xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision 79ac3c12a714bcd3f2354c52d948aed9575c46d6)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Rick Macklem at The University of Guelph.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  */
35 
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
38 
39 /*
40  * Rpc op calls, generally called from the vnode op calls or through the
41  * buffer cache, for NFS v2, 3 and 4.
42  * These do not normally make any changes to vnode arguments or use
43  * structures that might change between the VFS variants. The returned
44  * arguments are all at the end, after the NFSPROC_T *p one.
45  */
46 
47 #include "opt_inet6.h"
48 
49 #include <fs/nfs/nfsport.h>
50 #include <fs/nfsclient/nfs.h>
51 #include <sys/extattr.h>
52 #include <sys/sysctl.h>
53 #include <sys/taskqueue.h>
54 
55 SYSCTL_DECL(_vfs_nfs);
56 
57 static int	nfsignore_eexist = 0;
58 SYSCTL_INT(_vfs_nfs, OID_AUTO, ignore_eexist, CTLFLAG_RW,
59     &nfsignore_eexist, 0, "NFS ignore EEXIST replies for mkdir/symlink");
60 
61 static int	nfscl_dssameconn = 0;
62 SYSCTL_INT(_vfs_nfs, OID_AUTO, dssameconn, CTLFLAG_RW,
63     &nfscl_dssameconn, 0, "Use same TCP connection to multiple DSs");
64 
65 /*
66  * Global variables
67  */
68 extern struct nfsstatsv1 nfsstatsv1;
69 extern int nfs_numnfscbd;
70 extern struct timeval nfsboottime;
71 extern u_int32_t newnfs_false, newnfs_true;
72 extern nfstype nfsv34_type[9];
73 extern int nfsrv_useacl;
74 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
75 extern int nfscl_debuglevel;
76 extern int nfs_pnfsiothreads;
77 extern u_long sb_max_adj;
78 extern int nfs_maxcopyrange;
79 NFSCLSTATEMUTEX;
80 int nfstest_outofseq = 0;
81 int nfscl_assumeposixlocks = 1;
82 int nfscl_enablecallb = 0;
83 short nfsv4_cbport = NFSV4_CBPORT;
84 int nfstest_openallsetattr = 0;
85 
86 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
87 
88 /*
89  * nfscl_getsameserver() can return one of three values:
90  * NFSDSP_USETHISSESSION - Use this session for the DS.
91  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
92  *     session.
93  * NFSDSP_NOTFOUND - No matching server was found.
94  */
95 enum nfsclds_state {
96 	NFSDSP_USETHISSESSION = 0,
97 	NFSDSP_SEQTHISSESSION = 1,
98 	NFSDSP_NOTFOUND = 2,
99 };
100 
101 /*
102  * Do a write RPC on a DS data file, using this structure for the arguments,
103  * so that this function can be executed by a separate kernel process.
104  */
105 struct nfsclwritedsdorpc {
106 	int			done;
107 	int			inprog;
108 	struct task		tsk;
109 	struct vnode		*vp;
110 	int			iomode;
111 	int			must_commit;
112 	nfsv4stateid_t		*stateidp;
113 	struct nfsclds		*dsp;
114 	uint64_t		off;
115 	int			len;
116 #ifdef notyet
117 	int			advise;
118 #endif
119 	struct nfsfh		*fhp;
120 	struct mbuf		*m;
121 	int			vers;
122 	int			minorvers;
123 	struct ucred		*cred;
124 	NFSPROC_T		*p;
125 	int			err;
126 };
127 
128 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
129     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
130 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
131     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
132 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
133     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
134     void *);
135 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
136     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
137     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
138 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
139     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
140     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
141     int *, void *, int *);
142 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
143     struct nfscllockowner *, u_int64_t, u_int64_t,
144     u_int32_t, struct ucred *, NFSPROC_T *, int);
145 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
146     struct acl *, nfsv4stateid_t *, void *);
147 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
148     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
149     struct ucred *, NFSPROC_T *);
150 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
151     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
152     NFSPROC_T *);
153 static void nfscl_initsessionslots(struct nfsclsession *);
154 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
155     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
156     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
157     NFSPROC_T *);
158 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
159     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
160     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
161     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
162 static int nfsrpc_readds(vnode_t, struct uio *, nfsv4stateid_t *, int *,
163     struct nfsclds *, uint64_t, int, struct nfsfh *, int, int, int,
164     struct ucred *, NFSPROC_T *);
165 static int nfsrpc_writeds(vnode_t, struct uio *, int *, int *,
166     nfsv4stateid_t *, struct nfsclds *, uint64_t, int,
167     struct nfsfh *, int, int, int, int, struct ucred *, NFSPROC_T *);
168 static int nfsio_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
169     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
170     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
171 static int nfsrpc_writedsmir(vnode_t, int *, int *, nfsv4stateid_t *,
172     struct nfsclds *, uint64_t, int, struct nfsfh *, struct mbuf *, int, int,
173     struct ucred *, NFSPROC_T *);
174 static enum nfsclds_state nfscl_getsameserver(struct nfsmount *,
175     struct nfsclds *, struct nfsclds **, uint32_t *);
176 static int nfsio_commitds(vnode_t, uint64_t, int, struct nfsclds *,
177     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
178     NFSPROC_T *);
179 static int nfsrpc_commitds(vnode_t, uint64_t, int, struct nfsclds *,
180     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
181 #ifdef notyet
182 static int nfsio_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
183     struct nfsfh *, int, int, struct nfsclwritedsdorpc *, struct ucred *,
184     NFSPROC_T *);
185 static int nfsrpc_adviseds(vnode_t, uint64_t, int, int, struct nfsclds *,
186     struct nfsfh *, int, int, struct ucred *, NFSPROC_T *);
187 #endif
188 static int nfsrpc_allocaterpc(vnode_t, off_t, off_t, nfsv4stateid_t *,
189     struct nfsvattr *, int *, struct ucred *, NFSPROC_T *, void *);
190 static void nfsrv_setuplayoutget(struct nfsrv_descript *, int, uint64_t,
191     uint64_t, uint64_t, nfsv4stateid_t *, int, int, int);
192 static int nfsrv_parseug(struct nfsrv_descript *, int, uid_t *, gid_t *,
193     NFSPROC_T *);
194 static int nfsrv_parselayoutget(struct nfsmount *, struct nfsrv_descript *,
195     nfsv4stateid_t *, int *, struct nfsclflayouthead *);
196 static int nfsrpc_getopenlayout(struct nfsmount *, vnode_t, u_int8_t *,
197     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
198     struct nfscldeleg **, struct ucred *, NFSPROC_T *);
199 static int nfsrpc_getcreatelayout(vnode_t, char *, int, struct vattr *,
200     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
201     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
202     struct nfsfh **, int *, int *, void *, int *);
203 static int nfsrpc_openlayoutrpc(struct nfsmount *, vnode_t, u_int8_t *,
204     int, uint8_t *, int, uint32_t, struct nfsclopen *, uint8_t *, int,
205     struct nfscldeleg **, nfsv4stateid_t *, int, int, int, int *,
206     struct nfsclflayouthead *, int *, struct ucred *, NFSPROC_T *);
207 static int nfsrpc_createlayout(vnode_t, char *, int, struct vattr *,
208     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **,
209     struct ucred *, NFSPROC_T *, struct nfsvattr *, struct nfsvattr *,
210     struct nfsfh **, int *, int *, void *, int *, nfsv4stateid_t *,
211     int, int, int, int *, struct nfsclflayouthead *, int *);
212 static int nfsrpc_layoutget(struct nfsmount *, uint8_t *, int, int, uint64_t,
213     uint64_t, uint64_t, int, int, nfsv4stateid_t *, int *,
214     struct nfsclflayouthead *, struct ucred *, NFSPROC_T *, void *);
215 static int nfsrpc_layoutgetres(struct nfsmount *, vnode_t, uint8_t *,
216     int, nfsv4stateid_t *, int, uint32_t *, struct nfscllayout **,
217     struct nfsclflayouthead *, int, int, int *, struct ucred *, NFSPROC_T *);
218 static int nfsrpc_copyrpc(vnode_t, off_t, vnode_t, off_t, size_t *,
219     nfsv4stateid_t *, nfsv4stateid_t *, struct nfsvattr *, int *,
220     struct nfsvattr *, int *, bool, int *, struct ucred *, NFSPROC_T *);
221 static int nfsrpc_seekrpc(vnode_t, off_t *, nfsv4stateid_t *, bool *,
222     int, struct nfsvattr *, int *, struct ucred *);
223 static struct mbuf *nfsm_split(struct mbuf *, uint64_t);
224 
225 int nfs_pnfsio(task_fn_t *, void *);
226 
227 /*
228  * nfs null call from vfs.
229  */
230 int
231 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
232 {
233 	int error;
234 	struct nfsrv_descript nfsd, *nd = &nfsd;
235 
236 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
237 	error = nfscl_request(nd, vp, p, cred, NULL);
238 	if (nd->nd_repstat && !error)
239 		error = nd->nd_repstat;
240 	m_freem(nd->nd_mrep);
241 	return (error);
242 }
243 
244 /*
245  * nfs access rpc op.
246  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
247  * modes are changed on the server, accesses might still fail later.
248  */
249 int
250 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
251     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
252 {
253 	int error;
254 	u_int32_t mode, rmode;
255 
256 	if (acmode & VREAD)
257 		mode = NFSACCESS_READ;
258 	else
259 		mode = 0;
260 	if (vnode_vtype(vp) == VDIR) {
261 		if (acmode & VWRITE)
262 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
263 				 NFSACCESS_DELETE);
264 		if (acmode & VEXEC)
265 			mode |= NFSACCESS_LOOKUP;
266 	} else {
267 		if (acmode & VWRITE)
268 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
269 		if (acmode & VEXEC)
270 			mode |= NFSACCESS_EXECUTE;
271 	}
272 
273 	/*
274 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
275 	 */
276 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
277 	    NULL);
278 
279 	/*
280 	 * The NFS V3 spec does not clarify whether or not
281 	 * the returned access bits can be a superset of
282 	 * the ones requested, so...
283 	 */
284 	if (!error && (rmode & mode) != mode)
285 		error = EACCES;
286 	return (error);
287 }
288 
289 /*
290  * The actual rpc, separated out for Darwin.
291  */
292 int
293 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
294     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
295     void *stuff)
296 {
297 	u_int32_t *tl;
298 	u_int32_t supported, rmode;
299 	int error;
300 	struct nfsrv_descript nfsd, *nd = &nfsd;
301 	nfsattrbit_t attrbits;
302 
303 	*attrflagp = 0;
304 	supported = mode;
305 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
306 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
307 	*tl = txdr_unsigned(mode);
308 	if (nd->nd_flag & ND_NFSV4) {
309 		/*
310 		 * And do a Getattr op.
311 		 */
312 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
313 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
314 		NFSGETATTR_ATTRBIT(&attrbits);
315 		(void) nfsrv_putattrbit(nd, &attrbits);
316 	}
317 	error = nfscl_request(nd, vp, p, cred, stuff);
318 	if (error)
319 		return (error);
320 	if (nd->nd_flag & ND_NFSV3) {
321 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
322 		if (error)
323 			goto nfsmout;
324 	}
325 	if (!nd->nd_repstat) {
326 		if (nd->nd_flag & ND_NFSV4) {
327 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
328 			supported = fxdr_unsigned(u_int32_t, *tl++);
329 		} else {
330 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
331 		}
332 		rmode = fxdr_unsigned(u_int32_t, *tl);
333 		if (nd->nd_flag & ND_NFSV4)
334 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
335 
336 		/*
337 		 * It's not obvious what should be done about
338 		 * unsupported access modes. For now, be paranoid
339 		 * and clear the unsupported ones.
340 		 */
341 		rmode &= supported;
342 		*rmodep = rmode;
343 	} else
344 		error = nd->nd_repstat;
345 nfsmout:
346 	m_freem(nd->nd_mrep);
347 	return (error);
348 }
349 
350 /*
351  * nfs open rpc
352  */
353 int
354 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
355 {
356 	struct nfsclopen *op;
357 	struct nfscldeleg *dp;
358 	struct nfsfh *nfhp;
359 	struct nfsnode *np = VTONFS(vp);
360 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
361 	u_int32_t mode, clidrev;
362 	int ret, newone, error, expireret = 0, retrycnt;
363 
364 	/*
365 	 * For NFSv4, Open Ops are only done on Regular Files.
366 	 */
367 	if (vnode_vtype(vp) != VREG)
368 		return (0);
369 	mode = 0;
370 	if (amode & FREAD)
371 		mode |= NFSV4OPEN_ACCESSREAD;
372 	if (amode & FWRITE)
373 		mode |= NFSV4OPEN_ACCESSWRITE;
374 	nfhp = np->n_fhp;
375 
376 	retrycnt = 0;
377 #ifdef notdef
378 { char name[100]; int namel;
379 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
380 bcopy(NFS4NODENAME(np->n_v4), name, namel);
381 name[namel] = '\0';
382 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
383 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
384 else printf(" fhl=0\n");
385 }
386 #endif
387 	do {
388 	    dp = NULL;
389 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
390 		cred, p, NULL, &op, &newone, &ret, 1);
391 	    if (error) {
392 		return (error);
393 	    }
394 	    if (nmp->nm_clp != NULL)
395 		clidrev = nmp->nm_clp->nfsc_clientidrev;
396 	    else
397 		clidrev = 0;
398 	    if (ret == NFSCLOPEN_DOOPEN) {
399 		if (np->n_v4 != NULL) {
400 			/*
401 			 * For the first attempt, try and get a layout, if
402 			 * pNFS is enabled for the mount.
403 			 */
404 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
405 			    nfs_numnfscbd == 0 ||
406 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
407 				error = nfsrpc_openrpc(nmp, vp,
408 				    np->n_v4->n4_data,
409 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
410 				    np->n_fhp->nfh_len, mode, op,
411 				    NFS4NODENAME(np->n_v4),
412 				    np->n_v4->n4_namelen,
413 				    &dp, 0, 0x0, cred, p, 0, 0);
414 			else
415 				error = nfsrpc_getopenlayout(nmp, vp,
416 				    np->n_v4->n4_data,
417 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
418 				    np->n_fhp->nfh_len, mode, op,
419 				    NFS4NODENAME(np->n_v4),
420 				    np->n_v4->n4_namelen, &dp, cred, p);
421 			if (dp != NULL) {
422 #ifdef APPLE
423 				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
424 #else
425 				NFSLOCKNODE(np);
426 				np->n_flag &= ~NDELEGMOD;
427 				/*
428 				 * Invalidate the attribute cache, so that
429 				 * attributes that pre-date the issue of a
430 				 * delegation are not cached, since the
431 				 * cached attributes will remain valid while
432 				 * the delegation is held.
433 				 */
434 				NFSINVALATTRCACHE(np);
435 				NFSUNLOCKNODE(np);
436 #endif
437 				(void) nfscl_deleg(nmp->nm_mountp,
438 				    op->nfso_own->nfsow_clp,
439 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
440 			}
441 		} else {
442 			error = EIO;
443 		}
444 		newnfs_copyincred(cred, &op->nfso_cred);
445 	    } else if (ret == NFSCLOPEN_SETCRED)
446 		/*
447 		 * This is a new local open on a delegation. It needs
448 		 * to have credentials so that an open can be done
449 		 * against the server during recovery.
450 		 */
451 		newnfs_copyincred(cred, &op->nfso_cred);
452 
453 	    /*
454 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
455 	     * been done on this Open successfully and a VOP_CLOSE()
456 	     * is expected for each of these.
457 	     * If error is non-zero, don't increment it, since the Open
458 	     * hasn't succeeded yet.
459 	     */
460 	    if (!error)
461 		op->nfso_opencnt++;
462 	    nfscl_openrelease(nmp, op, error, newone);
463 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
464 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
465 		error == NFSERR_BADSESSION) {
466 		(void) nfs_catnap(PZERO, error, "nfs_open");
467 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
468 		&& clidrev != 0) {
469 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
470 		retrycnt++;
471 	    }
472 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
473 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
474 	    error == NFSERR_BADSESSION ||
475 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
476 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
477 	if (error && retrycnt >= 4)
478 		error = EIO;
479 	return (error);
480 }
481 
482 /*
483  * the actual open rpc
484  */
485 int
486 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
487     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
488     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
489     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
490     int syscred, int recursed)
491 {
492 	u_int32_t *tl;
493 	struct nfsrv_descript nfsd, *nd = &nfsd;
494 	struct nfscldeleg *dp, *ndp = NULL;
495 	struct nfsvattr nfsva;
496 	u_int32_t rflags, deleg;
497 	nfsattrbit_t attrbits;
498 	int error, ret, acesize, limitby;
499 	struct nfsclsession *tsep;
500 
501 	dp = *dpp;
502 	*dpp = NULL;
503 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
504 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
505 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
506 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
507 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
508 	tsep = nfsmnt_mdssession(nmp);
509 	*tl++ = tsep->nfsess_clientid.lval[0];
510 	*tl = tsep->nfsess_clientid.lval[1];
511 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
512 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
513 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
514 	if (reclaim) {
515 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
516 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
517 		*tl = txdr_unsigned(delegtype);
518 	} else {
519 		if (dp != NULL) {
520 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
521 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
522 			if (NFSHASNFSV4N(nmp))
523 				*tl++ = 0;
524 			else
525 				*tl++ = dp->nfsdl_stateid.seqid;
526 			*tl++ = dp->nfsdl_stateid.other[0];
527 			*tl++ = dp->nfsdl_stateid.other[1];
528 			*tl = dp->nfsdl_stateid.other[2];
529 		} else {
530 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
531 		}
532 		(void) nfsm_strtom(nd, name, namelen);
533 	}
534 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
535 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
536 	NFSZERO_ATTRBIT(&attrbits);
537 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
538 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
539 	(void) nfsrv_putattrbit(nd, &attrbits);
540 	if (syscred)
541 		nd->nd_flag |= ND_USEGSSNAME;
542 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
543 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
544 	if (error)
545 		return (error);
546 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
547 	if (!nd->nd_repstat) {
548 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
549 		    6 * NFSX_UNSIGNED);
550 		op->nfso_stateid.seqid = *tl++;
551 		op->nfso_stateid.other[0] = *tl++;
552 		op->nfso_stateid.other[1] = *tl++;
553 		op->nfso_stateid.other[2] = *tl;
554 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
555 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
556 		if (error)
557 			goto nfsmout;
558 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
559 		deleg = fxdr_unsigned(u_int32_t, *tl);
560 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
561 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
562 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
563 			      NFSCLFLAGS_FIRSTDELEG))
564 				op->nfso_own->nfsow_clp->nfsc_flags |=
565 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
566 			ndp = malloc(
567 			    sizeof (struct nfscldeleg) + newfhlen,
568 			    M_NFSCLDELEG, M_WAITOK);
569 			LIST_INIT(&ndp->nfsdl_owner);
570 			LIST_INIT(&ndp->nfsdl_lock);
571 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
572 			ndp->nfsdl_fhlen = newfhlen;
573 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
574 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
575 			nfscl_lockinit(&ndp->nfsdl_rwlock);
576 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
577 			    NFSX_UNSIGNED);
578 			ndp->nfsdl_stateid.seqid = *tl++;
579 			ndp->nfsdl_stateid.other[0] = *tl++;
580 			ndp->nfsdl_stateid.other[1] = *tl++;
581 			ndp->nfsdl_stateid.other[2] = *tl++;
582 			ret = fxdr_unsigned(int, *tl);
583 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
584 				ndp->nfsdl_flags = NFSCLDL_WRITE;
585 				/*
586 				 * Indicates how much the file can grow.
587 				 */
588 				NFSM_DISSECT(tl, u_int32_t *,
589 				    3 * NFSX_UNSIGNED);
590 				limitby = fxdr_unsigned(int, *tl++);
591 				switch (limitby) {
592 				case NFSV4OPEN_LIMITSIZE:
593 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
594 					break;
595 				case NFSV4OPEN_LIMITBLOCKS:
596 					ndp->nfsdl_sizelimit =
597 					    fxdr_unsigned(u_int64_t, *tl++);
598 					ndp->nfsdl_sizelimit *=
599 					    fxdr_unsigned(u_int64_t, *tl);
600 					break;
601 				default:
602 					error = NFSERR_BADXDR;
603 					goto nfsmout;
604 				}
605 			} else {
606 				ndp->nfsdl_flags = NFSCLDL_READ;
607 			}
608 			if (ret)
609 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
610 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
611 			    &acesize, p);
612 			if (error)
613 				goto nfsmout;
614 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
615 			error = NFSERR_BADXDR;
616 			goto nfsmout;
617 		}
618 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
619 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
620 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
621 		    NULL, NULL, NULL, p, cred);
622 		if (error)
623 			goto nfsmout;
624 		if (ndp != NULL) {
625 			ndp->nfsdl_change = nfsva.na_filerev;
626 			ndp->nfsdl_modtime = nfsva.na_mtime;
627 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
628 		}
629 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
630 		    do {
631 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
632 			    cred, p);
633 			if (ret == NFSERR_DELAY)
634 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
635 		    } while (ret == NFSERR_DELAY);
636 		    error = ret;
637 		}
638 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
639 		    nfscl_assumeposixlocks)
640 		    op->nfso_posixlock = 1;
641 		else
642 		    op->nfso_posixlock = 0;
643 
644 		/*
645 		 * If the server is handing out delegations, but we didn't
646 		 * get one because an OpenConfirm was required, try the
647 		 * Open again, to get a delegation. This is a harmless no-op,
648 		 * from a server's point of view.
649 		 */
650 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
651 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
652 		    && !error && dp == NULL && ndp == NULL && !recursed) {
653 		    do {
654 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
655 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
656 			    cred, p, syscred, 1);
657 			if (ret == NFSERR_DELAY)
658 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
659 		    } while (ret == NFSERR_DELAY);
660 		    if (ret) {
661 			if (ndp != NULL) {
662 				free(ndp, M_NFSCLDELEG);
663 				ndp = NULL;
664 			}
665 			if (ret == NFSERR_STALECLIENTID ||
666 			    ret == NFSERR_STALEDONTRECOVER ||
667 			    ret == NFSERR_BADSESSION)
668 				error = ret;
669 		    }
670 		}
671 	}
672 	if (nd->nd_repstat != 0 && error == 0)
673 		error = nd->nd_repstat;
674 	if (error == NFSERR_STALECLIENTID)
675 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
676 nfsmout:
677 	if (!error)
678 		*dpp = ndp;
679 	else if (ndp != NULL)
680 		free(ndp, M_NFSCLDELEG);
681 	m_freem(nd->nd_mrep);
682 	return (error);
683 }
684 
685 /*
686  * open downgrade rpc
687  */
688 int
689 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
690     struct ucred *cred, NFSPROC_T *p)
691 {
692 	u_int32_t *tl;
693 	struct nfsrv_descript nfsd, *nd = &nfsd;
694 	int error;
695 
696 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
697 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
698 	if (NFSHASNFSV4N(VFSTONFS(vp->v_mount)))
699 		*tl++ = 0;
700 	else
701 		*tl++ = op->nfso_stateid.seqid;
702 	*tl++ = op->nfso_stateid.other[0];
703 	*tl++ = op->nfso_stateid.other[1];
704 	*tl++ = op->nfso_stateid.other[2];
705 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
706 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
707 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
708 	error = nfscl_request(nd, vp, p, cred, NULL);
709 	if (error)
710 		return (error);
711 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
712 	if (!nd->nd_repstat) {
713 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
714 		op->nfso_stateid.seqid = *tl++;
715 		op->nfso_stateid.other[0] = *tl++;
716 		op->nfso_stateid.other[1] = *tl++;
717 		op->nfso_stateid.other[2] = *tl;
718 	}
719 	if (nd->nd_repstat && error == 0)
720 		error = nd->nd_repstat;
721 	if (error == NFSERR_STALESTATEID)
722 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
723 nfsmout:
724 	m_freem(nd->nd_mrep);
725 	return (error);
726 }
727 
728 /*
729  * V4 Close operation.
730  */
731 int
732 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
733 {
734 	struct nfsclclient *clp;
735 	int error;
736 
737 	if (vnode_vtype(vp) != VREG)
738 		return (0);
739 	if (doclose)
740 		error = nfscl_doclose(vp, &clp, p);
741 	else
742 		error = nfscl_getclose(vp, &clp);
743 	if (error)
744 		return (error);
745 
746 	nfscl_clientrelease(clp);
747 	return (0);
748 }
749 
750 /*
751  * Close the open.
752  */
753 void
754 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
755 {
756 	struct nfsrv_descript nfsd, *nd = &nfsd;
757 	struct nfscllockowner *lp, *nlp;
758 	struct nfscllock *lop, *nlop;
759 	struct ucred *tcred;
760 	u_int64_t off = 0, len = 0;
761 	u_int32_t type = NFSV4LOCKT_READ;
762 	int error, do_unlock, trycnt;
763 
764 	tcred = newnfs_getcred();
765 	newnfs_copycred(&op->nfso_cred, tcred);
766 	/*
767 	 * (Theoretically this could be done in the same
768 	 *  compound as the close, but having multiple
769 	 *  sequenced Ops in the same compound might be
770 	 *  too scary for some servers.)
771 	 */
772 	if (op->nfso_posixlock) {
773 		off = 0;
774 		len = NFS64BITSSET;
775 		type = NFSV4LOCKT_READ;
776 	}
777 
778 	/*
779 	 * Since this function is only called from VOP_INACTIVE(), no
780 	 * other thread will be manipulating this Open. As such, the
781 	 * lock lists are not being changed by other threads, so it should
782 	 * be safe to do this without locking.
783 	 */
784 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
785 		do_unlock = 1;
786 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
787 			if (op->nfso_posixlock == 0) {
788 				off = lop->nfslo_first;
789 				len = lop->nfslo_end - lop->nfslo_first;
790 				if (lop->nfslo_type == F_WRLCK)
791 					type = NFSV4LOCKT_WRITE;
792 				else
793 					type = NFSV4LOCKT_READ;
794 			}
795 			if (do_unlock) {
796 				trycnt = 0;
797 				do {
798 					error = nfsrpc_locku(nd, nmp, lp, off,
799 					    len, type, tcred, p, 0);
800 					if ((nd->nd_repstat == NFSERR_GRACE ||
801 					    nd->nd_repstat == NFSERR_DELAY) &&
802 					    error == 0)
803 						(void) nfs_catnap(PZERO,
804 						    (int)nd->nd_repstat,
805 						    "nfs_close");
806 				} while ((nd->nd_repstat == NFSERR_GRACE ||
807 				    nd->nd_repstat == NFSERR_DELAY) &&
808 				    error == 0 && trycnt++ < 5);
809 				if (op->nfso_posixlock)
810 					do_unlock = 0;
811 			}
812 			nfscl_freelock(lop, 0);
813 		}
814 		/*
815 		 * Do a ReleaseLockOwner.
816 		 * The lock owner name nfsl_owner may be used by other opens for
817 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
818 		 * puts on the wire has the file handle for this file appended
819 		 * to it, so it can be done now.
820 		 */
821 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
822 		    lp->nfsl_open->nfso_fhlen, tcred, p);
823 	}
824 
825 	/*
826 	 * There could be other Opens for different files on the same
827 	 * OpenOwner, so locking is required.
828 	 */
829 	NFSLOCKCLSTATE();
830 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
831 	NFSUNLOCKCLSTATE();
832 	do {
833 		error = nfscl_tryclose(op, tcred, nmp, p);
834 		if (error == NFSERR_GRACE)
835 			(void) nfs_catnap(PZERO, error, "nfs_close");
836 	} while (error == NFSERR_GRACE);
837 	NFSLOCKCLSTATE();
838 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
839 
840 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
841 		nfscl_freelockowner(lp, 0);
842 	nfscl_freeopen(op, 0);
843 	NFSUNLOCKCLSTATE();
844 	NFSFREECRED(tcred);
845 }
846 
847 /*
848  * The actual Close RPC.
849  */
850 int
851 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
852     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
853     int syscred)
854 {
855 	u_int32_t *tl;
856 	int error;
857 
858 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
859 	    op->nfso_fhlen, NULL, NULL, 0, 0);
860 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
861 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
862 	if (NFSHASNFSV4N(nmp))
863 		*tl++ = 0;
864 	else
865 		*tl++ = op->nfso_stateid.seqid;
866 	*tl++ = op->nfso_stateid.other[0];
867 	*tl++ = op->nfso_stateid.other[1];
868 	*tl = op->nfso_stateid.other[2];
869 	if (syscred)
870 		nd->nd_flag |= ND_USEGSSNAME;
871 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
872 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
873 	if (error)
874 		return (error);
875 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
876 	if (nd->nd_repstat == 0)
877 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
878 	error = nd->nd_repstat;
879 	if (error == NFSERR_STALESTATEID)
880 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
881 nfsmout:
882 	m_freem(nd->nd_mrep);
883 	return (error);
884 }
885 
886 /*
887  * V4 Open Confirm RPC.
888  */
889 int
890 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
891     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
892 {
893 	u_int32_t *tl;
894 	struct nfsrv_descript nfsd, *nd = &nfsd;
895 	struct nfsmount *nmp;
896 	int error;
897 
898 	nmp = VFSTONFS(vp->v_mount);
899 	if (NFSHASNFSV4N(nmp))
900 		return (0);		/* No confirmation for NFSv4.1. */
901 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
902 	    0, 0);
903 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
904 	*tl++ = op->nfso_stateid.seqid;
905 	*tl++ = op->nfso_stateid.other[0];
906 	*tl++ = op->nfso_stateid.other[1];
907 	*tl++ = op->nfso_stateid.other[2];
908 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
909 	error = nfscl_request(nd, vp, p, cred, NULL);
910 	if (error)
911 		return (error);
912 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
913 	if (!nd->nd_repstat) {
914 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
915 		op->nfso_stateid.seqid = *tl++;
916 		op->nfso_stateid.other[0] = *tl++;
917 		op->nfso_stateid.other[1] = *tl++;
918 		op->nfso_stateid.other[2] = *tl;
919 	}
920 	error = nd->nd_repstat;
921 	if (error == NFSERR_STALESTATEID)
922 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
923 nfsmout:
924 	m_freem(nd->nd_mrep);
925 	return (error);
926 }
927 
928 /*
929  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
930  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
931  */
932 int
933 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
934     bool *retokp, struct ucred *cred, NFSPROC_T *p)
935 {
936 	u_int32_t *tl;
937 	struct nfsrv_descript nfsd;
938 	struct nfsrv_descript *nd = &nfsd;
939 	u_int8_t *cp = NULL, *cp2, addr[INET6_ADDRSTRLEN + 9];
940 	u_short port;
941 	int error, isinet6 = 0, callblen;
942 	nfsquad_t confirm;
943 	static u_int32_t rev = 0;
944 	struct nfsclds *dsp, *odsp;
945 	struct in6_addr a6;
946 	struct nfsclsession *tsep;
947 	struct rpc_reconupcall recon;
948 	struct nfscl_reconarg *rcp;
949 
950 	if (nfsboottime.tv_sec == 0)
951 		NFSSETBOOTTIME(nfsboottime);
952 	if (NFSHASNFSV4N(nmp)) {
953 		error = NFSERR_BADSESSION;
954 		odsp = dsp = NULL;
955 		if (retokp != NULL) {
956 			NFSLOCKMNT(nmp);
957 			odsp = TAILQ_FIRST(&nmp->nm_sess);
958 			NFSUNLOCKMNT(nmp);
959 		}
960 		if (odsp != NULL) {
961 			/*
962 			 * When a session already exists, first try a
963 			 * CreateSession with the extant ClientID.
964 			 */
965 			dsp = malloc(sizeof(struct nfsclds) +
966 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
967 			    M_WAITOK | M_ZERO);
968 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
969 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
970 			dsp->nfsclds_sess.nfsess_clientid =
971 			    odsp->nfsclds_sess.nfsess_clientid;
972 			dsp->nfsclds_sess.nfsess_sequenceid =
973 			    odsp->nfsclds_sess.nfsess_sequenceid;
974 			dsp->nfsclds_flags = odsp->nfsclds_flags;
975 			if (dsp->nfsclds_servownlen > 0)
976 				memcpy(dsp->nfsclds_serverown,
977 				    odsp->nfsclds_serverown,
978 				    dsp->nfsclds_servownlen + 1);
979 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
980 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
981 			    NULL, MTX_DEF);
982 			nfscl_initsessionslots(&dsp->nfsclds_sess);
983 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
984 			    &nmp->nm_sockreq, NULL,
985 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
986 			NFSCL_DEBUG(1, "create session for extant "
987 			    "ClientID=%d\n", error);
988 			if (error != 0) {
989 				nfscl_freenfsclds(dsp);
990 				dsp = NULL;
991 				/*
992 				 * If *retokp is true, return any error other
993 				 * than NFSERR_STALECLIENTID,
994 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
995 				 * so that nfscl_recover() will not loop.
996 				 */
997 				if (*retokp)
998 					return (NFSERR_IO);
999 			} else
1000 				*retokp = true;
1001 		} else if (retokp != NULL && *retokp)
1002 			return (NFSERR_IO);
1003 		if (error != 0) {
1004 			/*
1005 			 * Either there was no previous session or the
1006 			 * CreateSession attempt failed, so...
1007 			 * do an ExchangeID followed by the CreateSession.
1008 			 */
1009 			clp->nfsc_rev = rev++;
1010 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1011 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1012 			    cred, p);
1013 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1014 			if (error == 0)
1015 				error = nfsrpc_createsession(nmp,
1016 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1017 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1018 				    cred, p);
1019 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1020 		}
1021 		if (error == 0) {
1022 			/*
1023 			 * If the session supports a backchannel, set up
1024 			 * the BindConnectionToSession call in the krpc
1025 			 * so that it is done on a reconnection.
1026 			 */
1027 			if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0) {
1028 				rcp = mem_alloc(sizeof(*rcp));
1029 				rcp->minorvers = nmp->nm_minorvers;
1030 				memcpy(rcp->sessionid,
1031 				    dsp->nfsclds_sess.nfsess_sessionid,
1032 				    NFSX_V4SESSIONID);
1033 				recon.call = nfsrpc_bindconnsess;
1034 				recon.arg = rcp;
1035 				CLNT_CONTROL(nmp->nm_client, CLSET_RECONUPCALL,
1036 				    &recon);
1037 			}
1038 
1039 			NFSLOCKMNT(nmp);
1040 			/*
1041 			 * The old sessions cannot be safely free'd
1042 			 * here, since they may still be used by
1043 			 * in-progress RPCs.
1044 			 */
1045 			tsep = NULL;
1046 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
1047 				tsep = NFSMNT_MDSSESSION(nmp);
1048 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1049 			    nfsclds_list);
1050 			/*
1051 			 * Wake up RPCs waiting for a slot on the
1052 			 * old session. These will then fail with
1053 			 * NFSERR_BADSESSION and be retried with the
1054 			 * new session by nfsv4_setsequence().
1055 			 * Also wakeup() processes waiting for the
1056 			 * new session.
1057 			 */
1058 			if (tsep != NULL)
1059 				wakeup(&tsep->nfsess_slots);
1060 			wakeup(&nmp->nm_sess);
1061 			NFSUNLOCKMNT(nmp);
1062 		} else if (dsp != NULL)
1063 			nfscl_freenfsclds(dsp);
1064 		if (error == 0 && reclaim == 0) {
1065 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1066 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1067 			if (error == NFSERR_COMPLETEALREADY ||
1068 			    error == NFSERR_NOTSUPP)
1069 				/* Ignore this error. */
1070 				error = 0;
1071 		}
1072 		return (error);
1073 	} else if (retokp != NULL && *retokp)
1074 		return (NFSERR_IO);
1075 	clp->nfsc_rev = rev++;
1076 
1077 	/*
1078 	 * Allocate a single session structure for NFSv4.0, because some of
1079 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1080 	 */
1081 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1082 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1083 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1084 	NFSLOCKMNT(nmp);
1085 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1086 	tsep = NFSMNT_MDSSESSION(nmp);
1087 	NFSUNLOCKMNT(nmp);
1088 
1089 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
1090 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1091 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1092 	*tl = txdr_unsigned(clp->nfsc_rev);
1093 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1094 
1095 	/*
1096 	 * set up the callback address
1097 	 */
1098 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1099 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1100 	callblen = strlen(nfsv4_callbackaddr);
1101 	if (callblen == 0)
1102 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1103 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1104 	    (callblen > 0 || cp != NULL)) {
1105 		port = htons(nfsv4_cbport);
1106 		cp2 = (u_int8_t *)&port;
1107 #ifdef INET6
1108 		if ((callblen > 0 &&
1109 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1110 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1111 
1112 			(void) nfsm_strtom(nd, "tcp6", 4);
1113 			if (callblen == 0) {
1114 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1115 				ip6add = ip6buf;
1116 			} else {
1117 				ip6add = nfsv4_callbackaddr;
1118 			}
1119 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1120 			    ip6add, cp2[0], cp2[1]);
1121 		} else
1122 #endif
1123 		{
1124 			(void) nfsm_strtom(nd, "tcp", 3);
1125 			if (callblen == 0)
1126 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1127 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1128 				    cp[2], cp[3], cp2[0], cp2[1]);
1129 			else
1130 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1131 				    "%s.%d.%d", nfsv4_callbackaddr,
1132 				    cp2[0], cp2[1]);
1133 		}
1134 		(void) nfsm_strtom(nd, addr, strlen(addr));
1135 	} else {
1136 		(void) nfsm_strtom(nd, "tcp", 3);
1137 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1138 	}
1139 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1140 	*tl = txdr_unsigned(clp->nfsc_cbident);
1141 	nd->nd_flag |= ND_USEGSSNAME;
1142 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1143 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1144 	if (error)
1145 		return (error);
1146 	if (nd->nd_repstat == 0) {
1147 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1148 	    tsep->nfsess_clientid.lval[0] = *tl++;
1149 	    tsep->nfsess_clientid.lval[1] = *tl++;
1150 	    confirm.lval[0] = *tl++;
1151 	    confirm.lval[1] = *tl;
1152 	    m_freem(nd->nd_mrep);
1153 	    nd->nd_mrep = NULL;
1154 
1155 	    /*
1156 	     * and confirm it.
1157 	     */
1158 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1159 		NULL, 0, 0);
1160 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1161 	    *tl++ = tsep->nfsess_clientid.lval[0];
1162 	    *tl++ = tsep->nfsess_clientid.lval[1];
1163 	    *tl++ = confirm.lval[0];
1164 	    *tl = confirm.lval[1];
1165 	    nd->nd_flag |= ND_USEGSSNAME;
1166 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1167 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1168 	    if (error)
1169 		return (error);
1170 	    m_freem(nd->nd_mrep);
1171 	    nd->nd_mrep = NULL;
1172 	}
1173 	error = nd->nd_repstat;
1174 nfsmout:
1175 	m_freem(nd->nd_mrep);
1176 	return (error);
1177 }
1178 
1179 /*
1180  * nfs getattr call.
1181  */
1182 int
1183 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1184     struct nfsvattr *nap, void *stuff)
1185 {
1186 	struct nfsrv_descript nfsd, *nd = &nfsd;
1187 	int error;
1188 	nfsattrbit_t attrbits;
1189 
1190 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1191 	if (nd->nd_flag & ND_NFSV4) {
1192 		NFSGETATTR_ATTRBIT(&attrbits);
1193 		(void) nfsrv_putattrbit(nd, &attrbits);
1194 	}
1195 	error = nfscl_request(nd, vp, p, cred, stuff);
1196 	if (error)
1197 		return (error);
1198 	if (!nd->nd_repstat)
1199 		error = nfsm_loadattr(nd, nap);
1200 	else
1201 		error = nd->nd_repstat;
1202 	m_freem(nd->nd_mrep);
1203 	return (error);
1204 }
1205 
1206 /*
1207  * nfs getattr call with non-vnode arguments.
1208  */
1209 int
1210 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1211     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1212     uint32_t *leasep)
1213 {
1214 	struct nfsrv_descript nfsd, *nd = &nfsd;
1215 	int error, vers = NFS_VER2;
1216 	nfsattrbit_t attrbits;
1217 
1218 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1219 	if (nd->nd_flag & ND_NFSV4) {
1220 		vers = NFS_VER4;
1221 		NFSGETATTR_ATTRBIT(&attrbits);
1222 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1223 		(void) nfsrv_putattrbit(nd, &attrbits);
1224 	} else if (nd->nd_flag & ND_NFSV3) {
1225 		vers = NFS_VER3;
1226 	}
1227 	if (syscred)
1228 		nd->nd_flag |= ND_USEGSSNAME;
1229 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1230 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1231 	if (error)
1232 		return (error);
1233 	if (nd->nd_repstat == 0) {
1234 		if ((nd->nd_flag & ND_NFSV4) != 0)
1235 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1236 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1237 			    NULL, NULL);
1238 		else
1239 			error = nfsm_loadattr(nd, nap);
1240 	} else
1241 		error = nd->nd_repstat;
1242 	m_freem(nd->nd_mrep);
1243 	return (error);
1244 }
1245 
1246 /*
1247  * Do an nfs setattr operation.
1248  */
1249 int
1250 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1251     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1252     void *stuff)
1253 {
1254 	int error, expireret = 0, openerr, retrycnt;
1255 	u_int32_t clidrev = 0, mode;
1256 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1257 	struct nfsfh *nfhp;
1258 	nfsv4stateid_t stateid;
1259 	void *lckp;
1260 
1261 	if (nmp->nm_clp != NULL)
1262 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1263 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1264 		mode = NFSV4OPEN_ACCESSWRITE;
1265 	else
1266 		mode = NFSV4OPEN_ACCESSREAD;
1267 	retrycnt = 0;
1268 	do {
1269 		lckp = NULL;
1270 		openerr = 1;
1271 		if (NFSHASNFSV4(nmp)) {
1272 			nfhp = VTONFS(vp)->n_fhp;
1273 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1274 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1275 			if (error && vnode_vtype(vp) == VREG &&
1276 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1277 			     nfstest_openallsetattr)) {
1278 				/*
1279 				 * No Open stateid, so try and open the file
1280 				 * now.
1281 				 */
1282 				if (mode == NFSV4OPEN_ACCESSWRITE)
1283 					openerr = nfsrpc_open(vp, FWRITE, cred,
1284 					    p);
1285 				else
1286 					openerr = nfsrpc_open(vp, FREAD, cred,
1287 					    p);
1288 				if (!openerr)
1289 					(void) nfscl_getstateid(vp,
1290 					    nfhp->nfh_fh, nfhp->nfh_len,
1291 					    mode, 0, cred, p, &stateid, &lckp);
1292 			}
1293 		}
1294 		if (vap != NULL)
1295 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1296 			    rnap, attrflagp, stuff);
1297 		else
1298 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1299 			    stuff);
1300 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1301 			NFSLOCKMNT(nmp);
1302 			nmp->nm_state |= NFSSTA_OPENMODE;
1303 			NFSUNLOCKMNT(nmp);
1304 		}
1305 		if (error == NFSERR_STALESTATEID)
1306 			nfscl_initiate_recovery(nmp->nm_clp);
1307 		if (lckp != NULL)
1308 			nfscl_lockderef(lckp);
1309 		if (!openerr)
1310 			(void) nfsrpc_close(vp, 0, p);
1311 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1312 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1313 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1314 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1315 		} else if ((error == NFSERR_EXPIRED ||
1316 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1317 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1318 		}
1319 		retrycnt++;
1320 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1321 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1322 	    error == NFSERR_BADSESSION ||
1323 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1324 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1325 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1326 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1327 	     retrycnt < 4));
1328 	if (error && retrycnt >= 4)
1329 		error = EIO;
1330 	return (error);
1331 }
1332 
1333 static int
1334 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1335     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1336     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1337 {
1338 	u_int32_t *tl;
1339 	struct nfsrv_descript nfsd, *nd = &nfsd;
1340 	int error;
1341 	nfsattrbit_t attrbits;
1342 
1343 	*attrflagp = 0;
1344 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1345 	if (nd->nd_flag & ND_NFSV4)
1346 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1347 	vap->va_type = vnode_vtype(vp);
1348 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1349 	if (nd->nd_flag & ND_NFSV3) {
1350 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1351 		*tl = newnfs_false;
1352 	} else if (nd->nd_flag & ND_NFSV4) {
1353 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1354 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1355 		NFSGETATTR_ATTRBIT(&attrbits);
1356 		(void) nfsrv_putattrbit(nd, &attrbits);
1357 	}
1358 	error = nfscl_request(nd, vp, p, cred, stuff);
1359 	if (error)
1360 		return (error);
1361 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1362 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1363 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1364 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1365 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1366 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1367 	m_freem(nd->nd_mrep);
1368 	if (nd->nd_repstat && !error)
1369 		error = nd->nd_repstat;
1370 	return (error);
1371 }
1372 
1373 /*
1374  * nfs lookup rpc
1375  */
1376 int
1377 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1378     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1379     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1380 {
1381 	u_int32_t *tl;
1382 	struct nfsrv_descript nfsd, *nd = &nfsd;
1383 	struct nfsmount *nmp;
1384 	struct nfsnode *np;
1385 	struct nfsfh *nfhp;
1386 	nfsattrbit_t attrbits;
1387 	int error = 0, lookupp = 0;
1388 
1389 	*attrflagp = 0;
1390 	*dattrflagp = 0;
1391 	if (vnode_vtype(dvp) != VDIR)
1392 		return (ENOTDIR);
1393 	nmp = VFSTONFS(dvp->v_mount);
1394 	if (len > NFS_MAXNAMLEN)
1395 		return (ENAMETOOLONG);
1396 	if (NFSHASNFSV4(nmp) && len == 1 &&
1397 		name[0] == '.') {
1398 		/*
1399 		 * Just return the current dir's fh.
1400 		 */
1401 		np = VTONFS(dvp);
1402 		nfhp = malloc(sizeof (struct nfsfh) +
1403 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1404 		nfhp->nfh_len = np->n_fhp->nfh_len;
1405 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1406 		*nfhpp = nfhp;
1407 		return (0);
1408 	}
1409 	if (NFSHASNFSV4(nmp) && len == 2 &&
1410 		name[0] == '.' && name[1] == '.') {
1411 		lookupp = 1;
1412 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1413 	} else {
1414 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1415 		(void) nfsm_strtom(nd, name, len);
1416 	}
1417 	if (nd->nd_flag & ND_NFSV4) {
1418 		NFSGETATTR_ATTRBIT(&attrbits);
1419 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1420 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1421 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1422 		(void) nfsrv_putattrbit(nd, &attrbits);
1423 	}
1424 	error = nfscl_request(nd, dvp, p, cred, stuff);
1425 	if (error)
1426 		return (error);
1427 	if (nd->nd_repstat) {
1428 		/*
1429 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1430 		 * the lookup is at the root of an fs, so return this dir.
1431 		 */
1432 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1433 		    np = VTONFS(dvp);
1434 		    nfhp = malloc(sizeof (struct nfsfh) +
1435 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1436 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1437 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1438 		    *nfhpp = nfhp;
1439 		    m_freem(nd->nd_mrep);
1440 		    return (0);
1441 		}
1442 		if (nd->nd_flag & ND_NFSV3)
1443 		    error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1444 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1445 		    ND_NFSV4) {
1446 			/* Load the directory attributes. */
1447 			error = nfsm_loadattr(nd, dnap);
1448 			if (error == 0)
1449 				*dattrflagp = 1;
1450 		}
1451 		goto nfsmout;
1452 	}
1453 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1454 		/* Load the directory attributes. */
1455 		error = nfsm_loadattr(nd, dnap);
1456 		if (error != 0)
1457 			goto nfsmout;
1458 		*dattrflagp = 1;
1459 		/* Skip over the Lookup and GetFH operation status values. */
1460 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1461 	}
1462 	error = nfsm_getfh(nd, nfhpp);
1463 	if (error)
1464 		goto nfsmout;
1465 
1466 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1467 	if ((nd->nd_flag & ND_NFSV3) && !error)
1468 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1469 nfsmout:
1470 	m_freem(nd->nd_mrep);
1471 	if (!error && nd->nd_repstat)
1472 		error = nd->nd_repstat;
1473 	return (error);
1474 }
1475 
1476 /*
1477  * Do a readlink rpc.
1478  */
1479 int
1480 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1481     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1482 {
1483 	u_int32_t *tl;
1484 	struct nfsrv_descript nfsd, *nd = &nfsd;
1485 	struct nfsnode *np = VTONFS(vp);
1486 	nfsattrbit_t attrbits;
1487 	int error, len, cangetattr = 1;
1488 
1489 	*attrflagp = 0;
1490 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1491 	if (nd->nd_flag & ND_NFSV4) {
1492 		/*
1493 		 * And do a Getattr op.
1494 		 */
1495 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1496 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1497 		NFSGETATTR_ATTRBIT(&attrbits);
1498 		(void) nfsrv_putattrbit(nd, &attrbits);
1499 	}
1500 	error = nfscl_request(nd, vp, p, cred, stuff);
1501 	if (error)
1502 		return (error);
1503 	if (nd->nd_flag & ND_NFSV3)
1504 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1505 	if (!nd->nd_repstat && !error) {
1506 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1507 		/*
1508 		 * This seems weird to me, but must have been added to
1509 		 * FreeBSD for some reason. The only thing I can think of
1510 		 * is that there was/is some server that replies with
1511 		 * more link data than it should?
1512 		 */
1513 		if (len == NFS_MAXPATHLEN) {
1514 			NFSLOCKNODE(np);
1515 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1516 				len = np->n_size;
1517 				cangetattr = 0;
1518 			}
1519 			NFSUNLOCKNODE(np);
1520 		}
1521 		error = nfsm_mbufuio(nd, uiop, len);
1522 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1523 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1524 	}
1525 	if (nd->nd_repstat && !error)
1526 		error = nd->nd_repstat;
1527 nfsmout:
1528 	m_freem(nd->nd_mrep);
1529 	return (error);
1530 }
1531 
1532 /*
1533  * Read operation.
1534  */
1535 int
1536 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1537     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1538 {
1539 	int error, expireret = 0, retrycnt;
1540 	u_int32_t clidrev = 0;
1541 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1542 	struct nfsnode *np = VTONFS(vp);
1543 	struct ucred *newcred;
1544 	struct nfsfh *nfhp = NULL;
1545 	nfsv4stateid_t stateid;
1546 	void *lckp;
1547 
1548 	if (nmp->nm_clp != NULL)
1549 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1550 	newcred = cred;
1551 	if (NFSHASNFSV4(nmp)) {
1552 		nfhp = np->n_fhp;
1553 		newcred = NFSNEWCRED(cred);
1554 	}
1555 	retrycnt = 0;
1556 	do {
1557 		lckp = NULL;
1558 		if (NFSHASNFSV4(nmp))
1559 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1560 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1561 			    &lckp);
1562 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1563 		    attrflagp, stuff);
1564 		if (error == NFSERR_OPENMODE) {
1565 			NFSLOCKMNT(nmp);
1566 			nmp->nm_state |= NFSSTA_OPENMODE;
1567 			NFSUNLOCKMNT(nmp);
1568 		}
1569 		if (error == NFSERR_STALESTATEID)
1570 			nfscl_initiate_recovery(nmp->nm_clp);
1571 		if (lckp != NULL)
1572 			nfscl_lockderef(lckp);
1573 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1574 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1575 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1576 			(void) nfs_catnap(PZERO, error, "nfs_read");
1577 		} else if ((error == NFSERR_EXPIRED ||
1578 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1579 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1580 		}
1581 		retrycnt++;
1582 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1583 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1584 	    error == NFSERR_BADSESSION ||
1585 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1586 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1587 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1588 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1589 	if (error && retrycnt >= 4)
1590 		error = EIO;
1591 	if (NFSHASNFSV4(nmp))
1592 		NFSFREECRED(newcred);
1593 	return (error);
1594 }
1595 
1596 /*
1597  * The actual read RPC.
1598  */
1599 static int
1600 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1601     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1602     int *attrflagp, void *stuff)
1603 {
1604 	u_int32_t *tl;
1605 	int error = 0, len, retlen, tsiz, eof = 0;
1606 	struct nfsrv_descript nfsd;
1607 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1608 	struct nfsrv_descript *nd = &nfsd;
1609 	int rsize;
1610 	off_t tmp_off;
1611 
1612 	*attrflagp = 0;
1613 	tsiz = uiop->uio_resid;
1614 	tmp_off = uiop->uio_offset + tsiz;
1615 	NFSLOCKMNT(nmp);
1616 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1617 		NFSUNLOCKMNT(nmp);
1618 		return (EFBIG);
1619 	}
1620 	rsize = nmp->nm_rsize;
1621 	NFSUNLOCKMNT(nmp);
1622 	nd->nd_mrep = NULL;
1623 	while (tsiz > 0) {
1624 		*attrflagp = 0;
1625 		len = (tsiz > rsize) ? rsize : tsiz;
1626 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1627 		if (nd->nd_flag & ND_NFSV4)
1628 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1629 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1630 		if (nd->nd_flag & ND_NFSV2) {
1631 			*tl++ = txdr_unsigned(uiop->uio_offset);
1632 			*tl++ = txdr_unsigned(len);
1633 			*tl = 0;
1634 		} else {
1635 			txdr_hyper(uiop->uio_offset, tl);
1636 			*(tl + 2) = txdr_unsigned(len);
1637 		}
1638 		/*
1639 		 * Since I can't do a Getattr for NFSv4 for Write, there
1640 		 * doesn't seem any point in doing one here, either.
1641 		 * (See the comment in nfsrpc_writerpc() for more info.)
1642 		 */
1643 		error = nfscl_request(nd, vp, p, cred, stuff);
1644 		if (error)
1645 			return (error);
1646 		if (nd->nd_flag & ND_NFSV3) {
1647 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1648 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1649 			error = nfsm_loadattr(nd, nap);
1650 			if (!error)
1651 				*attrflagp = 1;
1652 		}
1653 		if (nd->nd_repstat || error) {
1654 			if (!error)
1655 				error = nd->nd_repstat;
1656 			goto nfsmout;
1657 		}
1658 		if (nd->nd_flag & ND_NFSV3) {
1659 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1660 			eof = fxdr_unsigned(int, *(tl + 1));
1661 		} else if (nd->nd_flag & ND_NFSV4) {
1662 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1663 			eof = fxdr_unsigned(int, *tl);
1664 		}
1665 		NFSM_STRSIZ(retlen, len);
1666 		error = nfsm_mbufuio(nd, uiop, retlen);
1667 		if (error)
1668 			goto nfsmout;
1669 		m_freem(nd->nd_mrep);
1670 		nd->nd_mrep = NULL;
1671 		tsiz -= retlen;
1672 		if (!(nd->nd_flag & ND_NFSV2)) {
1673 			if (eof || retlen == 0)
1674 				tsiz = 0;
1675 		} else if (retlen < len)
1676 			tsiz = 0;
1677 	}
1678 	return (0);
1679 nfsmout:
1680 	if (nd->nd_mrep != NULL)
1681 		m_freem(nd->nd_mrep);
1682 	return (error);
1683 }
1684 
1685 /*
1686  * nfs write operation
1687  * When called_from_strategy != 0, it should return EIO for an error that
1688  * indicates recovery is in progress, so that the buffer will be left
1689  * dirty and be written back to the server later. If it loops around,
1690  * the recovery thread could get stuck waiting for the buffer and recovery
1691  * will then deadlock.
1692  */
1693 int
1694 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1695     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1696     void *stuff, int called_from_strategy)
1697 {
1698 	int error, expireret = 0, retrycnt, nostateid;
1699 	u_int32_t clidrev = 0;
1700 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1701 	struct nfsnode *np = VTONFS(vp);
1702 	struct ucred *newcred;
1703 	struct nfsfh *nfhp = NULL;
1704 	nfsv4stateid_t stateid;
1705 	void *lckp;
1706 
1707 	*must_commit = 0;
1708 	if (nmp->nm_clp != NULL)
1709 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1710 	newcred = cred;
1711 	if (NFSHASNFSV4(nmp)) {
1712 		newcred = NFSNEWCRED(cred);
1713 		nfhp = np->n_fhp;
1714 	}
1715 	retrycnt = 0;
1716 	do {
1717 		lckp = NULL;
1718 		nostateid = 0;
1719 		if (NFSHASNFSV4(nmp)) {
1720 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1721 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1722 			    &lckp);
1723 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1724 			    stateid.other[2] == 0) {
1725 				nostateid = 1;
1726 				NFSCL_DEBUG(1, "stateid0 in write\n");
1727 			}
1728 		}
1729 
1730 		/*
1731 		 * If there is no stateid for NFSv4, it means this is an
1732 		 * extraneous write after close. Basically a poorly
1733 		 * implemented buffer cache. Just don't do the write.
1734 		 */
1735 		if (nostateid)
1736 			error = 0;
1737 		else
1738 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1739 			    newcred, &stateid, p, nap, attrflagp, stuff);
1740 		if (error == NFSERR_STALESTATEID)
1741 			nfscl_initiate_recovery(nmp->nm_clp);
1742 		if (lckp != NULL)
1743 			nfscl_lockderef(lckp);
1744 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1745 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1746 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1747 			(void) nfs_catnap(PZERO, error, "nfs_write");
1748 		} else if ((error == NFSERR_EXPIRED ||
1749 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1750 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1751 		}
1752 		retrycnt++;
1753 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1754 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1755 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1756 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1757 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1758 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1759 	if (error != 0 && (retrycnt >= 4 ||
1760 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1761 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1762 		error = EIO;
1763 	if (NFSHASNFSV4(nmp))
1764 		NFSFREECRED(newcred);
1765 	return (error);
1766 }
1767 
1768 /*
1769  * The actual write RPC.
1770  */
1771 static int
1772 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1773     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1774     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1775 {
1776 	u_int32_t *tl;
1777 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1778 	struct nfsnode *np = VTONFS(vp);
1779 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1780 	int wccflag = 0, wsize;
1781 	int32_t backup;
1782 	struct nfsrv_descript nfsd;
1783 	struct nfsrv_descript *nd = &nfsd;
1784 	nfsattrbit_t attrbits;
1785 	off_t tmp_off;
1786 
1787 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1788 	*attrflagp = 0;
1789 	tsiz = uiop->uio_resid;
1790 	tmp_off = uiop->uio_offset + tsiz;
1791 	NFSLOCKMNT(nmp);
1792 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1793 		NFSUNLOCKMNT(nmp);
1794 		return (EFBIG);
1795 	}
1796 	wsize = nmp->nm_wsize;
1797 	NFSUNLOCKMNT(nmp);
1798 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1799 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1800 	while (tsiz > 0) {
1801 		*attrflagp = 0;
1802 		len = (tsiz > wsize) ? wsize : tsiz;
1803 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1804 		if (nd->nd_flag & ND_NFSV4) {
1805 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1806 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1807 			txdr_hyper(uiop->uio_offset, tl);
1808 			tl += 2;
1809 			*tl++ = txdr_unsigned(*iomode);
1810 			*tl = txdr_unsigned(len);
1811 		} else if (nd->nd_flag & ND_NFSV3) {
1812 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1813 			txdr_hyper(uiop->uio_offset, tl);
1814 			tl += 2;
1815 			*tl++ = txdr_unsigned(len);
1816 			*tl++ = txdr_unsigned(*iomode);
1817 			*tl = txdr_unsigned(len);
1818 		} else {
1819 			u_int32_t x;
1820 
1821 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1822 			/*
1823 			 * Not sure why someone changed this, since the
1824 			 * RFC clearly states that "beginoffset" and
1825 			 * "totalcount" are ignored, but it wouldn't
1826 			 * surprise me if there's a busted server out there.
1827 			 */
1828 			/* Set both "begin" and "current" to non-garbage. */
1829 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1830 			*tl++ = x;      /* "begin offset" */
1831 			*tl++ = x;      /* "current offset" */
1832 			x = txdr_unsigned(len);
1833 			*tl++ = x;      /* total to this offset */
1834 			*tl = x;        /* size of this write */
1835 		}
1836 		nfsm_uiombuf(nd, uiop, len);
1837 		/*
1838 		 * Although it is tempting to do a normal Getattr Op in the
1839 		 * NFSv4 compound, the result can be a nearly hung client
1840 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1841 		 * It occurs when the client can't map either the Owner or
1842 		 * Owner_group name in the Getattr reply to a uid/gid. When
1843 		 * there is a cache miss, the kernel does an upcall to the
1844 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1845 		 * or /etc/group file. It can then block in getnewbuf(),
1846 		 * waiting for dirty writes to be pushed to the NFS server.
1847 		 * The only reason this doesn't result in a complete
1848 		 * deadlock, is that the upcall times out and allows
1849 		 * the write to complete. However, progress is so slow
1850 		 * that it might just as well be deadlocked.
1851 		 * As such, we get the rest of the attributes, but not
1852 		 * Owner or Owner_group.
1853 		 * nb: nfscl_loadattrcache() needs to be told that these
1854 		 *     partial attributes from a write rpc are being
1855 		 *     passed in, via a argument flag.
1856 		 */
1857 		if (nd->nd_flag & ND_NFSV4) {
1858 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
1859 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1860 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
1861 			(void) nfsrv_putattrbit(nd, &attrbits);
1862 		}
1863 		error = nfscl_request(nd, vp, p, cred, stuff);
1864 		if (error)
1865 			return (error);
1866 		if (nd->nd_repstat) {
1867 			/*
1868 			 * In case the rpc gets retried, roll
1869 			 * the uio fields changed by nfsm_uiombuf()
1870 			 * back.
1871 			 */
1872 			uiop->uio_offset -= len;
1873 			uiop->uio_resid += len;
1874 			uiop->uio_iov->iov_base =
1875 			    (char *)uiop->uio_iov->iov_base - len;
1876 			uiop->uio_iov->iov_len += len;
1877 		}
1878 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1879 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1880 			    &wccflag, stuff);
1881 			if (error)
1882 				goto nfsmout;
1883 		}
1884 		if (!nd->nd_repstat) {
1885 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1886 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1887 					+ NFSX_VERF);
1888 				rlen = fxdr_unsigned(int, *tl++);
1889 				if (rlen == 0) {
1890 					error = NFSERR_IO;
1891 					goto nfsmout;
1892 				} else if (rlen < len) {
1893 					backup = len - rlen;
1894 					uiop->uio_iov->iov_base =
1895 					    (char *)uiop->uio_iov->iov_base -
1896 					    backup;
1897 					uiop->uio_iov->iov_len += backup;
1898 					uiop->uio_offset -= backup;
1899 					uiop->uio_resid += backup;
1900 					len = rlen;
1901 				}
1902 				commit = fxdr_unsigned(int, *tl++);
1903 
1904 				/*
1905 				 * Return the lowest commitment level
1906 				 * obtained by any of the RPCs.
1907 				 */
1908 				if (committed == NFSWRITE_FILESYNC)
1909 					committed = commit;
1910 				else if (committed == NFSWRITE_DATASYNC &&
1911 					commit == NFSWRITE_UNSTABLE)
1912 					committed = commit;
1913 				NFSLOCKMNT(nmp);
1914 				if (!NFSHASWRITEVERF(nmp)) {
1915 					NFSBCOPY((caddr_t)tl,
1916 					    (caddr_t)&nmp->nm_verf[0],
1917 					    NFSX_VERF);
1918 					NFSSETWRITEVERF(nmp);
1919 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1920 				    NFSX_VERF)) {
1921 					*must_commit = 1;
1922 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1923 				}
1924 				NFSUNLOCKMNT(nmp);
1925 			}
1926 			if (nd->nd_flag & ND_NFSV4)
1927 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1928 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1929 				error = nfsm_loadattr(nd, nap);
1930 				if (!error)
1931 					*attrflagp = NFS_LATTR_NOSHRINK;
1932 			}
1933 		} else {
1934 			error = nd->nd_repstat;
1935 		}
1936 		if (error)
1937 			goto nfsmout;
1938 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1939 		m_freem(nd->nd_mrep);
1940 		nd->nd_mrep = NULL;
1941 		tsiz -= len;
1942 	}
1943 nfsmout:
1944 	if (nd->nd_mrep != NULL)
1945 		m_freem(nd->nd_mrep);
1946 	*iomode = committed;
1947 	if (nd->nd_repstat && !error)
1948 		error = nd->nd_repstat;
1949 	return (error);
1950 }
1951 
1952 /*
1953  * nfs mknod rpc
1954  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1955  * mode set to specify the file type and the size field for rdev.
1956  */
1957 int
1958 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1959     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1960     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1961     int *attrflagp, int *dattrflagp, void *dstuff)
1962 {
1963 	u_int32_t *tl;
1964 	int error = 0;
1965 	struct nfsrv_descript nfsd, *nd = &nfsd;
1966 	nfsattrbit_t attrbits;
1967 
1968 	*nfhpp = NULL;
1969 	*attrflagp = 0;
1970 	*dattrflagp = 0;
1971 	if (namelen > NFS_MAXNAMLEN)
1972 		return (ENAMETOOLONG);
1973 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1974 	if (nd->nd_flag & ND_NFSV4) {
1975 		if (vtyp == VBLK || vtyp == VCHR) {
1976 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1977 			*tl++ = vtonfsv34_type(vtyp);
1978 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1979 			*tl = txdr_unsigned(NFSMINOR(rdev));
1980 		} else {
1981 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1982 			*tl = vtonfsv34_type(vtyp);
1983 		}
1984 	}
1985 	(void) nfsm_strtom(nd, name, namelen);
1986 	if (nd->nd_flag & ND_NFSV3) {
1987 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1988 		*tl = vtonfsv34_type(vtyp);
1989 	}
1990 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1991 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
1992 	if ((nd->nd_flag & ND_NFSV3) &&
1993 	    (vtyp == VCHR || vtyp == VBLK)) {
1994 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1995 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1996 		*tl = txdr_unsigned(NFSMINOR(rdev));
1997 	}
1998 	if (nd->nd_flag & ND_NFSV4) {
1999 		NFSGETATTR_ATTRBIT(&attrbits);
2000 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2001 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2002 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2003 		(void) nfsrv_putattrbit(nd, &attrbits);
2004 	}
2005 	if (nd->nd_flag & ND_NFSV2)
2006 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2007 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2008 	if (error)
2009 		return (error);
2010 	if (nd->nd_flag & ND_NFSV4)
2011 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2012 	if (!nd->nd_repstat) {
2013 		if (nd->nd_flag & ND_NFSV4) {
2014 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2015 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2016 			if (error)
2017 				goto nfsmout;
2018 		}
2019 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2020 		if (error)
2021 			goto nfsmout;
2022 	}
2023 	if (nd->nd_flag & ND_NFSV3)
2024 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2025 	if (!error && nd->nd_repstat)
2026 		error = nd->nd_repstat;
2027 nfsmout:
2028 	m_freem(nd->nd_mrep);
2029 	return (error);
2030 }
2031 
2032 /*
2033  * nfs file create call
2034  * Mostly just call the approriate routine. (I separated out v4, so that
2035  * error recovery wouldn't be as difficult.)
2036  */
2037 int
2038 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2039     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2040     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2041     int *attrflagp, int *dattrflagp, void *dstuff)
2042 {
2043 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2044 	struct nfsclowner *owp;
2045 	struct nfscldeleg *dp;
2046 	struct nfsmount *nmp = VFSTONFS(dvp->v_mount);
2047 	u_int32_t clidrev;
2048 
2049 	if (NFSHASNFSV4(nmp)) {
2050 	    retrycnt = 0;
2051 	    do {
2052 		dp = NULL;
2053 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2054 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2055 		    NULL, 1);
2056 		if (error)
2057 			return (error);
2058 		if (nmp->nm_clp != NULL)
2059 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2060 		else
2061 			clidrev = 0;
2062 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2063 		    nfs_numnfscbd == 0 || retrycnt > 0)
2064 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2065 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2066 			  attrflagp, dattrflagp, dstuff, &unlocked);
2067 		else
2068 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2069 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2070 			  attrflagp, dattrflagp, dstuff, &unlocked);
2071 		/*
2072 		 * There is no need to invalidate cached attributes here,
2073 		 * since new post-delegation issue attributes are always
2074 		 * returned by nfsrpc_createv4() and these will update the
2075 		 * attribute cache.
2076 		 */
2077 		if (dp != NULL)
2078 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2079 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2080 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2081 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2082 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2083 		    error == NFSERR_BADSESSION) {
2084 			(void) nfs_catnap(PZERO, error, "nfs_open");
2085 		} else if ((error == NFSERR_EXPIRED ||
2086 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2087 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2088 			retrycnt++;
2089 		}
2090 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2091 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2092 		error == NFSERR_BADSESSION ||
2093 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2094 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2095 	    if (error && retrycnt >= 4)
2096 		    error = EIO;
2097 	} else {
2098 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2099 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2100 		    dstuff);
2101 	}
2102 	return (error);
2103 }
2104 
2105 /*
2106  * The create rpc for v2 and 3.
2107  */
2108 static int
2109 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2110     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2111     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2112     int *attrflagp, int *dattrflagp, void *dstuff)
2113 {
2114 	u_int32_t *tl;
2115 	int error = 0;
2116 	struct nfsrv_descript nfsd, *nd = &nfsd;
2117 
2118 	*nfhpp = NULL;
2119 	*attrflagp = 0;
2120 	*dattrflagp = 0;
2121 	if (namelen > NFS_MAXNAMLEN)
2122 		return (ENAMETOOLONG);
2123 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2124 	(void) nfsm_strtom(nd, name, namelen);
2125 	if (nd->nd_flag & ND_NFSV3) {
2126 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2127 		if (fmode & O_EXCL) {
2128 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2129 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2130 			*tl++ = cverf.lval[0];
2131 			*tl = cverf.lval[1];
2132 		} else {
2133 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2134 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2135 		}
2136 	} else {
2137 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2138 	}
2139 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2140 	if (error)
2141 		return (error);
2142 	if (nd->nd_repstat == 0) {
2143 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2144 		if (error)
2145 			goto nfsmout;
2146 	}
2147 	if (nd->nd_flag & ND_NFSV3)
2148 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2149 	if (nd->nd_repstat != 0 && error == 0)
2150 		error = nd->nd_repstat;
2151 nfsmout:
2152 	m_freem(nd->nd_mrep);
2153 	return (error);
2154 }
2155 
2156 static int
2157 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2158     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2159     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2160     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2161     int *dattrflagp, void *dstuff, int *unlockedp)
2162 {
2163 	u_int32_t *tl;
2164 	int error = 0, deleg, newone, ret, acesize, limitby;
2165 	struct nfsrv_descript nfsd, *nd = &nfsd;
2166 	struct nfsclopen *op;
2167 	struct nfscldeleg *dp = NULL;
2168 	struct nfsnode *np;
2169 	struct nfsfh *nfhp;
2170 	nfsattrbit_t attrbits;
2171 	nfsv4stateid_t stateid;
2172 	u_int32_t rflags;
2173 	struct nfsmount *nmp;
2174 	struct nfsclsession *tsep;
2175 
2176 	nmp = VFSTONFS(dvp->v_mount);
2177 	np = VTONFS(dvp);
2178 	*unlockedp = 0;
2179 	*nfhpp = NULL;
2180 	*dpp = NULL;
2181 	*attrflagp = 0;
2182 	*dattrflagp = 0;
2183 	if (namelen > NFS_MAXNAMLEN)
2184 		return (ENAMETOOLONG);
2185 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2186 	/*
2187 	 * For V4, this is actually an Open op.
2188 	 */
2189 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2190 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2191 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2192 	    NFSV4OPEN_ACCESSREAD);
2193 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2194 	tsep = nfsmnt_mdssession(nmp);
2195 	*tl++ = tsep->nfsess_clientid.lval[0];
2196 	*tl = tsep->nfsess_clientid.lval[1];
2197 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2198 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2199 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2200 	if (fmode & O_EXCL) {
2201 		if (NFSHASNFSV4N(nmp)) {
2202 			if (NFSHASSESSPERSIST(nmp)) {
2203 				/* Use GUARDED for persistent sessions. */
2204 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2205 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2206 			} else {
2207 				/* Otherwise, use EXCLUSIVE4_1. */
2208 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2209 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2210 				*tl++ = cverf.lval[0];
2211 				*tl = cverf.lval[1];
2212 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2213 			}
2214 		} else {
2215 			/* NFSv4.0 */
2216 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2217 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2218 			*tl++ = cverf.lval[0];
2219 			*tl = cverf.lval[1];
2220 		}
2221 	} else {
2222 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2223 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2224 	}
2225 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2226 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2227 	(void) nfsm_strtom(nd, name, namelen);
2228 	/* Get the new file's handle and attributes. */
2229 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2230 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2231 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2232 	NFSGETATTR_ATTRBIT(&attrbits);
2233 	(void) nfsrv_putattrbit(nd, &attrbits);
2234 	/* Get the directory's post-op attributes. */
2235 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2236 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2237 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2238 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2239 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2240 	(void) nfsrv_putattrbit(nd, &attrbits);
2241 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2242 	if (error)
2243 		return (error);
2244 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2245 	if (nd->nd_repstat == 0) {
2246 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2247 		    6 * NFSX_UNSIGNED);
2248 		stateid.seqid = *tl++;
2249 		stateid.other[0] = *tl++;
2250 		stateid.other[1] = *tl++;
2251 		stateid.other[2] = *tl;
2252 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2253 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2254 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2255 		deleg = fxdr_unsigned(int, *tl);
2256 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2257 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2258 			if (!(owp->nfsow_clp->nfsc_flags &
2259 			      NFSCLFLAGS_FIRSTDELEG))
2260 				owp->nfsow_clp->nfsc_flags |=
2261 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2262 			dp = malloc(
2263 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2264 			    M_NFSCLDELEG, M_WAITOK);
2265 			LIST_INIT(&dp->nfsdl_owner);
2266 			LIST_INIT(&dp->nfsdl_lock);
2267 			dp->nfsdl_clp = owp->nfsow_clp;
2268 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2269 			nfscl_lockinit(&dp->nfsdl_rwlock);
2270 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2271 			    NFSX_UNSIGNED);
2272 			dp->nfsdl_stateid.seqid = *tl++;
2273 			dp->nfsdl_stateid.other[0] = *tl++;
2274 			dp->nfsdl_stateid.other[1] = *tl++;
2275 			dp->nfsdl_stateid.other[2] = *tl++;
2276 			ret = fxdr_unsigned(int, *tl);
2277 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2278 				dp->nfsdl_flags = NFSCLDL_WRITE;
2279 				/*
2280 				 * Indicates how much the file can grow.
2281 				 */
2282 				NFSM_DISSECT(tl, u_int32_t *,
2283 				    3 * NFSX_UNSIGNED);
2284 				limitby = fxdr_unsigned(int, *tl++);
2285 				switch (limitby) {
2286 				case NFSV4OPEN_LIMITSIZE:
2287 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2288 					break;
2289 				case NFSV4OPEN_LIMITBLOCKS:
2290 					dp->nfsdl_sizelimit =
2291 					    fxdr_unsigned(u_int64_t, *tl++);
2292 					dp->nfsdl_sizelimit *=
2293 					    fxdr_unsigned(u_int64_t, *tl);
2294 					break;
2295 				default:
2296 					error = NFSERR_BADXDR;
2297 					goto nfsmout;
2298 				}
2299 			} else {
2300 				dp->nfsdl_flags = NFSCLDL_READ;
2301 			}
2302 			if (ret)
2303 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2304 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2305 			    &acesize, p);
2306 			if (error)
2307 				goto nfsmout;
2308 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2309 			error = NFSERR_BADXDR;
2310 			goto nfsmout;
2311 		}
2312 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2313 		if (error)
2314 			goto nfsmout;
2315 		/* Get rid of the PutFH and Getattr status values. */
2316 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2317 		/* Load the directory attributes. */
2318 		error = nfsm_loadattr(nd, dnap);
2319 		if (error)
2320 			goto nfsmout;
2321 		*dattrflagp = 1;
2322 		if (dp != NULL && *attrflagp) {
2323 			dp->nfsdl_change = nnap->na_filerev;
2324 			dp->nfsdl_modtime = nnap->na_mtime;
2325 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2326 		}
2327 		/*
2328 		 * We can now complete the Open state.
2329 		 */
2330 		nfhp = *nfhpp;
2331 		if (dp != NULL) {
2332 			dp->nfsdl_fhlen = nfhp->nfh_len;
2333 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2334 		}
2335 		/*
2336 		 * Get an Open structure that will be
2337 		 * attached to the OpenOwner, acquired already.
2338 		 */
2339 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2340 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2341 		    cred, p, NULL, &op, &newone, NULL, 0);
2342 		if (error)
2343 			goto nfsmout;
2344 		op->nfso_stateid = stateid;
2345 		newnfs_copyincred(cred, &op->nfso_cred);
2346 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2347 		    do {
2348 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2349 			    nfhp->nfh_len, op, cred, p);
2350 			if (ret == NFSERR_DELAY)
2351 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2352 		    } while (ret == NFSERR_DELAY);
2353 		    error = ret;
2354 		}
2355 
2356 		/*
2357 		 * If the server is handing out delegations, but we didn't
2358 		 * get one because an OpenConfirm was required, try the
2359 		 * Open again, to get a delegation. This is a harmless no-op,
2360 		 * from a server's point of view.
2361 		 */
2362 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2363 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2364 		    !error && dp == NULL) {
2365 		    do {
2366 			ret = nfsrpc_openrpc(VFSTONFS(dvp->v_mount), dvp,
2367 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2368 			    nfhp->nfh_fh, nfhp->nfh_len,
2369 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2370 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2371 			if (ret == NFSERR_DELAY)
2372 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2373 		    } while (ret == NFSERR_DELAY);
2374 		    if (ret) {
2375 			if (dp != NULL) {
2376 				free(dp, M_NFSCLDELEG);
2377 				dp = NULL;
2378 			}
2379 			if (ret == NFSERR_STALECLIENTID ||
2380 			    ret == NFSERR_STALEDONTRECOVER ||
2381 			    ret == NFSERR_BADSESSION)
2382 				error = ret;
2383 		    }
2384 		}
2385 		nfscl_openrelease(nmp, op, error, newone);
2386 		*unlockedp = 1;
2387 	}
2388 	if (nd->nd_repstat != 0 && error == 0)
2389 		error = nd->nd_repstat;
2390 	if (error == NFSERR_STALECLIENTID)
2391 		nfscl_initiate_recovery(owp->nfsow_clp);
2392 nfsmout:
2393 	if (!error)
2394 		*dpp = dp;
2395 	else if (dp != NULL)
2396 		free(dp, M_NFSCLDELEG);
2397 	m_freem(nd->nd_mrep);
2398 	return (error);
2399 }
2400 
2401 /*
2402  * Nfs remove rpc
2403  */
2404 int
2405 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2406     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2407     void *dstuff)
2408 {
2409 	u_int32_t *tl;
2410 	struct nfsrv_descript nfsd, *nd = &nfsd;
2411 	struct nfsnode *np;
2412 	struct nfsmount *nmp;
2413 	nfsv4stateid_t dstateid;
2414 	int error, ret = 0, i;
2415 
2416 	*dattrflagp = 0;
2417 	if (namelen > NFS_MAXNAMLEN)
2418 		return (ENAMETOOLONG);
2419 	nmp = VFSTONFS(dvp->v_mount);
2420 tryagain:
2421 	if (NFSHASNFSV4(nmp) && ret == 0) {
2422 		ret = nfscl_removedeleg(vp, p, &dstateid);
2423 		if (ret == 1) {
2424 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2425 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2426 			    NFSX_UNSIGNED);
2427 			if (NFSHASNFSV4N(nmp))
2428 				*tl++ = 0;
2429 			else
2430 				*tl++ = dstateid.seqid;
2431 			*tl++ = dstateid.other[0];
2432 			*tl++ = dstateid.other[1];
2433 			*tl++ = dstateid.other[2];
2434 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2435 			np = VTONFS(dvp);
2436 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2437 			    np->n_fhp->nfh_len, 0);
2438 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2439 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2440 		}
2441 	} else {
2442 		ret = 0;
2443 	}
2444 	if (ret == 0)
2445 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2446 	(void) nfsm_strtom(nd, name, namelen);
2447 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2448 	if (error)
2449 		return (error);
2450 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2451 		/* For NFSv4, parse out any Delereturn replies. */
2452 		if (ret > 0 && nd->nd_repstat != 0 &&
2453 		    (nd->nd_flag & ND_NOMOREDATA)) {
2454 			/*
2455 			 * If the Delegreturn failed, try again without
2456 			 * it. The server will Recall, as required.
2457 			 */
2458 			m_freem(nd->nd_mrep);
2459 			goto tryagain;
2460 		}
2461 		for (i = 0; i < (ret * 2); i++) {
2462 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2463 			    ND_NFSV4) {
2464 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2465 			    if (*(tl + 1))
2466 				nd->nd_flag |= ND_NOMOREDATA;
2467 			}
2468 		}
2469 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2470 	}
2471 	if (nd->nd_repstat && !error)
2472 		error = nd->nd_repstat;
2473 nfsmout:
2474 	m_freem(nd->nd_mrep);
2475 	return (error);
2476 }
2477 
2478 /*
2479  * Do an nfs rename rpc.
2480  */
2481 int
2482 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2483     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2484     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2485     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2486 {
2487 	u_int32_t *tl;
2488 	struct nfsrv_descript nfsd, *nd = &nfsd;
2489 	struct nfsmount *nmp;
2490 	struct nfsnode *np;
2491 	nfsattrbit_t attrbits;
2492 	nfsv4stateid_t fdstateid, tdstateid;
2493 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2494 
2495 	*fattrflagp = 0;
2496 	*tattrflagp = 0;
2497 	nmp = VFSTONFS(fdvp->v_mount);
2498 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2499 		return (ENAMETOOLONG);
2500 tryagain:
2501 	if (NFSHASNFSV4(nmp) && ret == 0) {
2502 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2503 		    &tdstateid, &gottd, p);
2504 		if (gotfd && gottd) {
2505 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2506 		} else if (gotfd) {
2507 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2508 		} else if (gottd) {
2509 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2510 		}
2511 		if (gotfd) {
2512 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2513 			if (NFSHASNFSV4N(nmp))
2514 				*tl++ = 0;
2515 			else
2516 				*tl++ = fdstateid.seqid;
2517 			*tl++ = fdstateid.other[0];
2518 			*tl++ = fdstateid.other[1];
2519 			*tl = fdstateid.other[2];
2520 			if (gottd) {
2521 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2522 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2523 				np = VTONFS(tvp);
2524 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2525 				    np->n_fhp->nfh_len, 0);
2526 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2527 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2528 			}
2529 		}
2530 		if (gottd) {
2531 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2532 			if (NFSHASNFSV4N(nmp))
2533 				*tl++ = 0;
2534 			else
2535 				*tl++ = tdstateid.seqid;
2536 			*tl++ = tdstateid.other[0];
2537 			*tl++ = tdstateid.other[1];
2538 			*tl = tdstateid.other[2];
2539 		}
2540 		if (ret > 0) {
2541 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2542 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2543 			np = VTONFS(fdvp);
2544 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2545 			    np->n_fhp->nfh_len, 0);
2546 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2547 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2548 		}
2549 	} else {
2550 		ret = 0;
2551 	}
2552 	if (ret == 0)
2553 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2554 	if (nd->nd_flag & ND_NFSV4) {
2555 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2556 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2557 		NFSWCCATTR_ATTRBIT(&attrbits);
2558 		(void) nfsrv_putattrbit(nd, &attrbits);
2559 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2560 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2561 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2562 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2563 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2564 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2565 		(void) nfsrv_putattrbit(nd, &attrbits);
2566 		nd->nd_flag |= ND_V4WCCATTR;
2567 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2568 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2569 	}
2570 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2571 	if (!(nd->nd_flag & ND_NFSV4))
2572 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2573 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2574 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2575 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2576 	if (error)
2577 		return (error);
2578 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2579 		/* For NFSv4, parse out any Delereturn replies. */
2580 		if (ret > 0 && nd->nd_repstat != 0 &&
2581 		    (nd->nd_flag & ND_NOMOREDATA)) {
2582 			/*
2583 			 * If the Delegreturn failed, try again without
2584 			 * it. The server will Recall, as required.
2585 			 */
2586 			m_freem(nd->nd_mrep);
2587 			goto tryagain;
2588 		}
2589 		for (i = 0; i < (ret * 2); i++) {
2590 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2591 			    ND_NFSV4) {
2592 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2593 			    if (*(tl + 1)) {
2594 				if (i == 0 && ret > 1) {
2595 				    /*
2596 				     * If the Delegreturn failed, try again
2597 				     * without it. The server will Recall, as
2598 				     * required.
2599 				     * If ret > 1, the first iteration of this
2600 				     * loop is the second DelegReturn result.
2601 				     */
2602 				    m_freem(nd->nd_mrep);
2603 				    goto tryagain;
2604 				} else {
2605 				    nd->nd_flag |= ND_NOMOREDATA;
2606 				}
2607 			    }
2608 			}
2609 		}
2610 		/* Now, the first wcc attribute reply. */
2611 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2612 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2613 			if (*(tl + 1))
2614 				nd->nd_flag |= ND_NOMOREDATA;
2615 		}
2616 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2617 		    fstuff);
2618 		/* and the second wcc attribute reply. */
2619 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2620 		    !error) {
2621 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2622 			if (*(tl + 1))
2623 				nd->nd_flag |= ND_NOMOREDATA;
2624 		}
2625 		if (!error)
2626 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2627 			    NULL, tstuff);
2628 	}
2629 	if (nd->nd_repstat && !error)
2630 		error = nd->nd_repstat;
2631 nfsmout:
2632 	m_freem(nd->nd_mrep);
2633 	return (error);
2634 }
2635 
2636 /*
2637  * nfs hard link create rpc
2638  */
2639 int
2640 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2641     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2642     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2643 {
2644 	u_int32_t *tl;
2645 	struct nfsrv_descript nfsd, *nd = &nfsd;
2646 	nfsattrbit_t attrbits;
2647 	int error = 0;
2648 
2649 	*attrflagp = 0;
2650 	*dattrflagp = 0;
2651 	if (namelen > NFS_MAXNAMLEN)
2652 		return (ENAMETOOLONG);
2653 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2654 	if (nd->nd_flag & ND_NFSV4) {
2655 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2656 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2657 	}
2658 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2659 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2660 	if (nd->nd_flag & ND_NFSV4) {
2661 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2662 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2663 		NFSWCCATTR_ATTRBIT(&attrbits);
2664 		(void) nfsrv_putattrbit(nd, &attrbits);
2665 		nd->nd_flag |= ND_V4WCCATTR;
2666 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2667 		*tl = txdr_unsigned(NFSV4OP_LINK);
2668 	}
2669 	(void) nfsm_strtom(nd, name, namelen);
2670 	error = nfscl_request(nd, vp, p, cred, dstuff);
2671 	if (error)
2672 		return (error);
2673 	if (nd->nd_flag & ND_NFSV3) {
2674 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2675 		if (!error)
2676 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2677 			    NULL, dstuff);
2678 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2679 		/*
2680 		 * First, parse out the PutFH and Getattr result.
2681 		 */
2682 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2683 		if (!(*(tl + 1)))
2684 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2685 		if (*(tl + 1))
2686 			nd->nd_flag |= ND_NOMOREDATA;
2687 		/*
2688 		 * Get the pre-op attributes.
2689 		 */
2690 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2691 	}
2692 	if (nd->nd_repstat && !error)
2693 		error = nd->nd_repstat;
2694 nfsmout:
2695 	m_freem(nd->nd_mrep);
2696 	return (error);
2697 }
2698 
2699 /*
2700  * nfs symbolic link create rpc
2701  */
2702 int
2703 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
2704     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2705     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2706     int *dattrflagp, void *dstuff)
2707 {
2708 	u_int32_t *tl;
2709 	struct nfsrv_descript nfsd, *nd = &nfsd;
2710 	struct nfsmount *nmp;
2711 	int slen, error = 0;
2712 
2713 	*nfhpp = NULL;
2714 	*attrflagp = 0;
2715 	*dattrflagp = 0;
2716 	nmp = VFSTONFS(dvp->v_mount);
2717 	slen = strlen(target);
2718 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2719 		return (ENAMETOOLONG);
2720 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2721 	if (nd->nd_flag & ND_NFSV4) {
2722 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2723 		*tl = txdr_unsigned(NFLNK);
2724 		(void) nfsm_strtom(nd, target, slen);
2725 	}
2726 	(void) nfsm_strtom(nd, name, namelen);
2727 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2728 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2729 	if (!(nd->nd_flag & ND_NFSV4))
2730 		(void) nfsm_strtom(nd, target, slen);
2731 	if (nd->nd_flag & ND_NFSV2)
2732 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2733 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2734 	if (error)
2735 		return (error);
2736 	if (nd->nd_flag & ND_NFSV4)
2737 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2738 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2739 		if (!nd->nd_repstat)
2740 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2741 		if (!error)
2742 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2743 			    NULL, dstuff);
2744 	}
2745 	if (nd->nd_repstat && !error)
2746 		error = nd->nd_repstat;
2747 	m_freem(nd->nd_mrep);
2748 	/*
2749 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2750 	 * Only do this if vfs.nfs.ignore_eexist is set.
2751 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2752 	 * should guarantee "exactly once" RPC semantics.
2753 	 */
2754 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2755 	    nmp->nm_minorvers == 0))
2756 		error = 0;
2757 	return (error);
2758 }
2759 
2760 /*
2761  * nfs make dir rpc
2762  */
2763 int
2764 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2765     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2766     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2767     int *dattrflagp, void *dstuff)
2768 {
2769 	u_int32_t *tl;
2770 	struct nfsrv_descript nfsd, *nd = &nfsd;
2771 	nfsattrbit_t attrbits;
2772 	int error = 0;
2773 	struct nfsfh *fhp;
2774 	struct nfsmount *nmp;
2775 
2776 	*nfhpp = NULL;
2777 	*attrflagp = 0;
2778 	*dattrflagp = 0;
2779 	nmp = VFSTONFS(dvp->v_mount);
2780 	fhp = VTONFS(dvp)->n_fhp;
2781 	if (namelen > NFS_MAXNAMLEN)
2782 		return (ENAMETOOLONG);
2783 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2784 	if (nd->nd_flag & ND_NFSV4) {
2785 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2786 		*tl = txdr_unsigned(NFDIR);
2787 	}
2788 	(void) nfsm_strtom(nd, name, namelen);
2789 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2790 	if (nd->nd_flag & ND_NFSV4) {
2791 		NFSGETATTR_ATTRBIT(&attrbits);
2792 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2793 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2794 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2795 		(void) nfsrv_putattrbit(nd, &attrbits);
2796 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2797 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2798 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2799 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2800 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2801 		(void) nfsrv_putattrbit(nd, &attrbits);
2802 	}
2803 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2804 	if (error)
2805 		return (error);
2806 	if (nd->nd_flag & ND_NFSV4)
2807 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2808 	if (!nd->nd_repstat && !error) {
2809 		if (nd->nd_flag & ND_NFSV4) {
2810 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2811 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2812 		}
2813 		if (!error)
2814 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2815 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2816 			/* Get rid of the PutFH and Getattr status values. */
2817 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2818 			/* Load the directory attributes. */
2819 			error = nfsm_loadattr(nd, dnap);
2820 			if (error == 0)
2821 				*dattrflagp = 1;
2822 		}
2823 	}
2824 	if ((nd->nd_flag & ND_NFSV3) && !error)
2825 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2826 	if (nd->nd_repstat && !error)
2827 		error = nd->nd_repstat;
2828 nfsmout:
2829 	m_freem(nd->nd_mrep);
2830 	/*
2831 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2832 	 * Only do this if vfs.nfs.ignore_eexist is set.
2833 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2834 	 * should guarantee "exactly once" RPC semantics.
2835 	 */
2836 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2837 	    nmp->nm_minorvers == 0))
2838 		error = 0;
2839 	return (error);
2840 }
2841 
2842 /*
2843  * nfs remove directory call
2844  */
2845 int
2846 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2847     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2848 {
2849 	struct nfsrv_descript nfsd, *nd = &nfsd;
2850 	int error = 0;
2851 
2852 	*dattrflagp = 0;
2853 	if (namelen > NFS_MAXNAMLEN)
2854 		return (ENAMETOOLONG);
2855 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2856 	(void) nfsm_strtom(nd, name, namelen);
2857 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2858 	if (error)
2859 		return (error);
2860 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2861 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2862 	if (nd->nd_repstat && !error)
2863 		error = nd->nd_repstat;
2864 	m_freem(nd->nd_mrep);
2865 	/*
2866 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2867 	 */
2868 	if (error == ENOENT)
2869 		error = 0;
2870 	return (error);
2871 }
2872 
2873 /*
2874  * Readdir rpc.
2875  * Always returns with either uio_resid unchanged, if you are at the
2876  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2877  * filled in.
2878  * I felt this would allow caching of directory blocks more easily
2879  * than returning a pertially filled block.
2880  * Directory offset cookies:
2881  * Oh my, what to do with them...
2882  * I can think of three ways to deal with them:
2883  * 1 - have the layer above these RPCs maintain a map between logical
2884  *     directory byte offsets and the NFS directory offset cookies
2885  * 2 - pass the opaque directory offset cookies up into userland
2886  *     and let the libc functions deal with them, via the system call
2887  * 3 - return them to userland in the "struct dirent", so future versions
2888  *     of libc can use them and do whatever is necessary to make things work
2889  *     above these rpc calls, in the meantime
2890  * For now, I do #3 by "hiding" the directory offset cookies after the
2891  * d_name field in struct dirent. This is space inside d_reclen that
2892  * will be ignored by anything that doesn't know about them.
2893  * The directory offset cookies are filled in as the last 8 bytes of
2894  * each directory entry, after d_name. Someday, the userland libc
2895  * functions may be able to use these. In the meantime, it satisfies
2896  * OpenBSD's requirements for cookies being returned.
2897  * If expects the directory offset cookie for the read to be in uio_offset
2898  * and returns the one for the next entry after this directory block in
2899  * there, as well.
2900  */
2901 int
2902 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2903     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2904     int *eofp, void *stuff)
2905 {
2906 	int len, left;
2907 	struct dirent *dp = NULL;
2908 	u_int32_t *tl;
2909 	nfsquad_t cookie, ncookie;
2910 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
2911 	struct nfsnode *dnp = VTONFS(vp);
2912 	struct nfsvattr nfsva;
2913 	struct nfsrv_descript nfsd, *nd = &nfsd;
2914 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2915 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2916 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2917 	char *cp;
2918 	nfsattrbit_t attrbits, dattrbits;
2919 	u_int32_t rderr, *tl2 = NULL;
2920 	size_t tresid;
2921 
2922 	KASSERT(uiop->uio_iovcnt == 1 &&
2923 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
2924 	    ("nfs readdirrpc bad uio"));
2925 	ncookie.lval[0] = ncookie.lval[1] = 0;
2926 	/*
2927 	 * There is no point in reading a lot more than uio_resid, however
2928 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2929 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2930 	 * will never make readsize > nm_readdirsize.
2931 	 */
2932 	readsize = nmp->nm_readdirsize;
2933 	if (readsize > uiop->uio_resid)
2934 		readsize = uiop->uio_resid + DIRBLKSIZ;
2935 
2936 	*attrflagp = 0;
2937 	if (eofp)
2938 		*eofp = 0;
2939 	tresid = uiop->uio_resid;
2940 	cookie.lval[0] = cookiep->nfsuquad[0];
2941 	cookie.lval[1] = cookiep->nfsuquad[1];
2942 	nd->nd_mrep = NULL;
2943 
2944 	/*
2945 	 * For NFSv4, first create the "." and ".." entries.
2946 	 */
2947 	if (NFSHASNFSV4(nmp)) {
2948 		reqsize = 6 * NFSX_UNSIGNED;
2949 		NFSGETATTR_ATTRBIT(&dattrbits);
2950 		NFSZERO_ATTRBIT(&attrbits);
2951 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2952 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2953 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2954 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2955 			NFSSETBIT_ATTRBIT(&attrbits,
2956 			    NFSATTRBIT_MOUNTEDONFILEID);
2957 			gotmnton = 1;
2958 		} else {
2959 			/*
2960 			 * Must fake it. Use the fileno, except when the
2961 			 * fsid is != to that of the directory. For that
2962 			 * case, generate a fake fileno that is not the same.
2963 			 */
2964 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2965 			gotmnton = 0;
2966 		}
2967 
2968 		/*
2969 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2970 		 */
2971 		if (uiop->uio_offset == 0) {
2972 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2973 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2974 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2975 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2976 			(void) nfsrv_putattrbit(nd, &attrbits);
2977 			error = nfscl_request(nd, vp, p, cred, stuff);
2978 			if (error)
2979 			    return (error);
2980 			dotfileid = 0;	/* Fake out the compiler. */
2981 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2982 			    error = nfsm_loadattr(nd, &nfsva);
2983 			    if (error != 0)
2984 				goto nfsmout;
2985 			    dotfileid = nfsva.na_fileid;
2986 			}
2987 			if (nd->nd_repstat == 0) {
2988 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2989 			    len = fxdr_unsigned(int, *(tl + 4));
2990 			    if (len > 0 && len <= NFSX_V4FHMAX)
2991 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2992 			    else
2993 				error = EPERM;
2994 			    if (!error) {
2995 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
2996 				nfsva.na_mntonfileno = UINT64_MAX;
2997 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
2998 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
2999 				    NULL, NULL, NULL, p, cred);
3000 				if (error) {
3001 				    dotdotfileid = dotfileid;
3002 				} else if (gotmnton) {
3003 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3004 					dotdotfileid = nfsva.na_mntonfileno;
3005 				    else
3006 					dotdotfileid = nfsva.na_fileid;
3007 				} else if (nfsva.na_filesid[0] ==
3008 				    dnp->n_vattr.na_filesid[0] &&
3009 				    nfsva.na_filesid[1] ==
3010 				    dnp->n_vattr.na_filesid[1]) {
3011 				    dotdotfileid = nfsva.na_fileid;
3012 				} else {
3013 				    do {
3014 					fakefileno--;
3015 				    } while (fakefileno ==
3016 					nfsva.na_fileid);
3017 				    dotdotfileid = fakefileno;
3018 				}
3019 			    }
3020 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3021 			    /*
3022 			     * Lookupp returns NFSERR_NOENT when we are
3023 			     * at the root, so just use the current dir.
3024 			     */
3025 			    nd->nd_repstat = 0;
3026 			    dotdotfileid = dotfileid;
3027 			} else {
3028 			    error = nd->nd_repstat;
3029 			}
3030 			m_freem(nd->nd_mrep);
3031 			if (error)
3032 			    return (error);
3033 			nd->nd_mrep = NULL;
3034 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3035 			dp->d_pad0 = dp->d_pad1 = 0;
3036 			dp->d_off = 0;
3037 			dp->d_type = DT_DIR;
3038 			dp->d_fileno = dotfileid;
3039 			dp->d_namlen = 1;
3040 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3041 			dp->d_name[0] = '.';
3042 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3043 			/*
3044 			 * Just make these offset cookie 0.
3045 			 */
3046 			tl = (u_int32_t *)&dp->d_name[8];
3047 			*tl++ = 0;
3048 			*tl = 0;
3049 			blksiz += dp->d_reclen;
3050 			uiop->uio_resid -= dp->d_reclen;
3051 			uiop->uio_offset += dp->d_reclen;
3052 			uiop->uio_iov->iov_base =
3053 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3054 			uiop->uio_iov->iov_len -= dp->d_reclen;
3055 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3056 			dp->d_pad0 = dp->d_pad1 = 0;
3057 			dp->d_off = 0;
3058 			dp->d_type = DT_DIR;
3059 			dp->d_fileno = dotdotfileid;
3060 			dp->d_namlen = 2;
3061 			*((uint64_t *)dp->d_name) = 0;
3062 			dp->d_name[0] = '.';
3063 			dp->d_name[1] = '.';
3064 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3065 			/*
3066 			 * Just make these offset cookie 0.
3067 			 */
3068 			tl = (u_int32_t *)&dp->d_name[8];
3069 			*tl++ = 0;
3070 			*tl = 0;
3071 			blksiz += dp->d_reclen;
3072 			uiop->uio_resid -= dp->d_reclen;
3073 			uiop->uio_offset += dp->d_reclen;
3074 			uiop->uio_iov->iov_base =
3075 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3076 			uiop->uio_iov->iov_len -= dp->d_reclen;
3077 		}
3078 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3079 	} else {
3080 		reqsize = 5 * NFSX_UNSIGNED;
3081 	}
3082 
3083 	/*
3084 	 * Loop around doing readdir rpc's of size readsize.
3085 	 * The stopping criteria is EOF or buffer full.
3086 	 */
3087 	while (more_dirs && bigenough) {
3088 		*attrflagp = 0;
3089 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3090 		if (nd->nd_flag & ND_NFSV2) {
3091 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3092 			*tl++ = cookie.lval[1];
3093 			*tl = txdr_unsigned(readsize);
3094 		} else {
3095 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3096 			*tl++ = cookie.lval[0];
3097 			*tl++ = cookie.lval[1];
3098 			if (cookie.qval == 0) {
3099 				*tl++ = 0;
3100 				*tl++ = 0;
3101 			} else {
3102 				NFSLOCKNODE(dnp);
3103 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3104 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3105 				NFSUNLOCKNODE(dnp);
3106 			}
3107 			if (nd->nd_flag & ND_NFSV4) {
3108 				*tl++ = txdr_unsigned(readsize);
3109 				*tl = txdr_unsigned(readsize);
3110 				(void) nfsrv_putattrbit(nd, &attrbits);
3111 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3112 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3113 				(void) nfsrv_putattrbit(nd, &dattrbits);
3114 			} else {
3115 				*tl = txdr_unsigned(readsize);
3116 			}
3117 		}
3118 		error = nfscl_request(nd, vp, p, cred, stuff);
3119 		if (error)
3120 			return (error);
3121 		if (!(nd->nd_flag & ND_NFSV2)) {
3122 			if (nd->nd_flag & ND_NFSV3)
3123 				error = nfscl_postop_attr(nd, nap, attrflagp,
3124 				    stuff);
3125 			if (!nd->nd_repstat && !error) {
3126 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3127 				NFSLOCKNODE(dnp);
3128 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3129 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3130 				NFSUNLOCKNODE(dnp);
3131 			}
3132 		}
3133 		if (nd->nd_repstat || error) {
3134 			if (!error)
3135 				error = nd->nd_repstat;
3136 			goto nfsmout;
3137 		}
3138 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3139 		more_dirs = fxdr_unsigned(int, *tl);
3140 		if (!more_dirs)
3141 			tryformoredirs = 0;
3142 
3143 		/* loop through the dir entries, doctoring them to 4bsd form */
3144 		while (more_dirs && bigenough) {
3145 			if (nd->nd_flag & ND_NFSV4) {
3146 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3147 				ncookie.lval[0] = *tl++;
3148 				ncookie.lval[1] = *tl++;
3149 				len = fxdr_unsigned(int, *tl);
3150 			} else if (nd->nd_flag & ND_NFSV3) {
3151 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3152 				nfsva.na_fileid = fxdr_hyper(tl);
3153 				tl += 2;
3154 				len = fxdr_unsigned(int, *tl);
3155 			} else {
3156 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3157 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3158 				    *tl++);
3159 				len = fxdr_unsigned(int, *tl);
3160 			}
3161 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3162 				error = EBADRPC;
3163 				goto nfsmout;
3164 			}
3165 			tlen = roundup2(len, 8);
3166 			if (tlen == len)
3167 				tlen += 8;  /* To ensure null termination. */
3168 			left = DIRBLKSIZ - blksiz;
3169 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3170 				NFSBZERO(uiop->uio_iov->iov_base, left);
3171 				dp->d_reclen += left;
3172 				uiop->uio_iov->iov_base =
3173 				    (char *)uiop->uio_iov->iov_base + left;
3174 				uiop->uio_iov->iov_len -= left;
3175 				uiop->uio_resid -= left;
3176 				uiop->uio_offset += left;
3177 				blksiz = 0;
3178 			}
3179 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3180 			    uiop->uio_resid)
3181 				bigenough = 0;
3182 			if (bigenough) {
3183 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3184 				dp->d_pad0 = dp->d_pad1 = 0;
3185 				dp->d_off = 0;
3186 				dp->d_namlen = len;
3187 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3188 				    NFSX_HYPER;
3189 				dp->d_type = DT_UNKNOWN;
3190 				blksiz += dp->d_reclen;
3191 				if (blksiz == DIRBLKSIZ)
3192 					blksiz = 0;
3193 				uiop->uio_resid -= DIRHDSIZ;
3194 				uiop->uio_offset += DIRHDSIZ;
3195 				uiop->uio_iov->iov_base =
3196 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3197 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3198 				error = nfsm_mbufuio(nd, uiop, len);
3199 				if (error)
3200 					goto nfsmout;
3201 				cp = uiop->uio_iov->iov_base;
3202 				tlen -= len;
3203 				NFSBZERO(cp, tlen);
3204 				cp += tlen;	/* points to cookie storage */
3205 				tl2 = (u_int32_t *)cp;
3206 				uiop->uio_iov->iov_base =
3207 				    (char *)uiop->uio_iov->iov_base + tlen +
3208 				    NFSX_HYPER;
3209 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3210 				uiop->uio_resid -= tlen + NFSX_HYPER;
3211 				uiop->uio_offset += (tlen + NFSX_HYPER);
3212 			} else {
3213 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3214 				if (error)
3215 					goto nfsmout;
3216 			}
3217 			if (nd->nd_flag & ND_NFSV4) {
3218 				rderr = 0;
3219 				nfsva.na_mntonfileno = UINT64_MAX;
3220 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3221 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3222 				    NULL, NULL, &rderr, p, cred);
3223 				if (error)
3224 					goto nfsmout;
3225 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3226 			} else if (nd->nd_flag & ND_NFSV3) {
3227 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3228 				ncookie.lval[0] = *tl++;
3229 				ncookie.lval[1] = *tl++;
3230 			} else {
3231 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3232 				ncookie.lval[0] = 0;
3233 				ncookie.lval[1] = *tl++;
3234 			}
3235 			if (bigenough) {
3236 			    if (nd->nd_flag & ND_NFSV4) {
3237 				if (rderr) {
3238 				    dp->d_fileno = 0;
3239 				} else {
3240 				    if (gotmnton) {
3241 					if (nfsva.na_mntonfileno != UINT64_MAX)
3242 					    dp->d_fileno = nfsva.na_mntonfileno;
3243 					else
3244 					    dp->d_fileno = nfsva.na_fileid;
3245 				    } else if (nfsva.na_filesid[0] ==
3246 					dnp->n_vattr.na_filesid[0] &&
3247 					nfsva.na_filesid[1] ==
3248 					dnp->n_vattr.na_filesid[1]) {
3249 					dp->d_fileno = nfsva.na_fileid;
3250 				    } else {
3251 					do {
3252 					    fakefileno--;
3253 					} while (fakefileno ==
3254 					    nfsva.na_fileid);
3255 					dp->d_fileno = fakefileno;
3256 				    }
3257 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3258 				}
3259 			    } else {
3260 				dp->d_fileno = nfsva.na_fileid;
3261 			    }
3262 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3263 				ncookie.lval[0];
3264 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3265 				ncookie.lval[1];
3266 			}
3267 			more_dirs = fxdr_unsigned(int, *tl);
3268 		}
3269 		/*
3270 		 * If at end of rpc data, get the eof boolean
3271 		 */
3272 		if (!more_dirs) {
3273 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3274 			eof = fxdr_unsigned(int, *tl);
3275 			if (tryformoredirs)
3276 				more_dirs = !eof;
3277 			if (nd->nd_flag & ND_NFSV4) {
3278 				error = nfscl_postop_attr(nd, nap, attrflagp,
3279 				    stuff);
3280 				if (error)
3281 					goto nfsmout;
3282 			}
3283 		}
3284 		m_freem(nd->nd_mrep);
3285 		nd->nd_mrep = NULL;
3286 	}
3287 	/*
3288 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3289 	 * by increasing d_reclen for the last record.
3290 	 */
3291 	if (blksiz > 0) {
3292 		left = DIRBLKSIZ - blksiz;
3293 		NFSBZERO(uiop->uio_iov->iov_base, left);
3294 		dp->d_reclen += left;
3295 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3296 		    left;
3297 		uiop->uio_iov->iov_len -= left;
3298 		uiop->uio_resid -= left;
3299 		uiop->uio_offset += left;
3300 	}
3301 
3302 	/*
3303 	 * If returning no data, assume end of file.
3304 	 * If not bigenough, return not end of file, since you aren't
3305 	 *    returning all the data
3306 	 * Otherwise, return the eof flag from the server.
3307 	 */
3308 	if (eofp) {
3309 		if (tresid == ((size_t)(uiop->uio_resid)))
3310 			*eofp = 1;
3311 		else if (!bigenough)
3312 			*eofp = 0;
3313 		else
3314 			*eofp = eof;
3315 	}
3316 
3317 	/*
3318 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3319 	 */
3320 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3321 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3322 		NFSBZERO(dp, DIRBLKSIZ);
3323 		dp->d_type = DT_UNKNOWN;
3324 		tl = (u_int32_t *)&dp->d_name[4];
3325 		*tl++ = cookie.lval[0];
3326 		*tl = cookie.lval[1];
3327 		dp->d_reclen = DIRBLKSIZ;
3328 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3329 		    DIRBLKSIZ;
3330 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3331 		uiop->uio_resid -= DIRBLKSIZ;
3332 		uiop->uio_offset += DIRBLKSIZ;
3333 	}
3334 
3335 nfsmout:
3336 	if (nd->nd_mrep != NULL)
3337 		m_freem(nd->nd_mrep);
3338 	return (error);
3339 }
3340 
3341 #ifndef APPLE
3342 /*
3343  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3344  * (Also used for NFS V4 when mount flag set.)
3345  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3346  */
3347 int
3348 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3349     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3350     int *eofp, void *stuff)
3351 {
3352 	int len, left;
3353 	struct dirent *dp = NULL;
3354 	u_int32_t *tl;
3355 	vnode_t newvp = NULLVP;
3356 	struct nfsrv_descript nfsd, *nd = &nfsd;
3357 	struct nameidata nami, *ndp = &nami;
3358 	struct componentname *cnp = &ndp->ni_cnd;
3359 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3360 	struct nfsnode *dnp = VTONFS(vp), *np;
3361 	struct nfsvattr nfsva;
3362 	struct nfsfh *nfhp;
3363 	nfsquad_t cookie, ncookie;
3364 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3365 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3366 	int isdotdot = 0, unlocknewvp = 0;
3367 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3368 	u_int64_t fileno = 0;
3369 	char *cp;
3370 	nfsattrbit_t attrbits, dattrbits;
3371 	size_t tresid;
3372 	u_int32_t *tl2 = NULL, rderr;
3373 	struct timespec dctime;
3374 
3375 	KASSERT(uiop->uio_iovcnt == 1 &&
3376 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3377 	    ("nfs readdirplusrpc bad uio"));
3378 	ncookie.lval[0] = ncookie.lval[1] = 0;
3379 	timespecclear(&dctime);
3380 	*attrflagp = 0;
3381 	if (eofp != NULL)
3382 		*eofp = 0;
3383 	ndp->ni_dvp = vp;
3384 	nd->nd_mrep = NULL;
3385 	cookie.lval[0] = cookiep->nfsuquad[0];
3386 	cookie.lval[1] = cookiep->nfsuquad[1];
3387 	tresid = uiop->uio_resid;
3388 
3389 	/*
3390 	 * For NFSv4, first create the "." and ".." entries.
3391 	 */
3392 	if (NFSHASNFSV4(nmp)) {
3393 		NFSGETATTR_ATTRBIT(&dattrbits);
3394 		NFSZERO_ATTRBIT(&attrbits);
3395 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3396 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3397 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3398 			NFSSETBIT_ATTRBIT(&attrbits,
3399 			    NFSATTRBIT_MOUNTEDONFILEID);
3400 			gotmnton = 1;
3401 		} else {
3402 			/*
3403 			 * Must fake it. Use the fileno, except when the
3404 			 * fsid is != to that of the directory. For that
3405 			 * case, generate a fake fileno that is not the same.
3406 			 */
3407 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3408 			gotmnton = 0;
3409 		}
3410 
3411 		/*
3412 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3413 		 */
3414 		if (uiop->uio_offset == 0) {
3415 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3416 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3417 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3418 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3419 			(void) nfsrv_putattrbit(nd, &attrbits);
3420 			error = nfscl_request(nd, vp, p, cred, stuff);
3421 			if (error)
3422 			    return (error);
3423 			dotfileid = 0;	/* Fake out the compiler. */
3424 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3425 			    error = nfsm_loadattr(nd, &nfsva);
3426 			    if (error != 0)
3427 				goto nfsmout;
3428 			    dctime = nfsva.na_ctime;
3429 			    dotfileid = nfsva.na_fileid;
3430 			}
3431 			if (nd->nd_repstat == 0) {
3432 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3433 			    len = fxdr_unsigned(int, *(tl + 4));
3434 			    if (len > 0 && len <= NFSX_V4FHMAX)
3435 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3436 			    else
3437 				error = EPERM;
3438 			    if (!error) {
3439 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3440 				nfsva.na_mntonfileno = UINT64_MAX;
3441 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3442 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3443 				    NULL, NULL, NULL, p, cred);
3444 				if (error) {
3445 				    dotdotfileid = dotfileid;
3446 				} else if (gotmnton) {
3447 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3448 					dotdotfileid = nfsva.na_mntonfileno;
3449 				    else
3450 					dotdotfileid = nfsva.na_fileid;
3451 				} else if (nfsva.na_filesid[0] ==
3452 				    dnp->n_vattr.na_filesid[0] &&
3453 				    nfsva.na_filesid[1] ==
3454 				    dnp->n_vattr.na_filesid[1]) {
3455 				    dotdotfileid = nfsva.na_fileid;
3456 				} else {
3457 				    do {
3458 					fakefileno--;
3459 				    } while (fakefileno ==
3460 					nfsva.na_fileid);
3461 				    dotdotfileid = fakefileno;
3462 				}
3463 			    }
3464 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3465 			    /*
3466 			     * Lookupp returns NFSERR_NOENT when we are
3467 			     * at the root, so just use the current dir.
3468 			     */
3469 			    nd->nd_repstat = 0;
3470 			    dotdotfileid = dotfileid;
3471 			} else {
3472 			    error = nd->nd_repstat;
3473 			}
3474 			m_freem(nd->nd_mrep);
3475 			if (error)
3476 			    return (error);
3477 			nd->nd_mrep = NULL;
3478 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3479 			dp->d_pad0 = dp->d_pad1 = 0;
3480 			dp->d_off = 0;
3481 			dp->d_type = DT_DIR;
3482 			dp->d_fileno = dotfileid;
3483 			dp->d_namlen = 1;
3484 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3485 			dp->d_name[0] = '.';
3486 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3487 			/*
3488 			 * Just make these offset cookie 0.
3489 			 */
3490 			tl = (u_int32_t *)&dp->d_name[8];
3491 			*tl++ = 0;
3492 			*tl = 0;
3493 			blksiz += dp->d_reclen;
3494 			uiop->uio_resid -= dp->d_reclen;
3495 			uiop->uio_offset += dp->d_reclen;
3496 			uiop->uio_iov->iov_base =
3497 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3498 			uiop->uio_iov->iov_len -= dp->d_reclen;
3499 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3500 			dp->d_pad0 = dp->d_pad1 = 0;
3501 			dp->d_off = 0;
3502 			dp->d_type = DT_DIR;
3503 			dp->d_fileno = dotdotfileid;
3504 			dp->d_namlen = 2;
3505 			*((uint64_t *)dp->d_name) = 0;
3506 			dp->d_name[0] = '.';
3507 			dp->d_name[1] = '.';
3508 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3509 			/*
3510 			 * Just make these offset cookie 0.
3511 			 */
3512 			tl = (u_int32_t *)&dp->d_name[8];
3513 			*tl++ = 0;
3514 			*tl = 0;
3515 			blksiz += dp->d_reclen;
3516 			uiop->uio_resid -= dp->d_reclen;
3517 			uiop->uio_offset += dp->d_reclen;
3518 			uiop->uio_iov->iov_base =
3519 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3520 			uiop->uio_iov->iov_len -= dp->d_reclen;
3521 		}
3522 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3523 		if (gotmnton)
3524 			NFSSETBIT_ATTRBIT(&attrbits,
3525 			    NFSATTRBIT_MOUNTEDONFILEID);
3526 		if (!NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3527 		    NFSATTRBIT_TIMECREATE))
3528 			NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE);
3529 	}
3530 
3531 	/*
3532 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3533 	 * The stopping criteria is EOF or buffer full.
3534 	 */
3535 	while (more_dirs && bigenough) {
3536 		*attrflagp = 0;
3537 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3538  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3539 		*tl++ = cookie.lval[0];
3540 		*tl++ = cookie.lval[1];
3541 		if (cookie.qval == 0) {
3542 			*tl++ = 0;
3543 			*tl++ = 0;
3544 		} else {
3545 			NFSLOCKNODE(dnp);
3546 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3547 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3548 			NFSUNLOCKNODE(dnp);
3549 		}
3550 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3551 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3552 		if (nd->nd_flag & ND_NFSV4) {
3553 			(void) nfsrv_putattrbit(nd, &attrbits);
3554 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3555 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3556 			(void) nfsrv_putattrbit(nd, &dattrbits);
3557 		}
3558 		error = nfscl_request(nd, vp, p, cred, stuff);
3559 		if (error)
3560 			return (error);
3561 		if (nd->nd_flag & ND_NFSV3)
3562 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3563 		if (nd->nd_repstat || error) {
3564 			if (!error)
3565 				error = nd->nd_repstat;
3566 			goto nfsmout;
3567 		}
3568 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3569 			dctime = nap->na_ctime;
3570 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3571 		NFSLOCKNODE(dnp);
3572 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3573 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3574 		NFSUNLOCKNODE(dnp);
3575 		more_dirs = fxdr_unsigned(int, *tl);
3576 		if (!more_dirs)
3577 			tryformoredirs = 0;
3578 
3579 		/* loop through the dir entries, doctoring them to 4bsd form */
3580 		while (more_dirs && bigenough) {
3581 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3582 			if (nd->nd_flag & ND_NFSV4) {
3583 				ncookie.lval[0] = *tl++;
3584 				ncookie.lval[1] = *tl++;
3585 			} else {
3586 				fileno = fxdr_hyper(tl);
3587 				tl += 2;
3588 			}
3589 			len = fxdr_unsigned(int, *tl);
3590 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3591 				error = EBADRPC;
3592 				goto nfsmout;
3593 			}
3594 			tlen = roundup2(len, 8);
3595 			if (tlen == len)
3596 				tlen += 8;  /* To ensure null termination. */
3597 			left = DIRBLKSIZ - blksiz;
3598 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3599 				NFSBZERO(uiop->uio_iov->iov_base, left);
3600 				dp->d_reclen += left;
3601 				uiop->uio_iov->iov_base =
3602 				    (char *)uiop->uio_iov->iov_base + left;
3603 				uiop->uio_iov->iov_len -= left;
3604 				uiop->uio_resid -= left;
3605 				uiop->uio_offset += left;
3606 				blksiz = 0;
3607 			}
3608 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3609 			    uiop->uio_resid)
3610 				bigenough = 0;
3611 			if (bigenough) {
3612 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3613 				dp->d_pad0 = dp->d_pad1 = 0;
3614 				dp->d_off = 0;
3615 				dp->d_namlen = len;
3616 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3617 				    NFSX_HYPER;
3618 				dp->d_type = DT_UNKNOWN;
3619 				blksiz += dp->d_reclen;
3620 				if (blksiz == DIRBLKSIZ)
3621 					blksiz = 0;
3622 				uiop->uio_resid -= DIRHDSIZ;
3623 				uiop->uio_offset += DIRHDSIZ;
3624 				uiop->uio_iov->iov_base =
3625 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3626 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3627 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3628 				cnp->cn_namelen = len;
3629 				NFSCNHASHZERO(cnp);
3630 				error = nfsm_mbufuio(nd, uiop, len);
3631 				if (error)
3632 					goto nfsmout;
3633 				cp = uiop->uio_iov->iov_base;
3634 				tlen -= len;
3635 				NFSBZERO(cp, tlen);
3636 				cp += tlen;	/* points to cookie storage */
3637 				tl2 = (u_int32_t *)cp;
3638 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3639 				    cnp->cn_nameptr[1] == '.')
3640 					isdotdot = 1;
3641 				else
3642 					isdotdot = 0;
3643 				uiop->uio_iov->iov_base =
3644 				    (char *)uiop->uio_iov->iov_base + tlen +
3645 				    NFSX_HYPER;
3646 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3647 				uiop->uio_resid -= tlen + NFSX_HYPER;
3648 				uiop->uio_offset += (tlen + NFSX_HYPER);
3649 			} else {
3650 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3651 				if (error)
3652 					goto nfsmout;
3653 			}
3654 			nfhp = NULL;
3655 			if (nd->nd_flag & ND_NFSV3) {
3656 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3657 				ncookie.lval[0] = *tl++;
3658 				ncookie.lval[1] = *tl++;
3659 				attrflag = fxdr_unsigned(int, *tl);
3660 				if (attrflag) {
3661 				  error = nfsm_loadattr(nd, &nfsva);
3662 				  if (error)
3663 					goto nfsmout;
3664 				}
3665 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3666 				if (*tl) {
3667 					error = nfsm_getfh(nd, &nfhp);
3668 					if (error)
3669 					    goto nfsmout;
3670 				}
3671 				if (!attrflag && nfhp != NULL) {
3672 					free(nfhp, M_NFSFH);
3673 					nfhp = NULL;
3674 				}
3675 			} else {
3676 				rderr = 0;
3677 				nfsva.na_mntonfileno = 0xffffffff;
3678 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3679 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3680 				    NULL, NULL, &rderr, p, cred);
3681 				if (error)
3682 					goto nfsmout;
3683 			}
3684 
3685 			if (bigenough) {
3686 			    if (nd->nd_flag & ND_NFSV4) {
3687 				if (rderr) {
3688 				    dp->d_fileno = 0;
3689 				} else if (gotmnton) {
3690 				    if (nfsva.na_mntonfileno != 0xffffffff)
3691 					dp->d_fileno = nfsva.na_mntonfileno;
3692 				    else
3693 					dp->d_fileno = nfsva.na_fileid;
3694 				} else if (nfsva.na_filesid[0] ==
3695 				    dnp->n_vattr.na_filesid[0] &&
3696 				    nfsva.na_filesid[1] ==
3697 				    dnp->n_vattr.na_filesid[1]) {
3698 				    dp->d_fileno = nfsva.na_fileid;
3699 				} else {
3700 				    do {
3701 					fakefileno--;
3702 				    } while (fakefileno ==
3703 					nfsva.na_fileid);
3704 				    dp->d_fileno = fakefileno;
3705 				}
3706 			    } else {
3707 				dp->d_fileno = fileno;
3708 			    }
3709 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3710 				ncookie.lval[0];
3711 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3712 				ncookie.lval[1];
3713 
3714 			    if (nfhp != NULL) {
3715 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3716 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3717 				    VREF(vp);
3718 				    newvp = vp;
3719 				    unlocknewvp = 0;
3720 				    free(nfhp, M_NFSFH);
3721 				    np = dnp;
3722 				} else if (isdotdot != 0) {
3723 				    /*
3724 				     * Skip doing a nfscl_nget() call for "..".
3725 				     * There's a race between acquiring the nfs
3726 				     * node here and lookups that look for the
3727 				     * directory being read (in the parent).
3728 				     * It would try to get a lock on ".." here,
3729 				     * owning the lock on the directory being
3730 				     * read. Lookup will hold the lock on ".."
3731 				     * and try to acquire the lock on the
3732 				     * directory being read.
3733 				     * If the directory is unlocked/relocked,
3734 				     * then there is a LOR with the buflock
3735 				     * vp is relocked.
3736 				     */
3737 				    free(nfhp, M_NFSFH);
3738 				} else {
3739 				    error = nfscl_nget(vp->v_mount, vp,
3740 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3741 				    if (!error) {
3742 					newvp = NFSTOV(np);
3743 					unlocknewvp = 1;
3744 				    }
3745 				}
3746 				nfhp = NULL;
3747 				if (newvp != NULLVP) {
3748 				    error = nfscl_loadattrcache(&newvp,
3749 					&nfsva, NULL, NULL, 0, 0);
3750 				    if (error) {
3751 					if (unlocknewvp)
3752 					    vput(newvp);
3753 					else
3754 					    vrele(newvp);
3755 					goto nfsmout;
3756 				    }
3757 				    dp->d_type =
3758 					vtonfs_dtype(np->n_vattr.na_type);
3759 				    ndp->ni_vp = newvp;
3760 				    NFSCNHASH(cnp, HASHINIT);
3761 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3762 					ndp->ni_dvp != ndp->ni_vp &&
3763 					(newvp->v_type != VDIR ||
3764 					 dctime.tv_sec != 0)) {
3765 					cache_enter_time(ndp->ni_dvp,
3766 					    ndp->ni_vp, cnp,
3767 					    &nfsva.na_ctime,
3768 					    newvp->v_type != VDIR ? NULL :
3769 					    &dctime);
3770 				    }
3771 				    if (unlocknewvp)
3772 					vput(newvp);
3773 				    else
3774 					vrele(newvp);
3775 				    newvp = NULLVP;
3776 				}
3777 			    }
3778 			} else if (nfhp != NULL) {
3779 			    free(nfhp, M_NFSFH);
3780 			}
3781 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3782 			more_dirs = fxdr_unsigned(int, *tl);
3783 		}
3784 		/*
3785 		 * If at end of rpc data, get the eof boolean
3786 		 */
3787 		if (!more_dirs) {
3788 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3789 			eof = fxdr_unsigned(int, *tl);
3790 			if (tryformoredirs)
3791 				more_dirs = !eof;
3792 			if (nd->nd_flag & ND_NFSV4) {
3793 				error = nfscl_postop_attr(nd, nap, attrflagp,
3794 				    stuff);
3795 				if (error)
3796 					goto nfsmout;
3797 			}
3798 		}
3799 		m_freem(nd->nd_mrep);
3800 		nd->nd_mrep = NULL;
3801 	}
3802 	/*
3803 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3804 	 * by increasing d_reclen for the last record.
3805 	 */
3806 	if (blksiz > 0) {
3807 		left = DIRBLKSIZ - blksiz;
3808 		NFSBZERO(uiop->uio_iov->iov_base, left);
3809 		dp->d_reclen += left;
3810 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3811 		    left;
3812 		uiop->uio_iov->iov_len -= left;
3813 		uiop->uio_resid -= left;
3814 		uiop->uio_offset += left;
3815 	}
3816 
3817 	/*
3818 	 * If returning no data, assume end of file.
3819 	 * If not bigenough, return not end of file, since you aren't
3820 	 *    returning all the data
3821 	 * Otherwise, return the eof flag from the server.
3822 	 */
3823 	if (eofp != NULL) {
3824 		if (tresid == uiop->uio_resid)
3825 			*eofp = 1;
3826 		else if (!bigenough)
3827 			*eofp = 0;
3828 		else
3829 			*eofp = eof;
3830 	}
3831 
3832 	/*
3833 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3834 	 */
3835 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3836 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3837 		NFSBZERO(dp, DIRBLKSIZ);
3838 		dp->d_type = DT_UNKNOWN;
3839 		tl = (u_int32_t *)&dp->d_name[4];
3840 		*tl++ = cookie.lval[0];
3841 		*tl = cookie.lval[1];
3842 		dp->d_reclen = DIRBLKSIZ;
3843 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3844 		    DIRBLKSIZ;
3845 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3846 		uiop->uio_resid -= DIRBLKSIZ;
3847 		uiop->uio_offset += DIRBLKSIZ;
3848 	}
3849 
3850 nfsmout:
3851 	if (nd->nd_mrep != NULL)
3852 		m_freem(nd->nd_mrep);
3853 	return (error);
3854 }
3855 #endif	/* !APPLE */
3856 
3857 /*
3858  * Nfs commit rpc
3859  */
3860 int
3861 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3862     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3863 {
3864 	u_int32_t *tl;
3865 	struct nfsrv_descript nfsd, *nd = &nfsd;
3866 	nfsattrbit_t attrbits;
3867 	int error;
3868 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3869 
3870 	*attrflagp = 0;
3871 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3872 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3873 	txdr_hyper(offset, tl);
3874 	tl += 2;
3875 	*tl = txdr_unsigned(cnt);
3876 	if (nd->nd_flag & ND_NFSV4) {
3877 		/*
3878 		 * And do a Getattr op.
3879 		 */
3880 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3881 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3882 		NFSGETATTR_ATTRBIT(&attrbits);
3883 		(void) nfsrv_putattrbit(nd, &attrbits);
3884 	}
3885 	error = nfscl_request(nd, vp, p, cred, stuff);
3886 	if (error)
3887 		return (error);
3888 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3889 	if (!error && !nd->nd_repstat) {
3890 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3891 		NFSLOCKMNT(nmp);
3892 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3893 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3894 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3895 		}
3896 		NFSUNLOCKMNT(nmp);
3897 		if (nd->nd_flag & ND_NFSV4)
3898 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3899 	}
3900 nfsmout:
3901 	if (!error && nd->nd_repstat)
3902 		error = nd->nd_repstat;
3903 	m_freem(nd->nd_mrep);
3904 	return (error);
3905 }
3906 
3907 /*
3908  * NFS byte range lock rpc.
3909  * (Mostly just calls one of the three lower level RPC routines.)
3910  */
3911 int
3912 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3913     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3914 {
3915 	struct nfscllockowner *lp;
3916 	struct nfsclclient *clp;
3917 	struct nfsfh *nfhp;
3918 	struct nfsrv_descript nfsd, *nd = &nfsd;
3919 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
3920 	u_int64_t off, len;
3921 	off_t start, end;
3922 	u_int32_t clidrev = 0;
3923 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3924 	int callcnt, dorpc;
3925 
3926 	/*
3927 	 * Convert the flock structure into a start and end and do POSIX
3928 	 * bounds checking.
3929 	 */
3930 	switch (fl->l_whence) {
3931 	case SEEK_SET:
3932 	case SEEK_CUR:
3933 		/*
3934 		 * Caller is responsible for adding any necessary offset
3935 		 * when SEEK_CUR is used.
3936 		 */
3937 		start = fl->l_start;
3938 		off = fl->l_start;
3939 		break;
3940 	case SEEK_END:
3941 		start = size + fl->l_start;
3942 		off = size + fl->l_start;
3943 		break;
3944 	default:
3945 		return (EINVAL);
3946 	}
3947 	if (start < 0)
3948 		return (EINVAL);
3949 	if (fl->l_len != 0) {
3950 		end = start + fl->l_len - 1;
3951 		if (end < start)
3952 			return (EINVAL);
3953 	}
3954 
3955 	len = fl->l_len;
3956 	if (len == 0)
3957 		len = NFS64BITSSET;
3958 	retrycnt = 0;
3959 	do {
3960 	    nd->nd_repstat = 0;
3961 	    if (op == F_GETLK) {
3962 		error = nfscl_getcl(vp->v_mount, cred, p, false, &clp);
3963 		if (error)
3964 			return (error);
3965 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3966 		if (!error) {
3967 			clidrev = clp->nfsc_clientidrev;
3968 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3969 			    p, id, flags);
3970 		} else if (error == -1) {
3971 			error = 0;
3972 		}
3973 		nfscl_clientrelease(clp);
3974 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3975 		/*
3976 		 * We must loop around for all lockowner cases.
3977 		 */
3978 		callcnt = 0;
3979 		error = nfscl_getcl(vp->v_mount, cred, p, false, &clp);
3980 		if (error)
3981 			return (error);
3982 		do {
3983 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3984 			clp, id, flags, &lp, &dorpc);
3985 		    /*
3986 		     * If it returns a NULL lp, we're done.
3987 		     */
3988 		    if (lp == NULL) {
3989 			if (callcnt == 0)
3990 			    nfscl_clientrelease(clp);
3991 			else
3992 			    nfscl_releasealllocks(clp, vp, p, id, flags);
3993 			return (error);
3994 		    }
3995 		    if (nmp->nm_clp != NULL)
3996 			clidrev = nmp->nm_clp->nfsc_clientidrev;
3997 		    else
3998 			clidrev = 0;
3999 		    /*
4000 		     * If the server doesn't support Posix lock semantics,
4001 		     * only allow locks on the entire file, since it won't
4002 		     * handle overlapping byte ranges.
4003 		     * There might still be a problem when a lock
4004 		     * upgrade/downgrade (read<->write) occurs, since the
4005 		     * server "might" expect an unlock first?
4006 		     */
4007 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4008 			(off == 0 && len == NFS64BITSSET))) {
4009 			/*
4010 			 * Since the lock records will go away, we must
4011 			 * wait for grace and delay here.
4012 			 */
4013 			do {
4014 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4015 				NFSV4LOCKT_READ, cred, p, 0);
4016 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4017 				 nd->nd_repstat == NFSERR_DELAY) &&
4018 				error == 0)
4019 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4020 				    "nfs_advlock");
4021 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4022 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4023 		    }
4024 		    callcnt++;
4025 		} while (error == 0 && nd->nd_repstat == 0);
4026 		nfscl_releasealllocks(clp, vp, p, id, flags);
4027 	    } else if (op == F_SETLK) {
4028 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4029 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4030 		if (error || donelocally) {
4031 			return (error);
4032 		}
4033 		if (nmp->nm_clp != NULL)
4034 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4035 		else
4036 			clidrev = 0;
4037 		nfhp = VTONFS(vp)->n_fhp;
4038 		if (!lp->nfsl_open->nfso_posixlock &&
4039 		    (off != 0 || len != NFS64BITSSET)) {
4040 			error = EINVAL;
4041 		} else {
4042 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4043 			    nfhp->nfh_len, lp, newone, reclaim, off,
4044 			    len, fl->l_type, cred, p, 0);
4045 		}
4046 		if (!error)
4047 			error = nd->nd_repstat;
4048 		nfscl_lockrelease(lp, error, newone);
4049 	    } else {
4050 		error = EINVAL;
4051 	    }
4052 	    if (!error)
4053 	        error = nd->nd_repstat;
4054 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4055 		error == NFSERR_STALEDONTRECOVER ||
4056 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4057 		error == NFSERR_BADSESSION) {
4058 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4059 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4060 		&& clidrev != 0) {
4061 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4062 		retrycnt++;
4063 	    }
4064 	} while (error == NFSERR_GRACE ||
4065 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4066 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4067 	    error == NFSERR_BADSESSION ||
4068 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4069 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4070 	if (error && retrycnt >= 4)
4071 		error = EIO;
4072 	return (error);
4073 }
4074 
4075 /*
4076  * The lower level routine for the LockT case.
4077  */
4078 int
4079 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4080     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4081     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4082 {
4083 	u_int32_t *tl;
4084 	int error, type, size;
4085 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4086 	struct nfsnode *np;
4087 	struct nfsmount *nmp;
4088 	struct nfsclsession *tsep;
4089 
4090 	nmp = VFSTONFS(vp->v_mount);
4091 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4092 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4093 	if (fl->l_type == F_RDLCK)
4094 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4095 	else
4096 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4097 	txdr_hyper(off, tl);
4098 	tl += 2;
4099 	txdr_hyper(len, tl);
4100 	tl += 2;
4101 	tsep = nfsmnt_mdssession(nmp);
4102 	*tl++ = tsep->nfsess_clientid.lval[0];
4103 	*tl = tsep->nfsess_clientid.lval[1];
4104 	nfscl_filllockowner(id, own, flags);
4105 	np = VTONFS(vp);
4106 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4107 	    np->n_fhp->nfh_len);
4108 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4109 	error = nfscl_request(nd, vp, p, cred, NULL);
4110 	if (error)
4111 		return (error);
4112 	if (nd->nd_repstat == 0) {
4113 		fl->l_type = F_UNLCK;
4114 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4115 		nd->nd_repstat = 0;
4116 		fl->l_whence = SEEK_SET;
4117 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4118 		fl->l_start = fxdr_hyper(tl);
4119 		tl += 2;
4120 		len = fxdr_hyper(tl);
4121 		tl += 2;
4122 		if (len == NFS64BITSSET)
4123 			fl->l_len = 0;
4124 		else
4125 			fl->l_len = len;
4126 		type = fxdr_unsigned(int, *tl++);
4127 		if (type == NFSV4LOCKT_WRITE)
4128 			fl->l_type = F_WRLCK;
4129 		else
4130 			fl->l_type = F_RDLCK;
4131 		/*
4132 		 * XXX For now, I have no idea what to do with the
4133 		 * conflicting lock_owner, so I'll just set the pid == 0
4134 		 * and skip over the lock_owner.
4135 		 */
4136 		fl->l_pid = (pid_t)0;
4137 		tl += 2;
4138 		size = fxdr_unsigned(int, *tl);
4139 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4140 			error = EBADRPC;
4141 		if (!error)
4142 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4143 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4144 		nfscl_initiate_recovery(clp);
4145 nfsmout:
4146 	m_freem(nd->nd_mrep);
4147 	return (error);
4148 }
4149 
4150 /*
4151  * Lower level function that performs the LockU RPC.
4152  */
4153 static int
4154 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4155     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4156     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4157 {
4158 	u_int32_t *tl;
4159 	int error;
4160 
4161 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4162 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4163 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4164 	*tl++ = txdr_unsigned(type);
4165 	*tl = txdr_unsigned(lp->nfsl_seqid);
4166 	if (nfstest_outofseq &&
4167 	    (arc4random() % nfstest_outofseq) == 0)
4168 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4169 	tl++;
4170 	if (NFSHASNFSV4N(nmp))
4171 		*tl++ = 0;
4172 	else
4173 		*tl++ = lp->nfsl_stateid.seqid;
4174 	*tl++ = lp->nfsl_stateid.other[0];
4175 	*tl++ = lp->nfsl_stateid.other[1];
4176 	*tl++ = lp->nfsl_stateid.other[2];
4177 	txdr_hyper(off, tl);
4178 	tl += 2;
4179 	txdr_hyper(len, tl);
4180 	if (syscred)
4181 		nd->nd_flag |= ND_USEGSSNAME;
4182 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4183 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4184 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4185 	if (error)
4186 		return (error);
4187 	if (nd->nd_repstat == 0) {
4188 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4189 		lp->nfsl_stateid.seqid = *tl++;
4190 		lp->nfsl_stateid.other[0] = *tl++;
4191 		lp->nfsl_stateid.other[1] = *tl++;
4192 		lp->nfsl_stateid.other[2] = *tl;
4193 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4194 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4195 nfsmout:
4196 	m_freem(nd->nd_mrep);
4197 	return (error);
4198 }
4199 
4200 /*
4201  * The actual Lock RPC.
4202  */
4203 int
4204 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4205     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4206     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4207     NFSPROC_T *p, int syscred)
4208 {
4209 	u_int32_t *tl;
4210 	int error, size;
4211 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4212 	struct nfsclsession *tsep;
4213 
4214 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4215 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4216 	if (type == F_RDLCK)
4217 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4218 	else
4219 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4220 	*tl++ = txdr_unsigned(reclaim);
4221 	txdr_hyper(off, tl);
4222 	tl += 2;
4223 	txdr_hyper(len, tl);
4224 	tl += 2;
4225 	if (newone) {
4226 	    *tl = newnfs_true;
4227 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4228 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4229 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4230 	    if (NFSHASNFSV4N(nmp))
4231 		*tl++ = 0;
4232 	    else
4233 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4234 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4235 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4236 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4237 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4238 	    tsep = nfsmnt_mdssession(nmp);
4239 	    *tl++ = tsep->nfsess_clientid.lval[0];
4240 	    *tl = tsep->nfsess_clientid.lval[1];
4241 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4242 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4243 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4244 	} else {
4245 	    *tl = newnfs_false;
4246 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4247 	    if (NFSHASNFSV4N(nmp))
4248 		*tl++ = 0;
4249 	    else
4250 		*tl++ = lp->nfsl_stateid.seqid;
4251 	    *tl++ = lp->nfsl_stateid.other[0];
4252 	    *tl++ = lp->nfsl_stateid.other[1];
4253 	    *tl++ = lp->nfsl_stateid.other[2];
4254 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4255 	    if (nfstest_outofseq &&
4256 		(arc4random() % nfstest_outofseq) == 0)
4257 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4258 	}
4259 	if (syscred)
4260 		nd->nd_flag |= ND_USEGSSNAME;
4261 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4262 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4263 	if (error)
4264 		return (error);
4265 	if (newone)
4266 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4267 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4268 	if (nd->nd_repstat == 0) {
4269 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4270 		lp->nfsl_stateid.seqid = *tl++;
4271 		lp->nfsl_stateid.other[0] = *tl++;
4272 		lp->nfsl_stateid.other[1] = *tl++;
4273 		lp->nfsl_stateid.other[2] = *tl;
4274 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4275 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4276 		size = fxdr_unsigned(int, *(tl + 7));
4277 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4278 			error = EBADRPC;
4279 		if (!error)
4280 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4281 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4282 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4283 nfsmout:
4284 	m_freem(nd->nd_mrep);
4285 	return (error);
4286 }
4287 
4288 /*
4289  * nfs statfs rpc
4290  * (always called with the vp for the mount point)
4291  */
4292 int
4293 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4294     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4295     void *stuff)
4296 {
4297 	u_int32_t *tl = NULL;
4298 	struct nfsrv_descript nfsd, *nd = &nfsd;
4299 	struct nfsmount *nmp;
4300 	nfsattrbit_t attrbits;
4301 	int error;
4302 
4303 	*attrflagp = 0;
4304 	nmp = VFSTONFS(vp->v_mount);
4305 	if (NFSHASNFSV4(nmp)) {
4306 		/*
4307 		 * For V4, you actually do a getattr.
4308 		 */
4309 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4310 		NFSSTATFS_GETATTRBIT(&attrbits);
4311 		(void) nfsrv_putattrbit(nd, &attrbits);
4312 		nd->nd_flag |= ND_USEGSSNAME;
4313 		error = nfscl_request(nd, vp, p, cred, stuff);
4314 		if (error)
4315 			return (error);
4316 		if (nd->nd_repstat == 0) {
4317 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4318 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4319 			    cred);
4320 			if (!error) {
4321 				nmp->nm_fsid[0] = nap->na_filesid[0];
4322 				nmp->nm_fsid[1] = nap->na_filesid[1];
4323 				NFSSETHASSETFSID(nmp);
4324 				*attrflagp = 1;
4325 			}
4326 		} else {
4327 			error = nd->nd_repstat;
4328 		}
4329 		if (error)
4330 			goto nfsmout;
4331 	} else {
4332 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4333 		error = nfscl_request(nd, vp, p, cred, stuff);
4334 		if (error)
4335 			return (error);
4336 		if (nd->nd_flag & ND_NFSV3) {
4337 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4338 			if (error)
4339 				goto nfsmout;
4340 		}
4341 		if (nd->nd_repstat) {
4342 			error = nd->nd_repstat;
4343 			goto nfsmout;
4344 		}
4345 		NFSM_DISSECT(tl, u_int32_t *,
4346 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4347 	}
4348 	if (NFSHASNFSV3(nmp)) {
4349 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4350 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4351 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4352 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4353 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4354 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4355 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4356 	} else if (NFSHASNFSV4(nmp) == 0) {
4357 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4358 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4359 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4360 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4361 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4362 	}
4363 nfsmout:
4364 	m_freem(nd->nd_mrep);
4365 	return (error);
4366 }
4367 
4368 /*
4369  * nfs pathconf rpc
4370  */
4371 int
4372 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4373     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4374     void *stuff)
4375 {
4376 	struct nfsrv_descript nfsd, *nd = &nfsd;
4377 	struct nfsmount *nmp;
4378 	u_int32_t *tl;
4379 	nfsattrbit_t attrbits;
4380 	int error;
4381 
4382 	*attrflagp = 0;
4383 	nmp = VFSTONFS(vp->v_mount);
4384 	if (NFSHASNFSV4(nmp)) {
4385 		/*
4386 		 * For V4, you actually do a getattr.
4387 		 */
4388 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4389 		NFSPATHCONF_GETATTRBIT(&attrbits);
4390 		(void) nfsrv_putattrbit(nd, &attrbits);
4391 		nd->nd_flag |= ND_USEGSSNAME;
4392 		error = nfscl_request(nd, vp, p, cred, stuff);
4393 		if (error)
4394 			return (error);
4395 		if (nd->nd_repstat == 0) {
4396 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4397 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4398 			    cred);
4399 			if (!error)
4400 				*attrflagp = 1;
4401 		} else {
4402 			error = nd->nd_repstat;
4403 		}
4404 	} else {
4405 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4406 		error = nfscl_request(nd, vp, p, cred, stuff);
4407 		if (error)
4408 			return (error);
4409 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4410 		if (nd->nd_repstat && !error)
4411 			error = nd->nd_repstat;
4412 		if (!error) {
4413 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4414 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4415 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4416 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4417 			pc->pc_chownrestricted =
4418 			    fxdr_unsigned(u_int32_t, *tl++);
4419 			pc->pc_caseinsensitive =
4420 			    fxdr_unsigned(u_int32_t, *tl++);
4421 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4422 		}
4423 	}
4424 nfsmout:
4425 	m_freem(nd->nd_mrep);
4426 	return (error);
4427 }
4428 
4429 /*
4430  * nfs version 3 fsinfo rpc call
4431  */
4432 int
4433 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4434     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4435 {
4436 	u_int32_t *tl;
4437 	struct nfsrv_descript nfsd, *nd = &nfsd;
4438 	int error;
4439 
4440 	*attrflagp = 0;
4441 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4442 	error = nfscl_request(nd, vp, p, cred, stuff);
4443 	if (error)
4444 		return (error);
4445 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4446 	if (nd->nd_repstat && !error)
4447 		error = nd->nd_repstat;
4448 	if (!error) {
4449 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4450 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4451 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4452 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4453 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4454 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4455 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4456 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4457 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4458 		tl += 2;
4459 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4460 		tl += 2;
4461 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4462 	}
4463 nfsmout:
4464 	m_freem(nd->nd_mrep);
4465 	return (error);
4466 }
4467 
4468 /*
4469  * This function performs the Renew RPC.
4470  */
4471 int
4472 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4473     NFSPROC_T *p)
4474 {
4475 	u_int32_t *tl;
4476 	struct nfsrv_descript nfsd;
4477 	struct nfsrv_descript *nd = &nfsd;
4478 	struct nfsmount *nmp;
4479 	int error;
4480 	struct nfssockreq *nrp;
4481 	struct nfsclsession *tsep;
4482 
4483 	nmp = clp->nfsc_nmp;
4484 	if (nmp == NULL)
4485 		return (0);
4486 	if (dsp == NULL)
4487 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4488 		    0);
4489 	else
4490 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4491 		    &dsp->nfsclds_sess, 0, 0);
4492 	if (!NFSHASNFSV4N(nmp)) {
4493 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4494 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4495 		tsep = nfsmnt_mdssession(nmp);
4496 		*tl++ = tsep->nfsess_clientid.lval[0];
4497 		*tl = tsep->nfsess_clientid.lval[1];
4498 	}
4499 	nrp = NULL;
4500 	if (dsp != NULL)
4501 		nrp = dsp->nfsclds_sockp;
4502 	if (nrp == NULL)
4503 		/* If NULL, use the MDS socket. */
4504 		nrp = &nmp->nm_sockreq;
4505 	nd->nd_flag |= ND_USEGSSNAME;
4506 	if (dsp == NULL)
4507 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4508 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4509 	else {
4510 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4511 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4512 		if (error == ENXIO)
4513 			nfscl_cancelreqs(dsp);
4514 	}
4515 	if (error)
4516 		return (error);
4517 	error = nd->nd_repstat;
4518 	m_freem(nd->nd_mrep);
4519 	return (error);
4520 }
4521 
4522 /*
4523  * This function performs the Releaselockowner RPC.
4524  */
4525 int
4526 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4527     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4528 {
4529 	struct nfsrv_descript nfsd, *nd = &nfsd;
4530 	u_int32_t *tl;
4531 	int error;
4532 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4533 	struct nfsclsession *tsep;
4534 
4535 	if (NFSHASNFSV4N(nmp)) {
4536 		/* For NFSv4.1, do a FreeStateID. */
4537 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4538 		    NULL, 0, 0);
4539 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4540 	} else {
4541 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4542 		    NULL, 0, 0);
4543 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4544 		tsep = nfsmnt_mdssession(nmp);
4545 		*tl++ = tsep->nfsess_clientid.lval[0];
4546 		*tl = tsep->nfsess_clientid.lval[1];
4547 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4548 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4549 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4550 	}
4551 	nd->nd_flag |= ND_USEGSSNAME;
4552 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4553 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4554 	if (error)
4555 		return (error);
4556 	error = nd->nd_repstat;
4557 	m_freem(nd->nd_mrep);
4558 	return (error);
4559 }
4560 
4561 /*
4562  * This function performs the Compound to get the mount pt FH.
4563  */
4564 int
4565 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4566     NFSPROC_T *p)
4567 {
4568 	u_int32_t *tl;
4569 	struct nfsrv_descript nfsd;
4570 	struct nfsrv_descript *nd = &nfsd;
4571 	u_char *cp, *cp2;
4572 	int error, cnt, len, setnil;
4573 	u_int32_t *opcntp;
4574 
4575 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4576 	    0);
4577 	cp = dirpath;
4578 	cnt = 0;
4579 	do {
4580 		setnil = 0;
4581 		while (*cp == '/')
4582 			cp++;
4583 		cp2 = cp;
4584 		while (*cp2 != '\0' && *cp2 != '/')
4585 			cp2++;
4586 		if (*cp2 == '/') {
4587 			setnil = 1;
4588 			*cp2 = '\0';
4589 		}
4590 		if (cp2 != cp) {
4591 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4592 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4593 			nfsm_strtom(nd, cp, strlen(cp));
4594 			cnt++;
4595 		}
4596 		if (setnil)
4597 			*cp2++ = '/';
4598 		cp = cp2;
4599 	} while (*cp != '\0');
4600 	if (NFSHASNFSV4N(nmp))
4601 		/* Has a Sequence Op done by nfscl_reqstart(). */
4602 		*opcntp = txdr_unsigned(3 + cnt);
4603 	else
4604 		*opcntp = txdr_unsigned(2 + cnt);
4605 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4606 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4607 	nd->nd_flag |= ND_USEGSSNAME;
4608 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4609 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4610 	if (error)
4611 		return (error);
4612 	if (nd->nd_repstat == 0) {
4613 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4614 		tl += (2 + 2 * cnt);
4615 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4616 			len > NFSX_FHMAX) {
4617 			nd->nd_repstat = NFSERR_BADXDR;
4618 		} else {
4619 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4620 			if (nd->nd_repstat == 0)
4621 				nmp->nm_fhsize = len;
4622 		}
4623 	}
4624 	error = nd->nd_repstat;
4625 nfsmout:
4626 	m_freem(nd->nd_mrep);
4627 	return (error);
4628 }
4629 
4630 /*
4631  * This function performs the Delegreturn RPC.
4632  */
4633 int
4634 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4635     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4636 {
4637 	u_int32_t *tl;
4638 	struct nfsrv_descript nfsd;
4639 	struct nfsrv_descript *nd = &nfsd;
4640 	int error;
4641 
4642 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4643 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4644 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4645 	if (NFSHASNFSV4N(nmp))
4646 		*tl++ = 0;
4647 	else
4648 		*tl++ = dp->nfsdl_stateid.seqid;
4649 	*tl++ = dp->nfsdl_stateid.other[0];
4650 	*tl++ = dp->nfsdl_stateid.other[1];
4651 	*tl = dp->nfsdl_stateid.other[2];
4652 	if (syscred)
4653 		nd->nd_flag |= ND_USEGSSNAME;
4654 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4655 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4656 	if (error)
4657 		return (error);
4658 	error = nd->nd_repstat;
4659 	m_freem(nd->nd_mrep);
4660 	return (error);
4661 }
4662 
4663 /*
4664  * nfs getacl call.
4665  */
4666 int
4667 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4668     struct acl *aclp, void *stuff)
4669 {
4670 	struct nfsrv_descript nfsd, *nd = &nfsd;
4671 	int error;
4672 	nfsattrbit_t attrbits;
4673 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4674 
4675 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4676 		return (EOPNOTSUPP);
4677 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4678 	NFSZERO_ATTRBIT(&attrbits);
4679 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4680 	(void) nfsrv_putattrbit(nd, &attrbits);
4681 	error = nfscl_request(nd, vp, p, cred, stuff);
4682 	if (error)
4683 		return (error);
4684 	if (!nd->nd_repstat)
4685 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4686 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4687 	else
4688 		error = nd->nd_repstat;
4689 	m_freem(nd->nd_mrep);
4690 	return (error);
4691 }
4692 
4693 /*
4694  * nfs setacl call.
4695  */
4696 int
4697 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4698     struct acl *aclp, void *stuff)
4699 {
4700 	int error;
4701 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4702 
4703 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4704 		return (EOPNOTSUPP);
4705 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4706 	return (error);
4707 }
4708 
4709 /*
4710  * nfs setacl call.
4711  */
4712 static int
4713 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4714     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4715 {
4716 	struct nfsrv_descript nfsd, *nd = &nfsd;
4717 	int error;
4718 	nfsattrbit_t attrbits;
4719 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
4720 
4721 	if (!NFSHASNFSV4(nmp))
4722 		return (EOPNOTSUPP);
4723 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4724 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4725 	NFSZERO_ATTRBIT(&attrbits);
4726 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4727 	(void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0,
4728 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4729 	error = nfscl_request(nd, vp, p, cred, stuff);
4730 	if (error)
4731 		return (error);
4732 	/* Don't care about the pre/postop attributes */
4733 	m_freem(nd->nd_mrep);
4734 	return (nd->nd_repstat);
4735 }
4736 
4737 /*
4738  * Do the NFSv4.1 Exchange ID.
4739  */
4740 int
4741 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4742     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
4743     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
4744 {
4745 	uint32_t *tl, v41flags;
4746 	struct nfsrv_descript nfsd;
4747 	struct nfsrv_descript *nd = &nfsd;
4748 	struct nfsclds *dsp;
4749 	struct timespec verstime;
4750 	int error, len;
4751 
4752 	*dspp = NULL;
4753 	if (minorvers == 0)
4754 		minorvers = nmp->nm_minorvers;
4755 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
4756 	    NFS_VER4, minorvers);
4757 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4758 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4759 	*tl = txdr_unsigned(clp->nfsc_rev);
4760 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4761 
4762 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4763 	*tl++ = txdr_unsigned(exchflags);
4764 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4765 
4766 	/* Set the implementation id4 */
4767 	*tl = txdr_unsigned(1);
4768 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4769 	(void) nfsm_strtom(nd, version, strlen(version));
4770 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4771 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4772 	verstime.tv_nsec = 0;
4773 	txdr_nfsv4time(&verstime, tl);
4774 	nd->nd_flag |= ND_USEGSSNAME;
4775 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4776 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4777 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4778 	    (int)nd->nd_repstat);
4779 	if (error != 0)
4780 		return (error);
4781 	if (nd->nd_repstat == 0) {
4782 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4783 		len = fxdr_unsigned(int, *(tl + 7));
4784 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4785 			error = NFSERR_BADXDR;
4786 			goto nfsmout;
4787 		}
4788 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4789 		    M_WAITOK | M_ZERO);
4790 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4791 		dsp->nfsclds_servownlen = len;
4792 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4793 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4794 		dsp->nfsclds_sess.nfsess_sequenceid =
4795 		    fxdr_unsigned(uint32_t, *tl++);
4796 		v41flags = fxdr_unsigned(uint32_t, *tl);
4797 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4798 		    NFSHASPNFSOPT(nmp)) {
4799 			NFSCL_DEBUG(1, "set PNFS\n");
4800 			NFSLOCKMNT(nmp);
4801 			nmp->nm_state |= NFSSTA_PNFS;
4802 			NFSUNLOCKMNT(nmp);
4803 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4804 		}
4805 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4806 			dsp->nfsclds_flags |= NFSCLDS_DS;
4807 		if (minorvers == NFSV42_MINORVERSION)
4808 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
4809 		if (len > 0)
4810 			nd->nd_repstat = nfsrv_mtostr(nd,
4811 			    dsp->nfsclds_serverown, len);
4812 		if (nd->nd_repstat == 0) {
4813 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4814 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4815 			    NULL, MTX_DEF);
4816 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4817 			*dspp = dsp;
4818 		} else
4819 			free(dsp, M_NFSCLDS);
4820 	}
4821 	error = nd->nd_repstat;
4822 nfsmout:
4823 	m_freem(nd->nd_mrep);
4824 	return (error);
4825 }
4826 
4827 /*
4828  * Do the NFSv4.1 Create Session.
4829  */
4830 int
4831 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4832     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
4833     struct ucred *cred, NFSPROC_T *p)
4834 {
4835 	uint32_t crflags, maxval, *tl;
4836 	struct nfsrv_descript nfsd;
4837 	struct nfsrv_descript *nd = &nfsd;
4838 	int error, irdcnt, minorvers;
4839 
4840 	/* Make sure nm_rsize, nm_wsize is set. */
4841 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4842 		nmp->nm_rsize = NFS_MAXBSIZE;
4843 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4844 		nmp->nm_wsize = NFS_MAXBSIZE;
4845 	if (dsp == NULL)
4846 		minorvers = nmp->nm_minorvers;
4847 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
4848 		minorvers = NFSV42_MINORVERSION;
4849 	else
4850 		minorvers = NFSV41_MINORVERSION;
4851 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
4852 	    NFS_VER4, minorvers);
4853 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4854 	*tl++ = sep->nfsess_clientid.lval[0];
4855 	*tl++ = sep->nfsess_clientid.lval[1];
4856 	*tl++ = txdr_unsigned(sequenceid);
4857 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4858 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4859 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4860 	*tl = txdr_unsigned(crflags);
4861 
4862 	/* Fill in fore channel attributes. */
4863 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4864 	*tl++ = 0;				/* Header pad size */
4865 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
4866 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
4867 		/*
4868 		 * NFSv4.2 Extended Attribute operations may want to do
4869 		 * requests/replies that are larger than nm_rsize/nm_wsize.
4870 		 */
4871 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4872 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4873 	} else {
4874 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
4875 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
4876 	}
4877 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4878 	*tl++ = txdr_unsigned(20);		/* Max operations */
4879 	*tl++ = txdr_unsigned(64);		/* Max slots */
4880 	*tl = 0;				/* No rdma ird */
4881 
4882 	/* Fill in back channel attributes. */
4883 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4884 	*tl++ = 0;				/* Header pad size */
4885 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4886 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4887 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4888 	*tl++ = txdr_unsigned(4);		/* Max operations */
4889 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4890 	*tl = 0;				/* No rdma ird */
4891 
4892 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4893 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4894 
4895 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4896 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4897 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4898 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4899 	*tl++ = 0;				/* Null machine name */
4900 	*tl++ = 0;				/* Uid == 0 */
4901 	*tl++ = 0;				/* Gid == 0 */
4902 	*tl = 0;				/* No additional gids */
4903 	nd->nd_flag |= ND_USEGSSNAME;
4904 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4905 	    NFS_VER4, NULL, 1, NULL, NULL);
4906 	if (error != 0)
4907 		return (error);
4908 	if (nd->nd_repstat == 0) {
4909 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4910 		    2 * NFSX_UNSIGNED);
4911 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4912 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4913 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4914 		crflags = fxdr_unsigned(uint32_t, *tl);
4915 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4916 			NFSLOCKMNT(nmp);
4917 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4918 			NFSUNLOCKMNT(nmp);
4919 		}
4920 
4921 		/* Get the fore channel slot count. */
4922 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4923 		tl++;			/* Skip the header pad size. */
4924 
4925 		/* Make sure nm_wsize is small enough. */
4926 		maxval = fxdr_unsigned(uint32_t, *tl++);
4927 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4928 			if (nmp->nm_wsize > 8096)
4929 				nmp->nm_wsize /= 2;
4930 			else
4931 				break;
4932 		}
4933 		sep->nfsess_maxreq = maxval;
4934 
4935 		/* Make sure nm_rsize is small enough. */
4936 		maxval = fxdr_unsigned(uint32_t, *tl++);
4937 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4938 			if (nmp->nm_rsize > 8096)
4939 				nmp->nm_rsize /= 2;
4940 			else
4941 				break;
4942 		}
4943 		sep->nfsess_maxresp = maxval;
4944 
4945 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4946 		tl++;
4947 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4948 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4949 		irdcnt = fxdr_unsigned(int, *tl);
4950 		if (irdcnt > 0)
4951 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4952 
4953 		/* and the back channel slot count. */
4954 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4955 		tl += 5;
4956 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4957 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4958 	}
4959 	error = nd->nd_repstat;
4960 nfsmout:
4961 	m_freem(nd->nd_mrep);
4962 	return (error);
4963 }
4964 
4965 /*
4966  * Do the NFSv4.1 Destroy Session.
4967  */
4968 int
4969 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4970     struct ucred *cred, NFSPROC_T *p)
4971 {
4972 	uint32_t *tl;
4973 	struct nfsrv_descript nfsd;
4974 	struct nfsrv_descript *nd = &nfsd;
4975 	int error;
4976 	struct nfsclsession *tsep;
4977 
4978 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
4979 	    0);
4980 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4981 	tsep = nfsmnt_mdssession(nmp);
4982 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4983 	nd->nd_flag |= ND_USEGSSNAME;
4984 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4985 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4986 	if (error != 0)
4987 		return (error);
4988 	error = nd->nd_repstat;
4989 	m_freem(nd->nd_mrep);
4990 	return (error);
4991 }
4992 
4993 /*
4994  * Do the NFSv4.1 Destroy Client.
4995  */
4996 int
4997 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
4998     struct ucred *cred, NFSPROC_T *p)
4999 {
5000 	uint32_t *tl;
5001 	struct nfsrv_descript nfsd;
5002 	struct nfsrv_descript *nd = &nfsd;
5003 	int error;
5004 	struct nfsclsession *tsep;
5005 
5006 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5007 	    0);
5008 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5009 	tsep = nfsmnt_mdssession(nmp);
5010 	*tl++ = tsep->nfsess_clientid.lval[0];
5011 	*tl = tsep->nfsess_clientid.lval[1];
5012 	nd->nd_flag |= ND_USEGSSNAME;
5013 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5014 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5015 	if (error != 0)
5016 		return (error);
5017 	error = nd->nd_repstat;
5018 	m_freem(nd->nd_mrep);
5019 	return (error);
5020 }
5021 
5022 /*
5023  * Do the NFSv4.1 LayoutGet.
5024  */
5025 static int
5026 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5027     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5028     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5029     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
5030     void *stuff)
5031 {
5032 	struct nfsrv_descript nfsd, *nd = &nfsd;
5033 	int error;
5034 
5035 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5036 	    0);
5037 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5038 	    layouttype, layoutlen, 0);
5039 	nd->nd_flag |= ND_USEGSSNAME;
5040 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5041 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5042 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5043 	if (error != 0)
5044 		return (error);
5045 	if (nd->nd_repstat == 0)
5046 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5047 		    flhp);
5048 	if (error == 0 && nd->nd_repstat != 0)
5049 		error = nd->nd_repstat;
5050 	m_freem(nd->nd_mrep);
5051 	return (error);
5052 }
5053 
5054 /*
5055  * Do the NFSv4.1 Get Device Info.
5056  */
5057 int
5058 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5059     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5060     NFSPROC_T *p)
5061 {
5062 	uint32_t cnt, *tl, vers, minorvers;
5063 	struct nfsrv_descript nfsd;
5064 	struct nfsrv_descript *nd = &nfsd;
5065 	struct sockaddr_in sin, ssin;
5066 	struct sockaddr_in6 sin6, ssin6;
5067 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5068 	struct nfscldevinfo *ndi;
5069 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5070 	int stripecnt;
5071 	uint8_t stripeindex;
5072 	sa_family_t af, safilled;
5073 
5074 	ssin.sin_port = 0;		/* To shut up compiler. */
5075 	ssin.sin_addr.s_addr = 0;	/* ditto */
5076 	*ndip = NULL;
5077 	ndi = NULL;
5078 	gotdspp = NULL;
5079 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5080 	    0);
5081 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5082 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5083 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5084 	*tl++ = txdr_unsigned(layouttype);
5085 	*tl++ = txdr_unsigned(100000);
5086 	if (notifybitsp != NULL && *notifybitsp != 0) {
5087 		*tl = txdr_unsigned(1);		/* One word of bits. */
5088 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5089 		*tl = txdr_unsigned(*notifybitsp);
5090 	} else
5091 		*tl = txdr_unsigned(0);
5092 	nd->nd_flag |= ND_USEGSSNAME;
5093 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5094 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5095 	if (error != 0)
5096 		return (error);
5097 	if (nd->nd_repstat == 0) {
5098 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5099 		if (layouttype != fxdr_unsigned(int, *tl))
5100 			printf("EEK! devinfo layout type not same!\n");
5101 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5102 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5103 			stripecnt = fxdr_unsigned(int, *tl);
5104 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5105 			if (stripecnt < 1 || stripecnt > 4096) {
5106 				printf("pNFS File layout devinfo stripecnt %d:"
5107 				    " out of range\n", stripecnt);
5108 				error = NFSERR_BADXDR;
5109 				goto nfsmout;
5110 			}
5111 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5112 			    NFSX_UNSIGNED);
5113 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5114 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5115 			if (addrcnt < 1 || addrcnt > 128) {
5116 				printf("NFS devinfo addrcnt %d: out of range\n",
5117 				    addrcnt);
5118 				error = NFSERR_BADXDR;
5119 				goto nfsmout;
5120 			}
5121 
5122 			/*
5123 			 * Now we know how many stripe indices and addresses, so
5124 			 * we can allocate the structure the correct size.
5125 			 */
5126 			i = (stripecnt * sizeof(uint8_t)) /
5127 			    sizeof(struct nfsclds *) + 1;
5128 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5129 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5130 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5131 			    M_ZERO);
5132 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5133 			    NFSX_V4DEVICEID);
5134 			ndi->nfsdi_refcnt = 0;
5135 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5136 			ndi->nfsdi_stripecnt = stripecnt;
5137 			ndi->nfsdi_addrcnt = addrcnt;
5138 			/* Fill in the stripe indices. */
5139 			for (i = 0; i < stripecnt; i++) {
5140 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5141 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5142 				if (stripeindex >= addrcnt) {
5143 					printf("pNFS File Layout devinfo"
5144 					    " stripeindex %d: too big\n",
5145 					    (int)stripeindex);
5146 					error = NFSERR_BADXDR;
5147 					goto nfsmout;
5148 				}
5149 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5150 			}
5151 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5152 			/* For Flex File, we only get one address list. */
5153 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5154 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5155 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5156 			    NFSX_V4DEVICEID);
5157 			ndi->nfsdi_refcnt = 0;
5158 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5159 			addrcnt = ndi->nfsdi_addrcnt = 1;
5160 		}
5161 
5162 		/* Now, dissect the server address(es). */
5163 		safilled = AF_UNSPEC;
5164 		for (i = 0; i < addrcnt; i++) {
5165 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5166 			cnt = fxdr_unsigned(uint32_t, *tl);
5167 			if (cnt == 0) {
5168 				printf("NFS devinfo 0 len addrlist\n");
5169 				error = NFSERR_BADXDR;
5170 				goto nfsmout;
5171 			}
5172 			dspp = nfsfldi_addr(ndi, i);
5173 			safilled = AF_UNSPEC;
5174 			for (j = 0; j < cnt; j++) {
5175 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5176 				    &isudp);
5177 				if (error != 0 && error != EPERM) {
5178 					error = NFSERR_BADXDR;
5179 					goto nfsmout;
5180 				}
5181 				if (error == 0 && isudp == 0) {
5182 					/*
5183 					 * The priority is:
5184 					 * - Same address family.
5185 					 * Save the address and dspp, so that
5186 					 * the connection can be done after
5187 					 * parsing is complete.
5188 					 */
5189 					if (safilled == AF_UNSPEC ||
5190 					    (af == nmp->nm_nam->sa_family &&
5191 					     safilled != nmp->nm_nam->sa_family)
5192 					   ) {
5193 						if (af == AF_INET)
5194 							ssin = sin;
5195 						else
5196 							ssin6 = sin6;
5197 						safilled = af;
5198 						gotdspp = dspp;
5199 					}
5200 				}
5201 			}
5202 		}
5203 
5204 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5205 		gotminor = NFSV41_MINORVERSION;
5206 		/* For Flex File, we will take one of the versions to use. */
5207 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5208 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5209 			j = fxdr_unsigned(int, *tl);
5210 			if (j < 1 || j > NFSDEV_MAXVERS) {
5211 				printf("pNFS: too many versions\n");
5212 				error = NFSERR_BADXDR;
5213 				goto nfsmout;
5214 			}
5215 			gotvers = 0;
5216 			gotminor = 0;
5217 			for (i = 0; i < j; i++) {
5218 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5219 				vers = fxdr_unsigned(uint32_t, *tl++);
5220 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5221 				if (vers == NFS_VER3)
5222 					minorvers = 0;
5223 				if ((vers == NFS_VER4 && ((minorvers ==
5224 				    NFSV41_MINORVERSION && gotminor == 0) ||
5225 				    minorvers == NFSV42_MINORVERSION)) ||
5226 				    (vers == NFS_VER3 && gotvers == 0)) {
5227 					gotvers = vers;
5228 					gotminor = minorvers;
5229 					/* We'll take this one. */
5230 					ndi->nfsdi_versindex = i;
5231 					ndi->nfsdi_vers = vers;
5232 					ndi->nfsdi_minorvers = minorvers;
5233 					ndi->nfsdi_rsize = fxdr_unsigned(
5234 					    uint32_t, *tl++);
5235 					ndi->nfsdi_wsize = fxdr_unsigned(
5236 					    uint32_t, *tl++);
5237 					if (*tl == newnfs_true)
5238 						ndi->nfsdi_flags |=
5239 						    NFSDI_TIGHTCOUPLED;
5240 					else
5241 						ndi->nfsdi_flags &=
5242 						    ~NFSDI_TIGHTCOUPLED;
5243 				}
5244 			}
5245 			if (gotvers == 0) {
5246 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5247 				error = NFSERR_BADXDR;
5248 				goto nfsmout;
5249 			}
5250 		}
5251 
5252 		/* And the notify bits. */
5253 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5254 		bitcnt = fxdr_unsigned(int, *tl);
5255 		if (bitcnt > 0) {
5256 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5257 			if (notifybitsp != NULL)
5258 				*notifybitsp =
5259 				    fxdr_unsigned(uint32_t, *tl);
5260 		}
5261 		if (safilled != AF_UNSPEC) {
5262 			KASSERT(ndi != NULL, ("ndi is NULL"));
5263 			*ndip = ndi;
5264 		} else
5265 			error = EPERM;
5266 		if (error == 0) {
5267 			/*
5268 			 * Now we can do a TCP connection for the correct
5269 			 * NFS version and IP address.
5270 			 */
5271 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5272 			    gotvers, gotminor, &dsp, p);
5273 		}
5274 		if (error == 0) {
5275 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5276 			*gotdspp = dsp;
5277 		}
5278 	}
5279 	if (nd->nd_repstat != 0 && error == 0)
5280 		error = nd->nd_repstat;
5281 nfsmout:
5282 	if (error != 0 && ndi != NULL)
5283 		nfscl_freedevinfo(ndi);
5284 	m_freem(nd->nd_mrep);
5285 	return (error);
5286 }
5287 
5288 /*
5289  * Do the NFSv4.1 LayoutCommit.
5290  */
5291 int
5292 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5293     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5294     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5295 {
5296 	uint32_t *tl;
5297 	struct nfsrv_descript nfsd, *nd = &nfsd;
5298 	int error;
5299 
5300 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5301 	    0, 0);
5302 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5303 	    NFSX_STATEID);
5304 	txdr_hyper(off, tl);
5305 	tl += 2;
5306 	txdr_hyper(len, tl);
5307 	tl += 2;
5308 	if (reclaim != 0)
5309 		*tl++ = newnfs_true;
5310 	else
5311 		*tl++ = newnfs_false;
5312 	*tl++ = txdr_unsigned(stateidp->seqid);
5313 	*tl++ = stateidp->other[0];
5314 	*tl++ = stateidp->other[1];
5315 	*tl++ = stateidp->other[2];
5316 	*tl++ = newnfs_true;
5317 	if (lastbyte < off)
5318 		lastbyte = off;
5319 	else if (lastbyte >= (off + len))
5320 		lastbyte = off + len - 1;
5321 	txdr_hyper(lastbyte, tl);
5322 	tl += 2;
5323 	*tl++ = newnfs_false;
5324 	*tl++ = txdr_unsigned(layouttype);
5325 	/* All supported layouts are 0 length. */
5326 	*tl = txdr_unsigned(0);
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 LayoutReturn.
5339  */
5340 int
5341 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5342     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5343     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5344     uint32_t stat, uint32_t op, char *devid)
5345 {
5346 	uint32_t *tl;
5347 	struct nfsrv_descript nfsd, *nd = &nfsd;
5348 	uint64_t tu64;
5349 	int error;
5350 
5351 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5352 	    0, 0);
5353 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5354 	if (reclaim != 0)
5355 		*tl++ = newnfs_true;
5356 	else
5357 		*tl++ = newnfs_false;
5358 	*tl++ = txdr_unsigned(layouttype);
5359 	*tl++ = txdr_unsigned(iomode);
5360 	*tl = txdr_unsigned(layoutreturn);
5361 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5362 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5363 		    NFSX_UNSIGNED);
5364 		txdr_hyper(offset, tl);
5365 		tl += 2;
5366 		txdr_hyper(len, tl);
5367 		tl += 2;
5368 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5369 		*tl++ = txdr_unsigned(stateidp->seqid);
5370 		*tl++ = stateidp->other[0];
5371 		*tl++ = stateidp->other[1];
5372 		*tl++ = stateidp->other[2];
5373 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5374 			*tl = txdr_unsigned(0);
5375 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5376 			if (stat != 0) {
5377 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5378 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5379 				    NFSX_UNSIGNED);
5380 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5381 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5382 				    NFSX_UNSIGNED);
5383 				*tl++ = txdr_unsigned(1);	/* One error. */
5384 				tu64 = 0;			/* Offset. */
5385 				txdr_hyper(tu64, tl); tl += 2;
5386 				tu64 = UINT64_MAX;		/* Length. */
5387 				txdr_hyper(tu64, tl); tl += 2;
5388 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5389 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5390 				*tl++ = txdr_unsigned(1);	/* One error. */
5391 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5392 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5393 				*tl++ = txdr_unsigned(stat);
5394 				*tl++ = txdr_unsigned(op);
5395 			} else {
5396 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5397 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5398 				/* No ioerrs. */
5399 				*tl++ = 0;
5400 			}
5401 			*tl = 0;	/* No stats yet. */
5402 		}
5403 	}
5404 	nd->nd_flag |= ND_USEGSSNAME;
5405 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5406 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5407 	if (error != 0)
5408 		return (error);
5409 	if (nd->nd_repstat == 0) {
5410 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5411 		if (*tl != 0) {
5412 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5413 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5414 			stateidp->other[0] = *tl++;
5415 			stateidp->other[1] = *tl++;
5416 			stateidp->other[2] = *tl;
5417 		}
5418 	} else
5419 		error = nd->nd_repstat;
5420 nfsmout:
5421 	m_freem(nd->nd_mrep);
5422 	return (error);
5423 }
5424 
5425 /*
5426  * Acquire a layout and devinfo, if possible. The caller must have acquired
5427  * a reference count on the nfsclclient structure before calling this.
5428  * Return the layout in lypp with a reference count on it, if successful.
5429  */
5430 static int
5431 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5432     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5433     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5434 {
5435 	struct nfscllayout *lyp;
5436 	struct nfsclflayout *flp;
5437 	struct nfsclflayouthead flh;
5438 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5439 	nfsv4stateid_t stateid;
5440 	struct nfsclsession *tsep;
5441 
5442 	*lypp = NULL;
5443 	if (NFSHASFLEXFILE(nmp))
5444 		layouttype = NFSLAYOUT_FLEXFILE;
5445 	else
5446 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5447 	/*
5448 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5449 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5450 	 * flp == NULL.
5451 	 */
5452 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5453 	    off, &flp, &recalled);
5454 	islocked = 0;
5455 	if (lyp == NULL || flp == NULL) {
5456 		if (recalled != 0)
5457 			return (EIO);
5458 		LIST_INIT(&flh);
5459 		tsep = nfsmnt_mdssession(nmp);
5460 		layoutlen = tsep->nfsess_maxcache -
5461 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5462 		if (lyp == NULL) {
5463 			stateid.seqid = 0;
5464 			stateid.other[0] = stateidp->other[0];
5465 			stateid.other[1] = stateidp->other[1];
5466 			stateid.other[2] = stateidp->other[2];
5467 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5468 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5469 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5470 			    &retonclose, &flh, cred, p, NULL);
5471 		} else {
5472 			islocked = 1;
5473 			stateid.seqid = lyp->nfsly_stateid.seqid;
5474 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5475 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5476 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5477 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5478 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5479 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5480 			    &retonclose, &flh, cred, p, NULL);
5481 		}
5482 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5483 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5484 		    &flh, layouttype, error, NULL, cred, p);
5485 		if (error == 0)
5486 			*lypp = lyp;
5487 		else if (islocked != 0)
5488 			nfscl_rellayout(lyp, 1);
5489 	} else
5490 		*lypp = lyp;
5491 	return (error);
5492 }
5493 
5494 /*
5495  * Do a TCP connection plus exchange id and create session.
5496  * If successful, a "struct nfsclds" is linked into the list for the
5497  * mount point and a pointer to it is returned.
5498  */
5499 static int
5500 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5501     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5502     struct nfsclds **dspp, NFSPROC_T *p)
5503 {
5504 	struct sockaddr_in *msad, *sad;
5505 	struct sockaddr_in6 *msad6, *sad6;
5506 	struct nfsclclient *clp;
5507 	struct nfssockreq *nrp;
5508 	struct nfsclds *dsp, *tdsp;
5509 	int error, firsttry;
5510 	enum nfsclds_state retv;
5511 	uint32_t sequenceid = 0;
5512 
5513 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5514 	    ("nfsrpc_fillsa: NULL nr_cred"));
5515 	NFSLOCKCLSTATE();
5516 	clp = nmp->nm_clp;
5517 	NFSUNLOCKCLSTATE();
5518 	if (clp == NULL)
5519 		return (EPERM);
5520 	if (af == AF_INET) {
5521 		NFSLOCKMNT(nmp);
5522 		/*
5523 		 * Check to see if we already have a session for this
5524 		 * address that is usable for a DS.
5525 		 * Note that the MDS's address is in a different place
5526 		 * than the sessions already acquired for DS's.
5527 		 */
5528 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5529 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5530 		while (tdsp != NULL) {
5531 			if (msad != NULL && msad->sin_family == AF_INET &&
5532 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5533 			    sin->sin_port == msad->sin_port &&
5534 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5535 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5536 				*dspp = tdsp;
5537 				NFSUNLOCKMNT(nmp);
5538 				NFSCL_DEBUG(4, "fnd same addr\n");
5539 				return (0);
5540 			}
5541 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5542 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5543 				msad = (struct sockaddr_in *)
5544 				    tdsp->nfsclds_sockp->nr_nam;
5545 			else
5546 				msad = NULL;
5547 		}
5548 		NFSUNLOCKMNT(nmp);
5549 
5550 		/* No IP address match, so look for new/trunked one. */
5551 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5552 		sad->sin_len = sizeof(*sad);
5553 		sad->sin_family = AF_INET;
5554 		sad->sin_port = sin->sin_port;
5555 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5556 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5557 		nrp->nr_nam = (struct sockaddr *)sad;
5558 	} else if (af == AF_INET6) {
5559 		NFSLOCKMNT(nmp);
5560 		/*
5561 		 * Check to see if we already have a session for this
5562 		 * address that is usable for a DS.
5563 		 * Note that the MDS's address is in a different place
5564 		 * than the sessions already acquired for DS's.
5565 		 */
5566 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5567 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5568 		while (tdsp != NULL) {
5569 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5570 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5571 			    &msad6->sin6_addr) &&
5572 			    sin6->sin6_port == msad6->sin6_port &&
5573 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5574 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5575 				*dspp = tdsp;
5576 				NFSUNLOCKMNT(nmp);
5577 				return (0);
5578 			}
5579 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5580 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5581 				msad6 = (struct sockaddr_in6 *)
5582 				    tdsp->nfsclds_sockp->nr_nam;
5583 			else
5584 				msad6 = NULL;
5585 		}
5586 		NFSUNLOCKMNT(nmp);
5587 
5588 		/* No IP address match, so look for new/trunked one. */
5589 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5590 		sad6->sin6_len = sizeof(*sad6);
5591 		sad6->sin6_family = AF_INET6;
5592 		sad6->sin6_port = sin6->sin6_port;
5593 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5594 		    sizeof(struct in6_addr));
5595 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5596 		nrp->nr_nam = (struct sockaddr *)sad6;
5597 	} else
5598 		return (EPERM);
5599 
5600 	nrp->nr_sotype = SOCK_STREAM;
5601 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5602 	nrp->nr_prog = NFS_PROG;
5603 	nrp->nr_vers = vers;
5604 
5605 	/*
5606 	 * Use the credentials that were used for the mount, which are
5607 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5608 	 * Ref. counting the credentials with crhold() is probably not
5609 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5610 	 * unmount, but I did it anyhow.
5611 	 */
5612 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5613 	error = newnfs_connect(nmp, nrp, NULL, p, 0, false);
5614 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5615 
5616 	dsp = NULL;
5617 	/* Now, do the exchangeid and create session. */
5618 	if (error == 0) {
5619 		if (vers == NFS_VER4) {
5620 			firsttry = 0;
5621 			do {
5622 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5623 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5624 				    nrp->nr_cred, p);
5625 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5626 				if (error == NFSERR_MINORVERMISMATCH)
5627 					minorvers = NFSV42_MINORVERSION;
5628 			} while (error == NFSERR_MINORVERMISMATCH &&
5629 			    firsttry++ == 0);
5630 			if (error != 0)
5631 				newnfs_disconnect(nrp);
5632 		} else {
5633 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5634 			    M_WAITOK | M_ZERO);
5635 			dsp->nfsclds_flags |= NFSCLDS_DS;
5636 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5637 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5638 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5639 			    NULL, MTX_DEF);
5640 		}
5641 	}
5642 	if (error == 0) {
5643 		dsp->nfsclds_sockp = nrp;
5644 		if (vers == NFS_VER4) {
5645 			NFSLOCKMNT(nmp);
5646 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5647 			    &sequenceid);
5648 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5649 			if (retv == NFSDSP_USETHISSESSION &&
5650 			    nfscl_dssameconn != 0) {
5651 				NFSLOCKDS(tdsp);
5652 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5653 				NFSUNLOCKDS(tdsp);
5654 				NFSUNLOCKMNT(nmp);
5655 				/*
5656 				 * If there is already a session for this
5657 				 * server, use it.
5658 				 */
5659 				(void)newnfs_disconnect(nrp);
5660 				nfscl_freenfsclds(dsp);
5661 				*dspp = tdsp;
5662 				return (0);
5663 			}
5664 			if (retv == NFSDSP_NOTFOUND)
5665 				sequenceid =
5666 				    dsp->nfsclds_sess.nfsess_sequenceid;
5667 			NFSUNLOCKMNT(nmp);
5668 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5669 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
5670 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5671 		}
5672 	} else {
5673 		NFSFREECRED(nrp->nr_cred);
5674 		NFSFREEMUTEX(&nrp->nr_mtx);
5675 		free(nrp->nr_nam, M_SONAME);
5676 		free(nrp, M_NFSSOCKREQ);
5677 	}
5678 	if (error == 0) {
5679 		NFSCL_DEBUG(3, "add DS session\n");
5680 		/*
5681 		 * Put it at the end of the list. That way the list
5682 		 * is ordered by when the entry was added. This matters
5683 		 * since the one done first is the one that should be
5684 		 * used for sequencid'ing any subsequent create sessions.
5685 		 */
5686 		NFSLOCKMNT(nmp);
5687 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5688 		NFSUNLOCKMNT(nmp);
5689 		*dspp = dsp;
5690 	} else if (dsp != NULL) {
5691 		newnfs_disconnect(nrp);
5692 		nfscl_freenfsclds(dsp);
5693 	}
5694 	return (error);
5695 }
5696 
5697 /*
5698  * Do the NFSv4.1 Reclaim Complete.
5699  */
5700 int
5701 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5702 {
5703 	uint32_t *tl;
5704 	struct nfsrv_descript nfsd;
5705 	struct nfsrv_descript *nd = &nfsd;
5706 	int error;
5707 
5708 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5709 	    0);
5710 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5711 	*tl = newnfs_false;
5712 	nd->nd_flag |= ND_USEGSSNAME;
5713 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5714 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5715 	if (error != 0)
5716 		return (error);
5717 	error = nd->nd_repstat;
5718 	m_freem(nd->nd_mrep);
5719 	return (error);
5720 }
5721 
5722 /*
5723  * Initialize the slot tables for a session.
5724  */
5725 static void
5726 nfscl_initsessionslots(struct nfsclsession *sep)
5727 {
5728 	int i;
5729 
5730 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5731 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5732 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5733 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5734 	}
5735 	for (i = 0; i < 64; i++)
5736 		sep->nfsess_slotseq[i] = 0;
5737 	sep->nfsess_slots = 0;
5738 }
5739 
5740 /*
5741  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5742  */
5743 int
5744 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5745     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5746 {
5747 	struct nfsnode *np = VTONFS(vp);
5748 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
5749 	struct nfscllayout *layp;
5750 	struct nfscldevinfo *dip;
5751 	struct nfsclflayout *rflp;
5752 	struct mbuf *m, *m2;
5753 	struct nfsclwritedsdorpc *drpc, *tdrpc;
5754 	nfsv4stateid_t stateid;
5755 	struct ucred *newcred;
5756 	uint64_t lastbyte, len, off, oresid, xfer;
5757 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5758 	void *lckp;
5759 	uint8_t *dev;
5760 	void *iovbase = NULL;
5761 	size_t iovlen = 0;
5762 	off_t offs = 0;
5763 	ssize_t resid = 0;
5764 
5765 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5766 	    (np->n_flag & NNOLAYOUT) != 0)
5767 		return (EIO);
5768 	/* Now, get a reference cnt on the clientid for this mount. */
5769 	if (nfscl_getref(nmp) == 0)
5770 		return (EIO);
5771 
5772 	/* Find an appropriate stateid. */
5773 	newcred = NFSNEWCRED(cred);
5774 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5775 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5776 	if (error != 0) {
5777 		NFSFREECRED(newcred);
5778 		nfscl_relref(nmp);
5779 		return (error);
5780 	}
5781 	/* Search for a layout for this file. */
5782 	off = uiop->uio_offset;
5783 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5784 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5785 	if (layp == NULL || rflp == NULL) {
5786 		if (recalled != 0) {
5787 			NFSFREECRED(newcred);
5788 			nfscl_relref(nmp);
5789 			return (EIO);
5790 		}
5791 		if (layp != NULL) {
5792 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5793 			layp = NULL;
5794 		}
5795 		/* Try and get a Layout, if it is supported. */
5796 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5797 		    (np->n_flag & NWRITEOPENED) != 0)
5798 			iolaymode = NFSLAYOUTIOMODE_RW;
5799 		else
5800 			iolaymode = NFSLAYOUTIOMODE_READ;
5801 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5802 		    NULL, &stateid, off, &layp, newcred, p);
5803 		if (error != 0) {
5804 			NFSLOCKNODE(np);
5805 			np->n_flag |= NNOLAYOUT;
5806 			NFSUNLOCKNODE(np);
5807 			if (lckp != NULL)
5808 				nfscl_lockderef(lckp);
5809 			NFSFREECRED(newcred);
5810 			if (layp != NULL)
5811 				nfscl_rellayout(layp, 0);
5812 			nfscl_relref(nmp);
5813 			return (error);
5814 		}
5815 	}
5816 
5817 	/*
5818 	 * Loop around finding a layout that works for the first part of
5819 	 * this I/O operation, and then call the function that actually
5820 	 * does the RPC.
5821 	 */
5822 	eof = 0;
5823 	len = (uint64_t)uiop->uio_resid;
5824 	while (len > 0 && error == 0 && eof == 0) {
5825 		off = uiop->uio_offset;
5826 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5827 		if (error == 0) {
5828 			oresid = xfer = (uint64_t)uiop->uio_resid;
5829 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5830 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5831 			/*
5832 			 * For Flex File layout with mirrored DSs, select one
5833 			 * of them at random for reads. For writes and commits,
5834 			 * do all mirrors.
5835 			 */
5836 			m = NULL;
5837 			tdrpc = drpc = NULL;
5838 			firstmirror = 0;
5839 			mirrorcnt = 1;
5840 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5841 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5842 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5843 					firstmirror = arc4random() % mirrorcnt;
5844 					mirrorcnt = firstmirror + 1;
5845 				} else {
5846 					if (docommit == 0) {
5847 						/*
5848 						 * Save values, so uiop can be
5849 						 * rolled back upon a write
5850 						 * error.
5851 						 */
5852 						offs = uiop->uio_offset;
5853 						resid = uiop->uio_resid;
5854 						iovbase =
5855 						    uiop->uio_iov->iov_base;
5856 						iovlen = uiop->uio_iov->iov_len;
5857 						m = nfsm_uiombuflist(uiop, len,
5858 						    0);
5859 					}
5860 					tdrpc = drpc = malloc(sizeof(*drpc) *
5861 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
5862 					    M_ZERO);
5863 				}
5864 			}
5865 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5866 				m2 = NULL;
5867 				if (m != NULL && i < mirrorcnt - 1)
5868 					m2 = m_copym(m, 0, M_COPYALL, M_WAITOK);
5869 				else {
5870 					m2 = m;
5871 					m = NULL;
5872 				}
5873 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
5874 					dev = rflp->nfsfl_ffm[i].dev;
5875 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5876 					    rflp->nfsfl_ffm[i].devp);
5877 				} else {
5878 					dev = rflp->nfsfl_dev;
5879 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5880 					    rflp->nfsfl_devp);
5881 				}
5882 				if (dip != NULL) {
5883 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5884 					    != 0)
5885 						error = nfscl_dofflayoutio(vp,
5886 						    uiop, iomode, must_commit,
5887 						    &eof, &stateid, rwaccess,
5888 						    dip, layp, rflp, off, xfer,
5889 						    i, docommit, m2, tdrpc,
5890 						    newcred, p);
5891 					else
5892 						error = nfscl_doflayoutio(vp,
5893 						    uiop, iomode, must_commit,
5894 						    &eof, &stateid, rwaccess,
5895 						    dip, layp, rflp, off, xfer,
5896 						    docommit, newcred, p);
5897 					nfscl_reldevinfo(dip);
5898 				} else {
5899 					if (m2 != NULL)
5900 						m_freem(m2);
5901 					error = EIO;
5902 				}
5903 				tdrpc++;
5904 			}
5905 			if (m != NULL)
5906 				m_freem(m);
5907 			tdrpc = drpc;
5908 			timo = hz / 50;		/* Wait for 20msec. */
5909 			if (timo < 1)
5910 				timo = 1;
5911 			for (i = firstmirror; i < mirrorcnt - 1 &&
5912 			    tdrpc != NULL; i++, tdrpc++) {
5913 				/*
5914 				 * For the unused drpc entries, both inprog and
5915 				 * err == 0, so this loop won't break.
5916 				 */
5917 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
5918 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5919 					    timo);
5920 				if (error == 0 && tdrpc->err != 0)
5921 					error = tdrpc->err;
5922 			}
5923 			free(drpc, M_TEMP);
5924 			if (error == 0) {
5925 				if (mirrorcnt > 1 && rwaccess ==
5926 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5927 					NFSLOCKCLSTATE();
5928 					layp->nfsly_flags |= NFSLY_WRITTEN;
5929 					NFSUNLOCKCLSTATE();
5930 				}
5931 				lastbyte = off + xfer - 1;
5932 				NFSLOCKCLSTATE();
5933 				if (lastbyte > layp->nfsly_lastbyte)
5934 					layp->nfsly_lastbyte = lastbyte;
5935 				NFSUNLOCKCLSTATE();
5936 			} else if (error == NFSERR_OPENMODE &&
5937 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
5938 				NFSLOCKMNT(nmp);
5939 				nmp->nm_state |= NFSSTA_OPENMODE;
5940 				NFSUNLOCKMNT(nmp);
5941 			} else
5942 				error = EIO;
5943 			if (error == 0)
5944 				len -= (oresid - (uint64_t)uiop->uio_resid);
5945 			else if (mirrorcnt > 1 && rwaccess ==
5946 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5947 				/*
5948 				 * In case the rpc gets retried, roll the
5949 				 * uio fields changed by nfsm_uiombuflist()
5950 				 * back.
5951 				 */
5952 				uiop->uio_offset = offs;
5953 				uiop->uio_resid = resid;
5954 				uiop->uio_iov->iov_base = iovbase;
5955 				uiop->uio_iov->iov_len = iovlen;
5956 			}
5957 		}
5958 	}
5959 	if (lckp != NULL)
5960 		nfscl_lockderef(lckp);
5961 	NFSFREECRED(newcred);
5962 	nfscl_rellayout(layp, 0);
5963 	nfscl_relref(nmp);
5964 	return (error);
5965 }
5966 
5967 /*
5968  * Find a file layout that will handle the first bytes of the requested
5969  * range and return the information from it needed to the I/O operation.
5970  */
5971 int
5972 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5973     struct nfsclflayout **retflpp)
5974 {
5975 	struct nfsclflayout *flp, *nflp, *rflp;
5976 	uint32_t rw;
5977 
5978 	rflp = NULL;
5979 	rw = rwaccess;
5980 	/* For reading, do the Read list first and then the Write list. */
5981 	do {
5982 		if (rw == NFSV4OPEN_ACCESSREAD)
5983 			flp = LIST_FIRST(&lyp->nfsly_flayread);
5984 		else
5985 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
5986 		while (flp != NULL) {
5987 			nflp = LIST_NEXT(flp, nfsfl_list);
5988 			if (flp->nfsfl_off > off)
5989 				break;
5990 			if (flp->nfsfl_end > off &&
5991 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
5992 				rflp = flp;
5993 			flp = nflp;
5994 		}
5995 		if (rw == NFSV4OPEN_ACCESSREAD)
5996 			rw = NFSV4OPEN_ACCESSWRITE;
5997 		else
5998 			rw = 0;
5999 	} while (rw != 0);
6000 	if (rflp != NULL) {
6001 		/* This one covers the most bytes starting at off. */
6002 		*retflpp = rflp;
6003 		return (0);
6004 	}
6005 	return (EIO);
6006 }
6007 
6008 /*
6009  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6010  */
6011 static int
6012 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6013     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6014     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6015     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6016 {
6017 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6018 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6019 	struct nfsnode *np;
6020 	struct nfsfh *fhp;
6021 	struct nfsclds **dspp;
6022 
6023 	np = VTONFS(vp);
6024 	rel_off = off - flp->nfsfl_patoff;
6025 	stripe_unit_size = flp->nfsfl_util & NFSFLAYUTIL_STRIPE_MASK;
6026 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6027 	    dp->nfsdi_stripecnt;
6028 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6029 	error = 0;
6030 
6031 	/* Loop around, doing I/O for each stripe unit. */
6032 	while (len > 0 && error == 0) {
6033 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6034 		dspp = nfsfldi_addr(dp, stripe_index);
6035 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6036 			minorvers = NFSV42_MINORVERSION;
6037 		else
6038 			minorvers = NFSV41_MINORVERSION;
6039 		if (len > transfer && docommit == 0)
6040 			xfer = transfer;
6041 		else
6042 			xfer = len;
6043 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6044 			/* Dense layout. */
6045 			if (stripe_pos >= flp->nfsfl_fhcnt)
6046 				return (EIO);
6047 			fhp = flp->nfsfl_fh[stripe_pos];
6048 			io_off = (rel_off / (stripe_unit_size *
6049 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6050 			    rel_off % stripe_unit_size;
6051 		} else {
6052 			/* Sparse layout. */
6053 			if (flp->nfsfl_fhcnt > 1) {
6054 				if (stripe_index >= flp->nfsfl_fhcnt)
6055 					return (EIO);
6056 				fhp = flp->nfsfl_fh[stripe_index];
6057 			} else if (flp->nfsfl_fhcnt == 1)
6058 				fhp = flp->nfsfl_fh[0];
6059 			else
6060 				fhp = np->n_fhp;
6061 			io_off = off;
6062 		}
6063 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6064 			commit_thru_mds = 1;
6065 			if (docommit != 0)
6066 				error = EIO;
6067 		} else {
6068 			commit_thru_mds = 0;
6069 			NFSLOCKNODE(np);
6070 			np->n_flag |= NDSCOMMIT;
6071 			NFSUNLOCKNODE(np);
6072 		}
6073 		if (docommit != 0) {
6074 			if (error == 0)
6075 				error = nfsrpc_commitds(vp, io_off, xfer,
6076 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6077 			if (error == 0) {
6078 				/*
6079 				 * Set both eof and uio_resid = 0 to end any
6080 				 * loops.
6081 				 */
6082 				*eofp = 1;
6083 				uiop->uio_resid = 0;
6084 			} else {
6085 				NFSLOCKNODE(np);
6086 				np->n_flag &= ~NDSCOMMIT;
6087 				NFSUNLOCKNODE(np);
6088 			}
6089 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6090 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6091 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6092 		else {
6093 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6094 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6095 			    0, NFS_VER4, minorvers, cred, p);
6096 			if (error == 0) {
6097 				NFSLOCKCLSTATE();
6098 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6099 				NFSUNLOCKCLSTATE();
6100 			}
6101 		}
6102 		if (error == 0) {
6103 			transfer = stripe_unit_size;
6104 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6105 			len -= xfer;
6106 			off += xfer;
6107 		}
6108 	}
6109 	return (error);
6110 }
6111 
6112 /*
6113  * Do I/O using an NFSv4.1 flex file layout.
6114  */
6115 static int
6116 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6117     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6118     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6119     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6120     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6121 {
6122 	uint64_t xfer;
6123 	int error;
6124 	struct nfsnode *np;
6125 	struct nfsfh *fhp;
6126 	struct nfsclds **dspp;
6127 	struct ucred *tcred;
6128 	struct mbuf *m, *m2;
6129 	uint32_t copylen;
6130 
6131 	np = VTONFS(vp);
6132 	error = 0;
6133 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6134 	    (uintmax_t)len);
6135 	/* Loop around, doing I/O for each stripe unit. */
6136 	while (len > 0 && error == 0) {
6137 		dspp = nfsfldi_addr(dp, 0);
6138 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6139 		stateidp = &flp->nfsfl_ffm[mirror].st;
6140 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6141 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6142 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6143 			tcred = NFSNEWCRED(cred);
6144 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6145 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6146 			tcred->cr_ngroups = 1;
6147 		} else
6148 			tcred = cred;
6149 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6150 			copylen = dp->nfsdi_rsize;
6151 		else {
6152 			copylen = dp->nfsdi_wsize;
6153 			if (len > copylen && mp != NULL) {
6154 				/*
6155 				 * When a mirrored configuration needs to do
6156 				 * multiple writes to each mirror, all writes
6157 				 * except the last one must be a multiple of
6158 				 * 4 bytes.  This is required so that the XDR
6159 				 * does not need padding.
6160 				 * If possible, clip the size to an exact
6161 				 * multiple of the mbuf length, so that the
6162 				 * split will be on an mbuf boundary.
6163 				 */
6164 				copylen &= 0xfffffffc;
6165 				if (copylen > mp->m_len)
6166 					copylen = copylen / mp->m_len *
6167 					    mp->m_len;
6168 			}
6169 		}
6170 		NFSLOCKNODE(np);
6171 		np->n_flag |= NDSCOMMIT;
6172 		NFSUNLOCKNODE(np);
6173 		if (len > copylen && docommit == 0)
6174 			xfer = copylen;
6175 		else
6176 			xfer = len;
6177 		if (docommit != 0) {
6178 			if (error == 0) {
6179 				/*
6180 				 * Do last mirrored DS commit with this thread.
6181 				 */
6182 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6183 					error = nfsio_commitds(vp, off, xfer,
6184 					    *dspp, fhp, dp->nfsdi_vers,
6185 					    dp->nfsdi_minorvers, drpc, tcred,
6186 					    p);
6187 				else
6188 					error = nfsrpc_commitds(vp, off, xfer,
6189 					    *dspp, fhp, dp->nfsdi_vers,
6190 					    dp->nfsdi_minorvers, tcred, p);
6191 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6192 				if (error != 0 && error != EACCES && error !=
6193 				    ESTALE) {
6194 					NFSCL_DEBUG(4,
6195 					    "DS layreterr for commit\n");
6196 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6197 					    lyp, *dspp);
6198 				}
6199 			}
6200 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6201 			if (error == 0) {
6202 				/*
6203 				 * Set both eof and uio_resid = 0 to end any
6204 				 * loops.
6205 				 */
6206 				*eofp = 1;
6207 				uiop->uio_resid = 0;
6208 			} else {
6209 				NFSLOCKNODE(np);
6210 				np->n_flag &= ~NDSCOMMIT;
6211 				NFSUNLOCKNODE(np);
6212 			}
6213 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6214 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6215 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6216 			    dp->nfsdi_minorvers, tcred, p);
6217 			NFSCL_DEBUG(4, "readds=%d\n", error);
6218 			if (error != 0 && error != EACCES && error != ESTALE) {
6219 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6220 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6221 				    *dspp);
6222 			}
6223 		} else {
6224 			if (flp->nfsfl_mirrorcnt == 1) {
6225 				error = nfsrpc_writeds(vp, uiop, iomode,
6226 				    must_commit, stateidp, *dspp, off, xfer,
6227 				    fhp, 0, 1, dp->nfsdi_vers,
6228 				    dp->nfsdi_minorvers, tcred, p);
6229 				if (error == 0) {
6230 					NFSLOCKCLSTATE();
6231 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6232 					NFSUNLOCKCLSTATE();
6233 				}
6234 			} else {
6235 				m = mp;
6236 				if (xfer < len) {
6237 					/* The mbuf list must be split. */
6238 					m2 = nfsm_split(mp, xfer);
6239 					if (m2 != NULL)
6240 						mp = m2;
6241 					else {
6242 						m_freem(mp);
6243 						error = EIO;
6244 					}
6245 				}
6246 				NFSCL_DEBUG(4, "mcopy len=%jd xfer=%jd\n",
6247 				    (uintmax_t)len, (uintmax_t)xfer);
6248 				/*
6249 				 * Do last write to a mirrored DS with this
6250 				 * thread.
6251 				 */
6252 				if (error == 0) {
6253 					if (mirror < flp->nfsfl_mirrorcnt - 1)
6254 						error = nfsio_writedsmir(vp,
6255 						    iomode, must_commit,
6256 						    stateidp, *dspp, off,
6257 						    xfer, fhp, m,
6258 						    dp->nfsdi_vers,
6259 						    dp->nfsdi_minorvers, drpc,
6260 						    tcred, p);
6261 					else
6262 						error = nfsrpc_writedsmir(vp,
6263 						    iomode, must_commit,
6264 						    stateidp, *dspp, off,
6265 						    xfer, fhp, m,
6266 						    dp->nfsdi_vers,
6267 						    dp->nfsdi_minorvers, tcred,
6268 						    p);
6269 				}
6270 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6271 				if (error != 0 && error != EACCES && error !=
6272 				    ESTALE) {
6273 					NFSCL_DEBUG(4,
6274 					    "DS layreterr for write\n");
6275 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6276 					    lyp, *dspp);
6277 				}
6278 			}
6279 		}
6280 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6281 		if (error == 0) {
6282 			len -= xfer;
6283 			off += xfer;
6284 		}
6285 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6286 			NFSFREECRED(tcred);
6287 	}
6288 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6289 	return (error);
6290 }
6291 
6292 /*
6293  * The actual read RPC done to a DS.
6294  */
6295 static int
6296 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6297     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6298     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6299 {
6300 	uint32_t *tl;
6301 	int attrflag, error, retlen;
6302 	struct nfsrv_descript nfsd;
6303 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6304 	struct nfsrv_descript *nd = &nfsd;
6305 	struct nfssockreq *nrp;
6306 	struct nfsvattr na;
6307 
6308 	nd->nd_mrep = NULL;
6309 	if (vers == 0 || vers == NFS_VER4) {
6310 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6311 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6312 		vers = NFS_VER4;
6313 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6314 		if (flex != 0)
6315 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6316 		else
6317 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6318 	} else {
6319 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6320 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6321 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READ]);
6322 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_READDS]);
6323 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6324 	}
6325 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6326 	txdr_hyper(io_off, tl);
6327 	*(tl + 2) = txdr_unsigned(len);
6328 	nrp = dsp->nfsclds_sockp;
6329 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6330 	if (nrp == NULL)
6331 		/* If NULL, use the MDS socket. */
6332 		nrp = &nmp->nm_sockreq;
6333 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6334 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6335 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6336 	    error);
6337 	if (error != 0)
6338 		return (error);
6339 	if (vers == NFS_VER3) {
6340 		error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6341 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6342 		if (error != 0)
6343 			goto nfsmout;
6344 	}
6345 	if (nd->nd_repstat != 0) {
6346 		error = nd->nd_repstat;
6347 		goto nfsmout;
6348 	}
6349 	if (vers == NFS_VER3) {
6350 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6351 		*eofp = fxdr_unsigned(int, *(tl + 1));
6352 	} else {
6353 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6354 		*eofp = fxdr_unsigned(int, *tl);
6355 	}
6356 	NFSM_STRSIZ(retlen, len);
6357 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6358 	error = nfsm_mbufuio(nd, uiop, retlen);
6359 nfsmout:
6360 	if (nd->nd_mrep != NULL)
6361 		m_freem(nd->nd_mrep);
6362 	return (error);
6363 }
6364 
6365 /*
6366  * The actual write RPC done to a DS.
6367  */
6368 static int
6369 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6370     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6371     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6372     struct ucred *cred, NFSPROC_T *p)
6373 {
6374 	uint32_t *tl;
6375 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6376 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6377 	int32_t backup;
6378 	struct nfsrv_descript nfsd;
6379 	struct nfsrv_descript *nd = &nfsd;
6380 	struct nfssockreq *nrp;
6381 	struct nfsvattr na;
6382 
6383 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6384 	nd->nd_mrep = NULL;
6385 	if (vers == 0 || vers == NFS_VER4) {
6386 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6387 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6388 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6389 		vers = NFS_VER4;
6390 		if (flex != 0)
6391 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6392 		else
6393 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6394 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6395 	} else {
6396 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6397 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6398 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6399 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6400 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6401 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6402 	}
6403 	txdr_hyper(io_off, tl);
6404 	tl += 2;
6405 	if (vers == NFS_VER3)
6406 		*tl++ = txdr_unsigned(len);
6407 	*tl++ = txdr_unsigned(*iomode);
6408 	*tl = txdr_unsigned(len);
6409 	nfsm_uiombuf(nd, uiop, len);
6410 	nrp = dsp->nfsclds_sockp;
6411 	if (nrp == NULL)
6412 		/* If NULL, use the MDS socket. */
6413 		nrp = &nmp->nm_sockreq;
6414 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6415 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6416 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6417 	    nd->nd_repstat);
6418 	if (error != 0)
6419 		return (error);
6420 	if (nd->nd_repstat != 0) {
6421 		/*
6422 		 * In case the rpc gets retried, roll
6423 		 * the uio fields changed by nfsm_uiombuf()
6424 		 * back.
6425 		 */
6426 		uiop->uio_offset -= len;
6427 		uiop->uio_resid += len;
6428 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6429 		uiop->uio_iov->iov_len += len;
6430 		error = nd->nd_repstat;
6431 	} else {
6432 		if (vers == NFS_VER3) {
6433 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6434 			    NULL);
6435 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6436 			if (error != 0)
6437 				goto nfsmout;
6438 		}
6439 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6440 		rlen = fxdr_unsigned(int, *tl++);
6441 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6442 		if (rlen == 0) {
6443 			error = NFSERR_IO;
6444 			goto nfsmout;
6445 		} else if (rlen < len) {
6446 			backup = len - rlen;
6447 			uiop->uio_iov->iov_base =
6448 			    (char *)uiop->uio_iov->iov_base - backup;
6449 			uiop->uio_iov->iov_len += backup;
6450 			uiop->uio_offset -= backup;
6451 			uiop->uio_resid += backup;
6452 			len = rlen;
6453 		}
6454 		commit = fxdr_unsigned(int, *tl++);
6455 
6456 		/*
6457 		 * Return the lowest commitment level
6458 		 * obtained by any of the RPCs.
6459 		 */
6460 		if (committed == NFSWRITE_FILESYNC)
6461 			committed = commit;
6462 		else if (committed == NFSWRITE_DATASYNC &&
6463 		    commit == NFSWRITE_UNSTABLE)
6464 			committed = commit;
6465 		if (commit_thru_mds != 0) {
6466 			NFSLOCKMNT(nmp);
6467 			if (!NFSHASWRITEVERF(nmp)) {
6468 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6469 				NFSSETWRITEVERF(nmp);
6470 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6471 				*must_commit = 1;
6472 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6473 			}
6474 			NFSUNLOCKMNT(nmp);
6475 		} else {
6476 			NFSLOCKDS(dsp);
6477 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6478 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6479 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6480 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6481 				*must_commit = 1;
6482 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6483 			}
6484 			NFSUNLOCKDS(dsp);
6485 		}
6486 	}
6487 nfsmout:
6488 	if (nd->nd_mrep != NULL)
6489 		m_freem(nd->nd_mrep);
6490 	*iomode = committed;
6491 	if (nd->nd_repstat != 0 && error == 0)
6492 		error = nd->nd_repstat;
6493 	return (error);
6494 }
6495 
6496 /*
6497  * The actual write RPC done to a DS.
6498  * This variant is called from a separate kernel process for mirrors.
6499  * Any short write is considered an IO error.
6500  */
6501 static int
6502 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6503     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6504     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6505     struct ucred *cred, NFSPROC_T *p)
6506 {
6507 	uint32_t *tl;
6508 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6509 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6510 	struct nfsrv_descript nfsd;
6511 	struct nfsrv_descript *nd = &nfsd;
6512 	struct nfssockreq *nrp;
6513 	struct nfsvattr na;
6514 
6515 	nd->nd_mrep = NULL;
6516 	if (vers == 0 || vers == NFS_VER4) {
6517 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6518 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6519 		vers = NFS_VER4;
6520 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6521 		    minorvers);
6522 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6523 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6524 	} else {
6525 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6526 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6527 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITE]);
6528 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_WRITEDS]);
6529 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6530 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6531 	}
6532 	txdr_hyper(io_off, tl);
6533 	tl += 2;
6534 	if (vers == NFS_VER3)
6535 		*tl++ = txdr_unsigned(len);
6536 	*tl++ = txdr_unsigned(*iomode);
6537 	*tl = txdr_unsigned(len);
6538 	if (len > 0) {
6539 		/* Put data in mbuf chain. */
6540 		nd->nd_mb->m_next = m;
6541 	}
6542 	nrp = dsp->nfsclds_sockp;
6543 	if (nrp == NULL)
6544 		/* If NULL, use the MDS socket. */
6545 		nrp = &nmp->nm_sockreq;
6546 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6547 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6548 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6549 	    nd->nd_repstat);
6550 	if (error != 0)
6551 		return (error);
6552 	if (nd->nd_repstat != 0)
6553 		error = nd->nd_repstat;
6554 	else {
6555 		if (vers == NFS_VER3) {
6556 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6557 			    NULL);
6558 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6559 			    error);
6560 			if (error != 0)
6561 				goto nfsmout;
6562 		}
6563 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6564 		rlen = fxdr_unsigned(int, *tl++);
6565 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6566 		    rlen);
6567 		if (rlen != len) {
6568 			error = NFSERR_IO;
6569 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6570 			    len, rlen);
6571 			goto nfsmout;
6572 		}
6573 		commit = fxdr_unsigned(int, *tl++);
6574 
6575 		/*
6576 		 * Return the lowest commitment level
6577 		 * obtained by any of the RPCs.
6578 		 */
6579 		if (committed == NFSWRITE_FILESYNC)
6580 			committed = commit;
6581 		else if (committed == NFSWRITE_DATASYNC &&
6582 		    commit == NFSWRITE_UNSTABLE)
6583 			committed = commit;
6584 		NFSLOCKDS(dsp);
6585 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6586 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6587 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6588 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6589 			*must_commit = 1;
6590 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6591 		}
6592 		NFSUNLOCKDS(dsp);
6593 	}
6594 nfsmout:
6595 	if (nd->nd_mrep != NULL)
6596 		m_freem(nd->nd_mrep);
6597 	*iomode = committed;
6598 	if (nd->nd_repstat != 0 && error == 0)
6599 		error = nd->nd_repstat;
6600 	return (error);
6601 }
6602 
6603 /*
6604  * Start up the thread that will execute nfsrpc_writedsmir().
6605  */
6606 static void
6607 start_writedsmir(void *arg, int pending)
6608 {
6609 	struct nfsclwritedsdorpc *drpc;
6610 
6611 	drpc = (struct nfsclwritedsdorpc *)arg;
6612 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6613 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6614 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6615 	    drpc->p);
6616 	drpc->done = 1;
6617 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6618 }
6619 
6620 /*
6621  * Set up the write DS mirror call for the pNFS I/O thread.
6622  */
6623 static int
6624 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6625     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6626     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6627     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6628 {
6629 	int error, ret;
6630 
6631 	error = 0;
6632 	drpc->done = 0;
6633 	drpc->vp = vp;
6634 	drpc->iomode = *iomode;
6635 	drpc->must_commit = *must_commit;
6636 	drpc->stateidp = stateidp;
6637 	drpc->dsp = dsp;
6638 	drpc->off = off;
6639 	drpc->len = len;
6640 	drpc->fhp = fhp;
6641 	drpc->m = m;
6642 	drpc->vers = vers;
6643 	drpc->minorvers = minorvers;
6644 	drpc->cred = cred;
6645 	drpc->p = p;
6646 	drpc->inprog = 0;
6647 	ret = EIO;
6648 	if (nfs_pnfsiothreads != 0) {
6649 		ret = nfs_pnfsio(start_writedsmir, drpc);
6650 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6651 	}
6652 	if (ret != 0)
6653 		error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6654 		    dsp, off, len, fhp, m, vers, minorvers, cred, p);
6655 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6656 	return (error);
6657 }
6658 
6659 /*
6660  * Free up the nfsclds structure.
6661  */
6662 void
6663 nfscl_freenfsclds(struct nfsclds *dsp)
6664 {
6665 	int i;
6666 
6667 	if (dsp == NULL)
6668 		return;
6669 	if (dsp->nfsclds_sockp != NULL) {
6670 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6671 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6672 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6673 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6674 	}
6675 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
6676 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6677 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6678 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6679 			m_freem(
6680 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6681 	}
6682 	free(dsp, M_NFSCLDS);
6683 }
6684 
6685 static enum nfsclds_state
6686 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6687     struct nfsclds **retdspp, uint32_t *sequencep)
6688 {
6689 	struct nfsclds *dsp;
6690 	int fndseq;
6691 
6692 	/*
6693 	 * Search the list of nfsclds structures for one with the same
6694 	 * server.
6695 	 */
6696 	fndseq = 0;
6697 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6698 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6699 		    dsp->nfsclds_servownlen != 0 &&
6700 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6701 		    dsp->nfsclds_servownlen) &&
6702 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
6703 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6704 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
6705 			    dsp->nfsclds_flags);
6706 			if (fndseq == 0) {
6707 				/* Get sequenceid# from first entry. */
6708 				*sequencep =
6709 				    dsp->nfsclds_sess.nfsess_sequenceid;
6710 				fndseq = 1;
6711 			}
6712 			/* Server major id matches. */
6713 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6714 				*retdspp = dsp;
6715 				return (NFSDSP_USETHISSESSION);
6716 			}
6717 		}
6718 	}
6719 	if (fndseq != 0)
6720 		return (NFSDSP_SEQTHISSESSION);
6721 	return (NFSDSP_NOTFOUND);
6722 }
6723 
6724 /*
6725  * NFS commit rpc to a NFSv4.1 DS.
6726  */
6727 static int
6728 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6729     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6730     NFSPROC_T *p)
6731 {
6732 	uint32_t *tl;
6733 	struct nfsrv_descript nfsd, *nd = &nfsd;
6734 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6735 	struct nfssockreq *nrp;
6736 	struct nfsvattr na;
6737 	int attrflag, error;
6738 
6739 	nd->nd_mrep = NULL;
6740 	if (vers == 0 || vers == NFS_VER4) {
6741 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6742 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6743 		vers = NFS_VER4;
6744 	} else {
6745 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6746 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6747 		NFSDECRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMIT]);
6748 		NFSINCRGLOBAL(nfsstatsv1.rpccnt[NFSPROC_COMMITDS]);
6749 	}
6750 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6751 	    minorvers);
6752 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6753 	txdr_hyper(offset, tl);
6754 	tl += 2;
6755 	*tl = txdr_unsigned(cnt);
6756 	nrp = dsp->nfsclds_sockp;
6757 	if (nrp == NULL)
6758 		/* If NULL, use the MDS socket. */
6759 		nrp = &nmp->nm_sockreq;
6760 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6761 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6762 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6763 	    nd->nd_repstat);
6764 	if (error != 0)
6765 		return (error);
6766 	if (nd->nd_repstat == 0) {
6767 		if (vers == NFS_VER3) {
6768 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6769 			    NULL);
6770 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6771 			if (error != 0)
6772 				goto nfsmout;
6773 		}
6774 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6775 		NFSLOCKDS(dsp);
6776 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6777 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6778 			error = NFSERR_STALEWRITEVERF;
6779 		}
6780 		NFSUNLOCKDS(dsp);
6781 	}
6782 nfsmout:
6783 	if (error == 0 && nd->nd_repstat != 0)
6784 		error = nd->nd_repstat;
6785 	m_freem(nd->nd_mrep);
6786 	return (error);
6787 }
6788 
6789 /*
6790  * Start up the thread that will execute nfsrpc_commitds().
6791  */
6792 static void
6793 start_commitds(void *arg, int pending)
6794 {
6795 	struct nfsclwritedsdorpc *drpc;
6796 
6797 	drpc = (struct nfsclwritedsdorpc *)arg;
6798 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6799 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6800 	    drpc->p);
6801 	drpc->done = 1;
6802 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6803 }
6804 
6805 /*
6806  * Set up the commit DS mirror call for the pNFS I/O thread.
6807  */
6808 static int
6809 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6810     struct nfsfh *fhp, int vers, int minorvers,
6811     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6812 {
6813 	int error, ret;
6814 
6815 	error = 0;
6816 	drpc->done = 0;
6817 	drpc->vp = vp;
6818 	drpc->off = offset;
6819 	drpc->len = cnt;
6820 	drpc->dsp = dsp;
6821 	drpc->fhp = fhp;
6822 	drpc->vers = vers;
6823 	drpc->minorvers = minorvers;
6824 	drpc->cred = cred;
6825 	drpc->p = p;
6826 	drpc->inprog = 0;
6827 	ret = EIO;
6828 	if (nfs_pnfsiothreads != 0) {
6829 		ret = nfs_pnfsio(start_commitds, drpc);
6830 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6831 	}
6832 	if (ret != 0)
6833 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6834 		    minorvers, cred, p);
6835 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6836 	return (error);
6837 }
6838 
6839 /*
6840  * NFS Advise rpc
6841  */
6842 int
6843 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
6844     struct ucred *cred, NFSPROC_T *p)
6845 {
6846 	u_int32_t *tl;
6847 	struct nfsrv_descript nfsd, *nd = &nfsd;
6848 	nfsattrbit_t hints;
6849 	int error;
6850 
6851 	NFSZERO_ATTRBIT(&hints);
6852 	if (advise == POSIX_FADV_WILLNEED)
6853 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6854 	else if (advise == POSIX_FADV_DONTNEED)
6855 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6856 	else
6857 		return (0);
6858 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
6859 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6860 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
6861 	txdr_hyper(offset, tl);
6862 	tl += 2;
6863 	txdr_hyper(cnt, tl);
6864 	nfsrv_putattrbit(nd, &hints);
6865 	error = nfscl_request(nd, vp, p, cred, NULL);
6866 	if (error != 0)
6867 		return (error);
6868 	if (nd->nd_repstat != 0)
6869 		error = nd->nd_repstat;
6870 	m_freem(nd->nd_mrep);
6871 	return (error);
6872 }
6873 
6874 #ifdef notyet
6875 /*
6876  * NFS advise rpc to a NFSv4.2 DS.
6877  */
6878 static int
6879 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6880     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6881     struct ucred *cred, NFSPROC_T *p)
6882 {
6883 	uint32_t *tl;
6884 	struct nfsrv_descript nfsd, *nd = &nfsd;
6885 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6886 	struct nfssockreq *nrp;
6887 	nfsattrbit_t hints;
6888 	int error;
6889 
6890 	/* For NFS DSs prior to NFSv4.2, just return OK. */
6891 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
6892 		return (0);
6893 	NFSZERO_ATTRBIT(&hints);
6894 	if (advise == POSIX_FADV_WILLNEED)
6895 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6896 	else if (advise == POSIX_FADV_DONTNEED)
6897 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6898 	else
6899 		return (0);
6900 	nd->nd_mrep = NULL;
6901 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
6902 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6903 	vers = NFS_VER4;
6904 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
6905 	    minorvers);
6906 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6907 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6908 	txdr_hyper(offset, tl);
6909 	tl += 2;
6910 	*tl = txdr_unsigned(cnt);
6911 	nfsrv_putattrbit(nd, &hints);
6912 	nrp = dsp->nfsclds_sockp;
6913 	if (nrp == NULL)
6914 		/* If NULL, use the MDS socket. */
6915 		nrp = &nmp->nm_sockreq;
6916 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6917 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6918 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
6919 	    nd->nd_repstat);
6920 	if (error != 0)
6921 		return (error);
6922 	if (nd->nd_repstat != 0)
6923 		error = nd->nd_repstat;
6924 	m_freem(nd->nd_mrep);
6925 	return (error);
6926 }
6927 
6928 /*
6929  * Start up the thread that will execute nfsrpc_commitds().
6930  */
6931 static void
6932 start_adviseds(void *arg, int pending)
6933 {
6934 	struct nfsclwritedsdorpc *drpc;
6935 
6936 	drpc = (struct nfsclwritedsdorpc *)arg;
6937 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
6938 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
6939 	    drpc->cred, drpc->p);
6940 	drpc->done = 1;
6941 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
6942 }
6943 
6944 /*
6945  * Set up the commit DS mirror call for the pNFS I/O thread.
6946  */
6947 static int
6948 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6949     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6950     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6951 {
6952 	int error, ret;
6953 
6954 	error = 0;
6955 	drpc->done = 0;
6956 	drpc->vp = vp;
6957 	drpc->off = offset;
6958 	drpc->len = cnt;
6959 	drpc->advise = advise;
6960 	drpc->dsp = dsp;
6961 	drpc->fhp = fhp;
6962 	drpc->vers = vers;
6963 	drpc->minorvers = minorvers;
6964 	drpc->cred = cred;
6965 	drpc->p = p;
6966 	drpc->inprog = 0;
6967 	ret = EIO;
6968 	if (nfs_pnfsiothreads != 0) {
6969 		ret = nfs_pnfsio(start_adviseds, drpc);
6970 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
6971 	}
6972 	if (ret != 0)
6973 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
6974 		    minorvers, cred, p);
6975 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
6976 	return (error);
6977 }
6978 #endif	/* notyet */
6979 
6980 /*
6981  * Do the Allocate operation, retrying for recovery.
6982  */
6983 int
6984 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
6985     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
6986 {
6987 	int error, expireret = 0, retrycnt, nostateid;
6988 	uint32_t clidrev = 0;
6989 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
6990 	struct nfsfh *nfhp = NULL;
6991 	nfsv4stateid_t stateid;
6992 	off_t tmp_off;
6993 	void *lckp;
6994 
6995 	if (len < 0)
6996 		return (EINVAL);
6997 	if (len == 0)
6998 		return (0);
6999 	tmp_off = off + len;
7000 	NFSLOCKMNT(nmp);
7001 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
7002 		NFSUNLOCKMNT(nmp);
7003 		return (EFBIG);
7004 	}
7005 	if (nmp->nm_clp != NULL)
7006 		clidrev = nmp->nm_clp->nfsc_clientidrev;
7007 	NFSUNLOCKMNT(nmp);
7008 	nfhp = VTONFS(vp)->n_fhp;
7009 	retrycnt = 0;
7010 	do {
7011 		lckp = NULL;
7012 		nostateid = 0;
7013 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7014 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7015 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7016 		    stateid.other[2] == 0) {
7017 			nostateid = 1;
7018 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7019 		}
7020 
7021 		/*
7022 		 * Not finding a stateid should probably never happen,
7023 		 * but just return an error for this case.
7024 		 */
7025 		if (nostateid != 0)
7026 			error = EIO;
7027 		else
7028 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7029 			    nap, attrflagp, cred, p, stuff);
7030 		if (error == NFSERR_STALESTATEID)
7031 			nfscl_initiate_recovery(nmp->nm_clp);
7032 		if (lckp != NULL)
7033 			nfscl_lockderef(lckp);
7034 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7035 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7036 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7037 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7038 		} else if ((error == NFSERR_EXPIRED ||
7039 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7040 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7041 		}
7042 		retrycnt++;
7043 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7044 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7045 	    error == NFSERR_STALEDONTRECOVER ||
7046 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7047 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7048 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7049 	if (error != 0 && retrycnt >= 4)
7050 		error = EIO;
7051 	return (error);
7052 }
7053 
7054 /*
7055  * The allocate RPC.
7056  */
7057 static int
7058 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7059     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p,
7060     void *stuff)
7061 {
7062 	uint32_t *tl;
7063 	int error;
7064 	struct nfsrv_descript nfsd;
7065 	struct nfsrv_descript *nd = &nfsd;
7066 	nfsattrbit_t attrbits;
7067 
7068 	*attrflagp = 0;
7069 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7070 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7071 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7072 	txdr_hyper(off, tl); tl += 2;
7073 	txdr_hyper(len, tl); tl += 2;
7074 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7075 	NFSGETATTR_ATTRBIT(&attrbits);
7076 	nfsrv_putattrbit(nd, &attrbits);
7077 	error = nfscl_request(nd, vp, p, cred, stuff);
7078 	if (error != 0)
7079 		return (error);
7080 	if (nd->nd_repstat == 0) {
7081 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7082 		error = nfsm_loadattr(nd, nap);
7083 		if (error == 0)
7084 			*attrflagp = NFS_LATTR_NOSHRINK;
7085 	} else
7086 		error = nd->nd_repstat;
7087 nfsmout:
7088 	m_freem(nd->nd_mrep);
7089 	return (error);
7090 }
7091 
7092 /*
7093  * Set up the XDR arguments for the LayoutGet operation.
7094  */
7095 static void
7096 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7097     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7098     int layoutlen, int usecurstateid)
7099 {
7100 	uint32_t *tl;
7101 
7102 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7103 	    NFSX_STATEID);
7104 	*tl++ = newnfs_false;		/* Don't signal availability. */
7105 	*tl++ = txdr_unsigned(layouttype);
7106 	*tl++ = txdr_unsigned(iomode);
7107 	txdr_hyper(offset, tl);
7108 	tl += 2;
7109 	txdr_hyper(len, tl);
7110 	tl += 2;
7111 	txdr_hyper(minlen, tl);
7112 	tl += 2;
7113 	if (usecurstateid != 0) {
7114 		/* Special stateid for Current stateid. */
7115 		*tl++ = txdr_unsigned(1);
7116 		*tl++ = 0;
7117 		*tl++ = 0;
7118 		*tl++ = 0;
7119 	} else {
7120 		*tl++ = txdr_unsigned(stateidp->seqid);
7121 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7122 		*tl++ = stateidp->other[0];
7123 		*tl++ = stateidp->other[1];
7124 		*tl++ = stateidp->other[2];
7125 	}
7126 	*tl = txdr_unsigned(layoutlen);
7127 }
7128 
7129 /*
7130  * Parse the reply for a successful LayoutGet operation.
7131  */
7132 static int
7133 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7134     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7135 {
7136 	uint32_t *tl;
7137 	struct nfsclflayout *flp, *prevflp, *tflp;
7138 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7139 	int m, mirrorcnt;
7140 	uint64_t retlen, off;
7141 	struct nfsfh *nfhp;
7142 	uint8_t *cp;
7143 	uid_t user;
7144 	gid_t grp;
7145 
7146 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7147 	error = 0;
7148 	flp = NULL;
7149 	gotiomode = -1;
7150 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7151 	if (*tl++ != 0)
7152 		*retonclosep = 1;
7153 	else
7154 		*retonclosep = 0;
7155 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7156 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7157 	    (int)stateidp->seqid);
7158 	stateidp->other[0] = *tl++;
7159 	stateidp->other[1] = *tl++;
7160 	stateidp->other[2] = *tl++;
7161 	cnt = fxdr_unsigned(int, *tl);
7162 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7163 	if (cnt <= 0 || cnt > 10000) {
7164 		/* Don't accept more than 10000 layouts in reply. */
7165 		error = NFSERR_BADXDR;
7166 		goto nfsmout;
7167 	}
7168 	for (i = 0; i < cnt; i++) {
7169 		/* Dissect to the layout type. */
7170 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7171 		    3 * NFSX_UNSIGNED);
7172 		off = fxdr_hyper(tl); tl += 2;
7173 		retlen = fxdr_hyper(tl); tl += 2;
7174 		iomode = fxdr_unsigned(int, *tl++);
7175 		laytype = fxdr_unsigned(int, *tl);
7176 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7177 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7178 		/* Ignore length of layout body for now. */
7179 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7180 			/* Parse the File layout up to fhcnt. */
7181 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7182 			    NFSX_HYPER + NFSX_V4DEVICEID);
7183 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7184 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7185 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7186 			if (fhcnt < 0 || fhcnt > 100) {
7187 				/* Don't accept more than 100 file handles. */
7188 				error = NFSERR_BADXDR;
7189 				goto nfsmout;
7190 			}
7191 			if (fhcnt > 0)
7192 				flp = malloc(sizeof(*flp) + fhcnt *
7193 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7194 				    M_WAITOK);
7195 			else
7196 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7197 				    M_WAITOK);
7198 			flp->nfsfl_flags = NFSFL_FILE;
7199 			flp->nfsfl_fhcnt = 0;
7200 			flp->nfsfl_devp = NULL;
7201 			flp->nfsfl_off = off;
7202 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7203 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7204 			else
7205 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7206 			flp->nfsfl_iomode = iomode;
7207 			if (gotiomode == -1)
7208 				gotiomode = flp->nfsfl_iomode;
7209 			/* Ignore layout body length for now. */
7210 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7211 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7212 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7213 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7214 			mtx_lock(&nmp->nm_mtx);
7215 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7216 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7217 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7218 			mtx_unlock(&nmp->nm_mtx);
7219 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7220 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7221 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7222 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7223 			for (j = 0; j < fhcnt; j++) {
7224 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7225 				nfhlen = fxdr_unsigned(int, *tl);
7226 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7227 					error = NFSERR_BADXDR;
7228 					goto nfsmout;
7229 				}
7230 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7231 				    M_NFSFH, M_WAITOK);
7232 				flp->nfsfl_fh[j] = nfhp;
7233 				flp->nfsfl_fhcnt++;
7234 				nfhp->nfh_len = nfhlen;
7235 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7236 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7237 			}
7238 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7239 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7240 			    NFSX_HYPER);
7241 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7242 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7243 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7244 				error = NFSERR_BADXDR;
7245 				goto nfsmout;
7246 			}
7247 			flp = malloc(sizeof(*flp) + mirrorcnt *
7248 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7249 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7250 			flp->nfsfl_mirrorcnt = mirrorcnt;
7251 			for (j = 0; j < mirrorcnt; j++)
7252 				flp->nfsfl_ffm[j].devp = NULL;
7253 			flp->nfsfl_off = off;
7254 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7255 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7256 			else
7257 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7258 			flp->nfsfl_iomode = iomode;
7259 			if (gotiomode == -1)
7260 				gotiomode = flp->nfsfl_iomode;
7261 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7262 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7263 			    (uintmax_t)flp->nfsfl_stripeunit);
7264 			for (j = 0; j < mirrorcnt; j++) {
7265 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7266 				k = fxdr_unsigned(int, *tl);
7267 				if (k < 1 || k > 128) {
7268 					error = NFSERR_BADXDR;
7269 					goto nfsmout;
7270 				}
7271 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7272 				for (l = 0; l < k; l++) {
7273 					NFSM_DISSECT(tl, uint32_t *,
7274 					    NFSX_V4DEVICEID + NFSX_STATEID +
7275 					    2 * NFSX_UNSIGNED);
7276 					if (l == 0) {
7277 						/* Just use the first server. */
7278 						NFSBCOPY(tl,
7279 						    flp->nfsfl_ffm[j].dev,
7280 						    NFSX_V4DEVICEID);
7281 						tl += (NFSX_V4DEVICEID /
7282 						    NFSX_UNSIGNED);
7283 						tl++;
7284 						flp->nfsfl_ffm[j].st.seqid =
7285 						    *tl++;
7286 						flp->nfsfl_ffm[j].st.other[0] =
7287 						    *tl++;
7288 						flp->nfsfl_ffm[j].st.other[1] =
7289 						    *tl++;
7290 						flp->nfsfl_ffm[j].st.other[2] =
7291 						    *tl++;
7292 						NFSCL_DEBUG(4, "st.seqid=%u "
7293 						 "st.o0=0x%x st.o1=0x%x "
7294 						 "st.o2=0x%x\n",
7295 						 flp->nfsfl_ffm[j].st.seqid,
7296 						 flp->nfsfl_ffm[j].st.other[0],
7297 						 flp->nfsfl_ffm[j].st.other[1],
7298 						 flp->nfsfl_ffm[j].st.other[2]);
7299 					} else
7300 						tl += ((NFSX_V4DEVICEID +
7301 						    NFSX_STATEID +
7302 						    NFSX_UNSIGNED) /
7303 						    NFSX_UNSIGNED);
7304 					fhcnt = fxdr_unsigned(int, *tl);
7305 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7306 					if (fhcnt < 1 ||
7307 					    fhcnt > NFSDEV_MAXVERS) {
7308 						error = NFSERR_BADXDR;
7309 						goto nfsmout;
7310 					}
7311 					for (m = 0; m < fhcnt; m++) {
7312 						NFSM_DISSECT(tl, uint32_t *,
7313 						    NFSX_UNSIGNED);
7314 						nfhlen = fxdr_unsigned(int,
7315 						    *tl);
7316 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7317 						    nfhlen);
7318 						if (nfhlen <= 0 || nfhlen >
7319 						    NFSX_V4FHMAX) {
7320 							error = NFSERR_BADXDR;
7321 							goto nfsmout;
7322 						}
7323 						NFSM_DISSECT(cp, uint8_t *,
7324 						    NFSM_RNDUP(nfhlen));
7325 						if (l == 0) {
7326 							flp->nfsfl_ffm[j].fhcnt
7327 							    = fhcnt;
7328 							nfhp = malloc(
7329 							    sizeof(*nfhp) +
7330 							    nfhlen - 1, M_NFSFH,
7331 							    M_WAITOK);
7332 							flp->nfsfl_ffm[j].fh[m]
7333 							    = nfhp;
7334 							nfhp->nfh_len = nfhlen;
7335 							NFSBCOPY(cp,
7336 							    nfhp->nfh_fh,
7337 							    nfhlen);
7338 							NFSCL_DEBUG(4,
7339 							    "got fh\n");
7340 						}
7341 					}
7342 					/* Now, get the ffsd_user/ffds_group. */
7343 					error = nfsrv_parseug(nd, 0, &user,
7344 					    &grp, curthread);
7345 					NFSCL_DEBUG(4, "after parseu=%d\n",
7346 					    error);
7347 					if (error == 0)
7348 						error = nfsrv_parseug(nd, 1,
7349 						    &user, &grp, curthread);
7350 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7351 					    grp);
7352 					if (error != 0)
7353 						goto nfsmout;
7354 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7355 					    user, grp);
7356 					if (l == 0) {
7357 						flp->nfsfl_ffm[j].user = user;
7358 						flp->nfsfl_ffm[j].group = grp;
7359 						NFSCL_DEBUG(4,
7360 						    "usr=%d grp=%d\n", user,
7361 						    grp);
7362 					}
7363 				}
7364 			}
7365 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7366 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7367 #ifdef notnow
7368 			/*
7369 			 * At this time, there is no flag.
7370 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7371 			 * added, or it may never exist?
7372 			 */
7373 			mtx_lock(&nmp->nm_mtx);
7374 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7375 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7376 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7377 			mtx_unlock(&nmp->nm_mtx);
7378 #endif
7379 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7380 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7381 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7382 		} else {
7383 			error = NFSERR_BADXDR;
7384 			goto nfsmout;
7385 		}
7386 		if (flp->nfsfl_iomode == gotiomode) {
7387 			/* Keep the list in increasing offset order. */
7388 			tflp = LIST_FIRST(flhp);
7389 			prevflp = NULL;
7390 			while (tflp != NULL &&
7391 			    tflp->nfsfl_off < flp->nfsfl_off) {
7392 				prevflp = tflp;
7393 				tflp = LIST_NEXT(tflp, nfsfl_list);
7394 			}
7395 			if (prevflp == NULL)
7396 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7397 			else
7398 				LIST_INSERT_AFTER(prevflp, flp,
7399 				    nfsfl_list);
7400 			NFSCL_DEBUG(4, "flp inserted\n");
7401 		} else {
7402 			printf("nfscl_layoutget(): got wrong iomode\n");
7403 			nfscl_freeflayout(flp);
7404 		}
7405 		flp = NULL;
7406 	}
7407 nfsmout:
7408 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7409 	if (error != 0 && flp != NULL)
7410 		nfscl_freeflayout(flp);
7411 	return (error);
7412 }
7413 
7414 /*
7415  * Parse a user/group digit string.
7416  */
7417 static int
7418 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7419     NFSPROC_T *p)
7420 {
7421 	uint32_t *tl;
7422 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7423 	uint32_t len = 0;
7424 	int error = 0;
7425 
7426 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7427 	len = fxdr_unsigned(uint32_t, *tl);
7428 	str = NULL;
7429 	if (len > NFSV4_OPAQUELIMIT) {
7430 		error = NFSERR_BADXDR;
7431 		goto nfsmout;
7432 	}
7433 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7434 	if (len == 0) {
7435 		if (dogrp != 0)
7436 			*gidp = GID_NOGROUP;
7437 		else
7438 			*uidp = UID_NOBODY;
7439 		return (0);
7440 	}
7441 	if (len > NFSV4_SMALLSTR)
7442 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7443 	else
7444 		str = str0;
7445 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7446 	NFSBCOPY(cp, str, len);
7447 	str[len] = '\0';
7448 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7449 	if (dogrp != 0)
7450 		error = nfsv4_strtogid(nd, str, len, gidp);
7451 	else
7452 		error = nfsv4_strtouid(nd, str, len, uidp);
7453 nfsmout:
7454 	if (len > NFSV4_SMALLSTR)
7455 		free(str, M_TEMP);
7456 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7457 	return (error);
7458 }
7459 
7460 /*
7461  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7462  * so that it does both an Open and a Layoutget.
7463  */
7464 static int
7465 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7466     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7467     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7468     struct ucred *cred, NFSPROC_T *p)
7469 {
7470 	struct nfscllayout *lyp;
7471 	struct nfsclflayout *flp;
7472 	struct nfsclflayouthead flh;
7473 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7474 	int layouttype, laystat;
7475 	nfsv4stateid_t stateid;
7476 	struct nfsclsession *tsep;
7477 
7478 	error = 0;
7479 	if (NFSHASFLEXFILE(nmp))
7480 		layouttype = NFSLAYOUT_FLEXFILE;
7481 	else
7482 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7483 	/*
7484 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7485 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7486 	 * flp == NULL.
7487 	 */
7488 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7489 	    &recalled);
7490 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7491 	if (lyp == NULL)
7492 		islocked = 0;
7493 	else if (flp != NULL)
7494 		islocked = 1;
7495 	else
7496 		islocked = 2;
7497 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7498 		LIST_INIT(&flh);
7499 		tsep = nfsmnt_mdssession(nmp);
7500 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7501 		    3 * NFSX_UNSIGNED);
7502 		if (lyp == NULL)
7503 			usecurstateid = 1;
7504 		else {
7505 			usecurstateid = 0;
7506 			stateid.seqid = lyp->nfsly_stateid.seqid;
7507 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7508 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7509 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7510 		}
7511 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7512 		    newfhp, newfhlen, mode, op, name, namelen,
7513 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7514 		    &retonclose, &flh, &laystat, cred, p);
7515 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7516 		    laystat, error);
7517 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7518 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7519 		    &islocked, cred, p);
7520 	} else
7521 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7522 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7523 	if (islocked == 2)
7524 		nfscl_rellayout(lyp, 1);
7525 	else if (islocked == 1)
7526 		nfscl_rellayout(lyp, 0);
7527 	return (error);
7528 }
7529 
7530 /*
7531  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7532  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7533  * handled by nfsrpc_openrpc().
7534  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7535  * can be NULL.
7536  */
7537 static int
7538 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7539     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7540     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7541     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7542     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7543     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7544 {
7545 	uint32_t *tl;
7546 	struct nfsrv_descript nfsd, *nd = &nfsd;
7547 	struct nfscldeleg *ndp = NULL;
7548 	struct nfsvattr nfsva;
7549 	struct nfsclsession *tsep;
7550 	uint32_t rflags, deleg;
7551 	nfsattrbit_t attrbits;
7552 	int error, ret, acesize, limitby, iomode;
7553 
7554 	*dpp = NULL;
7555 	*laystatp = ENXIO;
7556 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7557 	    0, 0);
7558 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7559 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7560 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7561 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7562 	tsep = nfsmnt_mdssession(nmp);
7563 	*tl++ = tsep->nfsess_clientid.lval[0];
7564 	*tl = tsep->nfsess_clientid.lval[1];
7565 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7566 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7567 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7568 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7569 	nfsm_strtom(nd, name, namelen);
7570 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7571 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7572 	NFSZERO_ATTRBIT(&attrbits);
7573 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7574 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7575 	nfsrv_putattrbit(nd, &attrbits);
7576 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7577 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7578 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7579 		iomode = NFSLAYOUTIOMODE_RW;
7580 	else
7581 		iomode = NFSLAYOUTIOMODE_READ;
7582 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7583 	    layouttype, layoutlen, usecurstateid);
7584 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7585 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7586 	if (error != 0)
7587 		return (error);
7588 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7589 	if (nd->nd_repstat != 0)
7590 		*laystatp = nd->nd_repstat;
7591 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7592 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7593 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7594 		    6 * NFSX_UNSIGNED);
7595 		op->nfso_stateid.seqid = *tl++;
7596 		op->nfso_stateid.other[0] = *tl++;
7597 		op->nfso_stateid.other[1] = *tl++;
7598 		op->nfso_stateid.other[2] = *tl;
7599 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7600 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7601 		if (error != 0)
7602 			goto nfsmout;
7603 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7604 		deleg = fxdr_unsigned(u_int32_t, *tl);
7605 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7606 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7607 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7608 			      NFSCLFLAGS_FIRSTDELEG))
7609 				op->nfso_own->nfsow_clp->nfsc_flags |=
7610 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7611 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7612 			    M_NFSCLDELEG, M_WAITOK);
7613 			LIST_INIT(&ndp->nfsdl_owner);
7614 			LIST_INIT(&ndp->nfsdl_lock);
7615 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7616 			ndp->nfsdl_fhlen = newfhlen;
7617 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7618 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
7619 			nfscl_lockinit(&ndp->nfsdl_rwlock);
7620 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7621 			    NFSX_UNSIGNED);
7622 			ndp->nfsdl_stateid.seqid = *tl++;
7623 			ndp->nfsdl_stateid.other[0] = *tl++;
7624 			ndp->nfsdl_stateid.other[1] = *tl++;
7625 			ndp->nfsdl_stateid.other[2] = *tl++;
7626 			ret = fxdr_unsigned(int, *tl);
7627 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7628 				ndp->nfsdl_flags = NFSCLDL_WRITE;
7629 				/*
7630 				 * Indicates how much the file can grow.
7631 				 */
7632 				NFSM_DISSECT(tl, u_int32_t *,
7633 				    3 * NFSX_UNSIGNED);
7634 				limitby = fxdr_unsigned(int, *tl++);
7635 				switch (limitby) {
7636 				case NFSV4OPEN_LIMITSIZE:
7637 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7638 					break;
7639 				case NFSV4OPEN_LIMITBLOCKS:
7640 					ndp->nfsdl_sizelimit =
7641 					    fxdr_unsigned(u_int64_t, *tl++);
7642 					ndp->nfsdl_sizelimit *=
7643 					    fxdr_unsigned(u_int64_t, *tl);
7644 					break;
7645 				default:
7646 					error = NFSERR_BADXDR;
7647 					goto nfsmout;
7648 				};
7649 			} else
7650 				ndp->nfsdl_flags = NFSCLDL_READ;
7651 			if (ret != 0)
7652 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
7653 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7654 			    &acesize, p);
7655 			if (error != 0)
7656 				goto nfsmout;
7657 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7658 			error = NFSERR_BADXDR;
7659 			goto nfsmout;
7660 		}
7661 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7662 		    nfscl_assumeposixlocks)
7663 			op->nfso_posixlock = 1;
7664 		else
7665 			op->nfso_posixlock = 0;
7666 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7667 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
7668 		if (*++tl == 0) {
7669 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7670 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7671 			    NULL, NULL, NULL, p, cred);
7672 			if (error != 0)
7673 				goto nfsmout;
7674 			if (ndp != NULL) {
7675 				ndp->nfsdl_change = nfsva.na_filerev;
7676 				ndp->nfsdl_modtime = nfsva.na_mtime;
7677 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7678 				*dpp = ndp;
7679 				ndp = NULL;
7680 			}
7681 			/*
7682 			 * At this point, the Open has succeeded, so set
7683 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
7684 			 * this function just won't return a layout.
7685 			 */
7686 			if (nd->nd_repstat == 0) {
7687 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7688 				*laystatp = fxdr_unsigned(int, *++tl);
7689 				if (*laystatp == 0) {
7690 					error = nfsrv_parselayoutget(nmp, nd,
7691 					    stateidp, retonclosep, flhp);
7692 					if (error != 0)
7693 						*laystatp = error;
7694 				}
7695 			} else
7696 				nd->nd_repstat = 0;	/* Return 0 for Open. */
7697 		}
7698 	}
7699 	if (nd->nd_repstat != 0 && error == 0)
7700 		error = nd->nd_repstat;
7701 nfsmout:
7702 	free(ndp, M_NFSCLDELEG);
7703 	m_freem(nd->nd_mrep);
7704 	return (error);
7705 }
7706 
7707 /*
7708  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7709  * Used only for mounts with pNFS enabled.
7710  */
7711 static int
7712 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7713     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7714     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7715     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7716     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7717     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7718     struct nfsclflayouthead *flhp, int *laystatp)
7719 {
7720 	uint32_t *tl;
7721 	int error = 0, deleg, newone, ret, acesize, limitby;
7722 	struct nfsrv_descript nfsd, *nd = &nfsd;
7723 	struct nfsclopen *op;
7724 	struct nfscldeleg *dp = NULL;
7725 	struct nfsnode *np;
7726 	struct nfsfh *nfhp;
7727 	struct nfsclsession *tsep;
7728 	nfsattrbit_t attrbits;
7729 	nfsv4stateid_t stateid;
7730 	struct nfsmount *nmp;
7731 
7732 	nmp = VFSTONFS(dvp->v_mount);
7733 	np = VTONFS(dvp);
7734 	*laystatp = ENXIO;
7735 	*unlockedp = 0;
7736 	*nfhpp = NULL;
7737 	*dpp = NULL;
7738 	*attrflagp = 0;
7739 	*dattrflagp = 0;
7740 	if (namelen > NFS_MAXNAMLEN)
7741 		return (ENAMETOOLONG);
7742 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7743 	/*
7744 	 * For V4, this is actually an Open op.
7745 	 */
7746 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7747 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
7748 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7749 	    NFSV4OPEN_ACCESSREAD);
7750 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7751 	tsep = nfsmnt_mdssession(nmp);
7752 	*tl++ = tsep->nfsess_clientid.lval[0];
7753 	*tl = tsep->nfsess_clientid.lval[1];
7754 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7755 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7756 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7757 	if ((fmode & O_EXCL) != 0) {
7758 		if (NFSHASSESSPERSIST(nmp)) {
7759 			/* Use GUARDED for persistent sessions. */
7760 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
7761 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7762 		} else {
7763 			/* Otherwise, use EXCLUSIVE4_1. */
7764 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7765 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7766 			*tl++ = cverf.lval[0];
7767 			*tl = cverf.lval[1];
7768 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7769 		}
7770 	} else {
7771 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7772 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
7773 	}
7774 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7775 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7776 	nfsm_strtom(nd, name, namelen);
7777 	/* Get the new file's handle and attributes, plus save the FH. */
7778 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7779 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7780 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
7781 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7782 	NFSGETATTR_ATTRBIT(&attrbits);
7783 	nfsrv_putattrbit(nd, &attrbits);
7784 	/* Get the directory's post-op attributes. */
7785 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7786 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
7787 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7788 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7789 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7790 	nfsrv_putattrbit(nd, &attrbits);
7791 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7792 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7793 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7794 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7795 	    layouttype, layoutlen, usecurstateid);
7796 	error = nfscl_request(nd, dvp, p, cred, dstuff);
7797 	if (error != 0)
7798 		return (error);
7799 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7800 	    error);
7801 	if (nd->nd_repstat != 0)
7802 		*laystatp = nd->nd_repstat;
7803 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7804 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7805 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7806 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7807 		    6 * NFSX_UNSIGNED);
7808 		stateid.seqid = *tl++;
7809 		stateid.other[0] = *tl++;
7810 		stateid.other[1] = *tl++;
7811 		stateid.other[2] = *tl;
7812 		nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7813 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7814 		deleg = fxdr_unsigned(int, *tl);
7815 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7816 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7817 			if (!(owp->nfsow_clp->nfsc_flags &
7818 			      NFSCLFLAGS_FIRSTDELEG))
7819 				owp->nfsow_clp->nfsc_flags |=
7820 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7821 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7822 			    M_NFSCLDELEG, M_WAITOK);
7823 			LIST_INIT(&dp->nfsdl_owner);
7824 			LIST_INIT(&dp->nfsdl_lock);
7825 			dp->nfsdl_clp = owp->nfsow_clp;
7826 			newnfs_copyincred(cred, &dp->nfsdl_cred);
7827 			nfscl_lockinit(&dp->nfsdl_rwlock);
7828 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7829 			    NFSX_UNSIGNED);
7830 			dp->nfsdl_stateid.seqid = *tl++;
7831 			dp->nfsdl_stateid.other[0] = *tl++;
7832 			dp->nfsdl_stateid.other[1] = *tl++;
7833 			dp->nfsdl_stateid.other[2] = *tl++;
7834 			ret = fxdr_unsigned(int, *tl);
7835 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7836 				dp->nfsdl_flags = NFSCLDL_WRITE;
7837 				/*
7838 				 * Indicates how much the file can grow.
7839 				 */
7840 				NFSM_DISSECT(tl, u_int32_t *,
7841 				    3 * NFSX_UNSIGNED);
7842 				limitby = fxdr_unsigned(int, *tl++);
7843 				switch (limitby) {
7844 				case NFSV4OPEN_LIMITSIZE:
7845 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
7846 					break;
7847 				case NFSV4OPEN_LIMITBLOCKS:
7848 					dp->nfsdl_sizelimit =
7849 					    fxdr_unsigned(u_int64_t, *tl++);
7850 					dp->nfsdl_sizelimit *=
7851 					    fxdr_unsigned(u_int64_t, *tl);
7852 					break;
7853 				default:
7854 					error = NFSERR_BADXDR;
7855 					goto nfsmout;
7856 				};
7857 			} else {
7858 				dp->nfsdl_flags = NFSCLDL_READ;
7859 			}
7860 			if (ret != 0)
7861 				dp->nfsdl_flags |= NFSCLDL_RECALL;
7862 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7863 			    &acesize, p);
7864 			if (error != 0)
7865 				goto nfsmout;
7866 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7867 			error = NFSERR_BADXDR;
7868 			goto nfsmout;
7869 		}
7870 
7871 		/* Now, we should have the status for the SaveFH. */
7872 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7873 		if (*++tl == 0) {
7874 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7875 			/*
7876 			 * Now, process the GetFH and Getattr for the newly
7877 			 * created file. nfscl_mtofh() will set
7878 			 * ND_NOMOREDATA if these weren't successful.
7879 			 */
7880 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7881 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7882 			if (error != 0)
7883 				goto nfsmout;
7884 		} else
7885 			nd->nd_flag |= ND_NOMOREDATA;
7886 		/* Now we have the PutFH and Getattr for the directory. */
7887 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7888 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7889 			if (*++tl != 0)
7890 				nd->nd_flag |= ND_NOMOREDATA;
7891 			else {
7892 				NFSM_DISSECT(tl, uint32_t *, 2 *
7893 				    NFSX_UNSIGNED);
7894 				if (*++tl != 0)
7895 					nd->nd_flag |= ND_NOMOREDATA;
7896 			}
7897 		}
7898 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7899 			/* Load the directory attributes. */
7900 			error = nfsm_loadattr(nd, dnap);
7901 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7902 			if (error != 0)
7903 				goto nfsmout;
7904 			*dattrflagp = 1;
7905 			if (dp != NULL && *attrflagp != 0) {
7906 				dp->nfsdl_change = nnap->na_filerev;
7907 				dp->nfsdl_modtime = nnap->na_mtime;
7908 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7909 			}
7910 			/*
7911 			 * We can now complete the Open state.
7912 			 */
7913 			nfhp = *nfhpp;
7914 			if (dp != NULL) {
7915 				dp->nfsdl_fhlen = nfhp->nfh_len;
7916 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7917 				    nfhp->nfh_len);
7918 			}
7919 			/*
7920 			 * Get an Open structure that will be
7921 			 * attached to the OpenOwner, acquired already.
7922 			 */
7923 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7924 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7925 			    cred, p, NULL, &op, &newone, NULL, 0);
7926 			if (error != 0)
7927 				goto nfsmout;
7928 			op->nfso_stateid = stateid;
7929 			newnfs_copyincred(cred, &op->nfso_cred);
7930 
7931 			nfscl_openrelease(nmp, op, error, newone);
7932 			*unlockedp = 1;
7933 
7934 			/* Now, handle the RestoreFH and LayoutGet. */
7935 			if (nd->nd_repstat == 0) {
7936 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7937 				*laystatp = fxdr_unsigned(int, *(tl + 3));
7938 				if (*laystatp == 0) {
7939 					error = nfsrv_parselayoutget(nmp, nd,
7940 					    stateidp, retonclosep, flhp);
7941 					if (error != 0)
7942 						*laystatp = error;
7943 				}
7944 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7945 				    error);
7946 			} else
7947 				nd->nd_repstat = 0;
7948 		}
7949 	}
7950 	if (nd->nd_repstat != 0 && error == 0)
7951 		error = nd->nd_repstat;
7952 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7953 		nfscl_initiate_recovery(owp->nfsow_clp);
7954 nfsmout:
7955 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7956 	if (error == 0)
7957 		*dpp = dp;
7958 	else
7959 		free(dp, M_NFSCLDELEG);
7960 	m_freem(nd->nd_mrep);
7961 	return (error);
7962 }
7963 
7964 /*
7965  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7966  */
7967 static int
7968 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7969     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7970     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7971     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7972     int *dattrflagp, void *dstuff, int *unlockedp)
7973 {
7974 	struct nfscllayout *lyp;
7975 	struct nfsclflayouthead flh;
7976 	struct nfsfh *nfhp;
7977 	struct nfsclsession *tsep;
7978 	struct nfsmount *nmp;
7979 	nfsv4stateid_t stateid;
7980 	int error, layoutlen, layouttype, retonclose, laystat;
7981 
7982 	error = 0;
7983 	nmp = VFSTONFS(dvp->v_mount);
7984 	if (NFSHASFLEXFILE(nmp))
7985 		layouttype = NFSLAYOUT_FLEXFILE;
7986 	else
7987 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7988 	LIST_INIT(&flh);
7989 	tsep = nfsmnt_mdssession(nmp);
7990 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7991 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7992 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7993 	    dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7994 	    &flh, &laystat);
7995 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7996 	    laystat, error);
7997 	lyp = NULL;
7998 	if (laystat == 0) {
7999 		nfhp = *nfhpp;
8000 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
8001 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
8002 		    layouttype, laystat, NULL, cred, p);
8003 	} else
8004 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
8005 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
8006 		    cred, p);
8007 	if (laystat == 0)
8008 		nfscl_rellayout(lyp, 0);
8009 	return (error);
8010 }
8011 
8012 /*
8013  * Process the results of a layoutget() operation.
8014  */
8015 static int
8016 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8017     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8018     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8019     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8020 {
8021 	struct nfsclflayout *tflp;
8022 	struct nfscldevinfo *dip;
8023 	uint8_t *dev;
8024 	int i, mirrorcnt;
8025 
8026 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8027 		NFSLOCKMNT(nmp);
8028 		if (!NFSHASFLEXFILE(nmp)) {
8029 			/* Switch to using Flex File Layout. */
8030 			nmp->nm_state |= NFSSTA_FLEXFILE;
8031 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8032 			/* Disable pNFS. */
8033 			NFSCL_DEBUG(1, "disable PNFS\n");
8034 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8035 		}
8036 		NFSUNLOCKMNT(nmp);
8037 	}
8038 	if (laystat == 0) {
8039 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8040 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8041 			if (layouttype == NFSLAYOUT_FLEXFILE)
8042 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8043 			else
8044 				mirrorcnt = 1;
8045 			for (i = 0; i < mirrorcnt; i++) {
8046 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8047 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8048 				if (laystat != 0) {
8049 					if (layouttype == NFSLAYOUT_FLEXFILE)
8050 						dev = tflp->nfsfl_ffm[i].dev;
8051 					else
8052 						dev = tflp->nfsfl_dev;
8053 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8054 					    layouttype, notifybit, &dip, cred,
8055 					    p);
8056 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8057 					    laystat);
8058 					if (laystat != 0)
8059 						goto out;
8060 					laystat = nfscl_adddevinfo(nmp, dip, i,
8061 					    tflp);
8062 					if (laystat != 0)
8063 						printf("nfsrpc_layoutgetresout"
8064 						    ": cannot add\n");
8065 				}
8066 			}
8067 		}
8068 	}
8069 out:
8070 	if (laystat == 0) {
8071 		/*
8072 		 * nfscl_layout() always returns with the nfsly_lock
8073 		 * set to a refcnt (shared lock).
8074 		 * Passing in dvp is sufficient, since it is only used to
8075 		 * get the fsid for the file system.
8076 		 */
8077 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8078 		    layouttype, retonclose, flhp, lypp, cred, p);
8079 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8080 		    laystat);
8081 		if (laystat == 0 && islockedp != NULL)
8082 			*islockedp = 1;
8083 	}
8084 	return (laystat);
8085 }
8086 
8087 /*
8088  * nfs copy_file_range operation.
8089  */
8090 int
8091 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8092     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8093     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8094     struct ucred *cred, bool consecutive, bool *must_commitp)
8095 {
8096 	int commit, error, expireret = 0, retrycnt;
8097 	u_int32_t clidrev = 0;
8098 	struct nfsmount *nmp = VFSTONFS(invp->v_mount);
8099 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8100 	nfsv4stateid_t instateid, outstateid;
8101 	void *inlckp, *outlckp;
8102 
8103 	if (nmp->nm_clp != NULL)
8104 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8105 	innfhp = VTONFS(invp)->n_fhp;
8106 	outnfhp = VTONFS(outvp)->n_fhp;
8107 	retrycnt = 0;
8108 	do {
8109 		/* Get both stateids. */
8110 		inlckp = NULL;
8111 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8112 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8113 		    &inlckp);
8114 		outlckp = NULL;
8115 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8116 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8117 		    &outlckp);
8118 
8119 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8120 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8121 		    outattrflagp, consecutive, &commit, cred, curthread);
8122 		if (error == 0) {
8123 			if (commit != NFSWRITE_FILESYNC)
8124 				*must_commitp = true;
8125 			*inoffp += *lenp;
8126 			*outoffp += *lenp;
8127 		} else if (error == NFSERR_STALESTATEID)
8128 			nfscl_initiate_recovery(nmp->nm_clp);
8129 		if (inlckp != NULL)
8130 			nfscl_lockderef(inlckp);
8131 		if (outlckp != NULL)
8132 			nfscl_lockderef(outlckp);
8133 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8134 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8135 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8136 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8137 		} else if ((error == NFSERR_EXPIRED ||
8138 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8139 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8140 			    curthread);
8141 		}
8142 		retrycnt++;
8143 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8144 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8145 	      error == NFSERR_STALEDONTRECOVER ||
8146 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8147 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8148 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8149 	if (error != 0 && (retrycnt >= 4 ||
8150 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8151 	      error == NFSERR_STALEDONTRECOVER))
8152 		error = EIO;
8153 	return (error);
8154 }
8155 
8156 /*
8157  * The copy RPC.
8158  */
8159 static int
8160 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8161     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8162     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8163     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8164     NFSPROC_T *p)
8165 {
8166 	uint32_t *tl;
8167 	int error;
8168 	struct nfsrv_descript nfsd;
8169 	struct nfsrv_descript *nd = &nfsd;
8170 	struct nfsmount *nmp;
8171 	nfsattrbit_t attrbits;
8172 	uint64_t len;
8173 
8174 	nmp = VFSTONFS(outvp->v_mount);
8175 	*inattrflagp = *outattrflagp = 0;
8176 	*commitp = NFSWRITE_UNSTABLE;
8177 	len = *lenp;
8178 	*lenp = 0;
8179 	if (len > nfs_maxcopyrange)
8180 		len = nfs_maxcopyrange;
8181 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8182 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8183 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8184 	NFSGETATTR_ATTRBIT(&attrbits);
8185 	nfsrv_putattrbit(nd, &attrbits);
8186 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8187 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8188 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8189 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8190 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8191 	*tl = txdr_unsigned(NFSV4OP_COPY);
8192 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8193 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8194 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8195 	txdr_hyper(inoff, tl); tl += 2;
8196 	txdr_hyper(outoff, tl); tl += 2;
8197 	txdr_hyper(len, tl); tl += 2;
8198 	if (consecutive)
8199 		*tl++ = newnfs_true;
8200 	else
8201 		*tl++ = newnfs_false;
8202 	*tl++ = newnfs_true;
8203 	*tl++ = 0;
8204 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8205 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8206 	nfsrv_putattrbit(nd, &attrbits);
8207 	error = nfscl_request(nd, invp, p, cred, NULL);
8208 	if (error != 0)
8209 		return (error);
8210 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8211 		/* Get the input file's attributes. */
8212 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8213 		if (*(tl + 1) == 0) {
8214 			error = nfsm_loadattr(nd, innap);
8215 			if (error != 0)
8216 				goto nfsmout;
8217 			*inattrflagp = 1;
8218 		} else
8219 			nd->nd_flag |= ND_NOMOREDATA;
8220 	}
8221 	/* Skip over return stat for PutFH. */
8222 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8223 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8224 		if (*++tl != 0)
8225 			nd->nd_flag |= ND_NOMOREDATA;
8226 	}
8227 	/* Skip over return stat for Copy. */
8228 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8229 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8230 	if (nd->nd_repstat == 0) {
8231 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8232 		if (*tl != 0) {
8233 			/* There should be no callback ids. */
8234 			error = NFSERR_BADXDR;
8235 			goto nfsmout;
8236 		}
8237 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8238 		    NFSX_VERF);
8239 		len = fxdr_hyper(tl); tl += 2;
8240 		*commitp = fxdr_unsigned(int, *tl++);
8241 		NFSLOCKMNT(nmp);
8242 		if (!NFSHASWRITEVERF(nmp)) {
8243 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8244 			NFSSETWRITEVERF(nmp);
8245 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8246 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8247 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8248 		}
8249 		NFSUNLOCKMNT(nmp);
8250 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8251 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8252 			/* Must be a synchronous copy. */
8253 			nd->nd_repstat = NFSERR_NOTSUPP;
8254 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8255 		error = nfsm_loadattr(nd, outnap);
8256 		if (error == 0)
8257 			*outattrflagp = NFS_LATTR_NOSHRINK;
8258 		if (nd->nd_repstat == 0)
8259 			*lenp = len;
8260 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8261 		/*
8262 		 * For the case where consecutive is not supported, but
8263 		 * synchronous is supported, we can try consecutive == false
8264 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8265 		 * since Copy cannot be done.
8266 		 */
8267 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8268 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8269 			if (!consecutive || *++tl == newnfs_false)
8270 				nd->nd_repstat = NFSERR_NOTSUPP;
8271 		} else
8272 			nd->nd_repstat = NFSERR_BADXDR;
8273 	}
8274 	if (error == 0)
8275 		error = nd->nd_repstat;
8276 nfsmout:
8277 	m_freem(nd->nd_mrep);
8278 	return (error);
8279 }
8280 
8281 /*
8282  * Seek operation.
8283  */
8284 int
8285 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8286     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8287 {
8288 	int error, expireret = 0, retrycnt;
8289 	u_int32_t clidrev = 0;
8290 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
8291 	struct nfsnode *np = VTONFS(vp);
8292 	struct nfsfh *nfhp = NULL;
8293 	nfsv4stateid_t stateid;
8294 	void *lckp;
8295 
8296 	if (nmp->nm_clp != NULL)
8297 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8298 	nfhp = np->n_fhp;
8299 	retrycnt = 0;
8300 	do {
8301 		lckp = NULL;
8302 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8303 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8304 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8305 		    nap, attrflagp, cred);
8306 		if (error == NFSERR_STALESTATEID)
8307 			nfscl_initiate_recovery(nmp->nm_clp);
8308 		if (lckp != NULL)
8309 			nfscl_lockderef(lckp);
8310 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8311 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8312 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8313 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8314 		} else if ((error == NFSERR_EXPIRED ||
8315 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8316 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8317 			    curthread);
8318 		}
8319 		retrycnt++;
8320 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8321 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8322 	    error == NFSERR_BADSESSION ||
8323 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8324 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8325 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8326 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8327 	if (error && retrycnt >= 4)
8328 		error = EIO;
8329 	return (error);
8330 }
8331 
8332 /*
8333  * The seek RPC.
8334  */
8335 static int
8336 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8337     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8338 {
8339 	uint32_t *tl;
8340 	int error;
8341 	struct nfsrv_descript nfsd;
8342 	struct nfsrv_descript *nd = &nfsd;
8343 	nfsattrbit_t attrbits;
8344 
8345 	*attrflagp = 0;
8346 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8347 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8348 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8349 	txdr_hyper(*offp, tl); tl += 2;
8350 	*tl++ = txdr_unsigned(content);
8351 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8352 	NFSGETATTR_ATTRBIT(&attrbits);
8353 	nfsrv_putattrbit(nd, &attrbits);
8354 	error = nfscl_request(nd, vp, curthread, cred, NULL);
8355 	if (error != 0)
8356 		return (error);
8357 	if (nd->nd_repstat == 0) {
8358 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8359 		if (*tl++ == newnfs_true)
8360 			*eofp = true;
8361 		else
8362 			*eofp = false;
8363 		*offp = fxdr_hyper(tl);
8364 		/* Just skip over Getattr op status. */
8365 		error = nfsm_loadattr(nd, nap);
8366 		if (error == 0)
8367 			*attrflagp = 1;
8368 	}
8369 	error = nd->nd_repstat;
8370 nfsmout:
8371 	m_freem(nd->nd_mrep);
8372 	return (error);
8373 }
8374 
8375 /*
8376  * The getextattr RPC.
8377  */
8378 int
8379 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8380     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8381 {
8382 	uint32_t *tl;
8383 	int error;
8384 	struct nfsrv_descript nfsd;
8385 	struct nfsrv_descript *nd = &nfsd;
8386 	nfsattrbit_t attrbits;
8387 	uint32_t len, len2;
8388 
8389 	*attrflagp = 0;
8390 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8391 	nfsm_strtom(nd, name, strlen(name));
8392 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8393 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8394 	NFSGETATTR_ATTRBIT(&attrbits);
8395 	nfsrv_putattrbit(nd, &attrbits);
8396 	error = nfscl_request(nd, vp, p, cred, NULL);
8397 	if (error != 0)
8398 		return (error);
8399 	if (nd->nd_repstat == 0) {
8400 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8401 		len = fxdr_unsigned(uint32_t, *tl);
8402 		/* Sanity check lengths. */
8403 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8404 		    uiop->uio_resid <= UINT32_MAX) {
8405 			len2 = uiop->uio_resid;
8406 			if (len2 >= len)
8407 				error = nfsm_mbufuio(nd, uiop, len);
8408 			else {
8409 				error = nfsm_mbufuio(nd, uiop, len2);
8410 				if (error == 0) {
8411 					/*
8412 					 * nfsm_mbufuio() advances to a multiple
8413 					 * of 4, so round up len2 as well.  Then
8414 					 * we need to advance over the rest of
8415 					 * the data, rounding up the remaining
8416 					 * length.
8417 					 */
8418 					len2 = NFSM_RNDUP(len2);
8419 					len2 = NFSM_RNDUP(len - len2);
8420 					if (len2 > 0)
8421 						error = nfsm_advance(nd, len2,
8422 						    -1);
8423 				}
8424 			}
8425 		} else if (uiop == NULL && len > 0) {
8426 			/* Just wants the length and not the data. */
8427 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8428 		} else if (len > 0)
8429 			error = ENOATTR;
8430 		if (error != 0)
8431 			goto nfsmout;
8432 		*lenp = len;
8433 		/* Just skip over Getattr op status. */
8434 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8435 		error = nfsm_loadattr(nd, nap);
8436 		if (error == 0)
8437 			*attrflagp = 1;
8438 	}
8439 	if (error == 0)
8440 		error = nd->nd_repstat;
8441 nfsmout:
8442 	m_freem(nd->nd_mrep);
8443 	return (error);
8444 }
8445 
8446 /*
8447  * The setextattr RPC.
8448  */
8449 int
8450 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8451     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8452 {
8453 	uint32_t *tl;
8454 	int error;
8455 	struct nfsrv_descript nfsd;
8456 	struct nfsrv_descript *nd = &nfsd;
8457 	nfsattrbit_t attrbits;
8458 
8459 	*attrflagp = 0;
8460 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8461 	if (uiop->uio_resid > nd->nd_maxreq) {
8462 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8463 		m_freem(nd->nd_mreq);
8464 		return (EINVAL);
8465 	}
8466 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8467 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8468 	nfsm_strtom(nd, name, strlen(name));
8469 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8470 	*tl = txdr_unsigned(uiop->uio_resid);
8471 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8472 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8473 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8474 	NFSGETATTR_ATTRBIT(&attrbits);
8475 	nfsrv_putattrbit(nd, &attrbits);
8476 	error = nfscl_request(nd, vp, p, cred, NULL);
8477 	if (error != 0)
8478 		return (error);
8479 	if (nd->nd_repstat == 0) {
8480 		/* Just skip over the reply and Getattr op status. */
8481 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8482 		    NFSX_UNSIGNED);
8483 		error = nfsm_loadattr(nd, nap);
8484 		if (error == 0)
8485 			*attrflagp = 1;
8486 	}
8487 	if (error == 0)
8488 		error = nd->nd_repstat;
8489 nfsmout:
8490 	m_freem(nd->nd_mrep);
8491 	return (error);
8492 }
8493 
8494 /*
8495  * The removeextattr RPC.
8496  */
8497 int
8498 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8499     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8500 {
8501 	uint32_t *tl;
8502 	int error;
8503 	struct nfsrv_descript nfsd;
8504 	struct nfsrv_descript *nd = &nfsd;
8505 	nfsattrbit_t attrbits;
8506 
8507 	*attrflagp = 0;
8508 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8509 	nfsm_strtom(nd, name, strlen(name));
8510 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8511 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8512 	NFSGETATTR_ATTRBIT(&attrbits);
8513 	nfsrv_putattrbit(nd, &attrbits);
8514 	error = nfscl_request(nd, vp, p, cred, NULL);
8515 	if (error != 0)
8516 		return (error);
8517 	if (nd->nd_repstat == 0) {
8518 		/* Just skip over the reply and Getattr op status. */
8519 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8520 		    NFSX_UNSIGNED);
8521 		error = nfsm_loadattr(nd, nap);
8522 		if (error == 0)
8523 			*attrflagp = 1;
8524 	}
8525 	if (error == 0)
8526 		error = nd->nd_repstat;
8527 nfsmout:
8528 	m_freem(nd->nd_mrep);
8529 	return (error);
8530 }
8531 
8532 /*
8533  * The listextattr RPC.
8534  */
8535 int
8536 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8537     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8538     struct ucred *cred, NFSPROC_T *p)
8539 {
8540 	uint32_t *tl;
8541 	int cnt, error, i, len;
8542 	struct nfsrv_descript nfsd;
8543 	struct nfsrv_descript *nd = &nfsd;
8544 	nfsattrbit_t attrbits;
8545 	u_char c;
8546 
8547 	*attrflagp = 0;
8548 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8549 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8550 	txdr_hyper(*cookiep, tl); tl += 2;
8551 	*tl++ = txdr_unsigned(*lenp);
8552 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8553 	NFSGETATTR_ATTRBIT(&attrbits);
8554 	nfsrv_putattrbit(nd, &attrbits);
8555 	error = nfscl_request(nd, vp, p, cred, NULL);
8556 	if (error != 0)
8557 		return (error);
8558 	*eofp = true;
8559 	*lenp = 0;
8560 	if (nd->nd_repstat == 0) {
8561 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8562 		*cookiep = fxdr_hyper(tl); tl += 2;
8563 		cnt = fxdr_unsigned(int, *tl);
8564 		if (cnt < 0) {
8565 			error = EBADRPC;
8566 			goto nfsmout;
8567 		}
8568 		for (i = 0; i < cnt; i++) {
8569 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8570 			len = fxdr_unsigned(int, *tl);
8571 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8572 				error = EBADRPC;
8573 				goto nfsmout;
8574 			}
8575 			if (uiop == NULL)
8576 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8577 			else if (uiop->uio_resid >= len + 1) {
8578 				c = len;
8579 				error = uiomove(&c, sizeof(c), uiop);
8580 				if (error == 0)
8581 					error = nfsm_mbufuio(nd, uiop, len);
8582 			} else {
8583 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8584 				*eofp = false;
8585 			}
8586 			if (error != 0)
8587 				goto nfsmout;
8588 			*lenp += (len + 1);
8589 		}
8590 		/* Get the eof and skip over the Getattr op status. */
8591 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8592 		/*
8593 		 * *eofp is set false above, because it wasn't able to copy
8594 		 * all of the reply.
8595 		 */
8596 		if (*eofp && *tl == 0)
8597 			*eofp = false;
8598 		error = nfsm_loadattr(nd, nap);
8599 		if (error == 0)
8600 			*attrflagp = 1;
8601 	}
8602 	if (error == 0)
8603 		error = nd->nd_repstat;
8604 nfsmout:
8605 	m_freem(nd->nd_mrep);
8606 	return (error);
8607 }
8608 
8609 /*
8610  * Split an mbuf list.  For non-M_EXTPG mbufs, just use m_split().
8611  */
8612 static struct mbuf *
8613 nfsm_split(struct mbuf *mp, uint64_t xfer)
8614 {
8615 	struct mbuf *m, *m2;
8616 	vm_page_t pg;
8617 	int i, j, left, pgno, plen, trim;
8618 	char *cp, *cp2;
8619 
8620 	if ((mp->m_flags & M_EXTPG) == 0) {
8621 		m = m_split(mp, xfer, M_WAITOK);
8622 		return (m);
8623 	}
8624 
8625 	/* Find the correct mbuf to split at. */
8626 	for (m = mp; m != NULL && xfer > m->m_len; m = m->m_next)
8627 		xfer -= m->m_len;
8628 	if (m == NULL)
8629 		return (NULL);
8630 
8631 	/* If xfer == m->m_len, we can just split the mbuf list. */
8632 	if (xfer == m->m_len) {
8633 		m2 = m->m_next;
8634 		m->m_next = NULL;
8635 		return (m2);
8636 	}
8637 
8638 	/* Find the page to split at. */
8639 	pgno = 0;
8640 	left = xfer;
8641 	do {
8642 		if (pgno == 0)
8643 			plen = m_epg_pagelen(m, 0, m->m_epg_1st_off);
8644 		else
8645 			plen = m_epg_pagelen(m, pgno, 0);
8646 		if (left <= plen)
8647 			break;
8648 		left -= plen;
8649 		pgno++;
8650 	} while (pgno < m->m_epg_npgs);
8651 	if (pgno == m->m_epg_npgs)
8652 		panic("nfsm_split: eroneous ext_pgs mbuf");
8653 
8654 	m2 = mb_alloc_ext_pgs(M_WAITOK, mb_free_mext_pgs);
8655 	m2->m_epg_flags |= EPG_FLAG_ANON;
8656 
8657 	/*
8658 	 * If left < plen, allocate a new page for the new mbuf
8659 	 * and copy the data after left in the page to this new
8660 	 * page.
8661 	 */
8662 	if (left < plen) {
8663 		do {
8664 			pg = vm_page_alloc(NULL, 0, VM_ALLOC_NORMAL |
8665 			    VM_ALLOC_NOOBJ | VM_ALLOC_NODUMP |
8666 			    VM_ALLOC_WIRED);
8667 			if (pg == NULL)
8668 				vm_wait(NULL);
8669 		} while (pg == NULL);
8670 		m2->m_epg_pa[0] = VM_PAGE_TO_PHYS(pg);
8671 		m2->m_epg_npgs = 1;
8672 
8673 		/* Copy the data after left to the new page. */
8674 		trim = plen - left;
8675 		cp = (char *)(void *)PHYS_TO_DMAP(m->m_epg_pa[pgno]);
8676 		if (pgno == 0)
8677 			cp += m->m_epg_1st_off;
8678 		cp += left;
8679 		cp2 = (char *)(void *)PHYS_TO_DMAP(m2->m_epg_pa[0]);
8680 		if (pgno == m->m_epg_npgs - 1)
8681 			m2->m_epg_last_len = trim;
8682 		else {
8683 			cp2 += PAGE_SIZE - trim;
8684 			m2->m_epg_1st_off = PAGE_SIZE - trim;
8685 			m2->m_epg_last_len = m->m_epg_last_len;
8686 		}
8687 		memcpy(cp2, cp, trim);
8688 		m2->m_len = trim;
8689 	} else {
8690 		m2->m_len = 0;
8691 		m2->m_epg_last_len = m->m_epg_last_len;
8692 	}
8693 
8694 	/* Move the pages beyond pgno to the new mbuf. */
8695 	for (i = pgno + 1, j = m2->m_epg_npgs; i < m->m_epg_npgs; i++, j++) {
8696 		m2->m_epg_pa[j] = m->m_epg_pa[i];
8697 		/* Never moves page 0. */
8698 		m2->m_len += m_epg_pagelen(m, i, 0);
8699 	}
8700 	m2->m_epg_npgs = j;
8701 	m->m_epg_npgs = pgno + 1;
8702 	m->m_epg_last_len = left;
8703 	m->m_len = xfer;
8704 
8705 	m2->m_next = m->m_next;
8706 	m->m_next = NULL;
8707 	return (m2);
8708 }
8709 
8710 /*
8711  * Do the NFSv4.1 Bind Connection to Session.
8712  * Called from the reconnect layer of the krpc (sys/rpc/clnt_rc.c).
8713  */
8714 void
8715 nfsrpc_bindconnsess(CLIENT *cl, void *arg, struct ucred *cr)
8716 {
8717 	struct nfscl_reconarg *rcp = (struct nfscl_reconarg *)arg;
8718 	uint32_t res, *tl;
8719 	struct nfsrv_descript nfsd;
8720 	struct nfsrv_descript *nd = &nfsd;
8721 	struct rpc_callextra ext;
8722 	struct timeval utimeout;
8723 	enum clnt_stat stat;
8724 	int error;
8725 
8726 	nfscl_reqstart(nd, NFSPROC_BINDCONNTOSESS, NULL, NULL, 0, NULL, NULL,
8727 	    NFS_VER4, rcp->minorvers);
8728 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID + 2 * NFSX_UNSIGNED);
8729 	memcpy(tl, rcp->sessionid, NFSX_V4SESSIONID);
8730 	tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
8731 	*tl++ = txdr_unsigned(NFSCDFC4_FORE_OR_BOTH);
8732 	*tl = newnfs_false;
8733 
8734 	memset(&ext, 0, sizeof(ext));
8735 	utimeout.tv_sec = 30;
8736 	utimeout.tv_usec = 0;
8737 	ext.rc_auth = authunix_create(cr);
8738 	nd->nd_mrep = NULL;
8739 	stat = CLNT_CALL_MBUF(cl, &ext, NFSV4PROC_COMPOUND, nd->nd_mreq,
8740 	    &nd->nd_mrep, utimeout);
8741 	AUTH_DESTROY(ext.rc_auth);
8742 	if (stat != RPC_SUCCESS) {
8743 		printf("nfsrpc_bindconnsess: call failed stat=%d\n", stat);
8744 		return;
8745 	}
8746 	if (nd->nd_mrep == NULL) {
8747 		printf("nfsrpc_bindconnsess: no reply args\n");
8748 		return;
8749 	}
8750 	error = 0;
8751 	newnfs_realign(&nd->nd_mrep, M_WAITOK);
8752 	nd->nd_md = nd->nd_mrep;
8753 	nd->nd_dpos = mtod(nd->nd_md, char *);
8754 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8755 	nd->nd_repstat = fxdr_unsigned(uint32_t, *tl++);
8756 	if (nd->nd_repstat == NFSERR_OK) {
8757 		res = fxdr_unsigned(uint32_t, *tl);
8758 		if (res > 0 && (error = nfsm_advance(nd, NFSM_RNDUP(res),
8759 		    -1)) != 0)
8760 			goto nfsmout;
8761 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
8762 		    4 * NFSX_UNSIGNED);
8763 		tl += 3;
8764 		if (!NFSBCMP(tl, rcp->sessionid, NFSX_V4SESSIONID)) {
8765 			tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
8766 			res = fxdr_unsigned(uint32_t, *tl);
8767 			if (res != NFSCDFS4_BOTH)
8768 				printf("nfsrpc_bindconnsess: did not "
8769 				    "return FS4_BOTH\n");
8770 		} else
8771 			printf("nfsrpc_bindconnsess: not same "
8772 			    "sessionid\n");
8773 	} else if (nd->nd_repstat != NFSERR_BADSESSION)
8774 		printf("nfsrpc_bindconnsess: returned %d\n", nd->nd_repstat);
8775 nfsmout:
8776 	if (error != 0)
8777 		printf("nfsrpc_bindconnsess: reply bad xdr\n");
8778 	m_freem(nd->nd_mrep);
8779 }
8780