error = nfscl_doclose(vp, &clp, p);
750 	else {
751 		error = nfscl_getclose(vp, &clp);
752 		if (error == 0)
753 			nfscl_clientrelease(clp);
754 	}
755 	return (error);
756 }
757 
758 /*
759  * Close the open.
760  */
761 int
762 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p,
763     bool loop_on_delayed, bool freeop)
764 {
765 	struct nfsrv_descript nfsd, *nd = &nfsd;
766 	struct nfscllockowner *lp, *nlp;
767 	struct nfscllock *lop, *nlop;
768 	struct ucred *tcred;
769 	u_int64_t off = 0, len = 0;
770 	u_int32_t type = NFSV4LOCKT_READ;
771 	int error, do_unlock, trycnt;
772 
773 	tcred = newnfs_getcred();
774 	newnfs_copycred(&op->nfso_cred, tcred);
775 	/*
776 	 * (Theoretically this could be done in the same
777 	 *  compound as the close, but having multiple
778 	 *  sequenced Ops in the same compound might be
779 	 *  too scary for some servers.)
780 	 */
781 	if (op->nfso_posixlock) {
782 		off = 0;
783 		len = NFS64BITSSET;
784 		type = NFSV4LOCKT_READ;
785 	}
786 
787 	/*
788 	 * Since this function is only called from VOP_INACTIVE(), no
789 	 * other thread will be manipulating this Open. As such, the
790 	 * lock lists are not being changed by other threads, so it should
791 	 * be safe to do this without locking.
792 	 */
793 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
794 		do_unlock = 1;
795 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
796 			if (op->nfso_posixlock == 0) {
797 				off = lop->nfslo_first;
798 				len = lop->nfslo_end - lop->nfslo_first;
799 				if (lop->nfslo_type == F_WRLCK)
800 					type = NFSV4LOCKT_WRITE;
801 				else
802 					type = NFSV4LOCKT_READ;
803 			}
804 			if (do_unlock) {
805 				trycnt = 0;
806 				do {
807 					error = nfsrpc_locku(nd, nmp, lp, off,
808 					    len, type, tcred, p, 0);
809 					if ((nd->nd_repstat == NFSERR_GRACE ||
810 					    nd->nd_repstat == NFSERR_DELAY) &&
811 					    error == 0)
812 						(void) nfs_catnap(PZERO,
813 						    (int)nd->nd_repstat,
814 						    "nfs_close");
815 				} while ((nd->nd_repstat == NFSERR_GRACE ||
816 				    nd->nd_repstat == NFSERR_DELAY) &&
817 				    error == 0 && trycnt++ < 5);
818 				if (op->nfso_posixlock)
819 					do_unlock = 0;
820 			}
821 			nfscl_freelock(lop, 0);
822 		}
823 		/*
824 		 * Do a ReleaseLockOwner.
825 		 * The lock owner name nfsl_owner may be used by other opens for
826 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
827 		 * puts on the wire has the file handle for this file appended
828 		 * to it, so it can be done now.
829 		 */
830 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
831 		    lp->nfsl_open->nfso_fhlen, tcred, p);
832 	}
833 
834 	/*
835 	 * There could be other Opens for different files on the same
836 	 * OpenOwner, so locking is required.
837 	 */
838 	NFSLOCKCLSTATE();
839 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
840 	NFSUNLOCKCLSTATE();
841 	do {
842 		error = nfscl_tryclose(op, tcred, nmp, p, loop_on_delayed);
843 		if (error == NFSERR_GRACE)
844 			(void) nfs_catnap(PZERO, error, "nfs_close");
845 	} while (error == NFSERR_GRACE);
846 	NFSLOCKCLSTATE();
847 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
848 
849 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
850 		nfscl_freelockowner(lp, 0);
851 	if (freeop && error != NFSERR_DELAY)
852 		nfscl_freeopen(op, 0, true);
853 	NFSUNLOCKCLSTATE();
854 	NFSFREECRED(tcred);
855 	return (error);
856 }
857 
858 /*
859  * The actual Close RPC.
860  */
861 int
862 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
863     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
864     int syscred)
865 {
866 	u_int32_t *tl;
867 	int error;
868 
869 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
870 	    op->nfso_fhlen, NULL, NULL, 0, 0, cred);
871 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
872 	if (NFSHASNFSV4N(nmp)) {
873 		*tl++ = 0;
874 		*tl++ = 0;
875 	} else {
876 		*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
877 		*tl++ = op->nfso_stateid.seqid;
878 	}
879 	*tl++ = op->nfso_stateid.other[0];
880 	*tl++ = op->nfso_stateid.other[1];
881 	*tl = op->nfso_stateid.other[2];
882 	if (syscred)
883 		nd->nd_flag |= ND_USEGSSNAME;
884 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
885 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
886 	if (error)
887 		return (error);
888 	if (!NFSHASNFSV4N(nmp))
889 		NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
890 	if (nd->nd_repstat == 0)
891 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
892 	error = nd->nd_repstat;
893 	if (!NFSHASNFSV4N(nmp) && error == NFSERR_STALESTATEID)
894 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
895 nfsmout:
896 	m_freem(nd->nd_mrep);
897 	return (error);
898 }
899 
900 /*
901  * V4 Open Confirm RPC.
902  */
903 int
904 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
905     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
906 {
907 	u_int32_t *tl;
908 	struct nfsrv_descript nfsd, *nd = &nfsd;
909 	struct nfsmount *nmp;
910 	int error;
911 
912 	nmp = VFSTONFS(vp->v_mount);
913 	if (NFSHASNFSV4N(nmp))
914 		return (0);		/* No confirmation for NFSv4.1. */
915 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
916 	    0, 0, NULL);
917 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
918 	*tl++ = op->nfso_stateid.seqid;
919 	*tl++ = op->nfso_stateid.other[0];
920 	*tl++ = op->nfso_stateid.other[1];
921 	*tl++ = op->nfso_stateid.other[2];
922 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
923 	error = nfscl_request(nd, vp, p, cred);
924 	if (error)
925 		return (error);
926 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
927 	if (!nd->nd_repstat) {
928 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
929 		op->nfso_stateid.seqid = *tl++;
930 		op->nfso_stateid.other[0] = *tl++;
931 		op->nfso_stateid.other[1] = *tl++;
932 		op->nfso_stateid.other[2] = *tl;
933 	}
934 	error = nd->nd_repstat;
935 	if (error == NFSERR_STALESTATEID)
936 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
937 nfsmout:
938 	m_freem(nd->nd_mrep);
939 	return (error);
940 }
941 
942 /*
943  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
944  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
945  */
946 int
947 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
948     bool *retokp, struct ucred *cred, NFSPROC_T *p)
949 {
950 	u_int32_t *tl;
951 	struct nfsrv_descript nfsd;
952 	struct nfsrv_descript *nd = &nfsd;
953 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
954 	u_short port;
955 	int error, isinet6 = 0, callblen;
956 	nfsquad_t confirm;
957 	static u_int32_t rev = 0;
958 	struct nfsclds *dsp, *odsp;
959 	struct in6_addr a6;
960 	struct nfsclsession *tsep;
961 	struct rpc_reconupcall recon;
962 	struct nfscl_reconarg *rcp;
963 
964 	if (nfsboottime.tv_sec == 0)
965 		NFSSETBOOTTIME(nfsboottime);
966 	if (NFSHASNFSV4N(nmp)) {
967 		error = NFSERR_BADSESSION;
968 		odsp = dsp = NULL;
969 		if (retokp != NULL) {
970 			NFSLOCKMNT(nmp);
971 			odsp = TAILQ_FIRST(&nmp->nm_sess);
972 			NFSUNLOCKMNT(nmp);
973 		}
974 		if (odsp != NULL) {
975 			/*
976 			 * When a session already exists, first try a
977 			 * CreateSession with the extant ClientID.
978 			 */
979 			dsp = malloc(sizeof(struct nfsclds) +
980 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
981 			    M_WAITOK | M_ZERO);
982 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
983 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
984 			dsp->nfsclds_sess.nfsess_clientid =
985 			    odsp->nfsclds_sess.nfsess_clientid;
986 			dsp->nfsclds_sess.nfsess_sequenceid =
987 			    odsp->nfsclds_sess.nfsess_sequenceid + 1;
988 			dsp->nfsclds_flags = odsp->nfsclds_flags;
989 			if (dsp->nfsclds_servownlen > 0)
990 				memcpy(dsp->nfsclds_serverown,
991 				    odsp->nfsclds_serverown,
992 				    dsp->nfsclds_servownlen + 1);
993 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
994 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
995 			    NULL, MTX_DEF);
996 			nfscl_initsessionslots(&dsp->nfsclds_sess);
997 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
998 			    &nmp->nm_sockreq, NULL,
999 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
1000 			NFSCL_DEBUG(1, "create session for extant "
1001 			    "ClientID=%d\n", error);
1002 			if (error != 0) {
1003 				nfscl_freenfsclds(dsp);
1004 				dsp = NULL;
1005 				/*
1006 				 * If *retokp is true, return any error other
1007 				 * than NFSERR_STALECLIENTID,
1008 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
1009 				 * so that nfscl_recover() will not loop.
1010 				 */
1011 				if (*retokp)
1012 					return (NFSERR_IO);
1013 			} else
1014 				*retokp = true;
1015 		} else if (retokp != NULL && *retokp)
1016 			return (NFSERR_IO);
1017 		if (error != 0) {
1018 			/*
1019 			 * Either there was no previous session or the
1020 			 * CreateSession attempt failed, so...
1021 			 * do an ExchangeID followed by the CreateSession.
1022 			 */
1023 			clp->nfsc_rev = rev++;
1024 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1025 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1026 			    cred, p);
1027 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1028 			if (error == 0)
1029 				error = nfsrpc_createsession(nmp,
1030 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1031 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1032 				    cred, p);
1033 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1034 		}
1035 		if (error == 0) {
1036 			/*
1037 			 * If the session supports a backchannel, set up
1038 			 * the BindConnectionToSession call in the krpc
1039 			 * so that it is done on a reconnection.
1040 			 */
1041 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1042 				rcp = mem_alloc(sizeof(*rcp));
1043 				rcp->minorvers = nmp->nm_minorvers;
1044 				memcpy(rcp->sessionid,
1045 				    dsp->nfsclds_sess.nfsess_sessionid,
1046 				    NFSX_V4SESSIONID);
1047 				recon.call = nfsrpc_bindconnsess;
1048 				recon.arg = rcp;
1049 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1050 				    &recon);
1051 			}
1052 
1053 			NFSLOCKMNT(nmp);
1054 			/*
1055 			 * The old sessions cannot be safely free'd
1056 			 * here, since they may still be used by
1057 			 * in-progress RPCs.
1058 			 */
1059 			tsep = NULL;
1060 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL) {
1061 				/*
1062 				 * Mark the old session defunct.  Needed
1063 				 * when called from nfscl_hasexpired().
1064 				 */
1065 				tsep = NFSMNT_MDSSESSION(nmp);
1066 				tsep->nfsess_defunct = 1;
1067 			}
1068 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1069 			    nfsclds_list);
1070 			/*
1071 			 * Wake up RPCs waiting for a slot on the
1072 			 * old session. These will then fail with
1073 			 * NFSERR_BADSESSION and be retried with the
1074 			 * new session by nfsv4_setsequence().
1075 			 * Also wakeup() processes waiting for the
1076 			 * new session.
1077 			 */
1078 			if (tsep != NULL)
1079 				wakeup(&tsep->nfsess_slots);
1080 			wakeup(&nmp->nm_sess);
1081 			NFSUNLOCKMNT(nmp);
1082 		} else if (dsp != NULL)
1083 			nfscl_freenfsclds(dsp);
1084 		if (error == 0 && reclaim == 0) {
1085 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1086 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1087 			if (error == NFSERR_COMPLETEALREADY ||
1088 			    error == NFSERR_NOTSUPP)
1089 				/* Ignore this error. */
1090 				error = 0;
1091 		}
1092 		return (error);
1093 	} else if (retokp != NULL && *retokp)
1094 		return (NFSERR_IO);
1095 	clp->nfsc_rev = rev++;
1096 
1097 	/*
1098 	 * Allocate a single session structure for NFSv4.0, because some of
1099 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1100 	 */
1101 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1102 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1103 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1104 	NFSLOCKMNT(nmp);
1105 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1106 	tsep = NFSMNT_MDSSESSION(nmp);
1107 	NFSUNLOCKMNT(nmp);
1108 
1109 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0,
1110 	    NULL);
1111 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1112 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1113 	*tl = txdr_unsigned(clp->nfsc_rev);
1114 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1115 
1116 	/*
1117 	 * set up the callback address
1118 	 */
1119 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1120 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1121 	callblen = strlen(nfsv4_callbackaddr);
1122 	if (callblen == 0)
1123 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1124 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1125 	    (callblen > 0 || cp != NULL)) {
1126 		port = htons(nfsv4_cbport);
1127 		cp2 = (u_int8_t *)&port;
1128 #ifdef INET6
1129 		if ((callblen > 0 &&
1130 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1131 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1132 
1133 			(void) nfsm_strtom(nd, "tcp6", 4);
1134 			if (callblen == 0) {
1135 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1136 				ip6add = ip6buf;
1137 			} else {
1138 				ip6add = nfsv4_callbackaddr;
1139 			}
1140 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1141 			    ip6add, cp2[0], cp2[1]);
1142 		} else
1143 #endif
1144 		{
1145 			(void) nfsm_strtom(nd, "tcp", 3);
1146 			if (callblen == 0)
1147 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1148 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1149 				    cp[2], cp[3], cp2[0], cp2[1]);
1150 			else
1151 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1152 				    "%s.%d.%d", nfsv4_callbackaddr,
1153 				    cp2[0], cp2[1]);
1154 		}
1155 		(void) nfsm_strtom(nd, addr, strlen(addr));
1156 	} else {
1157 		(void) nfsm_strtom(nd, "tcp", 3);
1158 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1159 	}
1160 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1161 	*tl = txdr_unsigned(clp->nfsc_cbident);
1162 	nd->nd_flag |= ND_USEGSSNAME;
1163 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1164 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1165 	if (error)
1166 		return (error);
1167 	if (nd->nd_repstat == 0) {
1168 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1169 	    tsep->nfsess_clientid.lval[0] = *tl++;
1170 	    tsep->nfsess_clientid.lval[1] = *tl++;
1171 	    confirm.lval[0] = *tl++;
1172 	    confirm.lval[1] = *tl;
1173 	    m_freem(nd->nd_mrep);
1174 	    nd->nd_mrep = NULL;
1175 
1176 	    /*
1177 	     * and confirm it.
1178 	     */
1179 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1180 		NULL, 0, 0, NULL);
1181 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1182 	    *tl++ = tsep->nfsess_clientid.lval[0];
1183 	    *tl++ = tsep->nfsess_clientid.lval[1];
1184 	    *tl++ = confirm.lval[0];
1185 	    *tl = confirm.lval[1];
1186 	    nd->nd_flag |= ND_USEGSSNAME;
1187 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1188 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1189 	    if (error)
1190 		return (error);
1191 	    m_freem(nd->nd_mrep);
1192 	    nd->nd_mrep = NULL;
1193 	}
1194 	error = nd->nd_repstat;
1195 nfsmout:
1196 	m_freem(nd->nd_mrep);
1197 	return (error);
1198 }
1199 
1200 /*
1201  * nfs getattr call.
1202  */
1203 int
1204 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1205     struct nfsvattr *nap)
1206 {
1207 	struct nfsrv_descript nfsd, *nd = &nfsd;
1208 	int error;
1209 	nfsattrbit_t attrbits;
1210 
1211 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
1212 	if (nd->nd_flag & ND_NFSV4) {
1213 		NFSGETATTR_ATTRBIT(&attrbits);
1214 		(void) nfsrv_putattrbit(nd, &attrbits);
1215 	}
1216 	error = nfscl_request(nd, vp, p, cred);
1217 	if (error)
1218 		return (error);
1219 	if (!nd->nd_repstat)
1220 		error = nfsm_loadattr(nd, nap);
1221 	else
1222 		error = nd->nd_repstat;
1223 	m_freem(nd->nd_mrep);
1224 	return (error);
1225 }
1226 
1227 /*
1228  * nfs getattr call with non-vnode arguments.
1229  */
1230 int
1231 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1232     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1233     uint32_t *leasep)
1234 {
1235 	struct nfsrv_descript nfsd, *nd = &nfsd;
1236 	int error, vers = NFS_VER2;
1237 	nfsattrbit_t attrbits;
1238 
1239 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0,
1240 	    cred);
1241 	if (nd->nd_flag & ND_NFSV4) {
1242 		vers = NFS_VER4;
1243 		NFSGETATTR_ATTRBIT(&attrbits);
1244 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1245 		(void) nfsrv_putattrbit(nd, &attrbits);
1246 	} else if (nd->nd_flag & ND_NFSV3) {
1247 		vers = NFS_VER3;
1248 	}
1249 	if (syscred)
1250 		nd->nd_flag |= ND_USEGSSNAME;
1251 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1252 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1253 	if (error)
1254 		return (error);
1255 	if (nd->nd_repstat == 0) {
1256 		if ((nd->nd_flag & ND_NFSV4) != 0)
1257 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1258 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1259 			    NULL, NULL);
1260 		else
1261 			error = nfsm_loadattr(nd, nap);
1262 	} else
1263 		error = nd->nd_repstat;
1264 	m_freem(nd->nd_mrep);
1265 	return (error);
1266 }
1267 
1268 /*
1269  * Do an nfs setattr operation.
1270  */
1271 int
1272 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1273     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp)
1274 {
1275 	int error, expireret = 0, openerr, retrycnt;
1276 	u_int32_t clidrev = 0, mode;
1277 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1278 	struct nfsfh *nfhp;
1279 	nfsv4stateid_t stateid;
1280 	void *lckp;
1281 
1282 	if (nmp->nm_clp != NULL)
1283 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1284 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1285 		mode = NFSV4OPEN_ACCESSWRITE;
1286 	else
1287 		mode = NFSV4OPEN_ACCESSREAD;
1288 	retrycnt = 0;
1289 	do {
1290 		lckp = NULL;
1291 		openerr = 1;
1292 		if (NFSHASNFSV4(nmp)) {
1293 			nfhp = VTONFS(vp)->n_fhp;
1294 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1295 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1296 			if (error && vp->v_type == VREG &&
1297 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1298 			     nfstest_openallsetattr)) {
1299 				/*
1300 				 * No Open stateid, so try and open the file
1301 				 * now.
1302 				 */
1303 				if (mode == NFSV4OPEN_ACCESSWRITE)
1304 					openerr = nfsrpc_open(vp, FWRITE, cred,
1305 					    p);
1306 				else
1307 					openerr = nfsrpc_open(vp, FREAD, cred,
1308 					    p);
1309 				if (!openerr)
1310 					(void) nfscl_getstateid(vp,
1311 					    nfhp->nfh_fh, nfhp->nfh_len,
1312 					    mode, 0, cred, p, &stateid, &lckp);
1313 			}
1314 		}
1315 		if (vap != NULL)
1316 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1317 			    rnap, attrflagp);
1318 		else
1319 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid);
1320 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1321 			NFSLOCKMNT(nmp);
1322 			nmp->nm_state |= NFSSTA_OPENMODE;
1323 			NFSUNLOCKMNT(nmp);
1324 		}
1325 		if (error == NFSERR_STALESTATEID)
1326 			nfscl_initiate_recovery(nmp->nm_clp);
1327 		if (lckp != NULL)
1328 			nfscl_lockderef(lckp);
1329 		if (!openerr)
1330 			(void) nfsrpc_close(vp, 0, p);
1331 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1332 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1333 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1334 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1335 		} else if ((error == NFSERR_EXPIRED ||
1336 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1337 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1338 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1339 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1340 		    NFSHASNFSV4N(nmp)) {
1341 			error = EIO;
1342 		}
1343 		retrycnt++;
1344 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1345 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1346 	    error == NFSERR_BADSESSION ||
1347 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1348 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1349 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1350 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1351 	     retrycnt < 4));
1352 	if (error && retrycnt >= 4)
1353 		error = EIO;
1354 	return (error);
1355 }
1356 
1357 static int
1358 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1359     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1360     struct nfsvattr *rnap, int *attrflagp)
1361 {
1362 	u_int32_t *tl;
1363 	struct nfsrv_descript nfsd, *nd = &nfsd;
1364 	int error;
1365 	nfsattrbit_t attrbits;
1366 
1367 	*attrflagp = 0;
1368 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp, cred);
1369 	if (nd->nd_flag & ND_NFSV4)
1370 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1371 	vap->va_type = vp->v_type;
1372 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1373 	if (nd->nd_flag & ND_NFSV3) {
1374 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1375 		*tl = newnfs_false;
1376 	} else if (nd->nd_flag & ND_NFSV4) {
1377 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1378 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1379 		NFSGETATTR_ATTRBIT(&attrbits);
1380 		(void) nfsrv_putattrbit(nd, &attrbits);
1381 	}
1382 	error = nfscl_request(nd, vp, p, cred);
1383 	if (error)
1384 		return (error);
1385 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1386 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, NULL);
1387 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1388 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1389 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1390 		error = nfscl_postop_attr(nd, rnap, attrflagp);
1391 	m_freem(nd->nd_mrep);
1392 	if (nd->nd_repstat && !error)
1393 		error = nd->nd_repstat;
1394 	return (error);
1395 }
1396 
1397 /*
1398  * nfs lookup rpc
1399  */
1400 int
1401 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1402     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1403     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, uint32_t openmode)
1404 {
1405 	uint32_t deleg, rflags, *tl;
1406 	struct nfsrv_descript nfsd, *nd = &nfsd;
1407 	struct nfsmount *nmp;
1408 	struct nfsnode *np;
1409 	struct nfsfh *nfhp;
1410 	nfsattrbit_t attrbits;
1411 	int error = 0, lookupp = 0, newone, ret, retop;
1412 	uint8_t own[NFSV4CL_LOCKNAMELEN];
1413 	struct nfsclopen *op;
1414 	struct nfscldeleg *ndp;
1415 	nfsv4stateid_t stateid;
1416 
1417 	*attrflagp = 0;
1418 	*dattrflagp = 0;
1419 	if (dvp->v_type != VDIR)
1420 		return (ENOTDIR);
1421 	nmp = VFSTONFS(dvp->v_mount);
1422 	if (len > NFS_MAXNAMLEN)
1423 		return (ENAMETOOLONG);
1424 	if (NFSHASNFSV4(nmp) && len == 1 &&
1425 		name[0] == '.') {
1426 		/*
1427 		 * Just return the current dir's fh.
1428 		 */
1429 		np = VTONFS(dvp);
1430 		nfhp = malloc(sizeof (struct nfsfh) +
1431 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1432 		nfhp->nfh_len = np->n_fhp->nfh_len;
1433 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1434 		*nfhpp = nfhp;
1435 		return (0);
1436 	}
1437 	if (NFSHASNFSV4(nmp) && len == 2 &&
1438 		name[0] == '.' && name[1] == '.') {
1439 		lookupp = 1;
1440 		openmode = 0;
1441 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp, cred);
1442 	} else if (openmode != 0) {
1443 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPOPEN, dvp, cred);
1444 		nfsm_strtom(nd, name, len);
1445 	} else {
1446 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp, cred);
1447 		(void) nfsm_strtom(nd, name, len);
1448 	}
1449 	if (nd->nd_flag & ND_NFSV4) {
1450 		NFSGETATTR_ATTRBIT(&attrbits);
1451 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1452 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1453 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1454 		(void) nfsrv_putattrbit(nd, &attrbits);
1455 		if (openmode != 0) {
1456 			/* Test for a VREG file. */
1457 			NFSZERO_ATTRBIT(&attrbits);
1458 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
1459 			NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
1460 			*tl = txdr_unsigned(NFSV4OP_VERIFY);
1461 			nfsrv_putattrbit(nd, &attrbits);
1462 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1463 			*tl++ = txdr_unsigned(NFSX_UNSIGNED);
1464 			*tl = vtonfsv34_type(VREG);
1465 
1466 			/* Attempt the Open for VREG. */
1467 			nfscl_filllockowner(NULL, own, F_POSIX);
1468 			NFSM_BUILD(tl, uint32_t *, 6 * NFSX_UNSIGNED);
1469 			*tl++ = txdr_unsigned(NFSV4OP_OPEN);
1470 			*tl++ = 0;		/* seqid, ignored. */
1471 			*tl++ = txdr_unsigned(openmode);
1472 			*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
1473 			*tl++ = 0;		/* ClientID, ignored. */
1474 			*tl = 0;
1475 			nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN);
1476 			NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1477 			*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
1478 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMFH);
1479 		}
1480 	}
1481 	error = nfscl_request(nd, dvp, p, cred);
1482 	if (error)
1483 		return (error);
1484 	ndp = NULL;
1485 	if (nd->nd_repstat) {
1486 		/*
1487 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1488 		 * the lookup is at the root of an fs, so return this dir.
1489 		 */
1490 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1491 		    np = VTONFS(dvp);
1492 		    nfhp = malloc(sizeof (struct nfsfh) +
1493 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1494 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1495 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1496 		    *nfhpp = nfhp;
1497 		    m_freem(nd->nd_mrep);
1498 		    return (0);
1499 		}
1500 		if (nd->nd_flag & ND_NFSV3)
1501 		    error = nfscl_postop_attr(nd, dnap, dattrflagp);
1502 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1503 		    ND_NFSV4) {
1504 			/* Load the directory attributes. */
1505 			error = nfsm_loadattr(nd, dnap);
1506 			if (error != 0)
1507 				goto nfsmout;
1508 			*dattrflagp = 1;
1509 		}
1510 		/* Check Lookup operation reply status. */
1511 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1512 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1513 			if (*++tl != 0)
1514 				goto nfsmout;
1515 		}
1516 		/* Look for GetFH reply. */
1517 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1518 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1519 			if (*++tl != 0)
1520 				goto nfsmout;
1521 			error = nfsm_getfh(nd, nfhpp);
1522 			if (error)
1523 				goto nfsmout;
1524 		}
1525 		/* Look for Getattr reply. */
1526 		if (openmode != 0 && (nd->nd_flag & ND_NOMOREDATA) == 0) {
1527 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
1528 			if (*++tl != 0)
1529 				goto nfsmout;
1530 			error = nfsm_loadattr(nd, nap);
1531 			if (error == 0) {
1532 				/*
1533 				 * We have now successfully completed the
1534 				 * lookup, so set nd_repstat to 0.
1535 				 */
1536 				nd->nd_repstat = 0;
1537 				*attrflagp = 1;
1538 			}
1539 		}
1540 		goto nfsmout;
1541 	}
1542 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1543 		/* Load the directory attributes. */
1544 		error = nfsm_loadattr(nd, dnap);
1545 		if (error != 0)
1546 			goto nfsmout;
1547 		*dattrflagp = 1;
1548 		/* Skip over the Lookup and GetFH operation status values. */
1549 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1550 	}
1551 	error = nfsm_getfh(nd, nfhpp);
1552 	if (error)
1553 		goto nfsmout;
1554 
1555 	error = nfscl_postop_attr(nd, nap, attrflagp);
1556 	if (openmode != 0 && error == 0) {
1557 		NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID +
1558 		    10 * NFSX_UNSIGNED);
1559 		tl += 4;	/* Skip over Verify+Open status. */
1560 		stateid.seqid = *tl++;
1561 		stateid.other[0] = *tl++;
1562 		stateid.other[1] = *tl++;
1563 		stateid.other[2] = *tl;
1564 		rflags = fxdr_unsigned(uint32_t, *(tl + 6));
1565 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1566 		if (error != 0)
1567 			goto nfsmout;
1568 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
1569 		deleg = fxdr_unsigned(uint32_t, *tl);
1570 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
1571 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
1572 			/*
1573 			 * Just need to fill in the fields used by
1574 			 * nfscl_trydelegreturn().
1575 			 * Mark the mount point as acquiring
1576 			 * delegations, so NFSPROC_LOOKUPOPEN will
1577 			 * no longer be done.
1578 			 */
1579 			NFSLOCKMNT(nmp);
1580 			nmp->nm_privflag |= NFSMNTP_DELEGISSUED;
1581 			NFSUNLOCKMNT(nmp);
1582 			ndp = malloc(sizeof(struct nfscldeleg) +
1583 			    (*nfhpp)->nfh_len, M_NFSCLDELEG, M_WAITOK);
1584 			ndp->nfsdl_fhlen = (*nfhpp)->nfh_len;
1585 			NFSBCOPY((*nfhpp)->nfh_fh, ndp->nfsdl_fh,
1586 			    ndp->nfsdl_fhlen);
1587 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
1588 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
1589 			ndp->nfsdl_stateid.seqid = *tl++;
1590 			ndp->nfsdl_stateid.other[0] = *tl++;
1591 			ndp->nfsdl_stateid.other[1] = *tl++;
1592 			ndp->nfsdl_stateid.other[2] = *tl++;
1593 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
1594 			error = NFSERR_BADXDR;
1595 			goto nfsmout;
1596 		}
1597 		ret = nfscl_open(dvp, (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len,
1598 		    openmode, 0, cred, p, NULL, &op, &newone, &retop, 1, true);
1599 		if (ret != 0)
1600 			goto nfsmout;
1601 		if (newone != 0) {
1602 			op->nfso_stateid.seqid = stateid.seqid;
1603 			op->nfso_stateid.other[0] = stateid.other[0];
1604 			op->nfso_stateid.other[1] = stateid.other[1];
1605 			op->nfso_stateid.other[2] = stateid.other[2];
1606 			op->nfso_mode = openmode;
1607 		} else {
1608 			op->nfso_stateid.seqid = stateid.seqid;
1609 			if (retop == NFSCLOPEN_DOOPEN)
1610 				op->nfso_mode |= openmode;
1611 		}
1612 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
1613 		    nfscl_assumeposixlocks)
1614 			op->nfso_posixlock = 1;
1615 		else
1616 			op->nfso_posixlock = 0;
1617 		nfscl_openrelease(nmp, op, 0, 0);
1618 		if (ndp != NULL) {
1619 			/*
1620 			 * Since we do not have the vnode, we
1621 			 * cannot invalidate cached attributes.
1622 			 * Just return the delegation.
1623 			 */
1624 			nfscl_trydelegreturn(ndp, cred, nmp, p);
1625 		}
1626 	}
1627 	if ((nd->nd_flag & ND_NFSV3) && !error)
1628 		error = nfscl_postop_attr(nd, dnap, dattrflagp);
1629 nfsmout:
1630 	m_freem(nd->nd_mrep);
1631 	if (!error && nd->nd_repstat)
1632 		error = nd->nd_repstat;
1633 	free(ndp, M_NFSCLDELEG);
1634 	return (error);
1635 }
1636 
1637 /*
1638  * Do a readlink rpc.
1639  */
1640 int
1641 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1642     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1643 {
1644 	u_int32_t *tl;
1645 	struct nfsrv_descript nfsd, *nd = &nfsd;
1646 	struct nfsnode *np = VTONFS(vp);
1647 	nfsattrbit_t attrbits;
1648 	int error, len, cangetattr = 1;
1649 
1650 	*attrflagp = 0;
1651 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp, cred);
1652 	if (nd->nd_flag & ND_NFSV4) {
1653 		/*
1654 		 * And do a Getattr op.
1655 		 */
1656 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1657 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1658 		NFSGETATTR_ATTRBIT(&attrbits);
1659 		(void) nfsrv_putattrbit(nd, &attrbits);
1660 	}
1661 	error = nfscl_request(nd, vp, p, cred);
1662 	if (error)
1663 		return (error);
1664 	if (nd->nd_flag & ND_NFSV3)
1665 		error = nfscl_postop_attr(nd, nap, attrflagp);
1666 	if (!nd->nd_repstat && !error) {
1667 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1668 		/*
1669 		 * This seems weird to me, but must have been added to
1670 		 * FreeBSD for some reason. The only thing I can think of
1671 		 * is that there was/is some server that replies with
1672 		 * more link data than it should?
1673 		 */
1674 		if (len == NFS_MAXPATHLEN) {
1675 			NFSLOCKNODE(np);
1676 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1677 				len = np->n_size;
1678 				cangetattr = 0;
1679 			}
1680 			NFSUNLOCKNODE(np);
1681 		}
1682 		error = nfsm_mbufuio(nd, uiop, len);
1683 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1684 			error = nfscl_postop_attr(nd, nap, attrflagp);
1685 	}
1686 	if (nd->nd_repstat && !error)
1687 		error = nd->nd_repstat;
1688 nfsmout:
1689 	m_freem(nd->nd_mrep);
1690 	return (error);
1691 }
1692 
1693 /*
1694  * Read operation.
1695  */
1696 int
1697 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1698     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
1699 {
1700 	int error, expireret = 0, retrycnt;
1701 	u_int32_t clidrev = 0;
1702 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1703 	struct nfsnode *np = VTONFS(vp);
1704 	struct ucred *newcred;
1705 	struct nfsfh *nfhp = NULL;
1706 	nfsv4stateid_t stateid;
1707 	void *lckp;
1708 
1709 	if (nmp->nm_clp != NULL)
1710 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1711 	newcred = cred;
1712 	if (NFSHASNFSV4(nmp)) {
1713 		nfhp = np->n_fhp;
1714 		newcred = NFSNEWCRED(cred);
1715 	}
1716 	retrycnt = 0;
1717 	do {
1718 		lckp = NULL;
1719 		if (NFSHASNFSV4(nmp))
1720 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1721 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1722 			    &lckp);
1723 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1724 		    attrflagp);
1725 		if (error == NFSERR_OPENMODE) {
1726 			NFSLOCKMNT(nmp);
1727 			nmp->nm_state |= NFSSTA_OPENMODE;
1728 			NFSUNLOCKMNT(nmp);
1729 		}
1730 		if (error == NFSERR_STALESTATEID)
1731 			nfscl_initiate_recovery(nmp->nm_clp);
1732 		if (lckp != NULL)
1733 			nfscl_lockderef(lckp);
1734 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1735 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1736 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1737 			(void) nfs_catnap(PZERO, error, "nfs_read");
1738 		} else if ((error == NFSERR_EXPIRED ||
1739 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1740 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1741 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1742 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1743 		    NFSHASNFSV4N(nmp)) {
1744 			error = EIO;
1745 		}
1746 		retrycnt++;
1747 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1748 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1749 	    error == NFSERR_BADSESSION ||
1750 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1751 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1752 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1753 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1754 	if (error && retrycnt >= 4)
1755 		error = EIO;
1756 	if (NFSHASNFSV4(nmp))
1757 		NFSFREECRED(newcred);
1758 	return (error);
1759 }
1760 
1761 /*
1762  * The actual read RPC.
1763  */
1764 static int
1765 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1766     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1767     int *attrflagp)
1768 {
1769 	u_int32_t *tl;
1770 	int error = 0, len, retlen, tsiz, eof = 0;
1771 	struct nfsrv_descript nfsd;
1772 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1773 	struct nfsrv_descript *nd = &nfsd;
1774 	int rsize;
1775 	off_t tmp_off;
1776 
1777 	*attrflagp = 0;
1778 	tsiz = uiop->uio_resid;
1779 	tmp_off = uiop->uio_offset + tsiz;
1780 	NFSLOCKMNT(nmp);
1781 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1782 		NFSUNLOCKMNT(nmp);
1783 		return (EFBIG);
1784 	}
1785 	rsize = nmp->nm_rsize;
1786 	NFSUNLOCKMNT(nmp);
1787 	nd->nd_mrep = NULL;
1788 	while (tsiz > 0) {
1789 		*attrflagp = 0;
1790 		len = (tsiz > rsize) ? rsize : tsiz;
1791 		NFSCL_REQSTART(nd, NFSPROC_READ, vp, cred);
1792 		if (nd->nd_flag & ND_NFSV4)
1793 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1794 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1795 		if (nd->nd_flag & ND_NFSV2) {
1796 			*tl++ = txdr_unsigned(uiop->uio_offset);
1797 			*tl++ = txdr_unsigned(len);
1798 			*tl = 0;
1799 		} else {
1800 			txdr_hyper(uiop->uio_offset, tl);
1801 			*(tl + 2) = txdr_unsigned(len);
1802 		}
1803 		/*
1804 		 * Since I can't do a Getattr for NFSv4 for Write, there
1805 		 * doesn't seem any point in doing one here, either.
1806 		 * (See the comment in nfsrpc_writerpc() for more info.)
1807 		 */
1808 		error = nfscl_request(nd, vp, p, cred);
1809 		if (error)
1810 			return (error);
1811 		if (nd->nd_flag & ND_NFSV3) {
1812 			error = nfscl_postop_attr(nd, nap, attrflagp);
1813 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1814 			error = nfsm_loadattr(nd, nap);
1815 			if (!error)
1816 				*attrflagp = 1;
1817 		}
1818 		if (nd->nd_repstat || error) {
1819 			if (!error)
1820 				error = nd->nd_repstat;
1821 			goto nfsmout;
1822 		}
1823 		if (nd->nd_flag & ND_NFSV3) {
1824 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1825 			eof = fxdr_unsigned(int, *(tl + 1));
1826 		} else if (nd->nd_flag & ND_NFSV4) {
1827 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1828 			eof = fxdr_unsigned(int, *tl);
1829 		}
1830 		NFSM_STRSIZ(retlen, len);
1831 		error = nfsm_mbufuio(nd, uiop, retlen);
1832 		if (error)
1833 			goto nfsmout;
1834 		m_freem(nd->nd_mrep);
1835 		nd->nd_mrep = NULL;
1836 		tsiz -= retlen;
1837 		if (!(nd->nd_flag & ND_NFSV2)) {
1838 			if (eof || retlen == 0)
1839 				tsiz = 0;
1840 		} else if (retlen < len)
1841 			tsiz = 0;
1842 	}
1843 	return (0);
1844 nfsmout:
1845 	if (nd->nd_mrep != NULL)
1846 		m_freem(nd->nd_mrep);
1847 	return (error);
1848 }
1849 
1850 /*
1851  * nfs write operation
1852  * When called_from_strategy != 0, it should return EIO for an error that
1853  * indicates recovery is in progress, so that the buffer will be left
1854  * dirty and be written back to the server later. If it loops around,
1855  * the recovery thread could get stuck waiting for the buffer and recovery
1856  * will then deadlock.
1857  */
1858 int
1859 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1860     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1861     int called_from_strategy, int ioflag)
1862 {
1863 	int error, expireret = 0, retrycnt, nostateid;
1864 	u_int32_t clidrev = 0;
1865 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1866 	struct nfsnode *np = VTONFS(vp);
1867 	struct ucred *newcred;
1868 	struct nfsfh *nfhp = NULL;
1869 	nfsv4stateid_t stateid;
1870 	void *lckp;
1871 
1872 	KASSERT(*must_commit >= 0 && *must_commit <= 2,
1873 	    ("nfsrpc_write: must_commit out of range=%d", *must_commit));
1874 	if (nmp->nm_clp != NULL)
1875 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1876 	newcred = cred;
1877 	if (NFSHASNFSV4(nmp)) {
1878 		newcred = NFSNEWCRED(cred);
1879 		nfhp = np->n_fhp;
1880 	}
1881 	retrycnt = 0;
1882 	do {
1883 		lckp = NULL;
1884 		nostateid = 0;
1885 		if (NFSHASNFSV4(nmp)) {
1886 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1887 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1888 			    &lckp);
1889 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1890 			    stateid.other[2] == 0) {
1891 				nostateid = 1;
1892 				NFSCL_DEBUG(1, "stateid0 in write\n");
1893 			}
1894 		}
1895 
1896 		/*
1897 		 * If there is no stateid for NFSv4, it means this is an
1898 		 * extraneous write after close. Basically a poorly
1899 		 * implemented buffer cache. Just don't do the write.
1900 		 */
1901 		if (nostateid)
1902 			error = 0;
1903 		else
1904 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1905 			    newcred, &stateid, p, nap, attrflagp, ioflag);
1906 		if (error == NFSERR_STALESTATEID)
1907 			nfscl_initiate_recovery(nmp->nm_clp);
1908 		if (lckp != NULL)
1909 			nfscl_lockderef(lckp);
1910 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1911 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1912 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1913 			(void) nfs_catnap(PZERO, error, "nfs_write");
1914 		} else if ((error == NFSERR_EXPIRED ||
1915 		    ((!NFSHASINT(nmp) || !NFSHASNFSV4N(nmp)) &&
1916 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
1917 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1918 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp) &&
1919 		    NFSHASNFSV4N(nmp)) {
1920 			error = EIO;
1921 		}
1922 		retrycnt++;
1923 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1924 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1925 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1926 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1927 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1928 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1929 	if (error != 0 && (retrycnt >= 4 ||
1930 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1931 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1932 		error = EIO;
1933 	if (NFSHASNFSV4(nmp))
1934 		NFSFREECRED(newcred);
1935 	return (error);
1936 }
1937 
1938 /*
1939  * The actual write RPC.
1940  */
1941 static int
1942 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1943     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1944     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, int ioflag)
1945 {
1946 	u_int32_t *tl;
1947 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1948 	struct nfsnode *np = VTONFS(vp);
1949 	int error = 0, len, rlen, commit, committed = NFSWRITE_FILESYNC;
1950 	int wccflag = 0;
1951 	int32_t backup;
1952 	struct nfsrv_descript *nd;
1953 	nfsattrbit_t attrbits;
1954 	uint64_t tmp_off;
1955 	ssize_t tsiz, wsize;
1956 	bool do_append;
1957 
1958 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1959 	*attrflagp = 0;
1960 	tsiz = uiop->uio_resid;
1961 	tmp_off = uiop->uio_offset + tsiz;
1962 	NFSLOCKMNT(nmp);
1963 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1964 		NFSUNLOCKMNT(nmp);
1965 		return (EFBIG);
1966 	}
1967 	wsize = nmp->nm_wsize;
1968 	do_append = false;
1969 	if ((ioflag & IO_APPEND) != 0 && NFSHASNFSV4(nmp) && !NFSHASPNFS(nmp))
1970 		do_append = true;
1971 	NFSUNLOCKMNT(nmp);
1972 	nd = malloc(sizeof(*nd), M_TEMP, M_WAITOK);
1973 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1974 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1975 	while (tsiz > 0) {
1976 		*attrflagp = 0;
1977 		len = (tsiz > wsize) ? wsize : tsiz;
1978 		if (do_append)
1979 			NFSCL_REQSTART(nd, NFSPROC_APPENDWRITE, vp, cred);
1980 		else
1981 			NFSCL_REQSTART(nd, NFSPROC_WRITE, vp, cred);
1982 		if (nd->nd_flag & ND_NFSV4) {
1983 			if (do_append) {
1984 				NFSZERO_ATTRBIT(&attrbits);
1985 				NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_SIZE);
1986 				nfsrv_putattrbit(nd, &attrbits);
1987 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED +
1988 				    NFSX_HYPER);
1989 				*tl++ = txdr_unsigned(NFSX_HYPER);
1990 				txdr_hyper(uiop->uio_offset, tl); tl += 2;
1991 				*tl = txdr_unsigned(NFSV4OP_WRITE);
1992 			}
1993 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1994 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1995 			txdr_hyper(uiop->uio_offset, tl);
1996 			tl += 2;
1997 			*tl++ = txdr_unsigned(*iomode);
1998 			*tl = txdr_unsigned(len);
1999 		} else if (nd->nd_flag & ND_NFSV3) {
2000 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
2001 			txdr_hyper(uiop->uio_offset, tl);
2002 			tl += 2;
2003 			*tl++ = txdr_unsigned(len);
2004 			*tl++ = txdr_unsigned(*iomode);
2005 			*tl = txdr_unsigned(len);
2006 		} else {
2007 			u_int32_t x;
2008 
2009 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2010 			/*
2011 			 * Not sure why someone changed this, since the
2012 			 * RFC clearly states that "beginoffset" and
2013 			 * "totalcount" are ignored, but it wouldn't
2014 			 * surprise me if there's a busted server out there.
2015 			 */
2016 			/* Set both "begin" and "current" to non-garbage. */
2017 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
2018 			*tl++ = x;      /* "begin offset" */
2019 			*tl++ = x;      /* "current offset" */
2020 			x = txdr_unsigned(len);
2021 			*tl++ = x;      /* total to this offset */
2022 			*tl = x;        /* size of this write */
2023 		}
2024 		nfsm_uiombuf(nd, uiop, len);
2025 		/*
2026 		 * Although it is tempting to do a normal Getattr Op in the
2027 		 * NFSv4 compound, the result can be a nearly hung client
2028 		 * system if the Getattr asks for Owner and/or OwnerGroup.
2029 		 * It occurs when the client can't map either the Owner or
2030 		 * Owner_group name in the Getattr reply to a uid/gid. When
2031 		 * there is a cache miss, the kernel does an upcall to the
2032 		 * nfsuserd. Then, it can try and read the local /etc/passwd
2033 		 * or /etc/group file. It can then block in getnewbuf(),
2034 		 * waiting for dirty writes to be pushed to the NFS server.
2035 		 * The only reason this doesn't result in a complete
2036 		 * deadlock, is that the upcall times out and allows
2037 		 * the write to complete. However, progress is so slow
2038 		 * that it might just as well be deadlocked.
2039 		 * As such, we get the rest of the attributes, but not
2040 		 * Owner or Owner_group.
2041 		 * nb: nfscl_loadattrcache() needs to be told that these
2042 		 *     partial attributes from a write rpc are being
2043 		 *     passed in, via a argument flag.
2044 		 */
2045 		if (nd->nd_flag & ND_NFSV4) {
2046 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
2047 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2048 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2049 			(void) nfsrv_putattrbit(nd, &attrbits);
2050 		}
2051 		error = nfscl_request(nd, vp, p, cred);
2052 		if (error) {
2053 			free(nd, M_TEMP);
2054 			return (error);
2055 		}
2056 		if (nd->nd_repstat) {
2057 			/*
2058 			 * In case the rpc gets retried, roll
2059 			 * the uio fields changed by nfsm_uiombuf()
2060 			 * back.
2061 			 */
2062 			uiop->uio_offset -= len;
2063 			uiop->uio_resid += len;
2064 			uiop->uio_iov->iov_base =
2065 			    (char *)uiop->uio_iov->iov_base - len;
2066 			uiop->uio_iov->iov_len += len;
2067 		}
2068 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2069 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
2070 			    &wccflag, &tmp_off);
2071 			if (error)
2072 				goto nfsmout;
2073 		}
2074 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2075 		    (ND_NFSV4 | ND_NOMOREDATA) &&
2076 		    nd->nd_repstat == NFSERR_NOTSAME && do_append) {
2077 			/*
2078 			 * Verify of the file's size failed, so redo the
2079 			 * write using the file's size as returned in
2080 			 * the wcc attributes.
2081 			 */
2082 			if (tmp_off + tsiz <= nmp->nm_maxfilesize) {
2083 				do_append = false;
2084 				uiop->uio_offset = tmp_off;
2085 				m_freem(nd->nd_mrep);
2086 				nd->nd_mrep = NULL;
2087 				continue;
2088 			} else
2089 				nd->nd_repstat = EFBIG;
2090 		}
2091 		if (!nd->nd_repstat) {
2092 			if (do_append) {
2093 				/* Strip off the Write reply status. */
2094 				do_append = false;
2095 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2096 			}
2097 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2098 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
2099 					+ NFSX_VERF);
2100 				rlen = fxdr_unsigned(int, *tl++);
2101 				if (rlen == 0) {
2102 					error = NFSERR_IO;
2103 					goto nfsmout;
2104 				} else if (rlen < len) {
2105 					backup = len - rlen;
2106 					uiop->uio_iov->iov_base =
2107 					    (char *)uiop->uio_iov->iov_base -
2108 					    backup;
2109 					uiop->uio_iov->iov_len += backup;
2110 					uiop->uio_offset -= backup;
2111 					uiop->uio_resid += backup;
2112 					len = rlen;
2113 				}
2114 				commit = fxdr_unsigned(int, *tl++);
2115 
2116 				/*
2117 				 * Return the lowest commitment level
2118 				 * obtained by any of the RPCs.
2119 				 */
2120 				if (committed == NFSWRITE_FILESYNC)
2121 					committed = commit;
2122 				else if (committed == NFSWRITE_DATASYNC &&
2123 					commit == NFSWRITE_UNSTABLE)
2124 					committed = commit;
2125 				NFSLOCKMNT(nmp);
2126 				if (!NFSHASWRITEVERF(nmp)) {
2127 					NFSBCOPY((caddr_t)tl,
2128 					    (caddr_t)&nmp->nm_verf[0],
2129 					    NFSX_VERF);
2130 					NFSSETWRITEVERF(nmp);
2131 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
2132 				    NFSX_VERF) && *must_commit != 2) {
2133 					*must_commit = 1;
2134 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
2135 				}
2136 				NFSUNLOCKMNT(nmp);
2137 			}
2138 			if (nd->nd_flag & ND_NFSV4)
2139 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2140 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
2141 				error = nfsm_loadattr(nd, nap);
2142 				if (!error)
2143 					*attrflagp = NFS_LATTR_NOSHRINK;
2144 			}
2145 		} else {
2146 			error = nd->nd_repstat;
2147 		}
2148 		if (error)
2149 			goto nfsmout;
2150 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
2151 		m_freem(nd->nd_mrep);
2152 		nd->nd_mrep = NULL;
2153 		tsiz -= len;
2154 	}
2155 nfsmout:
2156 	if (nd->nd_mrep != NULL)
2157 		m_freem(nd->nd_mrep);
2158 	*iomode = committed;
2159 	if (nd->nd_repstat && !error)
2160 		error = nd->nd_repstat;
2161 	free(nd, M_TEMP);
2162 	return (error);
2163 }
2164 
2165 /*
2166  * Do an nfs deallocate operation.
2167  */
2168 int
2169 nfsrpc_deallocate(vnode_t vp, off_t offs, off_t len, struct nfsvattr *nap,
2170     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
2171 {
2172 	int error, expireret = 0, openerr, retrycnt;
2173 	uint32_t clidrev = 0;
2174 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2175 	struct nfsfh *nfhp;
2176 	nfsv4stateid_t stateid;
2177 	void *lckp;
2178 
2179 	if (nmp->nm_clp != NULL)
2180 		clidrev = nmp->nm_clp->nfsc_clientidrev;
2181 	retrycnt = 0;
2182 	do {
2183 		lckp = NULL;
2184 		openerr = 1;
2185 		nfhp = VTONFS(vp)->n_fhp;
2186 		error = nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
2187 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
2188 		if (error != 0) {
2189 			/*
2190 			 * No Open stateid, so try and open the file
2191 			 * now.
2192 			 */
2193 			openerr = nfsrpc_open(vp, FWRITE, cred, p);
2194 			if (openerr == 0)
2195 				nfscl_getstateid(vp, nfhp->nfh_fh,
2196 				    nfhp->nfh_len, NFSV4OPEN_ACCESSWRITE, 0,
2197 				    cred, p, &stateid, &lckp);
2198 		}
2199 		error = nfsrpc_deallocaterpc(vp, offs, len, &stateid, nap,
2200 		    attrflagp, cred, p);
2201 		if (error == NFSERR_STALESTATEID)
2202 			nfscl_initiate_recovery(nmp->nm_clp);
2203 		if (lckp != NULL)
2204 			nfscl_lockderef(lckp);
2205 		if (openerr == 0)
2206 			nfsrpc_close(vp, 0, p);
2207 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2208 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2209 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
2210 			(void) nfs_catnap(PZERO, error, "nfs_deallocate");
2211 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
2212 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
2213 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2214 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
2215 			error = EIO;
2216 		}
2217 		retrycnt++;
2218 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
2219 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2220 	    error == NFSERR_BADSESSION ||
2221 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
2222 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2223 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
2224 	if (error && retrycnt >= 4)
2225 		error = EIO;
2226 	return (error);
2227 }
2228 
2229 /*
2230  * The actual deallocate RPC.
2231  */
2232 static int
2233 nfsrpc_deallocaterpc(vnode_t vp, off_t offs, off_t len,
2234     nfsv4stateid_t *stateidp, struct nfsvattr *nap, int *attrflagp,
2235     struct ucred *cred, NFSPROC_T *p)
2236 {
2237 	uint32_t *tl;
2238 	struct nfsnode *np = VTONFS(vp);
2239 	int error, wccflag;
2240 	struct nfsrv_descript nfsd;
2241 	struct nfsrv_descript *nd = &nfsd;
2242 	nfsattrbit_t attrbits;
2243 
2244 	*attrflagp = 0;
2245 	NFSCL_REQSTART(nd, NFSPROC_DEALLOCATE, vp, cred);
2246 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
2247 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
2248 	txdr_hyper(offs, tl);
2249 	tl += 2;
2250 	txdr_hyper(len, tl);
2251 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
2252 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
2253 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2254 	nfsrv_putattrbit(nd, &attrbits);
2255 	error = nfscl_request(nd, vp, p, cred);
2256 	if (error != 0)
2257 		return (error);
2258 	wccflag = 0;
2259 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, &wccflag, NULL);
2260 	if (error != 0)
2261 		goto nfsmout;
2262 	if (nd->nd_repstat == 0) {
2263 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
2264 		error = nfsm_loadattr(nd, nap);
2265 		if (error != 0)
2266 			goto nfsmout;
2267 		*attrflagp = NFS_LATTR_NOSHRINK;
2268 	}
2269 	NFSWRITERPC_SETTIME(wccflag, np, nap, 1);
2270 nfsmout:
2271 	m_freem(nd->nd_mrep);
2272 	if (nd->nd_repstat != 0 && error == 0)
2273 		error = nd->nd_repstat;
2274 	return (error);
2275 }
2276 
2277 /*
2278  * nfs mknod rpc
2279  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
2280  * mode set to specify the file type and the size field for rdev.
2281  */
2282 int
2283 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2284     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
2285     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2286     int *attrflagp, int *dattrflagp)
2287 {
2288 	u_int32_t *tl;
2289 	int error = 0;
2290 	struct nfsrv_descript nfsd, *nd = &nfsd;
2291 	nfsattrbit_t attrbits;
2292 
2293 	*nfhpp = NULL;
2294 	*attrflagp = 0;
2295 	*dattrflagp = 0;
2296 	if (namelen > NFS_MAXNAMLEN)
2297 		return (ENAMETOOLONG);
2298 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp, cred);
2299 	if (nd->nd_flag & ND_NFSV4) {
2300 		if (vtyp == VBLK || vtyp == VCHR) {
2301 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
2302 			*tl++ = vtonfsv34_type(vtyp);
2303 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2304 			*tl = txdr_unsigned(NFSMINOR(rdev));
2305 		} else {
2306 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2307 			*tl = vtonfsv34_type(vtyp);
2308 		}
2309 	}
2310 	(void) nfsm_strtom(nd, name, namelen);
2311 	if (nd->nd_flag & ND_NFSV3) {
2312 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2313 		*tl = vtonfsv34_type(vtyp);
2314 	}
2315 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2316 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2317 	if ((nd->nd_flag & ND_NFSV3) &&
2318 	    (vtyp == VCHR || vtyp == VBLK)) {
2319 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2320 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2321 		*tl = txdr_unsigned(NFSMINOR(rdev));
2322 	}
2323 	if (nd->nd_flag & ND_NFSV4) {
2324 		NFSGETATTR_ATTRBIT(&attrbits);
2325 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2326 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2327 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2328 		(void) nfsrv_putattrbit(nd, &attrbits);
2329 	}
2330 	if (nd->nd_flag & ND_NFSV2)
2331 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2332 	error = nfscl_request(nd, dvp, p, cred);
2333 	if (error)
2334 		return (error);
2335 	if (nd->nd_flag & ND_NFSV4)
2336 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2337 	if (!nd->nd_repstat) {
2338 		if (nd->nd_flag & ND_NFSV4) {
2339 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2340 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2341 			if (error)
2342 				goto nfsmout;
2343 		}
2344 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2345 		if (error)
2346 			goto nfsmout;
2347 	}
2348 	if (nd->nd_flag & ND_NFSV3)
2349 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2350 	if (!error && nd->nd_repstat)
2351 		error = nd->nd_repstat;
2352 nfsmout:
2353 	m_freem(nd->nd_mrep);
2354 	return (error);
2355 }
2356 
2357 /*
2358  * nfs file create call
2359  * Mostly just call the approriate routine. (I separated out v4, so that
2360  * error recovery wouldn't be as difficult.)
2361  */
2362 int
2363 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2364     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2365     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2366     int *attrflagp, int *dattrflagp)
2367 {
2368 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2369 	struct nfsclowner *owp;
2370 	struct nfscldeleg *dp;
2371 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2372 	u_int32_t clidrev;
2373 
2374 	if (NFSHASNFSV4(nmp)) {
2375 	    retrycnt = 0;
2376 	    do {
2377 		dp = NULL;
2378 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2379 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2380 		    NULL, 1, true);
2381 		if (error)
2382 			return (error);
2383 		if (nmp->nm_clp != NULL)
2384 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2385 		else
2386 			clidrev = 0;
2387 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2388 		    nfs_numnfscbd == 0 || retrycnt > 0)
2389 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2390 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2391 			  attrflagp, dattrflagp, &unlocked);
2392 		else
2393 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2394 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2395 			  attrflagp, dattrflagp, &unlocked);
2396 		/*
2397 		 * There is no need to invalidate cached attributes here,
2398 		 * since new post-delegation issue attributes are always
2399 		 * returned by nfsrpc_createv4() and these will update the
2400 		 * attribute cache.
2401 		 */
2402 		if (dp != NULL)
2403 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2404 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2405 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2406 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2407 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2408 		    error == NFSERR_BADSESSION) {
2409 			(void) nfs_catnap(PZERO, error, "nfs_open");
2410 		} else if ((error == NFSERR_EXPIRED ||
2411 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2412 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2413 			retrycnt++;
2414 		}
2415 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2416 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2417 		error == NFSERR_BADSESSION ||
2418 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2419 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2420 	    if (error && retrycnt >= 4)
2421 		    error = EIO;
2422 	} else {
2423 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2424 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp);
2425 	}
2426 	return (error);
2427 }
2428 
2429 /*
2430  * The create rpc for v2 and 3.
2431  */
2432 static int
2433 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2434     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2435     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2436     int *attrflagp, int *dattrflagp)
2437 {
2438 	u_int32_t *tl;
2439 	int error = 0;
2440 	struct nfsrv_descript nfsd, *nd = &nfsd;
2441 
2442 	*nfhpp = NULL;
2443 	*attrflagp = 0;
2444 	*dattrflagp = 0;
2445 	if (namelen > NFS_MAXNAMLEN)
2446 		return (ENAMETOOLONG);
2447 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp, cred);
2448 	(void) nfsm_strtom(nd, name, namelen);
2449 	if (nd->nd_flag & ND_NFSV3) {
2450 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2451 		if (fmode & O_EXCL) {
2452 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2453 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2454 			*tl++ = cverf.lval[0];
2455 			*tl = cverf.lval[1];
2456 		} else {
2457 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2458 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2459 		}
2460 	} else {
2461 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2462 	}
2463 	error = nfscl_request(nd, dvp, p, cred);
2464 	if (error)
2465 		return (error);
2466 	if (nd->nd_repstat == 0) {
2467 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2468 		if (error)
2469 			goto nfsmout;
2470 	}
2471 	if (nd->nd_flag & ND_NFSV3)
2472 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2473 	if (nd->nd_repstat != 0 && error == 0)
2474 		error = nd->nd_repstat;
2475 nfsmout:
2476 	m_freem(nd->nd_mrep);
2477 	return (error);
2478 }
2479 
2480 static int
2481 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2482     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2483     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2484     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2485     int *dattrflagp, int *unlockedp)
2486 {
2487 	u_int32_t *tl;
2488 	int error = 0, deleg, newone, ret, acesize, limitby;
2489 	struct nfsrv_descript nfsd, *nd = &nfsd;
2490 	struct nfsclopen *op;
2491 	struct nfscldeleg *dp = NULL;
2492 	struct nfsnode *np;
2493 	struct nfsfh *nfhp;
2494 	nfsattrbit_t attrbits;
2495 	nfsv4stateid_t stateid;
2496 	u_int32_t rflags;
2497 	struct nfsmount *nmp;
2498 	struct nfsclsession *tsep;
2499 
2500 	nmp = VFSTONFS(dvp->v_mount);
2501 	np = VTONFS(dvp);
2502 	*unlockedp = 0;
2503 	*nfhpp = NULL;
2504 	*dpp = NULL;
2505 	*attrflagp = 0;
2506 	*dattrflagp = 0;
2507 	if (namelen > NFS_MAXNAMLEN)
2508 		return (ENAMETOOLONG);
2509 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp, cred);
2510 	/*
2511 	 * For V4, this is actually an Open op.
2512 	 */
2513 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2514 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2515 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2516 	    NFSV4OPEN_ACCESSREAD);
2517 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2518 	tsep = nfsmnt_mdssession(nmp);
2519 	*tl++ = tsep->nfsess_clientid.lval[0];
2520 	*tl = tsep->nfsess_clientid.lval[1];
2521 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2522 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2523 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2524 	if (fmode & O_EXCL) {
2525 		if (NFSHASNFSV4N(nmp)) {
2526 			if (NFSHASSESSPERSIST(nmp)) {
2527 				/* Use GUARDED for persistent sessions. */
2528 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2529 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2530 			} else {
2531 				/* Otherwise, use EXCLUSIVE4_1. */
2532 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2533 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2534 				*tl++ = cverf.lval[0];
2535 				*tl = cverf.lval[1];
2536 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2537 			}
2538 		} else {
2539 			/* NFSv4.0 */
2540 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2541 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2542 			*tl++ = cverf.lval[0];
2543 			*tl = cverf.lval[1];
2544 		}
2545 	} else {
2546 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2547 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2548 	}
2549 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2550 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2551 	(void) nfsm_strtom(nd, name, namelen);
2552 	/* Get the new file's handle and attributes. */
2553 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2554 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2555 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2556 	NFSGETATTR_ATTRBIT(&attrbits);
2557 	(void) nfsrv_putattrbit(nd, &attrbits);
2558 	/* Get the directory's post-op attributes. */
2559 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2560 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2561 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2562 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2563 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2564 	(void) nfsrv_putattrbit(nd, &attrbits);
2565 	error = nfscl_request(nd, dvp, p, cred);
2566 	if (error)
2567 		return (error);
2568 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2569 	if (nd->nd_repstat == 0) {
2570 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2571 		    6 * NFSX_UNSIGNED);
2572 		stateid.seqid = *tl++;
2573 		stateid.other[0] = *tl++;
2574 		stateid.other[1] = *tl++;
2575 		stateid.other[2] = *tl;
2576 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2577 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2578 		if (error)
2579 			goto nfsmout;
2580 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2581 		deleg = fxdr_unsigned(int, *tl);
2582 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2583 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2584 			if (!(owp->nfsow_clp->nfsc_flags &
2585 			      NFSCLFLAGS_FIRSTDELEG))
2586 				owp->nfsow_clp->nfsc_flags |=
2587 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2588 			dp = malloc(
2589 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2590 			    M_NFSCLDELEG, M_WAITOK);
2591 			LIST_INIT(&dp->nfsdl_owner);
2592 			LIST_INIT(&dp->nfsdl_lock);
2593 			dp->nfsdl_clp = owp->nfsow_clp;
2594 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2595 			nfscl_lockinit(&dp->nfsdl_rwlock);
2596 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2597 			    NFSX_UNSIGNED);
2598 			dp->nfsdl_stateid.seqid = *tl++;
2599 			dp->nfsdl_stateid.other[0] = *tl++;
2600 			dp->nfsdl_stateid.other[1] = *tl++;
2601 			dp->nfsdl_stateid.other[2] = *tl++;
2602 			ret = fxdr_unsigned(int, *tl);
2603 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2604 				dp->nfsdl_flags = NFSCLDL_WRITE;
2605 				/*
2606 				 * Indicates how much the file can grow.
2607 				 */
2608 				NFSM_DISSECT(tl, u_int32_t *,
2609 				    3 * NFSX_UNSIGNED);
2610 				limitby = fxdr_unsigned(int, *tl++);
2611 				switch (limitby) {
2612 				case NFSV4OPEN_LIMITSIZE:
2613 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2614 					break;
2615 				case NFSV4OPEN_LIMITBLOCKS:
2616 					dp->nfsdl_sizelimit =
2617 					    fxdr_unsigned(u_int64_t, *tl++);
2618 					dp->nfsdl_sizelimit *=
2619 					    fxdr_unsigned(u_int64_t, *tl);
2620 					break;
2621 				default:
2622 					error = NFSERR_BADXDR;
2623 					goto nfsmout;
2624 				}
2625 			} else {
2626 				dp->nfsdl_flags = NFSCLDL_READ;
2627 			}
2628 			if (ret)
2629 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2630 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
2631 			    &ret, &acesize, p);
2632 			if (error)
2633 				goto nfsmout;
2634 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2635 			error = NFSERR_BADXDR;
2636 			goto nfsmout;
2637 		}
2638 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2639 		if (error)
2640 			goto nfsmout;
2641 		/* Get rid of the PutFH and Getattr status values. */
2642 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2643 		/* Load the directory attributes. */
2644 		error = nfsm_loadattr(nd, dnap);
2645 		if (error)
2646 			goto nfsmout;
2647 		*dattrflagp = 1;
2648 		if (dp != NULL && *attrflagp) {
2649 			dp->nfsdl_change = nnap->na_filerev;
2650 			dp->nfsdl_modtime = nnap->na_mtime;
2651 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2652 		}
2653 		/*
2654 		 * We can now complete the Open state.
2655 		 */
2656 		nfhp = *nfhpp;
2657 		if (dp != NULL) {
2658 			dp->nfsdl_fhlen = nfhp->nfh_len;
2659 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2660 		}
2661 		/*
2662 		 * Get an Open structure that will be
2663 		 * attached to the OpenOwner, acquired already.
2664 		 */
2665 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2666 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2667 		    cred, p, NULL, &op, &newone, NULL, 0, false);
2668 		if (error)
2669 			goto nfsmout;
2670 		op->nfso_stateid = stateid;
2671 		newnfs_copyincred(cred, &op->nfso_cred);
2672 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2673 		    do {
2674 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2675 			    nfhp->nfh_len, op, cred, p);
2676 			if (ret == NFSERR_DELAY)
2677 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2678 		    } while (ret == NFSERR_DELAY);
2679 		    error = ret;
2680 		}
2681 
2682 		/*
2683 		 * If the server is handing out delegations, but we didn't
2684 		 * get one because an OpenConfirm was required, try the
2685 		 * Open again, to get a delegation. This is a harmless no-op,
2686 		 * from a server's point of view.
2687 		 */
2688 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2689 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2690 		    !error && dp == NULL) {
2691 		    do {
2692 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2693 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2694 			    nfhp->nfh_fh, nfhp->nfh_len,
2695 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2696 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2697 			if (ret == NFSERR_DELAY)
2698 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2699 		    } while (ret == NFSERR_DELAY);
2700 		    if (ret) {
2701 			if (dp != NULL) {
2702 				free(dp, M_NFSCLDELEG);
2703 				dp = NULL;
2704 			}
2705 			if (ret == NFSERR_STALECLIENTID ||
2706 			    ret == NFSERR_STALEDONTRECOVER ||
2707 			    ret == NFSERR_BADSESSION)
2708 				error = ret;
2709 		    }
2710 		}
2711 		nfscl_openrelease(nmp, op, error, newone);
2712 		*unlockedp = 1;
2713 	}
2714 	if (nd->nd_repstat != 0 && error == 0)
2715 		error = nd->nd_repstat;
2716 	if (error == NFSERR_STALECLIENTID)
2717 		nfscl_initiate_recovery(owp->nfsow_clp);
2718 nfsmout:
2719 	if (!error)
2720 		*dpp = dp;
2721 	else if (dp != NULL)
2722 		free(dp, M_NFSCLDELEG);
2723 	m_freem(nd->nd_mrep);
2724 	return (error);
2725 }
2726 
2727 /*
2728  * Nfs remove rpc
2729  */
2730 int
2731 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2732     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
2733 {
2734 	u_int32_t *tl;
2735 	struct nfsrv_descript nfsd, *nd = &nfsd;
2736 	struct nfsnode *np;
2737 	struct nfsmount *nmp;
2738 	nfsv4stateid_t dstateid;
2739 	int error, ret = 0, i;
2740 
2741 	*dattrflagp = 0;
2742 	if (namelen > NFS_MAXNAMLEN)
2743 		return (ENAMETOOLONG);
2744 	nmp = VFSTONFS(dvp->v_mount);
2745 tryagain:
2746 	if (NFSHASNFSV4(nmp) && ret == 0) {
2747 		ret = nfscl_removedeleg(vp, p, &dstateid);
2748 		if (ret == 1) {
2749 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp, cred);
2750 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2751 			    NFSX_UNSIGNED);
2752 			if (NFSHASNFSV4N(nmp))
2753 				*tl++ = 0;
2754 			else
2755 				*tl++ = dstateid.seqid;
2756 			*tl++ = dstateid.other[0];
2757 			*tl++ = dstateid.other[1];
2758 			*tl++ = dstateid.other[2];
2759 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2760 			np = VTONFS(dvp);
2761 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2762 			    np->n_fhp->nfh_len, 0);
2763 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2764 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2765 		}
2766 	} else {
2767 		ret = 0;
2768 	}
2769 	if (ret == 0)
2770 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp, cred);
2771 	(void) nfsm_strtom(nd, name, namelen);
2772 	error = nfscl_request(nd, dvp, p, cred);
2773 	if (error)
2774 		return (error);
2775 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2776 		/* For NFSv4, parse out any Delereturn replies. */
2777 		if (ret > 0 && nd->nd_repstat != 0 &&
2778 		    (nd->nd_flag & ND_NOMOREDATA)) {
2779 			/*
2780 			 * If the Delegreturn failed, try again without
2781 			 * it. The server will Recall, as required.
2782 			 */
2783 			m_freem(nd->nd_mrep);
2784 			goto tryagain;
2785 		}
2786 		for (i = 0; i < (ret * 2); i++) {
2787 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2788 			    ND_NFSV4) {
2789 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2790 			    if (*(tl + 1))
2791 				nd->nd_flag |= ND_NOMOREDATA;
2792 			}
2793 		}
2794 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
2795 	}
2796 	if (nd->nd_repstat && !error)
2797 		error = nd->nd_repstat;
2798 nfsmout:
2799 	m_freem(nd->nd_mrep);
2800 	return (error);
2801 }
2802 
2803 /*
2804  * Do an nfs rename rpc.
2805  */
2806 int
2807 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2808     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2809     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2810     int *fattrflagp, int *tattrflagp)
2811 {
2812 	u_int32_t *tl;
2813 	struct nfsrv_descript nfsd, *nd = &nfsd;
2814 	struct nfsmount *nmp;
2815 	struct nfsnode *np;
2816 	nfsattrbit_t attrbits;
2817 	nfsv4stateid_t fdstateid, tdstateid;
2818 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2819 
2820 	*fattrflagp = 0;
2821 	*tattrflagp = 0;
2822 	nmp = VFSTONFS(fdvp->v_mount);
2823 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2824 		return (ENAMETOOLONG);
2825 tryagain:
2826 	if (NFSHASNFSV4(nmp) && ret == 0) {
2827 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2828 		    &tdstateid, &gottd, p);
2829 		if (gotfd && gottd) {
2830 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp, cred);
2831 		} else if (gotfd) {
2832 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp, cred);
2833 		} else if (gottd) {
2834 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp, cred);
2835 		}
2836 		if (gotfd) {
2837 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2838 			if (NFSHASNFSV4N(nmp))
2839 				*tl++ = 0;
2840 			else
2841 				*tl++ = fdstateid.seqid;
2842 			*tl++ = fdstateid.other[0];
2843 			*tl++ = fdstateid.other[1];
2844 			*tl = fdstateid.other[2];
2845 			if (gottd) {
2846 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2847 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2848 				np = VTONFS(tvp);
2849 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2850 				    np->n_fhp->nfh_len, 0);
2851 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2852 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2853 			}
2854 		}
2855 		if (gottd) {
2856 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2857 			if (NFSHASNFSV4N(nmp))
2858 				*tl++ = 0;
2859 			else
2860 				*tl++ = tdstateid.seqid;
2861 			*tl++ = tdstateid.other[0];
2862 			*tl++ = tdstateid.other[1];
2863 			*tl = tdstateid.other[2];
2864 		}
2865 		if (ret > 0) {
2866 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2867 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2868 			np = VTONFS(fdvp);
2869 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2870 			    np->n_fhp->nfh_len, 0);
2871 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2872 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2873 		}
2874 	} else {
2875 		ret = 0;
2876 	}
2877 	if (ret == 0)
2878 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp, cred);
2879 	if (nd->nd_flag & ND_NFSV4) {
2880 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2881 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2882 		NFSWCCATTR_ATTRBIT(&attrbits);
2883 		(void) nfsrv_putattrbit(nd, &attrbits);
2884 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2885 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2886 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2887 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2888 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2889 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2890 		(void) nfsrv_putattrbit(nd, &attrbits);
2891 		nd->nd_flag |= ND_V4WCCATTR;
2892 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2893 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2894 	}
2895 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2896 	if (!(nd->nd_flag & ND_NFSV4))
2897 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2898 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2899 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2900 	error = nfscl_request(nd, fdvp, p, cred);
2901 	if (error)
2902 		return (error);
2903 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2904 		/* For NFSv4, parse out any Delereturn replies. */
2905 		if (ret > 0 && nd->nd_repstat != 0 &&
2906 		    (nd->nd_flag & ND_NOMOREDATA)) {
2907 			/*
2908 			 * If the Delegreturn failed, try again without
2909 			 * it. The server will Recall, as required.
2910 			 */
2911 			m_freem(nd->nd_mrep);
2912 			goto tryagain;
2913 		}
2914 		for (i = 0; i < (ret * 2); i++) {
2915 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2916 			    ND_NFSV4) {
2917 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2918 			    if (*(tl + 1)) {
2919 				if (i == 0 && ret > 1) {
2920 				    /*
2921 				     * If the Delegreturn failed, try again
2922 				     * without it. The server will Recall, as
2923 				     * required.
2924 				     * If ret > 1, the first iteration of this
2925 				     * loop is the second DelegReturn result.
2926 				     */
2927 				    m_freem(nd->nd_mrep);
2928 				    goto tryagain;
2929 				} else {
2930 				    nd->nd_flag |= ND_NOMOREDATA;
2931 				}
2932 			    }
2933 			}
2934 		}
2935 		/* Now, the first wcc attribute reply. */
2936 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2937 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2938 			if (*(tl + 1))
2939 				nd->nd_flag |= ND_NOMOREDATA;
2940 		}
2941 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL, NULL);
2942 		/* and the second wcc attribute reply. */
2943 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2944 		    !error) {
2945 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2946 			if (*(tl + 1))
2947 				nd->nd_flag |= ND_NOMOREDATA;
2948 		}
2949 		if (!error)
2950 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2951 			    NULL, NULL);
2952 	}
2953 	if (nd->nd_repstat && !error)
2954 		error = nd->nd_repstat;
2955 nfsmout:
2956 	m_freem(nd->nd_mrep);
2957 	return (error);
2958 }
2959 
2960 /*
2961  * nfs hard link create rpc
2962  */
2963 int
2964 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2965     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2966     struct nfsvattr *nap, int *attrflagp, int *dattrflagp)
2967 {
2968 	u_int32_t *tl;
2969 	struct nfsrv_descript nfsd, *nd = &nfsd;
2970 	nfsattrbit_t attrbits;
2971 	int error = 0;
2972 
2973 	*attrflagp = 0;
2974 	*dattrflagp = 0;
2975 	if (namelen > NFS_MAXNAMLEN)
2976 		return (ENAMETOOLONG);
2977 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp, cred);
2978 	if (nd->nd_flag & ND_NFSV4) {
2979 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2980 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2981 	}
2982 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2983 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2984 	if (nd->nd_flag & ND_NFSV4) {
2985 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2986 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2987 		NFSWCCATTR_ATTRBIT(&attrbits);
2988 		(void) nfsrv_putattrbit(nd, &attrbits);
2989 		nd->nd_flag |= ND_V4WCCATTR;
2990 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2991 		*tl = txdr_unsigned(NFSV4OP_LINK);
2992 	}
2993 	(void) nfsm_strtom(nd, name, namelen);
2994 	error = nfscl_request(nd, vp, p, cred);
2995 	if (error)
2996 		return (error);
2997 	if (nd->nd_flag & ND_NFSV3) {
2998 		error = nfscl_postop_attr(nd, nap, attrflagp);
2999 		if (!error)
3000 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3001 			    NULL, NULL);
3002 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
3003 		/*
3004 		 * First, parse out the PutFH and Getattr result.
3005 		 */
3006 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3007 		if (!(*(tl + 1)))
3008 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3009 		if (*(tl + 1))
3010 			nd->nd_flag |= ND_NOMOREDATA;
3011 		/*
3012 		 * Get the pre-op attributes.
3013 		 */
3014 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3015 	}
3016 	if (nd->nd_repstat && !error)
3017 		error = nd->nd_repstat;
3018 nfsmout:
3019 	m_freem(nd->nd_mrep);
3020 	return (error);
3021 }
3022 
3023 /*
3024  * nfs symbolic link create rpc
3025  */
3026 int
3027 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
3028     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3029     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3030     int *dattrflagp)
3031 {
3032 	u_int32_t *tl;
3033 	struct nfsrv_descript nfsd, *nd = &nfsd;
3034 	struct nfsmount *nmp;
3035 	int slen, error = 0;
3036 
3037 	*nfhpp = NULL;
3038 	*attrflagp = 0;
3039 	*dattrflagp = 0;
3040 	nmp = VFSTONFS(dvp->v_mount);
3041 	slen = strlen(target);
3042 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
3043 		return (ENAMETOOLONG);
3044 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp, cred);
3045 	if (nd->nd_flag & ND_NFSV4) {
3046 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3047 		*tl = txdr_unsigned(NFLNK);
3048 		(void) nfsm_strtom(nd, target, slen);
3049 	}
3050 	(void) nfsm_strtom(nd, name, namelen);
3051 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3052 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
3053 	if (!(nd->nd_flag & ND_NFSV4))
3054 		(void) nfsm_strtom(nd, target, slen);
3055 	if (nd->nd_flag & ND_NFSV2)
3056 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3057 	error = nfscl_request(nd, dvp, p, cred);
3058 	if (error)
3059 		return (error);
3060 	if (nd->nd_flag & ND_NFSV4)
3061 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3062 	if ((nd->nd_flag & ND_NFSV3) && !error) {
3063 		if (!nd->nd_repstat)
3064 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3065 		if (!error)
3066 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
3067 			    NULL, NULL);
3068 	}
3069 	if (nd->nd_repstat && !error)
3070 		error = nd->nd_repstat;
3071 	m_freem(nd->nd_mrep);
3072 	/*
3073 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3074 	 * Only do this if vfs.nfs.ignore_eexist is set.
3075 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3076 	 * should guarantee "exactly once" RPC semantics.
3077 	 */
3078 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3079 	    nmp->nm_minorvers == 0))
3080 		error = 0;
3081 	return (error);
3082 }
3083 
3084 /*
3085  * nfs make dir rpc
3086  */
3087 int
3088 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
3089     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
3090     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
3091     int *dattrflagp)
3092 {
3093 	u_int32_t *tl;
3094 	struct nfsrv_descript nfsd, *nd = &nfsd;
3095 	nfsattrbit_t attrbits;
3096 	int error = 0;
3097 	struct nfsfh *fhp;
3098 	struct nfsmount *nmp;
3099 
3100 	*nfhpp = NULL;
3101 	*attrflagp = 0;
3102 	*dattrflagp = 0;
3103 	nmp = VFSTONFS(dvp->v_mount);
3104 	fhp = VTONFS(dvp)->n_fhp;
3105 	if (namelen > NFS_MAXNAMLEN)
3106 		return (ENAMETOOLONG);
3107 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp, cred);
3108 	if (nd->nd_flag & ND_NFSV4) {
3109 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3110 		*tl = txdr_unsigned(NFDIR);
3111 	}
3112 	(void) nfsm_strtom(nd, name, namelen);
3113 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
3114 	if (nd->nd_flag & ND_NFSV4) {
3115 		NFSGETATTR_ATTRBIT(&attrbits);
3116 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3117 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3118 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3119 		(void) nfsrv_putattrbit(nd, &attrbits);
3120 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3121 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
3122 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
3123 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3124 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3125 		(void) nfsrv_putattrbit(nd, &attrbits);
3126 	}
3127 	error = nfscl_request(nd, dvp, p, cred);
3128 	if (error)
3129 		return (error);
3130 	if (nd->nd_flag & ND_NFSV4)
3131 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3132 	if (!nd->nd_repstat && !error) {
3133 		if (nd->nd_flag & ND_NFSV4) {
3134 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3135 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
3136 		}
3137 		if (!error)
3138 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
3139 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
3140 			/* Get rid of the PutFH and Getattr status values. */
3141 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
3142 			/* Load the directory attributes. */
3143 			error = nfsm_loadattr(nd, dnap);
3144 			if (error == 0)
3145 				*dattrflagp = 1;
3146 		}
3147 	}
3148 	if ((nd->nd_flag & ND_NFSV3) && !error)
3149 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3150 	if (nd->nd_repstat && !error)
3151 		error = nd->nd_repstat;
3152 nfsmout:
3153 	m_freem(nd->nd_mrep);
3154 	/*
3155 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
3156 	 * Only do this if vfs.nfs.ignore_eexist is set.
3157 	 * Never do this for NFSv4.1 or later minor versions, since sessions
3158 	 * should guarantee "exactly once" RPC semantics.
3159 	 */
3160 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
3161 	    nmp->nm_minorvers == 0))
3162 		error = 0;
3163 	return (error);
3164 }
3165 
3166 /*
3167  * nfs remove directory call
3168  */
3169 int
3170 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
3171     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp)
3172 {
3173 	struct nfsrv_descript nfsd, *nd = &nfsd;
3174 	int error = 0;
3175 
3176 	*dattrflagp = 0;
3177 	if (namelen > NFS_MAXNAMLEN)
3178 		return (ENAMETOOLONG);
3179 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp, cred);
3180 	(void) nfsm_strtom(nd, name, namelen);
3181 	error = nfscl_request(nd, dvp, p, cred);
3182 	if (error)
3183 		return (error);
3184 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
3185 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, NULL);
3186 	if (nd->nd_repstat && !error)
3187 		error = nd->nd_repstat;
3188 	m_freem(nd->nd_mrep);
3189 	/*
3190 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
3191 	 */
3192 	if (error == ENOENT)
3193 		error = 0;
3194 	return (error);
3195 }
3196 
3197 /*
3198  * Readdir rpc.
3199  * Always returns with either uio_resid unchanged, if you are at the
3200  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
3201  * filled in.
3202  * I felt this would allow caching of directory blocks more easily
3203  * than returning a pertially filled block.
3204  * Directory offset cookies:
3205  * Oh my, what to do with them...
3206  * I can think of three ways to deal with them:
3207  * 1 - have the layer above these RPCs maintain a map between logical
3208  *     directory byte offsets and the NFS directory offset cookies
3209  * 2 - pass the opaque directory offset cookies up into userland
3210  *     and let the libc functions deal with them, via the system call
3211  * 3 - return them to userland in the "struct dirent", so future versions
3212  *     of libc can use them and do whatever is necessary to make things work
3213  *     above these rpc calls, in the meantime
3214  * For now, I do #3 by "hiding" the directory offset cookies after the
3215  * d_name field in struct dirent. This is space inside d_reclen that
3216  * will be ignored by anything that doesn't know about them.
3217  * The directory offset cookies are filled in as the last 8 bytes of
3218  * each directory entry, after d_name. Someday, the userland libc
3219  * functions may be able to use these. In the meantime, it satisfies
3220  * OpenBSD's requirements for cookies being returned.
3221  * If expects the directory offset cookie for the read to be in uio_offset
3222  * and returns the one for the next entry after this directory block in
3223  * there, as well.
3224  */
3225 int
3226 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3227     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3228     int *eofp)
3229 {
3230 	int len, left;
3231 	struct dirent *dp = NULL;
3232 	u_int32_t *tl;
3233 	nfsquad_t cookie, ncookie;
3234 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3235 	struct nfsnode *dnp = VTONFS(vp);
3236 	struct nfsvattr nfsva;
3237 	struct nfsrv_descript nfsd, *nd = &nfsd;
3238 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3239 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
3240 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3241 	char *cp;
3242 	nfsattrbit_t attrbits, dattrbits;
3243 	u_int32_t rderr, *tl2 = NULL;
3244 	size_t tresid;
3245 
3246 	KASSERT(uiop->uio_iovcnt == 1 &&
3247 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3248 	    ("nfs readdirrpc bad uio"));
3249 	ncookie.lval[0] = ncookie.lval[1] = 0;
3250 	/*
3251 	 * There is no point in reading a lot more than uio_resid, however
3252 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
3253 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
3254 	 * will never make readsize > nm_readdirsize.
3255 	 */
3256 	readsize = nmp->nm_readdirsize;
3257 	if (readsize > uiop->uio_resid)
3258 		readsize = uiop->uio_resid + DIRBLKSIZ;
3259 
3260 	*attrflagp = 0;
3261 	if (eofp)
3262 		*eofp = 0;
3263 	tresid = uiop->uio_resid;
3264 	cookie.lval[0] = cookiep->nfsuquad[0];
3265 	cookie.lval[1] = cookiep->nfsuquad[1];
3266 	nd->nd_mrep = NULL;
3267 
3268 	/*
3269 	 * For NFSv4, first create the "." and ".." entries.
3270 	 */
3271 	if (NFSHASNFSV4(nmp)) {
3272 		reqsize = 6 * NFSX_UNSIGNED;
3273 		NFSGETATTR_ATTRBIT(&dattrbits);
3274 		NFSZERO_ATTRBIT(&attrbits);
3275 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3276 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
3277 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3278 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3279 			NFSSETBIT_ATTRBIT(&attrbits,
3280 			    NFSATTRBIT_MOUNTEDONFILEID);
3281 			gotmnton = 1;
3282 		} else {
3283 			/*
3284 			 * Must fake it. Use the fileno, except when the
3285 			 * fsid is != to that of the directory. For that
3286 			 * case, generate a fake fileno that is not the same.
3287 			 */
3288 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3289 			gotmnton = 0;
3290 		}
3291 
3292 		/*
3293 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3294 		 */
3295 		if (uiop->uio_offset == 0) {
3296 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp, cred);
3297 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3298 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3299 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3300 			(void) nfsrv_putattrbit(nd, &attrbits);
3301 			error = nfscl_request(nd, vp, p, cred);
3302 			if (error)
3303 			    return (error);
3304 			dotfileid = 0;	/* Fake out the compiler. */
3305 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3306 			    error = nfsm_loadattr(nd, &nfsva);
3307 			    if (error != 0)
3308 				goto nfsmout;
3309 			    dotfileid = nfsva.na_fileid;
3310 			}
3311 			if (nd->nd_repstat == 0) {
3312 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3313 			    len = fxdr_unsigned(int, *(tl + 4));
3314 			    if (len > 0 && len <= NFSX_V4FHMAX)
3315 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3316 			    else
3317 				error = EPERM;
3318 			    if (!error) {
3319 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3320 				nfsva.na_mntonfileno = UINT64_MAX;
3321 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3322 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3323 				    NULL, NULL, NULL, p, cred);
3324 				if (error) {
3325 				    dotdotfileid = dotfileid;
3326 				} else if (gotmnton) {
3327 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3328 					dotdotfileid = nfsva.na_mntonfileno;
3329 				    else
3330 					dotdotfileid = nfsva.na_fileid;
3331 				} else if (nfsva.na_filesid[0] ==
3332 				    dnp->n_vattr.na_filesid[0] &&
3333 				    nfsva.na_filesid[1] ==
3334 				    dnp->n_vattr.na_filesid[1]) {
3335 				    dotdotfileid = nfsva.na_fileid;
3336 				} else {
3337 				    do {
3338 					fakefileno--;
3339 				    } while (fakefileno ==
3340 					nfsva.na_fileid);
3341 				    dotdotfileid = fakefileno;
3342 				}
3343 			    }
3344 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3345 			    /*
3346 			     * Lookupp returns NFSERR_NOENT when we are
3347 			     * at the root, so just use the current dir.
3348 			     */
3349 			    nd->nd_repstat = 0;
3350 			    dotdotfileid = dotfileid;
3351 			} else {
3352 			    error = nd->nd_repstat;
3353 			}
3354 			m_freem(nd->nd_mrep);
3355 			if (error)
3356 			    return (error);
3357 			nd->nd_mrep = NULL;
3358 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3359 			dp->d_pad0 = dp->d_pad1 = 0;
3360 			dp->d_off = 0;
3361 			dp->d_type = DT_DIR;
3362 			dp->d_fileno = dotfileid;
3363 			dp->d_namlen = 1;
3364 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3365 			dp->d_name[0] = '.';
3366 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3367 			/*
3368 			 * Just make these offset cookie 0.
3369 			 */
3370 			tl = (u_int32_t *)&dp->d_name[8];
3371 			*tl++ = 0;
3372 			*tl = 0;
3373 			blksiz += dp->d_reclen;
3374 			uiop->uio_resid -= dp->d_reclen;
3375 			uiop->uio_offset += dp->d_reclen;
3376 			uiop->uio_iov->iov_base =
3377 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3378 			uiop->uio_iov->iov_len -= dp->d_reclen;
3379 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3380 			dp->d_pad0 = dp->d_pad1 = 0;
3381 			dp->d_off = 0;
3382 			dp->d_type = DT_DIR;
3383 			dp->d_fileno = dotdotfileid;
3384 			dp->d_namlen = 2;
3385 			*((uint64_t *)dp->d_name) = 0;
3386 			dp->d_name[0] = '.';
3387 			dp->d_name[1] = '.';
3388 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3389 			/*
3390 			 * Just make these offset cookie 0.
3391 			 */
3392 			tl = (u_int32_t *)&dp->d_name[8];
3393 			*tl++ = 0;
3394 			*tl = 0;
3395 			blksiz += dp->d_reclen;
3396 			uiop->uio_resid -= dp->d_reclen;
3397 			uiop->uio_offset += dp->d_reclen;
3398 			uiop->uio_iov->iov_base =
3399 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3400 			uiop->uio_iov->iov_len -= dp->d_reclen;
3401 		}
3402 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3403 	} else {
3404 		reqsize = 5 * NFSX_UNSIGNED;
3405 	}
3406 
3407 	/*
3408 	 * Loop around doing readdir rpc's of size readsize.
3409 	 * The stopping criteria is EOF or buffer full.
3410 	 */
3411 	while (more_dirs && bigenough) {
3412 		*attrflagp = 0;
3413 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp, cred);
3414 		if (nd->nd_flag & ND_NFSV2) {
3415 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3416 			*tl++ = cookie.lval[1];
3417 			*tl = txdr_unsigned(readsize);
3418 		} else {
3419 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3420 			*tl++ = cookie.lval[0];
3421 			*tl++ = cookie.lval[1];
3422 			if (cookie.qval == 0) {
3423 				*tl++ = 0;
3424 				*tl++ = 0;
3425 			} else {
3426 				NFSLOCKNODE(dnp);
3427 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3428 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3429 				NFSUNLOCKNODE(dnp);
3430 			}
3431 			if (nd->nd_flag & ND_NFSV4) {
3432 				*tl++ = txdr_unsigned(readsize);
3433 				*tl = txdr_unsigned(readsize);
3434 				(void) nfsrv_putattrbit(nd, &attrbits);
3435 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3436 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3437 				(void) nfsrv_putattrbit(nd, &dattrbits);
3438 			} else {
3439 				*tl = txdr_unsigned(readsize);
3440 			}
3441 		}
3442 		error = nfscl_request(nd, vp, p, cred);
3443 		if (error)
3444 			return (error);
3445 		if (!(nd->nd_flag & ND_NFSV2)) {
3446 			if (nd->nd_flag & ND_NFSV3)
3447 				error = nfscl_postop_attr(nd, nap, attrflagp);
3448 			if (!nd->nd_repstat && !error) {
3449 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3450 				NFSLOCKNODE(dnp);
3451 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3452 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3453 				NFSUNLOCKNODE(dnp);
3454 			}
3455 		}
3456 		if (nd->nd_repstat || error) {
3457 			if (!error)
3458 				error = nd->nd_repstat;
3459 			goto nfsmout;
3460 		}
3461 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3462 		more_dirs = fxdr_unsigned(int, *tl);
3463 		if (!more_dirs)
3464 			tryformoredirs = 0;
3465 
3466 		/* loop through the dir entries, doctoring them to 4bsd form */
3467 		while (more_dirs && bigenough) {
3468 			if (nd->nd_flag & ND_NFSV4) {
3469 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3470 				ncookie.lval[0] = *tl++;
3471 				ncookie.lval[1] = *tl++;
3472 				len = fxdr_unsigned(int, *tl);
3473 			} else if (nd->nd_flag & ND_NFSV3) {
3474 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3475 				nfsva.na_fileid = fxdr_hyper(tl);
3476 				tl += 2;
3477 				len = fxdr_unsigned(int, *tl);
3478 			} else {
3479 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3480 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3481 				    *tl++);
3482 				len = fxdr_unsigned(int, *tl);
3483 			}
3484 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3485 				error = EBADRPC;
3486 				goto nfsmout;
3487 			}
3488 			tlen = roundup2(len, 8);
3489 			if (tlen == len)
3490 				tlen += 8;  /* To ensure null termination. */
3491 			left = DIRBLKSIZ - blksiz;
3492 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3493 				NFSBZERO(uiop->uio_iov->iov_base, left);
3494 				dp->d_reclen += left;
3495 				uiop->uio_iov->iov_base =
3496 				    (char *)uiop->uio_iov->iov_base + left;
3497 				uiop->uio_iov->iov_len -= left;
3498 				uiop->uio_resid -= left;
3499 				uiop->uio_offset += left;
3500 				blksiz = 0;
3501 			}
3502 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3503 			    uiop->uio_resid)
3504 				bigenough = 0;
3505 			if (bigenough) {
3506 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3507 				dp->d_pad0 = dp->d_pad1 = 0;
3508 				dp->d_off = 0;
3509 				dp->d_namlen = len;
3510 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3511 				    NFSX_HYPER;
3512 				dp->d_type = DT_UNKNOWN;
3513 				blksiz += dp->d_reclen;
3514 				if (blksiz == DIRBLKSIZ)
3515 					blksiz = 0;
3516 				uiop->uio_resid -= DIRHDSIZ;
3517 				uiop->uio_offset += DIRHDSIZ;
3518 				uiop->uio_iov->iov_base =
3519 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3520 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3521 				error = nfsm_mbufuio(nd, uiop, len);
3522 				if (error)
3523 					goto nfsmout;
3524 				cp = uiop->uio_iov->iov_base;
3525 				tlen -= len;
3526 				NFSBZERO(cp, tlen);
3527 				cp += tlen;	/* points to cookie storage */
3528 				tl2 = (u_int32_t *)cp;
3529 				uiop->uio_iov->iov_base =
3530 				    (char *)uiop->uio_iov->iov_base + tlen +
3531 				    NFSX_HYPER;
3532 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3533 				uiop->uio_resid -= tlen + NFSX_HYPER;
3534 				uiop->uio_offset += (tlen + NFSX_HYPER);
3535 			} else {
3536 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3537 				if (error)
3538 					goto nfsmout;
3539 			}
3540 			if (nd->nd_flag & ND_NFSV4) {
3541 				rderr = 0;
3542 				nfsva.na_mntonfileno = UINT64_MAX;
3543 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3544 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3545 				    NULL, NULL, &rderr, p, cred);
3546 				if (error)
3547 					goto nfsmout;
3548 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3549 			} else if (nd->nd_flag & ND_NFSV3) {
3550 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3551 				ncookie.lval[0] = *tl++;
3552 				ncookie.lval[1] = *tl++;
3553 			} else {
3554 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3555 				ncookie.lval[0] = 0;
3556 				ncookie.lval[1] = *tl++;
3557 			}
3558 			if (bigenough) {
3559 			    if (nd->nd_flag & ND_NFSV4) {
3560 				if (rderr) {
3561 				    dp->d_fileno = 0;
3562 				} else {
3563 				    if (gotmnton) {
3564 					if (nfsva.na_mntonfileno != UINT64_MAX)
3565 					    dp->d_fileno = nfsva.na_mntonfileno;
3566 					else
3567 					    dp->d_fileno = nfsva.na_fileid;
3568 				    } else if (nfsva.na_filesid[0] ==
3569 					dnp->n_vattr.na_filesid[0] &&
3570 					nfsva.na_filesid[1] ==
3571 					dnp->n_vattr.na_filesid[1]) {
3572 					dp->d_fileno = nfsva.na_fileid;
3573 				    } else {
3574 					do {
3575 					    fakefileno--;
3576 					} while (fakefileno ==
3577 					    nfsva.na_fileid);
3578 					dp->d_fileno = fakefileno;
3579 				    }
3580 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3581 				}
3582 			    } else {
3583 				dp->d_fileno = nfsva.na_fileid;
3584 			    }
3585 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3586 				ncookie.lval[0];
3587 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3588 				ncookie.lval[1];
3589 			}
3590 			more_dirs = fxdr_unsigned(int, *tl);
3591 		}
3592 		/*
3593 		 * If at end of rpc data, get the eof boolean
3594 		 */
3595 		if (!more_dirs) {
3596 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3597 			eof = fxdr_unsigned(int, *tl);
3598 			if (tryformoredirs)
3599 				more_dirs = !eof;
3600 			if (nd->nd_flag & ND_NFSV4) {
3601 				error = nfscl_postop_attr(nd, nap, attrflagp);
3602 				if (error)
3603 					goto nfsmout;
3604 			}
3605 		}
3606 		m_freem(nd->nd_mrep);
3607 		nd->nd_mrep = NULL;
3608 	}
3609 	/*
3610 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3611 	 * by increasing d_reclen for the last record.
3612 	 */
3613 	if (blksiz > 0) {
3614 		left = DIRBLKSIZ - blksiz;
3615 		NFSBZERO(uiop->uio_iov->iov_base, left);
3616 		dp->d_reclen += left;
3617 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3618 		    left;
3619 		uiop->uio_iov->iov_len -= left;
3620 		uiop->uio_resid -= left;
3621 		uiop->uio_offset += left;
3622 	}
3623 
3624 	/*
3625 	 * If returning no data, assume end of file.
3626 	 * If not bigenough, return not end of file, since you aren't
3627 	 *    returning all the data
3628 	 * Otherwise, return the eof flag from the server.
3629 	 */
3630 	if (eofp) {
3631 		if (tresid == ((size_t)(uiop->uio_resid)))
3632 			*eofp = 1;
3633 		else if (!bigenough)
3634 			*eofp = 0;
3635 		else
3636 			*eofp = eof;
3637 	}
3638 
3639 	/*
3640 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3641 	 */
3642 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3643 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3644 		NFSBZERO(dp, DIRBLKSIZ);
3645 		dp->d_type = DT_UNKNOWN;
3646 		tl = (u_int32_t *)&dp->d_name[4];
3647 		*tl++ = cookie.lval[0];
3648 		*tl = cookie.lval[1];
3649 		dp->d_reclen = DIRBLKSIZ;
3650 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3651 		    DIRBLKSIZ;
3652 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3653 		uiop->uio_resid -= DIRBLKSIZ;
3654 		uiop->uio_offset += DIRBLKSIZ;
3655 	}
3656 
3657 nfsmout:
3658 	if (nd->nd_mrep != NULL)
3659 		m_freem(nd->nd_mrep);
3660 	return (error);
3661 }
3662 
3663 /*
3664  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3665  * (Also used for NFS V4 when mount flag set.)
3666  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3667  */
3668 int
3669 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3670     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3671     int *eofp)
3672 {
3673 	int len, left;
3674 	struct dirent *dp = NULL;
3675 	u_int32_t *tl;
3676 	vnode_t newvp = NULLVP;
3677 	struct nfsrv_descript nfsd, *nd = &nfsd;
3678 	struct nameidata nami, *ndp = &nami;
3679 	struct componentname *cnp = &ndp->ni_cnd;
3680 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3681 	struct nfsnode *dnp = VTONFS(vp), *np;
3682 	struct nfsvattr nfsva;
3683 	struct nfsfh *nfhp;
3684 	nfsquad_t cookie, ncookie;
3685 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3686 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3687 	int isdotdot = 0, unlocknewvp = 0;
3688 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3689 	u_int64_t fileno = 0;
3690 	char *cp;
3691 	nfsattrbit_t attrbits, dattrbits;
3692 	size_t tresid;
3693 	u_int32_t *tl2 = NULL, rderr;
3694 	struct timespec dctime, ts;
3695 	bool attr_ok;
3696 
3697 	KASSERT(uiop->uio_iovcnt == 1 &&
3698 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3699 	    ("nfs readdirplusrpc bad uio"));
3700 	ncookie.lval[0] = ncookie.lval[1] = 0;
3701 	timespecclear(&dctime);
3702 	*attrflagp = 0;
3703 	if (eofp != NULL)
3704 		*eofp = 0;
3705 	ndp->ni_dvp = vp;
3706 	nd->nd_mrep = NULL;
3707 	cookie.lval[0] = cookiep->nfsuquad[0];
3708 	cookie.lval[1] = cookiep->nfsuquad[1];
3709 	tresid = uiop->uio_resid;
3710 
3711 	/*
3712 	 * For NFSv4, first create the "." and ".." entries.
3713 	 */
3714 	if (NFSHASNFSV4(nmp)) {
3715 		NFSGETATTR_ATTRBIT(&dattrbits);
3716 		NFSZERO_ATTRBIT(&attrbits);
3717 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3718 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3719 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3720 			NFSSETBIT_ATTRBIT(&attrbits,
3721 			    NFSATTRBIT_MOUNTEDONFILEID);
3722 			gotmnton = 1;
3723 		} else {
3724 			/*
3725 			 * Must fake it. Use the fileno, except when the
3726 			 * fsid is != to that of the directory. For that
3727 			 * case, generate a fake fileno that is not the same.
3728 			 */
3729 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3730 			gotmnton = 0;
3731 		}
3732 
3733 		/*
3734 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3735 		 */
3736 		if (uiop->uio_offset == 0) {
3737 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp, cred);
3738 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3739 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3740 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3741 			(void) nfsrv_putattrbit(nd, &attrbits);
3742 			error = nfscl_request(nd, vp, p, cred);
3743 			if (error)
3744 			    return (error);
3745 			dotfileid = 0;	/* Fake out the compiler. */
3746 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3747 			    error = nfsm_loadattr(nd, &nfsva);
3748 			    if (error != 0)
3749 				goto nfsmout;
3750 			    dctime = nfsva.na_ctime;
3751 			    dotfileid = nfsva.na_fileid;
3752 			}
3753 			if (nd->nd_repstat == 0) {
3754 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3755 			    len = fxdr_unsigned(int, *(tl + 4));
3756 			    if (len > 0 && len <= NFSX_V4FHMAX)
3757 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3758 			    else
3759 				error = EPERM;
3760 			    if (!error) {
3761 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3762 				nfsva.na_mntonfileno = UINT64_MAX;
3763 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3764 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3765 				    NULL, NULL, NULL, p, cred);
3766 				if (error) {
3767 				    dotdotfileid = dotfileid;
3768 				} else if (gotmnton) {
3769 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3770 					dotdotfileid = nfsva.na_mntonfileno;
3771 				    else
3772 					dotdotfileid = nfsva.na_fileid;
3773 				} else if (nfsva.na_filesid[0] ==
3774 				    dnp->n_vattr.na_filesid[0] &&
3775 				    nfsva.na_filesid[1] ==
3776 				    dnp->n_vattr.na_filesid[1]) {
3777 				    dotdotfileid = nfsva.na_fileid;
3778 				} else {
3779 				    do {
3780 					fakefileno--;
3781 				    } while (fakefileno ==
3782 					nfsva.na_fileid);
3783 				    dotdotfileid = fakefileno;
3784 				}
3785 			    }
3786 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3787 			    /*
3788 			     * Lookupp returns NFSERR_NOENT when we are
3789 			     * at the root, so just use the current dir.
3790 			     */
3791 			    nd->nd_repstat = 0;
3792 			    dotdotfileid = dotfileid;
3793 			} else {
3794 			    error = nd->nd_repstat;
3795 			}
3796 			m_freem(nd->nd_mrep);
3797 			if (error)
3798 			    return (error);
3799 			nd->nd_mrep = NULL;
3800 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3801 			dp->d_pad0 = dp->d_pad1 = 0;
3802 			dp->d_off = 0;
3803 			dp->d_type = DT_DIR;
3804 			dp->d_fileno = dotfileid;
3805 			dp->d_namlen = 1;
3806 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3807 			dp->d_name[0] = '.';
3808 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3809 			/*
3810 			 * Just make these offset cookie 0.
3811 			 */
3812 			tl = (u_int32_t *)&dp->d_name[8];
3813 			*tl++ = 0;
3814 			*tl = 0;
3815 			blksiz += dp->d_reclen;
3816 			uiop->uio_resid -= dp->d_reclen;
3817 			uiop->uio_offset += dp->d_reclen;
3818 			uiop->uio_iov->iov_base =
3819 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3820 			uiop->uio_iov->iov_len -= dp->d_reclen;
3821 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3822 			dp->d_pad0 = dp->d_pad1 = 0;
3823 			dp->d_off = 0;
3824 			dp->d_type = DT_DIR;
3825 			dp->d_fileno = dotdotfileid;
3826 			dp->d_namlen = 2;
3827 			*((uint64_t *)dp->d_name) = 0;
3828 			dp->d_name[0] = '.';
3829 			dp->d_name[1] = '.';
3830 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3831 			/*
3832 			 * Just make these offset cookie 0.
3833 			 */
3834 			tl = (u_int32_t *)&dp->d_name[8];
3835 			*tl++ = 0;
3836 			*tl = 0;
3837 			blksiz += dp->d_reclen;
3838 			uiop->uio_resid -= dp->d_reclen;
3839 			uiop->uio_offset += dp->d_reclen;
3840 			uiop->uio_iov->iov_base =
3841 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3842 			uiop->uio_iov->iov_len -= dp->d_reclen;
3843 		}
3844 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3845 		if (gotmnton)
3846 			NFSSETBIT_ATTRBIT(&attrbits,
3847 			    NFSATTRBIT_MOUNTEDONFILEID);
3848 		if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3849 		    NFSATTRBIT_TIMECREATE))
3850 			NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
3851 	}
3852 
3853 	/*
3854 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3855 	 * The stopping criteria is EOF or buffer full.
3856 	 */
3857 	while (more_dirs && bigenough) {
3858 		*attrflagp = 0;
3859 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp, cred);
3860  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3861 		*tl++ = cookie.lval[0];
3862 		*tl++ = cookie.lval[1];
3863 		if (cookie.qval == 0) {
3864 			*tl++ = 0;
3865 			*tl++ = 0;
3866 		} else {
3867 			NFSLOCKNODE(dnp);
3868 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3869 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3870 			NFSUNLOCKNODE(dnp);
3871 		}
3872 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3873 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3874 		if (nd->nd_flag & ND_NFSV4) {
3875 			(void) nfsrv_putattrbit(nd, &attrbits);
3876 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3877 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3878 			(void) nfsrv_putattrbit(nd, &dattrbits);
3879 		}
3880 		nanouptime(&ts);
3881 		error = nfscl_request(nd, vp, p, cred);
3882 		if (error)
3883 			return (error);
3884 		if (nd->nd_flag & ND_NFSV3)
3885 			error = nfscl_postop_attr(nd, nap, attrflagp);
3886 		if (nd->nd_repstat || error) {
3887 			if (!error)
3888 				error = nd->nd_repstat;
3889 			goto nfsmout;
3890 		}
3891 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3892 			dctime = nap->na_ctime;
3893 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3894 		NFSLOCKNODE(dnp);
3895 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3896 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3897 		NFSUNLOCKNODE(dnp);
3898 		more_dirs = fxdr_unsigned(int, *tl);
3899 		if (!more_dirs)
3900 			tryformoredirs = 0;
3901 
3902 		/* loop through the dir entries, doctoring them to 4bsd form */
3903 		while (more_dirs && bigenough) {
3904 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3905 			if (nd->nd_flag & ND_NFSV4) {
3906 				ncookie.lval[0] = *tl++;
3907 				ncookie.lval[1] = *tl++;
3908 			} else {
3909 				fileno = fxdr_hyper(tl);
3910 				tl += 2;
3911 			}
3912 			len = fxdr_unsigned(int, *tl);
3913 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3914 				error = EBADRPC;
3915 				goto nfsmout;
3916 			}
3917 			tlen = roundup2(len, 8);
3918 			if (tlen == len)
3919 				tlen += 8;  /* To ensure null termination. */
3920 			left = DIRBLKSIZ - blksiz;
3921 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3922 				NFSBZERO(uiop->uio_iov->iov_base, left);
3923 				dp->d_reclen += left;
3924 				uiop->uio_iov->iov_base =
3925 				    (char *)uiop->uio_iov->iov_base + left;
3926 				uiop->uio_iov->iov_len -= left;
3927 				uiop->uio_resid -= left;
3928 				uiop->uio_offset += left;
3929 				blksiz = 0;
3930 			}
3931 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3932 			    uiop->uio_resid)
3933 				bigenough = 0;
3934 			if (bigenough) {
3935 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3936 				dp->d_pad0 = dp->d_pad1 = 0;
3937 				dp->d_off = 0;
3938 				dp->d_namlen = len;
3939 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3940 				    NFSX_HYPER;
3941 				dp->d_type = DT_UNKNOWN;
3942 				blksiz += dp->d_reclen;
3943 				if (blksiz == DIRBLKSIZ)
3944 					blksiz = 0;
3945 				uiop->uio_resid -= DIRHDSIZ;
3946 				uiop->uio_offset += DIRHDSIZ;
3947 				uiop->uio_iov->iov_base =
3948 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3949 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3950 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3951 				cnp->cn_namelen = len;
3952 				NFSCNHASHZERO(cnp);
3953 				error = nfsm_mbufuio(nd, uiop, len);
3954 				if (error)
3955 					goto nfsmout;
3956 				cp = uiop->uio_iov->iov_base;
3957 				tlen -= len;
3958 				NFSBZERO(cp, tlen);
3959 				cp += tlen;	/* points to cookie storage */
3960 				tl2 = (u_int32_t *)cp;
3961 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3962 				    cnp->cn_nameptr[1] == '.')
3963 					isdotdot = 1;
3964 				else
3965 					isdotdot = 0;
3966 				uiop->uio_iov->iov_base =
3967 				    (char *)uiop->uio_iov->iov_base + tlen +
3968 				    NFSX_HYPER;
3969 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3970 				uiop->uio_resid -= tlen + NFSX_HYPER;
3971 				uiop->uio_offset += (tlen + NFSX_HYPER);
3972 			} else {
3973 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3974 				if (error)
3975 					goto nfsmout;
3976 			}
3977 			nfhp = NULL;
3978 			if (nd->nd_flag & ND_NFSV3) {
3979 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3980 				ncookie.lval[0] = *tl++;
3981 				ncookie.lval[1] = *tl++;
3982 				attrflag = fxdr_unsigned(int, *tl);
3983 				if (attrflag) {
3984 				  error = nfsm_loadattr(nd, &nfsva);
3985 				  if (error)
3986 					goto nfsmout;
3987 				}
3988 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3989 				if (*tl) {
3990 					error = nfsm_getfh(nd, &nfhp);
3991 					if (error)
3992 					    goto nfsmout;
3993 				}
3994 				if (!attrflag && nfhp != NULL) {
3995 					free(nfhp, M_NFSFH);
3996 					nfhp = NULL;
3997 				}
3998 			} else {
3999 				rderr = 0;
4000 				nfsva.na_mntonfileno = 0xffffffff;
4001 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
4002 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
4003 				    NULL, NULL, &rderr, p, cred);
4004 				if (error)
4005 					goto nfsmout;
4006 			}
4007 
4008 			if (bigenough) {
4009 			    if (nd->nd_flag & ND_NFSV4) {
4010 				if (rderr) {
4011 				    dp->d_fileno = 0;
4012 				} else if (gotmnton) {
4013 				    if (nfsva.na_mntonfileno != 0xffffffff)
4014 					dp->d_fileno = nfsva.na_mntonfileno;
4015 				    else
4016 					dp->d_fileno = nfsva.na_fileid;
4017 				} else if (nfsva.na_filesid[0] ==
4018 				    dnp->n_vattr.na_filesid[0] &&
4019 				    nfsva.na_filesid[1] ==
4020 				    dnp->n_vattr.na_filesid[1]) {
4021 				    dp->d_fileno = nfsva.na_fileid;
4022 				} else {
4023 				    do {
4024 					fakefileno--;
4025 				    } while (fakefileno ==
4026 					nfsva.na_fileid);
4027 				    dp->d_fileno = fakefileno;
4028 				}
4029 			    } else {
4030 				dp->d_fileno = fileno;
4031 			    }
4032 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
4033 				ncookie.lval[0];
4034 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
4035 				ncookie.lval[1];
4036 
4037 			    if (nfhp != NULL) {
4038 				attr_ok = true;
4039 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
4040 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
4041 				    VREF(vp);
4042 				    newvp = vp;
4043 				    unlocknewvp = 0;
4044 				    free(nfhp, M_NFSFH);
4045 				    np = dnp;
4046 				} else if (isdotdot != 0) {
4047 				    /*
4048 				     * Skip doing a nfscl_nget() call for "..".
4049 				     * There's a race between acquiring the nfs
4050 				     * node here and lookups that look for the
4051 				     * directory being read (in the parent).
4052 				     * It would try to get a lock on ".." here,
4053 				     * owning the lock on the directory being
4054 				     * read. Lookup will hold the lock on ".."
4055 				     * and try to acquire the lock on the
4056 				     * directory being read.
4057 				     * If the directory is unlocked/relocked,
4058 				     * then there is a LOR with the buflock
4059 				     * vp is relocked.
4060 				     */
4061 				    free(nfhp, M_NFSFH);
4062 				} else {
4063 				    error = nfscl_nget(vp->v_mount, vp,
4064 				      nfhp, cnp, p, &np, LK_EXCLUSIVE);
4065 				    if (!error) {
4066 					newvp = NFSTOV(np);
4067 					unlocknewvp = 1;
4068 					/*
4069 					 * If n_localmodtime >= time before RPC,
4070 					 * then a file modification operation,
4071 					 * such as VOP_SETATTR() of size, has
4072 					 * occurred while the Lookup RPC and
4073 					 * acquisition of the vnode happened. As
4074 					 * such, the attributes might be stale,
4075 					 * with possibly an incorrect size.
4076 					 */
4077 					NFSLOCKNODE(np);
4078 					if (timespecisset(
4079 					    &np->n_localmodtime) &&
4080 					    timespeccmp(&np->n_localmodtime,
4081 					    &ts, >=)) {
4082 					    NFSCL_DEBUG(4, "nfsrpc_readdirplus:"
4083 						" localmod stale attributes\n");
4084 					    attr_ok = false;
4085 					}
4086 					NFSUNLOCKNODE(np);
4087 				    }
4088 				}
4089 				nfhp = NULL;
4090 				if (newvp != NULLVP) {
4091 				    if (attr_ok)
4092 					error = nfscl_loadattrcache(&newvp,
4093 					    &nfsva, NULL, 0, 0);
4094 				    if (error) {
4095 					if (unlocknewvp)
4096 					    vput(newvp);
4097 					else
4098 					    vrele(newvp);
4099 					goto nfsmout;
4100 				    }
4101 				    dp->d_type =
4102 					vtonfs_dtype(np->n_vattr.na_type);
4103 				    ndp->ni_vp = newvp;
4104 				    NFSCNHASH(cnp, HASHINIT);
4105 				    if (cnp->cn_namelen <= NCHNAMLEN &&
4106 					ndp->ni_dvp != ndp->ni_vp &&
4107 					(newvp->v_type != VDIR ||
4108 					 dctime.tv_sec != 0)) {
4109 					cache_enter_time_flags(ndp->ni_dvp,
4110 					    ndp->ni_vp, cnp,
4111 					    &nfsva.na_ctime,
4112 					    newvp->v_type != VDIR ? NULL :
4113 					    &dctime, VFS_CACHE_DROPOLD);
4114 				    }
4115 				    if (unlocknewvp)
4116 					vput(newvp);
4117 				    else
4118 					vrele(newvp);
4119 				    newvp = NULLVP;
4120 				}
4121 			    }
4122 			} else if (nfhp != NULL) {
4123 			    free(nfhp, M_NFSFH);
4124 			}
4125 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4126 			more_dirs = fxdr_unsigned(int, *tl);
4127 		}
4128 		/*
4129 		 * If at end of rpc data, get the eof boolean
4130 		 */
4131 		if (!more_dirs) {
4132 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
4133 			eof = fxdr_unsigned(int, *tl);
4134 			if (tryformoredirs)
4135 				more_dirs = !eof;
4136 			if (nd->nd_flag & ND_NFSV4) {
4137 				error = nfscl_postop_attr(nd, nap, attrflagp);
4138 				if (error)
4139 					goto nfsmout;
4140 			}
4141 		}
4142 		m_freem(nd->nd_mrep);
4143 		nd->nd_mrep = NULL;
4144 	}
4145 	/*
4146 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
4147 	 * by increasing d_reclen for the last record.
4148 	 */
4149 	if (blksiz > 0) {
4150 		left = DIRBLKSIZ - blksiz;
4151 		NFSBZERO(uiop->uio_iov->iov_base, left);
4152 		dp->d_reclen += left;
4153 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4154 		    left;
4155 		uiop->uio_iov->iov_len -= left;
4156 		uiop->uio_resid -= left;
4157 		uiop->uio_offset += left;
4158 	}
4159 
4160 	/*
4161 	 * If returning no data, assume end of file.
4162 	 * If not bigenough, return not end of file, since you aren't
4163 	 *    returning all the data
4164 	 * Otherwise, return the eof flag from the server.
4165 	 */
4166 	if (eofp != NULL) {
4167 		if (tresid == uiop->uio_resid)
4168 			*eofp = 1;
4169 		else if (!bigenough)
4170 			*eofp = 0;
4171 		else
4172 			*eofp = eof;
4173 	}
4174 
4175 	/*
4176 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
4177 	 */
4178 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
4179 		dp = (struct dirent *)uiop->uio_iov->iov_base;
4180 		NFSBZERO(dp, DIRBLKSIZ);
4181 		dp->d_type = DT_UNKNOWN;
4182 		tl = (u_int32_t *)&dp->d_name[4];
4183 		*tl++ = cookie.lval[0];
4184 		*tl = cookie.lval[1];
4185 		dp->d_reclen = DIRBLKSIZ;
4186 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
4187 		    DIRBLKSIZ;
4188 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
4189 		uiop->uio_resid -= DIRBLKSIZ;
4190 		uiop->uio_offset += DIRBLKSIZ;
4191 	}
4192 
4193 nfsmout:
4194 	if (nd->nd_mrep != NULL)
4195 		m_freem(nd->nd_mrep);
4196 	return (error);
4197 }
4198 
4199 /*
4200  * Nfs commit rpc
4201  */
4202 int
4203 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
4204     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4205 {
4206 	u_int32_t *tl;
4207 	struct nfsrv_descript nfsd, *nd = &nfsd;
4208 	nfsattrbit_t attrbits;
4209 	int error;
4210 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4211 
4212 	*attrflagp = 0;
4213 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp, cred);
4214 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
4215 	txdr_hyper(offset, tl);
4216 	tl += 2;
4217 	*tl = txdr_unsigned(cnt);
4218 	if (nd->nd_flag & ND_NFSV4) {
4219 		/*
4220 		 * And do a Getattr op.
4221 		 */
4222 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4223 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
4224 		NFSGETATTR_ATTRBIT(&attrbits);
4225 		(void) nfsrv_putattrbit(nd, &attrbits);
4226 	}
4227 	error = nfscl_request(nd, vp, p, cred);
4228 	if (error)
4229 		return (error);
4230 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, NULL);
4231 	if (!error && !nd->nd_repstat) {
4232 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
4233 		NFSLOCKMNT(nmp);
4234 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
4235 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
4236 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
4237 		}
4238 		NFSUNLOCKMNT(nmp);
4239 		if (nd->nd_flag & ND_NFSV4)
4240 			error = nfscl_postop_attr(nd, nap, attrflagp);
4241 	}
4242 nfsmout:
4243 	if (!error && nd->nd_repstat)
4244 		error = nd->nd_repstat;
4245 	m_freem(nd->nd_mrep);
4246 	return (error);
4247 }
4248 
4249 /*
4250  * NFS byte range lock rpc.
4251  * (Mostly just calls one of the three lower level RPC routines.)
4252  */
4253 int
4254 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
4255     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4256 {
4257 	struct nfscllockowner *lp;
4258 	struct nfsclclient *clp;
4259 	struct nfsfh *nfhp;
4260 	struct nfsrv_descript nfsd, *nd = &nfsd;
4261 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4262 	u_int64_t off, len;
4263 	off_t start, end;
4264 	u_int32_t clidrev = 0;
4265 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
4266 	int callcnt, dorpc;
4267 
4268 	/*
4269 	 * Convert the flock structure into a start and end and do POSIX
4270 	 * bounds checking.
4271 	 */
4272 	switch (fl->l_whence) {
4273 	case SEEK_SET:
4274 	case SEEK_CUR:
4275 		/*
4276 		 * Caller is responsible for adding any necessary offset
4277 		 * when SEEK_CUR is used.
4278 		 */
4279 		start = fl->l_start;
4280 		off = fl->l_start;
4281 		break;
4282 	case SEEK_END:
4283 		start = size + fl->l_start;
4284 		off = size + fl->l_start;
4285 		break;
4286 	default:
4287 		return (EINVAL);
4288 	}
4289 	if (start < 0)
4290 		return (EINVAL);
4291 	if (fl->l_len != 0) {
4292 		end = start + fl->l_len - 1;
4293 		if (end < start)
4294 			return (EINVAL);
4295 	}
4296 
4297 	len = fl->l_len;
4298 	if (len == 0)
4299 		len = NFS64BITSSET;
4300 	retrycnt = 0;
4301 	do {
4302 	    nd->nd_repstat = 0;
4303 	    if (op == F_GETLK) {
4304 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4305 		if (error)
4306 			return (error);
4307 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
4308 		if (!error) {
4309 			clidrev = clp->nfsc_clientidrev;
4310 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
4311 			    p, id, flags);
4312 		} else if (error == -1) {
4313 			error = 0;
4314 		}
4315 		nfscl_clientrelease(clp);
4316 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
4317 		/*
4318 		 * We must loop around for all lockowner cases.
4319 		 */
4320 		callcnt = 0;
4321 		error = nfscl_getcl(vp->v_mount, cred, p, false, true, &clp);
4322 		if (error)
4323 			return (error);
4324 		do {
4325 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
4326 			clp, id, flags, &lp, &dorpc);
4327 		    /*
4328 		     * If it returns a NULL lp, we're done.
4329 		     */
4330 		    if (lp == NULL) {
4331 			if (callcnt == 0)
4332 			    nfscl_clientrelease(clp);
4333 			else
4334 			    nfscl_releasealllocks(clp, vp, p, id, flags);
4335 			return (error);
4336 		    }
4337 		    if (nmp->nm_clp != NULL)
4338 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4339 		    else
4340 			clidrev = 0;
4341 		    /*
4342 		     * If the server doesn't support Posix lock semantics,
4343 		     * only allow locks on the entire file, since it won't
4344 		     * handle overlapping byte ranges.
4345 		     * There might still be a problem when a lock
4346 		     * upgrade/downgrade (read<->write) occurs, since the
4347 		     * server "might" expect an unlock first?
4348 		     */
4349 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4350 			(off == 0 && len == NFS64BITSSET))) {
4351 			/*
4352 			 * Since the lock records will go away, we must
4353 			 * wait for grace and delay here.
4354 			 */
4355 			do {
4356 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4357 				NFSV4LOCKT_READ, cred, p, 0);
4358 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4359 				 nd->nd_repstat == NFSERR_DELAY) &&
4360 				error == 0)
4361 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4362 				    "nfs_advlock");
4363 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4364 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4365 		    }
4366 		    callcnt++;
4367 		} while (error == 0 && nd->nd_repstat == 0);
4368 		nfscl_releasealllocks(clp, vp, p, id, flags);
4369 	    } else if (op == F_SETLK) {
4370 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4371 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4372 		if (error || donelocally) {
4373 			return (error);
4374 		}
4375 		if (nmp->nm_clp != NULL)
4376 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4377 		else
4378 			clidrev = 0;
4379 		nfhp = VTONFS(vp)->n_fhp;
4380 		if (!lp->nfsl_open->nfso_posixlock &&
4381 		    (off != 0 || len != NFS64BITSSET)) {
4382 			error = EINVAL;
4383 		} else {
4384 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4385 			    nfhp->nfh_len, lp, newone, reclaim, off,
4386 			    len, fl->l_type, cred, p, 0);
4387 		}
4388 		if (!error)
4389 			error = nd->nd_repstat;
4390 		nfscl_lockrelease(lp, error, newone);
4391 	    } else {
4392 		error = EINVAL;
4393 	    }
4394 	    if (!error)
4395 	        error = nd->nd_repstat;
4396 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4397 		error == NFSERR_STALEDONTRECOVER ||
4398 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4399 		error == NFSERR_BADSESSION) {
4400 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4401 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4402 		&& clidrev != 0) {
4403 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4404 		retrycnt++;
4405 	    }
4406 	} while (error == NFSERR_GRACE ||
4407 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4408 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4409 	    error == NFSERR_BADSESSION ||
4410 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4411 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4412 	if (error && retrycnt >= 4)
4413 		error = EIO;
4414 	return (error);
4415 }
4416 
4417 /*
4418  * The lower level routine for the LockT case.
4419  */
4420 int
4421 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4422     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4423     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4424 {
4425 	u_int32_t *tl;
4426 	int error, type, size;
4427 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4428 	struct nfsnode *np;
4429 	struct nfsmount *nmp;
4430 	struct nfsclsession *tsep;
4431 
4432 	nmp = VFSTONFS(vp->v_mount);
4433 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp, cred);
4434 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4435 	if (fl->l_type == F_RDLCK)
4436 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4437 	else
4438 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4439 	txdr_hyper(off, tl);
4440 	tl += 2;
4441 	txdr_hyper(len, tl);
4442 	tl += 2;
4443 	tsep = nfsmnt_mdssession(nmp);
4444 	*tl++ = tsep->nfsess_clientid.lval[0];
4445 	*tl = tsep->nfsess_clientid.lval[1];
4446 	nfscl_filllockowner(id, own, flags);
4447 	np = VTONFS(vp);
4448 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4449 	    np->n_fhp->nfh_len);
4450 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4451 	error = nfscl_request(nd, vp, p, cred);
4452 	if (error)
4453 		return (error);
4454 	if (nd->nd_repstat == 0) {
4455 		fl->l_type = F_UNLCK;
4456 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4457 		nd->nd_repstat = 0;
4458 		fl->l_whence = SEEK_SET;
4459 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4460 		fl->l_start = fxdr_hyper(tl);
4461 		tl += 2;
4462 		len = fxdr_hyper(tl);
4463 		tl += 2;
4464 		if (len == NFS64BITSSET)
4465 			fl->l_len = 0;
4466 		else
4467 			fl->l_len = len;
4468 		type = fxdr_unsigned(int, *tl++);
4469 		if (type == NFSV4LOCKT_WRITE)
4470 			fl->l_type = F_WRLCK;
4471 		else
4472 			fl->l_type = F_RDLCK;
4473 		/*
4474 		 * XXX For now, I have no idea what to do with the
4475 		 * conflicting lock_owner, so I'll just set the pid == 0
4476 		 * and skip over the lock_owner.
4477 		 */
4478 		fl->l_pid = (pid_t)0;
4479 		tl += 2;
4480 		size = fxdr_unsigned(int, *tl);
4481 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4482 			error = EBADRPC;
4483 		if (!error)
4484 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4485 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4486 		nfscl_initiate_recovery(clp);
4487 nfsmout:
4488 	m_freem(nd->nd_mrep);
4489 	return (error);
4490 }
4491 
4492 /*
4493  * Lower level function that performs the LockU RPC.
4494  */
4495 static int
4496 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4497     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4498     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4499 {
4500 	u_int32_t *tl;
4501 	int error;
4502 
4503 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4504 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0, cred);
4505 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4506 	*tl++ = txdr_unsigned(type);
4507 	*tl = txdr_unsigned(lp->nfsl_seqid);
4508 	if (nfstest_outofseq &&
4509 	    (arc4random() % nfstest_outofseq) == 0)
4510 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4511 	tl++;
4512 	if (NFSHASNFSV4N(nmp))
4513 		*tl++ = 0;
4514 	else
4515 		*tl++ = lp->nfsl_stateid.seqid;
4516 	*tl++ = lp->nfsl_stateid.other[0];
4517 	*tl++ = lp->nfsl_stateid.other[1];
4518 	*tl++ = lp->nfsl_stateid.other[2];
4519 	txdr_hyper(off, tl);
4520 	tl += 2;
4521 	txdr_hyper(len, tl);
4522 	if (syscred)
4523 		nd->nd_flag |= ND_USEGSSNAME;
4524 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4525 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4526 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4527 	if (error)
4528 		return (error);
4529 	if (nd->nd_repstat == 0) {
4530 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4531 		lp->nfsl_stateid.seqid = *tl++;
4532 		lp->nfsl_stateid.other[0] = *tl++;
4533 		lp->nfsl_stateid.other[1] = *tl++;
4534 		lp->nfsl_stateid.other[2] = *tl;
4535 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4536 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4537 nfsmout:
4538 	m_freem(nd->nd_mrep);
4539 	return (error);
4540 }
4541 
4542 /*
4543  * The actual Lock RPC.
4544  */
4545 int
4546 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4547     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4548     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4549     NFSPROC_T *p, int syscred)
4550 {
4551 	u_int32_t *tl;
4552 	int error, size;
4553 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4554 	struct nfsclsession *tsep;
4555 
4556 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0,
4557 	    cred);
4558 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4559 	if (type == F_RDLCK)
4560 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4561 	else
4562 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4563 	*tl++ = txdr_unsigned(reclaim);
4564 	txdr_hyper(off, tl);
4565 	tl += 2;
4566 	txdr_hyper(len, tl);
4567 	tl += 2;
4568 	if (newone) {
4569 	    *tl = newnfs_true;
4570 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4571 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4572 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4573 	    if (NFSHASNFSV4N(nmp))
4574 		*tl++ = 0;
4575 	    else
4576 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4577 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4578 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4579 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4580 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4581 	    tsep = nfsmnt_mdssession(nmp);
4582 	    *tl++ = tsep->nfsess_clientid.lval[0];
4583 	    *tl = tsep->nfsess_clientid.lval[1];
4584 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4585 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4586 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4587 	} else {
4588 	    *tl = newnfs_false;
4589 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4590 	    if (NFSHASNFSV4N(nmp))
4591 		*tl++ = 0;
4592 	    else
4593 		*tl++ = lp->nfsl_stateid.seqid;
4594 	    *tl++ = lp->nfsl_stateid.other[0];
4595 	    *tl++ = lp->nfsl_stateid.other[1];
4596 	    *tl++ = lp->nfsl_stateid.other[2];
4597 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4598 	    if (nfstest_outofseq &&
4599 		(arc4random() % nfstest_outofseq) == 0)
4600 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4601 	}
4602 	if (syscred)
4603 		nd->nd_flag |= ND_USEGSSNAME;
4604 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4605 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4606 	if (error)
4607 		return (error);
4608 	if (newone)
4609 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4610 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4611 	if (nd->nd_repstat == 0) {
4612 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4613 		lp->nfsl_stateid.seqid = *tl++;
4614 		lp->nfsl_stateid.other[0] = *tl++;
4615 		lp->nfsl_stateid.other[1] = *tl++;
4616 		lp->nfsl_stateid.other[2] = *tl;
4617 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4618 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4619 		size = fxdr_unsigned(int, *(tl + 7));
4620 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4621 			error = EBADRPC;
4622 		if (!error)
4623 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4624 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4625 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4626 nfsmout:
4627 	m_freem(nd->nd_mrep);
4628 	return (error);
4629 }
4630 
4631 /*
4632  * nfs statfs rpc
4633  * (always called with the vp for the mount point)
4634  */
4635 int
4636 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4637     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4638 {
4639 	u_int32_t *tl = NULL;
4640 	struct nfsrv_descript nfsd, *nd = &nfsd;
4641 	struct nfsmount *nmp;
4642 	nfsattrbit_t attrbits;
4643 	int error;
4644 
4645 	*attrflagp = 0;
4646 	nmp = VFSTONFS(vp->v_mount);
4647 	if (NFSHASNFSV4(nmp)) {
4648 		/*
4649 		 * For V4, you actually do a getattr.
4650 		 */
4651 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
4652 		NFSSTATFS_GETATTRBIT(&attrbits);
4653 		(void) nfsrv_putattrbit(nd, &attrbits);
4654 		nd->nd_flag |= ND_USEGSSNAME;
4655 		error = nfscl_request(nd, vp, p, cred);
4656 		if (error)
4657 			return (error);
4658 		if (nd->nd_repstat == 0) {
4659 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4660 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4661 			    cred);
4662 			if (!error) {
4663 				nmp->nm_fsid[0] = nap->na_filesid[0];
4664 				nmp->nm_fsid[1] = nap->na_filesid[1];
4665 				NFSSETHASSETFSID(nmp);
4666 				*attrflagp = 1;
4667 			}
4668 		} else {
4669 			error = nd->nd_repstat;
4670 		}
4671 		if (error)
4672 			goto nfsmout;
4673 	} else {
4674 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp, NULL);
4675 		error = nfscl_request(nd, vp, p, cred);
4676 		if (error)
4677 			return (error);
4678 		if (nd->nd_flag & ND_NFSV3) {
4679 			error = nfscl_postop_attr(nd, nap, attrflagp);
4680 			if (error)
4681 				goto nfsmout;
4682 		}
4683 		if (nd->nd_repstat) {
4684 			error = nd->nd_repstat;
4685 			goto nfsmout;
4686 		}
4687 		NFSM_DISSECT(tl, u_int32_t *,
4688 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4689 	}
4690 	if (NFSHASNFSV3(nmp)) {
4691 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4692 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4693 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4694 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4695 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4696 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4697 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4698 	} else if (NFSHASNFSV4(nmp) == 0) {
4699 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4700 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4701 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4702 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4703 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4704 	}
4705 nfsmout:
4706 	m_freem(nd->nd_mrep);
4707 	return (error);
4708 }
4709 
4710 /*
4711  * nfs pathconf rpc
4712  */
4713 int
4714 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4715     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4716 {
4717 	struct nfsrv_descript nfsd, *nd = &nfsd;
4718 	struct nfsmount *nmp;
4719 	u_int32_t *tl;
4720 	nfsattrbit_t attrbits;
4721 	int error;
4722 
4723 	*attrflagp = 0;
4724 	nmp = VFSTONFS(vp->v_mount);
4725 	if (NFSHASNFSV4(nmp)) {
4726 		/*
4727 		 * For V4, you actually do a getattr.
4728 		 */
4729 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp, cred);
4730 		NFSPATHCONF_GETATTRBIT(&attrbits);
4731 		(void) nfsrv_putattrbit(nd, &attrbits);
4732 		nd->nd_flag |= ND_USEGSSNAME;
4733 		error = nfscl_request(nd, vp, p, cred);
4734 		if (error)
4735 			return (error);
4736 		if (nd->nd_repstat == 0) {
4737 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4738 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4739 			    cred);
4740 			if (!error)
4741 				*attrflagp = 1;
4742 		} else {
4743 			error = nd->nd_repstat;
4744 		}
4745 	} else {
4746 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp, NULL);
4747 		error = nfscl_request(nd, vp, p, cred);
4748 		if (error)
4749 			return (error);
4750 		error = nfscl_postop_attr(nd, nap, attrflagp);
4751 		if (nd->nd_repstat && !error)
4752 			error = nd->nd_repstat;
4753 		if (!error) {
4754 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4755 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4756 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4757 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4758 			pc->pc_chownrestricted =
4759 			    fxdr_unsigned(u_int32_t, *tl++);
4760 			pc->pc_caseinsensitive =
4761 			    fxdr_unsigned(u_int32_t, *tl++);
4762 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4763 		}
4764 	}
4765 nfsmout:
4766 	m_freem(nd->nd_mrep);
4767 	return (error);
4768 }
4769 
4770 /*
4771  * nfs version 3 fsinfo rpc call
4772  */
4773 int
4774 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4775     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
4776 {
4777 	u_int32_t *tl;
4778 	struct nfsrv_descript nfsd, *nd = &nfsd;
4779 	int error;
4780 
4781 	*attrflagp = 0;
4782 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp, NULL);
4783 	error = nfscl_request(nd, vp, p, cred);
4784 	if (error)
4785 		return (error);
4786 	error = nfscl_postop_attr(nd, nap, attrflagp);
4787 	if (nd->nd_repstat && !error)
4788 		error = nd->nd_repstat;
4789 	if (!error) {
4790 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4791 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4792 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4793 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4794 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4795 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4796 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4797 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4798 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4799 		tl += 2;
4800 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4801 		tl += 2;
4802 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4803 	}
4804 nfsmout:
4805 	m_freem(nd->nd_mrep);
4806 	return (error);
4807 }
4808 
4809 /*
4810  * This function performs the Renew RPC.
4811  */
4812 int
4813 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4814     NFSPROC_T *p)
4815 {
4816 	u_int32_t *tl;
4817 	struct nfsrv_descript nfsd;
4818 	struct nfsrv_descript *nd = &nfsd;
4819 	struct nfsmount *nmp;
4820 	int error;
4821 	struct nfssockreq *nrp;
4822 	struct nfsclsession *tsep;
4823 
4824 	nmp = clp->nfsc_nmp;
4825 	if (nmp == NULL)
4826 		return (0);
4827 	if (dsp == NULL)
4828 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4829 		    0, cred);
4830 	else
4831 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4832 		    &dsp->nfsclds_sess, 0, 0, NULL);
4833 	if (!NFSHASNFSV4N(nmp)) {
4834 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4835 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4836 		tsep = nfsmnt_mdssession(nmp);
4837 		*tl++ = tsep->nfsess_clientid.lval[0];
4838 		*tl = tsep->nfsess_clientid.lval[1];
4839 	}
4840 	nrp = NULL;
4841 	if (dsp != NULL)
4842 		nrp = dsp->nfsclds_sockp;
4843 	if (nrp == NULL)
4844 		/* If NULL, use the MDS socket. */
4845 		nrp = &nmp->nm_sockreq;
4846 	nd->nd_flag |= ND_USEGSSNAME;
4847 	if (dsp == NULL)
4848 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4849 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4850 	else {
4851 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4852 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4853 		if (error == ENXIO)
4854 			nfscl_cancelreqs(dsp);
4855 	}
4856 	if (error)
4857 		return (error);
4858 	error = nd->nd_repstat;
4859 	m_freem(nd->nd_mrep);
4860 	return (error);
4861 }
4862 
4863 /*
4864  * This function performs the Releaselockowner RPC.
4865  */
4866 int
4867 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4868     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4869 {
4870 	struct nfsrv_descript nfsd, *nd = &nfsd;
4871 	u_int32_t *tl;
4872 	int error;
4873 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4874 	struct nfsclsession *tsep;
4875 
4876 	if (NFSHASNFSV4N(nmp)) {
4877 		/* For NFSv4.1, do a FreeStateID. */
4878 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4879 		    NULL, 0, 0, cred);
4880 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4881 	} else {
4882 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4883 		    NULL, 0, 0, NULL);
4884 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4885 		tsep = nfsmnt_mdssession(nmp);
4886 		*tl++ = tsep->nfsess_clientid.lval[0];
4887 		*tl = tsep->nfsess_clientid.lval[1];
4888 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4889 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4890 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4891 	}
4892 	nd->nd_flag |= ND_USEGSSNAME;
4893 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4894 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4895 	if (error)
4896 		return (error);
4897 	error = nd->nd_repstat;
4898 	m_freem(nd->nd_mrep);
4899 	return (error);
4900 }
4901 
4902 /*
4903  * This function performs the Compound to get the mount pt FH.
4904  */
4905 int
4906 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4907     NFSPROC_T *p)
4908 {
4909 	u_int32_t *tl;
4910 	struct nfsrv_descript nfsd;
4911 	struct nfsrv_descript *nd = &nfsd;
4912 	u_char *cp, *cp2;
4913 	int error, cnt, len, setnil;
4914 	u_int32_t *opcntp;
4915 
4916 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4917 	    0, NULL);
4918 	cp = dirpath;
4919 	cnt = 0;
4920 	do {
4921 		setnil = 0;
4922 		while (*cp == '/')
4923 			cp++;
4924 		cp2 = cp;
4925 		while (*cp2 != '\0' && *cp2 != '/')
4926 			cp2++;
4927 		if (*cp2 == '/') {
4928 			setnil = 1;
4929 			*cp2 = '\0';
4930 		}
4931 		if (cp2 != cp) {
4932 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4933 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4934 			nfsm_strtom(nd, cp, strlen(cp));
4935 			cnt++;
4936 		}
4937 		if (setnil)
4938 			*cp2++ = '/';
4939 		cp = cp2;
4940 	} while (*cp != '\0');
4941 	if (NFSHASNFSV4N(nmp))
4942 		/* Has a Sequence Op done by nfscl_reqstart(). */
4943 		*opcntp = txdr_unsigned(3 + cnt);
4944 	else
4945 		*opcntp = txdr_unsigned(2 + cnt);
4946 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4947 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4948 	nd->nd_flag |= ND_USEGSSNAME;
4949 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4950 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4951 	if (error)
4952 		return (error);
4953 	if (nd->nd_repstat == 0) {
4954 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4955 		tl += (2 + 2 * cnt);
4956 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4957 			len > NFSX_FHMAX) {
4958 			nd->nd_repstat = NFSERR_BADXDR;
4959 		} else {
4960 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4961 			if (nd->nd_repstat == 0)
4962 				nmp->nm_fhsize = len;
4963 		}
4964 	}
4965 	error = nd->nd_repstat;
4966 nfsmout:
4967 	m_freem(nd->nd_mrep);
4968 	return (error);
4969 }
4970 
4971 /*
4972  * This function performs the Delegreturn RPC.
4973  */
4974 int
4975 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4976     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4977 {
4978 	u_int32_t *tl;
4979 	struct nfsrv_descript nfsd;
4980 	struct nfsrv_descript *nd = &nfsd;
4981 	int error;
4982 
4983 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4984 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0, cred);
4985 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4986 	if (NFSHASNFSV4N(nmp))
4987 		*tl++ = 0;
4988 	else
4989 		*tl++ = dp->nfsdl_stateid.seqid;
4990 	*tl++ = dp->nfsdl_stateid.other[0];
4991 	*tl++ = dp->nfsdl_stateid.other[1];
4992 	*tl = dp->nfsdl_stateid.other[2];
4993 	if (syscred)
4994 		nd->nd_flag |= ND_USEGSSNAME;
4995 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4996 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4997 	if (error)
4998 		return (error);
4999 	error = nd->nd_repstat;
5000 	m_freem(nd->nd_mrep);
5001 	return (error);
5002 }
5003 
5004 /*
5005  * nfs getacl call.
5006  */
5007 int
5008 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
5009 {
5010 	struct nfsrv_descript nfsd, *nd = &nfsd;
5011 	int error;
5012 	nfsattrbit_t attrbits;
5013 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5014 
5015 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5016 		return (EOPNOTSUPP);
5017 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp, cred);
5018 	NFSZERO_ATTRBIT(&attrbits);
5019 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5020 	(void) nfsrv_putattrbit(nd, &attrbits);
5021 	error = nfscl_request(nd, vp, p, cred);
5022 	if (error)
5023 		return (error);
5024 	if (!nd->nd_repstat)
5025 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
5026 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
5027 	else
5028 		error = nd->nd_repstat;
5029 	m_freem(nd->nd_mrep);
5030 	return (error);
5031 }
5032 
5033 /*
5034  * nfs setacl call.
5035  */
5036 int
5037 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p, struct acl *aclp)
5038 {
5039 	int error;
5040 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5041 
5042 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
5043 		return (EOPNOTSUPP);
5044 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL);
5045 	return (error);
5046 }
5047 
5048 /*
5049  * nfs setacl call.
5050  */
5051 static int
5052 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
5053     struct acl *aclp, nfsv4stateid_t *stateidp)
5054 {
5055 	struct nfsrv_descript nfsd, *nd = &nfsd;
5056 	int error;
5057 	nfsattrbit_t attrbits;
5058 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5059 
5060 	if (!NFSHASNFSV4(nmp))
5061 		return (EOPNOTSUPP);
5062 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp, cred);
5063 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
5064 	NFSZERO_ATTRBIT(&attrbits);
5065 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
5066 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
5067 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
5068 	error = nfscl_request(nd, vp, p, cred);
5069 	if (error)
5070 		return (error);
5071 	/* Don't care about the pre/postop attributes */
5072 	m_freem(nd->nd_mrep);
5073 	return (nd->nd_repstat);
5074 }
5075 
5076 /*
5077  * Do the NFSv4.1 Exchange ID.
5078  */
5079 int
5080 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
5081     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
5082     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
5083 {
5084 	uint32_t *tl, v41flags;
5085 	struct nfsrv_descript nfsd;
5086 	struct nfsrv_descript *nd = &nfsd;
5087 	struct nfsclds *dsp;
5088 	struct timespec verstime;
5089 	int error, len;
5090 
5091 	*dspp = NULL;
5092 	if (minorvers == 0)
5093 		minorvers = nmp->nm_minorvers;
5094 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
5095 	    NFS_VER4, minorvers, NULL);
5096 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5097 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
5098 	*tl = txdr_unsigned(clp->nfsc_rev);
5099 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
5100 
5101 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
5102 	*tl++ = txdr_unsigned(exchflags);
5103 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
5104 
5105 	/* Set the implementation id4 */
5106 	*tl = txdr_unsigned(1);
5107 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
5108 	(void) nfsm_strtom(nd, version, strlen(version));
5109 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
5110 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
5111 	verstime.tv_nsec = 0;
5112 	txdr_nfsv4time(&verstime, tl);
5113 	nd->nd_flag |= ND_USEGSSNAME;
5114 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
5115 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5116 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
5117 	    (int)nd->nd_repstat);
5118 	if (error != 0)
5119 		return (error);
5120 	if (nd->nd_repstat == 0) {
5121 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
5122 		len = fxdr_unsigned(int, *(tl + 7));
5123 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
5124 			error = NFSERR_BADXDR;
5125 			goto nfsmout;
5126 		}
5127 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
5128 		    M_WAITOK | M_ZERO);
5129 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
5130 		dsp->nfsclds_servownlen = len;
5131 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
5132 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
5133 		dsp->nfsclds_sess.nfsess_sequenceid =
5134 		    fxdr_unsigned(uint32_t, *tl++);
5135 		v41flags = fxdr_unsigned(uint32_t, *tl);
5136 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
5137 		    NFSHASPNFSOPT(nmp)) {
5138 			NFSCL_DEBUG(1, "set PNFS\n");
5139 			NFSLOCKMNT(nmp);
5140 			nmp->nm_state |= NFSSTA_PNFS;
5141 			NFSUNLOCKMNT(nmp);
5142 			dsp->nfsclds_flags |= NFSCLDS_MDS;
5143 		}
5144 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
5145 			dsp->nfsclds_flags |= NFSCLDS_DS;
5146 		if (minorvers == NFSV42_MINORVERSION)
5147 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
5148 		if (len > 0)
5149 			nd->nd_repstat = nfsrv_mtostr(nd,
5150 			    dsp->nfsclds_serverown, len);
5151 		if (nd->nd_repstat == 0) {
5152 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5153 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5154 			    NULL, MTX_DEF);
5155 			nfscl_initsessionslots(&dsp->nfsclds_sess);
5156 			*dspp = dsp;
5157 		} else
5158 			free(dsp, M_NFSCLDS);
5159 	}
5160 	error = nd->nd_repstat;
5161 nfsmout:
5162 	m_freem(nd->nd_mrep);
5163 	return (error);
5164 }
5165 
5166 /*
5167  * Do the NFSv4.1 Create Session.
5168  */
5169 int
5170 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
5171     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
5172     struct ucred *cred, NFSPROC_T *p)
5173 {
5174 	uint32_t crflags, maxval, *tl;
5175 	struct nfsrv_descript nfsd;
5176 	struct nfsrv_descript *nd = &nfsd;
5177 	int error, irdcnt, minorvers;
5178 
5179 	/* Make sure nm_rsize, nm_wsize is set. */
5180 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
5181 		nmp->nm_rsize = NFS_MAXBSIZE;
5182 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
5183 		nmp->nm_wsize = NFS_MAXBSIZE;
5184 	if (dsp == NULL)
5185 		minorvers = nmp->nm_minorvers;
5186 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
5187 		minorvers = NFSV42_MINORVERSION;
5188 	else
5189 		minorvers = NFSV41_MINORVERSION;
5190 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
5191 	    NFS_VER4, minorvers, NULL);
5192 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5193 	*tl++ = sep->nfsess_clientid.lval[0];
5194 	*tl++ = sep->nfsess_clientid.lval[1];
5195 	*tl++ = txdr_unsigned(sequenceid);
5196 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
5197 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
5198 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
5199 	*tl = txdr_unsigned(crflags);
5200 
5201 	/* Fill in fore channel attributes. */
5202 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5203 	*tl++ = 0;				/* Header pad size */
5204 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
5205 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
5206 		/*
5207 		 * NFSv4.2 Extended Attribute operations may want to do
5208 		 * requests/replies that are larger than nm_rsize/nm_wsize.
5209 		 */
5210 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5211 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
5212 	} else {
5213 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
5214 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
5215 	}
5216 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5217 	*tl++ = txdr_unsigned(20);		/* Max operations */
5218 	*tl++ = txdr_unsigned(64);		/* Max slots */
5219 	*tl = 0;				/* No rdma ird */
5220 
5221 	/* Fill in back channel attributes. */
5222 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5223 	*tl++ = 0;				/* Header pad size */
5224 	*tl++ = txdr_unsigned(10000);		/* Max request size */
5225 	*tl++ = txdr_unsigned(10000);		/* Max response size */
5226 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
5227 	*tl++ = txdr_unsigned(4);		/* Max operations */
5228 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
5229 	*tl = 0;				/* No rdma ird */
5230 
5231 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
5232 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
5233 
5234 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
5235 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
5236 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
5237 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
5238 	*tl++ = 0;				/* Null machine name */
5239 	*tl++ = 0;				/* Uid == 0 */
5240 	*tl++ = 0;				/* Gid == 0 */
5241 	*tl = 0;				/* No additional gids */
5242 	nd->nd_flag |= ND_USEGSSNAME;
5243 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
5244 	    NFS_VER4, NULL, 1, NULL, NULL);
5245 	if (error != 0)
5246 		return (error);
5247 	if (nd->nd_repstat == 0) {
5248 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
5249 		    2 * NFSX_UNSIGNED);
5250 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
5251 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
5252 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
5253 		crflags = fxdr_unsigned(uint32_t, *tl);
5254 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
5255 			NFSLOCKMNT(nmp);
5256 			nmp->nm_state |= NFSSTA_SESSPERSIST;
5257 			NFSUNLOCKMNT(nmp);
5258 		}
5259 
5260 		/* Get the fore channel slot count. */
5261 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5262 		tl++;			/* Skip the header pad size. */
5263 
5264 		/* Make sure nm_wsize is small enough. */
5265 		maxval = fxdr_unsigned(uint32_t, *tl++);
5266 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
5267 			if (nmp->nm_wsize > 8096)
5268 				nmp->nm_wsize /= 2;
5269 			else
5270 				break;
5271 		}
5272 		sep->nfsess_maxreq = maxval;
5273 
5274 		/* Make sure nm_rsize is small enough. */
5275 		maxval = fxdr_unsigned(uint32_t, *tl++);
5276 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
5277 			if (nmp->nm_rsize > 8096)
5278 				nmp->nm_rsize /= 2;
5279 			else
5280 				break;
5281 		}
5282 		sep->nfsess_maxresp = maxval;
5283 
5284 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
5285 		tl++;
5286 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
5287 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
5288 		irdcnt = fxdr_unsigned(int, *tl);
5289 		if (irdcnt < 0 || irdcnt > 1) {
5290 			error = NFSERR_BADXDR;
5291 			goto nfsmout;
5292 		}
5293 		if (irdcnt > 0)
5294 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
5295 
5296 		/* and the back channel slot count. */
5297 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
5298 		tl += 5;
5299 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
5300 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
5301 	}
5302 	error = nd->nd_repstat;
5303 nfsmout:
5304 	m_freem(nd->nd_mrep);
5305 	return (error);
5306 }
5307 
5308 /*
5309  * Do the NFSv4.1 Destroy Client.
5310  */
5311 int
5312 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5313     struct ucred *cred, NFSPROC_T *p)
5314 {
5315 	uint32_t *tl;
5316 	struct nfsrv_descript nfsd;
5317 	struct nfsrv_descript *nd = &nfsd;
5318 	int error;
5319 	struct nfsclsession *tsep;
5320 
5321 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5322 	    0, NULL);
5323 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5324 	tsep = nfsmnt_mdssession(nmp);
5325 	*tl++ = tsep->nfsess_clientid.lval[0];
5326 	*tl = tsep->nfsess_clientid.lval[1];
5327 	nd->nd_flag |= ND_USEGSSNAME;
5328 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5329 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5330 	if (error != 0)
5331 		return (error);
5332 	error = nd->nd_repstat;
5333 	m_freem(nd->nd_mrep);
5334 	return (error);
5335 }
5336 
5337 /*
5338  * Do the NFSv4.1 LayoutGet.
5339  */
5340 static int
5341 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5342     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5343     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5344     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p)
5345 {
5346 	struct nfsrv_descript nfsd, *nd = &nfsd;
5347 	int error;
5348 
5349 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5350 	    0, cred);
5351 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5352 	    layouttype, layoutlen, 0);
5353 	nd->nd_flag |= ND_USEGSSNAME;
5354 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5355 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5356 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5357 	if (error != 0)
5358 		return (error);
5359 	if (nd->nd_repstat == 0)
5360 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5361 		    flhp);
5362 	if (error == 0 && nd->nd_repstat != 0)
5363 		error = nd->nd_repstat;
5364 	m_freem(nd->nd_mrep);
5365 	return (error);
5366 }
5367 
5368 /*
5369  * Do the NFSv4.1 Get Device Info.
5370  */
5371 int
5372 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5373     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5374     NFSPROC_T *p)
5375 {
5376 	uint32_t cnt, *tl, vers, minorvers;
5377 	struct nfsrv_descript nfsd;
5378 	struct nfsrv_descript *nd = &nfsd;
5379 	struct sockaddr_in sin, ssin;
5380 	struct sockaddr_in6 sin6, ssin6;
5381 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5382 	struct nfscldevinfo *ndi;
5383 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5384 	int stripecnt;
5385 	uint8_t stripeindex;
5386 	sa_family_t af, safilled;
5387 
5388 	ssin.sin_port = 0;		/* To shut up compiler. */
5389 	ssin.sin_addr.s_addr = 0;	/* ditto */
5390 	*ndip = NULL;
5391 	ndi = NULL;
5392 	gotdspp = NULL;
5393 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5394 	    0, cred);
5395 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5396 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5397 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5398 	*tl++ = txdr_unsigned(layouttype);
5399 	*tl++ = txdr_unsigned(100000);
5400 	if (notifybitsp != NULL && *notifybitsp != 0) {
5401 		*tl = txdr_unsigned(1);		/* One word of bits. */
5402 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5403 		*tl = txdr_unsigned(*notifybitsp);
5404 	} else
5405 		*tl = txdr_unsigned(0);
5406 	nd->nd_flag |= ND_USEGSSNAME;
5407 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5408 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5409 	if (error != 0)
5410 		return (error);
5411 	if (nd->nd_repstat == 0) {
5412 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5413 		if (layouttype != fxdr_unsigned(int, *tl))
5414 			printf("EEK! devinfo layout type not same!\n");
5415 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5416 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5417 			stripecnt = fxdr_unsigned(int, *tl);
5418 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5419 			if (stripecnt < 1 || stripecnt > 4096) {
5420 				printf("pNFS File layout devinfo stripecnt %d:"
5421 				    " out of range\n", stripecnt);
5422 				error = NFSERR_BADXDR;
5423 				goto nfsmout;
5424 			}
5425 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5426 			    NFSX_UNSIGNED);
5427 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5428 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5429 			if (addrcnt < 1 || addrcnt > 128) {
5430 				printf("NFS devinfo addrcnt %d: out of range\n",
5431 				    addrcnt);
5432 				error = NFSERR_BADXDR;
5433 				goto nfsmout;
5434 			}
5435 
5436 			/*
5437 			 * Now we know how many stripe indices and addresses, so
5438 			 * we can allocate the structure the correct size.
5439 			 */
5440 			i = (stripecnt * sizeof(uint8_t)) /
5441 			    sizeof(struct nfsclds *) + 1;
5442 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5443 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5444 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5445 			    M_ZERO);
5446 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5447 			    NFSX_V4DEVICEID);
5448 			ndi->nfsdi_refcnt = 0;
5449 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5450 			ndi->nfsdi_stripecnt = stripecnt;
5451 			ndi->nfsdi_addrcnt = addrcnt;
5452 			/* Fill in the stripe indices. */
5453 			for (i = 0; i < stripecnt; i++) {
5454 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5455 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5456 				if (stripeindex >= addrcnt) {
5457 					printf("pNFS File Layout devinfo"
5458 					    " stripeindex %d: too big\n",
5459 					    (int)stripeindex);
5460 					error = NFSERR_BADXDR;
5461 					goto nfsmout;
5462 				}
5463 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5464 			}
5465 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5466 			/* For Flex File, we only get one address list. */
5467 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5468 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5469 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5470 			    NFSX_V4DEVICEID);
5471 			ndi->nfsdi_refcnt = 0;
5472 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5473 			addrcnt = ndi->nfsdi_addrcnt = 1;
5474 		}
5475 
5476 		/* Now, dissect the server address(es). */
5477 		safilled = AF_UNSPEC;
5478 		for (i = 0; i < addrcnt; i++) {
5479 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5480 			cnt = fxdr_unsigned(uint32_t, *tl);
5481 			if (cnt == 0) {
5482 				printf("NFS devinfo 0 len addrlist\n");
5483 				error = NFSERR_BADXDR;
5484 				goto nfsmout;
5485 			}
5486 			dspp = nfsfldi_addr(ndi, i);
5487 			safilled = AF_UNSPEC;
5488 			for (j = 0; j < cnt; j++) {
5489 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5490 				    &isudp);
5491 				if (error != 0 && error != EPERM) {
5492 					error = NFSERR_BADXDR;
5493 					goto nfsmout;
5494 				}
5495 				if (error == 0 && isudp == 0) {
5496 					/*
5497 					 * The priority is:
5498 					 * - Same address family.
5499 					 * Save the address and dspp, so that
5500 					 * the connection can be done after
5501 					 * parsing is complete.
5502 					 */
5503 					if (safilled == AF_UNSPEC ||
5504 					    (af == nmp->nm_nam->sa_family &&
5505 					     safilled != nmp->nm_nam->sa_family)
5506 					   ) {
5507 						if (af == AF_INET)
5508 							ssin = sin;
5509 						else
5510 							ssin6 = sin6;
5511 						safilled = af;
5512 						gotdspp = dspp;
5513 					}
5514 				}
5515 			}
5516 		}
5517 
5518 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5519 		gotminor = NFSV41_MINORVERSION;
5520 		/* For Flex File, we will take one of the versions to use. */
5521 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5522 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5523 			j = fxdr_unsigned(int, *tl);
5524 			if (j < 1 || j > NFSDEV_MAXVERS) {
5525 				printf("pNFS: too many versions\n");
5526 				error = NFSERR_BADXDR;
5527 				goto nfsmout;
5528 			}
5529 			gotvers = 0;
5530 			gotminor = 0;
5531 			for (i = 0; i < j; i++) {
5532 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5533 				vers = fxdr_unsigned(uint32_t, *tl++);
5534 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5535 				if (vers == NFS_VER3)
5536 					minorvers = 0;
5537 				if ((vers == NFS_VER4 && ((minorvers ==
5538 				    NFSV41_MINORVERSION && gotminor == 0) ||
5539 				    minorvers == NFSV42_MINORVERSION)) ||
5540 				    (vers == NFS_VER3 && gotvers == 0)) {
5541 					gotvers = vers;
5542 					gotminor = minorvers;
5543 					/* We'll take this one. */
5544 					ndi->nfsdi_versindex = i;
5545 					ndi->nfsdi_vers = vers;
5546 					ndi->nfsdi_minorvers = minorvers;
5547 					ndi->nfsdi_rsize = fxdr_unsigned(
5548 					    uint32_t, *tl++);
5549 					ndi->nfsdi_wsize = fxdr_unsigned(
5550 					    uint32_t, *tl++);
5551 					if (*tl == newnfs_true)
5552 						ndi->nfsdi_flags |=
5553 						    NFSDI_TIGHTCOUPLED;
5554 					else
5555 						ndi->nfsdi_flags &=
5556 						    ~NFSDI_TIGHTCOUPLED;
5557 				}
5558 			}
5559 			if (gotvers == 0) {
5560 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5561 				error = NFSERR_BADXDR;
5562 				goto nfsmout;
5563 			}
5564 		}
5565 
5566 		/* And the notify bits. */
5567 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5568 		bitcnt = fxdr_unsigned(int, *tl);
5569 		if (bitcnt > 0) {
5570 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5571 			if (notifybitsp != NULL)
5572 				*notifybitsp =
5573 				    fxdr_unsigned(uint32_t, *tl);
5574 		}
5575 		if (safilled != AF_UNSPEC) {
5576 			KASSERT(ndi != NULL, ("ndi is NULL"));
5577 			*ndip = ndi;
5578 		} else
5579 			error = EPERM;
5580 		if (error == 0) {
5581 			/*
5582 			 * Now we can do a TCP connection for the correct
5583 			 * NFS version and IP address.
5584 			 */
5585 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5586 			    gotvers, gotminor, &dsp, p);
5587 		}
5588 		if (error == 0) {
5589 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5590 			*gotdspp = dsp;
5591 		}
5592 	}
5593 	if (nd->nd_repstat != 0 && error == 0)
5594 		error = nd->nd_repstat;
5595 nfsmout:
5596 	if (error != 0 && ndi != NULL)
5597 		nfscl_freedevinfo(ndi);
5598 	m_freem(nd->nd_mrep);
5599 	return (error);
5600 }
5601 
5602 /*
5603  * Do the NFSv4.1 LayoutCommit.
5604  */
5605 int
5606 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5607     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5608     int layouttype, struct ucred *cred, NFSPROC_T *p)
5609 {
5610 	uint32_t *tl;
5611 	struct nfsrv_descript nfsd, *nd = &nfsd;
5612 	int error;
5613 
5614 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5615 	    0, 0, cred);
5616 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5617 	    NFSX_STATEID);
5618 	txdr_hyper(off, tl);
5619 	tl += 2;
5620 	txdr_hyper(len, tl);
5621 	tl += 2;
5622 	if (reclaim != 0)
5623 		*tl++ = newnfs_true;
5624 	else
5625 		*tl++ = newnfs_false;
5626 	*tl++ = txdr_unsigned(stateidp->seqid);
5627 	*tl++ = stateidp->other[0];
5628 	*tl++ = stateidp->other[1];
5629 	*tl++ = stateidp->other[2];
5630 	*tl++ = newnfs_true;
5631 	if (lastbyte < off)
5632 		lastbyte = off;
5633 	else if (lastbyte >= (off + len))
5634 		lastbyte = off + len - 1;
5635 	txdr_hyper(lastbyte, tl);
5636 	tl += 2;
5637 	*tl++ = newnfs_false;
5638 	*tl++ = txdr_unsigned(layouttype);
5639 	/* All supported layouts are 0 length. */
5640 	*tl = txdr_unsigned(0);
5641 	nd->nd_flag |= ND_USEGSSNAME;
5642 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5643 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5644 	if (error != 0)
5645 		return (error);
5646 	error = nd->nd_repstat;
5647 	m_freem(nd->nd_mrep);
5648 	return (error);
5649 }
5650 
5651 /*
5652  * Do the NFSv4.1 LayoutReturn.
5653  */
5654 int
5655 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5656     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5657     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5658     uint32_t stat, uint32_t op, char *devid)
5659 {
5660 	uint32_t *tl;
5661 	struct nfsrv_descript nfsd, *nd = &nfsd;
5662 	uint64_t tu64;
5663 	int error;
5664 
5665 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5666 	    0, 0, cred);
5667 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5668 	if (reclaim != 0)
5669 		*tl++ = newnfs_true;
5670 	else
5671 		*tl++ = newnfs_false;
5672 	*tl++ = txdr_unsigned(layouttype);
5673 	*tl++ = txdr_unsigned(iomode);
5674 	*tl = txdr_unsigned(layoutreturn);
5675 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5676 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5677 		    NFSX_UNSIGNED);
5678 		txdr_hyper(offset, tl);
5679 		tl += 2;
5680 		txdr_hyper(len, tl);
5681 		tl += 2;
5682 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5683 		*tl++ = txdr_unsigned(stateidp->seqid);
5684 		*tl++ = stateidp->other[0];
5685 		*tl++ = stateidp->other[1];
5686 		*tl++ = stateidp->other[2];
5687 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5688 			*tl = txdr_unsigned(0);
5689 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5690 			if (stat != 0) {
5691 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5692 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5693 				    NFSX_UNSIGNED);
5694 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5695 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5696 				    NFSX_UNSIGNED);
5697 				*tl++ = txdr_unsigned(1);	/* One error. */
5698 				tu64 = 0;			/* Offset. */
5699 				txdr_hyper(tu64, tl); tl += 2;
5700 				tu64 = UINT64_MAX;		/* Length. */
5701 				txdr_hyper(tu64, tl); tl += 2;
5702 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5703 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5704 				*tl++ = txdr_unsigned(1);	/* One error. */
5705 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5706 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5707 				*tl++ = txdr_unsigned(stat);
5708 				*tl++ = txdr_unsigned(op);
5709 			} else {
5710 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5711 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5712 				/* No ioerrs. */
5713 				*tl++ = 0;
5714 			}
5715 			*tl = 0;	/* No stats yet. */
5716 		}
5717 	}
5718 	nd->nd_flag |= ND_USEGSSNAME;
5719 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5720 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5721 	if (error != 0)
5722 		return (error);
5723 	if (nd->nd_repstat == 0) {
5724 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5725 		if (*tl != 0) {
5726 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5727 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5728 			stateidp->other[0] = *tl++;
5729 			stateidp->other[1] = *tl++;
5730 			stateidp->other[2] = *tl;
5731 		}
5732 	} else
5733 		error = nd->nd_repstat;
5734 nfsmout:
5735 	m_freem(nd->nd_mrep);
5736 	return (error);
5737 }
5738 
5739 /*
5740  * Do the NFSv4.2 LayoutError.
5741  */
5742 static int
5743 nfsrpc_layouterror(struct nfsmount *nmp, uint8_t *fh, int fhlen, uint64_t offset,
5744     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5745     uint32_t stat, uint32_t op, char *devid)
5746 {
5747 	uint32_t *tl;
5748 	struct nfsrv_descript nfsd, *nd = &nfsd;
5749 	int error;
5750 
5751 	nfscl_reqstart(nd, NFSPROC_LAYOUTERROR, nmp, fh, fhlen, NULL, NULL,
5752 	    0, 0, cred);
5753 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5754 	    NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5755 	txdr_hyper(offset, tl); tl += 2;
5756 	txdr_hyper(len, tl); tl += 2;
5757 	*tl++ = txdr_unsigned(stateidp->seqid);
5758 	*tl++ = stateidp->other[0];
5759 	*tl++ = stateidp->other[1];
5760 	*tl++ = stateidp->other[2];
5761 	*tl++ = txdr_unsigned(1);
5762 	NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5763 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5764 	*tl++ = txdr_unsigned(stat);
5765 	*tl = txdr_unsigned(op);
5766 	nd->nd_flag |= ND_USEGSSNAME;
5767 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5768 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5769 	if (error != 0)
5770 		return (error);
5771 	if (nd->nd_repstat != 0)
5772 		error = nd->nd_repstat;
5773 	m_freem(nd->nd_mrep);
5774 	return (error);
5775 }
5776 
5777 /*
5778  * Acquire a layout and devinfo, if possible. The caller must have acquired
5779  * a reference count on the nfsclclient structure before calling this.
5780  * Return the layout in lypp with a reference count on it, if successful.
5781  */
5782 static int
5783 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5784     int iomode, uint32_t rw, uint32_t *notifybitsp, nfsv4stateid_t *stateidp,
5785     uint64_t off, struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5786 {
5787 	struct nfscllayout *lyp;
5788 	struct nfsclflayout *flp;
5789 	struct nfsclflayouthead flh;
5790 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5791 	nfsv4stateid_t stateid;
5792 	struct nfsclsession *tsep;
5793 
5794 	*lypp = NULL;
5795 	if (NFSHASFLEXFILE(nmp))
5796 		layouttype = NFSLAYOUT_FLEXFILE;
5797 	else
5798 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5799 	/*
5800 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5801 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5802 	 * flp == NULL.
5803 	 */
5804 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5805 	    off, rw, &flp, &recalled);
5806 	islocked = 0;
5807 	if (lyp == NULL || flp == NULL) {
5808 		if (recalled != 0)
5809 			return (EIO);
5810 		LIST_INIT(&flh);
5811 		tsep = nfsmnt_mdssession(nmp);
5812 		layoutlen = tsep->nfsess_maxcache -
5813 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5814 		if (lyp == NULL) {
5815 			stateid.seqid = 0;
5816 			stateid.other[0] = stateidp->other[0];
5817 			stateid.other[1] = stateidp->other[1];
5818 			stateid.other[2] = stateidp->other[2];
5819 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5820 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5821 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5822 			    &retonclose, &flh, cred, p);
5823 		} else {
5824 			islocked = 1;
5825 			stateid.seqid = lyp->nfsly_stateid.seqid;
5826 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5827 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5828 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5829 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5830 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5831 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5832 			    &retonclose, &flh, cred, p);
5833 		}
5834 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5835 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5836 		    &flh, layouttype, error, NULL, cred, p);
5837 		if (error == 0)
5838 			*lypp = lyp;
5839 		else if (islocked != 0)
5840 			nfscl_rellayout(lyp, 1);
5841 	} else
5842 		*lypp = lyp;
5843 	return (error);
5844 }
5845 
5846 /*
5847  * Do a TCP connection plus exchange id and create session.
5848  * If successful, a "struct nfsclds" is linked into the list for the
5849  * mount point and a pointer to it is returned.
5850  */
5851 static int
5852 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5853     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5854     struct nfsclds **dspp, NFSPROC_T *p)
5855 {
5856 	struct sockaddr_in *msad, *sad;
5857 	struct sockaddr_in6 *msad6, *sad6;
5858 	struct nfsclclient *clp;
5859 	struct nfssockreq *nrp;
5860 	struct nfsclds *dsp, *tdsp;
5861 	int error, firsttry;
5862 	enum nfsclds_state retv;
5863 	uint32_t sequenceid = 0;
5864 
5865 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5866 	    ("nfsrpc_fillsa: NULL nr_cred"));
5867 	NFSLOCKCLSTATE();
5868 	clp = nmp->nm_clp;
5869 	NFSUNLOCKCLSTATE();
5870 	if (clp == NULL)
5871 		return (EPERM);
5872 	if (af == AF_INET) {
5873 		NFSLOCKMNT(nmp);
5874 		/*
5875 		 * Check to see if we already have a session for this
5876 		 * address that is usable for a DS.
5877 		 * Note that the MDS's address is in a different place
5878 		 * than the sessions already acquired for DS's.
5879 		 */
5880 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5881 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5882 		while (tdsp != NULL) {
5883 			if (msad != NULL && msad->sin_family == AF_INET &&
5884 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5885 			    sin->sin_port == msad->sin_port &&
5886 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5887 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5888 				*dspp = tdsp;
5889 				NFSUNLOCKMNT(nmp);
5890 				NFSCL_DEBUG(4, "fnd same addr\n");
5891 				return (0);
5892 			}
5893 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5894 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5895 				msad = (struct sockaddr_in *)
5896 				    tdsp->nfsclds_sockp->nr_nam;
5897 			else
5898 				msad = NULL;
5899 		}
5900 		NFSUNLOCKMNT(nmp);
5901 
5902 		/* No IP address match, so look for new/trunked one. */
5903 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5904 		sad->sin_len = sizeof(*sad);
5905 		sad->sin_family = AF_INET;
5906 		sad->sin_port = sin->sin_port;
5907 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5908 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5909 		nrp->nr_nam = (struct sockaddr *)sad;
5910 	} else if (af == AF_INET6) {
5911 		NFSLOCKMNT(nmp);
5912 		/*
5913 		 * Check to see if we already have a session for this
5914 		 * address that is usable for a DS.
5915 		 * Note that the MDS's address is in a different place
5916 		 * than the sessions already acquired for DS's.
5917 		 */
5918 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5919 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5920 		while (tdsp != NULL) {
5921 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5922 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5923 			    &msad6->sin6_addr) &&
5924 			    sin6->sin6_port == msad6->sin6_port &&
5925 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5926 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5927 				*dspp = tdsp;
5928 				NFSUNLOCKMNT(nmp);
5929 				return (0);
5930 			}
5931 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5932 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5933 				msad6 = (struct sockaddr_in6 *)
5934 				    tdsp->nfsclds_sockp->nr_nam;
5935 			else
5936 				msad6 = NULL;
5937 		}
5938 		NFSUNLOCKMNT(nmp);
5939 
5940 		/* No IP address match, so look for new/trunked one. */
5941 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5942 		sad6->sin6_len = sizeof(*sad6);
5943 		sad6->sin6_family = AF_INET6;
5944 		sad6->sin6_port = sin6->sin6_port;
5945 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5946 		    sizeof(struct in6_addr));
5947 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5948 		nrp->nr_nam = (struct sockaddr *)sad6;
5949 	} else
5950 		return (EPERM);
5951 
5952 	nrp->nr_sotype = SOCK_STREAM;
5953 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5954 	nrp->nr_prog = NFS_PROG;
5955 	nrp->nr_vers = vers;
5956 
5957 	/*
5958 	 * Use the credentials that were used for the mount, which are
5959 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5960 	 * Ref. counting the credentials with crhold() is probably not
5961 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5962 	 * unmount, but I did it anyhow.
5963 	 */
5964 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5965 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false, &nrp->nr_client);
5966 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5967 
5968 	dsp = NULL;
5969 	/* Now, do the exchangeid and create session. */
5970 	if (error == 0) {
5971 		if (vers == NFS_VER4) {
5972 			firsttry = 0;
5973 			do {
5974 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5975 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5976 				    nrp->nr_cred, p);
5977 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5978 				if (error == NFSERR_MINORVERMISMATCH)
5979 					minorvers = NFSV42_MINORVERSION;
5980 			} while (error == NFSERR_MINORVERMISMATCH &&
5981 			    firsttry++ == 0);
5982 			if (error != 0)
5983 				newnfs_disconnect(NULL, nrp);
5984 		} else {
5985 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5986 			    M_WAITOK | M_ZERO);
5987 			dsp->nfsclds_flags |= NFSCLDS_DS;
5988 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5989 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5990 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5991 			    NULL, MTX_DEF);
5992 		}
5993 	}
5994 	if (error == 0) {
5995 		dsp->nfsclds_sockp = nrp;
5996 		if (vers == NFS_VER4) {
5997 			NFSLOCKMNT(nmp);
5998 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5999 			    &sequenceid);
6000 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
6001 			if (retv == NFSDSP_USETHISSESSION &&
6002 			    nfscl_dssameconn != 0) {
6003 				NFSLOCKDS(tdsp);
6004 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
6005 				NFSUNLOCKDS(tdsp);
6006 				NFSUNLOCKMNT(nmp);
6007 				/*
6008 				 * If there is already a session for this
6009 				 * server, use it.
6010 				 */
6011 				newnfs_disconnect(NULL, nrp);
6012 				nfscl_freenfsclds(dsp);
6013 				*dspp = tdsp;
6014 				return (0);
6015 			}
6016 			if (retv == NFSDSP_NOTFOUND)
6017 				sequenceid =
6018 				    dsp->nfsclds_sess.nfsess_sequenceid;
6019 			NFSUNLOCKMNT(nmp);
6020 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
6021 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
6022 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
6023 		}
6024 	} else {
6025 		NFSFREECRED(nrp->nr_cred);
6026 		NFSFREEMUTEX(&nrp->nr_mtx);
6027 		free(nrp->nr_nam, M_SONAME);
6028 		free(nrp, M_NFSSOCKREQ);
6029 	}
6030 	if (error == 0) {
6031 		NFSCL_DEBUG(3, "add DS session\n");
6032 		/*
6033 		 * Put it at the end of the list. That way the list
6034 		 * is ordered by when the entry was added. This matters
6035 		 * since the one done first is the one that should be
6036 		 * used for sequencid'ing any subsequent create sessions.
6037 		 */
6038 		NFSLOCKMNT(nmp);
6039 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
6040 		NFSUNLOCKMNT(nmp);
6041 		*dspp = dsp;
6042 	} else if (dsp != NULL) {
6043 		newnfs_disconnect(NULL, nrp);
6044 		nfscl_freenfsclds(dsp);
6045 	}
6046 	return (error);
6047 }
6048 
6049 /*
6050  * Do the NFSv4.1 Reclaim Complete.
6051  */
6052 int
6053 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
6054 {
6055 	uint32_t *tl;
6056 	struct nfsrv_descript nfsd;
6057 	struct nfsrv_descript *nd = &nfsd;
6058 	int error;
6059 
6060 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
6061 	    0, cred);
6062 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
6063 	*tl = newnfs_false;
6064 	nd->nd_flag |= ND_USEGSSNAME;
6065 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
6066 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
6067 	if (error != 0)
6068 		return (error);
6069 	error = nd->nd_repstat;
6070 	m_freem(nd->nd_mrep);
6071 	return (error);
6072 }
6073 
6074 /*
6075  * Initialize the slot tables for a session.
6076  */
6077 static void
6078 nfscl_initsessionslots(struct nfsclsession *sep)
6079 {
6080 	int i;
6081 
6082 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6083 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
6084 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
6085 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
6086 	}
6087 	for (i = 0; i < 64; i++)
6088 		sep->nfsess_slotseq[i] = 0;
6089 	sep->nfsess_slots = 0;
6090 	sep->nfsess_badslots = 0;
6091 }
6092 
6093 /*
6094  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
6095  */
6096 int
6097 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6098     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
6099 {
6100 	struct nfsnode *np = VTONFS(vp);
6101 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6102 	struct nfscllayout *layp;
6103 	struct nfscldevinfo *dip;
6104 	struct nfsclflayout *rflp;
6105 	struct mbuf *m, *m2;
6106 	struct nfsclwritedsdorpc *drpc, *tdrpc;
6107 	nfsv4stateid_t stateid;
6108 	struct ucred *newcred;
6109 	uint64_t lastbyte, len, off, oresid, xfer;
6110 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
6111 	void *lckp;
6112 	uint8_t *dev;
6113 	void *iovbase = NULL;
6114 	size_t iovlen = 0;
6115 	off_t offs = 0;
6116 	ssize_t resid = 0;
6117 	uint32_t op;
6118 
6119 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
6120 	    (np->n_flag & NNOLAYOUT) != 0)
6121 		return (EIO);
6122 	/* Now, get a reference cnt on the clientid for this mount. */
6123 	if (nfscl_getref(nmp) == 0)
6124 		return (EIO);
6125 
6126 	/* Find an appropriate stateid. */
6127 	newcred = NFSNEWCRED(cred);
6128 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
6129 	    rwaccess, 1, newcred, p, &stateid, &lckp);
6130 	if (error != 0) {
6131 		NFSFREECRED(newcred);
6132 		nfscl_relref(nmp);
6133 		return (error);
6134 	}
6135 	/* Search for a layout for this file. */
6136 	off = uiop->uio_offset;
6137 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
6138 	    np->n_fhp->nfh_len, off, rwaccess, &rflp, &recalled);
6139 	if (layp == NULL || rflp == NULL) {
6140 		if (recalled != 0) {
6141 			NFSFREECRED(newcred);
6142 			if (lckp != NULL)
6143 				nfscl_lockderef(lckp);
6144 			nfscl_relref(nmp);
6145 			return (EIO);
6146 		}
6147 		if (layp != NULL) {
6148 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
6149 			layp = NULL;
6150 		}
6151 		/* Try and get a Layout, if it is supported. */
6152 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
6153 		    (np->n_flag & NWRITEOPENED) != 0)
6154 			iolaymode = NFSLAYOUTIOMODE_RW;
6155 		else
6156 			iolaymode = NFSLAYOUTIOMODE_READ;
6157 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
6158 		    rwaccess, NULL, &stateid, off, &layp, newcred, p);
6159 		if (error != 0) {
6160 			NFSLOCKNODE(np);
6161 			np->n_flag |= NNOLAYOUT;
6162 			NFSUNLOCKNODE(np);
6163 			if (lckp != NULL)
6164 				nfscl_lockderef(lckp);
6165 			NFSFREECRED(newcred);
6166 			if (layp != NULL)
6167 				nfscl_rellayout(layp, 0);
6168 			nfscl_relref(nmp);
6169 			return (error);
6170 		}
6171 	}
6172 
6173 	/*
6174 	 * Loop around finding a layout that works for the first part of
6175 	 * this I/O operation, and then call the function that actually
6176 	 * does the RPC.
6177 	 */
6178 	eof = 0;
6179 	len = (uint64_t)uiop->uio_resid;
6180 	while (len > 0 && error == 0 && eof == 0) {
6181 		off = uiop->uio_offset;
6182 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
6183 		if (error == 0) {
6184 			oresid = xfer = (uint64_t)uiop->uio_resid;
6185 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
6186 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
6187 			/*
6188 			 * For Flex File layout with mirrored DSs, select one
6189 			 * of them at random for reads. For writes and commits,
6190 			 * do all mirrors.
6191 			 */
6192 			m = NULL;
6193 			tdrpc = drpc = NULL;
6194 			firstmirror = 0;
6195 			mirrorcnt = 1;
6196 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
6197 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
6198 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
6199 					firstmirror = arc4random() % mirrorcnt;
6200 					mirrorcnt = firstmirror + 1;
6201 				} else {
6202 					if (docommit == 0) {
6203 						/*
6204 						 * Save values, so uiop can be
6205 						 * rolled back upon a write
6206 						 * error.
6207 						 */
6208 						offs = uiop->uio_offset;
6209 						resid = uiop->uio_resid;
6210 						iovbase =
6211 						    uiop->uio_iov->iov_base;
6212 						iovlen = uiop->uio_iov->iov_len;
6213 						m = nfsm_uiombuflist(uiop, len,
6214 						    0);
6215 					}
6216 					tdrpc = drpc = malloc(sizeof(*drpc) *
6217 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
6218 					    M_ZERO);
6219 				}
6220 			}
6221 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
6222 				m2 = NULL;
6223 				if (m != NULL && i < mirrorcnt - 1)
6224 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
6225 				else {
6226 					m2 = m;
6227 					m = NULL;
6228 				}
6229 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
6230 					dev = rflp->nfsfl_ffm[i].dev;
6231 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6232 					    rflp->nfsfl_ffm[i].devp);
6233 				} else {
6234 					dev = rflp->nfsfl_dev;
6235 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
6236 					    rflp->nfsfl_devp);
6237 				}
6238 				if (dip != NULL) {
6239 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
6240 					    != 0)
6241 						error = nfscl_dofflayoutio(vp,
6242 						    uiop, iomode, must_commit,
6243 						    &eof, &stateid, rwaccess,
6244 						    dip, layp, rflp, off, xfer,
6245 						    i, docommit, m2, tdrpc,
6246 						    newcred, p);
6247 					else
6248 						error = nfscl_doflayoutio(vp,
6249 						    uiop, iomode, must_commit,
6250 						    &eof, &stateid, rwaccess,
6251 						    dip, layp, rflp, off, xfer,
6252 						    docommit, newcred, p);
6253 					nfscl_reldevinfo(dip);
6254 				} else {
6255 					if (m2 != NULL)
6256 						m_freem(m2);
6257 					error = EIO;
6258 				}
6259 				tdrpc++;
6260 			}
6261 			if (m != NULL)
6262 				m_freem(m);
6263 			tdrpc = drpc;
6264 			timo = hz / 50;		/* Wait for 20msec. */
6265 			if (timo < 1)
6266 				timo = 1;
6267 			for (i = firstmirror; i < mirrorcnt - 1 &&
6268 			    tdrpc != NULL; i++, tdrpc++) {
6269 				/*
6270 				 * For the unused drpc entries, both inprog and
6271 				 * err == 0, so this loop won't break.
6272 				 */
6273 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
6274 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
6275 					    timo);
6276 				if (error == 0 && tdrpc->err != 0)
6277 					error = tdrpc->err;
6278 				if (rwaccess != NFSV4OPEN_ACCESSREAD &&
6279 				    docommit == 0 && *must_commit == 0 &&
6280 				    tdrpc->must_commit == 1)
6281 					*must_commit = 1;
6282 			}
6283 			free(drpc, M_TEMP);
6284 			if (error == 0) {
6285 				if (mirrorcnt > 1 && rwaccess ==
6286 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6287 					NFSLOCKCLSTATE();
6288 					layp->nfsly_flags |= NFSLY_WRITTEN;
6289 					NFSUNLOCKCLSTATE();
6290 				}
6291 				lastbyte = off + xfer - 1;
6292 				NFSLOCKCLSTATE();
6293 				if (lastbyte > layp->nfsly_lastbyte)
6294 					layp->nfsly_lastbyte = lastbyte;
6295 				NFSUNLOCKCLSTATE();
6296 			} else if (error == NFSERR_OPENMODE &&
6297 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
6298 				NFSLOCKMNT(nmp);
6299 				nmp->nm_state |= NFSSTA_OPENMODE;
6300 				NFSUNLOCKMNT(nmp);
6301 			} else if ((error == NFSERR_NOSPC ||
6302 			    error == NFSERR_IO || error == NFSERR_NXIO) &&
6303 			    nmp->nm_minorvers == NFSV42_MINORVERSION) {
6304 				if (docommit != 0)
6305 					op = NFSV4OP_COMMIT;
6306 				else if (rwaccess == NFSV4OPEN_ACCESSREAD)
6307 					op = NFSV4OP_READ;
6308 				else
6309 					op = NFSV4OP_WRITE;
6310 				nfsrpc_layouterror(nmp, np->n_fhp->nfh_fh,
6311 				    np->n_fhp->nfh_len, off, xfer,
6312 				    &layp->nfsly_stateid, newcred, p, error, op,
6313 				    dip->nfsdi_deviceid);
6314 				error = EIO;
6315 			} else
6316 				error = EIO;
6317 			if (error == 0)
6318 				len -= (oresid - (uint64_t)uiop->uio_resid);
6319 			else if (mirrorcnt > 1 && rwaccess ==
6320 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
6321 				/*
6322 				 * In case the rpc gets retried, roll the
6323 				 * uio fields changed by nfsm_uiombuflist()
6324 				 * back.
6325 				 */
6326 				uiop->uio_offset = offs;
6327 				uiop->uio_resid = resid;
6328 				uiop->uio_iov->iov_base = iovbase;
6329 				uiop->uio_iov->iov_len = iovlen;
6330 			}
6331 		}
6332 	}
6333 	if (lckp != NULL)
6334 		nfscl_lockderef(lckp);
6335 	NFSFREECRED(newcred);
6336 	nfscl_rellayout(layp, 0);
6337 	nfscl_relref(nmp);
6338 	return (error);
6339 }
6340 
6341 /*
6342  * Find a file layout that will handle the first bytes of the requested
6343  * range and return the information from it needed to the I/O operation.
6344  */
6345 int
6346 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
6347     struct nfsclflayout **retflpp)
6348 {
6349 	struct nfsclflayout *flp, *nflp, *rflp;
6350 	uint32_t rw;
6351 
6352 	rflp = NULL;
6353 	rw = rwaccess;
6354 	/* For reading, do the Read list first and then the Write list. */
6355 	do {
6356 		if (rw == NFSV4OPEN_ACCESSREAD)
6357 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6358 		else
6359 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6360 		while (flp != NULL) {
6361 			nflp = LIST_NEXT(flp, nfsfl_list);
6362 			if (flp->nfsfl_off > off)
6363 				break;
6364 			if (flp->nfsfl_end > off &&
6365 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6366 				rflp = flp;
6367 			flp = nflp;
6368 		}
6369 		if (rw == NFSV4OPEN_ACCESSREAD)
6370 			rw = NFSV4OPEN_ACCESSWRITE;
6371 		else
6372 			rw = 0;
6373 	} while (rw != 0);
6374 	if (rflp != NULL) {
6375 		/* This one covers the most bytes starting at off. */
6376 		*retflpp = rflp;
6377 		return (0);
6378 	}
6379 	return (EIO);
6380 }
6381 
6382 /*
6383  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6384  */
6385 static int
6386 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6387     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6388     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6389     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6390 {
6391 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6392 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6393 	struct nfsnode *np;
6394 	struct nfsfh *fhp;
6395 	struct nfsclds **dspp;
6396 
6397 	np = VTONFS(vp);
6398 	rel_off = off - flp->nfsfl_patoff;
6399 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6400 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6401 	    dp->nfsdi_stripecnt;
6402 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6403 	error = 0;
6404 
6405 	/* Loop around, doing I/O for each stripe unit. */
6406 	while (len > 0 && error == 0) {
6407 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6408 		dspp = nfsfldi_addr(dp, stripe_index);
6409 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6410 			minorvers = NFSV42_MINORVERSION;
6411 		else
6412 			minorvers = NFSV41_MINORVERSION;
6413 		if (len > transfer && docommit == 0)
6414 			xfer = transfer;
6415 		else
6416 			xfer = len;
6417 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6418 			/* Dense layout. */
6419 			if (stripe_pos >= flp->nfsfl_fhcnt)
6420 				return (EIO);
6421 			fhp = flp->nfsfl_fh[stripe_pos];
6422 			io_off = (rel_off / (stripe_unit_size *
6423 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6424 			    rel_off % stripe_unit_size;
6425 		} else {
6426 			/* Sparse layout. */
6427 			if (flp->nfsfl_fhcnt > 1) {
6428 				if (stripe_index >= flp->nfsfl_fhcnt)
6429 					return (EIO);
6430 				fhp = flp->nfsfl_fh[stripe_index];
6431 			} else if (flp->nfsfl_fhcnt == 1)
6432 				fhp = flp->nfsfl_fh[0];
6433 			else
6434 				fhp = np->n_fhp;
6435 			io_off = off;
6436 		}
6437 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6438 			commit_thru_mds = 1;
6439 			if (docommit != 0)
6440 				error = EIO;
6441 		} else {
6442 			commit_thru_mds = 0;
6443 			NFSLOCKNODE(np);
6444 			np->n_flag |= NDSCOMMIT;
6445 			NFSUNLOCKNODE(np);
6446 		}
6447 		if (docommit != 0) {
6448 			if (error == 0)
6449 				error = nfsrpc_commitds(vp, io_off, xfer,
6450 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6451 			if (error == 0) {
6452 				/*
6453 				 * Set both eof and uio_resid = 0 to end any
6454 				 * loops.
6455 				 */
6456 				*eofp = 1;
6457 				uiop->uio_resid = 0;
6458 			} else {
6459 				NFSLOCKNODE(np);
6460 				np->n_flag &= ~NDSCOMMIT;
6461 				NFSUNLOCKNODE(np);
6462 			}
6463 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6464 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6465 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6466 		else {
6467 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6468 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6469 			    0, NFS_VER4, minorvers, cred, p);
6470 			if (error == 0) {
6471 				NFSLOCKCLSTATE();
6472 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6473 				NFSUNLOCKCLSTATE();
6474 			}
6475 		}
6476 		if (error == 0) {
6477 			transfer = stripe_unit_size;
6478 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6479 			len -= xfer;
6480 			off += xfer;
6481 		}
6482 	}
6483 	return (error);
6484 }
6485 
6486 /*
6487  * Do I/O using an NFSv4.1 flex file layout.
6488  */
6489 static int
6490 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6491     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6492     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6493     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6494     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6495 {
6496 	uint64_t xfer;
6497 	int error;
6498 	struct nfsnode *np;
6499 	struct nfsfh *fhp;
6500 	struct nfsclds **dspp;
6501 	struct ucred *tcred;
6502 	struct mbuf *m, *m2;
6503 	uint32_t copylen;
6504 
6505 	np = VTONFS(vp);
6506 	error = 0;
6507 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6508 	    (uintmax_t)len);
6509 	/* Loop around, doing I/O for each stripe unit. */
6510 	while (len > 0 && error == 0) {
6511 		dspp = nfsfldi_addr(dp, 0);
6512 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6513 		stateidp = &flp->nfsfl_ffm[mirror].st;
6514 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6515 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6516 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6517 			tcred = NFSNEWCRED(cred);
6518 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6519 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6520 			tcred->cr_ngroups = 1;
6521 		} else
6522 			tcred = cred;
6523 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6524 			copylen = dp->nfsdi_rsize;
6525 		else {
6526 			copylen = dp->nfsdi_wsize;
6527 			if (len > copylen && mp != NULL) {
6528 				/*
6529 				 * When a mirrored configuration needs to do
6530 				 * multiple writes to each mirror, all writes
6531 				 * except the last one must be a multiple of
6532 				 * 4 bytes.  This is required so that the XDR
6533 				 * does not need padding.
6534 				 * If possible, clip the size to an exact
6535 				 * multiple of the mbuf length, so that the
6536 				 * split will be on an mbuf boundary.
6537 				 */
6538 				copylen &= 0xfffffffc;
6539 				if (copylen > mp->m_len)
6540 					copylen = copylen / mp->m_len *
6541 					    mp->m_len;
6542 			}
6543 		}
6544 		NFSLOCKNODE(np);
6545 		np->n_flag |= NDSCOMMIT;
6546 		NFSUNLOCKNODE(np);
6547 		if (len > copylen && docommit == 0)
6548 			xfer = copylen;
6549 		else
6550 			xfer = len;
6551 		if (docommit != 0) {
6552 			if (error == 0) {
6553 				/*
6554 				 * Do last mirrored DS commit with this thread.
6555 				 */
6556 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6557 					error = nfsio_commitds(vp, off, xfer,
6558 					    *dspp, fhp, dp->nfsdi_vers,
6559 					    dp->nfsdi_minorvers, drpc, tcred,
6560 					    p);
6561 				else
6562 					error = nfsrpc_commitds(vp, off, xfer,
6563 					    *dspp, fhp, dp->nfsdi_vers,
6564 					    dp->nfsdi_minorvers, tcred, p);
6565 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6566 				if (error != 0 && error != EACCES && error !=
6567 				    ESTALE) {
6568 					NFSCL_DEBUG(4,
6569 					    "DS layreterr for commit\n");
6570 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6571 					    lyp, *dspp);
6572 				}
6573 			}
6574 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6575 			if (error == 0) {
6576 				/*
6577 				 * Set both eof and uio_resid = 0 to end any
6578 				 * loops.
6579 				 */
6580 				*eofp = 1;
6581 				uiop->uio_resid = 0;
6582 			} else {
6583 				NFSLOCKNODE(np);
6584 				np->n_flag &= ~NDSCOMMIT;
6585 				NFSUNLOCKNODE(np);
6586 			}
6587 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6588 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6589 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6590 			    dp->nfsdi_minorvers, tcred, p);
6591 			NFSCL_DEBUG(4, "readds=%d\n", error);
6592 			if (error != 0 && error != EACCES && error != ESTALE) {
6593 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6594 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6595 				    *dspp);
6596 			}
6597 		} else {
6598 			if (flp->nfsfl_mirrorcnt == 1) {
6599 				error = nfsrpc_writeds(vp, uiop, iomode,
6600 				    must_commit, stateidp, *dspp, off, xfer,
6601 				    fhp, 0, 1, dp->nfsdi_vers,
6602 				    dp->nfsdi_minorvers, tcred, p);
6603 				if (error == 0) {
6604 					NFSLOCKCLSTATE();
6605 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6606 					NFSUNLOCKCLSTATE();
6607 				}
6608 			} else {
6609 				m = mp;
6610 				if (xfer < len) {
6611 					/* The mbuf list must be split. */
6612 					m2 = nfsm_split(mp, xfer);
6613 					if (m2 != NULL)
6614 						mp = m2;
6615 					else {
6616 						m_freem(mp);
6617 						error = EIO;
6618 					}
6619 				}
6620 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6621 				    (uintmax_t)len, (uintmax_t)xfer);
6622 				/*
6623 				 * Do last write to a mirrored DS with this
6624 				 * thread.
6625 				 */
6626 				if (error == 0) {
6627 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6628 						error = nfsio_writedsmir(vp,
6629 						    iomode, must_commit,
6630 						    stateidp, *dspp, off,
6631 						    xfer, fhp, m,
6632 						    dp->nfsdi_vers,
6633 						    dp->nfsdi_minorvers, drpc,
6634 						    tcred, p);
6635 					else
6636 						error = nfsrpc_writedsmir(vp,
6637 						    iomode, must_commit,
6638 						    stateidp, *dspp, off,
6639 						    xfer, fhp, m,
6640 						    dp->nfsdi_vers,
6641 						    dp->nfsdi_minorvers, tcred,
6642 						    p);
6643 				}
6644 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6645 				if (error != 0 && error != EACCES && error !=
6646 				    ESTALE) {
6647 					NFSCL_DEBUG(4,
6648 					    "DS layreterr for write\n");
6649 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6650 					    lyp, *dspp);
6651 				}
6652 			}
6653 		}
6654 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6655 		if (error == 0) {
6656 			len -= xfer;
6657 			off += xfer;
6658 		}
6659 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6660 			NFSFREECRED(tcred);
6661 	}
6662 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6663 	return (error);
6664 }
6665 
6666 /*
6667  * The actual read RPC done to a DS.
6668  */
6669 static int
6670 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6671     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6672     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6673 {
6674 	uint32_t *tl;
6675 	int attrflag, error, retlen;
6676 	struct nfsrv_descript nfsd;
6677 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6678 	struct nfsrv_descript *nd = &nfsd;
6679 	struct nfssockreq *nrp;
6680 	struct nfsvattr na;
6681 
6682 	nd->nd_mrep = NULL;
6683 	if (vers == 0 || vers == NFS_VER4) {
6684 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6685 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6686 		    NULL);
6687 		vers = NFS_VER4;
6688 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6689 		if (flex != 0)
6690 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6691 		else
6692 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6693 	} else {
6694 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6695 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6696 		    NULL);
6697 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6698 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6699 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6700 	}
6701 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6702 	txdr_hyper(io_off, tl);
6703 	*(tl + 2) = txdr_unsigned(len);
6704 	nrp = dsp->nfsclds_sockp;
6705 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6706 	if (nrp == NULL)
6707 		/* If NULL, use the MDS socket. */
6708 		nrp = &nmp->nm_sockreq;
6709 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6710 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6711 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6712 	    error);
6713 	if (error != 0)
6714 		return (error);
6715 	if (vers == NFS_VER3) {
6716 		error = nfscl_postop_attr(nd, &na, &attrflag);
6717 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6718 		if (error != 0)
6719 			goto nfsmout;
6720 	}
6721 	if (nd->nd_repstat != 0) {
6722 		error = nd->nd_repstat;
6723 		goto nfsmout;
6724 	}
6725 	if (vers == NFS_VER3) {
6726 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6727 		*eofp = fxdr_unsigned(int, *(tl + 1));
6728 	} else {
6729 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6730 		*eofp = fxdr_unsigned(int, *tl);
6731 	}
6732 	NFSM_STRSIZ(retlen, len);
6733 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6734 	error = nfsm_mbufuio(nd, uiop, retlen);
6735 nfsmout:
6736 	if (nd->nd_mrep != NULL)
6737 		m_freem(nd->nd_mrep);
6738 	return (error);
6739 }
6740 
6741 /*
6742  * The actual write RPC done to a DS.
6743  */
6744 static int
6745 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6746     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6747     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6748     struct ucred *cred, NFSPROC_T *p)
6749 {
6750 	uint32_t *tl;
6751 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6752 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6753 	int32_t backup;
6754 	struct nfsrv_descript nfsd;
6755 	struct nfsrv_descript *nd = &nfsd;
6756 	struct nfssockreq *nrp;
6757 	struct nfsvattr na;
6758 
6759 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6760 	nd->nd_mrep = NULL;
6761 	if (vers == 0 || vers == NFS_VER4) {
6762 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6763 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6764 		    NULL);
6765 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6766 		vers = NFS_VER4;
6767 		if (flex != 0)
6768 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6769 		else
6770 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6771 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6772 	} else {
6773 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6774 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6775 		    NULL);
6776 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6777 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6778 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6779 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6780 	}
6781 	txdr_hyper(io_off, tl);
6782 	tl += 2;
6783 	if (vers == NFS_VER3)
6784 		*tl++ = txdr_unsigned(len);
6785 	*tl++ = txdr_unsigned(*iomode);
6786 	*tl = txdr_unsigned(len);
6787 	nfsm_uiombuf(nd, uiop, len);
6788 	nrp = dsp->nfsclds_sockp;
6789 	if (nrp == NULL)
6790 		/* If NULL, use the MDS socket. */
6791 		nrp = &nmp->nm_sockreq;
6792 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6793 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6794 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6795 	    nd->nd_repstat);
6796 	if (error != 0)
6797 		return (error);
6798 	if (nd->nd_repstat != 0) {
6799 		/*
6800 		 * In case the rpc gets retried, roll
6801 		 * the uio fields changed by nfsm_uiombuf()
6802 		 * back.
6803 		 */
6804 		uiop->uio_offset -= len;
6805 		uiop->uio_resid += len;
6806 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6807 		uiop->uio_iov->iov_len += len;
6808 		error = nd->nd_repstat;
6809 	} else {
6810 		if (vers == NFS_VER3) {
6811 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6812 			    NULL);
6813 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6814 			if (error != 0)
6815 				goto nfsmout;
6816 		}
6817 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6818 		rlen = fxdr_unsigned(int, *tl++);
6819 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6820 		if (rlen == 0) {
6821 			error = NFSERR_IO;
6822 			goto nfsmout;
6823 		} else if (rlen < len) {
6824 			backup = len - rlen;
6825 			uiop->uio_iov->iov_base =
6826 			    (char *)uiop->uio_iov->iov_base - backup;
6827 			uiop->uio_iov->iov_len += backup;
6828 			uiop->uio_offset -= backup;
6829 			uiop->uio_resid += backup;
6830 			len = rlen;
6831 		}
6832 		commit = fxdr_unsigned(int, *tl++);
6833 
6834 		/*
6835 		 * Return the lowest commitment level
6836 		 * obtained by any of the RPCs.
6837 		 */
6838 		if (committed == NFSWRITE_FILESYNC)
6839 			committed = commit;
6840 		else if (committed == NFSWRITE_DATASYNC &&
6841 		    commit == NFSWRITE_UNSTABLE)
6842 			committed = commit;
6843 		if (commit_thru_mds != 0) {
6844 			NFSLOCKMNT(nmp);
6845 			if (!NFSHASWRITEVERF(nmp)) {
6846 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6847 				NFSSETWRITEVERF(nmp);
6848 			} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF) &&
6849 			    *must_commit != 2) {
6850 				*must_commit = 1;
6851 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6852 			}
6853 			NFSUNLOCKMNT(nmp);
6854 		} else {
6855 			NFSLOCKDS(dsp);
6856 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6857 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6858 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6859 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
6860 			    *must_commit != 2) {
6861 				*must_commit = 1;
6862 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6863 			}
6864 			NFSUNLOCKDS(dsp);
6865 		}
6866 	}
6867 nfsmout:
6868 	if (nd->nd_mrep != NULL)
6869 		m_freem(nd->nd_mrep);
6870 	*iomode = committed;
6871 	if (nd->nd_repstat != 0 && error == 0)
6872 		error = nd->nd_repstat;
6873 	return (error);
6874 }
6875 
6876 /*
6877  * The actual write RPC done to a DS.
6878  * This variant is called from a separate kernel process for mirrors.
6879  * Any short write is considered an IO error.
6880  */
6881 static int
6882 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6883     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6884     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6885     struct ucred *cred, NFSPROC_T *p)
6886 {
6887 	uint32_t *tl;
6888 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6889 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6890 	struct nfsrv_descript nfsd;
6891 	struct nfsrv_descript *nd = &nfsd;
6892 	struct nfssockreq *nrp;
6893 	struct nfsvattr na;
6894 
6895 	nd->nd_mrep = NULL;
6896 	if (vers == 0 || vers == NFS_VER4) {
6897 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6898 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6899 		    NULL);
6900 		vers = NFS_VER4;
6901 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6902 		    minorvers);
6903 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6904 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6905 	} else {
6906 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6907 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
6908 		    NULL);
6909 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6910 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6911 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6912 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6913 	}
6914 	txdr_hyper(io_off, tl);
6915 	tl += 2;
6916 	if (vers == NFS_VER3)
6917 		*tl++ = txdr_unsigned(len);
6918 	*tl++ = txdr_unsigned(*iomode);
6919 	*tl = txdr_unsigned(len);
6920 	if (len > 0) {
6921 		/* Put data in mbuf chain. */
6922 		nd->nd_mb->m_next = m;
6923 	}
6924 	nrp = dsp->nfsclds_sockp;
6925 	if (nrp == NULL)
6926 		/* If NULL, use the MDS socket. */
6927 		nrp = &nmp->nm_sockreq;
6928 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6929 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6930 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6931 	    nd->nd_repstat);
6932 	if (error != 0)
6933 		return (error);
6934 	if (nd->nd_repstat != 0)
6935 		error = nd->nd_repstat;
6936 	else {
6937 		if (vers == NFS_VER3) {
6938 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6939 			    NULL);
6940 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6941 			    error);
6942 			if (error != 0)
6943 				goto nfsmout;
6944 		}
6945 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6946 		rlen = fxdr_unsigned(int, *tl++);
6947 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6948 		    rlen);
6949 		if (rlen != len) {
6950 			error = NFSERR_IO;
6951 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6952 			    len, rlen);
6953 			goto nfsmout;
6954 		}
6955 		commit = fxdr_unsigned(int, *tl++);
6956 
6957 		/*
6958 		 * Return the lowest commitment level
6959 		 * obtained by any of the RPCs.
6960 		 */
6961 		if (committed == NFSWRITE_FILESYNC)
6962 			committed = commit;
6963 		else if (committed == NFSWRITE_DATASYNC &&
6964 		    commit == NFSWRITE_UNSTABLE)
6965 			committed = commit;
6966 		NFSLOCKDS(dsp);
6967 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6968 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6969 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6970 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF) &&
6971 		    *must_commit != 2) {
6972 			*must_commit = 1;
6973 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6974 		}
6975 		NFSUNLOCKDS(dsp);
6976 	}
6977 nfsmout:
6978 	if (nd->nd_mrep != NULL)
6979 		m_freem(nd->nd_mrep);
6980 	*iomode = committed;
6981 	if (nd->nd_repstat != 0 && error == 0)
6982 		error = nd->nd_repstat;
6983 	return (error);
6984 }
6985 
6986 /*
6987  * Start up the thread that will execute nfsrpc_writedsmir().
6988  */
6989 static void
6990 start_writedsmir(void *arg, int pending)
6991 {
6992 	struct nfsclwritedsdorpc *drpc;
6993 
6994 	drpc = (struct nfsclwritedsdorpc *)arg;
6995 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6996 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6997 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6998 	    drpc->p);
6999 	drpc->done = 1;
7000 	crfree(drpc->cred);
7001 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
7002 }
7003 
7004 /*
7005  * Set up the write DS mirror call for the pNFS I/O thread.
7006  */
7007 static int
7008 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
7009     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
7010     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
7011     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7012 {
7013 	int error, ret;
7014 
7015 	error = 0;
7016 	drpc->done = 0;
7017 	drpc->vp = vp;
7018 	drpc->iomode = *iomode;
7019 	drpc->must_commit = *must_commit;
7020 	drpc->stateidp = stateidp;
7021 	drpc->dsp = dsp;
7022 	drpc->off = off;
7023 	drpc->len = len;
7024 	drpc->fhp = fhp;
7025 	drpc->m = m;
7026 	drpc->vers = vers;
7027 	drpc->minorvers = minorvers;
7028 	drpc->cred = crhold(cred);
7029 	drpc->p = p;
7030 	drpc->inprog = 0;
7031 	ret = EIO;
7032 	if (nfs_pnfsiothreads != 0) {
7033 		ret = nfs_pnfsio(start_writedsmir, drpc);
7034 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
7035 	}
7036 	if (ret != 0) {
7037 		error = nfsrpc_writedsmir(vp, iomode, &drpc->must_commit,
7038 		    stateidp, dsp, off, len, fhp, m, vers, minorvers, cred, p);
7039 		crfree(drpc->cred);
7040 	}
7041 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
7042 	return (error);
7043 }
7044 
7045 /*
7046  * Free up the nfsclds structure.
7047  */
7048 void
7049 nfscl_freenfsclds(struct nfsclds *dsp)
7050 {
7051 	int i;
7052 
7053 	if (dsp == NULL)
7054 		return;
7055 	if (dsp->nfsclds_sockp != NULL) {
7056 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
7057 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
7058 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
7059 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
7060 	}
7061 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
7062 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
7063 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
7064 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
7065 			m_freem(
7066 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
7067 	}
7068 	free(dsp, M_NFSCLDS);
7069 }
7070 
7071 static enum nfsclds_state
7072 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
7073     struct nfsclds **retdspp, uint32_t *sequencep)
7074 {
7075 	struct nfsclds *dsp;
7076 	int fndseq;
7077 
7078 	/*
7079 	 * Search the list of nfsclds structures for one with the same
7080 	 * server.
7081 	 */
7082 	fndseq = 0;
7083 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
7084 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
7085 		    dsp->nfsclds_servownlen != 0 &&
7086 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
7087 		    dsp->nfsclds_servownlen) &&
7088 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
7089 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
7090 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
7091 			    dsp->nfsclds_flags);
7092 			if (fndseq == 0) {
7093 				/* Get sequenceid# from first entry. */
7094 				*sequencep =
7095 				    dsp->nfsclds_sess.nfsess_sequenceid;
7096 				fndseq = 1;
7097 			}
7098 			/* Server major id matches. */
7099 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
7100 				*retdspp = dsp;
7101 				return (NFSDSP_USETHISSESSION);
7102 			}
7103 		}
7104 	}
7105 	if (fndseq != 0)
7106 		return (NFSDSP_SEQTHISSESSION);
7107 	return (NFSDSP_NOTFOUND);
7108 }
7109 
7110 /*
7111  * NFS commit rpc to a NFSv4.1 DS.
7112  */
7113 static int
7114 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7115     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
7116     NFSPROC_T *p)
7117 {
7118 	uint32_t *tl;
7119 	struct nfsrv_descript nfsd, *nd = &nfsd;
7120 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7121 	struct nfssockreq *nrp;
7122 	struct nfsvattr na;
7123 	int attrflag, error;
7124 
7125 	nd->nd_mrep = NULL;
7126 	if (vers == 0 || vers == NFS_VER4) {
7127 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
7128 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7129 		    NULL);
7130 		vers = NFS_VER4;
7131 	} else {
7132 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
7133 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers,
7134 		    NULL);
7135 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
7136 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
7137 	}
7138 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
7139 	    minorvers);
7140 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7141 	txdr_hyper(offset, tl);
7142 	tl += 2;
7143 	*tl = txdr_unsigned(cnt);
7144 	nrp = dsp->nfsclds_sockp;
7145 	if (nrp == NULL)
7146 		/* If NULL, use the MDS socket. */
7147 		nrp = &nmp->nm_sockreq;
7148 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7149 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7150 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
7151 	    nd->nd_repstat);
7152 	if (error != 0)
7153 		return (error);
7154 	if (nd->nd_repstat == 0) {
7155 		if (vers == NFS_VER3) {
7156 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
7157 			    NULL);
7158 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
7159 			if (error != 0)
7160 				goto nfsmout;
7161 		}
7162 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
7163 		NFSLOCKDS(dsp);
7164 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
7165 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
7166 			error = NFSERR_STALEWRITEVERF;
7167 		}
7168 		NFSUNLOCKDS(dsp);
7169 	}
7170 nfsmout:
7171 	if (error == 0 && nd->nd_repstat != 0)
7172 		error = nd->nd_repstat;
7173 	m_freem(nd->nd_mrep);
7174 	return (error);
7175 }
7176 
7177 /*
7178  * Start up the thread that will execute nfsrpc_commitds().
7179  */
7180 static void
7181 start_commitds(void *arg, int pending)
7182 {
7183 	struct nfsclwritedsdorpc *drpc;
7184 
7185 	drpc = (struct nfsclwritedsdorpc *)arg;
7186 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
7187 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
7188 	    drpc->p);
7189 	drpc->done = 1;
7190 	crfree(drpc->cred);
7191 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
7192 }
7193 
7194 /*
7195  * Set up the commit DS mirror call for the pNFS I/O thread.
7196  */
7197 static int
7198 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
7199     struct nfsfh *fhp, int vers, int minorvers,
7200     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7201 {
7202 	int error, ret;
7203 
7204 	error = 0;
7205 	drpc->done = 0;
7206 	drpc->vp = vp;
7207 	drpc->off = offset;
7208 	drpc->len = cnt;
7209 	drpc->dsp = dsp;
7210 	drpc->fhp = fhp;
7211 	drpc->vers = vers;
7212 	drpc->minorvers = minorvers;
7213 	drpc->cred = crhold(cred);
7214 	drpc->p = p;
7215 	drpc->inprog = 0;
7216 	ret = EIO;
7217 	if (nfs_pnfsiothreads != 0) {
7218 		ret = nfs_pnfsio(start_commitds, drpc);
7219 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
7220 	}
7221 	if (ret != 0) {
7222 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
7223 		    minorvers, cred, p);
7224 		crfree(drpc->cred);
7225 	}
7226 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
7227 	return (error);
7228 }
7229 
7230 /*
7231  * NFS Advise rpc
7232  */
7233 int
7234 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
7235     struct ucred *cred, NFSPROC_T *p)
7236 {
7237 	u_int32_t *tl;
7238 	struct nfsrv_descript nfsd, *nd = &nfsd;
7239 	nfsattrbit_t hints;
7240 	int error;
7241 
7242 	NFSZERO_ATTRBIT(&hints);
7243 	if (advise == POSIX_FADV_WILLNEED)
7244 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7245 	else if (advise == POSIX_FADV_DONTNEED)
7246 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7247 	else
7248 		return (0);
7249 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp, cred);
7250 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7251 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
7252 	txdr_hyper(offset, tl);
7253 	tl += 2;
7254 	txdr_hyper(cnt, tl);
7255 	nfsrv_putattrbit(nd, &hints);
7256 	error = nfscl_request(nd, vp, p, cred);
7257 	if (error != 0)
7258 		return (error);
7259 	if (nd->nd_repstat != 0)
7260 		error = nd->nd_repstat;
7261 	m_freem(nd->nd_mrep);
7262 	return (error);
7263 }
7264 
7265 #ifdef notyet
7266 /*
7267  * NFS advise rpc to a NFSv4.2 DS.
7268  */
7269 static int
7270 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7271     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7272     struct ucred *cred, NFSPROC_T *p)
7273 {
7274 	uint32_t *tl;
7275 	struct nfsrv_descript nfsd, *nd = &nfsd;
7276 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7277 	struct nfssockreq *nrp;
7278 	nfsattrbit_t hints;
7279 	int error;
7280 
7281 	/* For NFS DSs prior to NFSv4.2, just return OK. */
7282 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
7283 		return (0);
7284 	NFSZERO_ATTRBIT(&hints);
7285 	if (advise == POSIX_FADV_WILLNEED)
7286 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
7287 	else if (advise == POSIX_FADV_DONTNEED)
7288 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
7289 	else
7290 		return (0);
7291 	nd->nd_mrep = NULL;
7292 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
7293 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers, NULL);
7294 	vers = NFS_VER4;
7295 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
7296 	    minorvers);
7297 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
7298 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
7299 	txdr_hyper(offset, tl);
7300 	tl += 2;
7301 	*tl = txdr_unsigned(cnt);
7302 	nfsrv_putattrbit(nd, &hints);
7303 	nrp = dsp->nfsclds_sockp;
7304 	if (nrp == NULL)
7305 		/* If NULL, use the MDS socket. */
7306 		nrp = &nmp->nm_sockreq;
7307 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
7308 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
7309 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
7310 	    nd->nd_repstat);
7311 	if (error != 0)
7312 		return (error);
7313 	if (nd->nd_repstat != 0)
7314 		error = nd->nd_repstat;
7315 	m_freem(nd->nd_mrep);
7316 	return (error);
7317 }
7318 
7319 /*
7320  * Start up the thread that will execute nfsrpc_commitds().
7321  */
7322 static void
7323 start_adviseds(void *arg, int pending)
7324 {
7325 	struct nfsclwritedsdorpc *drpc;
7326 
7327 	drpc = (struct nfsclwritedsdorpc *)arg;
7328 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
7329 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
7330 	    drpc->cred, drpc->p);
7331 	drpc->done = 1;
7332 	crfree(drpc->cred);
7333 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
7334 }
7335 
7336 /*
7337  * Set up the advise DS mirror call for the pNFS I/O thread.
7338  */
7339 static int
7340 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
7341     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
7342     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
7343 {
7344 	int error, ret;
7345 
7346 	error = 0;
7347 	drpc->done = 0;
7348 	drpc->vp = vp;
7349 	drpc->off = offset;
7350 	drpc->len = cnt;
7351 	drpc->advise = advise;
7352 	drpc->dsp = dsp;
7353 	drpc->fhp = fhp;
7354 	drpc->vers = vers;
7355 	drpc->minorvers = minorvers;
7356 	drpc->cred = crhold(cred);
7357 	drpc->p = p;
7358 	drpc->inprog = 0;
7359 	ret = EIO;
7360 	if (nfs_pnfsiothreads != 0) {
7361 		ret = nfs_pnfsio(start_adviseds, drpc);
7362 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
7363 	}
7364 	if (ret != 0) {
7365 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
7366 		    minorvers, cred, p);
7367 		crfree(drpc->cred);
7368 	}
7369 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
7370 	return (error);
7371 }
7372 #endif	/* notyet */
7373 
7374 /*
7375  * Do the Allocate operation, retrying for recovery.
7376  */
7377 int
7378 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
7379     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7380 {
7381 	int error, expireret = 0, retrycnt, nostateid;
7382 	uint32_t clidrev = 0;
7383 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
7384 	struct nfsfh *nfhp = NULL;
7385 	nfsv4stateid_t stateid;
7386 	off_t tmp_off;
7387 	void *lckp;
7388 
7389 	if (len < 0)
7390 		return (EINVAL);
7391 	if (len == 0)
7392 		return (0);
7393 	tmp_off = off + len;
7394 	NFSLOCKMNT(nmp);
7395 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7396 		NFSUNLOCKMNT(nmp);
7397 		return (EFBIG);
7398 	}
7399 	if (nmp->nm_clp != NULL)
7400 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7401 	NFSUNLOCKMNT(nmp);
7402 	nfhp = VTONFS(vp)->n_fhp;
7403 	retrycnt = 0;
7404 	do {
7405 		lckp = NULL;
7406 		nostateid = 0;
7407 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7408 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7409 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7410 		    stateid.other[2] == 0) {
7411 			nostateid = 1;
7412 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7413 		}
7414 
7415 		/*
7416 		 * Not finding a stateid should probably never happen,
7417 		 * but just return an error for this case.
7418 		 */
7419 		if (nostateid != 0)
7420 			error = EIO;
7421 		else
7422 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7423 			    nap, attrflagp, cred, p);
7424 		if (error == NFSERR_STALESTATEID)
7425 			nfscl_initiate_recovery(nmp->nm_clp);
7426 		if (lckp != NULL)
7427 			nfscl_lockderef(lckp);
7428 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7429 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7430 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7431 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7432 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
7433 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
7434 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7435 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
7436 			error = EIO;
7437 		}
7438 		retrycnt++;
7439 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7440 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7441 	    error == NFSERR_STALEDONTRECOVER ||
7442 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7443 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7444 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7445 	if (error != 0 && retrycnt >= 4)
7446 		error = EIO;
7447 	return (error);
7448 }
7449 
7450 /*
7451  * The allocate RPC.
7452  */
7453 static int
7454 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7455     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
7456 {
7457 	uint32_t *tl;
7458 	int error;
7459 	struct nfsrv_descript nfsd;
7460 	struct nfsrv_descript *nd = &nfsd;
7461 	nfsattrbit_t attrbits;
7462 
7463 	*attrflagp = 0;
7464 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp, cred);
7465 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7466 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7467 	txdr_hyper(off, tl); tl += 2;
7468 	txdr_hyper(len, tl); tl += 2;
7469 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7470 	NFSGETATTR_ATTRBIT(&attrbits);
7471 	nfsrv_putattrbit(nd, &attrbits);
7472 	error = nfscl_request(nd, vp, p, cred);
7473 	if (error != 0)
7474 		return (error);
7475 	if (nd->nd_repstat == 0) {
7476 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7477 		error = nfsm_loadattr(nd, nap);
7478 		if (error == 0)
7479 			*attrflagp = NFS_LATTR_NOSHRINK;
7480 	} else
7481 		error = nd->nd_repstat;
7482 nfsmout:
7483 	m_freem(nd->nd_mrep);
7484 	return (error);
7485 }
7486 
7487 /*
7488  * Set up the XDR arguments for the LayoutGet operation.
7489  */
7490 static void
7491 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7492     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7493     int layoutlen, int usecurstateid)
7494 {
7495 	uint32_t *tl;
7496 
7497 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7498 	    NFSX_STATEID);
7499 	*tl++ = newnfs_false;		/* Don't signal availability. */
7500 	*tl++ = txdr_unsigned(layouttype);
7501 	*tl++ = txdr_unsigned(iomode);
7502 	txdr_hyper(offset, tl);
7503 	tl += 2;
7504 	txdr_hyper(len, tl);
7505 	tl += 2;
7506 	txdr_hyper(minlen, tl);
7507 	tl += 2;
7508 	if (usecurstateid != 0) {
7509 		/* Special stateid for Current stateid. */
7510 		*tl++ = txdr_unsigned(1);
7511 		*tl++ = 0;
7512 		*tl++ = 0;
7513 		*tl++ = 0;
7514 	} else {
7515 		*tl++ = txdr_unsigned(stateidp->seqid);
7516 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7517 		*tl++ = stateidp->other[0];
7518 		*tl++ = stateidp->other[1];
7519 		*tl++ = stateidp->other[2];
7520 	}
7521 	*tl = txdr_unsigned(layoutlen);
7522 }
7523 
7524 /*
7525  * Parse the reply for a successful LayoutGet operation.
7526  */
7527 static int
7528 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7529     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7530 {
7531 	uint32_t *tl;
7532 	struct nfsclflayout *flp, *prevflp, *tflp;
7533 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7534 	int m, mirrorcnt;
7535 	uint64_t retlen, off;
7536 	struct nfsfh *nfhp;
7537 	uint8_t *cp;
7538 	uid_t user;
7539 	gid_t grp;
7540 
7541 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7542 	error = 0;
7543 	flp = NULL;
7544 	gotiomode = -1;
7545 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7546 	if (*tl++ != 0)
7547 		*retonclosep = 1;
7548 	else
7549 		*retonclosep = 0;
7550 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7551 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7552 	    (int)stateidp->seqid);
7553 	stateidp->other[0] = *tl++;
7554 	stateidp->other[1] = *tl++;
7555 	stateidp->other[2] = *tl++;
7556 	cnt = fxdr_unsigned(int, *tl);
7557 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7558 	if (cnt <= 0 || cnt > 10000) {
7559 		/* Don't accept more than 10000 layouts in reply. */
7560 		error = NFSERR_BADXDR;
7561 		goto nfsmout;
7562 	}
7563 	for (i = 0; i < cnt; i++) {
7564 		/* Dissect to the layout type. */
7565 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7566 		    3 * NFSX_UNSIGNED);
7567 		off = fxdr_hyper(tl); tl += 2;
7568 		retlen = fxdr_hyper(tl); tl += 2;
7569 		iomode = fxdr_unsigned(int, *tl++);
7570 		laytype = fxdr_unsigned(int, *tl);
7571 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7572 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7573 		/* Ignore length of layout body for now. */
7574 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7575 			/* Parse the File layout up to fhcnt. */
7576 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7577 			    NFSX_HYPER + NFSX_V4DEVICEID);
7578 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7579 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7580 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7581 			if (fhcnt < 0 || fhcnt > 100) {
7582 				/* Don't accept more than 100 file handles. */
7583 				error = NFSERR_BADXDR;
7584 				goto nfsmout;
7585 			}
7586 			if (fhcnt > 0)
7587 				flp = malloc(sizeof(*flp) + fhcnt *
7588 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7589 				    M_WAITOK);
7590 			else
7591 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7592 				    M_WAITOK);
7593 			flp->nfsfl_flags = NFSFL_FILE;
7594 			flp->nfsfl_fhcnt = 0;
7595 			flp->nfsfl_devp = NULL;
7596 			flp->nfsfl_off = off;
7597 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7598 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7599 			else
7600 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7601 			flp->nfsfl_iomode = iomode;
7602 			if (gotiomode == -1)
7603 				gotiomode = flp->nfsfl_iomode;
7604 			/* Ignore layout body length for now. */
7605 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7606 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7607 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7608 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7609 			mtx_lock(&nmp->nm_mtx);
7610 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7611 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7612 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7613 			mtx_unlock(&nmp->nm_mtx);
7614 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7615 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7616 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7617 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7618 			for (j = 0; j < fhcnt; j++) {
7619 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7620 				nfhlen = fxdr_unsigned(int, *tl);
7621 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7622 					error = NFSERR_BADXDR;
7623 					goto nfsmout;
7624 				}
7625 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7626 				    M_NFSFH, M_WAITOK);
7627 				flp->nfsfl_fh[j] = nfhp;
7628 				flp->nfsfl_fhcnt++;
7629 				nfhp->nfh_len = nfhlen;
7630 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7631 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7632 			}
7633 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7634 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7635 			    NFSX_HYPER);
7636 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7637 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7638 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7639 				error = NFSERR_BADXDR;
7640 				goto nfsmout;
7641 			}
7642 			flp = malloc(sizeof(*flp) + mirrorcnt *
7643 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7644 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7645 			flp->nfsfl_mirrorcnt = mirrorcnt;
7646 			for (j = 0; j < mirrorcnt; j++)
7647 				flp->nfsfl_ffm[j].devp = NULL;
7648 			flp->nfsfl_off = off;
7649 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7650 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7651 			else
7652 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7653 			flp->nfsfl_iomode = iomode;
7654 			if (gotiomode == -1)
7655 				gotiomode = flp->nfsfl_iomode;
7656 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7657 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7658 			    (uintmax_t)flp->nfsfl_stripeunit);
7659 			for (j = 0; j < mirrorcnt; j++) {
7660 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7661 				k = fxdr_unsigned(int, *tl);
7662 				if (k < 1 || k > 128) {
7663 					error = NFSERR_BADXDR;
7664 					goto nfsmout;
7665 				}
7666 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7667 				for (l = 0; l < k; l++) {
7668 					NFSM_DISSECT(tl, uint32_t *,
7669 					    NFSX_V4DEVICEID + NFSX_STATEID +
7670 					    2 * NFSX_UNSIGNED);
7671 					if (l == 0) {
7672 						/* Just use the first server. */
7673 						NFSBCOPY(tl,
7674 						    flp->nfsfl_ffm[j].dev,
7675 						    NFSX_V4DEVICEID);
7676 						tl += (NFSX_V4DEVICEID /
7677 						    NFSX_UNSIGNED);
7678 						tl++;
7679 						flp->nfsfl_ffm[j].st.seqid =
7680 						    *tl++;
7681 						flp->nfsfl_ffm[j].st.other[0] =
7682 						    *tl++;
7683 						flp->nfsfl_ffm[j].st.other[1] =
7684 						    *tl++;
7685 						flp->nfsfl_ffm[j].st.other[2] =
7686 						    *tl++;
7687 						NFSCL_DEBUG(4, "st.seqid=%u "
7688 						 "st.o0=0x%x st.o1=0x%x "
7689 						 "st.o2=0x%x\n",
7690 						 flp->nfsfl_ffm[j].st.seqid,
7691 						 flp->nfsfl_ffm[j].st.other[0],
7692 						 flp->nfsfl_ffm[j].st.other[1],
7693 						 flp->nfsfl_ffm[j].st.other[2]);
7694 					} else
7695 						tl += ((NFSX_V4DEVICEID +
7696 						    NFSX_STATEID +
7697 						    NFSX_UNSIGNED) /
7698 						    NFSX_UNSIGNED);
7699 					fhcnt = fxdr_unsigned(int, *tl);
7700 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7701 					if (fhcnt < 1 ||
7702 					    fhcnt > NFSDEV_MAXVERS) {
7703 						error = NFSERR_BADXDR;
7704 						goto nfsmout;
7705 					}
7706 					for (m = 0; m < fhcnt; m++) {
7707 						NFSM_DISSECT(tl, uint32_t *,
7708 						    NFSX_UNSIGNED);
7709 						nfhlen = fxdr_unsigned(int,
7710 						    *tl);
7711 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7712 						    nfhlen);
7713 						if (nfhlen <= 0 || nfhlen >
7714 						    NFSX_V4FHMAX) {
7715 							error = NFSERR_BADXDR;
7716 							goto nfsmout;
7717 						}
7718 						NFSM_DISSECT(cp, uint8_t *,
7719 						    NFSM_RNDUP(nfhlen));
7720 						if (l == 0) {
7721 							flp->nfsfl_ffm[j].fhcnt
7722 							    = fhcnt;
7723 							nfhp = malloc(
7724 							    sizeof(*nfhp) +
7725 							    nfhlen - 1, M_NFSFH,
7726 							    M_WAITOK);
7727 							flp->nfsfl_ffm[j].fh[m]
7728 							    = nfhp;
7729 							nfhp->nfh_len = nfhlen;
7730 							NFSBCOPY(cp,
7731 							    nfhp->nfh_fh,
7732 							    nfhlen);
7733 							NFSCL_DEBUG(4,
7734 							    "got fh\n");
7735 						}
7736 					}
7737 					/* Now, get the ffsd_user/ffds_group. */
7738 					error = nfsrv_parseug(nd, 0, &user,
7739 					    &grp, curthread);
7740 					NFSCL_DEBUG(4, "after parseu=%d\n",
7741 					    error);
7742 					if (error == 0)
7743 						error = nfsrv_parseug(nd, 1,
7744 						    &user, &grp, curthread);
7745 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7746 					    grp);
7747 					if (error != 0)
7748 						goto nfsmout;
7749 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7750 					    user, grp);
7751 					if (l == 0) {
7752 						flp->nfsfl_ffm[j].user = user;
7753 						flp->nfsfl_ffm[j].group = grp;
7754 						NFSCL_DEBUG(4,
7755 						    "usr=%d grp=%d\n", user,
7756 						    grp);
7757 					}
7758 				}
7759 			}
7760 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7761 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7762 #ifdef notnow
7763 			/*
7764 			 * At this time, there is no flag.
7765 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7766 			 * added, or it may never exist?
7767 			 */
7768 			mtx_lock(&nmp->nm_mtx);
7769 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7770 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7771 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7772 			mtx_unlock(&nmp->nm_mtx);
7773 #endif
7774 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7775 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7776 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7777 		} else {
7778 			error = NFSERR_BADXDR;
7779 			goto nfsmout;
7780 		}
7781 		if (flp->nfsfl_iomode == gotiomode) {
7782 			/* Keep the list in increasing offset order. */
7783 			tflp = LIST_FIRST(flhp);
7784 			prevflp = NULL;
7785 			while (tflp != NULL &&
7786 			    tflp->nfsfl_off < flp->nfsfl_off) {
7787 				prevflp = tflp;
7788 				tflp = LIST_NEXT(tflp, nfsfl_list);
7789 			}
7790 			if (prevflp == NULL)
7791 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7792 			else
7793 				LIST_INSERT_AFTER(prevflp, flp,
7794 				    nfsfl_list);
7795 			NFSCL_DEBUG(4, "flp inserted\n");
7796 		} else {
7797 			printf("nfscl_layoutget(): got wrong iomode\n");
7798 			nfscl_freeflayout(flp);
7799 		}
7800 		flp = NULL;
7801 	}
7802 nfsmout:
7803 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7804 	if (error != 0 && flp != NULL)
7805 		nfscl_freeflayout(flp);
7806 	return (error);
7807 }
7808 
7809 /*
7810  * Parse a user/group digit string.
7811  */
7812 static int
7813 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7814     NFSPROC_T *p)
7815 {
7816 	uint32_t *tl;
7817 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7818 	uint32_t len = 0;
7819 	int error = 0;
7820 
7821 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7822 	len = fxdr_unsigned(uint32_t, *tl);
7823 	str = NULL;
7824 	if (len > NFSV4_OPAQUELIMIT) {
7825 		error = NFSERR_BADXDR;
7826 		goto nfsmout;
7827 	}
7828 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7829 	if (len == 0) {
7830 		if (dogrp != 0)
7831 			*gidp = GID_NOGROUP;
7832 		else
7833 			*uidp = UID_NOBODY;
7834 		return (0);
7835 	}
7836 	if (len > NFSV4_SMALLSTR)
7837 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7838 	else
7839 		str = str0;
7840 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7841 	NFSBCOPY(cp, str, len);
7842 	str[len] = '\0';
7843 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7844 	if (dogrp != 0)
7845 		error = nfsv4_strtogid(nd, str, len, gidp);
7846 	else
7847 		error = nfsv4_strtouid(nd, str, len, uidp);
7848 nfsmout:
7849 	if (len > NFSV4_SMALLSTR)
7850 		free(str, M_TEMP);
7851 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7852 	return (error);
7853 }
7854 
7855 /*
7856  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7857  * so that it does both an Open and a Layoutget.
7858  */
7859 static int
7860 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7861     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7862     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7863     struct ucred *cred, NFSPROC_T *p)
7864 {
7865 	struct nfscllayout *lyp;
7866 	struct nfsclflayout *flp;
7867 	struct nfsclflayouthead flh;
7868 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7869 	int layouttype, laystat;
7870 	nfsv4stateid_t stateid;
7871 	struct nfsclsession *tsep;
7872 
7873 	error = 0;
7874 	if (NFSHASFLEXFILE(nmp))
7875 		layouttype = NFSLAYOUT_FLEXFILE;
7876 	else
7877 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7878 	/*
7879 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7880 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7881 	 * flp == NULL.
7882 	 */
7883 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, mode, &flp,
7884 	    &recalled);
7885 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7886 	if (lyp == NULL)
7887 		islocked = 0;
7888 	else if (flp != NULL)
7889 		islocked = 1;
7890 	else
7891 		islocked = 2;
7892 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7893 		LIST_INIT(&flh);
7894 		tsep = nfsmnt_mdssession(nmp);
7895 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7896 		    3 * NFSX_UNSIGNED);
7897 		if (lyp == NULL)
7898 			usecurstateid = 1;
7899 		else {
7900 			usecurstateid = 0;
7901 			stateid.seqid = lyp->nfsly_stateid.seqid;
7902 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7903 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7904 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7905 		}
7906 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7907 		    newfhp, newfhlen, mode, op, name, namelen,
7908 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7909 		    &retonclose, &flh, &laystat, cred, p);
7910 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7911 		    laystat, error);
7912 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7913 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7914 		    &islocked, cred, p);
7915 	} else
7916 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7917 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7918 	if (islocked == 2)
7919 		nfscl_rellayout(lyp, 1);
7920 	else if (islocked == 1)
7921 		nfscl_rellayout(lyp, 0);
7922 	return (error);
7923 }
7924 
7925 /*
7926  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7927  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7928  * handled by nfsrpc_openrpc().
7929  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7930  * can be NULL.
7931  */
7932 static int
7933 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7934     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7935     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7936     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7937     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7938     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7939 {
7940 	uint32_t *tl;
7941 	struct nfsrv_descript nfsd, *nd = &nfsd;
7942 	struct nfscldeleg *ndp = NULL;
7943 	struct nfsvattr nfsva;
7944 	struct nfsclsession *tsep;
7945 	uint32_t rflags, deleg;
7946 	nfsattrbit_t attrbits;
7947 	int error, ret, acesize, limitby, iomode;
7948 
7949 	*dpp = NULL;
7950 	*laystatp = ENXIO;
7951 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7952 	    0, 0, cred);
7953 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7954 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7955 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7956 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7957 	tsep = nfsmnt_mdssession(nmp);
7958 	*tl++ = tsep->nfsess_clientid.lval[0];
7959 	*tl = tsep->nfsess_clientid.lval[1];
7960 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7961 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7962 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7963 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7964 	nfsm_strtom(nd, name, namelen);
7965 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7966 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7967 	NFSZERO_ATTRBIT(&attrbits);
7968 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7969 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7970 	nfsrv_putattrbit(nd, &attrbits);
7971 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7972 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7973 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7974 		iomode = NFSLAYOUTIOMODE_RW;
7975 	else
7976 		iomode = NFSLAYOUTIOMODE_READ;
7977 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7978 	    layouttype, layoutlen, usecurstateid);
7979 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7980 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7981 	if (error != 0)
7982 		return (error);
7983 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7984 	if (nd->nd_repstat != 0)
7985 		*laystatp = nd->nd_repstat;
7986 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7987 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7988 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7989 		    6 * NFSX_UNSIGNED);
7990 		op->nfso_stateid.seqid = *tl++;
7991 		op->nfso_stateid.other[0] = *tl++;
7992 		op->nfso_stateid.other[1] = *tl++;
7993 		op->nfso_stateid.other[2] = *tl;
7994 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7995 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7996 		if (error != 0)
7997 			goto nfsmout;
7998 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7999 		deleg = fxdr_unsigned(u_int32_t, *tl);
8000 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8001 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8002 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
8003 			      NFSCLFLAGS_FIRSTDELEG))
8004 				op->nfso_own->nfsow_clp->nfsc_flags |=
8005 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8006 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
8007 			    M_NFSCLDELEG, M_WAITOK);
8008 			LIST_INIT(&ndp->nfsdl_owner);
8009 			LIST_INIT(&ndp->nfsdl_lock);
8010 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
8011 			ndp->nfsdl_fhlen = newfhlen;
8012 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
8013 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
8014 			nfscl_lockinit(&ndp->nfsdl_rwlock);
8015 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8016 			    NFSX_UNSIGNED);
8017 			ndp->nfsdl_stateid.seqid = *tl++;
8018 			ndp->nfsdl_stateid.other[0] = *tl++;
8019 			ndp->nfsdl_stateid.other[1] = *tl++;
8020 			ndp->nfsdl_stateid.other[2] = *tl++;
8021 			ret = fxdr_unsigned(int, *tl);
8022 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8023 				ndp->nfsdl_flags = NFSCLDL_WRITE;
8024 				/*
8025 				 * Indicates how much the file can grow.
8026 				 */
8027 				NFSM_DISSECT(tl, u_int32_t *,
8028 				    3 * NFSX_UNSIGNED);
8029 				limitby = fxdr_unsigned(int, *tl++);
8030 				switch (limitby) {
8031 				case NFSV4OPEN_LIMITSIZE:
8032 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
8033 					break;
8034 				case NFSV4OPEN_LIMITBLOCKS:
8035 					ndp->nfsdl_sizelimit =
8036 					    fxdr_unsigned(u_int64_t, *tl++);
8037 					ndp->nfsdl_sizelimit *=
8038 					    fxdr_unsigned(u_int64_t, *tl);
8039 					break;
8040 				default:
8041 					error = NFSERR_BADXDR;
8042 					goto nfsmout;
8043 				};
8044 			} else
8045 				ndp->nfsdl_flags = NFSCLDL_READ;
8046 			if (ret != 0)
8047 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
8048 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, false,
8049 			    &ret, &acesize, p);
8050 			if (error != 0)
8051 				goto nfsmout;
8052 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8053 			error = NFSERR_BADXDR;
8054 			goto nfsmout;
8055 		}
8056 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
8057 		    nfscl_assumeposixlocks)
8058 			op->nfso_posixlock = 1;
8059 		else
8060 			op->nfso_posixlock = 0;
8061 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8062 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
8063 		if (*++tl == 0) {
8064 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
8065 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
8066 			    NULL, NULL, NULL, p, cred);
8067 			if (error != 0)
8068 				goto nfsmout;
8069 			if (ndp != NULL) {
8070 				ndp->nfsdl_change = nfsva.na_filerev;
8071 				ndp->nfsdl_modtime = nfsva.na_mtime;
8072 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8073 				*dpp = ndp;
8074 				ndp = NULL;
8075 			}
8076 			/*
8077 			 * At this point, the Open has succeeded, so set
8078 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
8079 			 * this function just won't return a layout.
8080 			 */
8081 			if (nd->nd_repstat == 0) {
8082 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8083 				*laystatp = fxdr_unsigned(int, *++tl);
8084 				if (*laystatp == 0) {
8085 					error = nfsrv_parselayoutget(nmp, nd,
8086 					    stateidp, retonclosep, flhp);
8087 					if (error != 0)
8088 						*laystatp = error;
8089 				}
8090 			} else
8091 				nd->nd_repstat = 0;	/* Return 0 for Open. */
8092 		}
8093 	}
8094 	if (nd->nd_repstat != 0 && error == 0)
8095 		error = nd->nd_repstat;
8096 nfsmout:
8097 	free(ndp, M_NFSCLDELEG);
8098 	m_freem(nd->nd_mrep);
8099 	return (error);
8100 }
8101 
8102 /*
8103  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
8104  * Used only for mounts with pNFS enabled.
8105  */
8106 static int
8107 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8108     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8109     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8110     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8111     int *dattrflagp, int *unlockedp, nfsv4stateid_t *stateidp,
8112     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
8113     struct nfsclflayouthead *flhp, int *laystatp)
8114 {
8115 	uint32_t *tl;
8116 	int error = 0, deleg, newone, ret, acesize, limitby;
8117 	struct nfsrv_descript nfsd, *nd = &nfsd;
8118 	struct nfsclopen *op;
8119 	struct nfscldeleg *dp = NULL;
8120 	struct nfsnode *np;
8121 	struct nfsfh *nfhp;
8122 	struct nfsclsession *tsep;
8123 	nfsattrbit_t attrbits;
8124 	nfsv4stateid_t stateid;
8125 	struct nfsmount *nmp;
8126 
8127 	nmp = VFSTONFS(dvp->v_mount);
8128 	np = VTONFS(dvp);
8129 	*laystatp = ENXIO;
8130 	*unlockedp = 0;
8131 	*nfhpp = NULL;
8132 	*dpp = NULL;
8133 	*attrflagp = 0;
8134 	*dattrflagp = 0;
8135 	if (namelen > NFS_MAXNAMLEN)
8136 		return (ENAMETOOLONG);
8137 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp, cred);
8138 	/*
8139 	 * For V4, this is actually an Open op.
8140 	 */
8141 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
8142 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
8143 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
8144 	    NFSV4OPEN_ACCESSREAD);
8145 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
8146 	tsep = nfsmnt_mdssession(nmp);
8147 	*tl++ = tsep->nfsess_clientid.lval[0];
8148 	*tl = tsep->nfsess_clientid.lval[1];
8149 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
8150 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8151 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
8152 	if ((fmode & O_EXCL) != 0) {
8153 		if (NFSHASSESSPERSIST(nmp)) {
8154 			/* Use GUARDED for persistent sessions. */
8155 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
8156 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8157 		} else {
8158 			/* Otherwise, use EXCLUSIVE4_1. */
8159 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
8160 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
8161 			*tl++ = cverf.lval[0];
8162 			*tl = cverf.lval[1];
8163 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
8164 		}
8165 	} else {
8166 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
8167 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
8168 	}
8169 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8170 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
8171 	nfsm_strtom(nd, name, namelen);
8172 	/* Get the new file's handle and attributes, plus save the FH. */
8173 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
8174 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
8175 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
8176 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8177 	NFSGETATTR_ATTRBIT(&attrbits);
8178 	nfsrv_putattrbit(nd, &attrbits);
8179 	/* Get the directory's post-op attributes. */
8180 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8181 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8182 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
8183 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
8184 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8185 	nfsrv_putattrbit(nd, &attrbits);
8186 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
8187 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
8188 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
8189 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
8190 	    layouttype, layoutlen, usecurstateid);
8191 	error = nfscl_request(nd, dvp, p, cred);
8192 	if (error != 0)
8193 		return (error);
8194 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
8195 	    error);
8196 	if (nd->nd_repstat != 0)
8197 		*laystatp = nd->nd_repstat;
8198 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
8199 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8200 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
8201 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8202 		    6 * NFSX_UNSIGNED);
8203 		stateid.seqid = *tl++;
8204 		stateid.other[0] = *tl++;
8205 		stateid.other[1] = *tl++;
8206 		stateid.other[2] = *tl;
8207 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
8208 		if (error != 0)
8209 			goto nfsmout;
8210 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
8211 		deleg = fxdr_unsigned(int, *tl);
8212 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
8213 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
8214 			if (!(owp->nfsow_clp->nfsc_flags &
8215 			      NFSCLFLAGS_FIRSTDELEG))
8216 				owp->nfsow_clp->nfsc_flags |=
8217 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
8218 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
8219 			    M_NFSCLDELEG, M_WAITOK);
8220 			LIST_INIT(&dp->nfsdl_owner);
8221 			LIST_INIT(&dp->nfsdl_lock);
8222 			dp->nfsdl_clp = owp->nfsow_clp;
8223 			newnfs_copyincred(cred, &dp->nfsdl_cred);
8224 			nfscl_lockinit(&dp->nfsdl_rwlock);
8225 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
8226 			    NFSX_UNSIGNED);
8227 			dp->nfsdl_stateid.seqid = *tl++;
8228 			dp->nfsdl_stateid.other[0] = *tl++;
8229 			dp->nfsdl_stateid.other[1] = *tl++;
8230 			dp->nfsdl_stateid.other[2] = *tl++;
8231 			ret = fxdr_unsigned(int, *tl);
8232 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
8233 				dp->nfsdl_flags = NFSCLDL_WRITE;
8234 				/*
8235 				 * Indicates how much the file can grow.
8236 				 */
8237 				NFSM_DISSECT(tl, u_int32_t *,
8238 				    3 * NFSX_UNSIGNED);
8239 				limitby = fxdr_unsigned(int, *tl++);
8240 				switch (limitby) {
8241 				case NFSV4OPEN_LIMITSIZE:
8242 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
8243 					break;
8244 				case NFSV4OPEN_LIMITBLOCKS:
8245 					dp->nfsdl_sizelimit =
8246 					    fxdr_unsigned(u_int64_t, *tl++);
8247 					dp->nfsdl_sizelimit *=
8248 					    fxdr_unsigned(u_int64_t, *tl);
8249 					break;
8250 				default:
8251 					error = NFSERR_BADXDR;
8252 					goto nfsmout;
8253 				};
8254 			} else {
8255 				dp->nfsdl_flags = NFSCLDL_READ;
8256 			}
8257 			if (ret != 0)
8258 				dp->nfsdl_flags |= NFSCLDL_RECALL;
8259 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, false,
8260 			    &ret, &acesize, p);
8261 			if (error != 0)
8262 				goto nfsmout;
8263 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
8264 			error = NFSERR_BADXDR;
8265 			goto nfsmout;
8266 		}
8267 
8268 		/* Now, we should have the status for the SaveFH. */
8269 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8270 		if (*++tl == 0) {
8271 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
8272 			/*
8273 			 * Now, process the GetFH and Getattr for the newly
8274 			 * created file. nfscl_mtofh() will set
8275 			 * ND_NOMOREDATA if these weren't successful.
8276 			 */
8277 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
8278 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
8279 			if (error != 0)
8280 				goto nfsmout;
8281 		} else
8282 			nd->nd_flag |= ND_NOMOREDATA;
8283 		/* Now we have the PutFH and Getattr for the directory. */
8284 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8285 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8286 			if (*++tl != 0)
8287 				nd->nd_flag |= ND_NOMOREDATA;
8288 			else {
8289 				NFSM_DISSECT(tl, uint32_t *, 2 *
8290 				    NFSX_UNSIGNED);
8291 				if (*++tl != 0)
8292 					nd->nd_flag |= ND_NOMOREDATA;
8293 			}
8294 		}
8295 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8296 			/* Load the directory attributes. */
8297 			error = nfsm_loadattr(nd, dnap);
8298 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
8299 			if (error != 0)
8300 				goto nfsmout;
8301 			*dattrflagp = 1;
8302 			if (dp != NULL && *attrflagp != 0) {
8303 				dp->nfsdl_change = nnap->na_filerev;
8304 				dp->nfsdl_modtime = nnap->na_mtime;
8305 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
8306 			}
8307 			/*
8308 			 * We can now complete the Open state.
8309 			 */
8310 			nfhp = *nfhpp;
8311 			if (dp != NULL) {
8312 				dp->nfsdl_fhlen = nfhp->nfh_len;
8313 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
8314 				    nfhp->nfh_len);
8315 			}
8316 			/*
8317 			 * Get an Open structure that will be
8318 			 * attached to the OpenOwner, acquired already.
8319 			 */
8320 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
8321 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
8322 			    cred, p, NULL, &op, &newone, NULL, 0, false);
8323 			if (error != 0)
8324 				goto nfsmout;
8325 			op->nfso_stateid = stateid;
8326 			newnfs_copyincred(cred, &op->nfso_cred);
8327 
8328 			nfscl_openrelease(nmp, op, error, newone);
8329 			*unlockedp = 1;
8330 
8331 			/* Now, handle the RestoreFH and LayoutGet. */
8332 			if (nd->nd_repstat == 0) {
8333 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
8334 				*laystatp = fxdr_unsigned(int, *(tl + 3));
8335 				if (*laystatp == 0) {
8336 					error = nfsrv_parselayoutget(nmp, nd,
8337 					    stateidp, retonclosep, flhp);
8338 					if (error != 0)
8339 						*laystatp = error;
8340 				}
8341 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
8342 				    error);
8343 			} else
8344 				nd->nd_repstat = 0;
8345 		}
8346 	}
8347 	if (nd->nd_repstat != 0 && error == 0)
8348 		error = nd->nd_repstat;
8349 	if (error == NFSERR_STALECLIENTID)
8350 		nfscl_initiate_recovery(owp->nfsow_clp);
8351 nfsmout:
8352 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
8353 	if (error == 0)
8354 		*dpp = dp;
8355 	else
8356 		free(dp, M_NFSCLDELEG);
8357 	m_freem(nd->nd_mrep);
8358 	return (error);
8359 }
8360 
8361 /*
8362  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
8363  */
8364 static int
8365 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
8366     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
8367     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
8368     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
8369     int *dattrflagp, int *unlockedp)
8370 {
8371 	struct nfscllayout *lyp;
8372 	struct nfsclflayouthead flh;
8373 	struct nfsfh *nfhp;
8374 	struct nfsclsession *tsep;
8375 	struct nfsmount *nmp;
8376 	nfsv4stateid_t stateid;
8377 	int error, layoutlen, layouttype, retonclose, laystat;
8378 
8379 	error = 0;
8380 	nmp = VFSTONFS(dvp->v_mount);
8381 	if (NFSHASFLEXFILE(nmp))
8382 		layouttype = NFSLAYOUT_FLEXFILE;
8383 	else
8384 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
8385 	LIST_INIT(&flh);
8386 	tsep = nfsmnt_mdssession(nmp);
8387 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
8388 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
8389 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
8390 	    unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
8391 	    &flh, &laystat);
8392 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
8393 	    laystat, error);
8394 	lyp = NULL;
8395 	if (laystat == 0) {
8396 		nfhp = *nfhpp;
8397 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8398 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8399 		    layouttype, laystat, NULL, cred, p);
8400 	} else
8401 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8402 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8403 		    cred, p);
8404 	if (laystat == 0)
8405 		nfscl_rellayout(lyp, 0);
8406 	return (error);
8407 }
8408 
8409 /*
8410  * Process the results of a layoutget() operation.
8411  */
8412 static int
8413 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8414     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8415     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8416     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8417 {
8418 	struct nfsclflayout *tflp;
8419 	struct nfscldevinfo *dip;
8420 	uint8_t *dev;
8421 	int i, mirrorcnt;
8422 
8423 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8424 		NFSLOCKMNT(nmp);
8425 		if (!NFSHASFLEXFILE(nmp)) {
8426 			/* Switch to using Flex File Layout. */
8427 			nmp->nm_state |= NFSSTA_FLEXFILE;
8428 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8429 			/* Disable pNFS. */
8430 			NFSCL_DEBUG(1, "disable PNFS\n");
8431 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8432 		}
8433 		NFSUNLOCKMNT(nmp);
8434 	}
8435 	if (laystat == 0) {
8436 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8437 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8438 			if (layouttype == NFSLAYOUT_FLEXFILE)
8439 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8440 			else
8441 				mirrorcnt = 1;
8442 			for (i = 0; i < mirrorcnt; i++) {
8443 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8444 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8445 				if (laystat != 0) {
8446 					if (layouttype == NFSLAYOUT_FLEXFILE)
8447 						dev = tflp->nfsfl_ffm[i].dev;
8448 					else
8449 						dev = tflp->nfsfl_dev;
8450 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8451 					    layouttype, notifybit, &dip, cred,
8452 					    p);
8453 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8454 					    laystat);
8455 					if (laystat != 0)
8456 						goto out;
8457 					laystat = nfscl_adddevinfo(nmp, dip, i,
8458 					    tflp);
8459 					if (laystat != 0)
8460 						printf("nfsrpc_layoutgetresout"
8461 						    ": cannot add\n");
8462 				}
8463 			}
8464 		}
8465 	}
8466 out:
8467 	if (laystat == 0) {
8468 		/*
8469 		 * nfscl_layout() always returns with the nfsly_lock
8470 		 * set to a refcnt (shared lock).
8471 		 * Passing in dvp is sufficient, since it is only used to
8472 		 * get the fsid for the file system.
8473 		 */
8474 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8475 		    layouttype, retonclose, flhp, lypp, cred, p);
8476 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8477 		    laystat);
8478 		if (laystat == 0 && islockedp != NULL)
8479 			*islockedp = 1;
8480 	}
8481 	return (laystat);
8482 }
8483 
8484 /*
8485  * nfs copy_file_range operation.
8486  */
8487 int
8488 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8489     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8490     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8491     struct ucred *cred, bool consecutive, bool *must_commitp)
8492 {
8493 	int commit, error, expireret = 0, retrycnt;
8494 	u_int32_t clidrev = 0;
8495 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8496 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8497 	nfsv4stateid_t instateid, outstateid;
8498 	void *inlckp, *outlckp;
8499 
8500 	if (nmp->nm_clp != NULL)
8501 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8502 	innfhp = VTONFS(invp)->n_fhp;
8503 	outnfhp = VTONFS(outvp)->n_fhp;
8504 	retrycnt = 0;
8505 	do {
8506 		/* Get both stateids. */
8507 		inlckp = NULL;
8508 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8509 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8510 		    &inlckp);
8511 		outlckp = NULL;
8512 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8513 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8514 		    &outlckp);
8515 
8516 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8517 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8518 		    outattrflagp, consecutive, &commit, cred, curthread);
8519 		if (error == 0) {
8520 			if (commit != NFSWRITE_FILESYNC)
8521 				*must_commitp = true;
8522 			*inoffp += *lenp;
8523 			*outoffp += *lenp;
8524 		} else if (error == NFSERR_STALESTATEID)
8525 			nfscl_initiate_recovery(nmp->nm_clp);
8526 		if (inlckp != NULL)
8527 			nfscl_lockderef(inlckp);
8528 		if (outlckp != NULL)
8529 			nfscl_lockderef(outlckp);
8530 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8531 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8532 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8533 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8534 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
8535 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
8536 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8537 			    curthread);
8538 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
8539 			error = EIO;
8540 		}
8541 		retrycnt++;
8542 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8543 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8544 	      error == NFSERR_STALEDONTRECOVER ||
8545 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8546 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8547 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8548 	if (error != 0 && (retrycnt >= 4 ||
8549 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8550 	      error == NFSERR_STALEDONTRECOVER))
8551 		error = EIO;
8552 	return (error);
8553 }
8554 
8555 /*
8556  * The copy RPC.
8557  */
8558 static int
8559 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8560     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8561     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8562     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8563     NFSPROC_T *p)
8564 {
8565 	uint32_t *tl;
8566 	int error;
8567 	struct nfsrv_descript nfsd;
8568 	struct nfsrv_descript *nd = &nfsd;
8569 	struct nfsmount *nmp;
8570 	nfsattrbit_t attrbits;
8571 	uint64_t len;
8572 
8573 	nmp = VFSTONFS(outvp->v_mount);
8574 	*inattrflagp = *outattrflagp = 0;
8575 	*commitp = NFSWRITE_UNSTABLE;
8576 	len = *lenp;
8577 	*lenp = 0;
8578 	if (len > nfs_maxcopyrange)
8579 		len = nfs_maxcopyrange;
8580 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp, cred);
8581 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8582 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8583 	NFSGETATTR_ATTRBIT(&attrbits);
8584 	nfsrv_putattrbit(nd, &attrbits);
8585 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8586 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8587 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8588 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8589 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8590 	*tl = txdr_unsigned(NFSV4OP_COPY);
8591 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8592 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8593 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8594 	txdr_hyper(inoff, tl); tl += 2;
8595 	txdr_hyper(outoff, tl); tl += 2;
8596 	txdr_hyper(len, tl); tl += 2;
8597 	if (consecutive)
8598 		*tl++ = newnfs_true;
8599 	else
8600 		*tl++ = newnfs_false;
8601 	*tl++ = newnfs_true;
8602 	*tl++ = 0;
8603 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8604 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8605 	nfsrv_putattrbit(nd, &attrbits);
8606 	error = nfscl_request(nd, invp, p, cred);
8607 	if (error != 0)
8608 		return (error);
8609 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8610 		/* Get the input file's attributes. */
8611 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8612 		if (*(tl + 1) == 0) {
8613 			error = nfsm_loadattr(nd, innap);
8614 			if (error != 0)
8615 				goto nfsmout;
8616 			*inattrflagp = 1;
8617 		} else
8618 			nd->nd_flag |= ND_NOMOREDATA;
8619 	}
8620 	/* Skip over return stat for PutFH. */
8621 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8622 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8623 		if (*++tl != 0)
8624 			nd->nd_flag |= ND_NOMOREDATA;
8625 	}
8626 	/* Skip over return stat for Copy. */
8627 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8628 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8629 	if (nd->nd_repstat == 0) {
8630 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8631 		if (*tl != 0) {
8632 			/* There should be no callback ids. */
8633 			error = NFSERR_BADXDR;
8634 			goto nfsmout;
8635 		}
8636 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8637 		    NFSX_VERF);
8638 		len = fxdr_hyper(tl); tl += 2;
8639 		*commitp = fxdr_unsigned(int, *tl++);
8640 		NFSLOCKMNT(nmp);
8641 		if (!NFSHASWRITEVERF(nmp)) {
8642 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8643 			NFSSETWRITEVERF(nmp);
8644 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8645 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8646 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8647 		}
8648 		NFSUNLOCKMNT(nmp);
8649 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8650 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8651 			/* Must be a synchronous copy. */
8652 			nd->nd_repstat = NFSERR_NOTSUPP;
8653 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8654 		error = nfsm_loadattr(nd, outnap);
8655 		if (error == 0)
8656 			*outattrflagp = NFS_LATTR_NOSHRINK;
8657 		if (nd->nd_repstat == 0)
8658 			*lenp = len;
8659 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8660 		/*
8661 		 * For the case where consecutive is not supported, but
8662 		 * synchronous is supported, we can try consecutive == false
8663 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8664 		 * since Copy cannot be done.
8665 		 */
8666 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8667 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8668 			if (!consecutive || *++tl == newnfs_false)
8669 				nd->nd_repstat = NFSERR_NOTSUPP;
8670 		} else
8671 			nd->nd_repstat = NFSERR_BADXDR;
8672 	}
8673 	if (error == 0)
8674 		error = nd->nd_repstat;
8675 nfsmout:
8676 	m_freem(nd->nd_mrep);
8677 	return (error);
8678 }
8679 
8680 /*
8681  * Seek operation.
8682  */
8683 int
8684 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8685     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8686 {
8687 	int error, expireret = 0, retrycnt;
8688 	u_int32_t clidrev = 0;
8689 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
8690 	struct nfsnode *np = VTONFS(vp);
8691 	struct nfsfh *nfhp = NULL;
8692 	nfsv4stateid_t stateid;
8693 	void *lckp;
8694 
8695 	if (nmp->nm_clp != NULL)
8696 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8697 	nfhp = np->n_fhp;
8698 	retrycnt = 0;
8699 	do {
8700 		lckp = NULL;
8701 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8702 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8703 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8704 		    nap, attrflagp, cred);
8705 		if (error == NFSERR_STALESTATEID)
8706 			nfscl_initiate_recovery(nmp->nm_clp);
8707 		if (lckp != NULL)
8708 			nfscl_lockderef(lckp);
8709 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8710 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8711 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8712 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8713 		} else if ((error == NFSERR_EXPIRED || (!NFSHASINT(nmp) &&
8714 		    error == NFSERR_BADSTATEID)) && clidrev != 0) {
8715 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8716 			    curthread);
8717 		} else if (error == NFSERR_BADSTATEID && NFSHASINT(nmp)) {
8718 			error = EIO;
8719 		}
8720 		retrycnt++;
8721 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8722 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8723 	    error == NFSERR_BADSESSION ||
8724 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8725 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8726 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8727 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8728 	if (error && retrycnt >= 4)
8729 		error = EIO;
8730 	return (error);
8731 }
8732 
8733 /*
8734  * The seek RPC.
8735  */
8736 static int
8737 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8738     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8739 {
8740 	uint32_t *tl;
8741 	int error;
8742 	struct nfsrv_descript nfsd;
8743 	struct nfsrv_descript *nd = &nfsd;
8744 	nfsattrbit_t attrbits;
8745 
8746 	*attrflagp = 0;
8747 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp, cred);
8748 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8749 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8750 	txdr_hyper(*offp, tl); tl += 2;
8751 	*tl++ = txdr_unsigned(content);
8752 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8753 	NFSGETATTR_ATTRBIT(&attrbits);
8754 	nfsrv_putattrbit(nd, &attrbits);
8755 	error = nfscl_request(nd, vp, curthread, cred);
8756 	if (error != 0)
8757 		return (error);
8758 	if (nd->nd_repstat == 0) {
8759 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8760 		if (*tl++ == newnfs_true)
8761 			*eofp = true;
8762 		else
8763 			*eofp = false;
8764 		*offp = fxdr_hyper(tl);
8765 		/* Just skip over Getattr op status. */
8766 		error = nfsm_loadattr(nd, nap);
8767 		if (error == 0)
8768 			*attrflagp = 1;
8769 	}
8770 	error = nd->nd_repstat;
8771 nfsmout:
8772 	m_freem(nd->nd_mrep);
8773 	return (error);
8774 }
8775 
8776 /*
8777  * The getextattr RPC.
8778  */
8779 int
8780 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8781     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8782 {
8783 	uint32_t *tl;
8784 	int error;
8785 	struct nfsrv_descript nfsd;
8786 	struct nfsrv_descript *nd = &nfsd;
8787 	nfsattrbit_t attrbits;
8788 	uint32_t len, len2;
8789 
8790 	*attrflagp = 0;
8791 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp, cred);
8792 	nfsm_strtom(nd, name, strlen(name));
8793 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8794 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8795 	NFSGETATTR_ATTRBIT(&attrbits);
8796 	nfsrv_putattrbit(nd, &attrbits);
8797 	error = nfscl_request(nd, vp, p, cred);
8798 	if (error != 0)
8799 		return (error);
8800 	if (nd->nd_repstat == 0) {
8801 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8802 		len = fxdr_unsigned(uint32_t, *tl);
8803 		/* Sanity check lengths. */
8804 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8805 		    uiop->uio_resid <= UINT32_MAX) {
8806 			len2 = uiop->uio_resid;
8807 			if (len2 >= len)
8808 				error = nfsm_mbufuio(nd, uiop, len);
8809 			else {
8810 				error = nfsm_mbufuio(nd, uiop, len2);
8811 				if (error == 0) {
8812 					/*
8813 					 * nfsm_mbufuio() advances to a multiple
8814 					 * of 4, so round up len2 as well.  Then
8815 					 * we need to advance over the rest of
8816 					 * the data, rounding up the remaining
8817 					 * length.
8818 					 */
8819 					len2 = NFSM_RNDUP(len2);
8820 					len2 = NFSM_RNDUP(len - len2);
8821 					if (len2 > 0)
8822 						error = nfsm_advance(nd, len2,
8823 						    -1);
8824 				}
8825 			}
8826 		} else if (uiop == NULL && len > 0) {
8827 			/* Just wants the length and not the data. */
8828 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8829 		} else if (len > 0)
8830 			error = ENOATTR;
8831 		if (error != 0)
8832 			goto nfsmout;
8833 		*lenp = len;
8834 		/* Just skip over Getattr op status. */
8835 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8836 		error = nfsm_loadattr(nd, nap);
8837 		if (error == 0)
8838 			*attrflagp = 1;
8839 	}
8840 	if (error == 0)
8841 		error = nd->nd_repstat;
8842 nfsmout:
8843 	m_freem(nd->nd_mrep);
8844 	return (error);
8845 }
8846 
8847 /*
8848  * The setextattr RPC.
8849  */
8850 int
8851 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8852     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8853 {
8854 	uint32_t *tl;
8855 	int error;
8856 	struct nfsrv_descript nfsd;
8857 	struct nfsrv_descript *nd = &nfsd;
8858 	nfsattrbit_t attrbits;
8859 
8860 	*attrflagp = 0;
8861 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp, cred);
8862 	if (uiop->uio_resid > nd->nd_maxreq) {
8863 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8864 		m_freem(nd->nd_mreq);
8865 		return (EINVAL);
8866 	}
8867 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8868 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8869 	nfsm_strtom(nd, name, strlen(name));
8870 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8871 	*tl = txdr_unsigned(uiop->uio_resid);
8872 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8873 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8874 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8875 	NFSGETATTR_ATTRBIT(&attrbits);
8876 	nfsrv_putattrbit(nd, &attrbits);
8877 	error = nfscl_request(nd, vp, p, cred);
8878 	if (error != 0)
8879 		return (error);
8880 	if (nd->nd_repstat == 0) {
8881 		/* Just skip over the reply and Getattr op status. */
8882 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8883 		    NFSX_UNSIGNED);
8884 		error = nfsm_loadattr(nd, nap);
8885 		if (error == 0)
8886 			*attrflagp = 1;
8887 	}
8888 	if (error == 0)
8889 		error = nd->nd_repstat;
8890 nfsmout:
8891 	m_freem(nd->nd_mrep);
8892 	return (error);
8893 }
8894 
8895 /*
8896  * The removeextattr RPC.
8897  */
8898 int
8899 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8900     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8901 {
8902 	uint32_t *tl;
8903 	int error;
8904 	struct nfsrv_descript nfsd;
8905 	struct nfsrv_descript *nd = &nfsd;
8906 	nfsattrbit_t attrbits;
8907 
8908 	*attrflagp = 0;
8909 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp, cred);
8910 	nfsm_strtom(nd, name, strlen(name));
8911 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8912 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8913 	NFSGETATTR_ATTRBIT(&attrbits);
8914 	nfsrv_putattrbit(nd, &attrbits);
8915 	error = nfscl_request(nd, vp, p, cred);
8916 	if (error != 0)
8917 		return (error);
8918 	if (nd->nd_repstat == 0) {
8919 		/* Just skip over the reply and Getattr op status. */
8920 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8921 		    NFSX_UNSIGNED);
8922 		error = nfsm_loadattr(nd, nap);
8923 		if (error == 0)
8924 			*attrflagp = 1;
8925 	}
8926 	if (error == 0)
8927 		error = nd->nd_repstat;
8928 nfsmout:
8929 	m_freem(nd->nd_mrep);
8930 	return (error);
8931 }
8932 
8933 /*
8934  * The listextattr RPC.
8935  */
8936 int
8937 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8938     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8939     struct ucred *cred, NFSPROC_T *p)
8940 {
8941 	uint32_t *tl;
8942 	int cnt, error, i, len;
8943 	struct nfsrv_descript nfsd;
8944 	struct nfsrv_descript *nd = &nfsd;
8945 	nfsattrbit_t attrbits;
8946 	u_char c;
8947 
8948 	*attrflagp = 0;
8949 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp, cred);
8950 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8951 	txdr_hyper(*cookiep, tl); tl += 2;
8952 	*tl++ = txdr_unsigned(*lenp);
8953 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8954 	NFSGETATTR_ATTRBIT(&attrbits);
8955 	nfsrv_putattrbit(nd, &attrbits);
8956 	error = nfscl_request(nd, vp, p, cred);
8957 	if (error != 0)
8958 		return (error);
8959 	*eofp = true;
8960 	*lenp = 0;
8961 	if (nd->nd_repstat == 0) {
8962 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8963 		*cookiep = fxdr_hyper(tl); tl += 2;
8964 		cnt = fxdr_unsigned(int, *tl);
8965 		if (cnt < 0) {
8966 			error = EBADRPC;
8967 			goto nfsmout;
8968 		}
8969 		for (i = 0; i < cnt; i++) {
8970 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8971 			len = fxdr_unsigned(int, *tl);
8972 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8973 				error = EBADRPC;
8974 				goto nfsmout;
8975 			}
8976 			if (uiop == NULL)
8977 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8978 			else if (uiop->uio_resid >= len + 1) {
8979 				c = len;
8980 				error = uiomove(&c, sizeof(c), uiop);
8981 				if (error == 0)
8982 					error = nfsm_mbufuio(nd, uiop, len);
8983 			} else {
8984 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8985 				*eofp = false;
8986 			}
8987 			if (error != 0)
8988 				goto nfsmout;
8989 			*lenp += (len + 1);
8990 		}
8991 		/* Get the eof and skip over the Getattr op status. */
8992 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8993 		/*
8994 		 * *eofp is set false above, because it wasn't able to copy
8995 		 * all of the reply.
8996 		 */
8997 		if (*eofp && *tl == 0)
8998 			*eofp = false;
8999 		error = nfsm_loadattr(nd, nap);
9000 		if (error == 0)
9001 			*attrflagp = 1;
9002 	}
9003 	if (error == 0)
9004 		error = nd->nd_repstat;
9005 nfsmout:
9006 	m_freem(nd->nd_mrep);
9007 	return (error);
9008 }
9009 
9010 /*
9011  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
9012  */
9013 static struct mbuf *
9014 nfsm_split(struct mbuf *mp, uint64_t xfer)
9015 {
9016 	struct mbuf *m, *m2;
9017 	vm_page_t pg;
9018 	int i, j, left, pgno, plen, trim;
9019 	char *cp, *cp2;
9020 
9021 	if ((mp->m_flags & M_EXTPG) == 0) {
9022 		m = m_split(mp, xfer, M_WAITOK);
9023 		return (m);
9024 	}
9025 
9026 	/* Find the correct mbuf to split at. */
9027 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
9028 		xfer -= m->m_len;
9029 	if (m == NULL)
9030 		return (NULL);
9031 
9032 	/* If xfer == m->m_len, we can just split the mbuf list. */
9033 	if (xfer == m->m_len) {
9034 		m2 = m->m_next;
9035 		m->m_next = NULL;
9036 		return (m2);
9037 	}
9038 
9039 	/* Find the page to split at. */
9040 	pgno = 0;
9041 	left = xfer;
9042 	do {
9043 		if (pgno == 0)
9044 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
9045 		else
9046 			plen = m_epg_pagelen(m, pgno, 0);
9047 		if (left <= plen)
9048 			break;
9049 		left -= plen;
9050 		pgno++;
9051 	} while (pgno < m->m_epg_npgs);
9052 	if (pgno == m->m_epg_npgs)
9053 		panic("nfsm_split: eroneous ext_pgs mbuf");
9054 
9055 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
9056 	m2->m_epg_flags |= EPG_FLAG_ANON;
9057 
9058 	/*
9059 	 * If left < plen, allocate a new page for the new mbuf
9060 	 * and copy the data after left in the page to this new
9061 	 * page.
9062 	 */
9063 	if (left < plen) {
9064 		pg = vm_page_alloc_noobj(VM_ALLOC_WAITOK | VM_ALLOC_NODUMP |
9065 		    VM_ALLOC_WIRED);
9066 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
9067 		m2->m_epg_npgs = 1;
9068 
9069 		/* Copy the data after left to the new page. */
9070 		trim = plen - left;
9071 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
9072 		if (pgno == 0)
9073 			cp += m->m_epg_1st_off;
9074 		cp += left;
9075 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
9076 		if (pgno == m->m_epg_npgs - 1)
9077 			m2->m_epg_last_len = trim;
9078 		else {
9079 			cp2 += PAGE_SIZE - trim;
9080 			m2->m_epg_1st_off = PAGE_SIZE - trim;
9081 			m2->m_epg_last_len = m->m_epg_last_len;
9082 		}
9083 		memcpy(cp2, cp, trim);
9084 		m2->m_len = trim;
9085 	} else {
9086 		m2->m_len = 0;
9087 		m2->m_epg_last_len = m->m_epg_last_len;
9088 	}
9089 
9090 	/* Move the pages beyond pgno to the new mbuf. */
9091 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
9092 		m2->m_epg_pa[j] = m->m_epg_pa[i];
9093 		/* Never moves page 0. */
9094 		m2->m_len += m_epg_pagelen(m, i, 0);
9095 	}
9096 	m2->m_epg_npgs = j;
9097 	m->m_epg_npgs = pgno + 1;
9098 	m->m_epg_last_len = left;
9099 	m->m_len = xfer;
9100 
9101 	m2->m_next = m->m_next;
9102 	m->m_next = NULL;
9103 	return (m2);
9104 }
9105 
9106 /*
9107  * Do the NFSv4.1 Bind Connection to Session.
9108  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
9109  */
9110 void
9111 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
9112 {
9113 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
9114 	uint32_t res, *tl;
9115 	struct nfsrv_descript nfsd;
9116 	struct nfsrv_descript *nd = &nfsd;
9117 	struct rpc_callextra ext;
9118 	struct timeval utimeout;
9119 	enum clnt_stat stat;
9120 	int error;
9121 
9122 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
9123 	    NFS_VER4, rcp->minorvers, NULL);
9124 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
9125 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
9126 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9127 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
9128 	*tl = newnfs_false;
9129 
9130 	memset(&ext, 0, sizeof(ext));
9131 	utimeout.tv_sec = 30;
9132 	utimeout.tv_usec = 0;
9133 	ext.rc_auth = authunix_create(cr);
9134 	nd->nd_mrep = NULL;
9135 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
9136 	    &nd->nd_mrep, utimeout);
9137 	AUTH_DESTROY(ext.rc_auth);
9138 	if (stat != RPC_SUCCESS) {
9139 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
9140 		return;
9141 	}
9142 	if (nd->nd_mrep == NULL) {
9143 		printf("nfsrpc_bindconnsess: no reply args\n");
9144 		return;
9145 	}
9146 	error = 0;
9147 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
9148 	nd->nd_md = nd->nd_mrep;
9149 	nd->nd_dpos = mtod(nd->nd_md, char *);
9150 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
9151 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
9152 	if (nd->nd_repstat == NFSERR_OK) {
9153 		res = fxdr_unsigned(uint32_t, *tl);
9154 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
9155 		    -1)) != 0)
9156 			goto nfsmout;
9157 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
9158 		    4 * NFSX_UNSIGNED);
9159 		tl += 3;
9160 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
9161 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
9162 			res = fxdr_unsigned(uint32_t, *tl);
9163 			if (res != NFSCDFS4_BOTH)
9164 				printf("nfsrpc_bindconnsess: did not "
9165 				    "return FS4_BOTH\n");
9166 		} else
9167 			printf("nfsrpc_bindconnsess: not same "
9168 			    "sessionid\n");
9169 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
9170 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
9171 nfsmout:
9172 	if (error != 0)
9173 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
9174 	m_freem(nd->nd_mrep);
9175 }
9176