xref: /freebsd/sys/fs/nfsclient/nfs_clrpcops.c (revision d65cd7a57bf0600b722afc770838a5d0c1c3a8e1)
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 int nfs_numnfscbd;
69 extern struct timeval nfsboottime;
70 extern u_int32_t newnfs_false, newnfs_true;
71 extern nfstype nfsv34_type[9];
72 extern int nfsrv_useacl;
73 extern char nfsv4_callbackaddr[INET6_ADDRSTRLEN];
74 extern int nfscl_debuglevel;
75 extern int nfs_pnfsiothreads;
76 extern u_long sb_max_adj;
77 extern int nfs_maxcopyrange;
78 NFSCLSTATEMUTEX;
79 int nfstest_outofseq = 0;
80 int nfscl_assumeposixlocks = 1;
81 int nfscl_enablecallb = 0;
82 short nfsv4_cbport = NFSV4_CBPORT;
83 int nfstest_openallsetattr = 0;
84 
85 #define	DIRHDSIZ	offsetof(struct dirent, d_name)
86 
87 /*
88  * nfscl_getsameserver() can return one of three values:
89  * NFSDSP_USETHISSESSION - Use this session for the DS.
90  * NFSDSP_SEQTHISSESSION - Use the nfsclds_sequence field of this dsp for new
91  *     session.
92  * NFSDSP_NOTFOUND - No matching server was found.
93  */
94 enum nfsclds_state {
95 	NFSDSP_USETHISSESSION = 0,
96 	NFSDSP_SEQTHISSESSION = 1,
97 	NFSDSP_NOTFOUND = 2,
98 };
99 
100 /*
101  * Do a write RPC on a DS data file, using this structure for the arguments,
102  * so that this function can be executed by a separate kernel process.
103  */
104 struct nfsclwritedsdorpc {
105 	int			done;
106 	int			inprog;
107 	struct task		tsk;
108 	struct vnode		*vp;
109 	int			iomode;
110 	int			must_commit;
111 	nfsv4stateid_t		*stateidp;
112 	struct nfsclds		*dsp;
113 	uint64_t		off;
114 	int			len;
115 #ifdef notyet
116 	int			advise;
117 #endif
118 	struct nfsfh		*fhp;
119 	struct mbuf		*m;
120 	int			vers;
121 	int			minorvers;
122 	struct ucred		*cred;
123 	NFSPROC_T		*p;
124 	int			err;
125 };
126 
127 static int nfsrpc_setattrrpc(vnode_t , struct vattr *, nfsv4stateid_t *,
128     struct ucred *, NFSPROC_T *, struct nfsvattr *, int *, void *);
129 static int nfsrpc_readrpc(vnode_t , struct uio *, struct ucred *,
130     nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *, void *);
131 static int nfsrpc_writerpc(vnode_t , struct uio *, int *, int *,
132     struct ucred *, nfsv4stateid_t *, NFSPROC_T *, struct nfsvattr *, int *,
133     void *);
134 static int nfsrpc_createv23(vnode_t , char *, int, struct vattr *,
135     nfsquad_t, int, struct ucred *, NFSPROC_T *, struct nfsvattr *,
136     struct nfsvattr *, struct nfsfh **, int *, int *, void *);
137 static int nfsrpc_createv4(vnode_t , char *, int, struct vattr *,
138     nfsquad_t, int, struct nfsclowner *, struct nfscldeleg **, struct ucred *,
139     NFSPROC_T *, struct nfsvattr *, struct nfsvattr *, struct nfsfh **, int *,
140     int *, void *, int *);
141 static int nfsrpc_locku(struct nfsrv_descript *, struct nfsmount *,
142     struct nfscllockowner *, u_int64_t, u_int64_t,
143     u_int32_t, struct ucred *, NFSPROC_T *, int);
144 static int nfsrpc_setaclrpc(vnode_t, struct ucred *, NFSPROC_T *,
145     struct acl *, nfsv4stateid_t *, void *);
146 static int nfsrpc_getlayout(struct nfsmount *, vnode_t, struct nfsfh *, int,
147     uint32_t *, nfsv4stateid_t *, uint64_t, struct nfscllayout **,
148     struct ucred *, NFSPROC_T *);
149 static int nfsrpc_fillsa(struct nfsmount *, struct sockaddr_in *,
150     struct sockaddr_in6 *, sa_family_t, int, int, struct nfsclds **,
151     NFSPROC_T *);
152 static void nfscl_initsessionslots(struct nfsclsession *);
153 static int nfscl_doflayoutio(vnode_t, struct uio *, int *, int *, int *,
154     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
155     struct nfsclflayout *, uint64_t, uint64_t, int, struct ucred *,
156     NFSPROC_T *);
157 static int nfscl_dofflayoutio(vnode_t, struct uio *, int *, int *, int *,
158     nfsv4stateid_t *, int, struct nfscldevinfo *, struct nfscllayout *,
159     struct nfsclflayout *, uint64_t, uint64_t, int, int, struct mbuf *,
160     struct nfsclwritedsdorpc *, struct ucred *, NFSPROC_T *);
161 static struct mbuf *nfsm_copym(struct mbuf *, int, int);
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 
224 int nfs_pnfsio(task_fn_t *, void *);
225 
226 /*
227  * nfs null call from vfs.
228  */
229 int
230 nfsrpc_null(vnode_t vp, struct ucred *cred, NFSPROC_T *p)
231 {
232 	int error;
233 	struct nfsrv_descript nfsd, *nd = &nfsd;
234 
235 	NFSCL_REQSTART(nd, NFSPROC_NULL, vp);
236 	error = nfscl_request(nd, vp, p, cred, NULL);
237 	if (nd->nd_repstat && !error)
238 		error = nd->nd_repstat;
239 	m_freem(nd->nd_mrep);
240 	return (error);
241 }
242 
243 /*
244  * nfs access rpc op.
245  * For nfs version 3 and 4, use the access rpc to check accessibility. If file
246  * modes are changed on the server, accesses might still fail later.
247  */
248 int
249 nfsrpc_access(vnode_t vp, int acmode, struct ucred *cred,
250     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp)
251 {
252 	int error;
253 	u_int32_t mode, rmode;
254 
255 	if (acmode & VREAD)
256 		mode = NFSACCESS_READ;
257 	else
258 		mode = 0;
259 	if (vnode_vtype(vp) == VDIR) {
260 		if (acmode & VWRITE)
261 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND |
262 				 NFSACCESS_DELETE);
263 		if (acmode & VEXEC)
264 			mode |= NFSACCESS_LOOKUP;
265 	} else {
266 		if (acmode & VWRITE)
267 			mode |= (NFSACCESS_MODIFY | NFSACCESS_EXTEND);
268 		if (acmode & VEXEC)
269 			mode |= NFSACCESS_EXECUTE;
270 	}
271 
272 	/*
273 	 * Now, just call nfsrpc_accessrpc() to do the actual RPC.
274 	 */
275 	error = nfsrpc_accessrpc(vp, mode, cred, p, nap, attrflagp, &rmode,
276 	    NULL);
277 
278 	/*
279 	 * The NFS V3 spec does not clarify whether or not
280 	 * the returned access bits can be a superset of
281 	 * the ones requested, so...
282 	 */
283 	if (!error && (rmode & mode) != mode)
284 		error = EACCES;
285 	return (error);
286 }
287 
288 /*
289  * The actual rpc, separated out for Darwin.
290  */
291 int
292 nfsrpc_accessrpc(vnode_t vp, u_int32_t mode, struct ucred *cred,
293     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, u_int32_t *rmodep,
294     void *stuff)
295 {
296 	u_int32_t *tl;
297 	u_int32_t supported, rmode;
298 	int error;
299 	struct nfsrv_descript nfsd, *nd = &nfsd;
300 	nfsattrbit_t attrbits;
301 
302 	*attrflagp = 0;
303 	supported = mode;
304 	NFSCL_REQSTART(nd, NFSPROC_ACCESS, vp);
305 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
306 	*tl = txdr_unsigned(mode);
307 	if (nd->nd_flag & ND_NFSV4) {
308 		/*
309 		 * And do a Getattr op.
310 		 */
311 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
312 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
313 		NFSGETATTR_ATTRBIT(&attrbits);
314 		(void) nfsrv_putattrbit(nd, &attrbits);
315 	}
316 	error = nfscl_request(nd, vp, p, cred, stuff);
317 	if (error)
318 		return (error);
319 	if (nd->nd_flag & ND_NFSV3) {
320 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
321 		if (error)
322 			goto nfsmout;
323 	}
324 	if (!nd->nd_repstat) {
325 		if (nd->nd_flag & ND_NFSV4) {
326 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
327 			supported = fxdr_unsigned(u_int32_t, *tl++);
328 		} else {
329 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
330 		}
331 		rmode = fxdr_unsigned(u_int32_t, *tl);
332 		if (nd->nd_flag & ND_NFSV4)
333 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
334 
335 		/*
336 		 * It's not obvious what should be done about
337 		 * unsupported access modes. For now, be paranoid
338 		 * and clear the unsupported ones.
339 		 */
340 		rmode &= supported;
341 		*rmodep = rmode;
342 	} else
343 		error = nd->nd_repstat;
344 nfsmout:
345 	m_freem(nd->nd_mrep);
346 	return (error);
347 }
348 
349 /*
350  * nfs open rpc
351  */
352 int
353 nfsrpc_open(vnode_t vp, int amode, struct ucred *cred, NFSPROC_T *p)
354 {
355 	struct nfsclopen *op;
356 	struct nfscldeleg *dp;
357 	struct nfsfh *nfhp;
358 	struct nfsnode *np = VTONFS(vp);
359 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
360 	u_int32_t mode, clidrev;
361 	int ret, newone, error, expireret = 0, retrycnt;
362 
363 	/*
364 	 * For NFSv4, Open Ops are only done on Regular Files.
365 	 */
366 	if (vnode_vtype(vp) != VREG)
367 		return (0);
368 	mode = 0;
369 	if (amode & FREAD)
370 		mode |= NFSV4OPEN_ACCESSREAD;
371 	if (amode & FWRITE)
372 		mode |= NFSV4OPEN_ACCESSWRITE;
373 	nfhp = np->n_fhp;
374 
375 	retrycnt = 0;
376 #ifdef notdef
377 { char name[100]; int namel;
378 namel = (np->n_v4->n4_namelen < 100) ? np->n_v4->n4_namelen : 99;
379 bcopy(NFS4NODENAME(np->n_v4), name, namel);
380 name[namel] = '\0';
381 printf("rpcopen p=0x%x name=%s",p->p_pid,name);
382 if (nfhp->nfh_len > 0) printf(" fh=0x%x\n",nfhp->nfh_fh[12]);
383 else printf(" fhl=0\n");
384 }
385 #endif
386 	do {
387 	    dp = NULL;
388 	    error = nfscl_open(vp, nfhp->nfh_fh, nfhp->nfh_len, mode, 1,
389 		cred, p, NULL, &op, &newone, &ret, 1);
390 	    if (error) {
391 		return (error);
392 	    }
393 	    if (nmp->nm_clp != NULL)
394 		clidrev = nmp->nm_clp->nfsc_clientidrev;
395 	    else
396 		clidrev = 0;
397 	    if (ret == NFSCLOPEN_DOOPEN) {
398 		if (np->n_v4 != NULL) {
399 			/*
400 			 * For the first attempt, try and get a layout, if
401 			 * pNFS is enabled for the mount.
402 			 */
403 			if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
404 			    nfs_numnfscbd == 0 ||
405 			    (np->n_flag & NNOLAYOUT) != 0 || retrycnt > 0)
406 				error = nfsrpc_openrpc(nmp, vp,
407 				    np->n_v4->n4_data,
408 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
409 				    np->n_fhp->nfh_len, mode, op,
410 				    NFS4NODENAME(np->n_v4),
411 				    np->n_v4->n4_namelen,
412 				    &dp, 0, 0x0, cred, p, 0, 0);
413 			else
414 				error = nfsrpc_getopenlayout(nmp, vp,
415 				    np->n_v4->n4_data,
416 				    np->n_v4->n4_fhlen, np->n_fhp->nfh_fh,
417 				    np->n_fhp->nfh_len, mode, op,
418 				    NFS4NODENAME(np->n_v4),
419 				    np->n_v4->n4_namelen, &dp, cred, p);
420 			if (dp != NULL) {
421 #ifdef APPLE
422 				OSBitAndAtomic((int32_t)~NDELEGMOD, (UInt32 *)&np->n_flag);
423 #else
424 				NFSLOCKNODE(np);
425 				np->n_flag &= ~NDELEGMOD;
426 				/*
427 				 * Invalidate the attribute cache, so that
428 				 * attributes that pre-date the issue of a
429 				 * delegation are not cached, since the
430 				 * cached attributes will remain valid while
431 				 * the delegation is held.
432 				 */
433 				NFSINVALATTRCACHE(np);
434 				NFSUNLOCKNODE(np);
435 #endif
436 				(void) nfscl_deleg(nmp->nm_mountp,
437 				    op->nfso_own->nfsow_clp,
438 				    nfhp->nfh_fh, nfhp->nfh_len, cred, p, &dp);
439 			}
440 		} else {
441 			error = EIO;
442 		}
443 		newnfs_copyincred(cred, &op->nfso_cred);
444 	    } else if (ret == NFSCLOPEN_SETCRED)
445 		/*
446 		 * This is a new local open on a delegation. It needs
447 		 * to have credentials so that an open can be done
448 		 * against the server during recovery.
449 		 */
450 		newnfs_copyincred(cred, &op->nfso_cred);
451 
452 	    /*
453 	     * nfso_opencnt is the count of how many VOP_OPEN()s have
454 	     * been done on this Open successfully and a VOP_CLOSE()
455 	     * is expected for each of these.
456 	     * If error is non-zero, don't increment it, since the Open
457 	     * hasn't succeeded yet.
458 	     */
459 	    if (!error)
460 		op->nfso_opencnt++;
461 	    nfscl_openrelease(nmp, op, error, newone);
462 	    if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
463 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
464 		error == NFSERR_BADSESSION) {
465 		(void) nfs_catnap(PZERO, error, "nfs_open");
466 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
467 		&& clidrev != 0) {
468 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
469 		retrycnt++;
470 	    }
471 	} while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
472 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
473 	    error == NFSERR_BADSESSION ||
474 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
475 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
476 	if (error && retrycnt >= 4)
477 		error = EIO;
478 	return (error);
479 }
480 
481 /*
482  * the actual open rpc
483  */
484 int
485 nfsrpc_openrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp, int fhlen,
486     u_int8_t *newfhp, int newfhlen, u_int32_t mode, struct nfsclopen *op,
487     u_int8_t *name, int namelen, struct nfscldeleg **dpp,
488     int reclaim, u_int32_t delegtype, struct ucred *cred, NFSPROC_T *p,
489     int syscred, int recursed)
490 {
491 	u_int32_t *tl;
492 	struct nfsrv_descript nfsd, *nd = &nfsd;
493 	struct nfscldeleg *dp, *ndp = NULL;
494 	struct nfsvattr nfsva;
495 	u_int32_t rflags, deleg;
496 	nfsattrbit_t attrbits;
497 	int error, ret, acesize, limitby;
498 	struct nfsclsession *tsep;
499 
500 	dp = *dpp;
501 	*dpp = NULL;
502 	nfscl_reqstart(nd, NFSPROC_OPEN, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
503 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
504 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
505 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
506 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
507 	tsep = nfsmnt_mdssession(nmp);
508 	*tl++ = tsep->nfsess_clientid.lval[0];
509 	*tl = tsep->nfsess_clientid.lval[1];
510 	(void) nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
511 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
512 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
513 	if (reclaim) {
514 		*tl = txdr_unsigned(NFSV4OPEN_CLAIMPREVIOUS);
515 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
516 		*tl = txdr_unsigned(delegtype);
517 	} else {
518 		if (dp != NULL) {
519 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMDELEGATECUR);
520 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
521 			if (NFSHASNFSV4N(nmp))
522 				*tl++ = 0;
523 			else
524 				*tl++ = dp->nfsdl_stateid.seqid;
525 			*tl++ = dp->nfsdl_stateid.other[0];
526 			*tl++ = dp->nfsdl_stateid.other[1];
527 			*tl = dp->nfsdl_stateid.other[2];
528 		} else {
529 			*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
530 		}
531 		(void) nfsm_strtom(nd, name, namelen);
532 	}
533 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
534 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
535 	NFSZERO_ATTRBIT(&attrbits);
536 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
537 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
538 	(void) nfsrv_putattrbit(nd, &attrbits);
539 	if (syscred)
540 		nd->nd_flag |= ND_USEGSSNAME;
541 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
542 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
543 	if (error)
544 		return (error);
545 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
546 	if (!nd->nd_repstat) {
547 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
548 		    6 * NFSX_UNSIGNED);
549 		op->nfso_stateid.seqid = *tl++;
550 		op->nfso_stateid.other[0] = *tl++;
551 		op->nfso_stateid.other[1] = *tl++;
552 		op->nfso_stateid.other[2] = *tl;
553 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
554 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
555 		if (error)
556 			goto nfsmout;
557 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
558 		deleg = fxdr_unsigned(u_int32_t, *tl);
559 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
560 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
561 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
562 			      NFSCLFLAGS_FIRSTDELEG))
563 				op->nfso_own->nfsow_clp->nfsc_flags |=
564 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
565 			ndp = malloc(
566 			    sizeof (struct nfscldeleg) + newfhlen,
567 			    M_NFSCLDELEG, M_WAITOK);
568 			LIST_INIT(&ndp->nfsdl_owner);
569 			LIST_INIT(&ndp->nfsdl_lock);
570 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
571 			ndp->nfsdl_fhlen = newfhlen;
572 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
573 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
574 			nfscl_lockinit(&ndp->nfsdl_rwlock);
575 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
576 			    NFSX_UNSIGNED);
577 			ndp->nfsdl_stateid.seqid = *tl++;
578 			ndp->nfsdl_stateid.other[0] = *tl++;
579 			ndp->nfsdl_stateid.other[1] = *tl++;
580 			ndp->nfsdl_stateid.other[2] = *tl++;
581 			ret = fxdr_unsigned(int, *tl);
582 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
583 				ndp->nfsdl_flags = NFSCLDL_WRITE;
584 				/*
585 				 * Indicates how much the file can grow.
586 				 */
587 				NFSM_DISSECT(tl, u_int32_t *,
588 				    3 * NFSX_UNSIGNED);
589 				limitby = fxdr_unsigned(int, *tl++);
590 				switch (limitby) {
591 				case NFSV4OPEN_LIMITSIZE:
592 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
593 					break;
594 				case NFSV4OPEN_LIMITBLOCKS:
595 					ndp->nfsdl_sizelimit =
596 					    fxdr_unsigned(u_int64_t, *tl++);
597 					ndp->nfsdl_sizelimit *=
598 					    fxdr_unsigned(u_int64_t, *tl);
599 					break;
600 				default:
601 					error = NFSERR_BADXDR;
602 					goto nfsmout;
603 				}
604 			} else {
605 				ndp->nfsdl_flags = NFSCLDL_READ;
606 			}
607 			if (ret)
608 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
609 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
610 			    &acesize, p);
611 			if (error)
612 				goto nfsmout;
613 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
614 			error = NFSERR_BADXDR;
615 			goto nfsmout;
616 		}
617 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
618 		error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
619 		    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
620 		    NULL, NULL, NULL, p, cred);
621 		if (error)
622 			goto nfsmout;
623 		if (ndp != NULL) {
624 			ndp->nfsdl_change = nfsva.na_filerev;
625 			ndp->nfsdl_modtime = nfsva.na_mtime;
626 			ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
627 		}
628 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM)) {
629 		    do {
630 			ret = nfsrpc_openconfirm(vp, newfhp, newfhlen, op,
631 			    cred, p);
632 			if (ret == NFSERR_DELAY)
633 			    (void) nfs_catnap(PZERO, ret, "nfs_open");
634 		    } while (ret == NFSERR_DELAY);
635 		    error = ret;
636 		}
637 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) ||
638 		    nfscl_assumeposixlocks)
639 		    op->nfso_posixlock = 1;
640 		else
641 		    op->nfso_posixlock = 0;
642 
643 		/*
644 		 * If the server is handing out delegations, but we didn't
645 		 * get one because an OpenConfirm was required, try the
646 		 * Open again, to get a delegation. This is a harmless no-op,
647 		 * from a server's point of view.
648 		 */
649 		if (!reclaim && (rflags & NFSV4OPEN_RESULTCONFIRM) &&
650 		    (op->nfso_own->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG)
651 		    && !error && dp == NULL && ndp == NULL && !recursed) {
652 		    do {
653 			ret = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp,
654 			    newfhlen, mode, op, name, namelen, &ndp, 0, 0x0,
655 			    cred, p, syscred, 1);
656 			if (ret == NFSERR_DELAY)
657 			    (void) nfs_catnap(PZERO, ret, "nfs_open2");
658 		    } while (ret == NFSERR_DELAY);
659 		    if (ret) {
660 			if (ndp != NULL) {
661 				free(ndp, M_NFSCLDELEG);
662 				ndp = NULL;
663 			}
664 			if (ret == NFSERR_STALECLIENTID ||
665 			    ret == NFSERR_STALEDONTRECOVER ||
666 			    ret == NFSERR_BADSESSION)
667 				error = ret;
668 		    }
669 		}
670 	}
671 	if (nd->nd_repstat != 0 && error == 0)
672 		error = nd->nd_repstat;
673 	if (error == NFSERR_STALECLIENTID)
674 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
675 nfsmout:
676 	if (!error)
677 		*dpp = ndp;
678 	else if (ndp != NULL)
679 		free(ndp, M_NFSCLDELEG);
680 	m_freem(nd->nd_mrep);
681 	return (error);
682 }
683 
684 /*
685  * open downgrade rpc
686  */
687 int
688 nfsrpc_opendowngrade(vnode_t vp, u_int32_t mode, struct nfsclopen *op,
689     struct ucred *cred, NFSPROC_T *p)
690 {
691 	u_int32_t *tl;
692 	struct nfsrv_descript nfsd, *nd = &nfsd;
693 	int error;
694 
695 	NFSCL_REQSTART(nd, NFSPROC_OPENDOWNGRADE, vp);
696 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 3 * NFSX_UNSIGNED);
697 	if (NFSHASNFSV4N(VFSTONFS(vnode_mount(vp))))
698 		*tl++ = 0;
699 	else
700 		*tl++ = op->nfso_stateid.seqid;
701 	*tl++ = op->nfso_stateid.other[0];
702 	*tl++ = op->nfso_stateid.other[1];
703 	*tl++ = op->nfso_stateid.other[2];
704 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
705 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
706 	*tl = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
707 	error = nfscl_request(nd, vp, p, cred, NULL);
708 	if (error)
709 		return (error);
710 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
711 	if (!nd->nd_repstat) {
712 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
713 		op->nfso_stateid.seqid = *tl++;
714 		op->nfso_stateid.other[0] = *tl++;
715 		op->nfso_stateid.other[1] = *tl++;
716 		op->nfso_stateid.other[2] = *tl;
717 	}
718 	if (nd->nd_repstat && error == 0)
719 		error = nd->nd_repstat;
720 	if (error == NFSERR_STALESTATEID)
721 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
722 nfsmout:
723 	m_freem(nd->nd_mrep);
724 	return (error);
725 }
726 
727 /*
728  * V4 Close operation.
729  */
730 int
731 nfsrpc_close(vnode_t vp, int doclose, NFSPROC_T *p)
732 {
733 	struct nfsclclient *clp;
734 	int error;
735 
736 	if (vnode_vtype(vp) != VREG)
737 		return (0);
738 	if (doclose)
739 		error = nfscl_doclose(vp, &clp, p);
740 	else
741 		error = nfscl_getclose(vp, &clp);
742 	if (error)
743 		return (error);
744 
745 	nfscl_clientrelease(clp);
746 	return (0);
747 }
748 
749 /*
750  * Close the open.
751  */
752 void
753 nfsrpc_doclose(struct nfsmount *nmp, struct nfsclopen *op, NFSPROC_T *p)
754 {
755 	struct nfsrv_descript nfsd, *nd = &nfsd;
756 	struct nfscllockowner *lp, *nlp;
757 	struct nfscllock *lop, *nlop;
758 	struct ucred *tcred;
759 	u_int64_t off = 0, len = 0;
760 	u_int32_t type = NFSV4LOCKT_READ;
761 	int error, do_unlock, trycnt;
762 
763 	tcred = newnfs_getcred();
764 	newnfs_copycred(&op->nfso_cred, tcred);
765 	/*
766 	 * (Theoretically this could be done in the same
767 	 *  compound as the close, but having multiple
768 	 *  sequenced Ops in the same compound might be
769 	 *  too scary for some servers.)
770 	 */
771 	if (op->nfso_posixlock) {
772 		off = 0;
773 		len = NFS64BITSSET;
774 		type = NFSV4LOCKT_READ;
775 	}
776 
777 	/*
778 	 * Since this function is only called from VOP_INACTIVE(), no
779 	 * other thread will be manipulating this Open. As such, the
780 	 * lock lists are not being changed by other threads, so it should
781 	 * be safe to do this without locking.
782 	 */
783 	LIST_FOREACH(lp, &op->nfso_lock, nfsl_list) {
784 		do_unlock = 1;
785 		LIST_FOREACH_SAFE(lop, &lp->nfsl_lock, nfslo_list, nlop) {
786 			if (op->nfso_posixlock == 0) {
787 				off = lop->nfslo_first;
788 				len = lop->nfslo_end - lop->nfslo_first;
789 				if (lop->nfslo_type == F_WRLCK)
790 					type = NFSV4LOCKT_WRITE;
791 				else
792 					type = NFSV4LOCKT_READ;
793 			}
794 			if (do_unlock) {
795 				trycnt = 0;
796 				do {
797 					error = nfsrpc_locku(nd, nmp, lp, off,
798 					    len, type, tcred, p, 0);
799 					if ((nd->nd_repstat == NFSERR_GRACE ||
800 					    nd->nd_repstat == NFSERR_DELAY) &&
801 					    error == 0)
802 						(void) nfs_catnap(PZERO,
803 						    (int)nd->nd_repstat,
804 						    "nfs_close");
805 				} while ((nd->nd_repstat == NFSERR_GRACE ||
806 				    nd->nd_repstat == NFSERR_DELAY) &&
807 				    error == 0 && trycnt++ < 5);
808 				if (op->nfso_posixlock)
809 					do_unlock = 0;
810 			}
811 			nfscl_freelock(lop, 0);
812 		}
813 		/*
814 		 * Do a ReleaseLockOwner.
815 		 * The lock owner name nfsl_owner may be used by other opens for
816 		 * other files but the lock_owner4 name that nfsrpc_rellockown()
817 		 * puts on the wire has the file handle for this file appended
818 		 * to it, so it can be done now.
819 		 */
820 		(void)nfsrpc_rellockown(nmp, lp, lp->nfsl_open->nfso_fh,
821 		    lp->nfsl_open->nfso_fhlen, tcred, p);
822 	}
823 
824 	/*
825 	 * There could be other Opens for different files on the same
826 	 * OpenOwner, so locking is required.
827 	 */
828 	NFSLOCKCLSTATE();
829 	nfscl_lockexcl(&op->nfso_own->nfsow_rwlock, NFSCLSTATEMUTEXPTR);
830 	NFSUNLOCKCLSTATE();
831 	do {
832 		error = nfscl_tryclose(op, tcred, nmp, p);
833 		if (error == NFSERR_GRACE)
834 			(void) nfs_catnap(PZERO, error, "nfs_close");
835 	} while (error == NFSERR_GRACE);
836 	NFSLOCKCLSTATE();
837 	nfscl_lockunlock(&op->nfso_own->nfsow_rwlock);
838 
839 	LIST_FOREACH_SAFE(lp, &op->nfso_lock, nfsl_list, nlp)
840 		nfscl_freelockowner(lp, 0);
841 	nfscl_freeopen(op, 0);
842 	NFSUNLOCKCLSTATE();
843 	NFSFREECRED(tcred);
844 }
845 
846 /*
847  * The actual Close RPC.
848  */
849 int
850 nfsrpc_closerpc(struct nfsrv_descript *nd, struct nfsmount *nmp,
851     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p,
852     int syscred)
853 {
854 	u_int32_t *tl;
855 	int error;
856 
857 	nfscl_reqstart(nd, NFSPROC_CLOSE, nmp, op->nfso_fh,
858 	    op->nfso_fhlen, NULL, NULL, 0, 0);
859 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
860 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
861 	if (NFSHASNFSV4N(nmp))
862 		*tl++ = 0;
863 	else
864 		*tl++ = op->nfso_stateid.seqid;
865 	*tl++ = op->nfso_stateid.other[0];
866 	*tl++ = op->nfso_stateid.other[1];
867 	*tl = op->nfso_stateid.other[2];
868 	if (syscred)
869 		nd->nd_flag |= ND_USEGSSNAME;
870 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
871 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
872 	if (error)
873 		return (error);
874 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
875 	if (nd->nd_repstat == 0)
876 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
877 	error = nd->nd_repstat;
878 	if (error == NFSERR_STALESTATEID)
879 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
880 nfsmout:
881 	m_freem(nd->nd_mrep);
882 	return (error);
883 }
884 
885 /*
886  * V4 Open Confirm RPC.
887  */
888 int
889 nfsrpc_openconfirm(vnode_t vp, u_int8_t *nfhp, int fhlen,
890     struct nfsclopen *op, struct ucred *cred, NFSPROC_T *p)
891 {
892 	u_int32_t *tl;
893 	struct nfsrv_descript nfsd, *nd = &nfsd;
894 	struct nfsmount *nmp;
895 	int error;
896 
897 	nmp = VFSTONFS(vnode_mount(vp));
898 	if (NFSHASNFSV4N(nmp))
899 		return (0);		/* No confirmation for NFSv4.1. */
900 	nfscl_reqstart(nd, NFSPROC_OPENCONFIRM, nmp, nfhp, fhlen, NULL, NULL,
901 	    0, 0);
902 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_STATEID);
903 	*tl++ = op->nfso_stateid.seqid;
904 	*tl++ = op->nfso_stateid.other[0];
905 	*tl++ = op->nfso_stateid.other[1];
906 	*tl++ = op->nfso_stateid.other[2];
907 	*tl = txdr_unsigned(op->nfso_own->nfsow_seqid);
908 	error = nfscl_request(nd, vp, p, cred, NULL);
909 	if (error)
910 		return (error);
911 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
912 	if (!nd->nd_repstat) {
913 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
914 		op->nfso_stateid.seqid = *tl++;
915 		op->nfso_stateid.other[0] = *tl++;
916 		op->nfso_stateid.other[1] = *tl++;
917 		op->nfso_stateid.other[2] = *tl;
918 	}
919 	error = nd->nd_repstat;
920 	if (error == NFSERR_STALESTATEID)
921 		nfscl_initiate_recovery(op->nfso_own->nfsow_clp);
922 nfsmout:
923 	m_freem(nd->nd_mrep);
924 	return (error);
925 }
926 
927 /*
928  * Do the setclientid and setclientid confirm RPCs. Called from nfs_statfs()
929  * when a mount has just occurred and when the server replies NFSERR_EXPIRED.
930  */
931 int
932 nfsrpc_setclient(struct nfsmount *nmp, struct nfsclclient *clp, int reclaim,
933     bool *retokp, struct ucred *cred, NFSPROC_T *p)
934 {
935 	u_int32_t *tl;
936 	struct nfsrv_descript nfsd;
937 	struct nfsrv_descript *nd = &nfsd;
938 	nfsattrbit_t attrbits;
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 	u_int32_t lease;
944 	static u_int32_t rev = 0;
945 	struct nfsclds *dsp, *odsp;
946 	struct in6_addr a6;
947 	struct nfsclsession *tsep;
948 
949 	if (nfsboottime.tv_sec == 0)
950 		NFSSETBOOTTIME(nfsboottime);
951 	if (NFSHASNFSV4N(nmp)) {
952 		error = NFSERR_BADSESSION;
953 		odsp = dsp = NULL;
954 		if (retokp != NULL) {
955 			NFSLOCKMNT(nmp);
956 			odsp = TAILQ_FIRST(&nmp->nm_sess);
957 			NFSUNLOCKMNT(nmp);
958 		}
959 		if (odsp != NULL) {
960 			/*
961 			 * When a session already exists, first try a
962 			 * CreateSession with the extant ClientID.
963 			 */
964 			dsp = malloc(sizeof(struct nfsclds) +
965 			    odsp->nfsclds_servownlen + 1, M_NFSCLDS,
966 			    M_WAITOK | M_ZERO);
967 			dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
968 			dsp->nfsclds_servownlen = odsp->nfsclds_servownlen;
969 			dsp->nfsclds_sess.nfsess_clientid =
970 			    odsp->nfsclds_sess.nfsess_clientid;
971 			dsp->nfsclds_sess.nfsess_sequenceid =
972 			    odsp->nfsclds_sess.nfsess_sequenceid;
973 			dsp->nfsclds_flags = odsp->nfsclds_flags;
974 			if (dsp->nfsclds_servownlen > 0)
975 				memcpy(dsp->nfsclds_serverown,
976 				    odsp->nfsclds_serverown,
977 				    dsp->nfsclds_servownlen + 1);
978 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
979 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
980 			    NULL, MTX_DEF);
981 			nfscl_initsessionslots(&dsp->nfsclds_sess);
982 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
983 			    &nmp->nm_sockreq, NULL,
984 			    dsp->nfsclds_sess.nfsess_sequenceid, 1, cred, p);
985 			NFSCL_DEBUG(1, "create session for extant "
986 			    "ClientID=%d\n", error);
987 			if (error != 0) {
988 				nfscl_freenfsclds(dsp);
989 				dsp = NULL;
990 				/*
991 				 * If *retokp is true, return any error other
992 				 * than NFSERR_STALECLIENTID,
993 				 * NFSERR_BADSESSION or NFSERR_STALEDONTRECOVER
994 				 * so that nfscl_recover() will not loop.
995 				 */
996 				if (*retokp)
997 					return (NFSERR_IO);
998 			} else
999 				*retokp = true;
1000 		} else if (retokp != NULL && *retokp)
1001 			return (NFSERR_IO);
1002 		if (error != 0) {
1003 			/*
1004 			 * Either there was no previous session or the
1005 			 * CreateSession attempt failed, so...
1006 			 * do an ExchangeID followed by the CreateSession.
1007 			 */
1008 			clp->nfsc_rev = rev++;
1009 			error = nfsrpc_exchangeid(nmp, clp, &nmp->nm_sockreq, 0,
1010 			    NFSV4EXCH_USEPNFSMDS | NFSV4EXCH_USENONPNFS, &dsp,
1011 			    cred, p);
1012 			NFSCL_DEBUG(1, "aft exch=%d\n", error);
1013 			if (error == 0)
1014 				error = nfsrpc_createsession(nmp,
1015 				    &dsp->nfsclds_sess, &nmp->nm_sockreq, NULL,
1016 				    dsp->nfsclds_sess.nfsess_sequenceid, 1,
1017 				    cred, p);
1018 			NFSCL_DEBUG(1, "aft createsess=%d\n", error);
1019 		}
1020 		if (error == 0) {
1021 			NFSLOCKMNT(nmp);
1022 			/*
1023 			 * The old sessions cannot be safely free'd
1024 			 * here, since they may still be used by
1025 			 * in-progress RPCs.
1026 			 */
1027 			tsep = NULL;
1028 			if (TAILQ_FIRST(&nmp->nm_sess) != NULL)
1029 				tsep = NFSMNT_MDSSESSION(nmp);
1030 			TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp,
1031 			    nfsclds_list);
1032 			/*
1033 			 * Wake up RPCs waiting for a slot on the
1034 			 * old session. These will then fail with
1035 			 * NFSERR_BADSESSION and be retried with the
1036 			 * new session by nfsv4_setsequence().
1037 			 * Also wakeup() processes waiting for the
1038 			 * new session.
1039 			 */
1040 			if (tsep != NULL)
1041 				wakeup(&tsep->nfsess_slots);
1042 			wakeup(&nmp->nm_sess);
1043 			NFSUNLOCKMNT(nmp);
1044 		} else if (dsp != NULL)
1045 			nfscl_freenfsclds(dsp);
1046 		if (error == 0 && reclaim == 0) {
1047 			error = nfsrpc_reclaimcomplete(nmp, cred, p);
1048 			NFSCL_DEBUG(1, "aft reclaimcomp=%d\n", error);
1049 			if (error == NFSERR_COMPLETEALREADY ||
1050 			    error == NFSERR_NOTSUPP)
1051 				/* Ignore this error. */
1052 				error = 0;
1053 		}
1054 		return (error);
1055 	} else if (retokp != NULL && *retokp)
1056 		return (NFSERR_IO);
1057 	clp->nfsc_rev = rev++;
1058 
1059 	/*
1060 	 * Allocate a single session structure for NFSv4.0, because some of
1061 	 * the fields are used by NFSv4.0 although it doesn't do a session.
1062 	 */
1063 	dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS, M_WAITOK | M_ZERO);
1064 	mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
1065 	mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession", NULL, MTX_DEF);
1066 	NFSLOCKMNT(nmp);
1067 	TAILQ_INSERT_HEAD(&nmp->nm_sess, dsp, nfsclds_list);
1068 	tsep = NFSMNT_MDSSESSION(nmp);
1069 	NFSUNLOCKMNT(nmp);
1070 
1071 	nfscl_reqstart(nd, NFSPROC_SETCLIENTID, nmp, NULL, 0, NULL, NULL, 0, 0);
1072 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1073 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);
1074 	*tl = txdr_unsigned(clp->nfsc_rev);
1075 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
1076 
1077 	/*
1078 	 * set up the callback address
1079 	 */
1080 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1081 	*tl = txdr_unsigned(NFS_CALLBCKPROG);
1082 	callblen = strlen(nfsv4_callbackaddr);
1083 	if (callblen == 0)
1084 		cp = nfscl_getmyip(nmp, &a6, &isinet6);
1085 	if (nfscl_enablecallb && nfs_numnfscbd > 0 &&
1086 	    (callblen > 0 || cp != NULL)) {
1087 		port = htons(nfsv4_cbport);
1088 		cp2 = (u_int8_t *)&port;
1089 #ifdef INET6
1090 		if ((callblen > 0 &&
1091 		     strchr(nfsv4_callbackaddr, ':')) || isinet6) {
1092 			char ip6buf[INET6_ADDRSTRLEN], *ip6add;
1093 
1094 			(void) nfsm_strtom(nd, "tcp6", 4);
1095 			if (callblen == 0) {
1096 				ip6_sprintf(ip6buf, (struct in6_addr *)cp);
1097 				ip6add = ip6buf;
1098 			} else {
1099 				ip6add = nfsv4_callbackaddr;
1100 			}
1101 			snprintf(addr, INET6_ADDRSTRLEN + 9, "%s.%d.%d",
1102 			    ip6add, cp2[0], cp2[1]);
1103 		} else
1104 #endif
1105 		{
1106 			(void) nfsm_strtom(nd, "tcp", 3);
1107 			if (callblen == 0)
1108 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1109 				    "%d.%d.%d.%d.%d.%d", cp[0], cp[1],
1110 				    cp[2], cp[3], cp2[0], cp2[1]);
1111 			else
1112 				snprintf(addr, INET6_ADDRSTRLEN + 9,
1113 				    "%s.%d.%d", nfsv4_callbackaddr,
1114 				    cp2[0], cp2[1]);
1115 		}
1116 		(void) nfsm_strtom(nd, addr, strlen(addr));
1117 	} else {
1118 		(void) nfsm_strtom(nd, "tcp", 3);
1119 		(void) nfsm_strtom(nd, "0.0.0.0.0.0", 11);
1120 	}
1121 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1122 	*tl = txdr_unsigned(clp->nfsc_cbident);
1123 	nd->nd_flag |= ND_USEGSSNAME;
1124 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1125 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1126 	if (error)
1127 		return (error);
1128 	if (nd->nd_repstat == 0) {
1129 	    NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1130 	    tsep->nfsess_clientid.lval[0] = *tl++;
1131 	    tsep->nfsess_clientid.lval[1] = *tl++;
1132 	    confirm.lval[0] = *tl++;
1133 	    confirm.lval[1] = *tl;
1134 	    m_freem(nd->nd_mrep);
1135 	    nd->nd_mrep = NULL;
1136 
1137 	    /*
1138 	     * and confirm it.
1139 	     */
1140 	    nfscl_reqstart(nd, NFSPROC_SETCLIENTIDCFRM, nmp, NULL, 0, NULL,
1141 		NULL, 0, 0);
1142 	    NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1143 	    *tl++ = tsep->nfsess_clientid.lval[0];
1144 	    *tl++ = tsep->nfsess_clientid.lval[1];
1145 	    *tl++ = confirm.lval[0];
1146 	    *tl = confirm.lval[1];
1147 	    nd->nd_flag |= ND_USEGSSNAME;
1148 	    error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1149 		cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1150 	    if (error)
1151 		return (error);
1152 	    m_freem(nd->nd_mrep);
1153 	    nd->nd_mrep = NULL;
1154 	    if (nd->nd_repstat == 0) {
1155 		nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, nmp->nm_fh,
1156 		    nmp->nm_fhsize, NULL, NULL, 0, 0);
1157 		NFSZERO_ATTRBIT(&attrbits);
1158 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1159 		(void) nfsrv_putattrbit(nd, &attrbits);
1160 		nd->nd_flag |= ND_USEGSSNAME;
1161 		error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p,
1162 		    cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
1163 		if (error)
1164 		    return (error);
1165 		if (nd->nd_repstat == 0) {
1166 		    error = nfsv4_loadattr(nd, NULL, NULL, NULL, NULL, 0, NULL,
1167 			NULL, NULL, NULL, NULL, 0, NULL, &lease, NULL, p, cred);
1168 		    if (error)
1169 			goto nfsmout;
1170 		    clp->nfsc_renew = NFSCL_RENEW(lease);
1171 		    clp->nfsc_expire = NFSD_MONOSEC + clp->nfsc_renew;
1172 		    clp->nfsc_clientidrev++;
1173 		    if (clp->nfsc_clientidrev == 0)
1174 			clp->nfsc_clientidrev++;
1175 		}
1176 	    }
1177 	}
1178 	error = nd->nd_repstat;
1179 nfsmout:
1180 	m_freem(nd->nd_mrep);
1181 	return (error);
1182 }
1183 
1184 /*
1185  * nfs getattr call.
1186  */
1187 int
1188 nfsrpc_getattr(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
1189     struct nfsvattr *nap, void *stuff)
1190 {
1191 	struct nfsrv_descript nfsd, *nd = &nfsd;
1192 	int error;
1193 	nfsattrbit_t attrbits;
1194 
1195 	NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
1196 	if (nd->nd_flag & ND_NFSV4) {
1197 		NFSGETATTR_ATTRBIT(&attrbits);
1198 		(void) nfsrv_putattrbit(nd, &attrbits);
1199 	}
1200 	error = nfscl_request(nd, vp, p, cred, stuff);
1201 	if (error)
1202 		return (error);
1203 	if (!nd->nd_repstat)
1204 		error = nfsm_loadattr(nd, nap);
1205 	else
1206 		error = nd->nd_repstat;
1207 	m_freem(nd->nd_mrep);
1208 	return (error);
1209 }
1210 
1211 /*
1212  * nfs getattr call with non-vnode arguemnts.
1213  */
1214 int
1215 nfsrpc_getattrnovp(struct nfsmount *nmp, u_int8_t *fhp, int fhlen, int syscred,
1216     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, u_int64_t *xidp,
1217     uint32_t *leasep)
1218 {
1219 	struct nfsrv_descript nfsd, *nd = &nfsd;
1220 	int error, vers = NFS_VER2;
1221 	nfsattrbit_t attrbits;
1222 
1223 	nfscl_reqstart(nd, NFSPROC_GETATTR, nmp, fhp, fhlen, NULL, NULL, 0, 0);
1224 	if (nd->nd_flag & ND_NFSV4) {
1225 		vers = NFS_VER4;
1226 		NFSGETATTR_ATTRBIT(&attrbits);
1227 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_LEASETIME);
1228 		(void) nfsrv_putattrbit(nd, &attrbits);
1229 	} else if (nd->nd_flag & ND_NFSV3) {
1230 		vers = NFS_VER3;
1231 	}
1232 	if (syscred)
1233 		nd->nd_flag |= ND_USEGSSNAME;
1234 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
1235 	    NFS_PROG, vers, NULL, 1, xidp, NULL);
1236 	if (error)
1237 		return (error);
1238 	if (nd->nd_repstat == 0) {
1239 		if ((nd->nd_flag & ND_NFSV4) != 0)
1240 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
1241 			    NULL, NULL, NULL, NULL, NULL, 0, NULL, leasep, NULL,
1242 			    NULL, NULL);
1243 		else
1244 			error = nfsm_loadattr(nd, nap);
1245 	} else
1246 		error = nd->nd_repstat;
1247 	m_freem(nd->nd_mrep);
1248 	return (error);
1249 }
1250 
1251 /*
1252  * Do an nfs setattr operation.
1253  */
1254 int
1255 nfsrpc_setattr(vnode_t vp, struct vattr *vap, NFSACL_T *aclp,
1256     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *rnap, int *attrflagp,
1257     void *stuff)
1258 {
1259 	int error, expireret = 0, openerr, retrycnt;
1260 	u_int32_t clidrev = 0, mode;
1261 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1262 	struct nfsfh *nfhp;
1263 	nfsv4stateid_t stateid;
1264 	void *lckp;
1265 
1266 	if (nmp->nm_clp != NULL)
1267 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1268 	if (vap != NULL && NFSATTRISSET(u_quad_t, vap, va_size))
1269 		mode = NFSV4OPEN_ACCESSWRITE;
1270 	else
1271 		mode = NFSV4OPEN_ACCESSREAD;
1272 	retrycnt = 0;
1273 	do {
1274 		lckp = NULL;
1275 		openerr = 1;
1276 		if (NFSHASNFSV4(nmp)) {
1277 			nfhp = VTONFS(vp)->n_fhp;
1278 			error = nfscl_getstateid(vp, nfhp->nfh_fh,
1279 			    nfhp->nfh_len, mode, 0, cred, p, &stateid, &lckp);
1280 			if (error && vnode_vtype(vp) == VREG &&
1281 			    (mode == NFSV4OPEN_ACCESSWRITE ||
1282 			     nfstest_openallsetattr)) {
1283 				/*
1284 				 * No Open stateid, so try and open the file
1285 				 * now.
1286 				 */
1287 				if (mode == NFSV4OPEN_ACCESSWRITE)
1288 					openerr = nfsrpc_open(vp, FWRITE, cred,
1289 					    p);
1290 				else
1291 					openerr = nfsrpc_open(vp, FREAD, cred,
1292 					    p);
1293 				if (!openerr)
1294 					(void) nfscl_getstateid(vp,
1295 					    nfhp->nfh_fh, nfhp->nfh_len,
1296 					    mode, 0, cred, p, &stateid, &lckp);
1297 			}
1298 		}
1299 		if (vap != NULL)
1300 			error = nfsrpc_setattrrpc(vp, vap, &stateid, cred, p,
1301 			    rnap, attrflagp, stuff);
1302 		else
1303 			error = nfsrpc_setaclrpc(vp, cred, p, aclp, &stateid,
1304 			    stuff);
1305 		if (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD) {
1306 			NFSLOCKMNT(nmp);
1307 			nmp->nm_state |= NFSSTA_OPENMODE;
1308 			NFSUNLOCKMNT(nmp);
1309 		}
1310 		if (error == NFSERR_STALESTATEID)
1311 			nfscl_initiate_recovery(nmp->nm_clp);
1312 		if (lckp != NULL)
1313 			nfscl_lockderef(lckp);
1314 		if (!openerr)
1315 			(void) nfsrpc_close(vp, 0, p);
1316 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1317 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1318 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1319 			(void) nfs_catnap(PZERO, error, "nfs_setattr");
1320 		} else if ((error == NFSERR_EXPIRED ||
1321 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1322 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1323 		}
1324 		retrycnt++;
1325 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1326 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1327 	    error == NFSERR_BADSESSION ||
1328 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1329 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1330 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1331 	    (error == NFSERR_OPENMODE && mode == NFSV4OPEN_ACCESSREAD &&
1332 	     retrycnt < 4));
1333 	if (error && retrycnt >= 4)
1334 		error = EIO;
1335 	return (error);
1336 }
1337 
1338 static int
1339 nfsrpc_setattrrpc(vnode_t vp, struct vattr *vap,
1340     nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
1341     struct nfsvattr *rnap, int *attrflagp, void *stuff)
1342 {
1343 	u_int32_t *tl;
1344 	struct nfsrv_descript nfsd, *nd = &nfsd;
1345 	int error;
1346 	nfsattrbit_t attrbits;
1347 
1348 	*attrflagp = 0;
1349 	NFSCL_REQSTART(nd, NFSPROC_SETATTR, vp);
1350 	if (nd->nd_flag & ND_NFSV4)
1351 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1352 	vap->va_type = vnode_vtype(vp);
1353 	nfscl_fillsattr(nd, vap, vp, NFSSATTR_FULL, 0);
1354 	if (nd->nd_flag & ND_NFSV3) {
1355 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1356 		*tl = newnfs_false;
1357 	} else if (nd->nd_flag & ND_NFSV4) {
1358 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1359 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1360 		NFSGETATTR_ATTRBIT(&attrbits);
1361 		(void) nfsrv_putattrbit(nd, &attrbits);
1362 	}
1363 	error = nfscl_request(nd, vp, p, cred, stuff);
1364 	if (error)
1365 		return (error);
1366 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1367 		error = nfscl_wcc_data(nd, vp, rnap, attrflagp, NULL, stuff);
1368 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 && !error)
1369 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
1370 	if (!(nd->nd_flag & ND_NFSV3) && !nd->nd_repstat && !error)
1371 		error = nfscl_postop_attr(nd, rnap, attrflagp, stuff);
1372 	m_freem(nd->nd_mrep);
1373 	if (nd->nd_repstat && !error)
1374 		error = nd->nd_repstat;
1375 	return (error);
1376 }
1377 
1378 /*
1379  * nfs lookup rpc
1380  */
1381 int
1382 nfsrpc_lookup(vnode_t dvp, char *name, int len, struct ucred *cred,
1383     NFSPROC_T *p, struct nfsvattr *dnap, struct nfsvattr *nap,
1384     struct nfsfh **nfhpp, int *attrflagp, int *dattrflagp, void *stuff)
1385 {
1386 	u_int32_t *tl;
1387 	struct nfsrv_descript nfsd, *nd = &nfsd;
1388 	struct nfsmount *nmp;
1389 	struct nfsnode *np;
1390 	struct nfsfh *nfhp;
1391 	nfsattrbit_t attrbits;
1392 	int error = 0, lookupp = 0;
1393 
1394 	*attrflagp = 0;
1395 	*dattrflagp = 0;
1396 	if (vnode_vtype(dvp) != VDIR)
1397 		return (ENOTDIR);
1398 	nmp = VFSTONFS(vnode_mount(dvp));
1399 	if (len > NFS_MAXNAMLEN)
1400 		return (ENAMETOOLONG);
1401 	if (NFSHASNFSV4(nmp) && len == 1 &&
1402 		name[0] == '.') {
1403 		/*
1404 		 * Just return the current dir's fh.
1405 		 */
1406 		np = VTONFS(dvp);
1407 		nfhp = malloc(sizeof (struct nfsfh) +
1408 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1409 		nfhp->nfh_len = np->n_fhp->nfh_len;
1410 		NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1411 		*nfhpp = nfhp;
1412 		return (0);
1413 	}
1414 	if (NFSHASNFSV4(nmp) && len == 2 &&
1415 		name[0] == '.' && name[1] == '.') {
1416 		lookupp = 1;
1417 		NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, dvp);
1418 	} else {
1419 		NFSCL_REQSTART(nd, NFSPROC_LOOKUP, dvp);
1420 		(void) nfsm_strtom(nd, name, len);
1421 	}
1422 	if (nd->nd_flag & ND_NFSV4) {
1423 		NFSGETATTR_ATTRBIT(&attrbits);
1424 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1425 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
1426 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1427 		(void) nfsrv_putattrbit(nd, &attrbits);
1428 	}
1429 	error = nfscl_request(nd, dvp, p, cred, stuff);
1430 	if (error)
1431 		return (error);
1432 	if (nd->nd_repstat) {
1433 		/*
1434 		 * When an NFSv4 Lookupp returns ENOENT, it means that
1435 		 * the lookup is at the root of an fs, so return this dir.
1436 		 */
1437 		if (nd->nd_repstat == NFSERR_NOENT && lookupp) {
1438 		    np = VTONFS(dvp);
1439 		    nfhp = malloc(sizeof (struct nfsfh) +
1440 			np->n_fhp->nfh_len, M_NFSFH, M_WAITOK);
1441 		    nfhp->nfh_len = np->n_fhp->nfh_len;
1442 		    NFSBCOPY(np->n_fhp->nfh_fh, nfhp->nfh_fh, nfhp->nfh_len);
1443 		    *nfhpp = nfhp;
1444 		    m_freem(nd->nd_mrep);
1445 		    return (0);
1446 		}
1447 		if (nd->nd_flag & ND_NFSV3)
1448 		    error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1449 		else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
1450 		    ND_NFSV4) {
1451 			/* Load the directory attributes. */
1452 			error = nfsm_loadattr(nd, dnap);
1453 			if (error == 0)
1454 				*dattrflagp = 1;
1455 		}
1456 		goto nfsmout;
1457 	}
1458 	if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
1459 		/* Load the directory attributes. */
1460 		error = nfsm_loadattr(nd, dnap);
1461 		if (error != 0)
1462 			goto nfsmout;
1463 		*dattrflagp = 1;
1464 		/* Skip over the Lookup and GetFH operation status values. */
1465 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1466 	}
1467 	error = nfsm_getfh(nd, nfhpp);
1468 	if (error)
1469 		goto nfsmout;
1470 
1471 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1472 	if ((nd->nd_flag & ND_NFSV3) && !error)
1473 		error = nfscl_postop_attr(nd, dnap, dattrflagp, stuff);
1474 nfsmout:
1475 	m_freem(nd->nd_mrep);
1476 	if (!error && nd->nd_repstat)
1477 		error = nd->nd_repstat;
1478 	return (error);
1479 }
1480 
1481 /*
1482  * Do a readlink rpc.
1483  */
1484 int
1485 nfsrpc_readlink(vnode_t vp, struct uio *uiop, struct ucred *cred,
1486     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1487 {
1488 	u_int32_t *tl;
1489 	struct nfsrv_descript nfsd, *nd = &nfsd;
1490 	struct nfsnode *np = VTONFS(vp);
1491 	nfsattrbit_t attrbits;
1492 	int error, len, cangetattr = 1;
1493 
1494 	*attrflagp = 0;
1495 	NFSCL_REQSTART(nd, NFSPROC_READLINK, vp);
1496 	if (nd->nd_flag & ND_NFSV4) {
1497 		/*
1498 		 * And do a Getattr op.
1499 		 */
1500 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1501 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
1502 		NFSGETATTR_ATTRBIT(&attrbits);
1503 		(void) nfsrv_putattrbit(nd, &attrbits);
1504 	}
1505 	error = nfscl_request(nd, vp, p, cred, stuff);
1506 	if (error)
1507 		return (error);
1508 	if (nd->nd_flag & ND_NFSV3)
1509 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1510 	if (!nd->nd_repstat && !error) {
1511 		NFSM_STRSIZ(len, NFS_MAXPATHLEN);
1512 		/*
1513 		 * This seems weird to me, but must have been added to
1514 		 * FreeBSD for some reason. The only thing I can think of
1515 		 * is that there was/is some server that replies with
1516 		 * more link data than it should?
1517 		 */
1518 		if (len == NFS_MAXPATHLEN) {
1519 			NFSLOCKNODE(np);
1520 			if (np->n_size > 0 && np->n_size < NFS_MAXPATHLEN) {
1521 				len = np->n_size;
1522 				cangetattr = 0;
1523 			}
1524 			NFSUNLOCKNODE(np);
1525 		}
1526 		error = nfsm_mbufuio(nd, uiop, len);
1527 		if ((nd->nd_flag & ND_NFSV4) && !error && cangetattr)
1528 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1529 	}
1530 	if (nd->nd_repstat && !error)
1531 		error = nd->nd_repstat;
1532 nfsmout:
1533 	m_freem(nd->nd_mrep);
1534 	return (error);
1535 }
1536 
1537 /*
1538  * Read operation.
1539  */
1540 int
1541 nfsrpc_read(vnode_t vp, struct uio *uiop, struct ucred *cred,
1542     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1543 {
1544 	int error, expireret = 0, retrycnt;
1545 	u_int32_t clidrev = 0;
1546 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1547 	struct nfsnode *np = VTONFS(vp);
1548 	struct ucred *newcred;
1549 	struct nfsfh *nfhp = NULL;
1550 	nfsv4stateid_t stateid;
1551 	void *lckp;
1552 
1553 	if (nmp->nm_clp != NULL)
1554 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1555 	newcred = cred;
1556 	if (NFSHASNFSV4(nmp)) {
1557 		nfhp = np->n_fhp;
1558 		newcred = NFSNEWCRED(cred);
1559 	}
1560 	retrycnt = 0;
1561 	do {
1562 		lckp = NULL;
1563 		if (NFSHASNFSV4(nmp))
1564 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1565 			    NFSV4OPEN_ACCESSREAD, 0, newcred, p, &stateid,
1566 			    &lckp);
1567 		error = nfsrpc_readrpc(vp, uiop, newcred, &stateid, p, nap,
1568 		    attrflagp, stuff);
1569 		if (error == NFSERR_OPENMODE) {
1570 			NFSLOCKMNT(nmp);
1571 			nmp->nm_state |= NFSSTA_OPENMODE;
1572 			NFSUNLOCKMNT(nmp);
1573 		}
1574 		if (error == NFSERR_STALESTATEID)
1575 			nfscl_initiate_recovery(nmp->nm_clp);
1576 		if (lckp != NULL)
1577 			nfscl_lockderef(lckp);
1578 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1579 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1580 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1581 			(void) nfs_catnap(PZERO, error, "nfs_read");
1582 		} else if ((error == NFSERR_EXPIRED ||
1583 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1584 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1585 		}
1586 		retrycnt++;
1587 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1588 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1589 	    error == NFSERR_BADSESSION ||
1590 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1591 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1592 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
1593 	    (error == NFSERR_OPENMODE && retrycnt < 4));
1594 	if (error && retrycnt >= 4)
1595 		error = EIO;
1596 	if (NFSHASNFSV4(nmp))
1597 		NFSFREECRED(newcred);
1598 	return (error);
1599 }
1600 
1601 /*
1602  * The actual read RPC.
1603  */
1604 static int
1605 nfsrpc_readrpc(vnode_t vp, struct uio *uiop, struct ucred *cred,
1606     nfsv4stateid_t *stateidp, NFSPROC_T *p, struct nfsvattr *nap,
1607     int *attrflagp, void *stuff)
1608 {
1609 	u_int32_t *tl;
1610 	int error = 0, len, retlen, tsiz, eof = 0;
1611 	struct nfsrv_descript nfsd;
1612 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1613 	struct nfsrv_descript *nd = &nfsd;
1614 	int rsize;
1615 	off_t tmp_off;
1616 
1617 	*attrflagp = 0;
1618 	tsiz = uiop->uio_resid;
1619 	tmp_off = uiop->uio_offset + tsiz;
1620 	NFSLOCKMNT(nmp);
1621 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1622 		NFSUNLOCKMNT(nmp);
1623 		return (EFBIG);
1624 	}
1625 	rsize = nmp->nm_rsize;
1626 	NFSUNLOCKMNT(nmp);
1627 	nd->nd_mrep = NULL;
1628 	while (tsiz > 0) {
1629 		*attrflagp = 0;
1630 		len = (tsiz > rsize) ? rsize : tsiz;
1631 		NFSCL_REQSTART(nd, NFSPROC_READ, vp);
1632 		if (nd->nd_flag & ND_NFSV4)
1633 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1634 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED * 3);
1635 		if (nd->nd_flag & ND_NFSV2) {
1636 			*tl++ = txdr_unsigned(uiop->uio_offset);
1637 			*tl++ = txdr_unsigned(len);
1638 			*tl = 0;
1639 		} else {
1640 			txdr_hyper(uiop->uio_offset, tl);
1641 			*(tl + 2) = txdr_unsigned(len);
1642 		}
1643 		/*
1644 		 * Since I can't do a Getattr for NFSv4 for Write, there
1645 		 * doesn't seem any point in doing one here, either.
1646 		 * (See the comment in nfsrpc_writerpc() for more info.)
1647 		 */
1648 		error = nfscl_request(nd, vp, p, cred, stuff);
1649 		if (error)
1650 			return (error);
1651 		if (nd->nd_flag & ND_NFSV3) {
1652 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
1653 		} else if (!nd->nd_repstat && (nd->nd_flag & ND_NFSV2)) {
1654 			error = nfsm_loadattr(nd, nap);
1655 			if (!error)
1656 				*attrflagp = 1;
1657 		}
1658 		if (nd->nd_repstat || error) {
1659 			if (!error)
1660 				error = nd->nd_repstat;
1661 			goto nfsmout;
1662 		}
1663 		if (nd->nd_flag & ND_NFSV3) {
1664 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1665 			eof = fxdr_unsigned(int, *(tl + 1));
1666 		} else if (nd->nd_flag & ND_NFSV4) {
1667 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
1668 			eof = fxdr_unsigned(int, *tl);
1669 		}
1670 		NFSM_STRSIZ(retlen, len);
1671 		error = nfsm_mbufuio(nd, uiop, retlen);
1672 		if (error)
1673 			goto nfsmout;
1674 		m_freem(nd->nd_mrep);
1675 		nd->nd_mrep = NULL;
1676 		tsiz -= retlen;
1677 		if (!(nd->nd_flag & ND_NFSV2)) {
1678 			if (eof || retlen == 0)
1679 				tsiz = 0;
1680 		} else if (retlen < len)
1681 			tsiz = 0;
1682 	}
1683 	return (0);
1684 nfsmout:
1685 	if (nd->nd_mrep != NULL)
1686 		m_freem(nd->nd_mrep);
1687 	return (error);
1688 }
1689 
1690 /*
1691  * nfs write operation
1692  * When called_from_strategy != 0, it should return EIO for an error that
1693  * indicates recovery is in progress, so that the buffer will be left
1694  * dirty and be written back to the server later. If it loops around,
1695  * the recovery thread could get stuck waiting for the buffer and recovery
1696  * will then deadlock.
1697  */
1698 int
1699 nfsrpc_write(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
1700     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
1701     void *stuff, int called_from_strategy)
1702 {
1703 	int error, expireret = 0, retrycnt, nostateid;
1704 	u_int32_t clidrev = 0;
1705 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1706 	struct nfsnode *np = VTONFS(vp);
1707 	struct ucred *newcred;
1708 	struct nfsfh *nfhp = NULL;
1709 	nfsv4stateid_t stateid;
1710 	void *lckp;
1711 
1712 	*must_commit = 0;
1713 	if (nmp->nm_clp != NULL)
1714 		clidrev = nmp->nm_clp->nfsc_clientidrev;
1715 	newcred = cred;
1716 	if (NFSHASNFSV4(nmp)) {
1717 		newcred = NFSNEWCRED(cred);
1718 		nfhp = np->n_fhp;
1719 	}
1720 	retrycnt = 0;
1721 	do {
1722 		lckp = NULL;
1723 		nostateid = 0;
1724 		if (NFSHASNFSV4(nmp)) {
1725 			(void)nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
1726 			    NFSV4OPEN_ACCESSWRITE, 0, newcred, p, &stateid,
1727 			    &lckp);
1728 			if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
1729 			    stateid.other[2] == 0) {
1730 				nostateid = 1;
1731 				NFSCL_DEBUG(1, "stateid0 in write\n");
1732 			}
1733 		}
1734 
1735 		/*
1736 		 * If there is no stateid for NFSv4, it means this is an
1737 		 * extraneous write after close. Basically a poorly
1738 		 * implemented buffer cache. Just don't do the write.
1739 		 */
1740 		if (nostateid)
1741 			error = 0;
1742 		else
1743 			error = nfsrpc_writerpc(vp, uiop, iomode, must_commit,
1744 			    newcred, &stateid, p, nap, attrflagp, stuff);
1745 		if (error == NFSERR_STALESTATEID)
1746 			nfscl_initiate_recovery(nmp->nm_clp);
1747 		if (lckp != NULL)
1748 			nfscl_lockderef(lckp);
1749 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
1750 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
1751 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
1752 			(void) nfs_catnap(PZERO, error, "nfs_write");
1753 		} else if ((error == NFSERR_EXPIRED ||
1754 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
1755 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
1756 		}
1757 		retrycnt++;
1758 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
1759 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1760 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy == 0) ||
1761 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
1762 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
1763 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
1764 	if (error != 0 && (retrycnt >= 4 ||
1765 	    ((error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
1766 	      error == NFSERR_STALEDONTRECOVER) && called_from_strategy != 0)))
1767 		error = EIO;
1768 	if (NFSHASNFSV4(nmp))
1769 		NFSFREECRED(newcred);
1770 	return (error);
1771 }
1772 
1773 /*
1774  * The actual write RPC.
1775  */
1776 static int
1777 nfsrpc_writerpc(vnode_t vp, struct uio *uiop, int *iomode,
1778     int *must_commit, struct ucred *cred, nfsv4stateid_t *stateidp,
1779     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
1780 {
1781 	u_int32_t *tl;
1782 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
1783 	struct nfsnode *np = VTONFS(vp);
1784 	int error = 0, len, tsiz, rlen, commit, committed = NFSWRITE_FILESYNC;
1785 	int wccflag = 0, wsize;
1786 	int32_t backup;
1787 	struct nfsrv_descript nfsd;
1788 	struct nfsrv_descript *nd = &nfsd;
1789 	nfsattrbit_t attrbits;
1790 	off_t tmp_off;
1791 
1792 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
1793 	*attrflagp = 0;
1794 	tsiz = uiop->uio_resid;
1795 	tmp_off = uiop->uio_offset + tsiz;
1796 	NFSLOCKMNT(nmp);
1797 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < uiop->uio_offset) {
1798 		NFSUNLOCKMNT(nmp);
1799 		return (EFBIG);
1800 	}
1801 	wsize = nmp->nm_wsize;
1802 	NFSUNLOCKMNT(nmp);
1803 	nd->nd_mrep = NULL;	/* NFSv2 sometimes does a write with */
1804 	nd->nd_repstat = 0;	/* uio_resid == 0, so the while is not done */
1805 	while (tsiz > 0) {
1806 		*attrflagp = 0;
1807 		len = (tsiz > wsize) ? wsize : tsiz;
1808 		NFSCL_REQSTART(nd, NFSPROC_WRITE, vp);
1809 		if (nd->nd_flag & ND_NFSV4) {
1810 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
1811 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+2*NFSX_UNSIGNED);
1812 			txdr_hyper(uiop->uio_offset, tl);
1813 			tl += 2;
1814 			*tl++ = txdr_unsigned(*iomode);
1815 			*tl = txdr_unsigned(len);
1816 		} else if (nd->nd_flag & ND_NFSV3) {
1817 			NFSM_BUILD(tl, u_int32_t *, NFSX_HYPER+3*NFSX_UNSIGNED);
1818 			txdr_hyper(uiop->uio_offset, tl);
1819 			tl += 2;
1820 			*tl++ = txdr_unsigned(len);
1821 			*tl++ = txdr_unsigned(*iomode);
1822 			*tl = txdr_unsigned(len);
1823 		} else {
1824 			u_int32_t x;
1825 
1826 			NFSM_BUILD(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1827 			/*
1828 			 * Not sure why someone changed this, since the
1829 			 * RFC clearly states that "beginoffset" and
1830 			 * "totalcount" are ignored, but it wouldn't
1831 			 * surprise me if there's a busted server out there.
1832 			 */
1833 			/* Set both "begin" and "current" to non-garbage. */
1834 			x = txdr_unsigned((u_int32_t)uiop->uio_offset);
1835 			*tl++ = x;      /* "begin offset" */
1836 			*tl++ = x;      /* "current offset" */
1837 			x = txdr_unsigned(len);
1838 			*tl++ = x;      /* total to this offset */
1839 			*tl = x;        /* size of this write */
1840 
1841 		}
1842 		nfsm_uiombuf(nd, uiop, len);
1843 		/*
1844 		 * Although it is tempting to do a normal Getattr Op in the
1845 		 * NFSv4 compound, the result can be a nearly hung client
1846 		 * system if the Getattr asks for Owner and/or OwnerGroup.
1847 		 * It occurs when the client can't map either the Owner or
1848 		 * Owner_group name in the Getattr reply to a uid/gid. When
1849 		 * there is a cache miss, the kernel does an upcall to the
1850 		 * nfsuserd. Then, it can try and read the local /etc/passwd
1851 		 * or /etc/group file. It can then block in getnewbuf(),
1852 		 * waiting for dirty writes to be pushed to the NFS server.
1853 		 * The only reason this doesn't result in a complete
1854 		 * deadlock, is that the upcall times out and allows
1855 		 * the write to complete. However, progress is so slow
1856 		 * that it might just as well be deadlocked.
1857 		 * As such, we get the rest of the attributes, but not
1858 		 * Owner or Owner_group.
1859 		 * nb: nfscl_loadattrcache() needs to be told that these
1860 		 *     partial attributes from a write rpc are being
1861 		 *     passed in, via a argument flag.
1862 		 */
1863 		if (nd->nd_flag & ND_NFSV4) {
1864 			NFSWRITEGETATTR_ATTRBIT(&attrbits);
1865 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1866 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
1867 			(void) nfsrv_putattrbit(nd, &attrbits);
1868 		}
1869 		error = nfscl_request(nd, vp, p, cred, stuff);
1870 		if (error)
1871 			return (error);
1872 		if (nd->nd_repstat) {
1873 			/*
1874 			 * In case the rpc gets retried, roll
1875 			 * the uio fileds changed by nfsm_uiombuf()
1876 			 * back.
1877 			 */
1878 			uiop->uio_offset -= len;
1879 			uiop->uio_resid += len;
1880 			uiop->uio_iov->iov_base =
1881 			    (char *)uiop->uio_iov->iov_base - len;
1882 			uiop->uio_iov->iov_len += len;
1883 		}
1884 		if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1885 			error = nfscl_wcc_data(nd, vp, nap, attrflagp,
1886 			    &wccflag, stuff);
1887 			if (error)
1888 				goto nfsmout;
1889 		}
1890 		if (!nd->nd_repstat) {
1891 			if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
1892 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED
1893 					+ NFSX_VERF);
1894 				rlen = fxdr_unsigned(int, *tl++);
1895 				if (rlen == 0) {
1896 					error = NFSERR_IO;
1897 					goto nfsmout;
1898 				} else if (rlen < len) {
1899 					backup = len - rlen;
1900 					uiop->uio_iov->iov_base =
1901 					    (char *)uiop->uio_iov->iov_base -
1902 					    backup;
1903 					uiop->uio_iov->iov_len += backup;
1904 					uiop->uio_offset -= backup;
1905 					uiop->uio_resid += backup;
1906 					len = rlen;
1907 				}
1908 				commit = fxdr_unsigned(int, *tl++);
1909 
1910 				/*
1911 				 * Return the lowest commitment level
1912 				 * obtained by any of the RPCs.
1913 				 */
1914 				if (committed == NFSWRITE_FILESYNC)
1915 					committed = commit;
1916 				else if (committed == NFSWRITE_DATASYNC &&
1917 					commit == NFSWRITE_UNSTABLE)
1918 					committed = commit;
1919 				NFSLOCKMNT(nmp);
1920 				if (!NFSHASWRITEVERF(nmp)) {
1921 					NFSBCOPY((caddr_t)tl,
1922 					    (caddr_t)&nmp->nm_verf[0],
1923 					    NFSX_VERF);
1924 					NFSSETWRITEVERF(nmp);
1925 	    			} else if (NFSBCMP(tl, nmp->nm_verf,
1926 				    NFSX_VERF)) {
1927 					*must_commit = 1;
1928 					NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
1929 				}
1930 				NFSUNLOCKMNT(nmp);
1931 			}
1932 			if (nd->nd_flag & ND_NFSV4)
1933 				NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1934 			if (nd->nd_flag & (ND_NFSV2 | ND_NFSV4)) {
1935 				error = nfsm_loadattr(nd, nap);
1936 				if (!error)
1937 					*attrflagp = NFS_LATTR_NOSHRINK;
1938 			}
1939 		} else {
1940 			error = nd->nd_repstat;
1941 		}
1942 		if (error)
1943 			goto nfsmout;
1944 		NFSWRITERPC_SETTIME(wccflag, np, nap, (nd->nd_flag & ND_NFSV4));
1945 		m_freem(nd->nd_mrep);
1946 		nd->nd_mrep = NULL;
1947 		tsiz -= len;
1948 	}
1949 nfsmout:
1950 	if (nd->nd_mrep != NULL)
1951 		m_freem(nd->nd_mrep);
1952 	*iomode = committed;
1953 	if (nd->nd_repstat && !error)
1954 		error = nd->nd_repstat;
1955 	return (error);
1956 }
1957 
1958 /*
1959  * nfs mknod rpc
1960  * For NFS v2 this is a kludge. Use a create rpc but with the IFMT bits of the
1961  * mode set to specify the file type and the size field for rdev.
1962  */
1963 int
1964 nfsrpc_mknod(vnode_t dvp, char *name, int namelen, struct vattr *vap,
1965     u_int32_t rdev, enum vtype vtyp, struct ucred *cred, NFSPROC_T *p,
1966     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
1967     int *attrflagp, int *dattrflagp, void *dstuff)
1968 {
1969 	u_int32_t *tl;
1970 	int error = 0;
1971 	struct nfsrv_descript nfsd, *nd = &nfsd;
1972 	nfsattrbit_t attrbits;
1973 
1974 	*nfhpp = NULL;
1975 	*attrflagp = 0;
1976 	*dattrflagp = 0;
1977 	if (namelen > NFS_MAXNAMLEN)
1978 		return (ENAMETOOLONG);
1979 	NFSCL_REQSTART(nd, NFSPROC_MKNOD, dvp);
1980 	if (nd->nd_flag & ND_NFSV4) {
1981 		if (vtyp == VBLK || vtyp == VCHR) {
1982 			NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
1983 			*tl++ = vtonfsv34_type(vtyp);
1984 			*tl++ = txdr_unsigned(NFSMAJOR(rdev));
1985 			*tl = txdr_unsigned(NFSMINOR(rdev));
1986 		} else {
1987 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1988 			*tl = vtonfsv34_type(vtyp);
1989 		}
1990 	}
1991 	(void) nfsm_strtom(nd, name, namelen);
1992 	if (nd->nd_flag & ND_NFSV3) {
1993 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
1994 		*tl = vtonfsv34_type(vtyp);
1995 	}
1996 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
1997 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
1998 	if ((nd->nd_flag & ND_NFSV3) &&
1999 	    (vtyp == VCHR || vtyp == VBLK)) {
2000 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2001 		*tl++ = txdr_unsigned(NFSMAJOR(rdev));
2002 		*tl = txdr_unsigned(NFSMINOR(rdev));
2003 	}
2004 	if (nd->nd_flag & ND_NFSV4) {
2005 		NFSGETATTR_ATTRBIT(&attrbits);
2006 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2007 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2008 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2009 		(void) nfsrv_putattrbit(nd, &attrbits);
2010 	}
2011 	if (nd->nd_flag & ND_NFSV2)
2012 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZERDEV, rdev);
2013 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2014 	if (error)
2015 		return (error);
2016 	if (nd->nd_flag & ND_NFSV4)
2017 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2018 	if (!nd->nd_repstat) {
2019 		if (nd->nd_flag & ND_NFSV4) {
2020 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2021 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2022 			if (error)
2023 				goto nfsmout;
2024 		}
2025 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2026 		if (error)
2027 			goto nfsmout;
2028 	}
2029 	if (nd->nd_flag & ND_NFSV3)
2030 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2031 	if (!error && nd->nd_repstat)
2032 		error = nd->nd_repstat;
2033 nfsmout:
2034 	m_freem(nd->nd_mrep);
2035 	return (error);
2036 }
2037 
2038 /*
2039  * nfs file create call
2040  * Mostly just call the approriate routine. (I separated out v4, so that
2041  * error recovery wouldn't be as difficult.)
2042  */
2043 int
2044 nfsrpc_create(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2045     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2046     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2047     int *attrflagp, int *dattrflagp, void *dstuff)
2048 {
2049 	int error = 0, newone, expireret = 0, retrycnt, unlocked;
2050 	struct nfsclowner *owp;
2051 	struct nfscldeleg *dp;
2052 	struct nfsmount *nmp = VFSTONFS(vnode_mount(dvp));
2053 	u_int32_t clidrev;
2054 
2055 	if (NFSHASNFSV4(nmp)) {
2056 	    retrycnt = 0;
2057 	    do {
2058 		dp = NULL;
2059 		error = nfscl_open(dvp, NULL, 0, (NFSV4OPEN_ACCESSWRITE |
2060 		    NFSV4OPEN_ACCESSREAD), 0, cred, p, &owp, NULL, &newone,
2061 		    NULL, 1);
2062 		if (error)
2063 			return (error);
2064 		if (nmp->nm_clp != NULL)
2065 			clidrev = nmp->nm_clp->nfsc_clientidrev;
2066 		else
2067 			clidrev = 0;
2068 		if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 ||
2069 		    nfs_numnfscbd == 0 || retrycnt > 0)
2070 			error = nfsrpc_createv4(dvp, name, namelen, vap, cverf,
2071 			  fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2072 			  attrflagp, dattrflagp, dstuff, &unlocked);
2073 		else
2074 			error = nfsrpc_getcreatelayout(dvp, name, namelen, vap,
2075 			  cverf, fmode, owp, &dp, cred, p, dnap, nnap, nfhpp,
2076 			  attrflagp, dattrflagp, dstuff, &unlocked);
2077 		/*
2078 		 * There is no need to invalidate cached attributes here,
2079 		 * since new post-delegation issue attributes are always
2080 		 * returned by nfsrpc_createv4() and these will update the
2081 		 * attribute cache.
2082 		 */
2083 		if (dp != NULL)
2084 			(void) nfscl_deleg(nmp->nm_mountp, owp->nfsow_clp,
2085 			    (*nfhpp)->nfh_fh, (*nfhpp)->nfh_len, cred, p, &dp);
2086 		nfscl_ownerrelease(nmp, owp, error, newone, unlocked);
2087 		if (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2088 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2089 		    error == NFSERR_BADSESSION) {
2090 			(void) nfs_catnap(PZERO, error, "nfs_open");
2091 		} else if ((error == NFSERR_EXPIRED ||
2092 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
2093 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
2094 			retrycnt++;
2095 		}
2096 	    } while (error == NFSERR_GRACE || error == NFSERR_STALECLIENTID ||
2097 		error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
2098 		error == NFSERR_BADSESSION ||
2099 		((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
2100 		 expireret == 0 && clidrev != 0 && retrycnt < 4));
2101 	    if (error && retrycnt >= 4)
2102 		    error = EIO;
2103 	} else {
2104 		error = nfsrpc_createv23(dvp, name, namelen, vap, cverf,
2105 		    fmode, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
2106 		    dstuff);
2107 	}
2108 	return (error);
2109 }
2110 
2111 /*
2112  * The create rpc for v2 and 3.
2113  */
2114 static int
2115 nfsrpc_createv23(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2116     nfsquad_t cverf, int fmode, struct ucred *cred, NFSPROC_T *p,
2117     struct nfsvattr *dnap, struct nfsvattr *nnap, struct nfsfh **nfhpp,
2118     int *attrflagp, int *dattrflagp, void *dstuff)
2119 {
2120 	u_int32_t *tl;
2121 	int error = 0;
2122 	struct nfsrv_descript nfsd, *nd = &nfsd;
2123 
2124 	*nfhpp = NULL;
2125 	*attrflagp = 0;
2126 	*dattrflagp = 0;
2127 	if (namelen > NFS_MAXNAMLEN)
2128 		return (ENAMETOOLONG);
2129 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2130 	(void) nfsm_strtom(nd, name, namelen);
2131 	if (nd->nd_flag & ND_NFSV3) {
2132 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2133 		if (fmode & O_EXCL) {
2134 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2135 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2136 			*tl++ = cverf.lval[0];
2137 			*tl = cverf.lval[1];
2138 		} else {
2139 			*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2140 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
2141 		}
2142 	} else {
2143 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZE0, 0);
2144 	}
2145 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2146 	if (error)
2147 		return (error);
2148 	if (nd->nd_repstat == 0) {
2149 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2150 		if (error)
2151 			goto nfsmout;
2152 	}
2153 	if (nd->nd_flag & ND_NFSV3)
2154 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2155 	if (nd->nd_repstat != 0 && error == 0)
2156 		error = nd->nd_repstat;
2157 nfsmout:
2158 	m_freem(nd->nd_mrep);
2159 	return (error);
2160 }
2161 
2162 static int
2163 nfsrpc_createv4(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2164     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
2165     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2166     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2167     int *dattrflagp, void *dstuff, int *unlockedp)
2168 {
2169 	u_int32_t *tl;
2170 	int error = 0, deleg, newone, ret, acesize, limitby;
2171 	struct nfsrv_descript nfsd, *nd = &nfsd;
2172 	struct nfsclopen *op;
2173 	struct nfscldeleg *dp = NULL;
2174 	struct nfsnode *np;
2175 	struct nfsfh *nfhp;
2176 	nfsattrbit_t attrbits;
2177 	nfsv4stateid_t stateid;
2178 	u_int32_t rflags;
2179 	struct nfsmount *nmp;
2180 	struct nfsclsession *tsep;
2181 
2182 	nmp = VFSTONFS(dvp->v_mount);
2183 	np = VTONFS(dvp);
2184 	*unlockedp = 0;
2185 	*nfhpp = NULL;
2186 	*dpp = NULL;
2187 	*attrflagp = 0;
2188 	*dattrflagp = 0;
2189 	if (namelen > NFS_MAXNAMLEN)
2190 		return (ENAMETOOLONG);
2191 	NFSCL_REQSTART(nd, NFSPROC_CREATE, dvp);
2192 	/*
2193 	 * For V4, this is actually an Open op.
2194 	 */
2195 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2196 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
2197 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
2198 	    NFSV4OPEN_ACCESSREAD);
2199 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
2200 	tsep = nfsmnt_mdssession(nmp);
2201 	*tl++ = tsep->nfsess_clientid.lval[0];
2202 	*tl = tsep->nfsess_clientid.lval[1];
2203 	(void) nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
2204 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2205 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
2206 	if (fmode & O_EXCL) {
2207 		if (NFSHASNFSV4N(nmp)) {
2208 			if (NFSHASSESSPERSIST(nmp)) {
2209 				/* Use GUARDED for persistent sessions. */
2210 				*tl = txdr_unsigned(NFSCREATE_GUARDED);
2211 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2212 			} else {
2213 				/* Otherwise, use EXCLUSIVE4_1. */
2214 				*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
2215 				NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2216 				*tl++ = cverf.lval[0];
2217 				*tl = cverf.lval[1];
2218 				nfscl_fillsattr(nd, vap, dvp, 0, 0);
2219 			}
2220 		} else {
2221 			/* NFSv4.0 */
2222 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE);
2223 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
2224 			*tl++ = cverf.lval[0];
2225 			*tl = cverf.lval[1];
2226 		}
2227 	} else {
2228 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
2229 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2230 	}
2231 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2232 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
2233 	(void) nfsm_strtom(nd, name, namelen);
2234 	/* Get the new file's handle and attributes. */
2235 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2236 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2237 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2238 	NFSGETATTR_ATTRBIT(&attrbits);
2239 	(void) nfsrv_putattrbit(nd, &attrbits);
2240 	/* Get the directory's post-op attributes. */
2241 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2242 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
2243 	(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
2244 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2245 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
2246 	(void) nfsrv_putattrbit(nd, &attrbits);
2247 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2248 	if (error)
2249 		return (error);
2250 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
2251 	if (nd->nd_repstat == 0) {
2252 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2253 		    6 * NFSX_UNSIGNED);
2254 		stateid.seqid = *tl++;
2255 		stateid.other[0] = *tl++;
2256 		stateid.other[1] = *tl++;
2257 		stateid.other[2] = *tl;
2258 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
2259 		(void) nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2260 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
2261 		deleg = fxdr_unsigned(int, *tl);
2262 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
2263 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
2264 			if (!(owp->nfsow_clp->nfsc_flags &
2265 			      NFSCLFLAGS_FIRSTDELEG))
2266 				owp->nfsow_clp->nfsc_flags |=
2267 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
2268 			dp = malloc(
2269 			    sizeof (struct nfscldeleg) + NFSX_V4FHMAX,
2270 			    M_NFSCLDELEG, M_WAITOK);
2271 			LIST_INIT(&dp->nfsdl_owner);
2272 			LIST_INIT(&dp->nfsdl_lock);
2273 			dp->nfsdl_clp = owp->nfsow_clp;
2274 			newnfs_copyincred(cred, &dp->nfsdl_cred);
2275 			nfscl_lockinit(&dp->nfsdl_rwlock);
2276 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
2277 			    NFSX_UNSIGNED);
2278 			dp->nfsdl_stateid.seqid = *tl++;
2279 			dp->nfsdl_stateid.other[0] = *tl++;
2280 			dp->nfsdl_stateid.other[1] = *tl++;
2281 			dp->nfsdl_stateid.other[2] = *tl++;
2282 			ret = fxdr_unsigned(int, *tl);
2283 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
2284 				dp->nfsdl_flags = NFSCLDL_WRITE;
2285 				/*
2286 				 * Indicates how much the file can grow.
2287 				 */
2288 				NFSM_DISSECT(tl, u_int32_t *,
2289 				    3 * NFSX_UNSIGNED);
2290 				limitby = fxdr_unsigned(int, *tl++);
2291 				switch (limitby) {
2292 				case NFSV4OPEN_LIMITSIZE:
2293 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
2294 					break;
2295 				case NFSV4OPEN_LIMITBLOCKS:
2296 					dp->nfsdl_sizelimit =
2297 					    fxdr_unsigned(u_int64_t, *tl++);
2298 					dp->nfsdl_sizelimit *=
2299 					    fxdr_unsigned(u_int64_t, *tl);
2300 					break;
2301 				default:
2302 					error = NFSERR_BADXDR;
2303 					goto nfsmout;
2304 				}
2305 			} else {
2306 				dp->nfsdl_flags = NFSCLDL_READ;
2307 			}
2308 			if (ret)
2309 				dp->nfsdl_flags |= NFSCLDL_RECALL;
2310 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
2311 			    &acesize, p);
2312 			if (error)
2313 				goto nfsmout;
2314 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
2315 			error = NFSERR_BADXDR;
2316 			goto nfsmout;
2317 		}
2318 		error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2319 		if (error)
2320 			goto nfsmout;
2321 		/* Get rid of the PutFH and Getattr status values. */
2322 		NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2323 		/* Load the directory attributes. */
2324 		error = nfsm_loadattr(nd, dnap);
2325 		if (error)
2326 			goto nfsmout;
2327 		*dattrflagp = 1;
2328 		if (dp != NULL && *attrflagp) {
2329 			dp->nfsdl_change = nnap->na_filerev;
2330 			dp->nfsdl_modtime = nnap->na_mtime;
2331 			dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
2332 		}
2333 		/*
2334 		 * We can now complete the Open state.
2335 		 */
2336 		nfhp = *nfhpp;
2337 		if (dp != NULL) {
2338 			dp->nfsdl_fhlen = nfhp->nfh_len;
2339 			NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh, nfhp->nfh_len);
2340 		}
2341 		/*
2342 		 * Get an Open structure that will be
2343 		 * attached to the OpenOwner, acquired already.
2344 		 */
2345 		error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
2346 		    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
2347 		    cred, p, NULL, &op, &newone, NULL, 0);
2348 		if (error)
2349 			goto nfsmout;
2350 		op->nfso_stateid = stateid;
2351 		newnfs_copyincred(cred, &op->nfso_cred);
2352 		if ((rflags & NFSV4OPEN_RESULTCONFIRM)) {
2353 		    do {
2354 			ret = nfsrpc_openconfirm(dvp, nfhp->nfh_fh,
2355 			    nfhp->nfh_len, op, cred, p);
2356 			if (ret == NFSERR_DELAY)
2357 			    (void) nfs_catnap(PZERO, ret, "nfs_create");
2358 		    } while (ret == NFSERR_DELAY);
2359 		    error = ret;
2360 		}
2361 
2362 		/*
2363 		 * If the server is handing out delegations, but we didn't
2364 		 * get one because an OpenConfirm was required, try the
2365 		 * Open again, to get a delegation. This is a harmless no-op,
2366 		 * from a server's point of view.
2367 		 */
2368 		if ((rflags & NFSV4OPEN_RESULTCONFIRM) &&
2369 		    (owp->nfsow_clp->nfsc_flags & NFSCLFLAGS_GOTDELEG) &&
2370 		    !error && dp == NULL) {
2371 		    do {
2372 			ret = nfsrpc_openrpc(VFSTONFS(vnode_mount(dvp)), dvp,
2373 			    np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
2374 			    nfhp->nfh_fh, nfhp->nfh_len,
2375 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), op,
2376 			    name, namelen, &dp, 0, 0x0, cred, p, 0, 1);
2377 			if (ret == NFSERR_DELAY)
2378 			    (void) nfs_catnap(PZERO, ret, "nfs_crt2");
2379 		    } while (ret == NFSERR_DELAY);
2380 		    if (ret) {
2381 			if (dp != NULL) {
2382 				free(dp, M_NFSCLDELEG);
2383 				dp = NULL;
2384 			}
2385 			if (ret == NFSERR_STALECLIENTID ||
2386 			    ret == NFSERR_STALEDONTRECOVER ||
2387 			    ret == NFSERR_BADSESSION)
2388 				error = ret;
2389 		    }
2390 		}
2391 		nfscl_openrelease(nmp, op, error, newone);
2392 		*unlockedp = 1;
2393 	}
2394 	if (nd->nd_repstat != 0 && error == 0)
2395 		error = nd->nd_repstat;
2396 	if (error == NFSERR_STALECLIENTID)
2397 		nfscl_initiate_recovery(owp->nfsow_clp);
2398 nfsmout:
2399 	if (!error)
2400 		*dpp = dp;
2401 	else if (dp != NULL)
2402 		free(dp, M_NFSCLDELEG);
2403 	m_freem(nd->nd_mrep);
2404 	return (error);
2405 }
2406 
2407 /*
2408  * Nfs remove rpc
2409  */
2410 int
2411 nfsrpc_remove(vnode_t dvp, char *name, int namelen, vnode_t vp,
2412     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp,
2413     void *dstuff)
2414 {
2415 	u_int32_t *tl;
2416 	struct nfsrv_descript nfsd, *nd = &nfsd;
2417 	struct nfsnode *np;
2418 	struct nfsmount *nmp;
2419 	nfsv4stateid_t dstateid;
2420 	int error, ret = 0, i;
2421 
2422 	*dattrflagp = 0;
2423 	if (namelen > NFS_MAXNAMLEN)
2424 		return (ENAMETOOLONG);
2425 	nmp = VFSTONFS(vnode_mount(dvp));
2426 tryagain:
2427 	if (NFSHASNFSV4(nmp) && ret == 0) {
2428 		ret = nfscl_removedeleg(vp, p, &dstateid);
2429 		if (ret == 1) {
2430 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGREMOVE, vp);
2431 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
2432 			    NFSX_UNSIGNED);
2433 			if (NFSHASNFSV4N(nmp))
2434 				*tl++ = 0;
2435 			else
2436 				*tl++ = dstateid.seqid;
2437 			*tl++ = dstateid.other[0];
2438 			*tl++ = dstateid.other[1];
2439 			*tl++ = dstateid.other[2];
2440 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2441 			np = VTONFS(dvp);
2442 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2443 			    np->n_fhp->nfh_len, 0);
2444 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2445 			*tl = txdr_unsigned(NFSV4OP_REMOVE);
2446 		}
2447 	} else {
2448 		ret = 0;
2449 	}
2450 	if (ret == 0)
2451 		NFSCL_REQSTART(nd, NFSPROC_REMOVE, dvp);
2452 	(void) nfsm_strtom(nd, name, namelen);
2453 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2454 	if (error)
2455 		return (error);
2456 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2457 		/* For NFSv4, parse out any Delereturn replies. */
2458 		if (ret > 0 && nd->nd_repstat != 0 &&
2459 		    (nd->nd_flag & ND_NOMOREDATA)) {
2460 			/*
2461 			 * If the Delegreturn failed, try again without
2462 			 * it. The server will Recall, as required.
2463 			 */
2464 			m_freem(nd->nd_mrep);
2465 			goto tryagain;
2466 		}
2467 		for (i = 0; i < (ret * 2); i++) {
2468 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2469 			    ND_NFSV4) {
2470 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2471 			    if (*(tl + 1))
2472 				nd->nd_flag |= ND_NOMOREDATA;
2473 			}
2474 		}
2475 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2476 	}
2477 	if (nd->nd_repstat && !error)
2478 		error = nd->nd_repstat;
2479 nfsmout:
2480 	m_freem(nd->nd_mrep);
2481 	return (error);
2482 }
2483 
2484 /*
2485  * Do an nfs rename rpc.
2486  */
2487 int
2488 nfsrpc_rename(vnode_t fdvp, vnode_t fvp, char *fnameptr, int fnamelen,
2489     vnode_t tdvp, vnode_t tvp, char *tnameptr, int tnamelen, struct ucred *cred,
2490     NFSPROC_T *p, struct nfsvattr *fnap, struct nfsvattr *tnap,
2491     int *fattrflagp, int *tattrflagp, void *fstuff, void *tstuff)
2492 {
2493 	u_int32_t *tl;
2494 	struct nfsrv_descript nfsd, *nd = &nfsd;
2495 	struct nfsmount *nmp;
2496 	struct nfsnode *np;
2497 	nfsattrbit_t attrbits;
2498 	nfsv4stateid_t fdstateid, tdstateid;
2499 	int error = 0, ret = 0, gottd = 0, gotfd = 0, i;
2500 
2501 	*fattrflagp = 0;
2502 	*tattrflagp = 0;
2503 	nmp = VFSTONFS(vnode_mount(fdvp));
2504 	if (fnamelen > NFS_MAXNAMLEN || tnamelen > NFS_MAXNAMLEN)
2505 		return (ENAMETOOLONG);
2506 tryagain:
2507 	if (NFSHASNFSV4(nmp) && ret == 0) {
2508 		ret = nfscl_renamedeleg(fvp, &fdstateid, &gotfd, tvp,
2509 		    &tdstateid, &gottd, p);
2510 		if (gotfd && gottd) {
2511 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME2, fvp);
2512 		} else if (gotfd) {
2513 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, fvp);
2514 		} else if (gottd) {
2515 			NFSCL_REQSTART(nd, NFSPROC_RETDELEGRENAME1, tvp);
2516 		}
2517 		if (gotfd) {
2518 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2519 			if (NFSHASNFSV4N(nmp))
2520 				*tl++ = 0;
2521 			else
2522 				*tl++ = fdstateid.seqid;
2523 			*tl++ = fdstateid.other[0];
2524 			*tl++ = fdstateid.other[1];
2525 			*tl = fdstateid.other[2];
2526 			if (gottd) {
2527 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2528 				*tl = txdr_unsigned(NFSV4OP_PUTFH);
2529 				np = VTONFS(tvp);
2530 				(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2531 				    np->n_fhp->nfh_len, 0);
2532 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2533 				*tl = txdr_unsigned(NFSV4OP_DELEGRETURN);
2534 			}
2535 		}
2536 		if (gottd) {
2537 			NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
2538 			if (NFSHASNFSV4N(nmp))
2539 				*tl++ = 0;
2540 			else
2541 				*tl++ = tdstateid.seqid;
2542 			*tl++ = tdstateid.other[0];
2543 			*tl++ = tdstateid.other[1];
2544 			*tl = tdstateid.other[2];
2545 		}
2546 		if (ret > 0) {
2547 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2548 			*tl = txdr_unsigned(NFSV4OP_PUTFH);
2549 			np = VTONFS(fdvp);
2550 			(void) nfsm_fhtom(nd, np->n_fhp->nfh_fh,
2551 			    np->n_fhp->nfh_len, 0);
2552 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2553 			*tl = txdr_unsigned(NFSV4OP_SAVEFH);
2554 		}
2555 	} else {
2556 		ret = 0;
2557 	}
2558 	if (ret == 0)
2559 		NFSCL_REQSTART(nd, NFSPROC_RENAME, fdvp);
2560 	if (nd->nd_flag & ND_NFSV4) {
2561 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2562 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2563 		NFSWCCATTR_ATTRBIT(&attrbits);
2564 		(void) nfsrv_putattrbit(nd, &attrbits);
2565 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2566 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2567 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2568 		    VTONFS(tdvp)->n_fhp->nfh_len, 0);
2569 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2570 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2571 		(void) nfsrv_putattrbit(nd, &attrbits);
2572 		nd->nd_flag |= ND_V4WCCATTR;
2573 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2574 		*tl = txdr_unsigned(NFSV4OP_RENAME);
2575 	}
2576 	(void) nfsm_strtom(nd, fnameptr, fnamelen);
2577 	if (!(nd->nd_flag & ND_NFSV4))
2578 		(void) nfsm_fhtom(nd, VTONFS(tdvp)->n_fhp->nfh_fh,
2579 			VTONFS(tdvp)->n_fhp->nfh_len, 0);
2580 	(void) nfsm_strtom(nd, tnameptr, tnamelen);
2581 	error = nfscl_request(nd, fdvp, p, cred, fstuff);
2582 	if (error)
2583 		return (error);
2584 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4)) {
2585 		/* For NFSv4, parse out any Delereturn replies. */
2586 		if (ret > 0 && nd->nd_repstat != 0 &&
2587 		    (nd->nd_flag & ND_NOMOREDATA)) {
2588 			/*
2589 			 * If the Delegreturn failed, try again without
2590 			 * it. The server will Recall, as required.
2591 			 */
2592 			m_freem(nd->nd_mrep);
2593 			goto tryagain;
2594 		}
2595 		for (i = 0; i < (ret * 2); i++) {
2596 			if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) ==
2597 			    ND_NFSV4) {
2598 			    NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2599 			    if (*(tl + 1)) {
2600 				if (i == 0 && ret > 1) {
2601 				    /*
2602 				     * If the Delegreturn failed, try again
2603 				     * without it. The server will Recall, as
2604 				     * required.
2605 				     * If ret > 1, the first iteration of this
2606 				     * loop is the second DelegReturn result.
2607 				     */
2608 				    m_freem(nd->nd_mrep);
2609 				    goto tryagain;
2610 				} else {
2611 				    nd->nd_flag |= ND_NOMOREDATA;
2612 				}
2613 			    }
2614 			}
2615 		}
2616 		/* Now, the first wcc attribute reply. */
2617 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2618 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2619 			if (*(tl + 1))
2620 				nd->nd_flag |= ND_NOMOREDATA;
2621 		}
2622 		error = nfscl_wcc_data(nd, fdvp, fnap, fattrflagp, NULL,
2623 		    fstuff);
2624 		/* and the second wcc attribute reply. */
2625 		if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4 &&
2626 		    !error) {
2627 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2628 			if (*(tl + 1))
2629 				nd->nd_flag |= ND_NOMOREDATA;
2630 		}
2631 		if (!error)
2632 			error = nfscl_wcc_data(nd, tdvp, tnap, tattrflagp,
2633 			    NULL, tstuff);
2634 	}
2635 	if (nd->nd_repstat && !error)
2636 		error = nd->nd_repstat;
2637 nfsmout:
2638 	m_freem(nd->nd_mrep);
2639 	return (error);
2640 }
2641 
2642 /*
2643  * nfs hard link create rpc
2644  */
2645 int
2646 nfsrpc_link(vnode_t dvp, vnode_t vp, char *name, int namelen,
2647     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2648     struct nfsvattr *nap, int *attrflagp, int *dattrflagp, void *dstuff)
2649 {
2650 	u_int32_t *tl;
2651 	struct nfsrv_descript nfsd, *nd = &nfsd;
2652 	nfsattrbit_t attrbits;
2653 	int error = 0;
2654 
2655 	*attrflagp = 0;
2656 	*dattrflagp = 0;
2657 	if (namelen > NFS_MAXNAMLEN)
2658 		return (ENAMETOOLONG);
2659 	NFSCL_REQSTART(nd, NFSPROC_LINK, vp);
2660 	if (nd->nd_flag & ND_NFSV4) {
2661 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2662 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2663 	}
2664 	(void) nfsm_fhtom(nd, VTONFS(dvp)->n_fhp->nfh_fh,
2665 		VTONFS(dvp)->n_fhp->nfh_len, 0);
2666 	if (nd->nd_flag & ND_NFSV4) {
2667 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2668 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2669 		NFSWCCATTR_ATTRBIT(&attrbits);
2670 		(void) nfsrv_putattrbit(nd, &attrbits);
2671 		nd->nd_flag |= ND_V4WCCATTR;
2672 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2673 		*tl = txdr_unsigned(NFSV4OP_LINK);
2674 	}
2675 	(void) nfsm_strtom(nd, name, namelen);
2676 	error = nfscl_request(nd, vp, p, cred, dstuff);
2677 	if (error)
2678 		return (error);
2679 	if (nd->nd_flag & ND_NFSV3) {
2680 		error = nfscl_postop_attr(nd, nap, attrflagp, dstuff);
2681 		if (!error)
2682 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2683 			    NULL, dstuff);
2684 	} else if ((nd->nd_flag & (ND_NFSV4 | ND_NOMOREDATA)) == ND_NFSV4) {
2685 		/*
2686 		 * First, parse out the PutFH and Getattr result.
2687 		 */
2688 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2689 		if (!(*(tl + 1)))
2690 			NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2691 		if (*(tl + 1))
2692 			nd->nd_flag |= ND_NOMOREDATA;
2693 		/*
2694 		 * Get the pre-op attributes.
2695 		 */
2696 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2697 	}
2698 	if (nd->nd_repstat && !error)
2699 		error = nd->nd_repstat;
2700 nfsmout:
2701 	m_freem(nd->nd_mrep);
2702 	return (error);
2703 }
2704 
2705 /*
2706  * nfs symbolic link create rpc
2707  */
2708 int
2709 nfsrpc_symlink(vnode_t dvp, char *name, int namelen, const char *target,
2710     struct vattr *vap, struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2711     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2712     int *dattrflagp, void *dstuff)
2713 {
2714 	u_int32_t *tl;
2715 	struct nfsrv_descript nfsd, *nd = &nfsd;
2716 	struct nfsmount *nmp;
2717 	int slen, error = 0;
2718 
2719 	*nfhpp = NULL;
2720 	*attrflagp = 0;
2721 	*dattrflagp = 0;
2722 	nmp = VFSTONFS(vnode_mount(dvp));
2723 	slen = strlen(target);
2724 	if (slen > NFS_MAXPATHLEN || namelen > NFS_MAXNAMLEN)
2725 		return (ENAMETOOLONG);
2726 	NFSCL_REQSTART(nd, NFSPROC_SYMLINK, dvp);
2727 	if (nd->nd_flag & ND_NFSV4) {
2728 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2729 		*tl = txdr_unsigned(NFLNK);
2730 		(void) nfsm_strtom(nd, target, slen);
2731 	}
2732 	(void) nfsm_strtom(nd, name, namelen);
2733 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2734 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
2735 	if (!(nd->nd_flag & ND_NFSV4))
2736 		(void) nfsm_strtom(nd, target, slen);
2737 	if (nd->nd_flag & ND_NFSV2)
2738 		nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2739 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2740 	if (error)
2741 		return (error);
2742 	if (nd->nd_flag & ND_NFSV4)
2743 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2744 	if ((nd->nd_flag & ND_NFSV3) && !error) {
2745 		if (!nd->nd_repstat)
2746 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2747 		if (!error)
2748 			error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp,
2749 			    NULL, dstuff);
2750 	}
2751 	if (nd->nd_repstat && !error)
2752 		error = nd->nd_repstat;
2753 	m_freem(nd->nd_mrep);
2754 	/*
2755 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2756 	 * Only do this if vfs.nfs.ignore_eexist is set.
2757 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2758 	 * should guarantee "exactly once" RPC semantics.
2759 	 */
2760 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2761 	    nmp->nm_minorvers == 0))
2762 		error = 0;
2763 	return (error);
2764 }
2765 
2766 /*
2767  * nfs make dir rpc
2768  */
2769 int
2770 nfsrpc_mkdir(vnode_t dvp, char *name, int namelen, struct vattr *vap,
2771     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
2772     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
2773     int *dattrflagp, void *dstuff)
2774 {
2775 	u_int32_t *tl;
2776 	struct nfsrv_descript nfsd, *nd = &nfsd;
2777 	nfsattrbit_t attrbits;
2778 	int error = 0;
2779 	struct nfsfh *fhp;
2780 	struct nfsmount *nmp;
2781 
2782 	*nfhpp = NULL;
2783 	*attrflagp = 0;
2784 	*dattrflagp = 0;
2785 	nmp = VFSTONFS(vnode_mount(dvp));
2786 	fhp = VTONFS(dvp)->n_fhp;
2787 	if (namelen > NFS_MAXNAMLEN)
2788 		return (ENAMETOOLONG);
2789 	NFSCL_REQSTART(nd, NFSPROC_MKDIR, dvp);
2790 	if (nd->nd_flag & ND_NFSV4) {
2791 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2792 		*tl = txdr_unsigned(NFDIR);
2793 	}
2794 	(void) nfsm_strtom(nd, name, namelen);
2795 	nfscl_fillsattr(nd, vap, dvp, NFSSATTR_SIZENEG1, 0);
2796 	if (nd->nd_flag & ND_NFSV4) {
2797 		NFSGETATTR_ATTRBIT(&attrbits);
2798 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2799 		*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2800 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2801 		(void) nfsrv_putattrbit(nd, &attrbits);
2802 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2803 		*tl = txdr_unsigned(NFSV4OP_PUTFH);
2804 		(void) nfsm_fhtom(nd, fhp->nfh_fh, fhp->nfh_len, 0);
2805 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
2806 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
2807 		(void) nfsrv_putattrbit(nd, &attrbits);
2808 	}
2809 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2810 	if (error)
2811 		return (error);
2812 	if (nd->nd_flag & ND_NFSV4)
2813 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2814 	if (!nd->nd_repstat && !error) {
2815 		if (nd->nd_flag & ND_NFSV4) {
2816 			NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2817 			error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
2818 		}
2819 		if (!error)
2820 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
2821 		if (error == 0 && (nd->nd_flag & ND_NFSV4) != 0) {
2822 			/* Get rid of the PutFH and Getattr status values. */
2823 			NFSM_DISSECT(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2824 			/* Load the directory attributes. */
2825 			error = nfsm_loadattr(nd, dnap);
2826 			if (error == 0)
2827 				*dattrflagp = 1;
2828 		}
2829 	}
2830 	if ((nd->nd_flag & ND_NFSV3) && !error)
2831 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2832 	if (nd->nd_repstat && !error)
2833 		error = nd->nd_repstat;
2834 nfsmout:
2835 	m_freem(nd->nd_mrep);
2836 	/*
2837 	 * Kludge: Map EEXIST => 0 assuming that it is a reply to a retry.
2838 	 * Only do this if vfs.nfs.ignore_eexist is set.
2839 	 * Never do this for NFSv4.1 or later minor versions, since sessions
2840 	 * should guarantee "exactly once" RPC semantics.
2841 	 */
2842 	if (error == EEXIST && nfsignore_eexist != 0 && (!NFSHASNFSV4(nmp) ||
2843 	    nmp->nm_minorvers == 0))
2844 		error = 0;
2845 	return (error);
2846 }
2847 
2848 /*
2849  * nfs remove directory call
2850  */
2851 int
2852 nfsrpc_rmdir(vnode_t dvp, char *name, int namelen, struct ucred *cred,
2853     NFSPROC_T *p, struct nfsvattr *dnap, int *dattrflagp, void *dstuff)
2854 {
2855 	struct nfsrv_descript nfsd, *nd = &nfsd;
2856 	int error = 0;
2857 
2858 	*dattrflagp = 0;
2859 	if (namelen > NFS_MAXNAMLEN)
2860 		return (ENAMETOOLONG);
2861 	NFSCL_REQSTART(nd, NFSPROC_RMDIR, dvp);
2862 	(void) nfsm_strtom(nd, name, namelen);
2863 	error = nfscl_request(nd, dvp, p, cred, dstuff);
2864 	if (error)
2865 		return (error);
2866 	if (nd->nd_flag & (ND_NFSV3 | ND_NFSV4))
2867 		error = nfscl_wcc_data(nd, dvp, dnap, dattrflagp, NULL, dstuff);
2868 	if (nd->nd_repstat && !error)
2869 		error = nd->nd_repstat;
2870 	m_freem(nd->nd_mrep);
2871 	/*
2872 	 * Kludge: Map ENOENT => 0 assuming that you have a reply to a retry.
2873 	 */
2874 	if (error == ENOENT)
2875 		error = 0;
2876 	return (error);
2877 }
2878 
2879 /*
2880  * Readdir rpc.
2881  * Always returns with either uio_resid unchanged, if you are at the
2882  * end of the directory, or uio_resid == 0, with all DIRBLKSIZ chunks
2883  * filled in.
2884  * I felt this would allow caching of directory blocks more easily
2885  * than returning a pertially filled block.
2886  * Directory offset cookies:
2887  * Oh my, what to do with them...
2888  * I can think of three ways to deal with them:
2889  * 1 - have the layer above these RPCs maintain a map between logical
2890  *     directory byte offsets and the NFS directory offset cookies
2891  * 2 - pass the opaque directory offset cookies up into userland
2892  *     and let the libc functions deal with them, via the system call
2893  * 3 - return them to userland in the "struct dirent", so future versions
2894  *     of libc can use them and do whatever is necessary to make things work
2895  *     above these rpc calls, in the meantime
2896  * For now, I do #3 by "hiding" the directory offset cookies after the
2897  * d_name field in struct dirent. This is space inside d_reclen that
2898  * will be ignored by anything that doesn't know about them.
2899  * The directory offset cookies are filled in as the last 8 bytes of
2900  * each directory entry, after d_name. Someday, the userland libc
2901  * functions may be able to use these. In the meantime, it satisfies
2902  * OpenBSD's requirements for cookies being returned.
2903  * If expects the directory offset cookie for the read to be in uio_offset
2904  * and returns the one for the next entry after this directory block in
2905  * there, as well.
2906  */
2907 int
2908 nfsrpc_readdir(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
2909     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
2910     int *eofp, void *stuff)
2911 {
2912 	int len, left;
2913 	struct dirent *dp = NULL;
2914 	u_int32_t *tl;
2915 	nfsquad_t cookie, ncookie;
2916 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
2917 	struct nfsnode *dnp = VTONFS(vp);
2918 	struct nfsvattr nfsva;
2919 	struct nfsrv_descript nfsd, *nd = &nfsd;
2920 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
2921 	int reqsize, tryformoredirs = 1, readsize, eof = 0, gotmnton = 0;
2922 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
2923 	char *cp;
2924 	nfsattrbit_t attrbits, dattrbits;
2925 	u_int32_t rderr, *tl2 = NULL;
2926 	size_t tresid;
2927 
2928 	KASSERT(uiop->uio_iovcnt == 1 &&
2929 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
2930 	    ("nfs readdirrpc bad uio"));
2931 	ncookie.lval[0] = ncookie.lval[1] = 0;
2932 	/*
2933 	 * There is no point in reading a lot more than uio_resid, however
2934 	 * adding one additional DIRBLKSIZ makes sense. Since uio_resid
2935 	 * and nm_readdirsize are both exact multiples of DIRBLKSIZ, this
2936 	 * will never make readsize > nm_readdirsize.
2937 	 */
2938 	readsize = nmp->nm_readdirsize;
2939 	if (readsize > uiop->uio_resid)
2940 		readsize = uiop->uio_resid + DIRBLKSIZ;
2941 
2942 	*attrflagp = 0;
2943 	if (eofp)
2944 		*eofp = 0;
2945 	tresid = uiop->uio_resid;
2946 	cookie.lval[0] = cookiep->nfsuquad[0];
2947 	cookie.lval[1] = cookiep->nfsuquad[1];
2948 	nd->nd_mrep = NULL;
2949 
2950 	/*
2951 	 * For NFSv4, first create the "." and ".." entries.
2952 	 */
2953 	if (NFSHASNFSV4(nmp)) {
2954 		reqsize = 6 * NFSX_UNSIGNED;
2955 		NFSGETATTR_ATTRBIT(&dattrbits);
2956 		NFSZERO_ATTRBIT(&attrbits);
2957 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
2958 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TYPE);
2959 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
2960 		    NFSATTRBIT_MOUNTEDONFILEID)) {
2961 			NFSSETBIT_ATTRBIT(&attrbits,
2962 			    NFSATTRBIT_MOUNTEDONFILEID);
2963 			gotmnton = 1;
2964 		} else {
2965 			/*
2966 			 * Must fake it. Use the fileno, except when the
2967 			 * fsid is != to that of the directory. For that
2968 			 * case, generate a fake fileno that is not the same.
2969 			 */
2970 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
2971 			gotmnton = 0;
2972 		}
2973 
2974 		/*
2975 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
2976 		 */
2977 		if (uiop->uio_offset == 0) {
2978 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
2979 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2980 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
2981 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
2982 			(void) nfsrv_putattrbit(nd, &attrbits);
2983 			error = nfscl_request(nd, vp, p, cred, stuff);
2984 			if (error)
2985 			    return (error);
2986 			dotfileid = 0;	/* Fake out the compiler. */
2987 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
2988 			    error = nfsm_loadattr(nd, &nfsva);
2989 			    if (error != 0)
2990 				goto nfsmout;
2991 			    dotfileid = nfsva.na_fileid;
2992 			}
2993 			if (nd->nd_repstat == 0) {
2994 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2995 			    len = fxdr_unsigned(int, *(tl + 4));
2996 			    if (len > 0 && len <= NFSX_V4FHMAX)
2997 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
2998 			    else
2999 				error = EPERM;
3000 			    if (!error) {
3001 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3002 				nfsva.na_mntonfileno = UINT64_MAX;
3003 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3004 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3005 				    NULL, NULL, NULL, p, cred);
3006 				if (error) {
3007 				    dotdotfileid = dotfileid;
3008 				} else if (gotmnton) {
3009 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3010 					dotdotfileid = nfsva.na_mntonfileno;
3011 				    else
3012 					dotdotfileid = nfsva.na_fileid;
3013 				} else if (nfsva.na_filesid[0] ==
3014 				    dnp->n_vattr.na_filesid[0] &&
3015 				    nfsva.na_filesid[1] ==
3016 				    dnp->n_vattr.na_filesid[1]) {
3017 				    dotdotfileid = nfsva.na_fileid;
3018 				} else {
3019 				    do {
3020 					fakefileno--;
3021 				    } while (fakefileno ==
3022 					nfsva.na_fileid);
3023 				    dotdotfileid = fakefileno;
3024 				}
3025 			    }
3026 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3027 			    /*
3028 			     * Lookupp returns NFSERR_NOENT when we are
3029 			     * at the root, so just use the current dir.
3030 			     */
3031 			    nd->nd_repstat = 0;
3032 			    dotdotfileid = dotfileid;
3033 			} else {
3034 			    error = nd->nd_repstat;
3035 			}
3036 			m_freem(nd->nd_mrep);
3037 			if (error)
3038 			    return (error);
3039 			nd->nd_mrep = NULL;
3040 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3041 			dp->d_pad0 = dp->d_pad1 = 0;
3042 			dp->d_off = 0;
3043 			dp->d_type = DT_DIR;
3044 			dp->d_fileno = dotfileid;
3045 			dp->d_namlen = 1;
3046 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3047 			dp->d_name[0] = '.';
3048 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3049 			/*
3050 			 * Just make these offset cookie 0.
3051 			 */
3052 			tl = (u_int32_t *)&dp->d_name[8];
3053 			*tl++ = 0;
3054 			*tl = 0;
3055 			blksiz += dp->d_reclen;
3056 			uiop->uio_resid -= dp->d_reclen;
3057 			uiop->uio_offset += dp->d_reclen;
3058 			uiop->uio_iov->iov_base =
3059 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3060 			uiop->uio_iov->iov_len -= dp->d_reclen;
3061 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3062 			dp->d_pad0 = dp->d_pad1 = 0;
3063 			dp->d_off = 0;
3064 			dp->d_type = DT_DIR;
3065 			dp->d_fileno = dotdotfileid;
3066 			dp->d_namlen = 2;
3067 			*((uint64_t *)dp->d_name) = 0;
3068 			dp->d_name[0] = '.';
3069 			dp->d_name[1] = '.';
3070 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3071 			/*
3072 			 * Just make these offset cookie 0.
3073 			 */
3074 			tl = (u_int32_t *)&dp->d_name[8];
3075 			*tl++ = 0;
3076 			*tl = 0;
3077 			blksiz += dp->d_reclen;
3078 			uiop->uio_resid -= dp->d_reclen;
3079 			uiop->uio_offset += dp->d_reclen;
3080 			uiop->uio_iov->iov_base =
3081 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3082 			uiop->uio_iov->iov_len -= dp->d_reclen;
3083 		}
3084 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_RDATTRERROR);
3085 	} else {
3086 		reqsize = 5 * NFSX_UNSIGNED;
3087 	}
3088 
3089 
3090 	/*
3091 	 * Loop around doing readdir rpc's of size readsize.
3092 	 * The stopping criteria is EOF or buffer full.
3093 	 */
3094 	while (more_dirs && bigenough) {
3095 		*attrflagp = 0;
3096 		NFSCL_REQSTART(nd, NFSPROC_READDIR, vp);
3097 		if (nd->nd_flag & ND_NFSV2) {
3098 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3099 			*tl++ = cookie.lval[1];
3100 			*tl = txdr_unsigned(readsize);
3101 		} else {
3102 			NFSM_BUILD(tl, u_int32_t *, reqsize);
3103 			*tl++ = cookie.lval[0];
3104 			*tl++ = cookie.lval[1];
3105 			if (cookie.qval == 0) {
3106 				*tl++ = 0;
3107 				*tl++ = 0;
3108 			} else {
3109 				NFSLOCKNODE(dnp);
3110 				*tl++ = dnp->n_cookieverf.nfsuquad[0];
3111 				*tl++ = dnp->n_cookieverf.nfsuquad[1];
3112 				NFSUNLOCKNODE(dnp);
3113 			}
3114 			if (nd->nd_flag & ND_NFSV4) {
3115 				*tl++ = txdr_unsigned(readsize);
3116 				*tl = txdr_unsigned(readsize);
3117 				(void) nfsrv_putattrbit(nd, &attrbits);
3118 				NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3119 				*tl = txdr_unsigned(NFSV4OP_GETATTR);
3120 				(void) nfsrv_putattrbit(nd, &dattrbits);
3121 			} else {
3122 				*tl = txdr_unsigned(readsize);
3123 			}
3124 		}
3125 		error = nfscl_request(nd, vp, p, cred, stuff);
3126 		if (error)
3127 			return (error);
3128 		if (!(nd->nd_flag & ND_NFSV2)) {
3129 			if (nd->nd_flag & ND_NFSV3)
3130 				error = nfscl_postop_attr(nd, nap, attrflagp,
3131 				    stuff);
3132 			if (!nd->nd_repstat && !error) {
3133 				NFSM_DISSECT(tl, u_int32_t *, NFSX_HYPER);
3134 				NFSLOCKNODE(dnp);
3135 				dnp->n_cookieverf.nfsuquad[0] = *tl++;
3136 				dnp->n_cookieverf.nfsuquad[1] = *tl;
3137 				NFSUNLOCKNODE(dnp);
3138 			}
3139 		}
3140 		if (nd->nd_repstat || error) {
3141 			if (!error)
3142 				error = nd->nd_repstat;
3143 			goto nfsmout;
3144 		}
3145 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3146 		more_dirs = fxdr_unsigned(int, *tl);
3147 		if (!more_dirs)
3148 			tryformoredirs = 0;
3149 
3150 		/* loop through the dir entries, doctoring them to 4bsd form */
3151 		while (more_dirs && bigenough) {
3152 			if (nd->nd_flag & ND_NFSV4) {
3153 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3154 				ncookie.lval[0] = *tl++;
3155 				ncookie.lval[1] = *tl++;
3156 				len = fxdr_unsigned(int, *tl);
3157 			} else if (nd->nd_flag & ND_NFSV3) {
3158 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3159 				nfsva.na_fileid = fxdr_hyper(tl);
3160 				tl += 2;
3161 				len = fxdr_unsigned(int, *tl);
3162 			} else {
3163 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3164 				nfsva.na_fileid = fxdr_unsigned(uint64_t,
3165 				    *tl++);
3166 				len = fxdr_unsigned(int, *tl);
3167 			}
3168 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3169 				error = EBADRPC;
3170 				goto nfsmout;
3171 			}
3172 			tlen = roundup2(len, 8);
3173 			if (tlen == len)
3174 				tlen += 8;  /* To ensure null termination. */
3175 			left = DIRBLKSIZ - blksiz;
3176 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3177 				NFSBZERO(uiop->uio_iov->iov_base, left);
3178 				dp->d_reclen += left;
3179 				uiop->uio_iov->iov_base =
3180 				    (char *)uiop->uio_iov->iov_base + left;
3181 				uiop->uio_iov->iov_len -= left;
3182 				uiop->uio_resid -= left;
3183 				uiop->uio_offset += left;
3184 				blksiz = 0;
3185 			}
3186 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3187 			    uiop->uio_resid)
3188 				bigenough = 0;
3189 			if (bigenough) {
3190 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3191 				dp->d_pad0 = dp->d_pad1 = 0;
3192 				dp->d_off = 0;
3193 				dp->d_namlen = len;
3194 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3195 				    NFSX_HYPER;
3196 				dp->d_type = DT_UNKNOWN;
3197 				blksiz += dp->d_reclen;
3198 				if (blksiz == DIRBLKSIZ)
3199 					blksiz = 0;
3200 				uiop->uio_resid -= DIRHDSIZ;
3201 				uiop->uio_offset += DIRHDSIZ;
3202 				uiop->uio_iov->iov_base =
3203 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3204 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3205 				error = nfsm_mbufuio(nd, uiop, len);
3206 				if (error)
3207 					goto nfsmout;
3208 				cp = uiop->uio_iov->iov_base;
3209 				tlen -= len;
3210 				NFSBZERO(cp, tlen);
3211 				cp += tlen;	/* points to cookie storage */
3212 				tl2 = (u_int32_t *)cp;
3213 				uiop->uio_iov->iov_base =
3214 				    (char *)uiop->uio_iov->iov_base + tlen +
3215 				    NFSX_HYPER;
3216 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3217 				uiop->uio_resid -= tlen + NFSX_HYPER;
3218 				uiop->uio_offset += (tlen + NFSX_HYPER);
3219 			} else {
3220 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3221 				if (error)
3222 					goto nfsmout;
3223 			}
3224 			if (nd->nd_flag & ND_NFSV4) {
3225 				rderr = 0;
3226 				nfsva.na_mntonfileno = UINT64_MAX;
3227 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3228 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3229 				    NULL, NULL, &rderr, p, cred);
3230 				if (error)
3231 					goto nfsmout;
3232 				NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3233 			} else if (nd->nd_flag & ND_NFSV3) {
3234 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3235 				ncookie.lval[0] = *tl++;
3236 				ncookie.lval[1] = *tl++;
3237 			} else {
3238 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3239 				ncookie.lval[0] = 0;
3240 				ncookie.lval[1] = *tl++;
3241 			}
3242 			if (bigenough) {
3243 			    if (nd->nd_flag & ND_NFSV4) {
3244 				if (rderr) {
3245 				    dp->d_fileno = 0;
3246 				} else {
3247 				    if (gotmnton) {
3248 					if (nfsva.na_mntonfileno != UINT64_MAX)
3249 					    dp->d_fileno = nfsva.na_mntonfileno;
3250 					else
3251 					    dp->d_fileno = nfsva.na_fileid;
3252 				    } else if (nfsva.na_filesid[0] ==
3253 					dnp->n_vattr.na_filesid[0] &&
3254 					nfsva.na_filesid[1] ==
3255 					dnp->n_vattr.na_filesid[1]) {
3256 					dp->d_fileno = nfsva.na_fileid;
3257 				    } else {
3258 					do {
3259 					    fakefileno--;
3260 					} while (fakefileno ==
3261 					    nfsva.na_fileid);
3262 					dp->d_fileno = fakefileno;
3263 				    }
3264 				    dp->d_type = vtonfs_dtype(nfsva.na_type);
3265 				}
3266 			    } else {
3267 				dp->d_fileno = nfsva.na_fileid;
3268 			    }
3269 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3270 				ncookie.lval[0];
3271 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3272 				ncookie.lval[1];
3273 			}
3274 			more_dirs = fxdr_unsigned(int, *tl);
3275 		}
3276 		/*
3277 		 * If at end of rpc data, get the eof boolean
3278 		 */
3279 		if (!more_dirs) {
3280 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3281 			eof = fxdr_unsigned(int, *tl);
3282 			if (tryformoredirs)
3283 				more_dirs = !eof;
3284 			if (nd->nd_flag & ND_NFSV4) {
3285 				error = nfscl_postop_attr(nd, nap, attrflagp,
3286 				    stuff);
3287 				if (error)
3288 					goto nfsmout;
3289 			}
3290 		}
3291 		m_freem(nd->nd_mrep);
3292 		nd->nd_mrep = NULL;
3293 	}
3294 	/*
3295 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3296 	 * by increasing d_reclen for the last record.
3297 	 */
3298 	if (blksiz > 0) {
3299 		left = DIRBLKSIZ - blksiz;
3300 		NFSBZERO(uiop->uio_iov->iov_base, left);
3301 		dp->d_reclen += left;
3302 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3303 		    left;
3304 		uiop->uio_iov->iov_len -= left;
3305 		uiop->uio_resid -= left;
3306 		uiop->uio_offset += left;
3307 	}
3308 
3309 	/*
3310 	 * If returning no data, assume end of file.
3311 	 * If not bigenough, return not end of file, since you aren't
3312 	 *    returning all the data
3313 	 * Otherwise, return the eof flag from the server.
3314 	 */
3315 	if (eofp) {
3316 		if (tresid == ((size_t)(uiop->uio_resid)))
3317 			*eofp = 1;
3318 		else if (!bigenough)
3319 			*eofp = 0;
3320 		else
3321 			*eofp = eof;
3322 	}
3323 
3324 	/*
3325 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3326 	 */
3327 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3328 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3329 		NFSBZERO(dp, DIRBLKSIZ);
3330 		dp->d_type = DT_UNKNOWN;
3331 		tl = (u_int32_t *)&dp->d_name[4];
3332 		*tl++ = cookie.lval[0];
3333 		*tl = cookie.lval[1];
3334 		dp->d_reclen = DIRBLKSIZ;
3335 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3336 		    DIRBLKSIZ;
3337 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3338 		uiop->uio_resid -= DIRBLKSIZ;
3339 		uiop->uio_offset += DIRBLKSIZ;
3340 	}
3341 
3342 nfsmout:
3343 	if (nd->nd_mrep != NULL)
3344 		m_freem(nd->nd_mrep);
3345 	return (error);
3346 }
3347 
3348 #ifndef APPLE
3349 /*
3350  * NFS V3 readdir plus RPC. Used in place of nfsrpc_readdir().
3351  * (Also used for NFS V4 when mount flag set.)
3352  * (ditto above w.r.t. multiple of DIRBLKSIZ, etc.)
3353  */
3354 int
3355 nfsrpc_readdirplus(vnode_t vp, struct uio *uiop, nfsuint64 *cookiep,
3356     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
3357     int *eofp, void *stuff)
3358 {
3359 	int len, left;
3360 	struct dirent *dp = NULL;
3361 	u_int32_t *tl;
3362 	vnode_t newvp = NULLVP;
3363 	struct nfsrv_descript nfsd, *nd = &nfsd;
3364 	struct nameidata nami, *ndp = &nami;
3365 	struct componentname *cnp = &ndp->ni_cnd;
3366 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3367 	struct nfsnode *dnp = VTONFS(vp), *np;
3368 	struct nfsvattr nfsva;
3369 	struct nfsfh *nfhp;
3370 	nfsquad_t cookie, ncookie;
3371 	int error = 0, tlen, more_dirs = 1, blksiz = 0, bigenough = 1;
3372 	int attrflag, tryformoredirs = 1, eof = 0, gotmnton = 0;
3373 	int isdotdot = 0, unlocknewvp = 0;
3374 	u_int64_t dotfileid, dotdotfileid = 0, fakefileno = UINT64_MAX;
3375 	u_int64_t fileno = 0;
3376 	char *cp;
3377 	nfsattrbit_t attrbits, dattrbits;
3378 	size_t tresid;
3379 	u_int32_t *tl2 = NULL, rderr;
3380 	struct timespec dctime;
3381 
3382 	KASSERT(uiop->uio_iovcnt == 1 &&
3383 	    (uiop->uio_resid & (DIRBLKSIZ - 1)) == 0,
3384 	    ("nfs readdirplusrpc bad uio"));
3385 	ncookie.lval[0] = ncookie.lval[1] = 0;
3386 	timespecclear(&dctime);
3387 	*attrflagp = 0;
3388 	if (eofp != NULL)
3389 		*eofp = 0;
3390 	ndp->ni_dvp = vp;
3391 	nd->nd_mrep = NULL;
3392 	cookie.lval[0] = cookiep->nfsuquad[0];
3393 	cookie.lval[1] = cookiep->nfsuquad[1];
3394 	tresid = uiop->uio_resid;
3395 
3396 	/*
3397 	 * For NFSv4, first create the "." and ".." entries.
3398 	 */
3399 	if (NFSHASNFSV4(nmp)) {
3400 		NFSGETATTR_ATTRBIT(&dattrbits);
3401 		NFSZERO_ATTRBIT(&attrbits);
3402 		NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FILEID);
3403 		if (NFSISSET_ATTRBIT(&dnp->n_vattr.na_suppattr,
3404 		    NFSATTRBIT_MOUNTEDONFILEID)) {
3405 			NFSSETBIT_ATTRBIT(&attrbits,
3406 			    NFSATTRBIT_MOUNTEDONFILEID);
3407 			gotmnton = 1;
3408 		} else {
3409 			/*
3410 			 * Must fake it. Use the fileno, except when the
3411 			 * fsid is != to that of the directory. For that
3412 			 * case, generate a fake fileno that is not the same.
3413 			 */
3414 			NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_FSID);
3415 			gotmnton = 0;
3416 		}
3417 
3418 		/*
3419 		 * Joy, oh joy. For V4 we get to hand craft '.' and '..'.
3420 		 */
3421 		if (uiop->uio_offset == 0) {
3422 			NFSCL_REQSTART(nd, NFSPROC_LOOKUPP, vp);
3423 			NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3424 			*tl++ = txdr_unsigned(NFSV4OP_GETFH);
3425 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3426 			(void) nfsrv_putattrbit(nd, &attrbits);
3427 			error = nfscl_request(nd, vp, p, cred, stuff);
3428 			if (error)
3429 			    return (error);
3430 			dotfileid = 0;	/* Fake out the compiler. */
3431 			if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
3432 			    error = nfsm_loadattr(nd, &nfsva);
3433 			    if (error != 0)
3434 				goto nfsmout;
3435 			    dctime = nfsva.na_ctime;
3436 			    dotfileid = nfsva.na_fileid;
3437 			}
3438 			if (nd->nd_repstat == 0) {
3439 			    NFSM_DISSECT(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
3440 			    len = fxdr_unsigned(int, *(tl + 4));
3441 			    if (len > 0 && len <= NFSX_V4FHMAX)
3442 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3443 			    else
3444 				error = EPERM;
3445 			    if (!error) {
3446 				NFSM_DISSECT(tl, u_int32_t *, 2*NFSX_UNSIGNED);
3447 				nfsva.na_mntonfileno = UINT64_MAX;
3448 				error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
3449 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3450 				    NULL, NULL, NULL, p, cred);
3451 				if (error) {
3452 				    dotdotfileid = dotfileid;
3453 				} else if (gotmnton) {
3454 				    if (nfsva.na_mntonfileno != UINT64_MAX)
3455 					dotdotfileid = nfsva.na_mntonfileno;
3456 				    else
3457 					dotdotfileid = nfsva.na_fileid;
3458 				} else if (nfsva.na_filesid[0] ==
3459 				    dnp->n_vattr.na_filesid[0] &&
3460 				    nfsva.na_filesid[1] ==
3461 				    dnp->n_vattr.na_filesid[1]) {
3462 				    dotdotfileid = nfsva.na_fileid;
3463 				} else {
3464 				    do {
3465 					fakefileno--;
3466 				    } while (fakefileno ==
3467 					nfsva.na_fileid);
3468 				    dotdotfileid = fakefileno;
3469 				}
3470 			    }
3471 			} else if (nd->nd_repstat == NFSERR_NOENT) {
3472 			    /*
3473 			     * Lookupp returns NFSERR_NOENT when we are
3474 			     * at the root, so just use the current dir.
3475 			     */
3476 			    nd->nd_repstat = 0;
3477 			    dotdotfileid = dotfileid;
3478 			} else {
3479 			    error = nd->nd_repstat;
3480 			}
3481 			m_freem(nd->nd_mrep);
3482 			if (error)
3483 			    return (error);
3484 			nd->nd_mrep = NULL;
3485 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3486 			dp->d_pad0 = dp->d_pad1 = 0;
3487 			dp->d_off = 0;
3488 			dp->d_type = DT_DIR;
3489 			dp->d_fileno = dotfileid;
3490 			dp->d_namlen = 1;
3491 			*((uint64_t *)dp->d_name) = 0;	/* Zero pad it. */
3492 			dp->d_name[0] = '.';
3493 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3494 			/*
3495 			 * Just make these offset cookie 0.
3496 			 */
3497 			tl = (u_int32_t *)&dp->d_name[8];
3498 			*tl++ = 0;
3499 			*tl = 0;
3500 			blksiz += dp->d_reclen;
3501 			uiop->uio_resid -= dp->d_reclen;
3502 			uiop->uio_offset += dp->d_reclen;
3503 			uiop->uio_iov->iov_base =
3504 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3505 			uiop->uio_iov->iov_len -= dp->d_reclen;
3506 			dp = (struct dirent *)uiop->uio_iov->iov_base;
3507 			dp->d_pad0 = dp->d_pad1 = 0;
3508 			dp->d_off = 0;
3509 			dp->d_type = DT_DIR;
3510 			dp->d_fileno = dotdotfileid;
3511 			dp->d_namlen = 2;
3512 			*((uint64_t *)dp->d_name) = 0;
3513 			dp->d_name[0] = '.';
3514 			dp->d_name[1] = '.';
3515 			dp->d_reclen = _GENERIC_DIRSIZ(dp) + NFSX_HYPER;
3516 			/*
3517 			 * Just make these offset cookie 0.
3518 			 */
3519 			tl = (u_int32_t *)&dp->d_name[8];
3520 			*tl++ = 0;
3521 			*tl = 0;
3522 			blksiz += dp->d_reclen;
3523 			uiop->uio_resid -= dp->d_reclen;
3524 			uiop->uio_offset += dp->d_reclen;
3525 			uiop->uio_iov->iov_base =
3526 			    (char *)uiop->uio_iov->iov_base + dp->d_reclen;
3527 			uiop->uio_iov->iov_len -= dp->d_reclen;
3528 		}
3529 		NFSREADDIRPLUS_ATTRBIT(&attrbits);
3530 		if (gotmnton)
3531 			NFSSETBIT_ATTRBIT(&attrbits,
3532 			    NFSATTRBIT_MOUNTEDONFILEID);
3533 	}
3534 
3535 	/*
3536 	 * Loop around doing readdir rpc's of size nm_readdirsize.
3537 	 * The stopping criteria is EOF or buffer full.
3538 	 */
3539 	while (more_dirs && bigenough) {
3540 		*attrflagp = 0;
3541 		NFSCL_REQSTART(nd, NFSPROC_READDIRPLUS, vp);
3542  		NFSM_BUILD(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
3543 		*tl++ = cookie.lval[0];
3544 		*tl++ = cookie.lval[1];
3545 		if (cookie.qval == 0) {
3546 			*tl++ = 0;
3547 			*tl++ = 0;
3548 		} else {
3549 			NFSLOCKNODE(dnp);
3550 			*tl++ = dnp->n_cookieverf.nfsuquad[0];
3551 			*tl++ = dnp->n_cookieverf.nfsuquad[1];
3552 			NFSUNLOCKNODE(dnp);
3553 		}
3554 		*tl++ = txdr_unsigned(nmp->nm_readdirsize);
3555 		*tl = txdr_unsigned(nmp->nm_readdirsize);
3556 		if (nd->nd_flag & ND_NFSV4) {
3557 			(void) nfsrv_putattrbit(nd, &attrbits);
3558 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3559 			*tl = txdr_unsigned(NFSV4OP_GETATTR);
3560 			(void) nfsrv_putattrbit(nd, &dattrbits);
3561 		}
3562 		error = nfscl_request(nd, vp, p, cred, stuff);
3563 		if (error)
3564 			return (error);
3565 		if (nd->nd_flag & ND_NFSV3)
3566 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3567 		if (nd->nd_repstat || error) {
3568 			if (!error)
3569 				error = nd->nd_repstat;
3570 			goto nfsmout;
3571 		}
3572 		if ((nd->nd_flag & ND_NFSV3) != 0 && *attrflagp != 0)
3573 			dctime = nap->na_ctime;
3574 		NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3575 		NFSLOCKNODE(dnp);
3576 		dnp->n_cookieverf.nfsuquad[0] = *tl++;
3577 		dnp->n_cookieverf.nfsuquad[1] = *tl++;
3578 		NFSUNLOCKNODE(dnp);
3579 		more_dirs = fxdr_unsigned(int, *tl);
3580 		if (!more_dirs)
3581 			tryformoredirs = 0;
3582 
3583 		/* loop through the dir entries, doctoring them to 4bsd form */
3584 		while (more_dirs && bigenough) {
3585 			NFSM_DISSECT(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3586 			if (nd->nd_flag & ND_NFSV4) {
3587 				ncookie.lval[0] = *tl++;
3588 				ncookie.lval[1] = *tl++;
3589 			} else {
3590 				fileno = fxdr_hyper(tl);
3591 				tl += 2;
3592 			}
3593 			len = fxdr_unsigned(int, *tl);
3594 			if (len <= 0 || len > NFS_MAXNAMLEN) {
3595 				error = EBADRPC;
3596 				goto nfsmout;
3597 			}
3598 			tlen = roundup2(len, 8);
3599 			if (tlen == len)
3600 				tlen += 8;  /* To ensure null termination. */
3601 			left = DIRBLKSIZ - blksiz;
3602 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER > left) {
3603 				NFSBZERO(uiop->uio_iov->iov_base, left);
3604 				dp->d_reclen += left;
3605 				uiop->uio_iov->iov_base =
3606 				    (char *)uiop->uio_iov->iov_base + left;
3607 				uiop->uio_iov->iov_len -= left;
3608 				uiop->uio_resid -= left;
3609 				uiop->uio_offset += left;
3610 				blksiz = 0;
3611 			}
3612 			if (_GENERIC_DIRLEN(len) + NFSX_HYPER >
3613 			    uiop->uio_resid)
3614 				bigenough = 0;
3615 			if (bigenough) {
3616 				dp = (struct dirent *)uiop->uio_iov->iov_base;
3617 				dp->d_pad0 = dp->d_pad1 = 0;
3618 				dp->d_off = 0;
3619 				dp->d_namlen = len;
3620 				dp->d_reclen = _GENERIC_DIRLEN(len) +
3621 				    NFSX_HYPER;
3622 				dp->d_type = DT_UNKNOWN;
3623 				blksiz += dp->d_reclen;
3624 				if (blksiz == DIRBLKSIZ)
3625 					blksiz = 0;
3626 				uiop->uio_resid -= DIRHDSIZ;
3627 				uiop->uio_offset += DIRHDSIZ;
3628 				uiop->uio_iov->iov_base =
3629 				    (char *)uiop->uio_iov->iov_base + DIRHDSIZ;
3630 				uiop->uio_iov->iov_len -= DIRHDSIZ;
3631 				cnp->cn_nameptr = uiop->uio_iov->iov_base;
3632 				cnp->cn_namelen = len;
3633 				NFSCNHASHZERO(cnp);
3634 				error = nfsm_mbufuio(nd, uiop, len);
3635 				if (error)
3636 					goto nfsmout;
3637 				cp = uiop->uio_iov->iov_base;
3638 				tlen -= len;
3639 				NFSBZERO(cp, tlen);
3640 				cp += tlen;	/* points to cookie storage */
3641 				tl2 = (u_int32_t *)cp;
3642 				if (len == 2 && cnp->cn_nameptr[0] == '.' &&
3643 				    cnp->cn_nameptr[1] == '.')
3644 					isdotdot = 1;
3645 				else
3646 					isdotdot = 0;
3647 				uiop->uio_iov->iov_base =
3648 				    (char *)uiop->uio_iov->iov_base + tlen +
3649 				    NFSX_HYPER;
3650 				uiop->uio_iov->iov_len -= tlen + NFSX_HYPER;
3651 				uiop->uio_resid -= tlen + NFSX_HYPER;
3652 				uiop->uio_offset += (tlen + NFSX_HYPER);
3653 			} else {
3654 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
3655 				if (error)
3656 					goto nfsmout;
3657 			}
3658 			nfhp = NULL;
3659 			if (nd->nd_flag & ND_NFSV3) {
3660 				NFSM_DISSECT(tl, u_int32_t *, 3*NFSX_UNSIGNED);
3661 				ncookie.lval[0] = *tl++;
3662 				ncookie.lval[1] = *tl++;
3663 				attrflag = fxdr_unsigned(int, *tl);
3664 				if (attrflag) {
3665 				  error = nfsm_loadattr(nd, &nfsva);
3666 				  if (error)
3667 					goto nfsmout;
3668 				}
3669 				NFSM_DISSECT(tl,u_int32_t *,NFSX_UNSIGNED);
3670 				if (*tl) {
3671 					error = nfsm_getfh(nd, &nfhp);
3672 					if (error)
3673 					    goto nfsmout;
3674 				}
3675 				if (!attrflag && nfhp != NULL) {
3676 					free(nfhp, M_NFSFH);
3677 					nfhp = NULL;
3678 				}
3679 			} else {
3680 				rderr = 0;
3681 				nfsva.na_mntonfileno = 0xffffffff;
3682 				error = nfsv4_loadattr(nd, NULL, &nfsva, &nfhp,
3683 				    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
3684 				    NULL, NULL, &rderr, p, cred);
3685 				if (error)
3686 					goto nfsmout;
3687 			}
3688 
3689 			if (bigenough) {
3690 			    if (nd->nd_flag & ND_NFSV4) {
3691 				if (rderr) {
3692 				    dp->d_fileno = 0;
3693 				} else if (gotmnton) {
3694 				    if (nfsva.na_mntonfileno != 0xffffffff)
3695 					dp->d_fileno = nfsva.na_mntonfileno;
3696 				    else
3697 					dp->d_fileno = nfsva.na_fileid;
3698 				} else if (nfsva.na_filesid[0] ==
3699 				    dnp->n_vattr.na_filesid[0] &&
3700 				    nfsva.na_filesid[1] ==
3701 				    dnp->n_vattr.na_filesid[1]) {
3702 				    dp->d_fileno = nfsva.na_fileid;
3703 				} else {
3704 				    do {
3705 					fakefileno--;
3706 				    } while (fakefileno ==
3707 					nfsva.na_fileid);
3708 				    dp->d_fileno = fakefileno;
3709 				}
3710 			    } else {
3711 				dp->d_fileno = fileno;
3712 			    }
3713 			    *tl2++ = cookiep->nfsuquad[0] = cookie.lval[0] =
3714 				ncookie.lval[0];
3715 			    *tl2 = cookiep->nfsuquad[1] = cookie.lval[1] =
3716 				ncookie.lval[1];
3717 
3718 			    if (nfhp != NULL) {
3719 				if (NFSRV_CMPFH(nfhp->nfh_fh, nfhp->nfh_len,
3720 				    dnp->n_fhp->nfh_fh, dnp->n_fhp->nfh_len)) {
3721 				    VREF(vp);
3722 				    newvp = vp;
3723 				    unlocknewvp = 0;
3724 				    free(nfhp, M_NFSFH);
3725 				    np = dnp;
3726 				} else if (isdotdot != 0) {
3727 				    /*
3728 				     * Skip doing a nfscl_nget() call for "..".
3729 				     * There's a race between acquiring the nfs
3730 				     * node here and lookups that look for the
3731 				     * directory being read (in the parent).
3732 				     * It would try to get a lock on ".." here,
3733 				     * owning the lock on the directory being
3734 				     * read. Lookup will hold the lock on ".."
3735 				     * and try to acquire the lock on the
3736 				     * directory being read.
3737 				     * If the directory is unlocked/relocked,
3738 				     * then there is a LOR with the buflock
3739 				     * vp is relocked.
3740 				     */
3741 				    free(nfhp, M_NFSFH);
3742 				} else {
3743 				    error = nfscl_nget(vnode_mount(vp), vp,
3744 				      nfhp, cnp, p, &np, NULL, LK_EXCLUSIVE);
3745 				    if (!error) {
3746 					newvp = NFSTOV(np);
3747 					unlocknewvp = 1;
3748 				    }
3749 				}
3750 				nfhp = NULL;
3751 				if (newvp != NULLVP) {
3752 				    error = nfscl_loadattrcache(&newvp,
3753 					&nfsva, NULL, NULL, 0, 0);
3754 				    if (error) {
3755 					if (unlocknewvp)
3756 					    vput(newvp);
3757 					else
3758 					    vrele(newvp);
3759 					goto nfsmout;
3760 				    }
3761 				    dp->d_type =
3762 					vtonfs_dtype(np->n_vattr.na_type);
3763 				    ndp->ni_vp = newvp;
3764 				    NFSCNHASH(cnp, HASHINIT);
3765 				    if (cnp->cn_namelen <= NCHNAMLEN &&
3766 					(newvp->v_type != VDIR ||
3767 					 dctime.tv_sec != 0)) {
3768 					cache_enter_time(ndp->ni_dvp,
3769 					    ndp->ni_vp, cnp,
3770 					    &nfsva.na_ctime,
3771 					    newvp->v_type != VDIR ? NULL :
3772 					    &dctime);
3773 				    }
3774 				    if (unlocknewvp)
3775 					vput(newvp);
3776 				    else
3777 					vrele(newvp);
3778 				    newvp = NULLVP;
3779 				}
3780 			    }
3781 			} else if (nfhp != NULL) {
3782 			    free(nfhp, M_NFSFH);
3783 			}
3784 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3785 			more_dirs = fxdr_unsigned(int, *tl);
3786 		}
3787 		/*
3788 		 * If at end of rpc data, get the eof boolean
3789 		 */
3790 		if (!more_dirs) {
3791 			NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
3792 			eof = fxdr_unsigned(int, *tl);
3793 			if (tryformoredirs)
3794 				more_dirs = !eof;
3795 			if (nd->nd_flag & ND_NFSV4) {
3796 				error = nfscl_postop_attr(nd, nap, attrflagp,
3797 				    stuff);
3798 				if (error)
3799 					goto nfsmout;
3800 			}
3801 		}
3802 		m_freem(nd->nd_mrep);
3803 		nd->nd_mrep = NULL;
3804 	}
3805 	/*
3806 	 * Fill last record, iff any, out to a multiple of DIRBLKSIZ
3807 	 * by increasing d_reclen for the last record.
3808 	 */
3809 	if (blksiz > 0) {
3810 		left = DIRBLKSIZ - blksiz;
3811 		NFSBZERO(uiop->uio_iov->iov_base, left);
3812 		dp->d_reclen += left;
3813 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3814 		    left;
3815 		uiop->uio_iov->iov_len -= left;
3816 		uiop->uio_resid -= left;
3817 		uiop->uio_offset += left;
3818 	}
3819 
3820 	/*
3821 	 * If returning no data, assume end of file.
3822 	 * If not bigenough, return not end of file, since you aren't
3823 	 *    returning all the data
3824 	 * Otherwise, return the eof flag from the server.
3825 	 */
3826 	if (eofp != NULL) {
3827 		if (tresid == uiop->uio_resid)
3828 			*eofp = 1;
3829 		else if (!bigenough)
3830 			*eofp = 0;
3831 		else
3832 			*eofp = eof;
3833 	}
3834 
3835 	/*
3836 	 * Add extra empty records to any remaining DIRBLKSIZ chunks.
3837 	 */
3838 	while (uiop->uio_resid > 0 && uiop->uio_resid != tresid) {
3839 		dp = (struct dirent *)uiop->uio_iov->iov_base;
3840 		NFSBZERO(dp, DIRBLKSIZ);
3841 		dp->d_type = DT_UNKNOWN;
3842 		tl = (u_int32_t *)&dp->d_name[4];
3843 		*tl++ = cookie.lval[0];
3844 		*tl = cookie.lval[1];
3845 		dp->d_reclen = DIRBLKSIZ;
3846 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base +
3847 		    DIRBLKSIZ;
3848 		uiop->uio_iov->iov_len -= DIRBLKSIZ;
3849 		uiop->uio_resid -= DIRBLKSIZ;
3850 		uiop->uio_offset += DIRBLKSIZ;
3851 	}
3852 
3853 nfsmout:
3854 	if (nd->nd_mrep != NULL)
3855 		m_freem(nd->nd_mrep);
3856 	return (error);
3857 }
3858 #endif	/* !APPLE */
3859 
3860 /*
3861  * Nfs commit rpc
3862  */
3863 int
3864 nfsrpc_commit(vnode_t vp, u_quad_t offset, int cnt, struct ucred *cred,
3865     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
3866 {
3867 	u_int32_t *tl;
3868 	struct nfsrv_descript nfsd, *nd = &nfsd;
3869 	nfsattrbit_t attrbits;
3870 	int error;
3871 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3872 
3873 	*attrflagp = 0;
3874 	NFSCL_REQSTART(nd, NFSPROC_COMMIT, vp);
3875 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3876 	txdr_hyper(offset, tl);
3877 	tl += 2;
3878 	*tl = txdr_unsigned(cnt);
3879 	if (nd->nd_flag & ND_NFSV4) {
3880 		/*
3881 		 * And do a Getattr op.
3882 		 */
3883 		NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
3884 		*tl = txdr_unsigned(NFSV4OP_GETATTR);
3885 		NFSGETATTR_ATTRBIT(&attrbits);
3886 		(void) nfsrv_putattrbit(nd, &attrbits);
3887 	}
3888 	error = nfscl_request(nd, vp, p, cred, stuff);
3889 	if (error)
3890 		return (error);
3891 	error = nfscl_wcc_data(nd, vp, nap, attrflagp, NULL, stuff);
3892 	if (!error && !nd->nd_repstat) {
3893 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
3894 		NFSLOCKMNT(nmp);
3895 		if (NFSBCMP(nmp->nm_verf, tl, NFSX_VERF)) {
3896 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
3897 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
3898 		}
3899 		NFSUNLOCKMNT(nmp);
3900 		if (nd->nd_flag & ND_NFSV4)
3901 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
3902 	}
3903 nfsmout:
3904 	if (!error && nd->nd_repstat)
3905 		error = nd->nd_repstat;
3906 	m_freem(nd->nd_mrep);
3907 	return (error);
3908 }
3909 
3910 /*
3911  * NFS byte range lock rpc.
3912  * (Mostly just calls one of the three lower level RPC routines.)
3913  */
3914 int
3915 nfsrpc_advlock(vnode_t vp, off_t size, int op, struct flock *fl,
3916     int reclaim, struct ucred *cred, NFSPROC_T *p, void *id, int flags)
3917 {
3918 	struct nfscllockowner *lp;
3919 	struct nfsclclient *clp;
3920 	struct nfsfh *nfhp;
3921 	struct nfsrv_descript nfsd, *nd = &nfsd;
3922 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
3923 	u_int64_t off, len;
3924 	off_t start, end;
3925 	u_int32_t clidrev = 0;
3926 	int error = 0, newone = 0, expireret = 0, retrycnt, donelocally;
3927 	int callcnt, dorpc;
3928 
3929 	/*
3930 	 * Convert the flock structure into a start and end and do POSIX
3931 	 * bounds checking.
3932 	 */
3933 	switch (fl->l_whence) {
3934 	case SEEK_SET:
3935 	case SEEK_CUR:
3936 		/*
3937 		 * Caller is responsible for adding any necessary offset
3938 		 * when SEEK_CUR is used.
3939 		 */
3940 		start = fl->l_start;
3941 		off = fl->l_start;
3942 		break;
3943 	case SEEK_END:
3944 		start = size + fl->l_start;
3945 		off = size + fl->l_start;
3946 		break;
3947 	default:
3948 		return (EINVAL);
3949 	}
3950 	if (start < 0)
3951 		return (EINVAL);
3952 	if (fl->l_len != 0) {
3953 		end = start + fl->l_len - 1;
3954 		if (end < start)
3955 			return (EINVAL);
3956 	}
3957 
3958 	len = fl->l_len;
3959 	if (len == 0)
3960 		len = NFS64BITSSET;
3961 	retrycnt = 0;
3962 	do {
3963 	    nd->nd_repstat = 0;
3964 	    if (op == F_GETLK) {
3965 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3966 		if (error)
3967 			return (error);
3968 		error = nfscl_lockt(vp, clp, off, len, fl, p, id, flags);
3969 		if (!error) {
3970 			clidrev = clp->nfsc_clientidrev;
3971 			error = nfsrpc_lockt(nd, vp, clp, off, len, fl, cred,
3972 			    p, id, flags);
3973 		} else if (error == -1) {
3974 			error = 0;
3975 		}
3976 		nfscl_clientrelease(clp);
3977 	    } else if (op == F_UNLCK && fl->l_type == F_UNLCK) {
3978 		/*
3979 		 * We must loop around for all lockowner cases.
3980 		 */
3981 		callcnt = 0;
3982 		error = nfscl_getcl(vnode_mount(vp), cred, p, 1, &clp);
3983 		if (error)
3984 			return (error);
3985 		do {
3986 		    error = nfscl_relbytelock(vp, off, len, cred, p, callcnt,
3987 			clp, id, flags, &lp, &dorpc);
3988 		    /*
3989 		     * If it returns a NULL lp, we're done.
3990 		     */
3991 		    if (lp == NULL) {
3992 			if (callcnt == 0)
3993 			    nfscl_clientrelease(clp);
3994 			else
3995 			    nfscl_releasealllocks(clp, vp, p, id, flags);
3996 			return (error);
3997 		    }
3998 		    if (nmp->nm_clp != NULL)
3999 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4000 		    else
4001 			clidrev = 0;
4002 		    /*
4003 		     * If the server doesn't support Posix lock semantics,
4004 		     * only allow locks on the entire file, since it won't
4005 		     * handle overlapping byte ranges.
4006 		     * There might still be a problem when a lock
4007 		     * upgrade/downgrade (read<->write) occurs, since the
4008 		     * server "might" expect an unlock first?
4009 		     */
4010 		    if (dorpc && (lp->nfsl_open->nfso_posixlock ||
4011 			(off == 0 && len == NFS64BITSSET))) {
4012 			/*
4013 			 * Since the lock records will go away, we must
4014 			 * wait for grace and delay here.
4015 			 */
4016 			do {
4017 			    error = nfsrpc_locku(nd, nmp, lp, off, len,
4018 				NFSV4LOCKT_READ, cred, p, 0);
4019 			    if ((nd->nd_repstat == NFSERR_GRACE ||
4020 				 nd->nd_repstat == NFSERR_DELAY) &&
4021 				error == 0)
4022 				(void) nfs_catnap(PZERO, (int)nd->nd_repstat,
4023 				    "nfs_advlock");
4024 			} while ((nd->nd_repstat == NFSERR_GRACE ||
4025 			    nd->nd_repstat == NFSERR_DELAY) && error == 0);
4026 		    }
4027 		    callcnt++;
4028 		} while (error == 0 && nd->nd_repstat == 0);
4029 		nfscl_releasealllocks(clp, vp, p, id, flags);
4030 	    } else if (op == F_SETLK) {
4031 		error = nfscl_getbytelock(vp, off, len, fl->l_type, cred, p,
4032 		    NULL, 0, id, flags, NULL, NULL, &lp, &newone, &donelocally);
4033 		if (error || donelocally) {
4034 			return (error);
4035 		}
4036 		if (nmp->nm_clp != NULL)
4037 			clidrev = nmp->nm_clp->nfsc_clientidrev;
4038 		else
4039 			clidrev = 0;
4040 		nfhp = VTONFS(vp)->n_fhp;
4041 		if (!lp->nfsl_open->nfso_posixlock &&
4042 		    (off != 0 || len != NFS64BITSSET)) {
4043 			error = EINVAL;
4044 		} else {
4045 			error = nfsrpc_lock(nd, nmp, vp, nfhp->nfh_fh,
4046 			    nfhp->nfh_len, lp, newone, reclaim, off,
4047 			    len, fl->l_type, cred, p, 0);
4048 		}
4049 		if (!error)
4050 			error = nd->nd_repstat;
4051 		nfscl_lockrelease(lp, error, newone);
4052 	    } else {
4053 		error = EINVAL;
4054 	    }
4055 	    if (!error)
4056 	        error = nd->nd_repstat;
4057 	    if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
4058 		error == NFSERR_STALEDONTRECOVER ||
4059 		error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4060 		error == NFSERR_BADSESSION) {
4061 		(void) nfs_catnap(PZERO, error, "nfs_advlock");
4062 	    } else if ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID)
4063 		&& clidrev != 0) {
4064 		expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
4065 		retrycnt++;
4066 	    }
4067 	} while (error == NFSERR_GRACE ||
4068 	    error == NFSERR_STALECLIENTID || error == NFSERR_DELAY ||
4069 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_STALESTATEID ||
4070 	    error == NFSERR_BADSESSION ||
4071 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
4072 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
4073 	if (error && retrycnt >= 4)
4074 		error = EIO;
4075 	return (error);
4076 }
4077 
4078 /*
4079  * The lower level routine for the LockT case.
4080  */
4081 int
4082 nfsrpc_lockt(struct nfsrv_descript *nd, vnode_t vp,
4083     struct nfsclclient *clp, u_int64_t off, u_int64_t len, struct flock *fl,
4084     struct ucred *cred, NFSPROC_T *p, void *id, int flags)
4085 {
4086 	u_int32_t *tl;
4087 	int error, type, size;
4088 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4089 	struct nfsnode *np;
4090 	struct nfsmount *nmp;
4091 	struct nfsclsession *tsep;
4092 
4093 	nmp = VFSTONFS(vp->v_mount);
4094 	NFSCL_REQSTART(nd, NFSPROC_LOCKT, vp);
4095 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4096 	if (fl->l_type == F_RDLCK)
4097 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4098 	else
4099 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4100 	txdr_hyper(off, tl);
4101 	tl += 2;
4102 	txdr_hyper(len, tl);
4103 	tl += 2;
4104 	tsep = nfsmnt_mdssession(nmp);
4105 	*tl++ = tsep->nfsess_clientid.lval[0];
4106 	*tl = tsep->nfsess_clientid.lval[1];
4107 	nfscl_filllockowner(id, own, flags);
4108 	np = VTONFS(vp);
4109 	NFSBCOPY(np->n_fhp->nfh_fh, &own[NFSV4CL_LOCKNAMELEN],
4110 	    np->n_fhp->nfh_len);
4111 	(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + np->n_fhp->nfh_len);
4112 	error = nfscl_request(nd, vp, p, cred, NULL);
4113 	if (error)
4114 		return (error);
4115 	if (nd->nd_repstat == 0) {
4116 		fl->l_type = F_UNLCK;
4117 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4118 		nd->nd_repstat = 0;
4119 		fl->l_whence = SEEK_SET;
4120 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4121 		fl->l_start = fxdr_hyper(tl);
4122 		tl += 2;
4123 		len = fxdr_hyper(tl);
4124 		tl += 2;
4125 		if (len == NFS64BITSSET)
4126 			fl->l_len = 0;
4127 		else
4128 			fl->l_len = len;
4129 		type = fxdr_unsigned(int, *tl++);
4130 		if (type == NFSV4LOCKT_WRITE)
4131 			fl->l_type = F_WRLCK;
4132 		else
4133 			fl->l_type = F_RDLCK;
4134 		/*
4135 		 * XXX For now, I have no idea what to do with the
4136 		 * conflicting lock_owner, so I'll just set the pid == 0
4137 		 * and skip over the lock_owner.
4138 		 */
4139 		fl->l_pid = (pid_t)0;
4140 		tl += 2;
4141 		size = fxdr_unsigned(int, *tl);
4142 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4143 			error = EBADRPC;
4144 		if (!error)
4145 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4146 	} else if (nd->nd_repstat == NFSERR_STALECLIENTID)
4147 		nfscl_initiate_recovery(clp);
4148 nfsmout:
4149 	m_freem(nd->nd_mrep);
4150 	return (error);
4151 }
4152 
4153 /*
4154  * Lower level function that performs the LockU RPC.
4155  */
4156 static int
4157 nfsrpc_locku(struct nfsrv_descript *nd, struct nfsmount *nmp,
4158     struct nfscllockowner *lp, u_int64_t off, u_int64_t len,
4159     u_int32_t type, struct ucred *cred, NFSPROC_T *p, int syscred)
4160 {
4161 	u_int32_t *tl;
4162 	int error;
4163 
4164 	nfscl_reqstart(nd, NFSPROC_LOCKU, nmp, lp->nfsl_open->nfso_fh,
4165 	    lp->nfsl_open->nfso_fhlen, NULL, NULL, 0, 0);
4166 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + 6 * NFSX_UNSIGNED);
4167 	*tl++ = txdr_unsigned(type);
4168 	*tl = txdr_unsigned(lp->nfsl_seqid);
4169 	if (nfstest_outofseq &&
4170 	    (arc4random() % nfstest_outofseq) == 0)
4171 		*tl = txdr_unsigned(lp->nfsl_seqid + 1);
4172 	tl++;
4173 	if (NFSHASNFSV4N(nmp))
4174 		*tl++ = 0;
4175 	else
4176 		*tl++ = lp->nfsl_stateid.seqid;
4177 	*tl++ = lp->nfsl_stateid.other[0];
4178 	*tl++ = lp->nfsl_stateid.other[1];
4179 	*tl++ = lp->nfsl_stateid.other[2];
4180 	txdr_hyper(off, tl);
4181 	tl += 2;
4182 	txdr_hyper(len, tl);
4183 	if (syscred)
4184 		nd->nd_flag |= ND_USEGSSNAME;
4185 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4186 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4187 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4188 	if (error)
4189 		return (error);
4190 	if (nd->nd_repstat == 0) {
4191 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4192 		lp->nfsl_stateid.seqid = *tl++;
4193 		lp->nfsl_stateid.other[0] = *tl++;
4194 		lp->nfsl_stateid.other[1] = *tl++;
4195 		lp->nfsl_stateid.other[2] = *tl;
4196 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4197 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4198 nfsmout:
4199 	m_freem(nd->nd_mrep);
4200 	return (error);
4201 }
4202 
4203 /*
4204  * The actual Lock RPC.
4205  */
4206 int
4207 nfsrpc_lock(struct nfsrv_descript *nd, struct nfsmount *nmp, vnode_t vp,
4208     u_int8_t *nfhp, int fhlen, struct nfscllockowner *lp, int newone,
4209     int reclaim, u_int64_t off, u_int64_t len, short type, struct ucred *cred,
4210     NFSPROC_T *p, int syscred)
4211 {
4212 	u_int32_t *tl;
4213 	int error, size;
4214 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4215 	struct nfsclsession *tsep;
4216 
4217 	nfscl_reqstart(nd, NFSPROC_LOCK, nmp, nfhp, fhlen, NULL, NULL, 0, 0);
4218 	NFSM_BUILD(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
4219 	if (type == F_RDLCK)
4220 		*tl++ = txdr_unsigned(NFSV4LOCKT_READ);
4221 	else
4222 		*tl++ = txdr_unsigned(NFSV4LOCKT_WRITE);
4223 	*tl++ = txdr_unsigned(reclaim);
4224 	txdr_hyper(off, tl);
4225 	tl += 2;
4226 	txdr_hyper(len, tl);
4227 	tl += 2;
4228 	if (newone) {
4229 	    *tl = newnfs_true;
4230 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID +
4231 		2 * NFSX_UNSIGNED + NFSX_HYPER);
4232 	    *tl++ = txdr_unsigned(lp->nfsl_open->nfso_own->nfsow_seqid);
4233 	    if (NFSHASNFSV4N(nmp))
4234 		*tl++ = 0;
4235 	    else
4236 		*tl++ = lp->nfsl_open->nfso_stateid.seqid;
4237 	    *tl++ = lp->nfsl_open->nfso_stateid.other[0];
4238 	    *tl++ = lp->nfsl_open->nfso_stateid.other[1];
4239 	    *tl++ = lp->nfsl_open->nfso_stateid.other[2];
4240 	    *tl++ = txdr_unsigned(lp->nfsl_seqid);
4241 	    tsep = nfsmnt_mdssession(nmp);
4242 	    *tl++ = tsep->nfsess_clientid.lval[0];
4243 	    *tl = tsep->nfsess_clientid.lval[1];
4244 	    NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4245 	    NFSBCOPY(nfhp, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4246 	    (void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4247 	} else {
4248 	    *tl = newnfs_false;
4249 	    NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID + NFSX_UNSIGNED);
4250 	    if (NFSHASNFSV4N(nmp))
4251 		*tl++ = 0;
4252 	    else
4253 		*tl++ = lp->nfsl_stateid.seqid;
4254 	    *tl++ = lp->nfsl_stateid.other[0];
4255 	    *tl++ = lp->nfsl_stateid.other[1];
4256 	    *tl++ = lp->nfsl_stateid.other[2];
4257 	    *tl = txdr_unsigned(lp->nfsl_seqid);
4258 	    if (nfstest_outofseq &&
4259 		(arc4random() % nfstest_outofseq) == 0)
4260 		    *tl = txdr_unsigned(lp->nfsl_seqid + 1);
4261 	}
4262 	if (syscred)
4263 		nd->nd_flag |= ND_USEGSSNAME;
4264 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
4265 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4266 	if (error)
4267 		return (error);
4268 	if (newone)
4269 	    NFSCL_INCRSEQID(lp->nfsl_open->nfso_own->nfsow_seqid, nd);
4270 	NFSCL_INCRSEQID(lp->nfsl_seqid, nd);
4271 	if (nd->nd_repstat == 0) {
4272 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID);
4273 		lp->nfsl_stateid.seqid = *tl++;
4274 		lp->nfsl_stateid.other[0] = *tl++;
4275 		lp->nfsl_stateid.other[1] = *tl++;
4276 		lp->nfsl_stateid.other[2] = *tl;
4277 	} else if (nd->nd_repstat == NFSERR_DENIED) {
4278 		NFSM_DISSECT(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
4279 		size = fxdr_unsigned(int, *(tl + 7));
4280 		if (size < 0 || size > NFSV4_OPAQUELIMIT)
4281 			error = EBADRPC;
4282 		if (!error)
4283 			error = nfsm_advance(nd, NFSM_RNDUP(size), -1);
4284 	} else if (nd->nd_repstat == NFSERR_STALESTATEID)
4285 		nfscl_initiate_recovery(lp->nfsl_open->nfso_own->nfsow_clp);
4286 nfsmout:
4287 	m_freem(nd->nd_mrep);
4288 	return (error);
4289 }
4290 
4291 /*
4292  * nfs statfs rpc
4293  * (always called with the vp for the mount point)
4294  */
4295 int
4296 nfsrpc_statfs(vnode_t vp, struct nfsstatfs *sbp, struct nfsfsinfo *fsp,
4297     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4298     void *stuff)
4299 {
4300 	u_int32_t *tl = NULL;
4301 	struct nfsrv_descript nfsd, *nd = &nfsd;
4302 	struct nfsmount *nmp;
4303 	nfsattrbit_t attrbits;
4304 	int error;
4305 
4306 	*attrflagp = 0;
4307 	nmp = VFSTONFS(vnode_mount(vp));
4308 	if (NFSHASNFSV4(nmp)) {
4309 		/*
4310 		 * For V4, you actually do a getattr.
4311 		 */
4312 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4313 		NFSSTATFS_GETATTRBIT(&attrbits);
4314 		(void) nfsrv_putattrbit(nd, &attrbits);
4315 		nd->nd_flag |= ND_USEGSSNAME;
4316 		error = nfscl_request(nd, vp, p, cred, stuff);
4317 		if (error)
4318 			return (error);
4319 		if (nd->nd_repstat == 0) {
4320 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4321 			    NULL, NULL, sbp, fsp, NULL, 0, NULL, NULL, NULL, p,
4322 			    cred);
4323 			if (!error) {
4324 				nmp->nm_fsid[0] = nap->na_filesid[0];
4325 				nmp->nm_fsid[1] = nap->na_filesid[1];
4326 				NFSSETHASSETFSID(nmp);
4327 				*attrflagp = 1;
4328 			}
4329 		} else {
4330 			error = nd->nd_repstat;
4331 		}
4332 		if (error)
4333 			goto nfsmout;
4334 	} else {
4335 		NFSCL_REQSTART(nd, NFSPROC_FSSTAT, vp);
4336 		error = nfscl_request(nd, vp, p, cred, stuff);
4337 		if (error)
4338 			return (error);
4339 		if (nd->nd_flag & ND_NFSV3) {
4340 			error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4341 			if (error)
4342 				goto nfsmout;
4343 		}
4344 		if (nd->nd_repstat) {
4345 			error = nd->nd_repstat;
4346 			goto nfsmout;
4347 		}
4348 		NFSM_DISSECT(tl, u_int32_t *,
4349 		    NFSX_STATFS(nd->nd_flag & ND_NFSV3));
4350 	}
4351 	if (NFSHASNFSV3(nmp)) {
4352 		sbp->sf_tbytes = fxdr_hyper(tl); tl += 2;
4353 		sbp->sf_fbytes = fxdr_hyper(tl); tl += 2;
4354 		sbp->sf_abytes = fxdr_hyper(tl); tl += 2;
4355 		sbp->sf_tfiles = fxdr_hyper(tl); tl += 2;
4356 		sbp->sf_ffiles = fxdr_hyper(tl); tl += 2;
4357 		sbp->sf_afiles = fxdr_hyper(tl); tl += 2;
4358 		sbp->sf_invarsec = fxdr_unsigned(u_int32_t, *tl);
4359 	} else if (NFSHASNFSV4(nmp) == 0) {
4360 		sbp->sf_tsize = fxdr_unsigned(u_int32_t, *tl++);
4361 		sbp->sf_bsize = fxdr_unsigned(u_int32_t, *tl++);
4362 		sbp->sf_blocks = fxdr_unsigned(u_int32_t, *tl++);
4363 		sbp->sf_bfree = fxdr_unsigned(u_int32_t, *tl++);
4364 		sbp->sf_bavail = fxdr_unsigned(u_int32_t, *tl);
4365 	}
4366 nfsmout:
4367 	m_freem(nd->nd_mrep);
4368 	return (error);
4369 }
4370 
4371 /*
4372  * nfs pathconf rpc
4373  */
4374 int
4375 nfsrpc_pathconf(vnode_t vp, struct nfsv3_pathconf *pc,
4376     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp,
4377     void *stuff)
4378 {
4379 	struct nfsrv_descript nfsd, *nd = &nfsd;
4380 	struct nfsmount *nmp;
4381 	u_int32_t *tl;
4382 	nfsattrbit_t attrbits;
4383 	int error;
4384 
4385 	*attrflagp = 0;
4386 	nmp = VFSTONFS(vnode_mount(vp));
4387 	if (NFSHASNFSV4(nmp)) {
4388 		/*
4389 		 * For V4, you actually do a getattr.
4390 		 */
4391 		NFSCL_REQSTART(nd, NFSPROC_GETATTR, vp);
4392 		NFSPATHCONF_GETATTRBIT(&attrbits);
4393 		(void) nfsrv_putattrbit(nd, &attrbits);
4394 		nd->nd_flag |= ND_USEGSSNAME;
4395 		error = nfscl_request(nd, vp, p, cred, stuff);
4396 		if (error)
4397 			return (error);
4398 		if (nd->nd_repstat == 0) {
4399 			error = nfsv4_loadattr(nd, NULL, nap, NULL, NULL, 0,
4400 			    pc, NULL, NULL, NULL, NULL, 0, NULL, NULL, NULL, p,
4401 			    cred);
4402 			if (!error)
4403 				*attrflagp = 1;
4404 		} else {
4405 			error = nd->nd_repstat;
4406 		}
4407 	} else {
4408 		NFSCL_REQSTART(nd, NFSPROC_PATHCONF, vp);
4409 		error = nfscl_request(nd, vp, p, cred, stuff);
4410 		if (error)
4411 			return (error);
4412 		error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4413 		if (nd->nd_repstat && !error)
4414 			error = nd->nd_repstat;
4415 		if (!error) {
4416 			NFSM_DISSECT(tl, u_int32_t *, NFSX_V3PATHCONF);
4417 			pc->pc_linkmax = fxdr_unsigned(u_int32_t, *tl++);
4418 			pc->pc_namemax = fxdr_unsigned(u_int32_t, *tl++);
4419 			pc->pc_notrunc = fxdr_unsigned(u_int32_t, *tl++);
4420 			pc->pc_chownrestricted =
4421 			    fxdr_unsigned(u_int32_t, *tl++);
4422 			pc->pc_caseinsensitive =
4423 			    fxdr_unsigned(u_int32_t, *tl++);
4424 			pc->pc_casepreserving = fxdr_unsigned(u_int32_t, *tl);
4425 		}
4426 	}
4427 nfsmout:
4428 	m_freem(nd->nd_mrep);
4429 	return (error);
4430 }
4431 
4432 /*
4433  * nfs version 3 fsinfo rpc call
4434  */
4435 int
4436 nfsrpc_fsinfo(vnode_t vp, struct nfsfsinfo *fsp, struct ucred *cred,
4437     NFSPROC_T *p, struct nfsvattr *nap, int *attrflagp, void *stuff)
4438 {
4439 	u_int32_t *tl;
4440 	struct nfsrv_descript nfsd, *nd = &nfsd;
4441 	int error;
4442 
4443 	*attrflagp = 0;
4444 	NFSCL_REQSTART(nd, NFSPROC_FSINFO, vp);
4445 	error = nfscl_request(nd, vp, p, cred, stuff);
4446 	if (error)
4447 		return (error);
4448 	error = nfscl_postop_attr(nd, nap, attrflagp, stuff);
4449 	if (nd->nd_repstat && !error)
4450 		error = nd->nd_repstat;
4451 	if (!error) {
4452 		NFSM_DISSECT(tl, u_int32_t *, NFSX_V3FSINFO);
4453 		fsp->fs_rtmax = fxdr_unsigned(u_int32_t, *tl++);
4454 		fsp->fs_rtpref = fxdr_unsigned(u_int32_t, *tl++);
4455 		fsp->fs_rtmult = fxdr_unsigned(u_int32_t, *tl++);
4456 		fsp->fs_wtmax = fxdr_unsigned(u_int32_t, *tl++);
4457 		fsp->fs_wtpref = fxdr_unsigned(u_int32_t, *tl++);
4458 		fsp->fs_wtmult = fxdr_unsigned(u_int32_t, *tl++);
4459 		fsp->fs_dtpref = fxdr_unsigned(u_int32_t, *tl++);
4460 		fsp->fs_maxfilesize = fxdr_hyper(tl);
4461 		tl += 2;
4462 		fxdr_nfsv3time(tl, &fsp->fs_timedelta);
4463 		tl += 2;
4464 		fsp->fs_properties = fxdr_unsigned(u_int32_t, *tl);
4465 	}
4466 nfsmout:
4467 	m_freem(nd->nd_mrep);
4468 	return (error);
4469 }
4470 
4471 /*
4472  * This function performs the Renew RPC.
4473  */
4474 int
4475 nfsrpc_renew(struct nfsclclient *clp, struct nfsclds *dsp, struct ucred *cred,
4476     NFSPROC_T *p)
4477 {
4478 	u_int32_t *tl;
4479 	struct nfsrv_descript nfsd;
4480 	struct nfsrv_descript *nd = &nfsd;
4481 	struct nfsmount *nmp;
4482 	int error;
4483 	struct nfssockreq *nrp;
4484 	struct nfsclsession *tsep;
4485 
4486 	nmp = clp->nfsc_nmp;
4487 	if (nmp == NULL)
4488 		return (0);
4489 	if (dsp == NULL)
4490 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL, NULL, 0,
4491 		    0);
4492 	else
4493 		nfscl_reqstart(nd, NFSPROC_RENEW, nmp, NULL, 0, NULL,
4494 		    &dsp->nfsclds_sess, 0, 0);
4495 	if (!NFSHASNFSV4N(nmp)) {
4496 		/* NFSv4.1 just uses a Sequence Op and not a Renew. */
4497 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4498 		tsep = nfsmnt_mdssession(nmp);
4499 		*tl++ = tsep->nfsess_clientid.lval[0];
4500 		*tl = tsep->nfsess_clientid.lval[1];
4501 	}
4502 	nrp = NULL;
4503 	if (dsp != NULL)
4504 		nrp = dsp->nfsclds_sockp;
4505 	if (nrp == NULL)
4506 		/* If NULL, use the MDS socket. */
4507 		nrp = &nmp->nm_sockreq;
4508 	nd->nd_flag |= ND_USEGSSNAME;
4509 	if (dsp == NULL)
4510 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4511 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4512 	else {
4513 		error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4514 		    NFS_PROG, NFS_VER4, NULL, 1, NULL, &dsp->nfsclds_sess);
4515 		if (error == ENXIO)
4516 			nfscl_cancelreqs(dsp);
4517 	}
4518 	if (error)
4519 		return (error);
4520 	error = nd->nd_repstat;
4521 	m_freem(nd->nd_mrep);
4522 	return (error);
4523 }
4524 
4525 /*
4526  * This function performs the Releaselockowner RPC.
4527  */
4528 int
4529 nfsrpc_rellockown(struct nfsmount *nmp, struct nfscllockowner *lp,
4530     uint8_t *fh, int fhlen, struct ucred *cred, NFSPROC_T *p)
4531 {
4532 	struct nfsrv_descript nfsd, *nd = &nfsd;
4533 	u_int32_t *tl;
4534 	int error;
4535 	uint8_t own[NFSV4CL_LOCKNAMELEN + NFSX_V4FHMAX];
4536 	struct nfsclsession *tsep;
4537 
4538 	if (NFSHASNFSV4N(nmp)) {
4539 		/* For NFSv4.1, do a FreeStateID. */
4540 		nfscl_reqstart(nd, NFSPROC_FREESTATEID, nmp, NULL, 0, NULL,
4541 		    NULL, 0, 0);
4542 		nfsm_stateidtom(nd, &lp->nfsl_stateid, NFSSTATEID_PUTSTATEID);
4543 	} else {
4544 		nfscl_reqstart(nd, NFSPROC_RELEASELCKOWN, nmp, NULL, 0, NULL,
4545 		    NULL, 0, 0);
4546 		NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
4547 		tsep = nfsmnt_mdssession(nmp);
4548 		*tl++ = tsep->nfsess_clientid.lval[0];
4549 		*tl = tsep->nfsess_clientid.lval[1];
4550 		NFSBCOPY(lp->nfsl_owner, own, NFSV4CL_LOCKNAMELEN);
4551 		NFSBCOPY(fh, &own[NFSV4CL_LOCKNAMELEN], fhlen);
4552 		(void)nfsm_strtom(nd, own, NFSV4CL_LOCKNAMELEN + fhlen);
4553 	}
4554 	nd->nd_flag |= ND_USEGSSNAME;
4555 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4556 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4557 	if (error)
4558 		return (error);
4559 	error = nd->nd_repstat;
4560 	m_freem(nd->nd_mrep);
4561 	return (error);
4562 }
4563 
4564 /*
4565  * This function performs the Compound to get the mount pt FH.
4566  */
4567 int
4568 nfsrpc_getdirpath(struct nfsmount *nmp, u_char *dirpath, struct ucred *cred,
4569     NFSPROC_T *p)
4570 {
4571 	u_int32_t *tl;
4572 	struct nfsrv_descript nfsd;
4573 	struct nfsrv_descript *nd = &nfsd;
4574 	u_char *cp, *cp2;
4575 	int error, cnt, len, setnil;
4576 	u_int32_t *opcntp;
4577 
4578 	nfscl_reqstart(nd, NFSPROC_PUTROOTFH, nmp, NULL, 0, &opcntp, NULL, 0,
4579 	    0);
4580 	cp = dirpath;
4581 	cnt = 0;
4582 	do {
4583 		setnil = 0;
4584 		while (*cp == '/')
4585 			cp++;
4586 		cp2 = cp;
4587 		while (*cp2 != '\0' && *cp2 != '/')
4588 			cp2++;
4589 		if (*cp2 == '/') {
4590 			setnil = 1;
4591 			*cp2 = '\0';
4592 		}
4593 		if (cp2 != cp) {
4594 			NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4595 			*tl = txdr_unsigned(NFSV4OP_LOOKUP);
4596 			nfsm_strtom(nd, cp, strlen(cp));
4597 			cnt++;
4598 		}
4599 		if (setnil)
4600 			*cp2++ = '/';
4601 		cp = cp2;
4602 	} while (*cp != '\0');
4603 	if (NFSHASNFSV4N(nmp))
4604 		/* Has a Sequence Op done by nfscl_reqstart(). */
4605 		*opcntp = txdr_unsigned(3 + cnt);
4606 	else
4607 		*opcntp = txdr_unsigned(2 + cnt);
4608 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
4609 	*tl = txdr_unsigned(NFSV4OP_GETFH);
4610 	nd->nd_flag |= ND_USEGSSNAME;
4611 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4612 		NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4613 	if (error)
4614 		return (error);
4615 	if (nd->nd_repstat == 0) {
4616 		NFSM_DISSECT(tl, u_int32_t *, (3 + 2 * cnt) * NFSX_UNSIGNED);
4617 		tl += (2 + 2 * cnt);
4618 		if ((len = fxdr_unsigned(int, *tl)) <= 0 ||
4619 			len > NFSX_FHMAX) {
4620 			nd->nd_repstat = NFSERR_BADXDR;
4621 		} else {
4622 			nd->nd_repstat = nfsrv_mtostr(nd, nmp->nm_fh, len);
4623 			if (nd->nd_repstat == 0)
4624 				nmp->nm_fhsize = len;
4625 		}
4626 	}
4627 	error = nd->nd_repstat;
4628 nfsmout:
4629 	m_freem(nd->nd_mrep);
4630 	return (error);
4631 }
4632 
4633 /*
4634  * This function performs the Delegreturn RPC.
4635  */
4636 int
4637 nfsrpc_delegreturn(struct nfscldeleg *dp, struct ucred *cred,
4638     struct nfsmount *nmp, NFSPROC_T *p, int syscred)
4639 {
4640 	u_int32_t *tl;
4641 	struct nfsrv_descript nfsd;
4642 	struct nfsrv_descript *nd = &nfsd;
4643 	int error;
4644 
4645 	nfscl_reqstart(nd, NFSPROC_DELEGRETURN, nmp, dp->nfsdl_fh,
4646 	    dp->nfsdl_fhlen, NULL, NULL, 0, 0);
4647 	NFSM_BUILD(tl, u_int32_t *, NFSX_STATEID);
4648 	if (NFSHASNFSV4N(nmp))
4649 		*tl++ = 0;
4650 	else
4651 		*tl++ = dp->nfsdl_stateid.seqid;
4652 	*tl++ = dp->nfsdl_stateid.other[0];
4653 	*tl++ = dp->nfsdl_stateid.other[1];
4654 	*tl = dp->nfsdl_stateid.other[2];
4655 	if (syscred)
4656 		nd->nd_flag |= ND_USEGSSNAME;
4657 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4658 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4659 	if (error)
4660 		return (error);
4661 	error = nd->nd_repstat;
4662 	m_freem(nd->nd_mrep);
4663 	return (error);
4664 }
4665 
4666 /*
4667  * nfs getacl call.
4668  */
4669 int
4670 nfsrpc_getacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4671     struct acl *aclp, void *stuff)
4672 {
4673 	struct nfsrv_descript nfsd, *nd = &nfsd;
4674 	int error;
4675 	nfsattrbit_t attrbits;
4676 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4677 
4678 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4679 		return (EOPNOTSUPP);
4680 	NFSCL_REQSTART(nd, NFSPROC_GETACL, vp);
4681 	NFSZERO_ATTRBIT(&attrbits);
4682 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4683 	(void) nfsrv_putattrbit(nd, &attrbits);
4684 	error = nfscl_request(nd, vp, p, cred, stuff);
4685 	if (error)
4686 		return (error);
4687 	if (!nd->nd_repstat)
4688 		error = nfsv4_loadattr(nd, vp, NULL, NULL, NULL, 0, NULL,
4689 		    NULL, NULL, NULL, aclp, 0, NULL, NULL, NULL, p, cred);
4690 	else
4691 		error = nd->nd_repstat;
4692 	m_freem(nd->nd_mrep);
4693 	return (error);
4694 }
4695 
4696 /*
4697  * nfs setacl call.
4698  */
4699 int
4700 nfsrpc_setacl(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4701     struct acl *aclp, void *stuff)
4702 {
4703 	int error;
4704 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4705 
4706 	if (nfsrv_useacl == 0 || !NFSHASNFSV4(nmp))
4707 		return (EOPNOTSUPP);
4708 	error = nfsrpc_setattr(vp, NULL, aclp, cred, p, NULL, NULL, stuff);
4709 	return (error);
4710 }
4711 
4712 /*
4713  * nfs setacl call.
4714  */
4715 static int
4716 nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p,
4717     struct acl *aclp, nfsv4stateid_t *stateidp, void *stuff)
4718 {
4719 	struct nfsrv_descript nfsd, *nd = &nfsd;
4720 	int error;
4721 	nfsattrbit_t attrbits;
4722 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
4723 
4724 	if (!NFSHASNFSV4(nmp))
4725 		return (EOPNOTSUPP);
4726 	NFSCL_REQSTART(nd, NFSPROC_SETACL, vp);
4727 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
4728 	NFSZERO_ATTRBIT(&attrbits);
4729 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL);
4730 	(void) nfsv4_fillattr(nd, vnode_mount(vp), vp, aclp, NULL, NULL, 0,
4731 	    &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL);
4732 	error = nfscl_request(nd, vp, p, cred, stuff);
4733 	if (error)
4734 		return (error);
4735 	/* Don't care about the pre/postop attributes */
4736 	m_freem(nd->nd_mrep);
4737 	return (nd->nd_repstat);
4738 }
4739 
4740 /*
4741  * Do the NFSv4.1 Exchange ID.
4742  */
4743 int
4744 nfsrpc_exchangeid(struct nfsmount *nmp, struct nfsclclient *clp,
4745     struct nfssockreq *nrp, int minorvers, uint32_t exchflags,
4746     struct nfsclds **dspp, struct ucred *cred, NFSPROC_T *p)
4747 {
4748 	uint32_t *tl, v41flags;
4749 	struct nfsrv_descript nfsd;
4750 	struct nfsrv_descript *nd = &nfsd;
4751 	struct nfsclds *dsp;
4752 	struct timespec verstime;
4753 	int error, len;
4754 
4755 	*dspp = NULL;
4756 	if (minorvers == 0)
4757 		minorvers = nmp->nm_minorvers;
4758 	nfscl_reqstart(nd, NFSPROC_EXCHANGEID, nmp, NULL, 0, NULL, NULL,
4759 	    NFS_VER4, minorvers);
4760 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
4761 	*tl++ = txdr_unsigned(nfsboottime.tv_sec);	/* Client owner */
4762 	*tl = txdr_unsigned(clp->nfsc_rev);
4763 	(void) nfsm_strtom(nd, clp->nfsc_id, clp->nfsc_idlen);
4764 
4765 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_UNSIGNED);
4766 	*tl++ = txdr_unsigned(exchflags);
4767 	*tl++ = txdr_unsigned(NFSV4EXCH_SP4NONE);
4768 
4769 	/* Set the implementation id4 */
4770 	*tl = txdr_unsigned(1);
4771 	(void) nfsm_strtom(nd, "freebsd.org", strlen("freebsd.org"));
4772 	(void) nfsm_strtom(nd, version, strlen(version));
4773 	NFSM_BUILD(tl, uint32_t *, NFSX_V4TIME);
4774 	verstime.tv_sec = 1293840000;		/* Jan 1, 2011 */
4775 	verstime.tv_nsec = 0;
4776 	txdr_nfsv4time(&verstime, tl);
4777 	nd->nd_flag |= ND_USEGSSNAME;
4778 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred,
4779 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4780 	NFSCL_DEBUG(1, "exchangeid err=%d reps=%d\n", error,
4781 	    (int)nd->nd_repstat);
4782 	if (error != 0)
4783 		return (error);
4784 	if (nd->nd_repstat == 0) {
4785 		NFSM_DISSECT(tl, uint32_t *, 6 * NFSX_UNSIGNED + NFSX_HYPER);
4786 		len = fxdr_unsigned(int, *(tl + 7));
4787 		if (len < 0 || len > NFSV4_OPAQUELIMIT) {
4788 			error = NFSERR_BADXDR;
4789 			goto nfsmout;
4790 		}
4791 		dsp = malloc(sizeof(struct nfsclds) + len + 1, M_NFSCLDS,
4792 		    M_WAITOK | M_ZERO);
4793 		dsp->nfsclds_expire = NFSD_MONOSEC + clp->nfsc_renew;
4794 		dsp->nfsclds_servownlen = len;
4795 		dsp->nfsclds_sess.nfsess_clientid.lval[0] = *tl++;
4796 		dsp->nfsclds_sess.nfsess_clientid.lval[1] = *tl++;
4797 		dsp->nfsclds_sess.nfsess_sequenceid =
4798 		    fxdr_unsigned(uint32_t, *tl++);
4799 		v41flags = fxdr_unsigned(uint32_t, *tl);
4800 		if ((v41flags & NFSV4EXCH_USEPNFSMDS) != 0 &&
4801 		    NFSHASPNFSOPT(nmp)) {
4802 			NFSCL_DEBUG(1, "set PNFS\n");
4803 			NFSLOCKMNT(nmp);
4804 			nmp->nm_state |= NFSSTA_PNFS;
4805 			NFSUNLOCKMNT(nmp);
4806 			dsp->nfsclds_flags |= NFSCLDS_MDS;
4807 		}
4808 		if ((v41flags & NFSV4EXCH_USEPNFSDS) != 0)
4809 			dsp->nfsclds_flags |= NFSCLDS_DS;
4810 		if (minorvers == NFSV42_MINORVERSION)
4811 			dsp->nfsclds_flags |= NFSCLDS_MINORV2;
4812 		if (len > 0)
4813 			nd->nd_repstat = nfsrv_mtostr(nd,
4814 			    dsp->nfsclds_serverown, len);
4815 		if (nd->nd_repstat == 0) {
4816 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
4817 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
4818 			    NULL, MTX_DEF);
4819 			nfscl_initsessionslots(&dsp->nfsclds_sess);
4820 			*dspp = dsp;
4821 		} else
4822 			free(dsp, M_NFSCLDS);
4823 	}
4824 	error = nd->nd_repstat;
4825 nfsmout:
4826 	m_freem(nd->nd_mrep);
4827 	return (error);
4828 }
4829 
4830 /*
4831  * Do the NFSv4.1 Create Session.
4832  */
4833 int
4834 nfsrpc_createsession(struct nfsmount *nmp, struct nfsclsession *sep,
4835     struct nfssockreq *nrp, struct nfsclds *dsp, uint32_t sequenceid, int mds,
4836     struct ucred *cred, NFSPROC_T *p)
4837 {
4838 	uint32_t crflags, maxval, *tl;
4839 	struct nfsrv_descript nfsd;
4840 	struct nfsrv_descript *nd = &nfsd;
4841 	int error, irdcnt, minorvers;
4842 
4843 	/* Make sure nm_rsize, nm_wsize is set. */
4844 	if (nmp->nm_rsize > NFS_MAXBSIZE || nmp->nm_rsize == 0)
4845 		nmp->nm_rsize = NFS_MAXBSIZE;
4846 	if (nmp->nm_wsize > NFS_MAXBSIZE || nmp->nm_wsize == 0)
4847 		nmp->nm_wsize = NFS_MAXBSIZE;
4848 	if (dsp == NULL)
4849 		minorvers = nmp->nm_minorvers;
4850 	else if ((dsp->nfsclds_flags & NFSCLDS_MINORV2) != 0)
4851 		minorvers = NFSV42_MINORVERSION;
4852 	else
4853 		minorvers = NFSV41_MINORVERSION;
4854 	nfscl_reqstart(nd, NFSPROC_CREATESESSION, nmp, NULL, 0, NULL, NULL,
4855 	    NFS_VER4, minorvers);
4856 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
4857 	*tl++ = sep->nfsess_clientid.lval[0];
4858 	*tl++ = sep->nfsess_clientid.lval[1];
4859 	*tl++ = txdr_unsigned(sequenceid);
4860 	crflags = (NFSMNT_RDONLY(nmp->nm_mountp) ? 0 : NFSV4CRSESS_PERSIST);
4861 	if (nfscl_enablecallb != 0 && nfs_numnfscbd > 0 && mds != 0)
4862 		crflags |= NFSV4CRSESS_CONNBACKCHAN;
4863 	*tl = txdr_unsigned(crflags);
4864 
4865 	/* Fill in fore channel attributes. */
4866 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4867 	*tl++ = 0;				/* Header pad size */
4868 	if ((nd->nd_flag & ND_NFSV42) != 0 && mds != 0 && sb_max_adj >=
4869 	    nmp->nm_wsize && sb_max_adj >= nmp->nm_rsize) {
4870 		/*
4871 		 * NFSv4.2 Extended Attribute operations may want to do
4872 		 * requests/replies that are larger than nm_rsize/nm_wsize.
4873 		 */
4874 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4875 		*tl++ = txdr_unsigned(sb_max_adj - NFS_MAXXDR);
4876 	} else {
4877 		*tl++ = txdr_unsigned(nmp->nm_wsize + NFS_MAXXDR);
4878 		*tl++ = txdr_unsigned(nmp->nm_rsize + NFS_MAXXDR);
4879 	}
4880 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4881 	*tl++ = txdr_unsigned(20);		/* Max operations */
4882 	*tl++ = txdr_unsigned(64);		/* Max slots */
4883 	*tl = 0;				/* No rdma ird */
4884 
4885 	/* Fill in back channel attributes. */
4886 	NFSM_BUILD(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4887 	*tl++ = 0;				/* Header pad size */
4888 	*tl++ = txdr_unsigned(10000);		/* Max request size */
4889 	*tl++ = txdr_unsigned(10000);		/* Max response size */
4890 	*tl++ = txdr_unsigned(4096);		/* Max response size cached */
4891 	*tl++ = txdr_unsigned(4);		/* Max operations */
4892 	*tl++ = txdr_unsigned(NFSV4_CBSLOTS);	/* Max slots */
4893 	*tl = 0;				/* No rdma ird */
4894 
4895 	NFSM_BUILD(tl, uint32_t *, 8 * NFSX_UNSIGNED);
4896 	*tl++ = txdr_unsigned(NFS_CALLBCKPROG);	/* Call back prog # */
4897 
4898 	/* Allow AUTH_SYS callbacks as uid, gid == 0. */
4899 	*tl++ = txdr_unsigned(1);		/* Auth_sys only */
4900 	*tl++ = txdr_unsigned(AUTH_SYS);	/* AUTH_SYS type */
4901 	*tl++ = txdr_unsigned(nfsboottime.tv_sec); /* time stamp */
4902 	*tl++ = 0;				/* Null machine name */
4903 	*tl++ = 0;				/* Uid == 0 */
4904 	*tl++ = 0;				/* Gid == 0 */
4905 	*tl = 0;				/* No additional gids */
4906 	nd->nd_flag |= ND_USEGSSNAME;
4907 	error = newnfs_request(nd, nmp, NULL, nrp, NULL, p, cred, NFS_PROG,
4908 	    NFS_VER4, NULL, 1, NULL, NULL);
4909 	if (error != 0)
4910 		return (error);
4911 	if (nd->nd_repstat == 0) {
4912 		NFSM_DISSECT(tl, uint32_t *, NFSX_V4SESSIONID +
4913 		    2 * NFSX_UNSIGNED);
4914 		bcopy(tl, sep->nfsess_sessionid, NFSX_V4SESSIONID);
4915 		tl += NFSX_V4SESSIONID / NFSX_UNSIGNED;
4916 		sep->nfsess_sequenceid = fxdr_unsigned(uint32_t, *tl++);
4917 		crflags = fxdr_unsigned(uint32_t, *tl);
4918 		if ((crflags & NFSV4CRSESS_PERSIST) != 0 && mds != 0) {
4919 			NFSLOCKMNT(nmp);
4920 			nmp->nm_state |= NFSSTA_SESSPERSIST;
4921 			NFSUNLOCKMNT(nmp);
4922 		}
4923 
4924 		/* Get the fore channel slot count. */
4925 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4926 		tl++;			/* Skip the header pad size. */
4927 
4928 		/* Make sure nm_wsize is small enough. */
4929 		maxval = fxdr_unsigned(uint32_t, *tl++);
4930 		while (maxval < nmp->nm_wsize + NFS_MAXXDR) {
4931 			if (nmp->nm_wsize > 8096)
4932 				nmp->nm_wsize /= 2;
4933 			else
4934 				break;
4935 		}
4936 		sep->nfsess_maxreq = maxval;
4937 
4938 		/* Make sure nm_rsize is small enough. */
4939 		maxval = fxdr_unsigned(uint32_t, *tl++);
4940 		while (maxval < nmp->nm_rsize + NFS_MAXXDR) {
4941 			if (nmp->nm_rsize > 8096)
4942 				nmp->nm_rsize /= 2;
4943 			else
4944 				break;
4945 		}
4946 		sep->nfsess_maxresp = maxval;
4947 
4948 		sep->nfsess_maxcache = fxdr_unsigned(int, *tl++);
4949 		tl++;
4950 		sep->nfsess_foreslots = fxdr_unsigned(uint16_t, *tl++);
4951 		NFSCL_DEBUG(4, "fore slots=%d\n", (int)sep->nfsess_foreslots);
4952 		irdcnt = fxdr_unsigned(int, *tl);
4953 		if (irdcnt > 0)
4954 			NFSM_DISSECT(tl, uint32_t *, irdcnt * NFSX_UNSIGNED);
4955 
4956 		/* and the back channel slot count. */
4957 		NFSM_DISSECT(tl, uint32_t *, 7 * NFSX_UNSIGNED);
4958 		tl += 5;
4959 		sep->nfsess_backslots = fxdr_unsigned(uint16_t, *tl);
4960 		NFSCL_DEBUG(4, "back slots=%d\n", (int)sep->nfsess_backslots);
4961 	}
4962 	error = nd->nd_repstat;
4963 nfsmout:
4964 	m_freem(nd->nd_mrep);
4965 	return (error);
4966 }
4967 
4968 /*
4969  * Do the NFSv4.1 Destroy Session.
4970  */
4971 int
4972 nfsrpc_destroysession(struct nfsmount *nmp, struct nfsclclient *clp,
4973     struct ucred *cred, NFSPROC_T *p)
4974 {
4975 	uint32_t *tl;
4976 	struct nfsrv_descript nfsd;
4977 	struct nfsrv_descript *nd = &nfsd;
4978 	int error;
4979 	struct nfsclsession *tsep;
4980 
4981 	nfscl_reqstart(nd, NFSPROC_DESTROYSESSION, nmp, NULL, 0, NULL, NULL, 0,
4982 	    0);
4983 	NFSM_BUILD(tl, uint32_t *, NFSX_V4SESSIONID);
4984 	tsep = nfsmnt_mdssession(nmp);
4985 	bcopy(tsep->nfsess_sessionid, tl, NFSX_V4SESSIONID);
4986 	nd->nd_flag |= ND_USEGSSNAME;
4987 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
4988 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
4989 	if (error != 0)
4990 		return (error);
4991 	error = nd->nd_repstat;
4992 	m_freem(nd->nd_mrep);
4993 	return (error);
4994 }
4995 
4996 /*
4997  * Do the NFSv4.1 Destroy Client.
4998  */
4999 int
5000 nfsrpc_destroyclient(struct nfsmount *nmp, struct nfsclclient *clp,
5001     struct ucred *cred, NFSPROC_T *p)
5002 {
5003 	uint32_t *tl;
5004 	struct nfsrv_descript nfsd;
5005 	struct nfsrv_descript *nd = &nfsd;
5006 	int error;
5007 	struct nfsclsession *tsep;
5008 
5009 	nfscl_reqstart(nd, NFSPROC_DESTROYCLIENT, nmp, NULL, 0, NULL, NULL, 0,
5010 	    0);
5011 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5012 	tsep = nfsmnt_mdssession(nmp);
5013 	*tl++ = tsep->nfsess_clientid.lval[0];
5014 	*tl = tsep->nfsess_clientid.lval[1];
5015 	nd->nd_flag |= ND_USEGSSNAME;
5016 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5017 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5018 	if (error != 0)
5019 		return (error);
5020 	error = nd->nd_repstat;
5021 	m_freem(nd->nd_mrep);
5022 	return (error);
5023 }
5024 
5025 /*
5026  * Do the NFSv4.1 LayoutGet.
5027  */
5028 static int
5029 nfsrpc_layoutget(struct nfsmount *nmp, uint8_t *fhp, int fhlen, int iomode,
5030     uint64_t offset, uint64_t len, uint64_t minlen, int layouttype,
5031     int layoutlen, nfsv4stateid_t *stateidp, int *retonclosep,
5032     struct nfsclflayouthead *flhp, struct ucred *cred, NFSPROC_T *p,
5033     void *stuff)
5034 {
5035 	struct nfsrv_descript nfsd, *nd = &nfsd;
5036 	int error;
5037 
5038 	nfscl_reqstart(nd, NFSPROC_LAYOUTGET, nmp, fhp, fhlen, NULL, NULL, 0,
5039 	    0);
5040 	nfsrv_setuplayoutget(nd, iomode, offset, len, minlen, stateidp,
5041 	    layouttype, layoutlen, 0);
5042 	nd->nd_flag |= ND_USEGSSNAME;
5043 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5044 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5045 	NFSCL_DEBUG(4, "layget err=%d st=%d\n", error, nd->nd_repstat);
5046 	if (error != 0)
5047 		return (error);
5048 	if (nd->nd_repstat == 0)
5049 		error = nfsrv_parselayoutget(nmp, nd, stateidp, retonclosep,
5050 		    flhp);
5051 	if (error == 0 && nd->nd_repstat != 0)
5052 		error = nd->nd_repstat;
5053 	m_freem(nd->nd_mrep);
5054 	return (error);
5055 }
5056 
5057 /*
5058  * Do the NFSv4.1 Get Device Info.
5059  */
5060 int
5061 nfsrpc_getdeviceinfo(struct nfsmount *nmp, uint8_t *deviceid, int layouttype,
5062     uint32_t *notifybitsp, struct nfscldevinfo **ndip, struct ucred *cred,
5063     NFSPROC_T *p)
5064 {
5065 	uint32_t cnt, *tl, vers, minorvers;
5066 	struct nfsrv_descript nfsd;
5067 	struct nfsrv_descript *nd = &nfsd;
5068 	struct sockaddr_in sin, ssin;
5069 	struct sockaddr_in6 sin6, ssin6;
5070 	struct nfsclds *dsp = NULL, **dspp, **gotdspp;
5071 	struct nfscldevinfo *ndi;
5072 	int addrcnt = 0, bitcnt, error, gotminor, gotvers, i, isudp, j;
5073 	int stripecnt;
5074 	uint8_t stripeindex;
5075 	sa_family_t af, safilled;
5076 
5077 	ssin.sin_port = 0;		/* To shut up compiler. */
5078 	ssin.sin_addr.s_addr = 0;	/* ditto */
5079 	*ndip = NULL;
5080 	ndi = NULL;
5081 	gotdspp = NULL;
5082 	nfscl_reqstart(nd, NFSPROC_GETDEVICEINFO, nmp, NULL, 0, NULL, NULL, 0,
5083 	    0);
5084 	NFSM_BUILD(tl, uint32_t *, NFSX_V4DEVICEID + 3 * NFSX_UNSIGNED);
5085 	NFSBCOPY(deviceid, tl, NFSX_V4DEVICEID);
5086 	tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5087 	*tl++ = txdr_unsigned(layouttype);
5088 	*tl++ = txdr_unsigned(100000);
5089 	if (notifybitsp != NULL && *notifybitsp != 0) {
5090 		*tl = txdr_unsigned(1);		/* One word of bits. */
5091 		NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5092 		*tl = txdr_unsigned(*notifybitsp);
5093 	} else
5094 		*tl = txdr_unsigned(0);
5095 	nd->nd_flag |= ND_USEGSSNAME;
5096 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5097 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5098 	if (error != 0)
5099 		return (error);
5100 	if (nd->nd_repstat == 0) {
5101 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5102 		if (layouttype != fxdr_unsigned(int, *tl))
5103 			printf("EEK! devinfo layout type not same!\n");
5104 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES) {
5105 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5106 			stripecnt = fxdr_unsigned(int, *tl);
5107 			NFSCL_DEBUG(4, "stripecnt=%d\n", stripecnt);
5108 			if (stripecnt < 1 || stripecnt > 4096) {
5109 				printf("pNFS File layout devinfo stripecnt %d:"
5110 				    " out of range\n", stripecnt);
5111 				error = NFSERR_BADXDR;
5112 				goto nfsmout;
5113 			}
5114 			NFSM_DISSECT(tl, uint32_t *, (stripecnt + 1) *
5115 			    NFSX_UNSIGNED);
5116 			addrcnt = fxdr_unsigned(int, *(tl + stripecnt));
5117 			NFSCL_DEBUG(4, "addrcnt=%d\n", addrcnt);
5118 			if (addrcnt < 1 || addrcnt > 128) {
5119 				printf("NFS devinfo addrcnt %d: out of range\n",
5120 				    addrcnt);
5121 				error = NFSERR_BADXDR;
5122 				goto nfsmout;
5123 			}
5124 
5125 			/*
5126 			 * Now we know how many stripe indices and addresses, so
5127 			 * we can allocate the structure the correct size.
5128 			 */
5129 			i = (stripecnt * sizeof(uint8_t)) /
5130 			    sizeof(struct nfsclds *) + 1;
5131 			NFSCL_DEBUG(4, "stripeindices=%d\n", i);
5132 			ndi = malloc(sizeof(*ndi) + (addrcnt + i) *
5133 			    sizeof(struct nfsclds *), M_NFSDEVINFO, M_WAITOK |
5134 			    M_ZERO);
5135 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5136 			    NFSX_V4DEVICEID);
5137 			ndi->nfsdi_refcnt = 0;
5138 			ndi->nfsdi_flags = NFSDI_FILELAYOUT;
5139 			ndi->nfsdi_stripecnt = stripecnt;
5140 			ndi->nfsdi_addrcnt = addrcnt;
5141 			/* Fill in the stripe indices. */
5142 			for (i = 0; i < stripecnt; i++) {
5143 				stripeindex = fxdr_unsigned(uint8_t, *tl++);
5144 				NFSCL_DEBUG(4, "stripeind=%d\n", stripeindex);
5145 				if (stripeindex >= addrcnt) {
5146 					printf("pNFS File Layout devinfo"
5147 					    " stripeindex %d: too big\n",
5148 					    (int)stripeindex);
5149 					error = NFSERR_BADXDR;
5150 					goto nfsmout;
5151 				}
5152 				nfsfldi_setstripeindex(ndi, i, stripeindex);
5153 			}
5154 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
5155 			/* For Flex File, we only get one address list. */
5156 			ndi = malloc(sizeof(*ndi) + sizeof(struct nfsclds *),
5157 			    M_NFSDEVINFO, M_WAITOK | M_ZERO);
5158 			NFSBCOPY(deviceid, ndi->nfsdi_deviceid,
5159 			    NFSX_V4DEVICEID);
5160 			ndi->nfsdi_refcnt = 0;
5161 			ndi->nfsdi_flags = NFSDI_FLEXFILE;
5162 			addrcnt = ndi->nfsdi_addrcnt = 1;
5163 		}
5164 
5165 		/* Now, dissect the server address(es). */
5166 		safilled = AF_UNSPEC;
5167 		for (i = 0; i < addrcnt; i++) {
5168 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5169 			cnt = fxdr_unsigned(uint32_t, *tl);
5170 			if (cnt == 0) {
5171 				printf("NFS devinfo 0 len addrlist\n");
5172 				error = NFSERR_BADXDR;
5173 				goto nfsmout;
5174 			}
5175 			dspp = nfsfldi_addr(ndi, i);
5176 			safilled = AF_UNSPEC;
5177 			for (j = 0; j < cnt; j++) {
5178 				error = nfsv4_getipaddr(nd, &sin, &sin6, &af,
5179 				    &isudp);
5180 				if (error != 0 && error != EPERM) {
5181 					error = NFSERR_BADXDR;
5182 					goto nfsmout;
5183 				}
5184 				if (error == 0 && isudp == 0) {
5185 					/*
5186 					 * The priority is:
5187 					 * - Same address family.
5188 					 * Save the address and dspp, so that
5189 					 * the connection can be done after
5190 					 * parsing is complete.
5191 					 */
5192 					if (safilled == AF_UNSPEC ||
5193 					    (af == nmp->nm_nam->sa_family &&
5194 					     safilled != nmp->nm_nam->sa_family)
5195 					   ) {
5196 						if (af == AF_INET)
5197 							ssin = sin;
5198 						else
5199 							ssin6 = sin6;
5200 						safilled = af;
5201 						gotdspp = dspp;
5202 					}
5203 				}
5204 			}
5205 		}
5206 
5207 		gotvers = NFS_VER4;	/* Default NFSv4.1 for File Layout. */
5208 		gotminor = NFSV41_MINORVERSION;
5209 		/* For Flex File, we will take one of the versions to use. */
5210 		if (layouttype == NFSLAYOUT_FLEXFILE) {
5211 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5212 			j = fxdr_unsigned(int, *tl);
5213 			if (j < 1 || j > NFSDEV_MAXVERS) {
5214 				printf("pNFS: too many versions\n");
5215 				error = NFSERR_BADXDR;
5216 				goto nfsmout;
5217 			}
5218 			gotvers = 0;
5219 			gotminor = 0;
5220 			for (i = 0; i < j; i++) {
5221 				NFSM_DISSECT(tl, uint32_t *, 5 * NFSX_UNSIGNED);
5222 				vers = fxdr_unsigned(uint32_t, *tl++);
5223 				minorvers = fxdr_unsigned(uint32_t, *tl++);
5224 				if (vers == NFS_VER3)
5225 					minorvers = 0;
5226 				if ((vers == NFS_VER4 && ((minorvers ==
5227 				    NFSV41_MINORVERSION && gotminor == 0) ||
5228 				    minorvers == NFSV42_MINORVERSION)) ||
5229 				    (vers == NFS_VER3 && gotvers == 0)) {
5230 					gotvers = vers;
5231 					gotminor = minorvers;
5232 					/* We'll take this one. */
5233 					ndi->nfsdi_versindex = i;
5234 					ndi->nfsdi_vers = vers;
5235 					ndi->nfsdi_minorvers = minorvers;
5236 					ndi->nfsdi_rsize = fxdr_unsigned(
5237 					    uint32_t, *tl++);
5238 					ndi->nfsdi_wsize = fxdr_unsigned(
5239 					    uint32_t, *tl++);
5240 					if (*tl == newnfs_true)
5241 						ndi->nfsdi_flags |=
5242 						    NFSDI_TIGHTCOUPLED;
5243 					else
5244 						ndi->nfsdi_flags &=
5245 						    ~NFSDI_TIGHTCOUPLED;
5246 				}
5247 			}
5248 			if (gotvers == 0) {
5249 				printf("pNFS: no NFSv3, NFSv4.1 or NFSv4.2\n");
5250 				error = NFSERR_BADXDR;
5251 				goto nfsmout;
5252 			}
5253 		}
5254 
5255 		/* And the notify bits. */
5256 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5257 		bitcnt = fxdr_unsigned(int, *tl);
5258 		if (bitcnt > 0) {
5259 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5260 			if (notifybitsp != NULL)
5261 				*notifybitsp =
5262 				    fxdr_unsigned(uint32_t, *tl);
5263 		}
5264 		if (safilled != AF_UNSPEC) {
5265 			KASSERT(ndi != NULL, ("ndi is NULL"));
5266 			*ndip = ndi;
5267 		} else
5268 			error = EPERM;
5269 		if (error == 0) {
5270 			/*
5271 			 * Now we can do a TCP connection for the correct
5272 			 * NFS version and IP address.
5273 			 */
5274 			error = nfsrpc_fillsa(nmp, &ssin, &ssin6, safilled,
5275 			    gotvers, gotminor, &dsp, p);
5276 		}
5277 		if (error == 0) {
5278 			KASSERT(gotdspp != NULL, ("gotdspp is NULL"));
5279 			*gotdspp = dsp;
5280 		}
5281 	}
5282 	if (nd->nd_repstat != 0 && error == 0)
5283 		error = nd->nd_repstat;
5284 nfsmout:
5285 	if (error != 0 && ndi != NULL)
5286 		nfscl_freedevinfo(ndi);
5287 	m_freem(nd->nd_mrep);
5288 	return (error);
5289 }
5290 
5291 /*
5292  * Do the NFSv4.1 LayoutCommit.
5293  */
5294 int
5295 nfsrpc_layoutcommit(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5296     uint64_t off, uint64_t len, uint64_t lastbyte, nfsv4stateid_t *stateidp,
5297     int layouttype, struct ucred *cred, NFSPROC_T *p, void *stuff)
5298 {
5299 	uint32_t *tl;
5300 	struct nfsrv_descript nfsd, *nd = &nfsd;
5301 	int error;
5302 
5303 	nfscl_reqstart(nd, NFSPROC_LAYOUTCOMMIT, nmp, fh, fhlen, NULL, NULL,
5304 	    0, 0);
5305 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
5306 	    NFSX_STATEID);
5307 	txdr_hyper(off, tl);
5308 	tl += 2;
5309 	txdr_hyper(len, tl);
5310 	tl += 2;
5311 	if (reclaim != 0)
5312 		*tl++ = newnfs_true;
5313 	else
5314 		*tl++ = newnfs_false;
5315 	*tl++ = txdr_unsigned(stateidp->seqid);
5316 	*tl++ = stateidp->other[0];
5317 	*tl++ = stateidp->other[1];
5318 	*tl++ = stateidp->other[2];
5319 	*tl++ = newnfs_true;
5320 	if (lastbyte < off)
5321 		lastbyte = off;
5322 	else if (lastbyte >= (off + len))
5323 		lastbyte = off + len - 1;
5324 	txdr_hyper(lastbyte, tl);
5325 	tl += 2;
5326 	*tl++ = newnfs_false;
5327 	*tl++ = txdr_unsigned(layouttype);
5328 	/* All supported layouts are 0 length. */
5329 	*tl = txdr_unsigned(0);
5330 	nd->nd_flag |= ND_USEGSSNAME;
5331 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5332 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5333 	if (error != 0)
5334 		return (error);
5335 	error = nd->nd_repstat;
5336 	m_freem(nd->nd_mrep);
5337 	return (error);
5338 }
5339 
5340 /*
5341  * Do the NFSv4.1 LayoutReturn.
5342  */
5343 int
5344 nfsrpc_layoutreturn(struct nfsmount *nmp, uint8_t *fh, int fhlen, int reclaim,
5345     int layouttype, uint32_t iomode, int layoutreturn, uint64_t offset,
5346     uint64_t len, nfsv4stateid_t *stateidp, struct ucred *cred, NFSPROC_T *p,
5347     uint32_t stat, uint32_t op, char *devid)
5348 {
5349 	uint32_t *tl;
5350 	struct nfsrv_descript nfsd, *nd = &nfsd;
5351 	uint64_t tu64;
5352 	int error;
5353 
5354 	nfscl_reqstart(nd, NFSPROC_LAYOUTRETURN, nmp, fh, fhlen, NULL, NULL,
5355 	    0, 0);
5356 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED);
5357 	if (reclaim != 0)
5358 		*tl++ = newnfs_true;
5359 	else
5360 		*tl++ = newnfs_false;
5361 	*tl++ = txdr_unsigned(layouttype);
5362 	*tl++ = txdr_unsigned(iomode);
5363 	*tl = txdr_unsigned(layoutreturn);
5364 	if (layoutreturn == NFSLAYOUTRETURN_FILE) {
5365 		NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_STATEID +
5366 		    NFSX_UNSIGNED);
5367 		txdr_hyper(offset, tl);
5368 		tl += 2;
5369 		txdr_hyper(len, tl);
5370 		tl += 2;
5371 		NFSCL_DEBUG(4, "layoutret stseq=%d\n", (int)stateidp->seqid);
5372 		*tl++ = txdr_unsigned(stateidp->seqid);
5373 		*tl++ = stateidp->other[0];
5374 		*tl++ = stateidp->other[1];
5375 		*tl++ = stateidp->other[2];
5376 		if (layouttype == NFSLAYOUT_NFSV4_1_FILES)
5377 			*tl = txdr_unsigned(0);
5378 		else if (layouttype == NFSLAYOUT_FLEXFILE) {
5379 			if (stat != 0) {
5380 				*tl = txdr_unsigned(2 * NFSX_HYPER +
5381 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5382 				    NFSX_UNSIGNED);
5383 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER +
5384 				    NFSX_STATEID + NFSX_V4DEVICEID + 5 *
5385 				    NFSX_UNSIGNED);
5386 				*tl++ = txdr_unsigned(1);	/* One error. */
5387 				tu64 = 0;			/* Offset. */
5388 				txdr_hyper(tu64, tl); tl += 2;
5389 				tu64 = UINT64_MAX;		/* Length. */
5390 				txdr_hyper(tu64, tl); tl += 2;
5391 				NFSBCOPY(stateidp, tl, NFSX_STATEID);
5392 				tl += (NFSX_STATEID / NFSX_UNSIGNED);
5393 				*tl++ = txdr_unsigned(1);	/* One error. */
5394 				NFSBCOPY(devid, tl, NFSX_V4DEVICEID);
5395 				tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
5396 				*tl++ = txdr_unsigned(stat);
5397 				*tl++ = txdr_unsigned(op);
5398 			} else {
5399 				*tl = txdr_unsigned(2 * NFSX_UNSIGNED);
5400 				NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
5401 				/* No ioerrs. */
5402 				*tl++ = 0;
5403 			}
5404 			*tl = 0;	/* No stats yet. */
5405 		}
5406 	}
5407 	nd->nd_flag |= ND_USEGSSNAME;
5408 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5409 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5410 	if (error != 0)
5411 		return (error);
5412 	if (nd->nd_repstat == 0) {
5413 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
5414 		if (*tl != 0) {
5415 			NFSM_DISSECT(tl, uint32_t *, NFSX_STATEID);
5416 			stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
5417 			stateidp->other[0] = *tl++;
5418 			stateidp->other[1] = *tl++;
5419 			stateidp->other[2] = *tl;
5420 		}
5421 	} else
5422 		error = nd->nd_repstat;
5423 nfsmout:
5424 	m_freem(nd->nd_mrep);
5425 	return (error);
5426 }
5427 
5428 /*
5429  * Acquire a layout and devinfo, if possible. The caller must have acquired
5430  * a reference count on the nfsclclient structure before calling this.
5431  * Return the layout in lypp with a reference count on it, if successful.
5432  */
5433 static int
5434 nfsrpc_getlayout(struct nfsmount *nmp, vnode_t vp, struct nfsfh *nfhp,
5435     int iomode, uint32_t *notifybitsp, nfsv4stateid_t *stateidp, uint64_t off,
5436     struct nfscllayout **lypp, struct ucred *cred, NFSPROC_T *p)
5437 {
5438 	struct nfscllayout *lyp;
5439 	struct nfsclflayout *flp;
5440 	struct nfsclflayouthead flh;
5441 	int error = 0, islocked, layoutlen, layouttype, recalled, retonclose;
5442 	nfsv4stateid_t stateid;
5443 	struct nfsclsession *tsep;
5444 
5445 	*lypp = NULL;
5446 	if (NFSHASFLEXFILE(nmp))
5447 		layouttype = NFSLAYOUT_FLEXFILE;
5448 	else
5449 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
5450 	/*
5451 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
5452 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
5453 	 * flp == NULL.
5454 	 */
5455 	lyp = nfscl_getlayout(nmp->nm_clp, nfhp->nfh_fh, nfhp->nfh_len,
5456 	    off, &flp, &recalled);
5457 	islocked = 0;
5458 	if (lyp == NULL || flp == NULL) {
5459 		if (recalled != 0)
5460 			return (EIO);
5461 		LIST_INIT(&flh);
5462 		tsep = nfsmnt_mdssession(nmp);
5463 		layoutlen = tsep->nfsess_maxcache -
5464 		    (NFSX_STATEID + 3 * NFSX_UNSIGNED);
5465 		if (lyp == NULL) {
5466 			stateid.seqid = 0;
5467 			stateid.other[0] = stateidp->other[0];
5468 			stateid.other[1] = stateidp->other[1];
5469 			stateid.other[2] = stateidp->other[2];
5470 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5471 			    nfhp->nfh_len, iomode, (uint64_t)0, UINT64_MAX,
5472 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5473 			    &retonclose, &flh, cred, p, NULL);
5474 		} else {
5475 			islocked = 1;
5476 			stateid.seqid = lyp->nfsly_stateid.seqid;
5477 			stateid.other[0] = lyp->nfsly_stateid.other[0];
5478 			stateid.other[1] = lyp->nfsly_stateid.other[1];
5479 			stateid.other[2] = lyp->nfsly_stateid.other[2];
5480 			error = nfsrpc_layoutget(nmp, nfhp->nfh_fh,
5481 			    nfhp->nfh_len, iomode, off, UINT64_MAX,
5482 			    (uint64_t)0, layouttype, layoutlen, &stateid,
5483 			    &retonclose, &flh, cred, p, NULL);
5484 		}
5485 		error = nfsrpc_layoutgetres(nmp, vp, nfhp->nfh_fh,
5486 		    nfhp->nfh_len, &stateid, retonclose, notifybitsp, &lyp,
5487 		    &flh, layouttype, error, NULL, cred, p);
5488 		if (error == 0)
5489 			*lypp = lyp;
5490 		else if (islocked != 0)
5491 			nfscl_rellayout(lyp, 1);
5492 	} else
5493 		*lypp = lyp;
5494 	return (error);
5495 }
5496 
5497 /*
5498  * Do a TCP connection plus exchange id and create session.
5499  * If successful, a "struct nfsclds" is linked into the list for the
5500  * mount point and a pointer to it is returned.
5501  */
5502 static int
5503 nfsrpc_fillsa(struct nfsmount *nmp, struct sockaddr_in *sin,
5504     struct sockaddr_in6 *sin6, sa_family_t af, int vers, int minorvers,
5505     struct nfsclds **dspp, NFSPROC_T *p)
5506 {
5507 	struct sockaddr_in *msad, *sad;
5508 	struct sockaddr_in6 *msad6, *sad6;
5509 	struct nfsclclient *clp;
5510 	struct nfssockreq *nrp;
5511 	struct nfsclds *dsp, *tdsp;
5512 	int error, firsttry;
5513 	enum nfsclds_state retv;
5514 	uint32_t sequenceid = 0;
5515 
5516 	KASSERT(nmp->nm_sockreq.nr_cred != NULL,
5517 	    ("nfsrpc_fillsa: NULL nr_cred"));
5518 	NFSLOCKCLSTATE();
5519 	clp = nmp->nm_clp;
5520 	NFSUNLOCKCLSTATE();
5521 	if (clp == NULL)
5522 		return (EPERM);
5523 	if (af == AF_INET) {
5524 		NFSLOCKMNT(nmp);
5525 		/*
5526 		 * Check to see if we already have a session for this
5527 		 * address that is usable for a DS.
5528 		 * Note that the MDS's address is in a different place
5529 		 * than the sessions already acquired for DS's.
5530 		 */
5531 		msad = (struct sockaddr_in *)nmp->nm_sockreq.nr_nam;
5532 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5533 		while (tdsp != NULL) {
5534 			if (msad != NULL && msad->sin_family == AF_INET &&
5535 			    sin->sin_addr.s_addr == msad->sin_addr.s_addr &&
5536 			    sin->sin_port == msad->sin_port &&
5537 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5538 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5539 				*dspp = tdsp;
5540 				NFSUNLOCKMNT(nmp);
5541 				NFSCL_DEBUG(4, "fnd same addr\n");
5542 				return (0);
5543 			}
5544 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5545 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5546 				msad = (struct sockaddr_in *)
5547 				    tdsp->nfsclds_sockp->nr_nam;
5548 			else
5549 				msad = NULL;
5550 		}
5551 		NFSUNLOCKMNT(nmp);
5552 
5553 		/* No IP address match, so look for new/trunked one. */
5554 		sad = malloc(sizeof(*sad), M_SONAME, M_WAITOK | M_ZERO);
5555 		sad->sin_len = sizeof(*sad);
5556 		sad->sin_family = AF_INET;
5557 		sad->sin_port = sin->sin_port;
5558 		sad->sin_addr.s_addr = sin->sin_addr.s_addr;
5559 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5560 		nrp->nr_nam = (struct sockaddr *)sad;
5561 	} else if (af == AF_INET6) {
5562 		NFSLOCKMNT(nmp);
5563 		/*
5564 		 * Check to see if we already have a session for this
5565 		 * address that is usable for a DS.
5566 		 * Note that the MDS's address is in a different place
5567 		 * than the sessions already acquired for DS's.
5568 		 */
5569 		msad6 = (struct sockaddr_in6 *)nmp->nm_sockreq.nr_nam;
5570 		tdsp = TAILQ_FIRST(&nmp->nm_sess);
5571 		while (tdsp != NULL) {
5572 			if (msad6 != NULL && msad6->sin6_family == AF_INET6 &&
5573 			    IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
5574 			    &msad6->sin6_addr) &&
5575 			    sin6->sin6_port == msad6->sin6_port &&
5576 			    (tdsp->nfsclds_flags & NFSCLDS_DS) != 0 &&
5577 			    tdsp->nfsclds_sess.nfsess_defunct == 0) {
5578 				*dspp = tdsp;
5579 				NFSUNLOCKMNT(nmp);
5580 				return (0);
5581 			}
5582 			tdsp = TAILQ_NEXT(tdsp, nfsclds_list);
5583 			if (tdsp != NULL && tdsp->nfsclds_sockp != NULL)
5584 				msad6 = (struct sockaddr_in6 *)
5585 				    tdsp->nfsclds_sockp->nr_nam;
5586 			else
5587 				msad6 = NULL;
5588 		}
5589 		NFSUNLOCKMNT(nmp);
5590 
5591 		/* No IP address match, so look for new/trunked one. */
5592 		sad6 = malloc(sizeof(*sad6), M_SONAME, M_WAITOK | M_ZERO);
5593 		sad6->sin6_len = sizeof(*sad6);
5594 		sad6->sin6_family = AF_INET6;
5595 		sad6->sin6_port = sin6->sin6_port;
5596 		NFSBCOPY(&sin6->sin6_addr, &sad6->sin6_addr,
5597 		    sizeof(struct in6_addr));
5598 		nrp = malloc(sizeof(*nrp), M_NFSSOCKREQ, M_WAITOK | M_ZERO);
5599 		nrp->nr_nam = (struct sockaddr *)sad6;
5600 	} else
5601 		return (EPERM);
5602 
5603 	nrp->nr_sotype = SOCK_STREAM;
5604 	mtx_init(&nrp->nr_mtx, "nfssock", NULL, MTX_DEF);
5605 	nrp->nr_prog = NFS_PROG;
5606 	nrp->nr_vers = vers;
5607 
5608 	/*
5609 	 * Use the credentials that were used for the mount, which are
5610 	 * in nmp->nm_sockreq.nr_cred for newnfs_connect() etc.
5611 	 * Ref. counting the credentials with crhold() is probably not
5612 	 * necessary, since nm_sockreq.nr_cred won't be crfree()'d until
5613 	 * unmount, but I did it anyhow.
5614 	 */
5615 	nrp->nr_cred = crhold(nmp->nm_sockreq.nr_cred);
5616 	error = newnfs_connect(nmp, nrp, NULL, p, 0);
5617 	NFSCL_DEBUG(3, "DS connect=%d\n", error);
5618 
5619 	dsp = NULL;
5620 	/* Now, do the exchangeid and create session. */
5621 	if (error == 0) {
5622 		if (vers == NFS_VER4) {
5623 			firsttry = 0;
5624 			do {
5625 				error = nfsrpc_exchangeid(nmp, clp, nrp,
5626 				    minorvers, NFSV4EXCH_USEPNFSDS, &dsp,
5627 				    nrp->nr_cred, p);
5628 				NFSCL_DEBUG(3, "DS exchangeid=%d\n", error);
5629 				if (error == NFSERR_MINORVERMISMATCH)
5630 					minorvers = NFSV42_MINORVERSION;
5631 			} while (error == NFSERR_MINORVERMISMATCH &&
5632 			    firsttry++ == 0);
5633 			if (error != 0)
5634 				newnfs_disconnect(nrp);
5635 		} else {
5636 			dsp = malloc(sizeof(struct nfsclds), M_NFSCLDS,
5637 			    M_WAITOK | M_ZERO);
5638 			dsp->nfsclds_flags |= NFSCLDS_DS;
5639 			dsp->nfsclds_expire = INT32_MAX; /* No renews needed. */
5640 			mtx_init(&dsp->nfsclds_mtx, "nfsds", NULL, MTX_DEF);
5641 			mtx_init(&dsp->nfsclds_sess.nfsess_mtx, "nfssession",
5642 			    NULL, MTX_DEF);
5643 		}
5644 	}
5645 	if (error == 0) {
5646 		dsp->nfsclds_sockp = nrp;
5647 		if (vers == NFS_VER4) {
5648 			NFSLOCKMNT(nmp);
5649 			retv = nfscl_getsameserver(nmp, dsp, &tdsp,
5650 			    &sequenceid);
5651 			NFSCL_DEBUG(3, "getsame ret=%d\n", retv);
5652 			if (retv == NFSDSP_USETHISSESSION &&
5653 			    nfscl_dssameconn != 0) {
5654 				NFSLOCKDS(tdsp);
5655 				tdsp->nfsclds_flags |= NFSCLDS_SAMECONN;
5656 				NFSUNLOCKDS(tdsp);
5657 				NFSUNLOCKMNT(nmp);
5658 				/*
5659 				 * If there is already a session for this
5660 				 * server, use it.
5661 				 */
5662 				(void)newnfs_disconnect(nrp);
5663 				nfscl_freenfsclds(dsp);
5664 				*dspp = tdsp;
5665 				return (0);
5666 			}
5667 			if (retv == NFSDSP_NOTFOUND)
5668 				sequenceid =
5669 				    dsp->nfsclds_sess.nfsess_sequenceid;
5670 			NFSUNLOCKMNT(nmp);
5671 			error = nfsrpc_createsession(nmp, &dsp->nfsclds_sess,
5672 			    nrp, dsp, sequenceid, 0, nrp->nr_cred, p);
5673 			NFSCL_DEBUG(3, "DS createsess=%d\n", error);
5674 		}
5675 	} else {
5676 		NFSFREECRED(nrp->nr_cred);
5677 		NFSFREEMUTEX(&nrp->nr_mtx);
5678 		free(nrp->nr_nam, M_SONAME);
5679 		free(nrp, M_NFSSOCKREQ);
5680 	}
5681 	if (error == 0) {
5682 		NFSCL_DEBUG(3, "add DS session\n");
5683 		/*
5684 		 * Put it at the end of the list. That way the list
5685 		 * is ordered by when the entry was added. This matters
5686 		 * since the one done first is the one that should be
5687 		 * used for sequencid'ing any subsequent create sessions.
5688 		 */
5689 		NFSLOCKMNT(nmp);
5690 		TAILQ_INSERT_TAIL(&nmp->nm_sess, dsp, nfsclds_list);
5691 		NFSUNLOCKMNT(nmp);
5692 		*dspp = dsp;
5693 	} else if (dsp != NULL) {
5694 		newnfs_disconnect(nrp);
5695 		nfscl_freenfsclds(dsp);
5696 	}
5697 	return (error);
5698 }
5699 
5700 /*
5701  * Do the NFSv4.1 Reclaim Complete.
5702  */
5703 int
5704 nfsrpc_reclaimcomplete(struct nfsmount *nmp, struct ucred *cred, NFSPROC_T *p)
5705 {
5706 	uint32_t *tl;
5707 	struct nfsrv_descript nfsd;
5708 	struct nfsrv_descript *nd = &nfsd;
5709 	int error;
5710 
5711 	nfscl_reqstart(nd, NFSPROC_RECLAIMCOMPL, nmp, NULL, 0, NULL, NULL, 0,
5712 	    0);
5713 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
5714 	*tl = newnfs_false;
5715 	nd->nd_flag |= ND_USEGSSNAME;
5716 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred,
5717 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
5718 	if (error != 0)
5719 		return (error);
5720 	error = nd->nd_repstat;
5721 	m_freem(nd->nd_mrep);
5722 	return (error);
5723 }
5724 
5725 /*
5726  * Initialize the slot tables for a session.
5727  */
5728 static void
5729 nfscl_initsessionslots(struct nfsclsession *sep)
5730 {
5731 	int i;
5732 
5733 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
5734 		if (sep->nfsess_cbslots[i].nfssl_reply != NULL)
5735 			m_freem(sep->nfsess_cbslots[i].nfssl_reply);
5736 		NFSBZERO(&sep->nfsess_cbslots[i], sizeof(struct nfsslot));
5737 	}
5738 	for (i = 0; i < 64; i++)
5739 		sep->nfsess_slotseq[i] = 0;
5740 	sep->nfsess_slots = 0;
5741 }
5742 
5743 /*
5744  * Called to try and do an I/O operation via an NFSv4.1 Data Server (DS).
5745  */
5746 int
5747 nfscl_doiods(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
5748     uint32_t rwaccess, int docommit, struct ucred *cred, NFSPROC_T *p)
5749 {
5750 	struct nfsnode *np = VTONFS(vp);
5751 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
5752 	struct nfscllayout *layp;
5753 	struct nfscldevinfo *dip;
5754 	struct nfsclflayout *rflp;
5755 	struct mbuf *m;
5756 	struct nfsclwritedsdorpc *drpc, *tdrpc;
5757 	nfsv4stateid_t stateid;
5758 	struct ucred *newcred;
5759 	uint64_t lastbyte, len, off, oresid, xfer;
5760 	int eof, error, firstmirror, i, iolaymode, mirrorcnt, recalled, timo;
5761 	void *lckp;
5762 	uint8_t *dev;
5763 	void *iovbase = NULL;
5764 	size_t iovlen = 0;
5765 	off_t offs = 0;
5766 	ssize_t resid = 0;
5767 
5768 	if (!NFSHASPNFS(nmp) || nfscl_enablecallb == 0 || nfs_numnfscbd == 0 ||
5769 	    (np->n_flag & NNOLAYOUT) != 0)
5770 		return (EIO);
5771 	/* Now, get a reference cnt on the clientid for this mount. */
5772 	if (nfscl_getref(nmp) == 0)
5773 		return (EIO);
5774 
5775 	/* Find an appropriate stateid. */
5776 	newcred = NFSNEWCRED(cred);
5777 	error = nfscl_getstateid(vp, np->n_fhp->nfh_fh, np->n_fhp->nfh_len,
5778 	    rwaccess, 1, newcred, p, &stateid, &lckp);
5779 	if (error != 0) {
5780 		NFSFREECRED(newcred);
5781 		nfscl_relref(nmp);
5782 		return (error);
5783 	}
5784 	/* Search for a layout for this file. */
5785 	off = uiop->uio_offset;
5786 	layp = nfscl_getlayout(nmp->nm_clp, np->n_fhp->nfh_fh,
5787 	    np->n_fhp->nfh_len, off, &rflp, &recalled);
5788 	if (layp == NULL || rflp == NULL) {
5789 		if (recalled != 0) {
5790 			NFSFREECRED(newcred);
5791 			nfscl_relref(nmp);
5792 			return (EIO);
5793 		}
5794 		if (layp != NULL) {
5795 			nfscl_rellayout(layp, (rflp == NULL) ? 1 : 0);
5796 			layp = NULL;
5797 		}
5798 		/* Try and get a Layout, if it is supported. */
5799 		if (rwaccess == NFSV4OPEN_ACCESSWRITE ||
5800 		    (np->n_flag & NWRITEOPENED) != 0)
5801 			iolaymode = NFSLAYOUTIOMODE_RW;
5802 		else
5803 			iolaymode = NFSLAYOUTIOMODE_READ;
5804 		error = nfsrpc_getlayout(nmp, vp, np->n_fhp, iolaymode,
5805 		    NULL, &stateid, off, &layp, newcred, p);
5806 		if (error != 0) {
5807 			NFSLOCKNODE(np);
5808 			np->n_flag |= NNOLAYOUT;
5809 			NFSUNLOCKNODE(np);
5810 			if (lckp != NULL)
5811 				nfscl_lockderef(lckp);
5812 			NFSFREECRED(newcred);
5813 			if (layp != NULL)
5814 				nfscl_rellayout(layp, 0);
5815 			nfscl_relref(nmp);
5816 			return (error);
5817 		}
5818 	}
5819 
5820 	/*
5821 	 * Loop around finding a layout that works for the first part of
5822 	 * this I/O operation, and then call the function that actually
5823 	 * does the RPC.
5824 	 */
5825 	eof = 0;
5826 	len = (uint64_t)uiop->uio_resid;
5827 	while (len > 0 && error == 0 && eof == 0) {
5828 		off = uiop->uio_offset;
5829 		error = nfscl_findlayoutforio(layp, off, rwaccess, &rflp);
5830 		if (error == 0) {
5831 			oresid = xfer = (uint64_t)uiop->uio_resid;
5832 			if (xfer > (rflp->nfsfl_end - rflp->nfsfl_off))
5833 				xfer = rflp->nfsfl_end - rflp->nfsfl_off;
5834 			/*
5835 			 * For Flex File layout with mirrored DSs, select one
5836 			 * of them at random for reads. For writes and commits,
5837 			 * do all mirrors.
5838 			 */
5839 			m = NULL;
5840 			tdrpc = drpc = NULL;
5841 			firstmirror = 0;
5842 			mirrorcnt = 1;
5843 			if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0 &&
5844 			    (mirrorcnt = rflp->nfsfl_mirrorcnt) > 1) {
5845 				if (rwaccess == NFSV4OPEN_ACCESSREAD) {
5846 					firstmirror = arc4random() % mirrorcnt;
5847 					mirrorcnt = firstmirror + 1;
5848 				} else {
5849 					if (docommit == 0) {
5850 						/*
5851 						 * Save values, so uiop can be
5852 						 * rolled back upon a write
5853 						 * error.
5854 						 */
5855 						offs = uiop->uio_offset;
5856 						resid = uiop->uio_resid;
5857 						iovbase =
5858 						    uiop->uio_iov->iov_base;
5859 						iovlen = uiop->uio_iov->iov_len;
5860 						m = nfsm_uiombuflist(uiop, len,
5861 						    NULL, NULL);
5862 					}
5863 					tdrpc = drpc = malloc(sizeof(*drpc) *
5864 					    (mirrorcnt - 1), M_TEMP, M_WAITOK |
5865 					    M_ZERO);
5866 				}
5867 			}
5868 			for (i = firstmirror; i < mirrorcnt && error == 0; i++){
5869 				if ((layp->nfsly_flags & NFSLY_FLEXFILE) != 0) {
5870 					dev = rflp->nfsfl_ffm[i].dev;
5871 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5872 					    rflp->nfsfl_ffm[i].devp);
5873 				} else {
5874 					dev = rflp->nfsfl_dev;
5875 					dip = nfscl_getdevinfo(nmp->nm_clp, dev,
5876 					    rflp->nfsfl_devp);
5877 				}
5878 				if (dip != NULL) {
5879 					if ((rflp->nfsfl_flags & NFSFL_FLEXFILE)
5880 					    != 0)
5881 						error = nfscl_dofflayoutio(vp,
5882 						    uiop, iomode, must_commit,
5883 						    &eof, &stateid, rwaccess,
5884 						    dip, layp, rflp, off, xfer,
5885 						    i, docommit, m, tdrpc,
5886 						    newcred, p);
5887 					else
5888 						error = nfscl_doflayoutio(vp,
5889 						    uiop, iomode, must_commit,
5890 						    &eof, &stateid, rwaccess,
5891 						    dip, layp, rflp, off, xfer,
5892 						    docommit, newcred, p);
5893 					nfscl_reldevinfo(dip);
5894 				} else
5895 					error = EIO;
5896 				tdrpc++;
5897 			}
5898 			if (m != NULL)
5899 				m_freem(m);
5900 			tdrpc = drpc;
5901 			timo = hz / 50;		/* Wait for 20msec. */
5902 			if (timo < 1)
5903 				timo = 1;
5904 			for (i = firstmirror; i < mirrorcnt - 1 &&
5905 			    tdrpc != NULL; i++, tdrpc++) {
5906 				/*
5907 				 * For the unused drpc entries, both inprog and
5908 				 * err == 0, so this loop won't break.
5909 				 */
5910 				while (tdrpc->inprog != 0 && tdrpc->done == 0)
5911 					tsleep(&tdrpc->tsk, PVFS, "clrpcio",
5912 					    timo);
5913 				if (error == 0 && tdrpc->err != 0)
5914 					error = tdrpc->err;
5915 			}
5916 			free(drpc, M_TEMP);
5917 			if (error == 0) {
5918 				if (mirrorcnt > 1 && rwaccess ==
5919 				    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5920 					NFSLOCKCLSTATE();
5921 					layp->nfsly_flags |= NFSLY_WRITTEN;
5922 					NFSUNLOCKCLSTATE();
5923 				}
5924 				lastbyte = off + xfer - 1;
5925 				NFSLOCKCLSTATE();
5926 				if (lastbyte > layp->nfsly_lastbyte)
5927 					layp->nfsly_lastbyte = lastbyte;
5928 				NFSUNLOCKCLSTATE();
5929 			} else if (error == NFSERR_OPENMODE &&
5930 			    rwaccess == NFSV4OPEN_ACCESSREAD) {
5931 				NFSLOCKMNT(nmp);
5932 				nmp->nm_state |= NFSSTA_OPENMODE;
5933 				NFSUNLOCKMNT(nmp);
5934 			} else
5935 				error = EIO;
5936 			if (error == 0)
5937 				len -= (oresid - (uint64_t)uiop->uio_resid);
5938 			else if (mirrorcnt > 1 && rwaccess ==
5939 			    NFSV4OPEN_ACCESSWRITE && docommit == 0) {
5940 				/*
5941 				 * In case the rpc gets retried, roll the
5942 				 * uio fields changed by nfsm_uiombuflist()
5943 				 * back.
5944 				 */
5945 				uiop->uio_offset = offs;
5946 				uiop->uio_resid = resid;
5947 				uiop->uio_iov->iov_base = iovbase;
5948 				uiop->uio_iov->iov_len = iovlen;
5949 			}
5950 		}
5951 	}
5952 	if (lckp != NULL)
5953 		nfscl_lockderef(lckp);
5954 	NFSFREECRED(newcred);
5955 	nfscl_rellayout(layp, 0);
5956 	nfscl_relref(nmp);
5957 	return (error);
5958 }
5959 
5960 /*
5961  * Make a copy of the mbuf chain and add an mbuf for null padding, as required.
5962  */
5963 static struct mbuf *
5964 nfsm_copym(struct mbuf *m, int off, int xfer)
5965 {
5966 	struct mbuf *m2, *m3, *m4;
5967 	uint32_t *tl;
5968 	int rem;
5969 
5970 	m2 = m_copym(m, off, xfer, M_WAITOK);
5971 	rem = NFSM_RNDUP(xfer) - xfer;
5972 	if (rem > 0) {
5973 		/*
5974 		 * The zero padding to a multiple of 4 bytes is required by
5975 		 * the XDR. So that the mbufs copied by reference aren't
5976 		 * modified, add an mbuf with the zero'd bytes to the list.
5977 		 * rem will be a maximum of 3, so one zero'd uint32_t is
5978 		 * sufficient.
5979 		 */
5980 		m3 = m2;
5981 		while (m3->m_next != NULL)
5982 			m3 = m3->m_next;
5983 		NFSMGET(m4);
5984 		tl = mtod(m4, uint32_t *);
5985 		*tl = 0;
5986 		m4->m_len = rem;
5987 		m3->m_next = m4;
5988 	}
5989 	return (m2);
5990 }
5991 
5992 /*
5993  * Find a file layout that will handle the first bytes of the requested
5994  * range and return the information from it needed to the I/O operation.
5995  */
5996 int
5997 nfscl_findlayoutforio(struct nfscllayout *lyp, uint64_t off, uint32_t rwaccess,
5998     struct nfsclflayout **retflpp)
5999 {
6000 	struct nfsclflayout *flp, *nflp, *rflp;
6001 	uint32_t rw;
6002 
6003 	rflp = NULL;
6004 	rw = rwaccess;
6005 	/* For reading, do the Read list first and then the Write list. */
6006 	do {
6007 		if (rw == NFSV4OPEN_ACCESSREAD)
6008 			flp = LIST_FIRST(&lyp->nfsly_flayread);
6009 		else
6010 			flp = LIST_FIRST(&lyp->nfsly_flayrw);
6011 		while (flp != NULL) {
6012 			nflp = LIST_NEXT(flp, nfsfl_list);
6013 			if (flp->nfsfl_off > off)
6014 				break;
6015 			if (flp->nfsfl_end > off &&
6016 			    (rflp == NULL || rflp->nfsfl_end < flp->nfsfl_end))
6017 				rflp = flp;
6018 			flp = nflp;
6019 		}
6020 		if (rw == NFSV4OPEN_ACCESSREAD)
6021 			rw = NFSV4OPEN_ACCESSWRITE;
6022 		else
6023 			rw = 0;
6024 	} while (rw != 0);
6025 	if (rflp != NULL) {
6026 		/* This one covers the most bytes starting at off. */
6027 		*retflpp = rflp;
6028 		return (0);
6029 	}
6030 	return (EIO);
6031 }
6032 
6033 /*
6034  * Do I/O using an NFSv4.1 or NFSv4.2 file layout.
6035  */
6036 static int
6037 nfscl_doflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6038     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6039     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6040     uint64_t len, int docommit, struct ucred *cred, NFSPROC_T *p)
6041 {
6042 	uint64_t io_off, rel_off, stripe_unit_size, transfer, xfer;
6043 	int commit_thru_mds, error, stripe_index, stripe_pos, minorvers;
6044 	struct nfsnode *np;
6045 	struct nfsfh *fhp;
6046 	struct nfsclds **dspp;
6047 
6048 	np = VTONFS(vp);
6049 	rel_off = off - flp->nfsfl_patoff;
6050 	stripe_unit_size = (flp->nfsfl_util >> 6) & 0x3ffffff;
6051 	stripe_pos = (rel_off / stripe_unit_size + flp->nfsfl_stripe1) %
6052 	    dp->nfsdi_stripecnt;
6053 	transfer = stripe_unit_size - (rel_off % stripe_unit_size);
6054 	error = 0;
6055 
6056 	/* Loop around, doing I/O for each stripe unit. */
6057 	while (len > 0 && error == 0) {
6058 		stripe_index = nfsfldi_stripeindex(dp, stripe_pos);
6059 		dspp = nfsfldi_addr(dp, stripe_index);
6060 		if (((*dspp)->nfsclds_flags & NFSCLDS_MINORV2) != 0)
6061 			minorvers = NFSV42_MINORVERSION;
6062 		else
6063 			minorvers = NFSV41_MINORVERSION;
6064 		if (len > transfer && docommit == 0)
6065 			xfer = transfer;
6066 		else
6067 			xfer = len;
6068 		if ((flp->nfsfl_util & NFSFLAYUTIL_DENSE) != 0) {
6069 			/* Dense layout. */
6070 			if (stripe_pos >= flp->nfsfl_fhcnt)
6071 				return (EIO);
6072 			fhp = flp->nfsfl_fh[stripe_pos];
6073 			io_off = (rel_off / (stripe_unit_size *
6074 			    dp->nfsdi_stripecnt)) * stripe_unit_size +
6075 			    rel_off % stripe_unit_size;
6076 		} else {
6077 			/* Sparse layout. */
6078 			if (flp->nfsfl_fhcnt > 1) {
6079 				if (stripe_index >= flp->nfsfl_fhcnt)
6080 					return (EIO);
6081 				fhp = flp->nfsfl_fh[stripe_index];
6082 			} else if (flp->nfsfl_fhcnt == 1)
6083 				fhp = flp->nfsfl_fh[0];
6084 			else
6085 				fhp = np->n_fhp;
6086 			io_off = off;
6087 		}
6088 		if ((flp->nfsfl_util & NFSFLAYUTIL_COMMIT_THRU_MDS) != 0) {
6089 			commit_thru_mds = 1;
6090 			if (docommit != 0)
6091 				error = EIO;
6092 		} else {
6093 			commit_thru_mds = 0;
6094 			NFSLOCKNODE(np);
6095 			np->n_flag |= NDSCOMMIT;
6096 			NFSUNLOCKNODE(np);
6097 		}
6098 		if (docommit != 0) {
6099 			if (error == 0)
6100 				error = nfsrpc_commitds(vp, io_off, xfer,
6101 				    *dspp, fhp, NFS_VER4, minorvers, cred, p);
6102 			if (error == 0) {
6103 				/*
6104 				 * Set both eof and uio_resid = 0 to end any
6105 				 * loops.
6106 				 */
6107 				*eofp = 1;
6108 				uiop->uio_resid = 0;
6109 			} else {
6110 				NFSLOCKNODE(np);
6111 				np->n_flag &= ~NDSCOMMIT;
6112 				NFSUNLOCKNODE(np);
6113 			}
6114 		} else if (rwflag == NFSV4OPEN_ACCESSREAD)
6115 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6116 			    io_off, xfer, fhp, 0, NFS_VER4, minorvers, cred, p);
6117 		else {
6118 			error = nfsrpc_writeds(vp, uiop, iomode, must_commit,
6119 			    stateidp, *dspp, io_off, xfer, fhp, commit_thru_mds,
6120 			    0, NFS_VER4, minorvers, cred, p);
6121 			if (error == 0) {
6122 				NFSLOCKCLSTATE();
6123 				lyp->nfsly_flags |= NFSLY_WRITTEN;
6124 				NFSUNLOCKCLSTATE();
6125 			}
6126 		}
6127 		if (error == 0) {
6128 			transfer = stripe_unit_size;
6129 			stripe_pos = (stripe_pos + 1) % dp->nfsdi_stripecnt;
6130 			len -= xfer;
6131 			off += xfer;
6132 		}
6133 	}
6134 	return (error);
6135 }
6136 
6137 /*
6138  * Do I/O using an NFSv4.1 flex file layout.
6139  */
6140 static int
6141 nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6142     int *eofp, nfsv4stateid_t *stateidp, int rwflag, struct nfscldevinfo *dp,
6143     struct nfscllayout *lyp, struct nfsclflayout *flp, uint64_t off,
6144     uint64_t len, int mirror, int docommit, struct mbuf *mp,
6145     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6146 {
6147 	uint64_t transfer, xfer;
6148 	int error, rel_off;
6149 	struct nfsnode *np;
6150 	struct nfsfh *fhp;
6151 	struct nfsclds **dspp;
6152 	struct ucred *tcred;
6153 	struct mbuf *m;
6154 
6155 	np = VTONFS(vp);
6156 	error = 0;
6157 	rel_off = 0;
6158 	NFSCL_DEBUG(4, "nfscl_dofflayoutio: off=%ju len=%ju\n", (uintmax_t)off,
6159 	    (uintmax_t)len);
6160 	/* Loop around, doing I/O for each stripe unit. */
6161 	while (len > 0 && error == 0) {
6162 		dspp = nfsfldi_addr(dp, 0);
6163 		fhp = flp->nfsfl_ffm[mirror].fh[dp->nfsdi_versindex];
6164 		stateidp = &flp->nfsfl_ffm[mirror].st;
6165 		NFSCL_DEBUG(4, "mirror=%d vind=%d fhlen=%d st.seqid=0x%x\n",
6166 		    mirror, dp->nfsdi_versindex, fhp->nfh_len, stateidp->seqid);
6167 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) {
6168 			tcred = NFSNEWCRED(cred);
6169 			tcred->cr_uid = flp->nfsfl_ffm[mirror].user;
6170 			tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group;
6171 			tcred->cr_ngroups = 1;
6172 		} else
6173 			tcred = cred;
6174 		if (rwflag == NFSV4OPEN_ACCESSREAD)
6175 			transfer = dp->nfsdi_rsize;
6176 		else
6177 			transfer = dp->nfsdi_wsize;
6178 		NFSLOCKNODE(np);
6179 		np->n_flag |= NDSCOMMIT;
6180 		NFSUNLOCKNODE(np);
6181 		if (len > transfer && docommit == 0)
6182 			xfer = transfer;
6183 		else
6184 			xfer = len;
6185 		if (docommit != 0) {
6186 			if (error == 0) {
6187 				/*
6188 				 * Do last mirrored DS commit with this thread.
6189 				 */
6190 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6191 					error = nfsio_commitds(vp, off, xfer,
6192 					    *dspp, fhp, dp->nfsdi_vers,
6193 					    dp->nfsdi_minorvers, drpc, tcred,
6194 					    p);
6195 				else
6196 					error = nfsrpc_commitds(vp, off, xfer,
6197 					    *dspp, fhp, dp->nfsdi_vers,
6198 					    dp->nfsdi_minorvers, tcred, p);
6199 				NFSCL_DEBUG(4, "commitds=%d\n", error);
6200 				if (error != 0 && error != EACCES && error !=
6201 				    ESTALE) {
6202 					NFSCL_DEBUG(4,
6203 					    "DS layreterr for commit\n");
6204 					nfscl_dserr(NFSV4OP_COMMIT, error, dp,
6205 					    lyp, *dspp);
6206 				}
6207 			}
6208 			NFSCL_DEBUG(4, "aft nfsio_commitds=%d\n", error);
6209 			if (error == 0) {
6210 				/*
6211 				 * Set both eof and uio_resid = 0 to end any
6212 				 * loops.
6213 				 */
6214 				*eofp = 1;
6215 				uiop->uio_resid = 0;
6216 			} else {
6217 				NFSLOCKNODE(np);
6218 				np->n_flag &= ~NDSCOMMIT;
6219 				NFSUNLOCKNODE(np);
6220 			}
6221 		} else if (rwflag == NFSV4OPEN_ACCESSREAD) {
6222 			error = nfsrpc_readds(vp, uiop, stateidp, eofp, *dspp,
6223 			    off, xfer, fhp, 1, dp->nfsdi_vers,
6224 			    dp->nfsdi_minorvers, tcred, p);
6225 			NFSCL_DEBUG(4, "readds=%d\n", error);
6226 			if (error != 0 && error != EACCES && error != ESTALE) {
6227 				NFSCL_DEBUG(4, "DS layreterr for read\n");
6228 				nfscl_dserr(NFSV4OP_READ, error, dp, lyp,
6229 				    *dspp);
6230 			}
6231 		} else {
6232 			if (flp->nfsfl_mirrorcnt == 1) {
6233 				error = nfsrpc_writeds(vp, uiop, iomode,
6234 				    must_commit, stateidp, *dspp, off, xfer,
6235 				    fhp, 0, 1, dp->nfsdi_vers,
6236 				    dp->nfsdi_minorvers, tcred, p);
6237 				if (error == 0) {
6238 					NFSLOCKCLSTATE();
6239 					lyp->nfsly_flags |= NFSLY_WRITTEN;
6240 					NFSUNLOCKCLSTATE();
6241 				}
6242 			} else {
6243 				m = nfsm_copym(mp, rel_off, xfer);
6244 				NFSCL_DEBUG(4, "mcopy reloff=%d xfer=%jd\n",
6245 				    rel_off, (uintmax_t)xfer);
6246 				/*
6247 				 * Do last write to a mirrored DS with this
6248 				 * thread.
6249 				 */
6250 				if (mirror < flp->nfsfl_mirrorcnt - 1)
6251 					error = nfsio_writedsmir(vp, iomode,
6252 					    must_commit, stateidp, *dspp, off,
6253 					    xfer, fhp, m, dp->nfsdi_vers,
6254 					    dp->nfsdi_minorvers, drpc, tcred,
6255 					    p);
6256 				else
6257 					error = nfsrpc_writedsmir(vp, iomode,
6258 					    must_commit, stateidp, *dspp, off,
6259 					    xfer, fhp, m, dp->nfsdi_vers,
6260 					    dp->nfsdi_minorvers, tcred, p);
6261 				NFSCL_DEBUG(4, "nfsio_writedsmir=%d\n", error);
6262 				if (error != 0 && error != EACCES && error !=
6263 				    ESTALE) {
6264 					NFSCL_DEBUG(4,
6265 					    "DS layreterr for write\n");
6266 					nfscl_dserr(NFSV4OP_WRITE, error, dp,
6267 					    lyp, *dspp);
6268 				}
6269 			}
6270 		}
6271 		NFSCL_DEBUG(4, "aft read/writeds=%d\n", error);
6272 		if (error == 0) {
6273 			len -= xfer;
6274 			off += xfer;
6275 			rel_off += xfer;
6276 		}
6277 		if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0)
6278 			NFSFREECRED(tcred);
6279 	}
6280 	NFSCL_DEBUG(4, "eo nfscl_dofflayoutio=%d\n", error);
6281 	return (error);
6282 }
6283 
6284 /*
6285  * The actual read RPC done to a DS.
6286  */
6287 static int
6288 nfsrpc_readds(vnode_t vp, struct uio *uiop, nfsv4stateid_t *stateidp, int *eofp,
6289     struct nfsclds *dsp, uint64_t io_off, int len, struct nfsfh *fhp, int flex,
6290     int vers, int minorvers, struct ucred *cred, NFSPROC_T *p)
6291 {
6292 	uint32_t *tl;
6293 	int attrflag, error, retlen;
6294 	struct nfsrv_descript nfsd;
6295 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6296 	struct nfsrv_descript *nd = &nfsd;
6297 	struct nfssockreq *nrp;
6298 	struct nfsvattr na;
6299 
6300 	nd->nd_mrep = NULL;
6301 	if (vers == 0 || vers == NFS_VER4) {
6302 		nfscl_reqstart(nd, NFSPROC_READDS, nmp, fhp->nfh_fh,
6303 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6304 		vers = NFS_VER4;
6305 		NFSCL_DEBUG(4, "nfsrpc_readds: vers4 minvers=%d\n", minorvers);
6306 		if (flex != 0)
6307 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6308 		else
6309 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6310 	} else {
6311 		nfscl_reqstart(nd, NFSPROC_READ, nmp, fhp->nfh_fh,
6312 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6313 		NFSCL_DEBUG(4, "nfsrpc_readds: vers3\n");
6314 	}
6315 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED * 3);
6316 	txdr_hyper(io_off, tl);
6317 	*(tl + 2) = txdr_unsigned(len);
6318 	nrp = dsp->nfsclds_sockp;
6319 	NFSCL_DEBUG(4, "nfsrpc_readds: nrp=%p\n", nrp);
6320 	if (nrp == NULL)
6321 		/* If NULL, use the MDS socket. */
6322 		nrp = &nmp->nm_sockreq;
6323 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6324 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6325 	NFSCL_DEBUG(4, "nfsrpc_readds: stat=%d err=%d\n", nd->nd_repstat,
6326 	    error);
6327 	if (error != 0)
6328 		return (error);
6329 	if (vers == NFS_VER3) {
6330 		error = nfscl_postop_attr(nd, &na, &attrflag, NULL);
6331 		NFSCL_DEBUG(4, "nfsrpc_readds: postop=%d\n", error);
6332 		if (error != 0)
6333 			goto nfsmout;
6334 	}
6335 	if (nd->nd_repstat != 0) {
6336 		error = nd->nd_repstat;
6337 		goto nfsmout;
6338 	}
6339 	if (vers == NFS_VER3) {
6340 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
6341 		*eofp = fxdr_unsigned(int, *(tl + 1));
6342 	} else {
6343 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
6344 		*eofp = fxdr_unsigned(int, *tl);
6345 	}
6346 	NFSM_STRSIZ(retlen, len);
6347 	NFSCL_DEBUG(4, "nfsrpc_readds: retlen=%d eof=%d\n", retlen, *eofp);
6348 	error = nfsm_mbufuio(nd, uiop, retlen);
6349 nfsmout:
6350 	if (nd->nd_mrep != NULL)
6351 		m_freem(nd->nd_mrep);
6352 	return (error);
6353 }
6354 
6355 /*
6356  * The actual write RPC done to a DS.
6357  */
6358 static int
6359 nfsrpc_writeds(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit,
6360     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6361     struct nfsfh *fhp, int commit_thru_mds, int flex, int vers, int minorvers,
6362     struct ucred *cred, NFSPROC_T *p)
6363 {
6364 	uint32_t *tl;
6365 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6366 	int attrflag, error, rlen, commit, committed = NFSWRITE_FILESYNC;
6367 	int32_t backup;
6368 	struct nfsrv_descript nfsd;
6369 	struct nfsrv_descript *nd = &nfsd;
6370 	struct nfssockreq *nrp;
6371 	struct nfsvattr na;
6372 
6373 	KASSERT(uiop->uio_iovcnt == 1, ("nfs: writerpc iovcnt > 1"));
6374 	nd->nd_mrep = NULL;
6375 	if (vers == 0 || vers == NFS_VER4) {
6376 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6377 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6378 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers4 minvers=%d\n", minorvers);
6379 		vers = NFS_VER4;
6380 		if (flex != 0)
6381 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6382 		else
6383 			nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSEQIDZERO);
6384 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6385 	} else {
6386 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6387 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6388 		NFSCL_DEBUG(4, "nfsrpc_writeds: vers3\n");
6389 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6390 	}
6391 	txdr_hyper(io_off, tl);
6392 	tl += 2;
6393 	if (vers == NFS_VER3)
6394 		*tl++ = txdr_unsigned(len);
6395 	*tl++ = txdr_unsigned(*iomode);
6396 	*tl = txdr_unsigned(len);
6397 	nfsm_uiombuf(nd, uiop, len);
6398 	nrp = dsp->nfsclds_sockp;
6399 	if (nrp == NULL)
6400 		/* If NULL, use the MDS socket. */
6401 		nrp = &nmp->nm_sockreq;
6402 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6403 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6404 	NFSCL_DEBUG(4, "nfsrpc_writeds: err=%d stat=%d\n", error,
6405 	    nd->nd_repstat);
6406 	if (error != 0)
6407 		return (error);
6408 	if (nd->nd_repstat != 0) {
6409 		/*
6410 		 * In case the rpc gets retried, roll
6411 		 * the uio fileds changed by nfsm_uiombuf()
6412 		 * back.
6413 		 */
6414 		uiop->uio_offset -= len;
6415 		uiop->uio_resid += len;
6416 		uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base - len;
6417 		uiop->uio_iov->iov_len += len;
6418 		error = nd->nd_repstat;
6419 	} else {
6420 		if (vers == NFS_VER3) {
6421 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6422 			    NULL);
6423 			NFSCL_DEBUG(4, "nfsrpc_writeds: wcc_data=%d\n", error);
6424 			if (error != 0)
6425 				goto nfsmout;
6426 		}
6427 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6428 		rlen = fxdr_unsigned(int, *tl++);
6429 		NFSCL_DEBUG(4, "nfsrpc_writeds: len=%d rlen=%d\n", len, rlen);
6430 		if (rlen == 0) {
6431 			error = NFSERR_IO;
6432 			goto nfsmout;
6433 		} else if (rlen < len) {
6434 			backup = len - rlen;
6435 			uiop->uio_iov->iov_base =
6436 			    (char *)uiop->uio_iov->iov_base - backup;
6437 			uiop->uio_iov->iov_len += backup;
6438 			uiop->uio_offset -= backup;
6439 			uiop->uio_resid += backup;
6440 			len = rlen;
6441 		}
6442 		commit = fxdr_unsigned(int, *tl++);
6443 
6444 		/*
6445 		 * Return the lowest commitment level
6446 		 * obtained by any of the RPCs.
6447 		 */
6448 		if (committed == NFSWRITE_FILESYNC)
6449 			committed = commit;
6450 		else if (committed == NFSWRITE_DATASYNC &&
6451 		    commit == NFSWRITE_UNSTABLE)
6452 			committed = commit;
6453 		if (commit_thru_mds != 0) {
6454 			NFSLOCKMNT(nmp);
6455 			if (!NFSHASWRITEVERF(nmp)) {
6456 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6457 				NFSSETWRITEVERF(nmp);
6458 	    		} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
6459 				*must_commit = 1;
6460 				NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
6461 			}
6462 			NFSUNLOCKMNT(nmp);
6463 		} else {
6464 			NFSLOCKDS(dsp);
6465 			if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6466 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6467 				dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6468 			} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6469 				*must_commit = 1;
6470 				NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6471 			}
6472 			NFSUNLOCKDS(dsp);
6473 		}
6474 	}
6475 nfsmout:
6476 	if (nd->nd_mrep != NULL)
6477 		m_freem(nd->nd_mrep);
6478 	*iomode = committed;
6479 	if (nd->nd_repstat != 0 && error == 0)
6480 		error = nd->nd_repstat;
6481 	return (error);
6482 }
6483 
6484 /*
6485  * The actual write RPC done to a DS.
6486  * This variant is called from a separate kernel process for mirrors.
6487  * Any short write is considered an IO error.
6488  */
6489 static int
6490 nfsrpc_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6491     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t io_off, int len,
6492     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6493     struct ucred *cred, NFSPROC_T *p)
6494 {
6495 	uint32_t *tl;
6496 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6497 	int attrflag, error, commit, committed = NFSWRITE_FILESYNC, rlen;
6498 	struct nfsrv_descript nfsd;
6499 	struct nfsrv_descript *nd = &nfsd;
6500 	struct nfssockreq *nrp;
6501 	struct nfsvattr na;
6502 
6503 	nd->nd_mrep = NULL;
6504 	if (vers == 0 || vers == NFS_VER4) {
6505 		nfscl_reqstart(nd, NFSPROC_WRITEDS, nmp, fhp->nfh_fh,
6506 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6507 		vers = NFS_VER4;
6508 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers4 minvers=%d\n",
6509 		    minorvers);
6510 		nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
6511 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
6512 	} else {
6513 		nfscl_reqstart(nd, NFSPROC_WRITE, nmp, fhp->nfh_fh,
6514 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6515 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: vers3\n");
6516 		NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED);
6517 	}
6518 	txdr_hyper(io_off, tl);
6519 	tl += 2;
6520 	if (vers == NFS_VER3)
6521 		*tl++ = txdr_unsigned(len);
6522 	*tl++ = txdr_unsigned(*iomode);
6523 	*tl = txdr_unsigned(len);
6524 	if (len > 0) {
6525 		/* Put data in mbuf chain. */
6526 		nd->nd_mb->m_next = m;
6527 		/* Set nd_mb and nd_bpos to end of data. */
6528 		while (m->m_next != NULL)
6529 			m = m->m_next;
6530 		nd->nd_mb = m;
6531 		nd->nd_bpos = mtod(m, char *) + m->m_len;
6532 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: lastmb len=%d\n", m->m_len);
6533 	}
6534 	nrp = dsp->nfsclds_sockp;
6535 	if (nrp == NULL)
6536 		/* If NULL, use the MDS socket. */
6537 		nrp = &nmp->nm_sockreq;
6538 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6539 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6540 	NFSCL_DEBUG(4, "nfsrpc_writedsmir: err=%d stat=%d\n", error,
6541 	    nd->nd_repstat);
6542 	if (error != 0)
6543 		return (error);
6544 	if (nd->nd_repstat != 0)
6545 		error = nd->nd_repstat;
6546 	else {
6547 		if (vers == NFS_VER3) {
6548 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6549 			    NULL);
6550 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: wcc_data=%d\n",
6551 			    error);
6552 			if (error != 0)
6553 				goto nfsmout;
6554 		}
6555 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_VERF);
6556 		rlen = fxdr_unsigned(int, *tl++);
6557 		NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n", len,
6558 		    rlen);
6559 		if (rlen != len) {
6560 			error = NFSERR_IO;
6561 			NFSCL_DEBUG(4, "nfsrpc_writedsmir: len=%d rlen=%d\n",
6562 			    len, rlen);
6563 			goto nfsmout;
6564 		}
6565 		commit = fxdr_unsigned(int, *tl++);
6566 
6567 		/*
6568 		 * Return the lowest commitment level
6569 		 * obtained by any of the RPCs.
6570 		 */
6571 		if (committed == NFSWRITE_FILESYNC)
6572 			committed = commit;
6573 		else if (committed == NFSWRITE_DATASYNC &&
6574 		    commit == NFSWRITE_UNSTABLE)
6575 			committed = commit;
6576 		NFSLOCKDS(dsp);
6577 		if ((dsp->nfsclds_flags & NFSCLDS_HASWRITEVERF) == 0) {
6578 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6579 			dsp->nfsclds_flags |= NFSCLDS_HASWRITEVERF;
6580 		} else if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6581 			*must_commit = 1;
6582 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6583 		}
6584 		NFSUNLOCKDS(dsp);
6585 	}
6586 nfsmout:
6587 	if (nd->nd_mrep != NULL)
6588 		m_freem(nd->nd_mrep);
6589 	*iomode = committed;
6590 	if (nd->nd_repstat != 0 && error == 0)
6591 		error = nd->nd_repstat;
6592 	return (error);
6593 }
6594 
6595 /*
6596  * Start up the thread that will execute nfsrpc_writedsmir().
6597  */
6598 static void
6599 start_writedsmir(void *arg, int pending)
6600 {
6601 	struct nfsclwritedsdorpc *drpc;
6602 
6603 	drpc = (struct nfsclwritedsdorpc *)arg;
6604 	drpc->err = nfsrpc_writedsmir(drpc->vp, &drpc->iomode,
6605 	    &drpc->must_commit, drpc->stateidp, drpc->dsp, drpc->off, drpc->len,
6606 	    drpc->fhp, drpc->m, drpc->vers, drpc->minorvers, drpc->cred,
6607 	    drpc->p);
6608 	drpc->done = 1;
6609 	NFSCL_DEBUG(4, "start_writedsmir: err=%d\n", drpc->err);
6610 }
6611 
6612 /*
6613  * Set up the write DS mirror call for the pNFS I/O thread.
6614  */
6615 static int
6616 nfsio_writedsmir(vnode_t vp, int *iomode, int *must_commit,
6617     nfsv4stateid_t *stateidp, struct nfsclds *dsp, uint64_t off, int len,
6618     struct nfsfh *fhp, struct mbuf *m, int vers, int minorvers,
6619     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6620 {
6621 	int error, ret;
6622 
6623 	error = 0;
6624 	drpc->done = 0;
6625 	drpc->vp = vp;
6626 	drpc->iomode = *iomode;
6627 	drpc->must_commit = *must_commit;
6628 	drpc->stateidp = stateidp;
6629 	drpc->dsp = dsp;
6630 	drpc->off = off;
6631 	drpc->len = len;
6632 	drpc->fhp = fhp;
6633 	drpc->m = m;
6634 	drpc->vers = vers;
6635 	drpc->minorvers = minorvers;
6636 	drpc->cred = cred;
6637 	drpc->p = p;
6638 	drpc->inprog = 0;
6639 	ret = EIO;
6640 	if (nfs_pnfsiothreads != 0) {
6641 		ret = nfs_pnfsio(start_writedsmir, drpc);
6642 		NFSCL_DEBUG(4, "nfsio_writedsmir: nfs_pnfsio=%d\n", ret);
6643 	}
6644 	if (ret != 0)
6645 		error = nfsrpc_writedsmir(vp, iomode, must_commit, stateidp,
6646 		    dsp, off, len, fhp, m, vers, minorvers, cred, p);
6647 	NFSCL_DEBUG(4, "nfsio_writedsmir: error=%d\n", error);
6648 	return (error);
6649 }
6650 
6651 /*
6652  * Free up the nfsclds structure.
6653  */
6654 void
6655 nfscl_freenfsclds(struct nfsclds *dsp)
6656 {
6657 	int i;
6658 
6659 	if (dsp == NULL)
6660 		return;
6661 	if (dsp->nfsclds_sockp != NULL) {
6662 		NFSFREECRED(dsp->nfsclds_sockp->nr_cred);
6663 		NFSFREEMUTEX(&dsp->nfsclds_sockp->nr_mtx);
6664 		free(dsp->nfsclds_sockp->nr_nam, M_SONAME);
6665 		free(dsp->nfsclds_sockp, M_NFSSOCKREQ);
6666 	}
6667 	NFSFREEMUTEX(&dsp->nfsclds_mtx);
6668 	NFSFREEMUTEX(&dsp->nfsclds_sess.nfsess_mtx);
6669 	for (i = 0; i < NFSV4_CBSLOTS; i++) {
6670 		if (dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply != NULL)
6671 			m_freem(
6672 			    dsp->nfsclds_sess.nfsess_cbslots[i].nfssl_reply);
6673 	}
6674 	free(dsp, M_NFSCLDS);
6675 }
6676 
6677 static enum nfsclds_state
6678 nfscl_getsameserver(struct nfsmount *nmp, struct nfsclds *newdsp,
6679     struct nfsclds **retdspp, uint32_t *sequencep)
6680 {
6681 	struct nfsclds *dsp;
6682 	int fndseq;
6683 
6684 	/*
6685 	 * Search the list of nfsclds structures for one with the same
6686 	 * server.
6687 	 */
6688 	fndseq = 0;
6689 	TAILQ_FOREACH(dsp, &nmp->nm_sess, nfsclds_list) {
6690 		if (dsp->nfsclds_servownlen == newdsp->nfsclds_servownlen &&
6691 		    dsp->nfsclds_servownlen != 0 &&
6692 		    !NFSBCMP(dsp->nfsclds_serverown, newdsp->nfsclds_serverown,
6693 		    dsp->nfsclds_servownlen) &&
6694 		    dsp->nfsclds_sess.nfsess_defunct == 0) {
6695 			NFSCL_DEBUG(4, "fnd same fdsp=%p dsp=%p flg=0x%x\n",
6696 			    TAILQ_FIRST(&nmp->nm_sess), dsp,
6697 			    dsp->nfsclds_flags);
6698 			if (fndseq == 0) {
6699 				/* Get sequenceid# from first entry. */
6700 				*sequencep =
6701 				    dsp->nfsclds_sess.nfsess_sequenceid;
6702 				fndseq = 1;
6703 			}
6704 			/* Server major id matches. */
6705 			if ((dsp->nfsclds_flags & NFSCLDS_DS) != 0) {
6706 				*retdspp = dsp;
6707 				return (NFSDSP_USETHISSESSION);
6708 			}
6709 
6710 		}
6711 	}
6712 	if (fndseq != 0)
6713 		return (NFSDSP_SEQTHISSESSION);
6714 	return (NFSDSP_NOTFOUND);
6715 }
6716 
6717 /*
6718  * NFS commit rpc to a NFSv4.1 DS.
6719  */
6720 static int
6721 nfsrpc_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6722     struct nfsfh *fhp, int vers, int minorvers, struct ucred *cred,
6723     NFSPROC_T *p)
6724 {
6725 	uint32_t *tl;
6726 	struct nfsrv_descript nfsd, *nd = &nfsd;
6727 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6728 	struct nfssockreq *nrp;
6729 	struct nfsvattr na;
6730 	int attrflag, error;
6731 
6732 	nd->nd_mrep = NULL;
6733 	if (vers == 0 || vers == NFS_VER4) {
6734 		nfscl_reqstart(nd, NFSPROC_COMMITDS, nmp, fhp->nfh_fh,
6735 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6736 		vers = NFS_VER4;
6737 	} else
6738 		nfscl_reqstart(nd, NFSPROC_COMMIT, nmp, fhp->nfh_fh,
6739 		    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6740 	NFSCL_DEBUG(4, "nfsrpc_commitds: vers=%d minvers=%d\n", vers,
6741 	    minorvers);
6742 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6743 	txdr_hyper(offset, tl);
6744 	tl += 2;
6745 	*tl = txdr_unsigned(cnt);
6746 	nrp = dsp->nfsclds_sockp;
6747 	if (nrp == NULL)
6748 		/* If NULL, use the MDS socket. */
6749 		nrp = &nmp->nm_sockreq;
6750 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6751 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6752 	NFSCL_DEBUG(4, "nfsrpc_commitds: err=%d stat=%d\n", error,
6753 	    nd->nd_repstat);
6754 	if (error != 0)
6755 		return (error);
6756 	if (nd->nd_repstat == 0) {
6757 		if (vers == NFS_VER3) {
6758 			error = nfscl_wcc_data(nd, vp, &na, &attrflag, NULL,
6759 			    NULL);
6760 			NFSCL_DEBUG(4, "nfsrpc_commitds: wccdata=%d\n", error);
6761 			if (error != 0)
6762 				goto nfsmout;
6763 		}
6764 		NFSM_DISSECT(tl, u_int32_t *, NFSX_VERF);
6765 		NFSLOCKDS(dsp);
6766 		if (NFSBCMP(tl, dsp->nfsclds_verf, NFSX_VERF)) {
6767 			NFSBCOPY(tl, dsp->nfsclds_verf, NFSX_VERF);
6768 			error = NFSERR_STALEWRITEVERF;
6769 		}
6770 		NFSUNLOCKDS(dsp);
6771 	}
6772 nfsmout:
6773 	if (error == 0 && nd->nd_repstat != 0)
6774 		error = nd->nd_repstat;
6775 	m_freem(nd->nd_mrep);
6776 	return (error);
6777 }
6778 
6779 /*
6780  * Start up the thread that will execute nfsrpc_commitds().
6781  */
6782 static void
6783 start_commitds(void *arg, int pending)
6784 {
6785 	struct nfsclwritedsdorpc *drpc;
6786 
6787 	drpc = (struct nfsclwritedsdorpc *)arg;
6788 	drpc->err = nfsrpc_commitds(drpc->vp, drpc->off, drpc->len,
6789 	    drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers, drpc->cred,
6790 	    drpc->p);
6791 	drpc->done = 1;
6792 	NFSCL_DEBUG(4, "start_commitds: err=%d\n", drpc->err);
6793 }
6794 
6795 /*
6796  * Set up the commit DS mirror call for the pNFS I/O thread.
6797  */
6798 static int
6799 nfsio_commitds(vnode_t vp, uint64_t offset, int cnt, struct nfsclds *dsp,
6800     struct nfsfh *fhp, int vers, int minorvers,
6801     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6802 {
6803 	int error, ret;
6804 
6805 	error = 0;
6806 	drpc->done = 0;
6807 	drpc->vp = vp;
6808 	drpc->off = offset;
6809 	drpc->len = cnt;
6810 	drpc->dsp = dsp;
6811 	drpc->fhp = fhp;
6812 	drpc->vers = vers;
6813 	drpc->minorvers = minorvers;
6814 	drpc->cred = cred;
6815 	drpc->p = p;
6816 	drpc->inprog = 0;
6817 	ret = EIO;
6818 	if (nfs_pnfsiothreads != 0) {
6819 		ret = nfs_pnfsio(start_commitds, drpc);
6820 		NFSCL_DEBUG(4, "nfsio_commitds: nfs_pnfsio=%d\n", ret);
6821 	}
6822 	if (ret != 0)
6823 		error = nfsrpc_commitds(vp, offset, cnt, dsp, fhp, vers,
6824 		    minorvers, cred, p);
6825 	NFSCL_DEBUG(4, "nfsio_commitds: error=%d\n", error);
6826 	return (error);
6827 }
6828 
6829 /*
6830  * NFS Advise rpc
6831  */
6832 int
6833 nfsrpc_advise(vnode_t vp, off_t offset, uint64_t cnt, int advise,
6834     struct ucred *cred, NFSPROC_T *p)
6835 {
6836 	u_int32_t *tl;
6837 	struct nfsrv_descript nfsd, *nd = &nfsd;
6838 	nfsattrbit_t hints;
6839 	int error;
6840 
6841 	NFSZERO_ATTRBIT(&hints);
6842 	if (advise == POSIX_FADV_WILLNEED)
6843 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6844 	else if (advise == POSIX_FADV_DONTNEED)
6845 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6846 	else
6847 		return (0);
6848 	NFSCL_REQSTART(nd, NFSPROC_IOADVISE, vp);
6849 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6850 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER);
6851 	txdr_hyper(offset, tl);
6852 	tl += 2;
6853 	txdr_hyper(cnt, tl);
6854 	nfsrv_putattrbit(nd, &hints);
6855 	error = nfscl_request(nd, vp, p, cred, NULL);
6856 	if (error != 0)
6857 		return (error);
6858 	if (nd->nd_repstat != 0)
6859 		error = nd->nd_repstat;
6860 	m_freem(nd->nd_mrep);
6861 	return (error);
6862 }
6863 
6864 #ifdef notyet
6865 /*
6866  * NFS advise rpc to a NFSv4.2 DS.
6867  */
6868 static int
6869 nfsrpc_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6870     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6871     struct ucred *cred, NFSPROC_T *p)
6872 {
6873 	uint32_t *tl;
6874 	struct nfsrv_descript nfsd, *nd = &nfsd;
6875 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6876 	struct nfssockreq *nrp;
6877 	nfsattrbit_t hints;
6878 	int error;
6879 
6880 	/* For NFS DSs prior to NFSv4.2, just return OK. */
6881 	if (vers == NFS_VER3 || minorversion < NFSV42_MINORVERSION)
6882 		return (0);
6883 	NFSZERO_ATTRBIT(&hints);
6884 	if (advise == POSIX_FADV_WILLNEED)
6885 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_WILLNEED);
6886 	else if (advise == POSIX_FADV_DONTNEED)
6887 		NFSSETBIT_ATTRBIT(&hints, NFSV4IOHINT_DONTNEED);
6888 	else
6889 		return (0);
6890 	nd->nd_mrep = NULL;
6891 	nfscl_reqstart(nd, NFSPROC_IOADVISEDS, nmp, fhp->nfh_fh,
6892 	    fhp->nfh_len, NULL, &dsp->nfsclds_sess, vers, minorvers);
6893 	vers = NFS_VER4;
6894 	NFSCL_DEBUG(4, "nfsrpc_adviseds: vers=%d minvers=%d\n", vers,
6895 	    minorvers);
6896 	nfsm_stateidtom(nd, NULL, NFSSTATEID_PUTALLZERO);
6897 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
6898 	txdr_hyper(offset, tl);
6899 	tl += 2;
6900 	*tl = txdr_unsigned(cnt);
6901 	nfsrv_putattrbit(nd, &hints);
6902 	nrp = dsp->nfsclds_sockp;
6903 	if (nrp == NULL)
6904 		/* If NULL, use the MDS socket. */
6905 		nrp = &nmp->nm_sockreq;
6906 	error = newnfs_request(nd, nmp, NULL, nrp, vp, p, cred,
6907 	    NFS_PROG, vers, NULL, 1, NULL, &dsp->nfsclds_sess);
6908 	NFSCL_DEBUG(4, "nfsrpc_adviseds: err=%d stat=%d\n", error,
6909 	    nd->nd_repstat);
6910 	if (error != 0)
6911 		return (error);
6912 	if (nd->nd_repstat != 0)
6913 		error = nd->nd_repstat;
6914 	m_freem(nd->nd_mrep);
6915 	return (error);
6916 }
6917 
6918 /*
6919  * Start up the thread that will execute nfsrpc_commitds().
6920  */
6921 static void
6922 start_adviseds(void *arg, int pending)
6923 {
6924 	struct nfsclwritedsdorpc *drpc;
6925 
6926 	drpc = (struct nfsclwritedsdorpc *)arg;
6927 	drpc->err = nfsrpc_adviseds(drpc->vp, drpc->off, drpc->len,
6928 	    drpc->advise, drpc->dsp, drpc->fhp, drpc->vers, drpc->minorvers,
6929 	    drpc->cred, drpc->p);
6930 	drpc->done = 1;
6931 	NFSCL_DEBUG(4, "start_adviseds: err=%d\n", drpc->err);
6932 }
6933 
6934 /*
6935  * Set up the commit DS mirror call for the pNFS I/O thread.
6936  */
6937 static int
6938 nfsio_adviseds(vnode_t vp, uint64_t offset, int cnt, int advise,
6939     struct nfsclds *dsp, struct nfsfh *fhp, int vers, int minorvers,
6940     struct nfsclwritedsdorpc *drpc, struct ucred *cred, NFSPROC_T *p)
6941 {
6942 	int error, ret;
6943 
6944 	error = 0;
6945 	drpc->done = 0;
6946 	drpc->vp = vp;
6947 	drpc->off = offset;
6948 	drpc->len = cnt;
6949 	drpc->advise = advise;
6950 	drpc->dsp = dsp;
6951 	drpc->fhp = fhp;
6952 	drpc->vers = vers;
6953 	drpc->minorvers = minorvers;
6954 	drpc->cred = cred;
6955 	drpc->p = p;
6956 	drpc->inprog = 0;
6957 	ret = EIO;
6958 	if (nfs_pnfsiothreads != 0) {
6959 		ret = nfs_pnfsio(start_adviseds, drpc);
6960 		NFSCL_DEBUG(4, "nfsio_adviseds: nfs_pnfsio=%d\n", ret);
6961 	}
6962 	if (ret != 0)
6963 		error = nfsrpc_adviseds(vp, offset, cnt, advise, dsp, fhp, vers,
6964 		    minorvers, cred, p);
6965 	NFSCL_DEBUG(4, "nfsio_adviseds: error=%d\n", error);
6966 	return (error);
6967 }
6968 #endif	/* notyet */
6969 
6970 /*
6971  * Do the Allocate operation, retrying for recovery.
6972  */
6973 int
6974 nfsrpc_allocate(vnode_t vp, off_t off, off_t len, struct nfsvattr *nap,
6975     int *attrflagp, struct ucred *cred, NFSPROC_T *p, void *stuff)
6976 {
6977 	int error, expireret = 0, retrycnt, nostateid;
6978 	uint32_t clidrev = 0;
6979 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
6980 	struct nfsfh *nfhp = NULL;
6981 	nfsv4stateid_t stateid;
6982 	off_t tmp_off;
6983 	void *lckp;
6984 
6985 	if (len < 0)
6986 		return (EINVAL);
6987 	if (len == 0)
6988 		return (0);
6989 	tmp_off = off + len;
6990 	NFSLOCKMNT(nmp);
6991 	if (tmp_off > nmp->nm_maxfilesize || tmp_off < off) {
6992 		NFSUNLOCKMNT(nmp);
6993 		return (EFBIG);
6994 	}
6995 	if (nmp->nm_clp != NULL)
6996 		clidrev = nmp->nm_clp->nfsc_clientidrev;
6997 	NFSUNLOCKMNT(nmp);
6998 	nfhp = VTONFS(vp)->n_fhp;
6999 	retrycnt = 0;
7000 	do {
7001 		lckp = NULL;
7002 		nostateid = 0;
7003 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
7004 		    NFSV4OPEN_ACCESSWRITE, 0, cred, p, &stateid, &lckp);
7005 		if (stateid.other[0] == 0 && stateid.other[1] == 0 &&
7006 		    stateid.other[2] == 0) {
7007 			nostateid = 1;
7008 			NFSCL_DEBUG(1, "stateid0 in allocate\n");
7009 		}
7010 
7011 		/*
7012 		 * Not finding a stateid should probably never happen,
7013 		 * but just return an error for this case.
7014 		 */
7015 		if (nostateid != 0)
7016 			error = EIO;
7017 		else
7018 			error = nfsrpc_allocaterpc(vp, off, len, &stateid,
7019 			    nap, attrflagp, cred, p, stuff);
7020 		if (error == NFSERR_STALESTATEID)
7021 			nfscl_initiate_recovery(nmp->nm_clp);
7022 		if (lckp != NULL)
7023 			nfscl_lockderef(lckp);
7024 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
7025 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
7026 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
7027 			(void) nfs_catnap(PZERO, error, "nfs_allocate");
7028 		} else if ((error == NFSERR_EXPIRED ||
7029 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
7030 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev, p);
7031 		}
7032 		retrycnt++;
7033 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
7034 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
7035 	    error == NFSERR_STALEDONTRECOVER ||
7036 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
7037 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
7038 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
7039 	if (error != 0 && retrycnt >= 4)
7040 		error = EIO;
7041 	return (error);
7042 }
7043 
7044 /*
7045  * The allocate RPC.
7046  */
7047 static int
7048 nfsrpc_allocaterpc(vnode_t vp, off_t off, off_t len, nfsv4stateid_t *stateidp,
7049     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p,
7050     void *stuff)
7051 {
7052 	uint32_t *tl;
7053 	int error;
7054 	struct nfsrv_descript nfsd;
7055 	struct nfsrv_descript *nd = &nfsd;
7056 	nfsattrbit_t attrbits;
7057 
7058 	*attrflagp = 0;
7059 	NFSCL_REQSTART(nd, NFSPROC_ALLOCATE, vp);
7060 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
7061 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_HYPER + NFSX_UNSIGNED);
7062 	txdr_hyper(off, tl); tl += 2;
7063 	txdr_hyper(len, tl); tl += 2;
7064 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7065 	NFSGETATTR_ATTRBIT(&attrbits);
7066 	nfsrv_putattrbit(nd, &attrbits);
7067 	error = nfscl_request(nd, vp, p, cred, stuff);
7068 	if (error != 0)
7069 		return (error);
7070 	if (nd->nd_repstat == 0) {
7071 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7072 		error = nfsm_loadattr(nd, nap);
7073 		if (error == 0)
7074 			*attrflagp = NFS_LATTR_NOSHRINK;
7075 	} else
7076 		error = nd->nd_repstat;
7077 nfsmout:
7078 	m_freem(nd->nd_mrep);
7079 	return (error);
7080 }
7081 
7082 /*
7083  * Set up the XDR arguments for the LayoutGet operation.
7084  */
7085 static void
7086 nfsrv_setuplayoutget(struct nfsrv_descript *nd, int iomode, uint64_t offset,
7087     uint64_t len, uint64_t minlen, nfsv4stateid_t *stateidp, int layouttype,
7088     int layoutlen, int usecurstateid)
7089 {
7090 	uint32_t *tl;
7091 
7092 	NFSM_BUILD(tl, uint32_t *, 4 * NFSX_UNSIGNED + 3 * NFSX_HYPER +
7093 	    NFSX_STATEID);
7094 	*tl++ = newnfs_false;		/* Don't signal availability. */
7095 	*tl++ = txdr_unsigned(layouttype);
7096 	*tl++ = txdr_unsigned(iomode);
7097 	txdr_hyper(offset, tl);
7098 	tl += 2;
7099 	txdr_hyper(len, tl);
7100 	tl += 2;
7101 	txdr_hyper(minlen, tl);
7102 	tl += 2;
7103 	if (usecurstateid != 0) {
7104 		/* Special stateid for Current stateid. */
7105 		*tl++ = txdr_unsigned(1);
7106 		*tl++ = 0;
7107 		*tl++ = 0;
7108 		*tl++ = 0;
7109 	} else {
7110 		*tl++ = txdr_unsigned(stateidp->seqid);
7111 		NFSCL_DEBUG(4, "layget seq=%d\n", (int)stateidp->seqid);
7112 		*tl++ = stateidp->other[0];
7113 		*tl++ = stateidp->other[1];
7114 		*tl++ = stateidp->other[2];
7115 	}
7116 	*tl = txdr_unsigned(layoutlen);
7117 }
7118 
7119 /*
7120  * Parse the reply for a successful LayoutGet operation.
7121  */
7122 static int
7123 nfsrv_parselayoutget(struct nfsmount *nmp, struct nfsrv_descript *nd,
7124     nfsv4stateid_t *stateidp, int *retonclosep, struct nfsclflayouthead *flhp)
7125 {
7126 	uint32_t *tl;
7127 	struct nfsclflayout *flp, *prevflp, *tflp;
7128 	int cnt, error, fhcnt, gotiomode, i, iomode, j, k, l, laytype, nfhlen;
7129 	int m, mirrorcnt;
7130 	uint64_t retlen, off;
7131 	struct nfsfh *nfhp;
7132 	uint8_t *cp;
7133 	uid_t user;
7134 	gid_t grp;
7135 
7136 	NFSCL_DEBUG(4, "in nfsrv_parselayoutget\n");
7137 	error = 0;
7138 	flp = NULL;
7139 	gotiomode = -1;
7140 	NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED + NFSX_STATEID);
7141 	if (*tl++ != 0)
7142 		*retonclosep = 1;
7143 	else
7144 		*retonclosep = 0;
7145 	stateidp->seqid = fxdr_unsigned(uint32_t, *tl++);
7146 	NFSCL_DEBUG(4, "retoncls=%d stseq=%d\n", *retonclosep,
7147 	    (int)stateidp->seqid);
7148 	stateidp->other[0] = *tl++;
7149 	stateidp->other[1] = *tl++;
7150 	stateidp->other[2] = *tl++;
7151 	cnt = fxdr_unsigned(int, *tl);
7152 	NFSCL_DEBUG(4, "layg cnt=%d\n", cnt);
7153 	if (cnt <= 0 || cnt > 10000) {
7154 		/* Don't accept more than 10000 layouts in reply. */
7155 		error = NFSERR_BADXDR;
7156 		goto nfsmout;
7157 	}
7158 	for (i = 0; i < cnt; i++) {
7159 		/* Dissect to the layout type. */
7160 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER +
7161 		    3 * NFSX_UNSIGNED);
7162 		off = fxdr_hyper(tl); tl += 2;
7163 		retlen = fxdr_hyper(tl); tl += 2;
7164 		iomode = fxdr_unsigned(int, *tl++);
7165 		laytype = fxdr_unsigned(int, *tl);
7166 		NFSCL_DEBUG(4, "layt=%d off=%ju len=%ju iom=%d\n", laytype,
7167 		    (uintmax_t)off, (uintmax_t)retlen, iomode);
7168 		/* Ignore length of layout body for now. */
7169 		if (laytype == NFSLAYOUT_NFSV4_1_FILES) {
7170 			/* Parse the File layout up to fhcnt. */
7171 			NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED +
7172 			    NFSX_HYPER + NFSX_V4DEVICEID);
7173 			fhcnt = fxdr_unsigned(int, *(tl + 4 +
7174 			    NFSX_V4DEVICEID / NFSX_UNSIGNED));
7175 			NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7176 			if (fhcnt < 0 || fhcnt > 100) {
7177 				/* Don't accept more than 100 file handles. */
7178 				error = NFSERR_BADXDR;
7179 				goto nfsmout;
7180 			}
7181 			if (fhcnt > 0)
7182 				flp = malloc(sizeof(*flp) + fhcnt *
7183 				    sizeof(struct nfsfh *), M_NFSFLAYOUT,
7184 				    M_WAITOK);
7185 			else
7186 				flp = malloc(sizeof(*flp), M_NFSFLAYOUT,
7187 				    M_WAITOK);
7188 			flp->nfsfl_flags = NFSFL_FILE;
7189 			flp->nfsfl_fhcnt = 0;
7190 			flp->nfsfl_devp = NULL;
7191 			flp->nfsfl_off = off;
7192 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7193 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7194 			else
7195 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7196 			flp->nfsfl_iomode = iomode;
7197 			if (gotiomode == -1)
7198 				gotiomode = flp->nfsfl_iomode;
7199 			/* Ignore layout body length for now. */
7200 			NFSBCOPY(tl, flp->nfsfl_dev, NFSX_V4DEVICEID);
7201 			tl += (NFSX_V4DEVICEID / NFSX_UNSIGNED);
7202 			flp->nfsfl_util = fxdr_unsigned(uint32_t, *tl++);
7203 			NFSCL_DEBUG(4, "flutil=0x%x\n", flp->nfsfl_util);
7204 			mtx_lock(&nmp->nm_mtx);
7205 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_util &
7206 			    NFSFLAYUTIL_IOADVISE_THRU_MDS) != 0)
7207 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7208 			mtx_unlock(&nmp->nm_mtx);
7209 			flp->nfsfl_stripe1 = fxdr_unsigned(uint32_t, *tl++);
7210 			flp->nfsfl_patoff = fxdr_hyper(tl); tl += 2;
7211 			NFSCL_DEBUG(4, "stripe1=%u poff=%ju\n",
7212 			    flp->nfsfl_stripe1, (uintmax_t)flp->nfsfl_patoff);
7213 			for (j = 0; j < fhcnt; j++) {
7214 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7215 				nfhlen = fxdr_unsigned(int, *tl);
7216 				if (nfhlen <= 0 || nfhlen > NFSX_V4FHMAX) {
7217 					error = NFSERR_BADXDR;
7218 					goto nfsmout;
7219 				}
7220 				nfhp = malloc(sizeof(*nfhp) + nfhlen - 1,
7221 				    M_NFSFH, M_WAITOK);
7222 				flp->nfsfl_fh[j] = nfhp;
7223 				flp->nfsfl_fhcnt++;
7224 				nfhp->nfh_len = nfhlen;
7225 				NFSM_DISSECT(cp, uint8_t *, NFSM_RNDUP(nfhlen));
7226 				NFSBCOPY(cp, nfhp->nfh_fh, nfhlen);
7227 			}
7228 		} else if (laytype == NFSLAYOUT_FLEXFILE) {
7229 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED +
7230 			    NFSX_HYPER);
7231 			mirrorcnt = fxdr_unsigned(int, *(tl + 2));
7232 			NFSCL_DEBUG(4, "mirrorcnt=%d\n", mirrorcnt);
7233 			if (mirrorcnt < 1 || mirrorcnt > NFSDEV_MAXMIRRORS) {
7234 				error = NFSERR_BADXDR;
7235 				goto nfsmout;
7236 			}
7237 			flp = malloc(sizeof(*flp) + mirrorcnt *
7238 			    sizeof(struct nfsffm), M_NFSFLAYOUT, M_WAITOK);
7239 			flp->nfsfl_flags = NFSFL_FLEXFILE;
7240 			flp->nfsfl_mirrorcnt = mirrorcnt;
7241 			for (j = 0; j < mirrorcnt; j++)
7242 				flp->nfsfl_ffm[j].devp = NULL;
7243 			flp->nfsfl_off = off;
7244 			if (flp->nfsfl_off + retlen < flp->nfsfl_off)
7245 				flp->nfsfl_end = UINT64_MAX - flp->nfsfl_off;
7246 			else
7247 				flp->nfsfl_end = flp->nfsfl_off + retlen;
7248 			flp->nfsfl_iomode = iomode;
7249 			if (gotiomode == -1)
7250 				gotiomode = flp->nfsfl_iomode;
7251 			flp->nfsfl_stripeunit = fxdr_hyper(tl);
7252 			NFSCL_DEBUG(4, "stripeunit=%ju\n",
7253 			    (uintmax_t)flp->nfsfl_stripeunit);
7254 			for (j = 0; j < mirrorcnt; j++) {
7255 				NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7256 				k = fxdr_unsigned(int, *tl);
7257 				if (k < 1 || k > 128) {
7258 					error = NFSERR_BADXDR;
7259 					goto nfsmout;
7260 				}
7261 				NFSCL_DEBUG(4, "servercnt=%d\n", k);
7262 				for (l = 0; l < k; l++) {
7263 					NFSM_DISSECT(tl, uint32_t *,
7264 					    NFSX_V4DEVICEID + NFSX_STATEID +
7265 					    2 * NFSX_UNSIGNED);
7266 					if (l == 0) {
7267 						/* Just use the first server. */
7268 						NFSBCOPY(tl,
7269 						    flp->nfsfl_ffm[j].dev,
7270 						    NFSX_V4DEVICEID);
7271 						tl += (NFSX_V4DEVICEID /
7272 						    NFSX_UNSIGNED);
7273 						tl++;
7274 						flp->nfsfl_ffm[j].st.seqid =
7275 						    *tl++;
7276 						flp->nfsfl_ffm[j].st.other[0] =
7277 						    *tl++;
7278 						flp->nfsfl_ffm[j].st.other[1] =
7279 						    *tl++;
7280 						flp->nfsfl_ffm[j].st.other[2] =
7281 						    *tl++;
7282 						NFSCL_DEBUG(4, "st.seqid=%u "
7283 						 "st.o0=0x%x st.o1=0x%x "
7284 						 "st.o2=0x%x\n",
7285 						 flp->nfsfl_ffm[j].st.seqid,
7286 						 flp->nfsfl_ffm[j].st.other[0],
7287 						 flp->nfsfl_ffm[j].st.other[1],
7288 						 flp->nfsfl_ffm[j].st.other[2]);
7289 					} else
7290 						tl += ((NFSX_V4DEVICEID +
7291 						    NFSX_STATEID +
7292 						    NFSX_UNSIGNED) /
7293 						    NFSX_UNSIGNED);
7294 					fhcnt = fxdr_unsigned(int, *tl);
7295 					NFSCL_DEBUG(4, "fhcnt=%d\n", fhcnt);
7296 					if (fhcnt < 1 ||
7297 					    fhcnt > NFSDEV_MAXVERS) {
7298 						error = NFSERR_BADXDR;
7299 						goto nfsmout;
7300 					}
7301 					for (m = 0; m < fhcnt; m++) {
7302 						NFSM_DISSECT(tl, uint32_t *,
7303 						    NFSX_UNSIGNED);
7304 						nfhlen = fxdr_unsigned(int,
7305 						    *tl);
7306 						NFSCL_DEBUG(4, "nfhlen=%d\n",
7307 						    nfhlen);
7308 						if (nfhlen <= 0 || nfhlen >
7309 						    NFSX_V4FHMAX) {
7310 							error = NFSERR_BADXDR;
7311 							goto nfsmout;
7312 						}
7313 						NFSM_DISSECT(cp, uint8_t *,
7314 						    NFSM_RNDUP(nfhlen));
7315 						if (l == 0) {
7316 							flp->nfsfl_ffm[j].fhcnt
7317 							    = fhcnt;
7318 							nfhp = malloc(
7319 							    sizeof(*nfhp) +
7320 							    nfhlen - 1, M_NFSFH,
7321 							    M_WAITOK);
7322 							flp->nfsfl_ffm[j].fh[m]
7323 							    = nfhp;
7324 							nfhp->nfh_len = nfhlen;
7325 							NFSBCOPY(cp,
7326 							    nfhp->nfh_fh,
7327 							    nfhlen);
7328 							NFSCL_DEBUG(4,
7329 							    "got fh\n");
7330 						}
7331 					}
7332 					/* Now, get the ffsd_user/ffds_group. */
7333 					error = nfsrv_parseug(nd, 0, &user,
7334 					    &grp, curthread);
7335 					NFSCL_DEBUG(4, "after parseu=%d\n",
7336 					    error);
7337 					if (error == 0)
7338 						error = nfsrv_parseug(nd, 1,
7339 						    &user, &grp, curthread);
7340 					NFSCL_DEBUG(4, "aft parseg=%d\n",
7341 					    grp);
7342 					if (error != 0)
7343 						goto nfsmout;
7344 					NFSCL_DEBUG(4, "user=%d group=%d\n",
7345 					    user, grp);
7346 					if (l == 0) {
7347 						flp->nfsfl_ffm[j].user = user;
7348 						flp->nfsfl_ffm[j].group = grp;
7349 						NFSCL_DEBUG(4,
7350 						    "usr=%d grp=%d\n", user,
7351 						    grp);
7352 					}
7353 				}
7354 			}
7355 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7356 			flp->nfsfl_fflags = fxdr_unsigned(uint32_t, *tl++);
7357 #ifdef notnow
7358 			/*
7359 			 * At this time, there is no flag.
7360 			 * NFSFLEXFLAG_IOADVISE_THRU_MDS might need to be
7361 			 * added, or it may never exist?
7362 			 */
7363 			mtx_lock(&nmp->nm_mtx);
7364 			if (nmp->nm_minorvers > 1 && (flp->nfsfl_fflags &
7365 			    NFSFLEXFLAG_IOADVISE_THRU_MDS) != 0)
7366 				nmp->nm_privflag |= NFSMNTP_IOADVISETHRUMDS;
7367 			mtx_unlock(&nmp->nm_mtx);
7368 #endif
7369 			flp->nfsfl_statshint = fxdr_unsigned(uint32_t, *tl);
7370 			NFSCL_DEBUG(4, "fflags=0x%x statshint=%d\n",
7371 			    flp->nfsfl_fflags, flp->nfsfl_statshint);
7372 		} else {
7373 			error = NFSERR_BADXDR;
7374 			goto nfsmout;
7375 		}
7376 		if (flp->nfsfl_iomode == gotiomode) {
7377 			/* Keep the list in increasing offset order. */
7378 			tflp = LIST_FIRST(flhp);
7379 			prevflp = NULL;
7380 			while (tflp != NULL &&
7381 			    tflp->nfsfl_off < flp->nfsfl_off) {
7382 				prevflp = tflp;
7383 				tflp = LIST_NEXT(tflp, nfsfl_list);
7384 			}
7385 			if (prevflp == NULL)
7386 				LIST_INSERT_HEAD(flhp, flp, nfsfl_list);
7387 			else
7388 				LIST_INSERT_AFTER(prevflp, flp,
7389 				    nfsfl_list);
7390 			NFSCL_DEBUG(4, "flp inserted\n");
7391 		} else {
7392 			printf("nfscl_layoutget(): got wrong iomode\n");
7393 			nfscl_freeflayout(flp);
7394 		}
7395 		flp = NULL;
7396 	}
7397 nfsmout:
7398 	NFSCL_DEBUG(4, "eo nfsrv_parselayoutget=%d\n", error);
7399 	if (error != 0 && flp != NULL)
7400 		nfscl_freeflayout(flp);
7401 	return (error);
7402 }
7403 
7404 /*
7405  * Parse a user/group digit string.
7406  */
7407 static int
7408 nfsrv_parseug(struct nfsrv_descript *nd, int dogrp, uid_t *uidp, gid_t *gidp,
7409     NFSPROC_T *p)
7410 {
7411 	uint32_t *tl;
7412 	char *cp, *str, str0[NFSV4_SMALLSTR + 1];
7413 	uint32_t len = 0;
7414 	int error = 0;
7415 
7416 	NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
7417 	len = fxdr_unsigned(uint32_t, *tl);
7418 	str = NULL;
7419 	if (len > NFSV4_OPAQUELIMIT) {
7420 		error = NFSERR_BADXDR;
7421 		goto nfsmout;
7422 	}
7423 	NFSCL_DEBUG(4, "nfsrv_parseug: len=%d\n", len);
7424 	if (len == 0) {
7425 		if (dogrp != 0)
7426 			*gidp = GID_NOGROUP;
7427 		else
7428 			*uidp = UID_NOBODY;
7429 		return (0);
7430 	}
7431 	if (len > NFSV4_SMALLSTR)
7432 		str = malloc(len + 1, M_TEMP, M_WAITOK);
7433 	else
7434 		str = str0;
7435 	NFSM_DISSECT(cp, char *, NFSM_RNDUP(len));
7436 	NFSBCOPY(cp, str, len);
7437 	str[len] = '\0';
7438 	NFSCL_DEBUG(4, "nfsrv_parseug: str=%s\n", str);
7439 	if (dogrp != 0)
7440 		error = nfsv4_strtogid(nd, str, len, gidp);
7441 	else
7442 		error = nfsv4_strtouid(nd, str, len, uidp);
7443 nfsmout:
7444 	if (len > NFSV4_SMALLSTR)
7445 		free(str, M_TEMP);
7446 	NFSCL_DEBUG(4, "eo nfsrv_parseug=%d\n", error);
7447 	return (error);
7448 }
7449 
7450 /*
7451  * Similar to nfsrpc_getlayout(), except that it uses nfsrpc_openlayget(),
7452  * so that it does both an Open and a Layoutget.
7453  */
7454 static int
7455 nfsrpc_getopenlayout(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7456     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7457     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7458     struct ucred *cred, NFSPROC_T *p)
7459 {
7460 	struct nfscllayout *lyp;
7461 	struct nfsclflayout *flp;
7462 	struct nfsclflayouthead flh;
7463 	int error, islocked, layoutlen, recalled, retonclose, usecurstateid;
7464 	int layouttype, laystat;
7465 	nfsv4stateid_t stateid;
7466 	struct nfsclsession *tsep;
7467 
7468 	error = 0;
7469 	if (NFSHASFLEXFILE(nmp))
7470 		layouttype = NFSLAYOUT_FLEXFILE;
7471 	else
7472 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7473 	/*
7474 	 * If lyp is returned non-NULL, there will be a refcnt (shared lock)
7475 	 * on it, iff flp != NULL or a lock (exclusive lock) on it iff
7476 	 * flp == NULL.
7477 	 */
7478 	lyp = nfscl_getlayout(nmp->nm_clp, newfhp, newfhlen, 0, &flp,
7479 	    &recalled);
7480 	NFSCL_DEBUG(4, "nfsrpc_getopenlayout nfscl_getlayout lyp=%p\n", lyp);
7481 	if (lyp == NULL)
7482 		islocked = 0;
7483 	else if (flp != NULL)
7484 		islocked = 1;
7485 	else
7486 		islocked = 2;
7487 	if ((lyp == NULL || flp == NULL) && recalled == 0) {
7488 		LIST_INIT(&flh);
7489 		tsep = nfsmnt_mdssession(nmp);
7490 		layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID +
7491 		    3 * NFSX_UNSIGNED);
7492 		if (lyp == NULL)
7493 			usecurstateid = 1;
7494 		else {
7495 			usecurstateid = 0;
7496 			stateid.seqid = lyp->nfsly_stateid.seqid;
7497 			stateid.other[0] = lyp->nfsly_stateid.other[0];
7498 			stateid.other[1] = lyp->nfsly_stateid.other[1];
7499 			stateid.other[2] = lyp->nfsly_stateid.other[2];
7500 		}
7501 		error = nfsrpc_openlayoutrpc(nmp, vp, nfhp, fhlen,
7502 		    newfhp, newfhlen, mode, op, name, namelen,
7503 		    dpp, &stateid, usecurstateid, layouttype, layoutlen,
7504 		    &retonclose, &flh, &laystat, cred, p);
7505 		NFSCL_DEBUG(4, "aft nfsrpc_openlayoutrpc laystat=%d err=%d\n",
7506 		    laystat, error);
7507 		laystat = nfsrpc_layoutgetres(nmp, vp, newfhp, newfhlen,
7508 		    &stateid, retonclose, NULL, &lyp, &flh, layouttype, laystat,
7509 		    &islocked, cred, p);
7510 	} else
7511 		error = nfsrpc_openrpc(nmp, vp, nfhp, fhlen, newfhp, newfhlen,
7512 		    mode, op, name, namelen, dpp, 0, 0, cred, p, 0, 0);
7513 	if (islocked == 2)
7514 		nfscl_rellayout(lyp, 1);
7515 	else if (islocked == 1)
7516 		nfscl_rellayout(lyp, 0);
7517 	return (error);
7518 }
7519 
7520 /*
7521  * This function does an Open+LayoutGet for an NFSv4.1 mount with pNFS
7522  * enabled, only for the CLAIM_NULL case.  All other NFSv4 Opens are
7523  * handled by nfsrpc_openrpc().
7524  * For the case where op == NULL, dvp is the directory.  When op != NULL, it
7525  * can be NULL.
7526  */
7527 static int
7528 nfsrpc_openlayoutrpc(struct nfsmount *nmp, vnode_t vp, u_int8_t *nfhp,
7529     int fhlen, uint8_t *newfhp, int newfhlen, uint32_t mode,
7530     struct nfsclopen *op, uint8_t *name, int namelen, struct nfscldeleg **dpp,
7531     nfsv4stateid_t *stateidp, int usecurstateid, int layouttype,
7532     int layoutlen, int *retonclosep, struct nfsclflayouthead *flhp,
7533     int *laystatp, struct ucred *cred, NFSPROC_T *p)
7534 {
7535 	uint32_t *tl;
7536 	struct nfsrv_descript nfsd, *nd = &nfsd;
7537 	struct nfscldeleg *ndp = NULL;
7538 	struct nfsvattr nfsva;
7539 	struct nfsclsession *tsep;
7540 	uint32_t rflags, deleg;
7541 	nfsattrbit_t attrbits;
7542 	int error, ret, acesize, limitby, iomode;
7543 
7544 	*dpp = NULL;
7545 	*laystatp = ENXIO;
7546 	nfscl_reqstart(nd, NFSPROC_OPENLAYGET, nmp, nfhp, fhlen, NULL, NULL,
7547 	    0, 0);
7548 	NFSM_BUILD(tl, uint32_t *, 5 * NFSX_UNSIGNED);
7549 	*tl++ = txdr_unsigned(op->nfso_own->nfsow_seqid);
7550 	*tl++ = txdr_unsigned(mode & NFSV4OPEN_ACCESSBOTH);
7551 	*tl++ = txdr_unsigned((mode >> NFSLCK_SHIFT) & NFSV4OPEN_DENYBOTH);
7552 	tsep = nfsmnt_mdssession(nmp);
7553 	*tl++ = tsep->nfsess_clientid.lval[0];
7554 	*tl = tsep->nfsess_clientid.lval[1];
7555 	nfsm_strtom(nd, op->nfso_own->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7556 	NFSM_BUILD(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7557 	*tl++ = txdr_unsigned(NFSV4OPEN_NOCREATE);
7558 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7559 	nfsm_strtom(nd, name, namelen);
7560 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7561 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7562 	NFSZERO_ATTRBIT(&attrbits);
7563 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_CHANGE);
7564 	NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMEMODIFY);
7565 	nfsrv_putattrbit(nd, &attrbits);
7566 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
7567 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7568 	if ((mode & NFSV4OPEN_ACCESSWRITE) != 0)
7569 		iomode = NFSLAYOUTIOMODE_RW;
7570 	else
7571 		iomode = NFSLAYOUTIOMODE_READ;
7572 	nfsrv_setuplayoutget(nd, iomode, 0, UINT64_MAX, 0, stateidp,
7573 	    layouttype, layoutlen, usecurstateid);
7574 	error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, vp, p, cred,
7575 	    NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL);
7576 	if (error != 0)
7577 		return (error);
7578 	NFSCL_INCRSEQID(op->nfso_own->nfsow_seqid, nd);
7579 	if (nd->nd_repstat != 0)
7580 		*laystatp = nd->nd_repstat;
7581 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7582 		/* ND_NOMOREDATA will be set if the Open operation failed. */
7583 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7584 		    6 * NFSX_UNSIGNED);
7585 		op->nfso_stateid.seqid = *tl++;
7586 		op->nfso_stateid.other[0] = *tl++;
7587 		op->nfso_stateid.other[1] = *tl++;
7588 		op->nfso_stateid.other[2] = *tl;
7589 		rflags = fxdr_unsigned(u_int32_t, *(tl + 6));
7590 		error = nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7591 		if (error != 0)
7592 			goto nfsmout;
7593 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7594 		deleg = fxdr_unsigned(u_int32_t, *tl);
7595 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7596 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7597 			if (!(op->nfso_own->nfsow_clp->nfsc_flags &
7598 			      NFSCLFLAGS_FIRSTDELEG))
7599 				op->nfso_own->nfsow_clp->nfsc_flags |=
7600 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7601 			ndp = malloc(sizeof(struct nfscldeleg) + newfhlen,
7602 			    M_NFSCLDELEG, M_WAITOK);
7603 			LIST_INIT(&ndp->nfsdl_owner);
7604 			LIST_INIT(&ndp->nfsdl_lock);
7605 			ndp->nfsdl_clp = op->nfso_own->nfsow_clp;
7606 			ndp->nfsdl_fhlen = newfhlen;
7607 			NFSBCOPY(newfhp, ndp->nfsdl_fh, newfhlen);
7608 			newnfs_copyincred(cred, &ndp->nfsdl_cred);
7609 			nfscl_lockinit(&ndp->nfsdl_rwlock);
7610 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7611 			    NFSX_UNSIGNED);
7612 			ndp->nfsdl_stateid.seqid = *tl++;
7613 			ndp->nfsdl_stateid.other[0] = *tl++;
7614 			ndp->nfsdl_stateid.other[1] = *tl++;
7615 			ndp->nfsdl_stateid.other[2] = *tl++;
7616 			ret = fxdr_unsigned(int, *tl);
7617 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7618 				ndp->nfsdl_flags = NFSCLDL_WRITE;
7619 				/*
7620 				 * Indicates how much the file can grow.
7621 				 */
7622 				NFSM_DISSECT(tl, u_int32_t *,
7623 				    3 * NFSX_UNSIGNED);
7624 				limitby = fxdr_unsigned(int, *tl++);
7625 				switch (limitby) {
7626 				case NFSV4OPEN_LIMITSIZE:
7627 					ndp->nfsdl_sizelimit = fxdr_hyper(tl);
7628 					break;
7629 				case NFSV4OPEN_LIMITBLOCKS:
7630 					ndp->nfsdl_sizelimit =
7631 					    fxdr_unsigned(u_int64_t, *tl++);
7632 					ndp->nfsdl_sizelimit *=
7633 					    fxdr_unsigned(u_int64_t, *tl);
7634 					break;
7635 				default:
7636 					error = NFSERR_BADXDR;
7637 					goto nfsmout;
7638 				};
7639 			} else
7640 				ndp->nfsdl_flags = NFSCLDL_READ;
7641 			if (ret != 0)
7642 				ndp->nfsdl_flags |= NFSCLDL_RECALL;
7643 			error = nfsrv_dissectace(nd, &ndp->nfsdl_ace, &ret,
7644 			    &acesize, p);
7645 			if (error != 0)
7646 				goto nfsmout;
7647 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7648 			error = NFSERR_BADXDR;
7649 			goto nfsmout;
7650 		}
7651 		if ((rflags & NFSV4OPEN_LOCKTYPEPOSIX) != 0 ||
7652 		    nfscl_assumeposixlocks)
7653 			op->nfso_posixlock = 1;
7654 		else
7655 			op->nfso_posixlock = 0;
7656 		NFSM_DISSECT(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7657 		/* If the 2nd element == NFS_OK, the Getattr succeeded. */
7658 		if (*++tl == 0) {
7659 			error = nfsv4_loadattr(nd, NULL, &nfsva, NULL,
7660 			    NULL, 0, NULL, NULL, NULL, NULL, NULL, 0,
7661 			    NULL, NULL, NULL, p, cred);
7662 			if (error != 0)
7663 				goto nfsmout;
7664 			if (ndp != NULL) {
7665 				ndp->nfsdl_change = nfsva.na_filerev;
7666 				ndp->nfsdl_modtime = nfsva.na_mtime;
7667 				ndp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7668 				*dpp = ndp;
7669 				ndp = NULL;
7670 			}
7671 			/*
7672 			 * At this point, the Open has succeeded, so set
7673 			 * nd_repstat = NFS_OK.  If the Layoutget failed,
7674 			 * this function just won't return a layout.
7675 			 */
7676 			if (nd->nd_repstat == 0) {
7677 				NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7678 				*laystatp = fxdr_unsigned(int, *++tl);
7679 				if (*laystatp == 0) {
7680 					error = nfsrv_parselayoutget(nmp, nd,
7681 					    stateidp, retonclosep, flhp);
7682 					if (error != 0)
7683 						*laystatp = error;
7684 				}
7685 			} else
7686 				nd->nd_repstat = 0;	/* Return 0 for Open. */
7687 		}
7688 	}
7689 	if (nd->nd_repstat != 0 && error == 0)
7690 		error = nd->nd_repstat;
7691 nfsmout:
7692 	free(ndp, M_NFSCLDELEG);
7693 	m_freem(nd->nd_mrep);
7694 	return (error);
7695 }
7696 
7697 /*
7698  * Similar nfsrpc_createv4(), but also does the LayoutGet operation.
7699  * Used only for mounts with pNFS enabled.
7700  */
7701 static int
7702 nfsrpc_createlayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7703     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7704     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7705     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7706     int *dattrflagp, void *dstuff, int *unlockedp, nfsv4stateid_t *stateidp,
7707     int usecurstateid, int layouttype, int layoutlen, int *retonclosep,
7708     struct nfsclflayouthead *flhp, int *laystatp)
7709 {
7710 	uint32_t *tl;
7711 	int error = 0, deleg, newone, ret, acesize, limitby;
7712 	struct nfsrv_descript nfsd, *nd = &nfsd;
7713 	struct nfsclopen *op;
7714 	struct nfscldeleg *dp = NULL;
7715 	struct nfsnode *np;
7716 	struct nfsfh *nfhp;
7717 	struct nfsclsession *tsep;
7718 	nfsattrbit_t attrbits;
7719 	nfsv4stateid_t stateid;
7720 	struct nfsmount *nmp;
7721 
7722 	nmp = VFSTONFS(dvp->v_mount);
7723 	np = VTONFS(dvp);
7724 	*laystatp = ENXIO;
7725 	*unlockedp = 0;
7726 	*nfhpp = NULL;
7727 	*dpp = NULL;
7728 	*attrflagp = 0;
7729 	*dattrflagp = 0;
7730 	if (namelen > NFS_MAXNAMLEN)
7731 		return (ENAMETOOLONG);
7732 	NFSCL_REQSTART(nd, NFSPROC_CREATELAYGET, dvp);
7733 	/*
7734 	 * For V4, this is actually an Open op.
7735 	 */
7736 	NFSM_BUILD(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
7737 	*tl++ = txdr_unsigned(owp->nfsow_seqid);
7738 	*tl++ = txdr_unsigned(NFSV4OPEN_ACCESSWRITE |
7739 	    NFSV4OPEN_ACCESSREAD);
7740 	*tl++ = txdr_unsigned(NFSV4OPEN_DENYNONE);
7741 	tsep = nfsmnt_mdssession(nmp);
7742 	*tl++ = tsep->nfsess_clientid.lval[0];
7743 	*tl = tsep->nfsess_clientid.lval[1];
7744 	nfsm_strtom(nd, owp->nfsow_owner, NFSV4CL_LOCKNAMELEN);
7745 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7746 	*tl++ = txdr_unsigned(NFSV4OPEN_CREATE);
7747 	if ((fmode & O_EXCL) != 0) {
7748 		if (NFSHASSESSPERSIST(nmp)) {
7749 			/* Use GUARDED for persistent sessions. */
7750 			*tl = txdr_unsigned(NFSCREATE_GUARDED);
7751 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7752 		} else {
7753 			/* Otherwise, use EXCLUSIVE4_1. */
7754 			*tl = txdr_unsigned(NFSCREATE_EXCLUSIVE41);
7755 			NFSM_BUILD(tl, u_int32_t *, NFSX_VERF);
7756 			*tl++ = cverf.lval[0];
7757 			*tl = cverf.lval[1];
7758 			nfscl_fillsattr(nd, vap, dvp, 0, 0);
7759 		}
7760 	} else {
7761 		*tl = txdr_unsigned(NFSCREATE_UNCHECKED);
7762 		nfscl_fillsattr(nd, vap, dvp, 0, 0);
7763 	}
7764 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7765 	*tl = txdr_unsigned(NFSV4OPEN_CLAIMNULL);
7766 	nfsm_strtom(nd, name, namelen);
7767 	/* Get the new file's handle and attributes, plus save the FH. */
7768 	NFSM_BUILD(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
7769 	*tl++ = txdr_unsigned(NFSV4OP_SAVEFH);
7770 	*tl++ = txdr_unsigned(NFSV4OP_GETFH);
7771 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7772 	NFSGETATTR_ATTRBIT(&attrbits);
7773 	nfsrv_putattrbit(nd, &attrbits);
7774 	/* Get the directory's post-op attributes. */
7775 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7776 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
7777 	nfsm_fhtom(nd, np->n_fhp->nfh_fh, np->n_fhp->nfh_len, 0);
7778 	NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED);
7779 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
7780 	nfsrv_putattrbit(nd, &attrbits);
7781 	NFSM_BUILD(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
7782 	*tl++ = txdr_unsigned(NFSV4OP_RESTOREFH);
7783 	*tl = txdr_unsigned(NFSV4OP_LAYOUTGET);
7784 	nfsrv_setuplayoutget(nd, NFSLAYOUTIOMODE_RW, 0, UINT64_MAX, 0, stateidp,
7785 	    layouttype, layoutlen, usecurstateid);
7786 	error = nfscl_request(nd, dvp, p, cred, dstuff);
7787 	if (error != 0)
7788 		return (error);
7789 	NFSCL_DEBUG(4, "nfsrpc_createlayout stat=%d err=%d\n", nd->nd_repstat,
7790 	    error);
7791 	if (nd->nd_repstat != 0)
7792 		*laystatp = nd->nd_repstat;
7793 	NFSCL_INCRSEQID(owp->nfsow_seqid, nd);
7794 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7795 		NFSCL_DEBUG(4, "nfsrpc_createlayout open succeeded\n");
7796 		NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7797 		    6 * NFSX_UNSIGNED);
7798 		stateid.seqid = *tl++;
7799 		stateid.other[0] = *tl++;
7800 		stateid.other[1] = *tl++;
7801 		stateid.other[2] = *tl;
7802 		nfsrv_getattrbits(nd, &attrbits, NULL, NULL);
7803 		NFSM_DISSECT(tl, u_int32_t *, NFSX_UNSIGNED);
7804 		deleg = fxdr_unsigned(int, *tl);
7805 		if (deleg == NFSV4OPEN_DELEGATEREAD ||
7806 		    deleg == NFSV4OPEN_DELEGATEWRITE) {
7807 			if (!(owp->nfsow_clp->nfsc_flags &
7808 			      NFSCLFLAGS_FIRSTDELEG))
7809 				owp->nfsow_clp->nfsc_flags |=
7810 				  (NFSCLFLAGS_FIRSTDELEG | NFSCLFLAGS_GOTDELEG);
7811 			dp = malloc(sizeof(struct nfscldeleg) + NFSX_V4FHMAX,
7812 			    M_NFSCLDELEG, M_WAITOK);
7813 			LIST_INIT(&dp->nfsdl_owner);
7814 			LIST_INIT(&dp->nfsdl_lock);
7815 			dp->nfsdl_clp = owp->nfsow_clp;
7816 			newnfs_copyincred(cred, &dp->nfsdl_cred);
7817 			nfscl_lockinit(&dp->nfsdl_rwlock);
7818 			NFSM_DISSECT(tl, u_int32_t *, NFSX_STATEID +
7819 			    NFSX_UNSIGNED);
7820 			dp->nfsdl_stateid.seqid = *tl++;
7821 			dp->nfsdl_stateid.other[0] = *tl++;
7822 			dp->nfsdl_stateid.other[1] = *tl++;
7823 			dp->nfsdl_stateid.other[2] = *tl++;
7824 			ret = fxdr_unsigned(int, *tl);
7825 			if (deleg == NFSV4OPEN_DELEGATEWRITE) {
7826 				dp->nfsdl_flags = NFSCLDL_WRITE;
7827 				/*
7828 				 * Indicates how much the file can grow.
7829 				 */
7830 				NFSM_DISSECT(tl, u_int32_t *,
7831 				    3 * NFSX_UNSIGNED);
7832 				limitby = fxdr_unsigned(int, *tl++);
7833 				switch (limitby) {
7834 				case NFSV4OPEN_LIMITSIZE:
7835 					dp->nfsdl_sizelimit = fxdr_hyper(tl);
7836 					break;
7837 				case NFSV4OPEN_LIMITBLOCKS:
7838 					dp->nfsdl_sizelimit =
7839 					    fxdr_unsigned(u_int64_t, *tl++);
7840 					dp->nfsdl_sizelimit *=
7841 					    fxdr_unsigned(u_int64_t, *tl);
7842 					break;
7843 				default:
7844 					error = NFSERR_BADXDR;
7845 					goto nfsmout;
7846 				};
7847 			} else {
7848 				dp->nfsdl_flags = NFSCLDL_READ;
7849 			}
7850 			if (ret != 0)
7851 				dp->nfsdl_flags |= NFSCLDL_RECALL;
7852 			error = nfsrv_dissectace(nd, &dp->nfsdl_ace, &ret,
7853 			    &acesize, p);
7854 			if (error != 0)
7855 				goto nfsmout;
7856 		} else if (deleg != NFSV4OPEN_DELEGATENONE) {
7857 			error = NFSERR_BADXDR;
7858 			goto nfsmout;
7859 		}
7860 
7861 		/* Now, we should have the status for the SaveFH. */
7862 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7863 		if (*++tl == 0) {
7864 			NFSCL_DEBUG(4, "nfsrpc_createlayout SaveFH ok\n");
7865 			/*
7866 			 * Now, process the GetFH and Getattr for the newly
7867 			 * created file. nfscl_mtofh() will set
7868 			 * ND_NOMOREDATA if these weren't successful.
7869 			 */
7870 			error = nfscl_mtofh(nd, nfhpp, nnap, attrflagp);
7871 			NFSCL_DEBUG(4, "aft nfscl_mtofh err=%d\n", error);
7872 			if (error != 0)
7873 				goto nfsmout;
7874 		} else
7875 			nd->nd_flag |= ND_NOMOREDATA;
7876 		/* Now we have the PutFH and Getattr for the directory. */
7877 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7878 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
7879 			if (*++tl != 0)
7880 				nd->nd_flag |= ND_NOMOREDATA;
7881 			else {
7882 				NFSM_DISSECT(tl, uint32_t *, 2 *
7883 				    NFSX_UNSIGNED);
7884 				if (*++tl != 0)
7885 					nd->nd_flag |= ND_NOMOREDATA;
7886 			}
7887 		}
7888 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
7889 			/* Load the directory attributes. */
7890 			error = nfsm_loadattr(nd, dnap);
7891 			NFSCL_DEBUG(4, "aft nfsm_loadattr err=%d\n", error);
7892 			if (error != 0)
7893 				goto nfsmout;
7894 			*dattrflagp = 1;
7895 			if (dp != NULL && *attrflagp != 0) {
7896 				dp->nfsdl_change = nnap->na_filerev;
7897 				dp->nfsdl_modtime = nnap->na_mtime;
7898 				dp->nfsdl_flags |= NFSCLDL_MODTIMESET;
7899 			}
7900 			/*
7901 			 * We can now complete the Open state.
7902 			 */
7903 			nfhp = *nfhpp;
7904 			if (dp != NULL) {
7905 				dp->nfsdl_fhlen = nfhp->nfh_len;
7906 				NFSBCOPY(nfhp->nfh_fh, dp->nfsdl_fh,
7907 				    nfhp->nfh_len);
7908 			}
7909 			/*
7910 			 * Get an Open structure that will be
7911 			 * attached to the OpenOwner, acquired already.
7912 			 */
7913 			error = nfscl_open(dvp, nfhp->nfh_fh, nfhp->nfh_len,
7914 			    (NFSV4OPEN_ACCESSWRITE | NFSV4OPEN_ACCESSREAD), 0,
7915 			    cred, p, NULL, &op, &newone, NULL, 0);
7916 			if (error != 0)
7917 				goto nfsmout;
7918 			op->nfso_stateid = stateid;
7919 			newnfs_copyincred(cred, &op->nfso_cred);
7920 
7921 			nfscl_openrelease(nmp, op, error, newone);
7922 			*unlockedp = 1;
7923 
7924 			/* Now, handle the RestoreFH and LayoutGet. */
7925 			if (nd->nd_repstat == 0) {
7926 				NFSM_DISSECT(tl, uint32_t *, 4 * NFSX_UNSIGNED);
7927 				*laystatp = fxdr_unsigned(int, *(tl + 3));
7928 				if (*laystatp == 0) {
7929 					error = nfsrv_parselayoutget(nmp, nd,
7930 					    stateidp, retonclosep, flhp);
7931 					if (error != 0)
7932 						*laystatp = error;
7933 				}
7934 				NFSCL_DEBUG(4, "aft nfsrv_parselayout err=%d\n",
7935 				    error);
7936 			} else
7937 				nd->nd_repstat = 0;
7938 		}
7939 	}
7940 	if (nd->nd_repstat != 0 && error == 0)
7941 		error = nd->nd_repstat;
7942 	if (error == NFSERR_STALECLIENTID || error == NFSERR_BADSESSION)
7943 		nfscl_initiate_recovery(owp->nfsow_clp);
7944 nfsmout:
7945 	NFSCL_DEBUG(4, "eo nfsrpc_createlayout err=%d\n", error);
7946 	if (error == 0)
7947 		*dpp = dp;
7948 	else
7949 		free(dp, M_NFSCLDELEG);
7950 	m_freem(nd->nd_mrep);
7951 	return (error);
7952 }
7953 
7954 /*
7955  * Similar to nfsrpc_getopenlayout(), except that it used for the Create case.
7956  */
7957 static int
7958 nfsrpc_getcreatelayout(vnode_t dvp, char *name, int namelen, struct vattr *vap,
7959     nfsquad_t cverf, int fmode, struct nfsclowner *owp, struct nfscldeleg **dpp,
7960     struct ucred *cred, NFSPROC_T *p, struct nfsvattr *dnap,
7961     struct nfsvattr *nnap, struct nfsfh **nfhpp, int *attrflagp,
7962     int *dattrflagp, void *dstuff, int *unlockedp)
7963 {
7964 	struct nfscllayout *lyp;
7965 	struct nfsclflayouthead flh;
7966 	struct nfsfh *nfhp;
7967 	struct nfsclsession *tsep;
7968 	struct nfsmount *nmp;
7969 	nfsv4stateid_t stateid;
7970 	int error, layoutlen, layouttype, retonclose, laystat;
7971 
7972 	error = 0;
7973 	nmp = VFSTONFS(dvp->v_mount);
7974 	if (NFSHASFLEXFILE(nmp))
7975 		layouttype = NFSLAYOUT_FLEXFILE;
7976 	else
7977 		layouttype = NFSLAYOUT_NFSV4_1_FILES;
7978 	LIST_INIT(&flh);
7979 	tsep = nfsmnt_mdssession(nmp);
7980 	layoutlen = tsep->nfsess_maxcache - (NFSX_STATEID + 3 * NFSX_UNSIGNED);
7981 	error = nfsrpc_createlayout(dvp, name, namelen, vap, cverf, fmode,
7982 	    owp, dpp, cred, p, dnap, nnap, nfhpp, attrflagp, dattrflagp,
7983 	    dstuff, unlockedp, &stateid, 1, layouttype, layoutlen, &retonclose,
7984 	    &flh, &laystat);
7985 	NFSCL_DEBUG(4, "aft nfsrpc_createlayoutrpc laystat=%d err=%d\n",
7986 	    laystat, error);
7987 	lyp = NULL;
7988 	if (laystat == 0) {
7989 		nfhp = *nfhpp;
7990 		laystat = nfsrpc_layoutgetres(nmp, dvp, nfhp->nfh_fh,
7991 		    nfhp->nfh_len, &stateid, retonclose, NULL, &lyp, &flh,
7992 		    layouttype, laystat, NULL, cred, p);
7993 	} else
7994 		laystat = nfsrpc_layoutgetres(nmp, dvp, NULL, 0, &stateid,
7995 		    retonclose, NULL, &lyp, &flh, layouttype, laystat, NULL,
7996 		    cred, p);
7997 	if (laystat == 0)
7998 		nfscl_rellayout(lyp, 0);
7999 	return (error);
8000 }
8001 
8002 /*
8003  * Process the results of a layoutget() operation.
8004  */
8005 static int
8006 nfsrpc_layoutgetres(struct nfsmount *nmp, vnode_t vp, uint8_t *newfhp,
8007     int newfhlen, nfsv4stateid_t *stateidp, int retonclose, uint32_t *notifybit,
8008     struct nfscllayout **lypp, struct nfsclflayouthead *flhp, int layouttype,
8009     int laystat, int *islockedp, struct ucred *cred, NFSPROC_T *p)
8010 {
8011 	struct nfsclflayout *tflp;
8012 	struct nfscldevinfo *dip;
8013 	uint8_t *dev;
8014 	int i, mirrorcnt;
8015 
8016 	if (laystat == NFSERR_UNKNLAYOUTTYPE) {
8017 		NFSLOCKMNT(nmp);
8018 		if (!NFSHASFLEXFILE(nmp)) {
8019 			/* Switch to using Flex File Layout. */
8020 			nmp->nm_state |= NFSSTA_FLEXFILE;
8021 		} else if (layouttype == NFSLAYOUT_FLEXFILE) {
8022 			/* Disable pNFS. */
8023 			NFSCL_DEBUG(1, "disable PNFS\n");
8024 			nmp->nm_state &= ~(NFSSTA_PNFS | NFSSTA_FLEXFILE);
8025 		}
8026 		NFSUNLOCKMNT(nmp);
8027 	}
8028 	if (laystat == 0) {
8029 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres at FOREACH\n");
8030 		LIST_FOREACH(tflp, flhp, nfsfl_list) {
8031 			if (layouttype == NFSLAYOUT_FLEXFILE)
8032 				mirrorcnt = tflp->nfsfl_mirrorcnt;
8033 			else
8034 				mirrorcnt = 1;
8035 			for (i = 0; i < mirrorcnt; i++) {
8036 				laystat = nfscl_adddevinfo(nmp, NULL, i, tflp);
8037 				NFSCL_DEBUG(4, "aft adddev=%d\n", laystat);
8038 				if (laystat != 0) {
8039 					if (layouttype == NFSLAYOUT_FLEXFILE)
8040 						dev = tflp->nfsfl_ffm[i].dev;
8041 					else
8042 						dev = tflp->nfsfl_dev;
8043 					laystat = nfsrpc_getdeviceinfo(nmp, dev,
8044 					    layouttype, notifybit, &dip, cred,
8045 					    p);
8046 					NFSCL_DEBUG(4, "aft nfsrpc_gdi=%d\n",
8047 					    laystat);
8048 					if (laystat != 0)
8049 						goto out;
8050 					laystat = nfscl_adddevinfo(nmp, dip, i,
8051 					    tflp);
8052 					if (laystat != 0)
8053 						printf("nfsrpc_layoutgetresout"
8054 						    ": cannot add\n");
8055 				}
8056 			}
8057 		}
8058 	}
8059 out:
8060 	if (laystat == 0) {
8061 		/*
8062 		 * nfscl_layout() always returns with the nfsly_lock
8063 		 * set to a refcnt (shared lock).
8064 		 * Passing in dvp is sufficient, since it is only used to
8065 		 * get the fsid for the file system.
8066 		 */
8067 		laystat = nfscl_layout(nmp, vp, newfhp, newfhlen, stateidp,
8068 		    layouttype, retonclose, flhp, lypp, cred, p);
8069 		NFSCL_DEBUG(4, "nfsrpc_layoutgetres: aft nfscl_layout=%d\n",
8070 		    laystat);
8071 		if (laystat == 0 && islockedp != NULL)
8072 			*islockedp = 1;
8073 	}
8074 	return (laystat);
8075 }
8076 
8077 /*
8078  * nfs copy_file_range operation.
8079  */
8080 int
8081 nfsrpc_copy_file_range(vnode_t invp, off_t *inoffp, vnode_t outvp,
8082     off_t *outoffp, size_t *lenp, unsigned int flags, int *inattrflagp,
8083     struct nfsvattr *innap, int *outattrflagp, struct nfsvattr *outnap,
8084     struct ucred *cred, bool consecutive, bool *must_commitp)
8085 {
8086 	int commit, error, expireret = 0, retrycnt;
8087 	u_int32_t clidrev = 0;
8088 	struct nfsmount *nmp = VFSTONFS(vnode_mount(invp));
8089 	struct nfsfh *innfhp = NULL, *outnfhp = NULL;
8090 	nfsv4stateid_t instateid, outstateid;
8091 	void *inlckp, *outlckp;
8092 
8093 	if (nmp->nm_clp != NULL)
8094 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8095 	innfhp = VTONFS(invp)->n_fhp;
8096 	outnfhp = VTONFS(outvp)->n_fhp;
8097 	retrycnt = 0;
8098 	do {
8099 		/* Get both stateids. */
8100 		inlckp = NULL;
8101 		nfscl_getstateid(invp, innfhp->nfh_fh, innfhp->nfh_len,
8102 		    NFSV4OPEN_ACCESSREAD, 0, NULL, curthread, &instateid,
8103 		    &inlckp);
8104 		outlckp = NULL;
8105 		nfscl_getstateid(outvp, outnfhp->nfh_fh, outnfhp->nfh_len,
8106 		    NFSV4OPEN_ACCESSWRITE, 0, NULL, curthread, &outstateid,
8107 		    &outlckp);
8108 
8109 		error = nfsrpc_copyrpc(invp, *inoffp, outvp, *outoffp, lenp,
8110 		    &instateid, &outstateid, innap, inattrflagp, outnap,
8111 		    outattrflagp, consecutive, &commit, cred, curthread);
8112 		if (error == 0) {
8113 			if (commit != NFSWRITE_FILESYNC)
8114 				*must_commitp = true;
8115 			*inoffp += *lenp;
8116 			*outoffp += *lenp;
8117 		} else if (error == NFSERR_STALESTATEID)
8118 			nfscl_initiate_recovery(nmp->nm_clp);
8119 		if (inlckp != NULL)
8120 			nfscl_lockderef(inlckp);
8121 		if (outlckp != NULL)
8122 			nfscl_lockderef(outlckp);
8123 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8124 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8125 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8126 			(void) nfs_catnap(PZERO, error, "nfs_cfr");
8127 		} else if ((error == NFSERR_EXPIRED ||
8128 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8129 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8130 			    curthread);
8131 		}
8132 		retrycnt++;
8133 	} while (error == NFSERR_GRACE || error == NFSERR_DELAY ||
8134 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8135 	      error == NFSERR_STALEDONTRECOVER ||
8136 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8137 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8138 	     expireret == 0 && clidrev != 0 && retrycnt < 4));
8139 	if (error != 0 && (retrycnt >= 4 ||
8140 	    error == NFSERR_STALESTATEID || error == NFSERR_BADSESSION ||
8141 	      error == NFSERR_STALEDONTRECOVER))
8142 		error = EIO;
8143 	return (error);
8144 }
8145 
8146 /*
8147  * The copy RPC.
8148  */
8149 static int
8150 nfsrpc_copyrpc(vnode_t invp, off_t inoff, vnode_t outvp, off_t outoff,
8151     size_t *lenp, nfsv4stateid_t *instateidp, nfsv4stateid_t *outstateidp,
8152     struct nfsvattr *innap, int *inattrflagp, struct nfsvattr *outnap,
8153     int *outattrflagp, bool consecutive, int *commitp, struct ucred *cred,
8154     NFSPROC_T *p)
8155 {
8156 	uint32_t *tl;
8157 	int error;
8158 	struct nfsrv_descript nfsd;
8159 	struct nfsrv_descript *nd = &nfsd;
8160 	struct nfsmount *nmp;
8161 	nfsattrbit_t attrbits;
8162 	uint64_t len;
8163 
8164 	nmp = VFSTONFS(outvp->v_mount);
8165 	*inattrflagp = *outattrflagp = 0;
8166 	*commitp = NFSWRITE_UNSTABLE;
8167 	len = *lenp;
8168 	*lenp = 0;
8169 	if (len > nfs_maxcopyrange)
8170 		len = nfs_maxcopyrange;
8171 	NFSCL_REQSTART(nd, NFSPROC_COPY, invp);
8172 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8173 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8174 	NFSGETATTR_ATTRBIT(&attrbits);
8175 	nfsrv_putattrbit(nd, &attrbits);
8176 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8177 	*tl = txdr_unsigned(NFSV4OP_PUTFH);
8178 	nfsm_fhtom(nd, VTONFS(outvp)->n_fhp->nfh_fh,
8179 	    VTONFS(outvp)->n_fhp->nfh_len, 0);
8180 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8181 	*tl = txdr_unsigned(NFSV4OP_COPY);
8182 	nfsm_stateidtom(nd, instateidp, NFSSTATEID_PUTSTATEID);
8183 	nfsm_stateidtom(nd, outstateidp, NFSSTATEID_PUTSTATEID);
8184 	NFSM_BUILD(tl, uint32_t *, 3 * NFSX_HYPER + 4 * NFSX_UNSIGNED);
8185 	txdr_hyper(inoff, tl); tl += 2;
8186 	txdr_hyper(outoff, tl); tl += 2;
8187 	txdr_hyper(len, tl); tl += 2;
8188 	if (consecutive)
8189 		*tl++ = newnfs_true;
8190 	else
8191 		*tl++ = newnfs_false;
8192 	*tl++ = newnfs_true;
8193 	*tl++ = 0;
8194 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8195 	NFSWRITEGETATTR_ATTRBIT(&attrbits);
8196 	nfsrv_putattrbit(nd, &attrbits);
8197 	error = nfscl_request(nd, invp, p, cred, NULL);
8198 	if (error != 0)
8199 		return (error);
8200 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8201 		/* Get the input file's attributes. */
8202 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8203 		if (*(tl + 1) == 0) {
8204 			error = nfsm_loadattr(nd, innap);
8205 			if (error != 0)
8206 				goto nfsmout;
8207 			*inattrflagp = 1;
8208 		} else
8209 			nd->nd_flag |= ND_NOMOREDATA;
8210 	}
8211 	/* Skip over return stat for PutFH. */
8212 	if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8213 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8214 		if (*++tl != 0)
8215 			nd->nd_flag |= ND_NOMOREDATA;
8216 	}
8217 	/* Skip over return stat for Copy. */
8218 	if ((nd->nd_flag & ND_NOMOREDATA) == 0)
8219 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8220 	if (nd->nd_repstat == 0) {
8221 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8222 		if (*tl != 0) {
8223 			/* There should be no callback ids. */
8224 			error = NFSERR_BADXDR;
8225 			goto nfsmout;
8226 		}
8227 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + 3 * NFSX_UNSIGNED +
8228 		    NFSX_VERF);
8229 		len = fxdr_hyper(tl); tl += 2;
8230 		*commitp = fxdr_unsigned(int, *tl++);
8231 		NFSLOCKMNT(nmp);
8232 		if (!NFSHASWRITEVERF(nmp)) {
8233 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8234 			NFSSETWRITEVERF(nmp);
8235 	    	} else if (NFSBCMP(tl, nmp->nm_verf, NFSX_VERF)) {
8236 			NFSBCOPY(tl, nmp->nm_verf, NFSX_VERF);
8237 			nd->nd_repstat = NFSERR_STALEWRITEVERF;
8238 		}
8239 		NFSUNLOCKMNT(nmp);
8240 		tl += (NFSX_VERF / NFSX_UNSIGNED);
8241 		if (nd->nd_repstat == 0 && *++tl != newnfs_true)
8242 			/* Must be a synchronous copy. */
8243 			nd->nd_repstat = NFSERR_NOTSUPP;
8244 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8245 		error = nfsm_loadattr(nd, outnap);
8246 		if (error == 0)
8247 			*outattrflagp = NFS_LATTR_NOSHRINK;
8248 		if (nd->nd_repstat == 0)
8249 			*lenp = len;
8250 	} else if (nd->nd_repstat == NFSERR_OFFLOADNOREQS) {
8251 		/*
8252 		 * For the case where consecutive is not supported, but
8253 		 * synchronous is supported, we can try consecutive == false
8254 		 * by returning this error.  Otherwise, return NFSERR_NOTSUPP,
8255 		 * since Copy cannot be done.
8256 		 */
8257 		if ((nd->nd_flag & ND_NOMOREDATA) == 0) {
8258 			NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8259 			if (!consecutive || *++tl == newnfs_false)
8260 				nd->nd_repstat = NFSERR_NOTSUPP;
8261 		} else
8262 			nd->nd_repstat = NFSERR_BADXDR;
8263 	}
8264 	if (error == 0)
8265 		error = nd->nd_repstat;
8266 nfsmout:
8267 	m_freem(nd->nd_mrep);
8268 	return (error);
8269 }
8270 
8271 /*
8272  * Seek operation.
8273  */
8274 int
8275 nfsrpc_seek(vnode_t vp, off_t *offp, bool *eofp, int content,
8276     struct ucred *cred, struct nfsvattr *nap, int *attrflagp)
8277 {
8278 	int error, expireret = 0, retrycnt;
8279 	u_int32_t clidrev = 0;
8280 	struct nfsmount *nmp = VFSTONFS(vnode_mount(vp));
8281 	struct nfsnode *np = VTONFS(vp);
8282 	struct nfsfh *nfhp = NULL;
8283 	nfsv4stateid_t stateid;
8284 	void *lckp;
8285 
8286 	if (nmp->nm_clp != NULL)
8287 		clidrev = nmp->nm_clp->nfsc_clientidrev;
8288 	nfhp = np->n_fhp;
8289 	retrycnt = 0;
8290 	do {
8291 		lckp = NULL;
8292 		nfscl_getstateid(vp, nfhp->nfh_fh, nfhp->nfh_len,
8293 		    NFSV4OPEN_ACCESSREAD, 0, cred, curthread, &stateid, &lckp);
8294 		error = nfsrpc_seekrpc(vp, offp, &stateid, eofp, content,
8295 		    nap, attrflagp, cred);
8296 		if (error == NFSERR_STALESTATEID)
8297 			nfscl_initiate_recovery(nmp->nm_clp);
8298 		if (lckp != NULL)
8299 			nfscl_lockderef(lckp);
8300 		if (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8301 		    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8302 		    error == NFSERR_OLDSTATEID || error == NFSERR_BADSESSION) {
8303 			(void) nfs_catnap(PZERO, error, "nfs_seek");
8304 		} else if ((error == NFSERR_EXPIRED ||
8305 		    error == NFSERR_BADSTATEID) && clidrev != 0) {
8306 			expireret = nfscl_hasexpired(nmp->nm_clp, clidrev,
8307 			    curthread);
8308 		}
8309 		retrycnt++;
8310 	} while (error == NFSERR_GRACE || error == NFSERR_STALESTATEID ||
8311 	    error == NFSERR_STALEDONTRECOVER || error == NFSERR_DELAY ||
8312 	    error == NFSERR_BADSESSION ||
8313 	    (error == NFSERR_OLDSTATEID && retrycnt < 20) ||
8314 	    ((error == NFSERR_EXPIRED || error == NFSERR_BADSTATEID) &&
8315 	     expireret == 0 && clidrev != 0 && retrycnt < 4) ||
8316 	    (error == NFSERR_OPENMODE && retrycnt < 4));
8317 	if (error && retrycnt >= 4)
8318 		error = EIO;
8319 	return (error);
8320 }
8321 
8322 /*
8323  * The seek RPC.
8324  */
8325 static int
8326 nfsrpc_seekrpc(vnode_t vp, off_t *offp, nfsv4stateid_t *stateidp, bool *eofp,
8327     int content, struct nfsvattr *nap, int *attrflagp, struct ucred *cred)
8328 {
8329 	uint32_t *tl;
8330 	int error;
8331 	struct nfsrv_descript nfsd;
8332 	struct nfsrv_descript *nd = &nfsd;
8333 	nfsattrbit_t attrbits;
8334 
8335 	*attrflagp = 0;
8336 	NFSCL_REQSTART(nd, NFSPROC_SEEK, vp);
8337 	nfsm_stateidtom(nd, stateidp, NFSSTATEID_PUTSTATEID);
8338 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8339 	txdr_hyper(*offp, tl); tl += 2;
8340 	*tl++ = txdr_unsigned(content);
8341 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8342 	NFSGETATTR_ATTRBIT(&attrbits);
8343 	nfsrv_putattrbit(nd, &attrbits);
8344 	error = nfscl_request(nd, vp, curthread, cred, NULL);
8345 	if (error != 0)
8346 		return (error);
8347 	if (nd->nd_repstat == 0) {
8348 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED + NFSX_HYPER);
8349 		if (*tl++ == newnfs_true)
8350 			*eofp = true;
8351 		else
8352 			*eofp = false;
8353 		*offp = fxdr_hyper(tl);
8354 		/* Just skip over Getattr op status. */
8355 		error = nfsm_loadattr(nd, nap);
8356 		if (error == 0)
8357 			*attrflagp = 1;
8358 	}
8359 	error = nd->nd_repstat;
8360 nfsmout:
8361 	m_freem(nd->nd_mrep);
8362 	return (error);
8363 }
8364 
8365 /*
8366  * The getextattr RPC.
8367  */
8368 int
8369 nfsrpc_getextattr(vnode_t vp, const char *name, struct uio *uiop, ssize_t *lenp,
8370     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8371 {
8372 	uint32_t *tl;
8373 	int error;
8374 	struct nfsrv_descript nfsd;
8375 	struct nfsrv_descript *nd = &nfsd;
8376 	nfsattrbit_t attrbits;
8377 	uint32_t len, len2;
8378 
8379 	*attrflagp = 0;
8380 	NFSCL_REQSTART(nd, NFSPROC_GETEXTATTR, vp);
8381 	nfsm_strtom(nd, name, strlen(name));
8382 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8383 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8384 	NFSGETATTR_ATTRBIT(&attrbits);
8385 	nfsrv_putattrbit(nd, &attrbits);
8386 	error = nfscl_request(nd, vp, p, cred, NULL);
8387 	if (error != 0)
8388 		return (error);
8389 	if (nd->nd_repstat == 0) {
8390 		NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8391 		len = fxdr_unsigned(uint32_t, *tl);
8392 		/* Sanity check lengths. */
8393 		if (uiop != NULL && len > 0 && len <= IOSIZE_MAX &&
8394 		    uiop->uio_resid <= UINT32_MAX) {
8395 			len2 = uiop->uio_resid;
8396 			if (len2 >= len)
8397 				error = nfsm_mbufuio(nd, uiop, len);
8398 			else {
8399 				error = nfsm_mbufuio(nd, uiop, len2);
8400 				if (error == 0) {
8401 					/*
8402 					 * nfsm_mbufuio() advances to a multiple
8403 					 * of 4, so round up len2 as well.  Then
8404 					 * we need to advance over the rest of
8405 					 * the data, rounding up the remaining
8406 					 * length.
8407 					 */
8408 					len2 = NFSM_RNDUP(len2);
8409 					len2 = NFSM_RNDUP(len - len2);
8410 					if (len2 > 0)
8411 						error = nfsm_advance(nd, len2,
8412 						    -1);
8413 				}
8414 			}
8415 		} else if (uiop == NULL && len > 0) {
8416 			/* Just wants the length and not the data. */
8417 			error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8418 		} else if (len > 0)
8419 			error = ENOATTR;
8420 		if (error != 0)
8421 			goto nfsmout;
8422 		*lenp = len;
8423 		/* Just skip over Getattr op status. */
8424 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_UNSIGNED);
8425 		error = nfsm_loadattr(nd, nap);
8426 		if (error == 0)
8427 			*attrflagp = 1;
8428 	}
8429 	if (error == 0)
8430 		error = nd->nd_repstat;
8431 nfsmout:
8432 	m_freem(nd->nd_mrep);
8433 	return (error);
8434 }
8435 
8436 /*
8437  * The setextattr RPC.
8438  */
8439 int
8440 nfsrpc_setextattr(vnode_t vp, const char *name, struct uio *uiop,
8441     struct nfsvattr *nap, int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8442 {
8443 	uint32_t *tl;
8444 	int error;
8445 	struct nfsrv_descript nfsd;
8446 	struct nfsrv_descript *nd = &nfsd;
8447 	nfsattrbit_t attrbits;
8448 
8449 	*attrflagp = 0;
8450 	NFSCL_REQSTART(nd, NFSPROC_SETEXTATTR, vp);
8451 	if (uiop->uio_resid > nd->nd_maxreq) {
8452 		/* nd_maxreq is set by NFSCL_REQSTART(). */
8453 		m_freem(nd->nd_mreq);
8454 		return (EINVAL);
8455 	}
8456 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8457 	*tl = txdr_unsigned(NFSV4SXATTR_EITHER);
8458 	nfsm_strtom(nd, name, strlen(name));
8459 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8460 	*tl = txdr_unsigned(uiop->uio_resid);
8461 	nfsm_uiombuf(nd, uiop, uiop->uio_resid);
8462 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8463 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8464 	NFSGETATTR_ATTRBIT(&attrbits);
8465 	nfsrv_putattrbit(nd, &attrbits);
8466 	error = nfscl_request(nd, vp, p, cred, NULL);
8467 	if (error != 0)
8468 		return (error);
8469 	if (nd->nd_repstat == 0) {
8470 		/* Just skip over the reply and Getattr op status. */
8471 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8472 		    NFSX_UNSIGNED);
8473 		error = nfsm_loadattr(nd, nap);
8474 		if (error == 0)
8475 			*attrflagp = 1;
8476 	}
8477 	if (error == 0)
8478 		error = nd->nd_repstat;
8479 nfsmout:
8480 	m_freem(nd->nd_mrep);
8481 	return (error);
8482 }
8483 
8484 /*
8485  * The removeextattr RPC.
8486  */
8487 int
8488 nfsrpc_rmextattr(vnode_t vp, const char *name, struct nfsvattr *nap,
8489     int *attrflagp, struct ucred *cred, NFSPROC_T *p)
8490 {
8491 	uint32_t *tl;
8492 	int error;
8493 	struct nfsrv_descript nfsd;
8494 	struct nfsrv_descript *nd = &nfsd;
8495 	nfsattrbit_t attrbits;
8496 
8497 	*attrflagp = 0;
8498 	NFSCL_REQSTART(nd, NFSPROC_RMEXTATTR, vp);
8499 	nfsm_strtom(nd, name, strlen(name));
8500 	NFSM_BUILD(tl, uint32_t *, NFSX_UNSIGNED);
8501 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8502 	NFSGETATTR_ATTRBIT(&attrbits);
8503 	nfsrv_putattrbit(nd, &attrbits);
8504 	error = nfscl_request(nd, vp, p, cred, NULL);
8505 	if (error != 0)
8506 		return (error);
8507 	if (nd->nd_repstat == 0) {
8508 		/* Just skip over the reply and Getattr op status. */
8509 		NFSM_DISSECT(tl, uint32_t *, 2 * NFSX_HYPER + 3 *
8510 		    NFSX_UNSIGNED);
8511 		error = nfsm_loadattr(nd, nap);
8512 		if (error == 0)
8513 			*attrflagp = 1;
8514 	}
8515 	if (error == 0)
8516 		error = nd->nd_repstat;
8517 nfsmout:
8518 	m_freem(nd->nd_mrep);
8519 	return (error);
8520 }
8521 
8522 /*
8523  * The listextattr RPC.
8524  */
8525 int
8526 nfsrpc_listextattr(vnode_t vp, uint64_t *cookiep, struct uio *uiop,
8527     size_t *lenp, bool *eofp, struct nfsvattr *nap, int *attrflagp,
8528     struct ucred *cred, NFSPROC_T *p)
8529 {
8530 	uint32_t *tl;
8531 	int cnt, error, i, len;
8532 	struct nfsrv_descript nfsd;
8533 	struct nfsrv_descript *nd = &nfsd;
8534 	nfsattrbit_t attrbits;
8535 	u_char c;
8536 
8537 	*attrflagp = 0;
8538 	NFSCL_REQSTART(nd, NFSPROC_LISTEXTATTR, vp);
8539 	NFSM_BUILD(tl, uint32_t *, NFSX_HYPER + 2 * NFSX_UNSIGNED);
8540 	txdr_hyper(*cookiep, tl); tl += 2;
8541 	*tl++ = txdr_unsigned(*lenp);
8542 	*tl = txdr_unsigned(NFSV4OP_GETATTR);
8543 	NFSGETATTR_ATTRBIT(&attrbits);
8544 	nfsrv_putattrbit(nd, &attrbits);
8545 	error = nfscl_request(nd, vp, p, cred, NULL);
8546 	if (error != 0)
8547 		return (error);
8548 	*eofp = true;
8549 	*lenp = 0;
8550 	if (nd->nd_repstat == 0) {
8551 		NFSM_DISSECT(tl, uint32_t *, NFSX_HYPER + NFSX_UNSIGNED);
8552 		*cookiep = fxdr_hyper(tl); tl += 2;
8553 		cnt = fxdr_unsigned(int, *tl);
8554 		if (cnt < 0) {
8555 			error = EBADRPC;
8556 			goto nfsmout;
8557 		}
8558 		for (i = 0; i < cnt; i++) {
8559 			NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED);
8560 			len = fxdr_unsigned(int, *tl);
8561 			if (len <= 0 || len > EXTATTR_MAXNAMELEN) {
8562 				error = EBADRPC;
8563 				goto nfsmout;
8564 			}
8565 			if (uiop == NULL)
8566 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8567 			else if (uiop->uio_resid >= len + 1) {
8568 				c = len;
8569 				error = uiomove(&c, sizeof(c), uiop);
8570 				if (error == 0)
8571 					error = nfsm_mbufuio(nd, uiop, len);
8572 			} else {
8573 				error = nfsm_advance(nd, NFSM_RNDUP(len), -1);
8574 				*eofp = false;
8575 			}
8576 			if (error != 0)
8577 				goto nfsmout;
8578 			*lenp += (len + 1);
8579 		}
8580 		/* Get the eof and skip over the Getattr op status. */
8581 		NFSM_DISSECT(tl, uint32_t *, 3 * NFSX_UNSIGNED);
8582 		/*
8583 		 * *eofp is set false above, because it wasn't able to copy
8584 		 * all of the reply.
8585 		 */
8586 		if (*eofp && *tl == 0)
8587 			*eofp = false;
8588 		error = nfsm_loadattr(nd, nap);
8589 		if (error == 0)
8590 			*attrflagp = 1;
8591 	}
8592 	if (error == 0)
8593 		error = nd->nd_repstat;
8594 nfsmout:
8595 	m_freem(nd->nd_mrep);
8596 	return (error);
8597 }
8598 
8599