xref: /netbsd-src/sys/nfs/nfs_serv.c (revision fad4c9f71477ae11cea2ee75ec82151ac770a534)
1 /*	$NetBSD: nfs_serv.c,v 1.113 2006/06/30 09:56:03 yamt Exp $	*/
2 
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  *	@(#)nfs_serv.c	8.8 (Berkeley) 7/31/95
35  */
36 
37 /*
38  * nfs version 2 and 3 server calls to vnode ops
39  * - these routines generally have 3 phases
40  *   1 - break down and validate rpc request in mbuf list
41  *   2 - do the vnode ops for the request
42  *       (surprisingly ?? many are very similar to syscalls in vfs_syscalls.c)
43  *   3 - build the rpc reply in an mbuf list
44  *   nb:
45  *	- do not mix the phases, since the nfsm_?? macros can return failures
46  *	  on a bad rpc or similar and do not do any vrele() or vput()'s
47  *
48  *      - the nfsm_reply() macro generates an nfs rpc reply with the nfs
49  *	error number iff error != 0 whereas
50  *	returning an error from the server function implies a fatal error
51  *	such as a badly constructed rpc request that should be dropped without
52  *	a reply.
53  *	For Version 3, nfsm_reply() does not return for the error case, since
54  *	most version 3 rpcs return more than the status for error cases.
55  */
56 
57 #include <sys/cdefs.h>
58 __KERNEL_RCSID(0, "$NetBSD: nfs_serv.c,v 1.113 2006/06/30 09:56:03 yamt Exp $");
59 
60 #include <sys/param.h>
61 #include <sys/systm.h>
62 #include <sys/proc.h>
63 #include <sys/file.h>
64 #include <sys/namei.h>
65 #include <sys/vnode.h>
66 #include <sys/mount.h>
67 #include <sys/socket.h>
68 #include <sys/socketvar.h>
69 #include <sys/mbuf.h>
70 #include <sys/dirent.h>
71 #include <sys/stat.h>
72 #include <sys/kernel.h>
73 #include <sys/kauth.h>
74 
75 #include <uvm/uvm.h>
76 
77 #include <nfs/nfsproto.h>
78 #include <nfs/rpcv2.h>
79 #include <nfs/nfs.h>
80 #include <nfs/xdr_subs.h>
81 #include <nfs/nfsm_subs.h>
82 #include <nfs/nqnfs.h>
83 #include <nfs/nfs_var.h>
84 
85 /* Global vars */
86 extern u_int32_t nfs_xdrneg1;
87 extern u_int32_t nfs_false, nfs_true;
88 extern enum vtype nv3tov_type[8];
89 extern struct nfsstats nfsstats;
90 extern const nfstype nfsv2_type[9];
91 extern const nfstype nfsv3_type[9];
92 int nfsrvw_procrastinate = NFS_GATHERDELAY * 1000;
93 int nfsd_use_loan = 1;	/* use page-loan for READ OP */
94 
95 /*
96  * nfs v3 access service
97  */
98 int
99 nfsrv3_access(nfsd, slp, lwp, mrq)
100 	struct nfsrv_descript *nfsd;
101 	struct nfssvc_sock *slp;
102 	struct lwp *lwp;
103 	struct mbuf **mrq;
104 {
105 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
106 	struct mbuf *nam = nfsd->nd_nam;
107 	caddr_t dpos = nfsd->nd_dpos;
108 	kauth_cred_t cred = nfsd->nd_cr;
109 	struct vnode *vp;
110 	nfsfh_t nfh;
111 	fhandle_t *fhp;
112 	u_int32_t *tl;
113 	int32_t t1;
114 	caddr_t bpos;
115 	int error = 0, rdonly, cache = 0, getret;
116 	char *cp2;
117 	struct mbuf *mb, *mreq;
118 	struct vattr va;
119 	u_long inmode, testmode, outmode;
120 	u_quad_t frev;
121 
122 	fhp = &nfh.fh_generic;
123 	nfsm_srvmtofh(fhp);
124 	nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
125 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
126 	    (nfsd->nd_flag & ND_KERBAUTH), FALSE);
127 	if (error) {
128 		nfsm_reply(NFSX_UNSIGNED);
129 		nfsm_srvpostop_attr(1, (struct vattr *)0);
130 		return (0);
131 	}
132 	inmode = fxdr_unsigned(u_int32_t, *tl);
133 	outmode = 0;
134 	if ((inmode & NFSV3ACCESS_READ) &&
135 	    nfsrv_access(vp, VREAD, cred, rdonly, lwp, 0) == 0)
136 		outmode |= NFSV3ACCESS_READ;
137 	if (vp->v_type != VDIR) {
138 		testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND);
139 		if (testmode &&
140 		    nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
141 			outmode |= testmode;
142 		if ((inmode & NFSV3ACCESS_EXECUTE) &&
143 		    nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
144 			outmode |= NFSV3ACCESS_EXECUTE;
145 	} else {
146 		testmode = inmode & (NFSV3ACCESS_MODIFY | NFSV3ACCESS_EXTEND |
147 		    NFSV3ACCESS_DELETE);
148 		if (testmode &&
149 		    nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 0) == 0)
150 			outmode |= testmode;
151 		if ((inmode & NFSV3ACCESS_LOOKUP) &&
152 		    nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0) == 0)
153 			outmode |= NFSV3ACCESS_LOOKUP;
154 	}
155 	getret = VOP_GETATTR(vp, &va, cred, lwp);
156 	vput(vp);
157 	nfsm_reply(NFSX_POSTOPATTR(1) + NFSX_UNSIGNED);
158 	nfsm_srvpostop_attr(getret, &va);
159 	nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
160 	*tl = txdr_unsigned(outmode);
161 	nfsm_srvdone;
162 }
163 
164 /*
165  * nfs getattr service
166  */
167 int
168 nfsrv_getattr(nfsd, slp, lwp, mrq)
169 	struct nfsrv_descript *nfsd;
170 	struct nfssvc_sock *slp;
171 	struct lwp *lwp;
172 	struct mbuf **mrq;
173 {
174 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
175 	struct mbuf *nam = nfsd->nd_nam;
176 	caddr_t dpos = nfsd->nd_dpos;
177 	kauth_cred_t cred = nfsd->nd_cr;
178 	struct nfs_fattr *fp;
179 	struct vattr va;
180 	struct vnode *vp;
181 	nfsfh_t nfh;
182 	fhandle_t *fhp;
183 	u_int32_t *tl;
184 	int32_t t1;
185 	caddr_t bpos;
186 	int error = 0, rdonly, cache = 0;
187 	char *cp2;
188 	struct mbuf *mb, *mreq;
189 	u_quad_t frev;
190 
191 	fhp = &nfh.fh_generic;
192 	nfsm_srvmtofh(fhp);
193 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
194 	    (nfsd->nd_flag & ND_KERBAUTH), FALSE);
195 	if (error) {
196 		nfsm_reply(0);
197 		return (0);
198 	}
199 	nqsrv_getl(vp, ND_READ);
200 	error = VOP_GETATTR(vp, &va, cred, lwp);
201 	vput(vp);
202 	nfsm_reply(NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
203 	if (error)
204 		return (0);
205 	nfsm_build(fp, struct nfs_fattr *, NFSX_FATTR(nfsd->nd_flag & ND_NFSV3));
206 	nfsm_srvfillattr(&va, fp);
207 	nfsm_srvdone;
208 }
209 
210 /*
211  * nfs setattr service
212  */
213 int
214 nfsrv_setattr(nfsd, slp, lwp, mrq)
215 	struct nfsrv_descript *nfsd;
216 	struct nfssvc_sock *slp;
217 	struct lwp *lwp;
218 	struct mbuf **mrq;
219 {
220 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
221 	struct mbuf *nam = nfsd->nd_nam;
222 	caddr_t dpos = nfsd->nd_dpos;
223 	kauth_cred_t cred = nfsd->nd_cr;
224 	struct vattr va, preat;
225 	struct nfsv2_sattr *sp;
226 	struct nfs_fattr *fp;
227 	struct vnode *vp;
228 	nfsfh_t nfh;
229 	fhandle_t *fhp;
230 	u_int32_t *tl;
231 	int32_t t1;
232 	caddr_t bpos;
233 	int error = 0, rdonly, cache = 0, preat_ret = 1, postat_ret = 1;
234 	int v3 = (nfsd->nd_flag & ND_NFSV3), gcheck = 0;
235 	char *cp2;
236 	struct mbuf *mb, *mreq;
237 	u_quad_t frev;
238 	struct timespec guard;
239 	struct mount *mp = NULL;
240 
241 	memset(&guard, 0, sizeof guard);	/* XXX gcc */
242 
243 	fhp = &nfh.fh_generic;
244 	nfsm_srvmtofh(fhp);
245 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
246 		return (ESTALE);
247 	vn_start_write(NULL, &mp, V_WAIT);
248 	VATTR_NULL(&va);
249 	if (v3) {
250 		nfsm_srvsattr(&va);
251 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
252 		gcheck = fxdr_unsigned(int, *tl);
253 		if (gcheck) {
254 			nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
255 			fxdr_nfsv3time(tl, &guard);
256 		}
257 	} else {
258 		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
259 		/*
260 		 * Nah nah nah nah na nah
261 		 * There is a bug in the Sun client that puts 0xffff in the mode
262 		 * field of sattr when it should put in 0xffffffff. The u_short
263 		 * doesn't sign extend.
264 		 * --> check the low order 2 bytes for 0xffff
265 		 */
266 		if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
267 			va.va_mode = nfstov_mode(sp->sa_mode);
268 		if (sp->sa_uid != nfs_xdrneg1)
269 			va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
270 		if (sp->sa_gid != nfs_xdrneg1)
271 			va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
272 		if (sp->sa_size != nfs_xdrneg1)
273 			va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
274 		if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
275 #ifdef notyet
276 			fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
277 #else
278 			va.va_atime.tv_sec =
279 				fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
280 			va.va_atime.tv_nsec = 0;
281 #endif
282 		}
283 		if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
284 			fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
285 
286 	}
287 
288 	/*
289 	 * Now that we have all the fields, lets do it.
290 	 */
291 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
292 	    (nfsd->nd_flag & ND_KERBAUTH), FALSE);
293 	if (error) {
294 		nfsm_reply(2 * NFSX_UNSIGNED);
295 		nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
296 		vn_finished_write(mp, 0);
297 		return (0);
298 	}
299 	nqsrv_getl(vp, ND_WRITE);
300 	if (v3) {
301 		error = preat_ret = VOP_GETATTR(vp, &preat, cred, lwp);
302 		if (!error && gcheck &&
303 			(preat.va_ctime.tv_sec != guard.tv_sec ||
304 			 preat.va_ctime.tv_nsec != guard.tv_nsec))
305 			error = NFSERR_NOT_SYNC;
306 		if (error) {
307 			vput(vp);
308 			nfsm_reply(NFSX_WCCDATA(v3));
309 			nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
310 			vn_finished_write(mp, 0);
311 			return (0);
312 		}
313 	}
314 
315 	/*
316 	 * If the size is being changed write acces is required, otherwise
317 	 * just check for a read only file system.
318 	 */
319 	if (va.va_size == ((u_quad_t)((quad_t) -1))) {
320 		if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
321 			error = EROFS;
322 			goto out;
323 		}
324 	} else {
325 		if (vp->v_type == VDIR) {
326 			error = EISDIR;
327 			goto out;
328 		} else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
329 			lwp, 0)) != 0)
330 			goto out;
331 	}
332 	error = VOP_SETATTR(vp, &va, cred, lwp);
333 	postat_ret = VOP_GETATTR(vp, &va, cred, lwp);
334 	if (!error)
335 		error = postat_ret;
336 out:
337 	vput(vp);
338 	vn_finished_write(mp, 0);
339 	nfsm_reply(NFSX_WCCORFATTR(v3));
340 	if (v3) {
341 		nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
342 		return (0);
343 	} else {
344 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
345 		nfsm_srvfillattr(&va, fp);
346 	}
347 	nfsm_srvdone;
348 }
349 
350 /*
351  * nfs lookup rpc
352  */
353 int
354 nfsrv_lookup(nfsd, slp, lwp, mrq)
355 	struct nfsrv_descript *nfsd;
356 	struct nfssvc_sock *slp;
357 	struct lwp *lwp;
358 	struct mbuf **mrq;
359 {
360 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
361 	struct mbuf *nam = nfsd->nd_nam;
362 	caddr_t dpos = nfsd->nd_dpos;
363 	kauth_cred_t cred = nfsd->nd_cr;
364 	struct nfs_fattr *fp;
365 	struct nameidata nd, ind, *ndp = &nd;
366 	struct vnode *vp, *dirp;
367 	nfsfh_t nfh;
368 	fhandle_t *fhp;
369 	caddr_t cp;
370 	u_int32_t *tl;
371 	int32_t t1;
372 	caddr_t bpos;
373 	int error = 0, cache = 0, dirattr_ret = 1;
374 	uint32_t len;
375 	int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
376 	char *cp2;
377 	struct mbuf *mb, *mreq;
378 	struct vattr va, dirattr;
379 	u_quad_t frev;
380 
381 	fhp = &nfh.fh_generic;
382 	nfsm_srvmtofh(fhp);
383 	nfsm_srvnamesiz(len);
384 
385 	pubflag = nfs_ispublicfh(fhp);
386 
387 	nd.ni_cnd.cn_cred = cred;
388 	nd.ni_cnd.cn_nameiop = LOOKUP;
389 	nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
390 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
391 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
392 
393 	if (!error && pubflag) {
394 		if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
395 			/*
396 			 * Setup call to lookup() to see if we can find
397 			 * the index file. Arguably, this doesn't belong
398 			 * in a kernel.. Ugh.
399 			 */
400 			ind = nd;
401 			VOP_UNLOCK(nd.ni_vp, 0);
402 			ind.ni_pathlen = strlen(nfs_pub.np_index);
403 			ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
404 			    nfs_pub.np_index;
405 			ind.ni_startdir = nd.ni_vp;
406 			VREF(ind.ni_startdir);
407 			error = lookup(&ind);
408 			if (!error) {
409 				/*
410 				 * Found an index file. Get rid of
411 				 * the old references.
412 				 */
413 				if (dirp)
414 					vrele(dirp);
415 				dirp = nd.ni_vp;
416 				vrele(nd.ni_startdir);
417 				ndp = &ind;
418 			} else
419 				error = 0;
420 		}
421 		/*
422 		 * If the public filehandle was used, check that this lookup
423 		 * didn't result in a filehandle outside the publicly exported
424 		 * filesystem.
425 		 */
426 
427 		if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
428 			vput(nd.ni_vp);
429 			error = EPERM;
430 		}
431 	}
432 
433 	if (dirp) {
434 		if (v3)
435 			dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred, lwp);
436 		vrele(dirp);
437 	}
438 
439 	if (error) {
440 		nfsm_reply(NFSX_POSTOPATTR(v3));
441 		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
442 		return (0);
443 	}
444 
445 	nqsrv_getl(ndp->ni_startdir, ND_READ);
446 	vrele(ndp->ni_startdir);
447 	PNBUF_PUT(nd.ni_cnd.cn_pnbuf);
448 	vp = ndp->ni_vp;
449 	error = vfs_composefh(vp, fhp);
450 	if (!error)
451 		error = VOP_GETATTR(vp, &va, cred, lwp);
452 	vput(vp);
453 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
454 	if (error) {
455 		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
456 		return (0);
457 	}
458 	nfsm_srvfhtom(fhp, v3);
459 	if (v3) {
460 		nfsm_srvpostop_attr(0, &va);
461 		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
462 	} else {
463 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
464 		nfsm_srvfillattr(&va, fp);
465 	}
466 	nfsm_srvdone;
467 }
468 
469 /*
470  * nfs readlink service
471  */
472 int
473 nfsrv_readlink(nfsd, slp, lwp, mrq)
474 	struct nfsrv_descript *nfsd;
475 	struct nfssvc_sock *slp;
476 	struct lwp *lwp;
477 	struct mbuf **mrq;
478 {
479 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
480 	struct mbuf *nam = nfsd->nd_nam;
481 	caddr_t dpos = nfsd->nd_dpos;
482 	kauth_cred_t cred = nfsd->nd_cr;
483 	struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
484 	struct iovec *ivp = iv;
485 	struct mbuf *mp;
486 	u_int32_t *tl;
487 	int32_t t1;
488 	caddr_t bpos;
489 	int error = 0, rdonly, cache = 0, i, padlen, getret;
490 	uint32_t len;
491 	int v3 = (nfsd->nd_flag & ND_NFSV3);
492 	char *cp2;
493 	struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq;
494 	struct vnode *vp;
495 	struct vattr attr;
496 	nfsfh_t nfh;
497 	fhandle_t *fhp;
498 	struct uio io, *uiop = &io;
499 	u_quad_t frev;
500 
501 	fhp = &nfh.fh_generic;
502 	nfsm_srvmtofh(fhp);
503 	len = 0;
504 	i = 0;
505 	while (len < NFS_MAXPATHLEN) {
506 		mp = m_get(M_WAIT, MT_DATA);
507 		MCLAIM(mp, &nfs_mowner);
508 		m_clget(mp, M_WAIT);
509 		mp->m_len = NFSMSIZ(mp);
510 		if (len == 0)
511 			mp3 = mp2 = mp;
512 		else {
513 			mp2->m_next = mp;
514 			mp2 = mp;
515 		}
516 		if ((len+mp->m_len) > NFS_MAXPATHLEN) {
517 			mp->m_len = NFS_MAXPATHLEN-len;
518 			len = NFS_MAXPATHLEN;
519 		} else
520 			len += mp->m_len;
521 		ivp->iov_base = mtod(mp, caddr_t);
522 		ivp->iov_len = mp->m_len;
523 		i++;
524 		ivp++;
525 	}
526 	uiop->uio_iov = iv;
527 	uiop->uio_iovcnt = i;
528 	uiop->uio_offset = 0;
529 	uiop->uio_resid = len;
530 	uiop->uio_rw = UIO_READ;
531 	UIO_SETUP_SYSSPACE(uiop);
532 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
533 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
534 	if (error) {
535 		m_freem(mp3);
536 		nfsm_reply(2 * NFSX_UNSIGNED);
537 		nfsm_srvpostop_attr(1, (struct vattr *)0);
538 		return (0);
539 	}
540 	if (vp->v_type != VLNK) {
541 		if (v3)
542 			error = EINVAL;
543 		else
544 			error = ENXIO;
545 		goto out;
546 	}
547 	nqsrv_getl(vp, ND_READ);
548 	error = VOP_READLINK(vp, uiop, cred);
549 out:
550 	getret = VOP_GETATTR(vp, &attr, cred, lwp);
551 	vput(vp);
552 	if (error)
553 		m_freem(mp3);
554 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
555 	if (v3) {
556 		nfsm_srvpostop_attr(getret, &attr);
557 		if (error)
558 			return (0);
559 	}
560 	len -= uiop->uio_resid;
561 	padlen = nfsm_padlen(len);
562 	if (uiop->uio_resid || padlen)
563 		nfs_zeropad(mp3, uiop->uio_resid, padlen);
564 	nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
565 	*tl = txdr_unsigned(len);
566 	mb->m_next = mp3;
567 	nfsm_srvdone;
568 }
569 
570 /*
571  * nfs read service
572  */
573 int
574 nfsrv_read(nfsd, slp, lwp, mrq)
575 	struct nfsrv_descript *nfsd;
576 	struct nfssvc_sock *slp;
577 	struct lwp *lwp;
578 	struct mbuf **mrq;
579 {
580 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
581 	struct mbuf *nam = nfsd->nd_nam;
582 	caddr_t dpos = nfsd->nd_dpos;
583 	kauth_cred_t cred = nfsd->nd_cr;
584 	struct mbuf *m;
585 	struct nfs_fattr *fp;
586 	u_int32_t *tl;
587 	int32_t t1;
588 	int i;
589 	caddr_t bpos;
590 	int error = 0, rdonly, cache = 0, getret;
591 	int v3 = (nfsd->nd_flag & ND_NFSV3);
592 	uint32_t reqlen, len, cnt, left;
593 	int padlen;
594 	char *cp2;
595 	struct mbuf *mb, *mreq;
596 	struct vnode *vp;
597 	nfsfh_t nfh;
598 	fhandle_t *fhp;
599 	struct uio io, *uiop = &io;
600 	struct vattr va;
601 	off_t off;
602 	u_quad_t frev;
603 
604 	fhp = &nfh.fh_generic;
605 	nfsm_srvmtofh(fhp);
606 	if (v3) {
607 		nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
608 		off = fxdr_hyper(tl);
609 	} else {
610 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
611 		off = (off_t)fxdr_unsigned(u_int32_t, *tl);
612 	}
613 	nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
614 	reqlen = fxdr_unsigned(uint32_t, *tl);
615 	reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd));
616 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
617 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
618 	if (error) {
619 		nfsm_reply(2 * NFSX_UNSIGNED);
620 		nfsm_srvpostop_attr(1, (struct vattr *)0);
621 		return (0);
622 	}
623 	if (vp->v_type != VREG) {
624 		if (v3)
625 			error = EINVAL;
626 		else
627 			error = (vp->v_type == VDIR) ? EISDIR : EACCES;
628 	}
629 	if (!error) {
630 	    nqsrv_getl(vp, ND_READ);
631 	    if ((error = nfsrv_access(vp, VREAD, cred, rdonly, lwp, 1)) != 0)
632 		error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 1);
633 	}
634 	getret = VOP_GETATTR(vp, &va, cred, lwp);
635 	if (!error)
636 		error = getret;
637 	if (error) {
638 		vput(vp);
639 		nfsm_reply(NFSX_POSTOPATTR(v3));
640 		nfsm_srvpostop_attr(getret, &va);
641 		return (0);
642 	}
643 	if (off >= va.va_size)
644 		cnt = 0;
645 	else if ((off + reqlen) > va.va_size)
646 		cnt = va.va_size - off;
647 	else
648 		cnt = reqlen;
649 	nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
650 	if (v3) {
651 		nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
652 		*tl++ = nfs_true;
653 		fp = (struct nfs_fattr *)tl;
654 		tl += (NFSX_V3FATTR / sizeof (u_int32_t));
655 	} else {
656 		nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
657 		fp = (struct nfs_fattr *)tl;
658 		tl += (NFSX_V2FATTR / sizeof (u_int32_t));
659 	}
660 	len = left = cnt;
661 	if (cnt > 0) {
662 		if (nfsd_use_loan) {
663 			struct vm_page **pgpp;
664 			voff_t pgoff = trunc_page(off);
665 			int npages;
666 			vaddr_t lva;
667 
668 			npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT;
669 			KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */
670 
671 			/* allocate kva for mbuf data */
672 			lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so);
673 			if (lva == 0) {
674 				/* fall back to VOP_READ */
675 				goto loan_fail;
676 			}
677 
678 			/* allocate mbuf */
679 			m = m_get(M_WAIT, MT_DATA);
680 			MCLAIM(m, &nfs_mowner);
681 			pgpp = m->m_ext.ext_pgs;
682 
683 			/* loan pages */
684 			error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages,
685 			    pgpp);
686 			if (error) {
687 				sokvafree(lva, npages << PAGE_SHIFT);
688 				m_free(m);
689 				if (error == EBUSY)
690 					goto loan_fail;
691 				goto read_error;
692 			}
693 
694 			/* associate kva to mbuf */
695 			MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)),
696 			    cnt, M_MBUF, soloanfree, slp->ns_so);
697 			m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
698 			m->m_len = cnt;
699 
700 			/* map pages */
701 			for (i = 0; i < npages; i++) {
702 				pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]),
703 				    VM_PROT_READ);
704 				lva += PAGE_SIZE;
705 			}
706 
707 			pmap_update(pmap_kernel());
708 
709 			mb->m_next = m;
710 			mb = m;
711 			error = 0;
712 			uiop->uio_resid = 0;
713 		} else {
714 			struct iovec *iv;
715 			struct iovec *iv2;
716 			struct mbuf *m2;
717 			int siz;
718 loan_fail:
719 			/*
720 			 * Generate the mbuf list with the uio_iov ref. to it.
721 			 */
722 			i = 0;
723 			m = m2 = mb;
724 			while (left > 0) {
725 				siz = min(M_TRAILINGSPACE(m), left);
726 				if (siz > 0) {
727 					left -= siz;
728 					i++;
729 				}
730 				if (left > 0) {
731 					m = m_get(M_WAIT, MT_DATA);
732 					MCLAIM(m, &nfs_mowner);
733 					m_clget(m, M_WAIT);
734 					m->m_len = 0;
735 					m2->m_next = m;
736 					m2 = m;
737 				}
738 			}
739 			iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK);
740 			uiop->uio_iov = iv2 = iv;
741 			m = mb;
742 			left = cnt;
743 			i = 0;
744 			while (left > 0) {
745 				if (m == NULL)
746 					panic("nfsrv_read iov");
747 				siz = min(M_TRAILINGSPACE(m), left);
748 				if (siz > 0) {
749 					iv->iov_base = mtod(m, caddr_t) +
750 					    m->m_len;
751 					iv->iov_len = siz;
752 					m->m_len += siz;
753 					left -= siz;
754 					iv++;
755 					i++;
756 				}
757 				m = m->m_next;
758 			}
759 			uiop->uio_iovcnt = i;
760 			uiop->uio_offset = off;
761 			uiop->uio_resid = cnt;
762 			uiop->uio_rw = UIO_READ;
763 			UIO_SETUP_SYSSPACE(uiop);
764 			error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
765 			free((caddr_t)iv2, M_TEMP);
766 		}
767 read_error:
768 		if (error || (getret = VOP_GETATTR(vp, &va, cred, lwp)) != 0){
769 			if (!error)
770 				error = getret;
771 			m_freem(mreq);
772 			vput(vp);
773 			nfsm_reply(NFSX_POSTOPATTR(v3));
774 			nfsm_srvpostop_attr(getret, &va);
775 			return (0);
776 		}
777 	} else {
778 		uiop->uio_resid = 0;
779 	}
780 	vput(vp);
781 	nfsm_srvfillattr(&va, fp);
782 	len -= uiop->uio_resid;
783 	padlen = nfsm_padlen(len);
784 	if (uiop->uio_resid || padlen)
785 		nfs_zeropad(mb, uiop->uio_resid, padlen);
786 	if (v3) {
787 		/* count */
788 		*tl++ = txdr_unsigned(len);
789 		/* eof */
790 		if (off + len >= va.va_size)
791 			*tl++ = nfs_true;
792 		else
793 			*tl++ = nfs_false;
794 	}
795 	*tl = txdr_unsigned(len);
796 	nfsm_srvdone;
797 }
798 
799 /*
800  * nfs write service
801  */
802 int
803 nfsrv_write(nfsd, slp, lwp, mrq)
804 	struct nfsrv_descript *nfsd;
805 	struct nfssvc_sock *slp;
806 	struct lwp *lwp;
807 	struct mbuf **mrq;
808 {
809 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
810 	struct mbuf *nam = nfsd->nd_nam;
811 	caddr_t dpos = nfsd->nd_dpos;
812 	kauth_cred_t cred = nfsd->nd_cr;
813 	struct iovec *ivp;
814 	int i, cnt;
815 	struct mbuf *mp;
816 	struct nfs_fattr *fp;
817 	struct iovec *iv;
818 	struct vattr va, forat;
819 	u_int32_t *tl;
820 	int32_t t1;
821 	caddr_t bpos;
822 	int error = 0, rdonly, cache = 0, len, forat_ret = 1;
823 	int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
824 	int stable = NFSV3WRITE_FILESYNC;
825 	int v3 = (nfsd->nd_flag & ND_NFSV3);
826 	char *cp2;
827 	struct mbuf *mb, *mreq;
828 	struct vnode *vp;
829 	nfsfh_t nfh;
830 	fhandle_t *fhp;
831 	struct uio io, *uiop = &io;
832 	off_t off;
833 	u_quad_t frev;
834 	struct mount *mntp = NULL;
835 
836 	if (mrep == NULL) {
837 		*mrq = NULL;
838 		return (0);
839 	}
840 	fhp = &nfh.fh_generic;
841 	nfsm_srvmtofh(fhp);
842 	if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
843 		return (ESTALE);
844 	vn_start_write(NULL, &mntp, V_WAIT);
845 	if (v3) {
846 		nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
847 		off = fxdr_hyper(tl);
848 		tl += 3;
849 		stable = fxdr_unsigned(int, *tl++);
850 	} else {
851 		nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
852 		off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
853 		tl += 2;
854 	}
855 	retlen = len = fxdr_unsigned(int32_t, *tl);
856 	cnt = i = 0;
857 
858 	/*
859 	 * For NFS Version 2, it is not obvious what a write of zero length
860 	 * should do, but I might as well be consistent with Version 3,
861 	 * which is to return ok so long as there are no permission problems.
862 	 */
863 	if (len > 0) {
864 		zeroing = 1;
865 		mp = mrep;
866 		while (mp) {
867 			if (mp == md) {
868 				zeroing = 0;
869 				adjust = dpos - mtod(mp, caddr_t);
870 				mp->m_len -= adjust;
871 				if (mp->m_len > 0 && adjust > 0)
872 					NFSMADV(mp, adjust);
873 			}
874 			if (zeroing)
875 				mp->m_len = 0;
876 			else if (mp->m_len > 0) {
877 				i += mp->m_len;
878 				if (i > len) {
879 					mp->m_len -= (i - len);
880 					zeroing	= 1;
881 				}
882 				if (mp->m_len > 0)
883 					cnt++;
884 			}
885 			mp = mp->m_next;
886 		}
887 	}
888 	if (len > NFS_MAXDATA || len < 0 || i < len) {
889 		error = EIO;
890 		nfsm_reply(2 * NFSX_UNSIGNED);
891 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
892 		vn_finished_write(mntp, 0);
893 		return (0);
894 	}
895 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
896 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
897 	if (error) {
898 		nfsm_reply(2 * NFSX_UNSIGNED);
899 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
900 		vn_finished_write(mntp, 0);
901 		return (0);
902 	}
903 	if (v3)
904 		forat_ret = VOP_GETATTR(vp, &forat, cred, lwp);
905 	if (vp->v_type != VREG) {
906 		if (v3)
907 			error = EINVAL;
908 		else
909 			error = (vp->v_type == VDIR) ? EISDIR : EACCES;
910 	}
911 	if (!error) {
912 		nqsrv_getl(vp, ND_WRITE);
913 		error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
914 	}
915 	if (error) {
916 		vput(vp);
917 		nfsm_reply(NFSX_WCCDATA(v3));
918 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
919 		vn_finished_write(mntp, 0);
920 		return (0);
921 	}
922 
923 	if (len > 0) {
924 		ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK);
925 		uiop->uio_iov = iv = ivp;
926 		uiop->uio_iovcnt = cnt;
927 		mp = mrep;
928 		while (mp) {
929 			if (mp->m_len > 0) {
930 				ivp->iov_base = mtod(mp, caddr_t);
931 				ivp->iov_len = mp->m_len;
932 				ivp++;
933 			}
934 			mp = mp->m_next;
935 		}
936 
937 		/*
938 		 * XXX
939 		 * The IO_METASYNC flag indicates that all metadata (and not
940 		 * just enough to ensure data integrity) must be written to
941 		 * stable storage synchronously.
942 		 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
943 		 */
944 		if (stable == NFSV3WRITE_UNSTABLE)
945 			ioflags = IO_NODELOCKED;
946 		else if (stable == NFSV3WRITE_DATASYNC)
947 			ioflags = (IO_SYNC | IO_NODELOCKED);
948 		else
949 			ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
950 		uiop->uio_resid = len;
951 		uiop->uio_rw = UIO_WRITE;
952 		uiop->uio_offset = off;
953 		UIO_SETUP_SYSSPACE(uiop);
954 		error = VOP_WRITE(vp, uiop, ioflags, cred);
955 		nfsstats.srvvop_writes++;
956 		free(iv, M_TEMP);
957 	}
958 	aftat_ret = VOP_GETATTR(vp, &va, cred, lwp);
959 	vput(vp);
960 	vn_finished_write(mntp, 0);
961 	if (!error)
962 		error = aftat_ret;
963 	nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
964 		2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
965 	if (v3) {
966 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
967 		if (error)
968 			return (0);
969 		nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
970 		*tl++ = txdr_unsigned(retlen);
971 		if (stable == NFSV3WRITE_UNSTABLE)
972 			*tl++ = txdr_unsigned(stable);
973 		else
974 			*tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
975 		/*
976 		 * Actually, there is no need to txdr these fields,
977 		 * but it may make the values more human readable,
978 		 * for debugging purposes.
979 		 */
980 		*tl++ = txdr_unsigned(boottime.tv_sec);
981 		*tl = txdr_unsigned(boottime.tv_usec);
982 	} else {
983 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
984 		nfsm_srvfillattr(&va, fp);
985 	}
986 	nfsm_srvdone;
987 }
988 
989 /*
990  * XXX elad: the original NFSW_SAMECRED() macro also made sure the
991  *	     two nd_flag fields of the descriptors contained
992  *	     ND_KERBAUTH.
993  */
994 static int
995 nfsrv_samecred(kauth_cred_t cred1, kauth_cred_t cred2)
996 {
997 	int i, do_ngroups;
998 
999 	if (kauth_cred_geteuid(cred1) != kauth_cred_geteuid(cred2))
1000 		return (0);
1001 	if (kauth_cred_ngroups(cred1) != kauth_cred_ngroups(cred2))
1002 		return (0);
1003 	do_ngroups = kauth_cred_ngroups(cred1);
1004 	for (i = 0; i < do_ngroups; i++)
1005 		if (kauth_cred_group(cred1, i) !=
1006 		    kauth_cred_group(cred2, i))
1007 			return (0);
1008 
1009 	return (1);
1010 }
1011 
1012 /*
1013  * NFS write service with write gathering support. Called when
1014  * nfsrvw_procrastinate > 0.
1015  * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
1016  * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
1017  * Jan. 1994.
1018  */
1019 int
1020 nfsrv_writegather(ndp, slp, lwp, mrq)
1021 	struct nfsrv_descript **ndp;
1022 	struct nfssvc_sock *slp;
1023 	struct lwp *lwp;
1024 	struct mbuf **mrq;
1025 {
1026 	struct timeval now;
1027 	struct iovec *ivp;
1028 	struct mbuf *mp;
1029 	struct nfsrv_descript *wp, *nfsd, *owp, *swp;
1030 	struct nfs_fattr *fp;
1031 	int i = 0;
1032 	struct iovec *iov;
1033 	struct nfsrvw_delayhash *wpp;
1034 	kauth_cred_t cred;
1035 	struct vattr va, forat;
1036 	u_int32_t *tl;
1037 	int32_t t1;
1038 	caddr_t bpos, dpos;
1039 	int error = 0, rdonly, cache = 0, len = 0, forat_ret = 1;
1040 	int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
1041 	char *cp2;
1042 	struct mbuf *mb, *mreq, *mrep, *md;
1043 	struct vnode *vp;
1044 	struct uio io, *uiop = &io;
1045 	u_quad_t frev, cur_usec;
1046 	struct mount *mntp = NULL;
1047 
1048 	*mrq = NULL;
1049 	if (*ndp) {
1050 	    nfsd = *ndp;
1051 	    *ndp = NULL;
1052 	    mrep = nfsd->nd_mrep;
1053 	    md = nfsd->nd_md;
1054 	    dpos = nfsd->nd_dpos;
1055 	    cred = nfsd->nd_cr;
1056 	    v3 = (nfsd->nd_flag & ND_NFSV3);
1057 	    LIST_INIT(&nfsd->nd_coalesce);
1058 	    nfsd->nd_mreq = NULL;
1059 	    nfsd->nd_stable = NFSV3WRITE_FILESYNC;
1060 	    getmicrotime(&now);
1061 	    cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec;
1062 	    nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
1063 
1064 	    /*
1065 	     * Now, get the write header..
1066 	     */
1067 	    nfsm_srvmtofh(&nfsd->nd_fh);
1068 	    if (v3) {
1069 		nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1070 		nfsd->nd_off = fxdr_hyper(tl);
1071 		tl += 3;
1072 		nfsd->nd_stable = fxdr_unsigned(int, *tl++);
1073 	    } else {
1074 		nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1075 		nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
1076 		tl += 2;
1077 	    }
1078 	    len = fxdr_unsigned(int32_t, *tl);
1079 	    nfsd->nd_len = len;
1080 	    nfsd->nd_eoff = nfsd->nd_off + len;
1081 
1082 	    /*
1083 	     * Trim the header out of the mbuf list and trim off any trailing
1084 	     * junk so that the mbuf list has only the write data.
1085 	     */
1086 	    zeroing = 1;
1087 	    i = 0;
1088 	    mp = mrep;
1089 	    while (mp) {
1090 		if (mp == md) {
1091 		    zeroing = 0;
1092 		    adjust = dpos - mtod(mp, caddr_t);
1093 		    mp->m_len -= adjust;
1094 		    if (mp->m_len > 0 && adjust > 0)
1095 			NFSMADV(mp, adjust);
1096 		}
1097 		if (zeroing)
1098 		    mp->m_len = 0;
1099 		else {
1100 		    i += mp->m_len;
1101 		    if (i > len) {
1102 			mp->m_len -= (i - len);
1103 			zeroing = 1;
1104 		    }
1105 		}
1106 		mp = mp->m_next;
1107 	    }
1108 	    if (len > NFS_MAXDATA || len < 0  || i < len) {
1109 nfsmout:
1110 		m_freem(mrep);
1111 		error = EIO;
1112 		nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1113 		if (v3)
1114 		    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1115 		nfsd->nd_mreq = mreq;
1116 		nfsd->nd_mrep = NULL;
1117 		nfsd->nd_time = 0;
1118 	    }
1119 
1120 	    /*
1121 	     * Add this entry to the hash and time queues.
1122 	     */
1123 	    s = splsoftclock();
1124 	    owp = NULL;
1125 	    wp = LIST_FIRST(&slp->ns_tq);
1126 	    while (wp && wp->nd_time < nfsd->nd_time) {
1127 		owp = wp;
1128 		wp = LIST_NEXT(wp, nd_tq);
1129 	    }
1130 	    if (owp) {
1131 		LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1132 	    } else {
1133 		LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1134 	    }
1135 	    if (nfsd->nd_mrep) {
1136 		wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1137 		owp = NULL;
1138 		wp = LIST_FIRST(wpp);
1139 		while (wp &&
1140 		    memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) {
1141 		    owp = wp;
1142 		    wp = LIST_NEXT(wp, nd_hash);
1143 		}
1144 		while (wp && wp->nd_off < nfsd->nd_off &&
1145 		    !memcmp(&nfsd->nd_fh, &wp->nd_fh, NFSX_V3FH)) {
1146 		    owp = wp;
1147 		    wp = LIST_NEXT(wp, nd_hash);
1148 		}
1149 		if (owp) {
1150 		    LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1151 
1152 		    /*
1153 		     * Search the hash list for overlapping entries and
1154 		     * coalesce.
1155 		     */
1156 		    for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1157 			wp = LIST_NEXT(nfsd, nd_hash);
1158 			if (nfsrv_samecred(owp->nd_cr, nfsd->nd_cr))
1159 			    nfsrvw_coalesce(owp, nfsd);
1160 		    }
1161 		} else {
1162 		    LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1163 		}
1164 	    }
1165 	    splx(s);
1166 	}
1167 
1168 	/*
1169 	 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1170 	 * and generate the associated reply mbuf list(s).
1171 	 */
1172 loop1:
1173 	getmicrotime(&now);
1174 	cur_usec = (u_quad_t)now.tv_sec * 1000000 + (u_quad_t)now.tv_usec;
1175 	s = splsoftclock();
1176 	for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) {
1177 		owp = LIST_NEXT(nfsd, nd_tq);
1178 		if (nfsd->nd_time > cur_usec)
1179 		    break;
1180 		if (nfsd->nd_mreq)
1181 		    continue;
1182 		LIST_REMOVE(nfsd, nd_tq);
1183 		LIST_REMOVE(nfsd, nd_hash);
1184 		splx(s);
1185 		mrep = nfsd->nd_mrep;
1186 		nfsd->nd_mrep = NULL;
1187 		cred = nfsd->nd_cr;
1188 		v3 = (nfsd->nd_flag & ND_NFSV3);
1189 		forat_ret = aftat_ret = 1;
1190 		error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1191 		    nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1192 		    FALSE);
1193 		if (!error) {
1194 		    if (v3)
1195 			forat_ret = VOP_GETATTR(vp, &forat, cred, lwp);
1196 		    if (vp->v_type != VREG) {
1197 			if (v3)
1198 			    error = EINVAL;
1199 			else
1200 			    error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1201 		    }
1202 		} else
1203 		    vp = NULL;
1204 		if (!error) {
1205 		    nqsrv_getl(vp, ND_WRITE);
1206 		    error = nfsrv_access(vp, VWRITE, cred, rdonly, lwp, 1);
1207 		}
1208 
1209 		if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1210 		    ioflags = IO_NODELOCKED;
1211 		else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1212 		    ioflags = (IO_SYNC | IO_NODELOCKED);
1213 		else
1214 		    ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1215 		uiop->uio_rw = UIO_WRITE;
1216 		uiop->uio_offset = nfsd->nd_off;
1217 		uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1218 		UIO_SETUP_SYSSPACE(uiop);
1219 		if (uiop->uio_resid > 0) {
1220 		    mp = mrep;
1221 		    i = 0;
1222 		    while (mp) {
1223 			if (mp->m_len > 0)
1224 			    i++;
1225 			mp = mp->m_next;
1226 		    }
1227 		    uiop->uio_iovcnt = i;
1228 		    iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
1229 		    uiop->uio_iov = ivp = iov;
1230 		    mp = mrep;
1231 		    while (mp) {
1232 			if (mp->m_len > 0) {
1233 			    ivp->iov_base = mtod(mp, caddr_t);
1234 			    ivp->iov_len = mp->m_len;
1235 			    ivp++;
1236 			}
1237 			mp = mp->m_next;
1238 		    }
1239 		    if (!error) {
1240 			if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) {
1241 			    VOP_UNLOCK(vp, 0);
1242 			    vn_start_write(NULL, &mntp, V_WAIT);
1243 			    vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1244 			}
1245 			if (!error) {
1246 			    error = VOP_WRITE(vp, uiop, ioflags, cred);
1247 			    nfsstats.srvvop_writes++;
1248 			    vn_finished_write(mntp, 0);
1249 			}
1250 		    }
1251 		    free((caddr_t)iov, M_TEMP);
1252 		}
1253 		m_freem(mrep);
1254 		if (vp) {
1255 		    aftat_ret = VOP_GETATTR(vp, &va, cred, lwp);
1256 		    vput(vp);
1257 		}
1258 
1259 		/*
1260 		 * Loop around generating replies for all write rpcs that have
1261 		 * now been completed.
1262 		 */
1263 		swp = nfsd;
1264 		do {
1265 		    if (error) {
1266 			nfsm_writereply(NFSX_WCCDATA(v3), v3);
1267 			if (v3) {
1268 			    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1269 			}
1270 		    } else {
1271 			nfsm_writereply(NFSX_PREOPATTR(v3) +
1272 			    NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1273 			    NFSX_WRITEVERF(v3), v3);
1274 			if (v3) {
1275 			    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1276 			    nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1277 			    *tl++ = txdr_unsigned(nfsd->nd_len);
1278 			    *tl++ = txdr_unsigned(swp->nd_stable);
1279 			    /*
1280 			     * Actually, there is no need to txdr these fields,
1281 			     * but it may make the values more human readable,
1282 			     * for debugging purposes.
1283 			     */
1284 			    *tl++ = txdr_unsigned(boottime.tv_sec);
1285 			    *tl = txdr_unsigned(boottime.tv_usec);
1286 			} else {
1287 			    nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1288 			    nfsm_srvfillattr(&va, fp);
1289 			}
1290 		    }
1291 		    nfsd->nd_mreq = mreq;
1292 		    if (nfsd->nd_mrep)
1293 			panic("nfsrv_write: nd_mrep not free");
1294 
1295 		    /*
1296 		     * Done. Put it at the head of the timer queue so that
1297 		     * the final phase can return the reply.
1298 		     */
1299 		    s = splsoftclock();
1300 		    if (nfsd != swp) {
1301 			nfsd->nd_time = 0;
1302 			LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1303 		    }
1304 		    nfsd = LIST_FIRST(&swp->nd_coalesce);
1305 		    if (nfsd) {
1306 			LIST_REMOVE(nfsd, nd_tq);
1307 		    }
1308 		    splx(s);
1309 		} while (nfsd);
1310 		s = splsoftclock();
1311 		swp->nd_time = 0;
1312 		LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1313 		splx(s);
1314 		goto loop1;
1315 	}
1316 	splx(s);
1317 
1318 	/*
1319 	 * Search for a reply to return.
1320 	 */
1321 	s = splsoftclock();
1322 	LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) {
1323 		if (nfsd->nd_mreq) {
1324 		    LIST_REMOVE(nfsd, nd_tq);
1325 		    *mrq = nfsd->nd_mreq;
1326 		    *ndp = nfsd;
1327 		    break;
1328 		}
1329 	}
1330 	splx(s);
1331 	return (0);
1332 }
1333 
1334 /*
1335  * Coalesce the write request nfsd into owp. To do this we must:
1336  * - remove nfsd from the queues
1337  * - merge nfsd->nd_mrep into owp->nd_mrep
1338  * - update the nd_eoff and nd_stable for owp
1339  * - put nfsd on owp's nd_coalesce list
1340  * NB: Must be called at splsoftclock().
1341  */
1342 void
1343 nfsrvw_coalesce(owp, nfsd)
1344         struct nfsrv_descript *owp;
1345         struct nfsrv_descript *nfsd;
1346 {
1347         int overlap;
1348         struct mbuf *mp;
1349 	struct nfsrv_descript *m;
1350 
1351         LIST_REMOVE(nfsd, nd_hash);
1352         LIST_REMOVE(nfsd, nd_tq);
1353         if (owp->nd_eoff < nfsd->nd_eoff) {
1354             overlap = owp->nd_eoff - nfsd->nd_off;
1355             if (overlap < 0)
1356                 panic("nfsrv_coalesce: bad off");
1357             if (overlap > 0)
1358                 m_adj(nfsd->nd_mrep, overlap);
1359             mp = owp->nd_mrep;
1360             while (mp->m_next)
1361                 mp = mp->m_next;
1362             mp->m_next = nfsd->nd_mrep;
1363             owp->nd_eoff = nfsd->nd_eoff;
1364         } else
1365             m_freem(nfsd->nd_mrep);
1366         nfsd->nd_mrep = NULL;
1367         if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1368             owp->nd_stable = NFSV3WRITE_FILESYNC;
1369         else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1370             owp->nd_stable == NFSV3WRITE_UNSTABLE)
1371             owp->nd_stable = NFSV3WRITE_DATASYNC;
1372         LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1373  	/*
1374  	 * nfsd might hold coalesce elements! Move them to owp.
1375  	 * Otherwise, requests may be lost and clients will be stuck.
1376  	 */
1377 	while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) {
1378 		LIST_REMOVE(m, nd_tq);
1379 		LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1380 	}
1381 }
1382 
1383 /*
1384  * nfs create service
1385  * now does a truncate to 0 length via. setattr if it already exists
1386  */
1387 int
1388 nfsrv_create(nfsd, slp, lwp, mrq)
1389 	struct nfsrv_descript *nfsd;
1390 	struct nfssvc_sock *slp;
1391 	struct lwp *lwp;
1392 	struct mbuf **mrq;
1393 {
1394 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1395 	struct mbuf *nam = nfsd->nd_nam;
1396 	caddr_t dpos = nfsd->nd_dpos;
1397 	kauth_cred_t cred = nfsd->nd_cr;
1398 	struct nfs_fattr *fp;
1399 	struct vattr va, dirfor, diraft;
1400 	struct nfsv2_sattr *sp;
1401 	u_int32_t *tl;
1402 	struct nameidata nd;
1403 	caddr_t cp;
1404 	int32_t t1;
1405 	caddr_t bpos;
1406 	int error = 0, cache = 0, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1407 	int rdev = 0;
1408 	int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1409 	char *cp2;
1410 	struct mbuf *mb, *mreq;
1411 	struct vnode *vp = NULL, *dirp = NULL;
1412 	nfsfh_t nfh;
1413 	fhandle_t *fhp;
1414 	u_quad_t frev, tempsize;
1415 	u_char cverf[NFSX_V3CREATEVERF];
1416 	struct mount *mp = NULL;
1417 
1418 	nd.ni_cnd.cn_nameiop = 0;
1419 	fhp = &nfh.fh_generic;
1420 	nfsm_srvmtofh(fhp);
1421 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1422 		return (ESTALE);
1423 	vn_start_write(NULL, &mp, V_WAIT);
1424 	nfsm_srvnamesiz(len);
1425 	nd.ni_cnd.cn_cred = cred;
1426 	nd.ni_cnd.cn_nameiop = CREATE;
1427 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1428 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1429 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1430 	if (dirp) {
1431 		if (v3)
1432 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
1433 		else {
1434 			vrele(dirp);
1435 			dirp = (struct vnode *)0;
1436 		}
1437 	}
1438 	if (error) {
1439 		nfsm_reply(NFSX_WCCDATA(v3));
1440 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1441 		if (dirp)
1442 			vrele(dirp);
1443 		vn_finished_write(mp, 0);
1444 		return (0);
1445 	}
1446 	VATTR_NULL(&va);
1447 	if (v3) {
1448 		va.va_mode = 0;
1449 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1450 		how = fxdr_unsigned(int, *tl);
1451 		switch (how) {
1452 		case NFSV3CREATE_GUARDED:
1453 			if (nd.ni_vp) {
1454 				error = EEXIST;
1455 				break;
1456 			}
1457 		case NFSV3CREATE_UNCHECKED:
1458 			nfsm_srvsattr(&va);
1459 			break;
1460 		case NFSV3CREATE_EXCLUSIVE:
1461 			nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1462 			memcpy(cverf, cp, NFSX_V3CREATEVERF);
1463 			exclusive_flag = 1;
1464 			break;
1465 		};
1466 		va.va_type = VREG;
1467 	} else {
1468 		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1469 		va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1470 		if (va.va_type == VNON)
1471 			va.va_type = VREG;
1472 		va.va_mode = nfstov_mode(sp->sa_mode);
1473 		switch (va.va_type) {
1474 		case VREG:
1475 			tsize = fxdr_unsigned(int32_t, sp->sa_size);
1476 			if (tsize != -1)
1477 				va.va_size = (u_quad_t)tsize;
1478 			break;
1479 		case VCHR:
1480 		case VBLK:
1481 		case VFIFO:
1482 			rdev = fxdr_unsigned(int32_t, sp->sa_size);
1483 			break;
1484 		default:
1485 			break;
1486 		};
1487 	}
1488 
1489 	/*
1490 	 * Iff doesn't exist, create it
1491 	 * otherwise just truncate to 0 length
1492 	 *   should I set the mode too ??
1493 	 */
1494 	if (nd.ni_vp == NULL) {
1495 		if (va.va_type == VREG || va.va_type == VSOCK) {
1496 			nqsrv_getl(nd.ni_dvp, ND_WRITE);
1497 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1498 			if (!error) {
1499 				if (exclusive_flag) {
1500 					exclusive_flag = 0;
1501 					VATTR_NULL(&va);
1502 					/*
1503 					 * XXX
1504 					 * assuming NFSX_V3CREATEVERF
1505 					 * == sizeof(nfstime3)
1506 					 */
1507 					fxdr_nfsv3time(cverf, &va.va_atime);
1508 					error = VOP_SETATTR(nd.ni_vp, &va, cred,
1509 						lwp);
1510 				}
1511 			}
1512 		} else if (va.va_type == VCHR || va.va_type == VBLK ||
1513 			va.va_type == VFIFO) {
1514 			if (va.va_type == VCHR && rdev == 0xffffffff)
1515 				va.va_type = VFIFO;
1516 			if (va.va_type != VFIFO &&
1517 			    (error = kauth_authorize_generic(cred,
1518 			    KAUTH_GENERIC_ISSUSER, (u_short *)0))) {
1519 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1520 				vput(nd.ni_dvp);
1521 				nfsm_reply(0);
1522 				vn_finished_write(mp, 0);
1523 				return (error);
1524 			} else
1525 				va.va_rdev = (dev_t)rdev;
1526 			nqsrv_getl(nd.ni_dvp, ND_WRITE);
1527 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1528 			    &va);
1529 			if (error) {
1530 				nfsm_reply(0);
1531 			}
1532 			if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1533 				vrele(nd.ni_dvp);
1534 				vput(nd.ni_vp);
1535 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1536 				error = EINVAL;
1537 				nfsm_reply(0);
1538 			}
1539 		} else {
1540 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1541 			vput(nd.ni_dvp);
1542 			error = ENXIO;
1543 		}
1544 		vp = nd.ni_vp;
1545 	} else {
1546 		vp = nd.ni_vp;
1547 		if (nd.ni_dvp == vp)
1548 			vrele(nd.ni_dvp);
1549 		else
1550 			vput(nd.ni_dvp);
1551 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1552 		if (!error && va.va_size != -1) {
1553 			error = nfsrv_access(vp, VWRITE, cred,
1554 			    (nd.ni_cnd.cn_flags & RDONLY), lwp, 0);
1555 			if (!error) {
1556 				nqsrv_getl(vp, ND_WRITE);
1557 				tempsize = va.va_size;
1558 				VATTR_NULL(&va);
1559 				va.va_size = tempsize;
1560 				error = VOP_SETATTR(vp, &va, cred, lwp);
1561 			}
1562 		}
1563 		if (error)
1564 			vput(vp);
1565 	}
1566 	if (!error) {
1567 		error = vfs_composefh(vp, fhp);
1568 		if (!error)
1569 			error = VOP_GETATTR(vp, &va, cred, lwp);
1570 		vput(vp);
1571 	}
1572 	if (v3) {
1573 		if (exclusive_flag && !error) {
1574 			/*
1575 			 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3)
1576 			 */
1577 			char oldverf[NFSX_V3CREATEVERF];
1578 
1579 			txdr_nfsv3time(&va.va_atime, oldverf);
1580 			if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF))
1581 				error = EEXIST;
1582 		}
1583 		if (dirp) {
1584 			diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
1585 			vrele(dirp);
1586 		}
1587 	}
1588 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1589 	if (v3) {
1590 		if (!error) {
1591 			nfsm_srvpostop_fh(fhp);
1592 			nfsm_srvpostop_attr(0, &va);
1593 		}
1594 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1595 	} else {
1596 		nfsm_srvfhtom(fhp, v3);
1597 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1598 		nfsm_srvfillattr(&va, fp);
1599 	}
1600 	vn_finished_write(mp, 0);
1601 	return (0);
1602 nfsmout:
1603 	if (dirp)
1604 		vrele(dirp);
1605 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1606 	if (nd.ni_dvp == nd.ni_vp)
1607 		vrele(nd.ni_dvp);
1608 	else
1609 		vput(nd.ni_dvp);
1610 	if (nd.ni_vp)
1611 		vput(nd.ni_vp);
1612 	vn_finished_write(mp, 0);
1613 	return (error);
1614 }
1615 
1616 /*
1617  * nfs v3 mknod service
1618  */
1619 int
1620 nfsrv_mknod(nfsd, slp, lwp, mrq)
1621 	struct nfsrv_descript *nfsd;
1622 	struct nfssvc_sock *slp;
1623 	struct lwp *lwp;
1624 	struct mbuf **mrq;
1625 {
1626 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1627 	struct mbuf *nam = nfsd->nd_nam;
1628 	caddr_t dpos = nfsd->nd_dpos;
1629 	kauth_cred_t cred = nfsd->nd_cr;
1630 	struct vattr va, dirfor, diraft;
1631 	u_int32_t *tl;
1632 	struct nameidata nd;
1633 	int32_t t1;
1634 	caddr_t bpos;
1635 	int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1636 	u_int32_t major, minor;
1637 	enum vtype vtyp;
1638 	char *cp2;
1639 	struct mbuf *mb, *mreq;
1640 	struct vnode *vp, *dirp = (struct vnode *)0;
1641 	nfsfh_t nfh;
1642 	fhandle_t *fhp;
1643 	u_quad_t frev;
1644 	struct mount *mp = NULL;
1645 
1646 	nd.ni_cnd.cn_nameiop = 0;
1647 	fhp = &nfh.fh_generic;
1648 	nfsm_srvmtofh(fhp);
1649 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1650 		return (ESTALE);
1651 	vn_start_write(NULL, &mp, V_WAIT);
1652 	nfsm_srvnamesiz(len);
1653 	nd.ni_cnd.cn_cred = cred;
1654 	nd.ni_cnd.cn_nameiop = CREATE;
1655 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1656 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1657 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1658 	if (dirp)
1659 		dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
1660 	if (error) {
1661 		nfsm_reply(NFSX_WCCDATA(1));
1662 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1663 		if (dirp)
1664 			vrele(dirp);
1665 		vn_finished_write(mp, 0);
1666 		return (0);
1667 	}
1668 	nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1669 	vtyp = nfsv3tov_type(*tl);
1670 	if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1671 		error = NFSERR_BADTYPE;
1672 		goto abort;
1673 	}
1674 	VATTR_NULL(&va);
1675 	va.va_mode = 0;
1676 	nfsm_srvsattr(&va);
1677 	if (vtyp == VCHR || vtyp == VBLK) {
1678 		dev_t rdev;
1679 
1680 		nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1681 		major = fxdr_unsigned(u_int32_t, *tl++);
1682 		minor = fxdr_unsigned(u_int32_t, *tl);
1683 		rdev = makedev(major, minor);
1684 		if (major(rdev) != major || minor(rdev) != minor) {
1685 			error = EINVAL;
1686 			goto abort;
1687 		}
1688 		va.va_rdev = rdev;
1689 	}
1690 
1691 	/*
1692 	 * Iff doesn't exist, create it.
1693 	 */
1694 	if (nd.ni_vp) {
1695 		error = EEXIST;
1696 abort:
1697 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1698 		if (nd.ni_dvp == nd.ni_vp)
1699 			vrele(nd.ni_dvp);
1700 		else
1701 			vput(nd.ni_dvp);
1702 		if (nd.ni_vp)
1703 			vput(nd.ni_vp);
1704 		goto out;
1705 	}
1706 	va.va_type = vtyp;
1707 	if (vtyp == VSOCK) {
1708 		nqsrv_getl(nd.ni_dvp, ND_WRITE);
1709 		error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1710 	} else {
1711 		if (va.va_type != VFIFO &&
1712 		    (error = kauth_authorize_generic(cred,
1713 		    KAUTH_GENERIC_ISSUSER, (u_short *)0))) {
1714 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1715 			vput(nd.ni_dvp);
1716 			goto out;
1717 		}
1718 		nqsrv_getl(nd.ni_dvp, ND_WRITE);
1719 		error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1720 		if (error)
1721 			goto out;
1722 		if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1723 			vput(nd.ni_vp);
1724 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1725 			error = EINVAL;
1726 		}
1727 	}
1728 out:
1729 	vp = nd.ni_vp;
1730 	if (!error) {
1731 		error = vfs_composefh(vp, fhp);
1732 		if (!error)
1733 			error = VOP_GETATTR(vp, &va, cred, lwp);
1734 		vput(vp);
1735 	}
1736 	if (dirp) {
1737 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
1738 		vrele(dirp);
1739 	}
1740 	nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
1741 	if (!error) {
1742 		nfsm_srvpostop_fh(fhp);
1743 		nfsm_srvpostop_attr(0, &va);
1744 	}
1745 	nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1746 	vn_finished_write(mp, 0);
1747 	return (0);
1748 nfsmout:
1749 	if (dirp)
1750 		vrele(dirp);
1751 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1752 	if (nd.ni_dvp == nd.ni_vp)
1753 		vrele(nd.ni_dvp);
1754 	else
1755 		vput(nd.ni_dvp);
1756 	if (nd.ni_vp)
1757 		vput(nd.ni_vp);
1758 	vn_finished_write(mp, 0);
1759 	return (error);
1760 }
1761 
1762 /*
1763  * nfs remove service
1764  */
1765 int
1766 nfsrv_remove(nfsd, slp, lwp, mrq)
1767 	struct nfsrv_descript *nfsd;
1768 	struct nfssvc_sock *slp;
1769 	struct lwp *lwp;
1770 	struct mbuf **mrq;
1771 {
1772 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1773 	struct mbuf *nam = nfsd->nd_nam;
1774 	caddr_t dpos = nfsd->nd_dpos;
1775 	kauth_cred_t cred = nfsd->nd_cr;
1776 	struct nameidata nd;
1777 	u_int32_t *tl;
1778 	int32_t t1;
1779 	caddr_t bpos;
1780 	int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
1781 	int v3 = (nfsd->nd_flag & ND_NFSV3);
1782 	char *cp2;
1783 	struct mbuf *mb, *mreq;
1784 	struct vnode *vp, *dirp;
1785 	struct vattr dirfor, diraft;
1786 	nfsfh_t nfh;
1787 	fhandle_t *fhp;
1788 	u_quad_t frev;
1789 	struct mount *mp = NULL;
1790 
1791 #ifndef nolint
1792 	vp = (struct vnode *)0;
1793 #endif
1794 	fhp = &nfh.fh_generic;
1795 	nfsm_srvmtofh(fhp);
1796 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1797 		return (ESTALE);
1798 	vn_start_write(NULL, &mp, V_WAIT);
1799 	nfsm_srvnamesiz(len);
1800 	nd.ni_cnd.cn_cred = cred;
1801 	nd.ni_cnd.cn_nameiop = DELETE;
1802 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1803 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1804 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1805 	if (dirp) {
1806 		if (v3)
1807 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
1808 		else
1809 			vrele(dirp);
1810 	}
1811 	if (!error) {
1812 		vp = nd.ni_vp;
1813 		if (vp->v_type == VDIR &&
1814 		    (error = kauth_authorize_generic(cred,
1815 		    KAUTH_GENERIC_ISSUSER, (u_short *)0)) != 0)
1816 			goto out;
1817 		/*
1818 		 * The root of a mounted filesystem cannot be deleted.
1819 		 */
1820 		if (vp->v_flag & VROOT) {
1821 			error = EBUSY;
1822 			goto out;
1823 		}
1824 out:
1825 		if (!error) {
1826 			nqsrv_getl(nd.ni_dvp, ND_WRITE);
1827 			nqsrv_getl(vp, ND_WRITE);
1828 			error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1829 		} else {
1830 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1831 			if (nd.ni_dvp == vp)
1832 				vrele(nd.ni_dvp);
1833 			else
1834 				vput(nd.ni_dvp);
1835 			vput(vp);
1836 		}
1837 	}
1838 	if (dirp && v3) {
1839 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
1840 		vrele(dirp);
1841 	}
1842 	nfsm_reply(NFSX_WCCDATA(v3));
1843 	if (v3) {
1844 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1845 		vn_finished_write(mp, 0);
1846 		return (0);
1847 	}
1848 	vn_finished_write(mp, 0);
1849 	nfsm_srvdone;
1850 }
1851 
1852 /*
1853  * nfs rename service
1854  */
1855 int
1856 nfsrv_rename(nfsd, slp, lwp, mrq)
1857 	struct nfsrv_descript *nfsd;
1858 	struct nfssvc_sock *slp;
1859 	struct lwp *lwp;
1860 	struct mbuf **mrq;
1861 {
1862 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1863 	struct mbuf *nam = nfsd->nd_nam;
1864 	caddr_t dpos = nfsd->nd_dpos;
1865 	kauth_cred_t cred = nfsd->nd_cr;
1866 	u_int32_t *tl;
1867 	int32_t t1;
1868 	caddr_t bpos;
1869 	int error = 0, cache = 0, fdirfor_ret = 1, fdiraft_ret = 1;
1870 	uint32_t len, len2;
1871 	int tdirfor_ret = 1, tdiraft_ret = 1;
1872 	int v3 = (nfsd->nd_flag & ND_NFSV3);
1873 	char *cp2;
1874 	struct mbuf *mb, *mreq;
1875 	struct nameidata fromnd, tond;
1876 	struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
1877 	struct vnode *tdirp = (struct vnode *)0;
1878 	struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1879 	nfsfh_t fnfh, tnfh;
1880 	fhandle_t *ffhp, *tfhp;
1881 	u_quad_t frev;
1882 	uid_t saved_uid;
1883 	struct mount *mp = NULL;
1884 
1885 #ifndef nolint
1886 	fvp = (struct vnode *)0;
1887 #endif
1888 	ffhp = &fnfh.fh_generic;
1889 	tfhp = &tnfh.fh_generic;
1890 	fromnd.ni_cnd.cn_nameiop = 0;
1891 	tond.ni_cnd.cn_nameiop = 0;
1892 	nfsm_srvmtofh(ffhp);
1893 	if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL)
1894 		return (ESTALE);
1895 	vn_start_write(NULL, &mp, V_WAIT);
1896 	nfsm_srvnamesiz(len);
1897 	/*
1898 	 * Remember our original uid so that we can reset cr_uid before
1899 	 * the second nfs_namei() call, in case it is remapped.
1900 	 */
1901 	saved_uid = kauth_cred_geteuid(cred);
1902 	fromnd.ni_cnd.cn_cred = cred;
1903 	fromnd.ni_cnd.cn_nameiop = DELETE;
1904 	fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
1905 	error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
1906 		&dpos, &fdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1907 	if (fdirp) {
1908 		if (v3)
1909 			fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred, lwp);
1910 		else {
1911 			vrele(fdirp);
1912 			fdirp = (struct vnode *)0;
1913 		}
1914 	}
1915 	if (error) {
1916 		nfsm_reply(2 * NFSX_WCCDATA(v3));
1917 		nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1918 		nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1919 		if (fdirp)
1920 			vrele(fdirp);
1921 		vn_finished_write(mp, 0);
1922 		return (0);
1923 	}
1924 	fvp = fromnd.ni_vp;
1925 	nfsm_srvmtofh(tfhp);
1926 	if (v3) {
1927 		nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
1928 		len2 = fxdr_unsigned(uint32_t, *tl);
1929 		/* len2 will be checked by nfs_namei */
1930 	}
1931 	else {
1932 		/* NFSv2 */
1933 		nfsm_strsiz(len2, NFS_MAXNAMLEN);
1934 	}
1935 	kauth_cred_seteuid(cred, saved_uid);
1936 	tond.ni_cnd.cn_cred = cred;
1937 	tond.ni_cnd.cn_nameiop = RENAME;
1938 	tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
1939 	error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
1940 		&dpos, &tdirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1941 	if (tdirp) {
1942 		if (v3)
1943 			tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred, lwp);
1944 		else {
1945 			vrele(tdirp);
1946 			tdirp = (struct vnode *)0;
1947 		}
1948 	}
1949 	if (error) {
1950 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1951 		vrele(fromnd.ni_dvp);
1952 		vrele(fvp);
1953 		goto out1;
1954 	}
1955 	tdvp = tond.ni_dvp;
1956 	tvp = tond.ni_vp;
1957 	if (tvp != NULL) {
1958 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1959 			if (v3)
1960 				error = EEXIST;
1961 			else
1962 				error = EISDIR;
1963 			goto out;
1964 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1965 			if (v3)
1966 				error = EEXIST;
1967 			else
1968 				error = ENOTDIR;
1969 			goto out;
1970 		}
1971 		if (tvp->v_type == VDIR && tvp->v_mountedhere) {
1972 			if (v3)
1973 				error = EXDEV;
1974 			else
1975 				error = ENOTEMPTY;
1976 			goto out;
1977 		}
1978 	}
1979 	if (fvp->v_type == VDIR && fvp->v_mountedhere) {
1980 		if (v3)
1981 			error = EXDEV;
1982 		else
1983 			error = ENOTEMPTY;
1984 		goto out;
1985 	}
1986 	if (fvp->v_mount != tdvp->v_mount) {
1987 		if (v3)
1988 			error = EXDEV;
1989 		else
1990 			error = ENOTEMPTY;
1991 		goto out;
1992 	}
1993 	if (fvp == tdvp) {
1994 		if (v3)
1995 			error = EINVAL;
1996 		else
1997 			error = ENOTEMPTY;
1998 	}
1999 	/*
2000 	 * If source is the same as the destination (that is the
2001 	 * same vnode with the same name in the same directory),
2002 	 * then there is nothing to do.
2003 	 */
2004 	if (fvp == tvp && fromnd.ni_dvp == tdvp &&
2005 	    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
2006 	    !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
2007 	      fromnd.ni_cnd.cn_namelen))
2008 		error = -1;
2009 out:
2010 	if (!error) {
2011 		nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
2012 		nqsrv_getl(tdvp, ND_WRITE);
2013 		if (tvp) {
2014 			nqsrv_getl(tvp, ND_WRITE);
2015 		}
2016 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2017 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2018 	} else {
2019 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2020 		if (tdvp == tvp)
2021 			vrele(tdvp);
2022 		else
2023 			vput(tdvp);
2024 		if (tvp)
2025 			vput(tvp);
2026 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2027 		vrele(fromnd.ni_dvp);
2028 		vrele(fvp);
2029 		if (error == -1)
2030 			error = 0;
2031 	}
2032 	vrele(tond.ni_startdir);
2033 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2034 out1:
2035 	if (fdirp) {
2036 		fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, lwp);
2037 		vrele(fdirp);
2038 	}
2039 	if (tdirp) {
2040 		tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, lwp);
2041 		vrele(tdirp);
2042 	}
2043 	vrele(fromnd.ni_startdir);
2044 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2045 	nfsm_reply(2 * NFSX_WCCDATA(v3));
2046 	if (v3) {
2047 		nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
2048 		nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
2049 	}
2050 	vn_finished_write(mp, 0);
2051 	return (0);
2052 
2053 nfsmout:
2054 	if (fdirp)
2055 		vrele(fdirp);
2056 #ifdef notdef
2057 	if (tdirp)
2058 		vrele(tdirp);
2059 #endif
2060 	if (tond.ni_cnd.cn_nameiop) {
2061 		vrele(tond.ni_startdir);
2062 		PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2063 	}
2064 	if (fromnd.ni_cnd.cn_nameiop) {
2065 		vrele(fromnd.ni_startdir);
2066 		PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2067 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2068 		vrele(fromnd.ni_dvp);
2069 		vrele(fvp);
2070 	}
2071 	return (error);
2072 }
2073 
2074 /*
2075  * nfs link service
2076  */
2077 int
2078 nfsrv_link(nfsd, slp, lwp, mrq)
2079 	struct nfsrv_descript *nfsd;
2080 	struct nfssvc_sock *slp;
2081 	struct lwp *lwp;
2082 	struct mbuf **mrq;
2083 {
2084 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2085 	struct mbuf *nam = nfsd->nd_nam;
2086 	caddr_t dpos = nfsd->nd_dpos;
2087 	kauth_cred_t cred = nfsd->nd_cr;
2088 	struct nameidata nd;
2089 	u_int32_t *tl;
2090 	int32_t t1;
2091 	caddr_t bpos;
2092 	int error = 0, rdonly, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2093 	int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
2094 	char *cp2;
2095 	struct mbuf *mb, *mreq;
2096 	struct vnode *vp, *xp, *dirp = (struct vnode *)0;
2097 	struct vattr dirfor, diraft, at;
2098 	nfsfh_t nfh, dnfh;
2099 	fhandle_t *fhp, *dfhp;
2100 	u_quad_t frev;
2101 	struct mount *mp = NULL;
2102 
2103 	fhp = &nfh.fh_generic;
2104 	dfhp = &dnfh.fh_generic;
2105 	nfsm_srvmtofh(fhp);
2106 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2107 		return (ESTALE);
2108 	vn_start_write(NULL, &mp, V_WAIT);
2109 	nfsm_srvmtofh(dfhp);
2110 	nfsm_srvnamesiz(len);
2111 	error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
2112 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2113 	if (error) {
2114 		nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2115 		nfsm_srvpostop_attr(getret, &at);
2116 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2117 		vn_finished_write(mp, 0);
2118 		return (0);
2119 	}
2120 	if (vp->v_type == VDIR && (error = kauth_authorize_generic(cred,
2121 	    KAUTH_GENERIC_ISSUSER, (u_short *)0)) != 0)
2122 		goto out1;
2123 	nd.ni_cnd.cn_cred = cred;
2124 	nd.ni_cnd.cn_nameiop = CREATE;
2125 	nd.ni_cnd.cn_flags = LOCKPARENT;
2126 	error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
2127 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2128 	if (dirp) {
2129 		if (v3)
2130 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2131 		else {
2132 			vrele(dirp);
2133 			dirp = (struct vnode *)0;
2134 		}
2135 	}
2136 	if (error)
2137 		goto out1;
2138 	xp = nd.ni_vp;
2139 	if (xp != NULL) {
2140 		error = EEXIST;
2141 		goto out;
2142 	}
2143 	xp = nd.ni_dvp;
2144 	if (vp->v_mount != xp->v_mount)
2145 		error = EXDEV;
2146 out:
2147 	if (!error) {
2148 		nqsrv_getl(vp, ND_WRITE);
2149 		nqsrv_getl(xp, ND_WRITE);
2150 		error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2151 	} else {
2152 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2153 		if (nd.ni_dvp == nd.ni_vp)
2154 			vrele(nd.ni_dvp);
2155 		else
2156 			vput(nd.ni_dvp);
2157 		if (nd.ni_vp)
2158 			vrele(nd.ni_vp);
2159 	}
2160 out1:
2161 	if (v3)
2162 		getret = VOP_GETATTR(vp, &at, cred, lwp);
2163 	if (dirp) {
2164 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2165 		vrele(dirp);
2166 	}
2167 	vrele(vp);
2168 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2169 	if (v3) {
2170 		nfsm_srvpostop_attr(getret, &at);
2171 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2172 		vn_finished_write(mp, 0);
2173 		return (0);
2174 	}
2175 	vn_finished_write(mp, 0);
2176 	nfsm_srvdone;
2177 }
2178 
2179 /*
2180  * nfs symbolic link service
2181  */
2182 int
2183 nfsrv_symlink(nfsd, slp, lwp, mrq)
2184 	struct nfsrv_descript *nfsd;
2185 	struct nfssvc_sock *slp;
2186 	struct lwp *lwp;
2187 	struct mbuf **mrq;
2188 {
2189 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2190 	struct mbuf *nam = nfsd->nd_nam;
2191 	caddr_t dpos = nfsd->nd_dpos;
2192 	kauth_cred_t cred = nfsd->nd_cr;
2193 	struct vattr va, dirfor, diraft;
2194 	struct nameidata nd;
2195 	u_int32_t *tl;
2196 	int32_t t1;
2197 	struct nfsv2_sattr *sp;
2198 	char *bpos, *pathcp = NULL, *cp2;
2199 	struct uio io;
2200 	struct iovec iv;
2201 	int error = 0, cache = 0, dirfor_ret = 1, diraft_ret = 1;
2202 	uint32_t len, len2;
2203 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2204 	struct mbuf *mb, *mreq;
2205 	struct vnode *dirp = (struct vnode *)0;
2206 	nfsfh_t nfh;
2207 	fhandle_t *fhp;
2208 	u_quad_t frev;
2209 	struct mount *mp = NULL;
2210 
2211 	nd.ni_cnd.cn_nameiop = 0;
2212 	fhp = &nfh.fh_generic;
2213 	nfsm_srvmtofh(fhp);
2214 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2215 		return (ESTALE);
2216 	vn_start_write(NULL, &mp, V_WAIT);
2217 	nfsm_srvnamesiz(len);
2218 	nd.ni_cnd.cn_cred = cred;
2219 	nd.ni_cnd.cn_nameiop = CREATE;
2220 	nd.ni_cnd.cn_flags = LOCKPARENT;
2221 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2222 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2223 	if (dirp) {
2224 		if (v3)
2225 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2226 		else {
2227 			vrele(dirp);
2228 			dirp = (struct vnode *)0;
2229 		}
2230 	}
2231 	if (error)
2232 		goto out;
2233 	VATTR_NULL(&va);
2234 	va.va_type = VLNK;
2235 	if (v3) {
2236 		va.va_mode = 0;
2237 		nfsm_srvsattr(&va);
2238 		nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
2239 		len2 = fxdr_unsigned(uint32_t, *tl);
2240 		if (len2 > PATH_MAX) {
2241 			/* XXX should check _PC_NO_TRUNC */
2242 			error = ENAMETOOLONG;
2243 			goto abortop;
2244 		}
2245 	}
2246 	else {
2247 		/* NFSv2 */
2248 		nfsm_strsiz(len2, NFS_MAXPATHLEN);
2249 	}
2250 	pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK);
2251 	iv.iov_base = pathcp;
2252 	iv.iov_len = len2;
2253 	io.uio_resid = len2;
2254 	io.uio_offset = 0;
2255 	io.uio_iov = &iv;
2256 	io.uio_iovcnt = 1;
2257 	io.uio_rw = UIO_READ;
2258 	UIO_SETUP_SYSSPACE(&io);
2259 	nfsm_mtouio(&io, len2);
2260 	if (!v3) {
2261 		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2262 		va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2263 	}
2264 	*(pathcp + len2) = '\0';
2265 	if (nd.ni_vp) {
2266 		error = EEXIST;
2267 abortop:
2268 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2269 		if (nd.ni_dvp == nd.ni_vp)
2270 			vrele(nd.ni_dvp);
2271 		else
2272 			vput(nd.ni_dvp);
2273 		if (nd.ni_vp)
2274 			vrele(nd.ni_vp);
2275 		goto out;
2276 	}
2277 	nqsrv_getl(nd.ni_dvp, ND_WRITE);
2278 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2279 	if (!error) {
2280 	    if (v3) {
2281 		error = vfs_composefh(nd.ni_vp, fhp);
2282 		if (!error)
2283 		    error = VOP_GETATTR(nd.ni_vp, &va, cred, lwp);
2284 		vput(nd.ni_vp);
2285 	    } else {
2286 		vput(nd.ni_vp);
2287 	    }
2288 	}
2289 out:
2290 	if (pathcp)
2291 		free(pathcp, M_TEMP);
2292 	if (dirp) {
2293 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2294 		vrele(dirp);
2295 	}
2296 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2297 	if (v3) {
2298 		if (!error) {
2299 			nfsm_srvpostop_fh(fhp);
2300 			nfsm_srvpostop_attr(0, &va);
2301 		}
2302 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2303 	}
2304 	vn_finished_write(mp, 0);
2305 	return (0);
2306 nfsmout:
2307 	if (dirp)
2308 		vrele(dirp);
2309 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2310 	if (nd.ni_dvp == nd.ni_vp)
2311 		vrele(nd.ni_dvp);
2312 	else
2313 		vput(nd.ni_dvp);
2314 	if (nd.ni_vp)
2315 		vrele(nd.ni_vp);
2316 	if (pathcp)
2317 		free(pathcp, M_TEMP);
2318 	vn_finished_write(mp, 0);
2319 	return (error);
2320 }
2321 
2322 /*
2323  * nfs mkdir service
2324  */
2325 int
2326 nfsrv_mkdir(nfsd, slp, lwp, mrq)
2327 	struct nfsrv_descript *nfsd;
2328 	struct nfssvc_sock *slp;
2329 	struct lwp *lwp;
2330 	struct mbuf **mrq;
2331 {
2332 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2333 	struct mbuf *nam = nfsd->nd_nam;
2334 	caddr_t dpos = nfsd->nd_dpos;
2335 	kauth_cred_t cred = nfsd->nd_cr;
2336 	struct vattr va, dirfor, diraft;
2337 	struct nfs_fattr *fp;
2338 	struct nameidata nd;
2339 	caddr_t cp;
2340 	u_int32_t *tl;
2341 	int32_t t1;
2342 	caddr_t bpos;
2343 	int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2344 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2345 	char *cp2;
2346 	struct mbuf *mb, *mreq;
2347 	struct vnode *vp, *dirp = (struct vnode *)0;
2348 	nfsfh_t nfh;
2349 	fhandle_t *fhp;
2350 	u_quad_t frev;
2351 	struct mount *mp = NULL;
2352 
2353 	fhp = &nfh.fh_generic;
2354 	nfsm_srvmtofh(fhp);
2355 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2356 		return (ESTALE);
2357 	vn_start_write(NULL, &mp, V_WAIT);
2358 	nfsm_srvnamesiz(len);
2359 	nd.ni_cnd.cn_cred = cred;
2360 	nd.ni_cnd.cn_nameiop = CREATE;
2361 	nd.ni_cnd.cn_flags = LOCKPARENT;
2362 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2363 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2364 	if (dirp) {
2365 		if (v3)
2366 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2367 		else {
2368 			vrele(dirp);
2369 			dirp = (struct vnode *)0;
2370 		}
2371 	}
2372 	if (error) {
2373 		nfsm_reply(NFSX_WCCDATA(v3));
2374 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2375 		if (dirp)
2376 			vrele(dirp);
2377 		vn_finished_write(mp, 0);
2378 		return (0);
2379 	}
2380 	VATTR_NULL(&va);
2381 	if (v3) {
2382 		va.va_mode = 0;
2383 		nfsm_srvsattr(&va);
2384 	} else {
2385 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2386 		va.va_mode = nfstov_mode(*tl++);
2387 	}
2388 	va.va_type = VDIR;
2389 	vp = nd.ni_vp;
2390 	if (vp != NULL) {
2391 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2392 		if (nd.ni_dvp == vp)
2393 			vrele(nd.ni_dvp);
2394 		else
2395 			vput(nd.ni_dvp);
2396 		vrele(vp);
2397 		error = EEXIST;
2398 		goto out;
2399 	}
2400 	nqsrv_getl(nd.ni_dvp, ND_WRITE);
2401 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2402 	if (!error) {
2403 		vp = nd.ni_vp;
2404 		error = vfs_composefh(vp, fhp);
2405 		if (!error)
2406 			error = VOP_GETATTR(vp, &va, cred, lwp);
2407 		vput(vp);
2408 	}
2409 out:
2410 	if (dirp) {
2411 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2412 		vrele(dirp);
2413 	}
2414 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2415 	if (v3) {
2416 		if (!error) {
2417 			nfsm_srvpostop_fh(fhp);
2418 			nfsm_srvpostop_attr(0, &va);
2419 		}
2420 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2421 	} else {
2422 		nfsm_srvfhtom(fhp, v3);
2423 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2424 		nfsm_srvfillattr(&va, fp);
2425 	}
2426 	vn_finished_write(mp, 0);
2427 	return (0);
2428 nfsmout:
2429 	if (dirp)
2430 		vrele(dirp);
2431 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2432 	if (nd.ni_dvp == nd.ni_vp)
2433 		vrele(nd.ni_dvp);
2434 	else
2435 		vput(nd.ni_dvp);
2436 	if (nd.ni_vp)
2437 		vrele(nd.ni_vp);
2438 	vn_finished_write(mp, 0);
2439 	return (error);
2440 }
2441 
2442 /*
2443  * nfs rmdir service
2444  */
2445 int
2446 nfsrv_rmdir(nfsd, slp, lwp, mrq)
2447 	struct nfsrv_descript *nfsd;
2448 	struct nfssvc_sock *slp;
2449 	struct lwp *lwp;
2450 	struct mbuf **mrq;
2451 {
2452 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2453 	struct mbuf *nam = nfsd->nd_nam;
2454 	caddr_t dpos = nfsd->nd_dpos;
2455 	kauth_cred_t cred = nfsd->nd_cr;
2456 	u_int32_t *tl;
2457 	int32_t t1;
2458 	caddr_t bpos;
2459 	int error = 0, cache = 0, len, dirfor_ret = 1, diraft_ret = 1;
2460 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2461 	char *cp2;
2462 	struct mbuf *mb, *mreq;
2463 	struct vnode *vp, *dirp = (struct vnode *)0;
2464 	struct vattr dirfor, diraft;
2465 	nfsfh_t nfh;
2466 	fhandle_t *fhp;
2467 	struct nameidata nd;
2468 	u_quad_t frev;
2469 	struct mount *mp = NULL;
2470 
2471 	fhp = &nfh.fh_generic;
2472 	nfsm_srvmtofh(fhp);
2473 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2474 		return (ESTALE);
2475 	vn_start_write(NULL, &mp, V_WAIT);
2476 	nfsm_srvnamesiz(len);
2477 	nd.ni_cnd.cn_cred = cred;
2478 	nd.ni_cnd.cn_nameiop = DELETE;
2479 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2480 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2481 		&dirp, lwp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2482 	if (dirp) {
2483 		if (v3)
2484 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, lwp);
2485 		else {
2486 			vrele(dirp);
2487 			dirp = (struct vnode *)0;
2488 		}
2489 	}
2490 	if (error) {
2491 		nfsm_reply(NFSX_WCCDATA(v3));
2492 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2493 		if (dirp)
2494 			vrele(dirp);
2495 		vn_finished_write(mp, 0);
2496 		return (0);
2497 	}
2498 	vp = nd.ni_vp;
2499 	if (vp->v_type != VDIR) {
2500 		error = ENOTDIR;
2501 		goto out;
2502 	}
2503 	/*
2504 	 * No rmdir "." please.
2505 	 */
2506 	if (nd.ni_dvp == vp) {
2507 		error = EINVAL;
2508 		goto out;
2509 	}
2510 	/*
2511 	 * The root of a mounted filesystem cannot be deleted.
2512 	 */
2513 	if (vp->v_flag & VROOT)
2514 		error = EBUSY;
2515 out:
2516 	if (!error) {
2517 		nqsrv_getl(nd.ni_dvp, ND_WRITE);
2518 		nqsrv_getl(vp, ND_WRITE);
2519 		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2520 	} else {
2521 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2522 		if (nd.ni_dvp == nd.ni_vp)
2523 			vrele(nd.ni_dvp);
2524 		else
2525 			vput(nd.ni_dvp);
2526 		vput(vp);
2527 	}
2528 	if (dirp) {
2529 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, lwp);
2530 		vrele(dirp);
2531 	}
2532 	nfsm_reply(NFSX_WCCDATA(v3));
2533 	if (v3) {
2534 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2535 		vn_finished_write(mp, 0);
2536 		return (0);
2537 	}
2538 	vn_finished_write(mp, 0);
2539 	nfsm_srvdone;
2540 }
2541 
2542 /*
2543  * nfs readdir service
2544  * - mallocs what it thinks is enough to read
2545  *	count rounded up to a multiple of NFS_SRVDIRBLKSIZ <= NFS_MAXREADDIR
2546  * - calls VOP_READDIR()
2547  * - loops around building the reply
2548  *	if the output generated exceeds count break out of loop
2549  *	The nfsm_clget macro is used here so that the reply will be packed
2550  *	tightly in mbuf clusters.
2551  * - it only knows that it has encountered eof when the VOP_READDIR()
2552  *	reads nothing
2553  * - as such one readdir rpc will return eof false although you are there
2554  *	and then the next will return eof
2555  * - it trims out records with d_fileno == 0
2556  *	this doesn't matter for Unix clients, but they might confuse clients
2557  *	for other os'.
2558  * - it trims out records with d_type == DT_WHT
2559  *	these cannot be seen through NFS (unless we extend the protocol)
2560  * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2561  *	than requested, but this may not apply to all filesystems. For
2562  *	example, client NFS does not { although it is never remote mounted
2563  *	anyhow }
2564  *     The alternate call nfsrv_readdirplus() does lookups as well.
2565  * PS: The NFS protocol spec. does not clarify what the "count" byte
2566  *	argument is a count of.. just name strings and file id's or the
2567  *	entire reply rpc or ...
2568  *	I tried just file name and id sizes and it confused the Sun client,
2569  *	so I am using the full rpc size now. The "paranoia.." comment refers
2570  *	to including the status longwords that are not a part of the dir.
2571  *	"entry" structures, but are in the rpc.
2572  */
2573 
2574 #define	NFS_SRVDIRBLKSIZ	1024
2575 
2576 struct flrep {
2577 	nfsuint64 fl_off;
2578 	u_int32_t fl_postopok;
2579 	u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2580 	u_int32_t fl_fhok;
2581 	u_int32_t fl_fhsize;
2582 	u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2583 };
2584 
2585 int
2586 nfsrv_readdir(nfsd, slp, lwp, mrq)
2587 	struct nfsrv_descript *nfsd;
2588 	struct nfssvc_sock *slp;
2589 	struct lwp *lwp;
2590 	struct mbuf **mrq;
2591 {
2592 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2593 	struct mbuf *nam = nfsd->nd_nam;
2594 	caddr_t dpos = nfsd->nd_dpos;
2595 	kauth_cred_t cred = nfsd->nd_cr;
2596 	char *bp, *be;
2597 	struct mbuf *mp;
2598 	struct dirent *dp;
2599 	caddr_t cp;
2600 	u_int32_t *tl;
2601 	int32_t t1;
2602 	caddr_t bpos;
2603 	struct mbuf *mb, *mreq, *mp2;
2604 	char *cpos, *cend, *cp2, *rbuf;
2605 	struct vnode *vp;
2606 	struct vattr at;
2607 	nfsfh_t nfh;
2608 	fhandle_t *fhp;
2609 	struct uio io;
2610 	struct iovec iv;
2611 	int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2612 	int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, ncookies;
2613 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2614 	u_quad_t frev, off, toff, verf;
2615 	off_t *cookies = NULL, *cookiep;
2616 	nfsuint64 jar;
2617 
2618 	fhp = &nfh.fh_generic;
2619 	nfsm_srvmtofh(fhp);
2620 	if (v3) {
2621 		nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2622 		toff = fxdr_hyper(tl);
2623 		tl += 2;
2624 		verf = fxdr_hyper(tl);
2625 		tl += 2;
2626 	} else {
2627 		nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2628 		toff = fxdr_unsigned(u_quad_t, *tl++);
2629 	}
2630 	off = toff;
2631 	cnt = fxdr_unsigned(int, *tl);
2632 	siz = ((cnt + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2633 	xfer = NFS_SRVMAXDATA(nfsd);
2634 	if (siz > xfer)
2635 		siz = xfer;
2636 	fullsiz = siz;
2637 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2638 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2639 	if (!error && vp->v_type != VDIR) {
2640 		error = ENOTDIR;
2641 		vput(vp);
2642 	}
2643 	if (error) {
2644 		nfsm_reply(NFSX_UNSIGNED);
2645 		nfsm_srvpostop_attr(getret, &at);
2646 		return (0);
2647 	}
2648 	nqsrv_getl(vp, ND_READ);
2649 	if (v3) {
2650 		error = getret = VOP_GETATTR(vp, &at, cred, lwp);
2651 #ifdef NFS3_STRICTVERF
2652 		/*
2653 		 * XXX This check is too strict for Solaris 2.5 clients.
2654 		 */
2655 		if (!error && toff && verf != at.va_filerev)
2656 			error = NFSERR_BAD_COOKIE;
2657 #endif
2658 	}
2659 	if (!error)
2660 		error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
2661 	if (error) {
2662 		vput(vp);
2663 		nfsm_reply(NFSX_POSTOPATTR(v3));
2664 		nfsm_srvpostop_attr(getret, &at);
2665 		return (0);
2666 	}
2667 	VOP_UNLOCK(vp, 0);
2668 	rbuf = malloc(siz, M_TEMP, M_WAITOK);
2669 again:
2670 	iv.iov_base = rbuf;
2671 	iv.iov_len = fullsiz;
2672 	io.uio_iov = &iv;
2673 	io.uio_iovcnt = 1;
2674 	io.uio_offset = (off_t)off;
2675 	io.uio_resid = fullsiz;
2676 	io.uio_rw = UIO_READ;
2677 	UIO_SETUP_SYSSPACE(&io);
2678 	eofflag = 0;
2679 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2680 
2681 	error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2682 
2683 	off = (off_t)io.uio_offset;
2684 	if (!cookies && !error)
2685 		error = NFSERR_PERM;
2686 	if (v3) {
2687 		getret = VOP_GETATTR(vp, &at, cred, lwp);
2688 		if (!error)
2689 			error = getret;
2690 	}
2691 
2692 	VOP_UNLOCK(vp, 0);
2693 	if (error) {
2694 		vrele(vp);
2695 		free((caddr_t)rbuf, M_TEMP);
2696 		if (cookies)
2697 			free((caddr_t)cookies, M_TEMP);
2698 		nfsm_reply(NFSX_POSTOPATTR(v3));
2699 		nfsm_srvpostop_attr(getret, &at);
2700 		return (0);
2701 	}
2702 	if (io.uio_resid) {
2703 		siz -= io.uio_resid;
2704 
2705 		/*
2706 		 * If nothing read, return eof
2707 		 * rpc reply
2708 		 */
2709 		if (siz == 0) {
2710 			vrele(vp);
2711 			nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2712 				2 * NFSX_UNSIGNED);
2713 			if (v3) {
2714 				nfsm_srvpostop_attr(getret, &at);
2715 				nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2716 				txdr_hyper(at.va_filerev, tl);
2717 				tl += 2;
2718 			} else
2719 				nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2720 			*tl++ = nfs_false;
2721 			*tl = nfs_true;
2722 			free((caddr_t)rbuf, M_TEMP);
2723 			free((caddr_t)cookies, M_TEMP);
2724 			return (0);
2725 		}
2726 	}
2727 
2728 	/*
2729 	 * Check for degenerate cases of nothing useful read.
2730 	 * If so go try again
2731 	 */
2732 	cpos = rbuf;
2733 	cend = rbuf + siz;
2734 	dp = (struct dirent *)cpos;
2735 	cookiep = cookies;
2736 
2737 	while (cpos < cend && ncookies > 0 &&
2738 		(dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2739 		cpos += dp->d_reclen;
2740 		dp = (struct dirent *)cpos;
2741 		cookiep++;
2742 		ncookies--;
2743 	}
2744 	if (cpos >= cend || ncookies == 0) {
2745 		toff = off;
2746 		siz = fullsiz;
2747 		goto again;
2748 	}
2749 
2750 	len = 3 * NFSX_UNSIGNED;	/* paranoia, probably can be 0 */
2751 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2752 	if (v3) {
2753 		nfsm_srvpostop_attr(getret, &at);
2754 		nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2755 		txdr_hyper(at.va_filerev, tl);
2756 	}
2757 	mp = mp2 = mb;
2758 	bp = bpos;
2759 	be = bp + M_TRAILINGSPACE(mp);
2760 
2761 	/* Loop through the records and build reply */
2762 	while (cpos < cend && ncookies > 0) {
2763 		if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2764 			nlen = dp->d_namlen;
2765 			rem = nfsm_rndup(nlen)-nlen;
2766 			len += (4 * NFSX_UNSIGNED + nlen + rem);
2767 			if (v3)
2768 				len += 2 * NFSX_UNSIGNED;
2769 			if (len > cnt) {
2770 				eofflag = 0;
2771 				break;
2772 			}
2773 			/*
2774 			 * Build the directory record xdr from
2775 			 * the dirent entry.
2776 			 */
2777 			nfsm_clget;
2778 			*tl = nfs_true;
2779 			bp += NFSX_UNSIGNED;
2780 			if (v3) {
2781 				nfsm_clget;
2782 				*tl = txdr_unsigned(dp->d_fileno >> 32);
2783 				bp += NFSX_UNSIGNED;
2784 			}
2785 			nfsm_clget;
2786 			*tl = txdr_unsigned(dp->d_fileno);
2787 			bp += NFSX_UNSIGNED;
2788 			nfsm_clget;
2789 			*tl = txdr_unsigned(nlen);
2790 			bp += NFSX_UNSIGNED;
2791 
2792 			/* And loop around copying the name */
2793 			xfer = nlen;
2794 			cp = dp->d_name;
2795 			while (xfer > 0) {
2796 				nfsm_clget;
2797 				if ((bp+xfer) > be)
2798 					tsiz = be-bp;
2799 				else
2800 					tsiz = xfer;
2801 				memcpy(bp, cp, tsiz);
2802 				bp += tsiz;
2803 				xfer -= tsiz;
2804 				if (xfer > 0)
2805 					cp += tsiz;
2806 			}
2807 			/* And null pad to an int32_t boundary */
2808 			for (i = 0; i < rem; i++)
2809 				*bp++ = '\0';
2810 			nfsm_clget;
2811 
2812 			/* Finish off the record */
2813 			txdr_hyper(*cookiep, &jar);
2814 			if (v3) {
2815 				*tl = jar.nfsuquad[0];
2816 				bp += NFSX_UNSIGNED;
2817 				nfsm_clget;
2818 			}
2819 			*tl = jar.nfsuquad[1];
2820 			bp += NFSX_UNSIGNED;
2821 		}
2822 		cpos += dp->d_reclen;
2823 		dp = (struct dirent *)cpos;
2824 		cookiep++;
2825 		ncookies--;
2826 	}
2827 	vrele(vp);
2828 	nfsm_clget;
2829 	*tl = nfs_false;
2830 	bp += NFSX_UNSIGNED;
2831 	nfsm_clget;
2832 	if (eofflag)
2833 		*tl = nfs_true;
2834 	else
2835 		*tl = nfs_false;
2836 	bp += NFSX_UNSIGNED;
2837 	if (mp != mb) {
2838 		if (bp < be)
2839 			mp->m_len = bp - mtod(mp, caddr_t);
2840 	} else
2841 		mp->m_len += bp - bpos;
2842 	free((caddr_t)rbuf, M_TEMP);
2843 	free((caddr_t)cookies, M_TEMP);
2844 	nfsm_srvdone;
2845 }
2846 
2847 int
2848 nfsrv_readdirplus(nfsd, slp, lwp, mrq)
2849 	struct nfsrv_descript *nfsd;
2850 	struct nfssvc_sock *slp;
2851 	struct lwp *lwp;
2852 	struct mbuf **mrq;
2853 {
2854 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2855 	struct mbuf *nam = nfsd->nd_nam;
2856 	caddr_t dpos = nfsd->nd_dpos;
2857 	kauth_cred_t cred = nfsd->nd_cr;
2858 	char *bp, *be;
2859 	struct mbuf *mp;
2860 	struct dirent *dp;
2861 	caddr_t cp;
2862 	u_int32_t *tl;
2863 	int32_t t1;
2864 	caddr_t bpos;
2865 	struct mbuf *mb, *mreq, *mp2;
2866 	char *cpos, *cend, *cp2, *rbuf;
2867 	struct vnode *vp, *nvp;
2868 	struct flrep fl;
2869 	nfsfh_t nfh;
2870 	fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2871 	struct uio io;
2872 	struct iovec iv;
2873 	struct vattr va, at, *vap = &va;
2874 	struct nfs_fattr *fp;
2875 	int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2876 	int siz, cnt, fullsiz, eofflag, rdonly, cache = 0, dirlen, ncookies;
2877 	u_quad_t frev, off, toff, verf;
2878 	off_t *cookies = NULL, *cookiep;
2879 
2880 	fhp = &nfh.fh_generic;
2881 	nfsm_srvmtofh(fhp);
2882 	nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2883 	toff = fxdr_hyper(tl);
2884 	tl += 2;
2885 	verf = fxdr_hyper(tl);
2886 	tl += 2;
2887 	siz = fxdr_unsigned(int, *tl++);
2888 	cnt = fxdr_unsigned(int, *tl);
2889 	off = toff;
2890 	siz = ((siz + NFS_SRVDIRBLKSIZ - 1) & ~(NFS_SRVDIRBLKSIZ - 1));
2891 	xfer = NFS_SRVMAXDATA(nfsd);
2892 	if (siz > xfer)
2893 		siz = xfer;
2894 	fullsiz = siz;
2895 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2896 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2897 	if (!error && vp->v_type != VDIR) {
2898 		error = ENOTDIR;
2899 		vput(vp);
2900 	}
2901 	if (error) {
2902 		nfsm_reply(NFSX_UNSIGNED);
2903 		nfsm_srvpostop_attr(getret, &at);
2904 		return (0);
2905 	}
2906 	error = getret = VOP_GETATTR(vp, &at, cred, lwp);
2907 #ifdef NFS3_STRICTVERF
2908 	/*
2909 	 * XXX This check is too strict for Solaris 2.5 clients.
2910 	 */
2911 	if (!error && toff && verf != at.va_filerev)
2912 		error = NFSERR_BAD_COOKIE;
2913 #endif
2914 	if (!error) {
2915 		nqsrv_getl(vp, ND_READ);
2916 		error = nfsrv_access(vp, VEXEC, cred, rdonly, lwp, 0);
2917 	}
2918 	if (error) {
2919 		vput(vp);
2920 		nfsm_reply(NFSX_V3POSTOPATTR);
2921 		nfsm_srvpostop_attr(getret, &at);
2922 		return (0);
2923 	}
2924 	VOP_UNLOCK(vp, 0);
2925 
2926 	rbuf = malloc(siz, M_TEMP, M_WAITOK);
2927 again:
2928 	iv.iov_base = rbuf;
2929 	iv.iov_len = fullsiz;
2930 	io.uio_iov = &iv;
2931 	io.uio_iovcnt = 1;
2932 	io.uio_offset = (off_t)off;
2933 	io.uio_resid = fullsiz;
2934 	io.uio_rw = UIO_READ;
2935 	UIO_SETUP_SYSSPACE(&io);
2936 	eofflag = 0;
2937 
2938 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2939 
2940 	error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2941 
2942 	off = (u_quad_t)io.uio_offset;
2943 	getret = VOP_GETATTR(vp, &at, cred, lwp);
2944 
2945 	VOP_UNLOCK(vp, 0);
2946 
2947 	/*
2948 	 * If the VGET operation doesn't work for this filesystem,
2949 	 * we can't support readdirplus. Returning NOTSUPP should
2950 	 * make clients fall back to plain readdir.
2951 	 * There's no need to check for VPTOFH as well, we wouldn't
2952 	 * even be here otherwise.
2953 	 */
2954 	if (!getret) {
2955 		if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2956 			getret = (getret == EOPNOTSUPP) ?
2957 				NFSERR_NOTSUPP : NFSERR_IO;
2958 		else
2959 			vput(nvp);
2960 	}
2961 
2962 	if (!cookies && !error)
2963 		error = NFSERR_PERM;
2964 	if (!error)
2965 		error = getret;
2966 	if (error) {
2967 		vrele(vp);
2968 		if (cookies)
2969 			free((caddr_t)cookies, M_TEMP);
2970 		free((caddr_t)rbuf, M_TEMP);
2971 		nfsm_reply(NFSX_V3POSTOPATTR);
2972 		nfsm_srvpostop_attr(getret, &at);
2973 		return (0);
2974 	}
2975 	if (io.uio_resid) {
2976 		siz -= io.uio_resid;
2977 
2978 		/*
2979 		 * If nothing read, return eof
2980 		 * rpc reply
2981 		 */
2982 		if (siz == 0) {
2983 			vrele(vp);
2984 			nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2985 				2 * NFSX_UNSIGNED);
2986 			nfsm_srvpostop_attr(getret, &at);
2987 			nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2988 			txdr_hyper(at.va_filerev, tl);
2989 			tl += 2;
2990 			*tl++ = nfs_false;
2991 			*tl = nfs_true;
2992 			free((caddr_t)cookies, M_TEMP);
2993 			free((caddr_t)rbuf, M_TEMP);
2994 			return (0);
2995 		}
2996 	}
2997 
2998 	/*
2999 	 * Check for degenerate cases of nothing useful read.
3000 	 * If so go try again
3001 	 */
3002 	cpos = rbuf;
3003 	cend = rbuf + siz;
3004 	dp = (struct dirent *)cpos;
3005 	cookiep = cookies;
3006 
3007 	while (cpos < cend && ncookies > 0 &&
3008 		(dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
3009 		cpos += dp->d_reclen;
3010 		dp = (struct dirent *)cpos;
3011 		cookiep++;
3012 		ncookies--;
3013 	}
3014 	if (cpos >= cend || ncookies == 0) {
3015 		toff = off;
3016 		siz = fullsiz;
3017 		goto again;
3018 	}
3019 
3020 	dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
3021 	nfsm_reply(cnt);
3022 	nfsm_srvpostop_attr(getret, &at);
3023 	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3024 	txdr_hyper(at.va_filerev, tl);
3025 	mp = mp2 = mb;
3026 	bp = bpos;
3027 	be = bp + M_TRAILINGSPACE(mp);
3028 
3029 	/* Loop through the records and build reply */
3030 	while (cpos < cend && ncookies > 0) {
3031 		if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
3032 			nlen = dp->d_namlen;
3033 			rem = nfsm_rndup(nlen)-nlen;
3034 
3035 			/*
3036 			 * For readdir_and_lookup get the vnode using
3037 			 * the file number.
3038 			 */
3039 			if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
3040 				goto invalid;
3041 			if (vfs_composefh(nvp, nfhp)) {
3042 				vput(nvp);
3043 				goto invalid;
3044 			}
3045 			if (VOP_GETATTR(nvp, vap, cred, lwp)) {
3046 				vput(nvp);
3047 				goto invalid;
3048 			}
3049 			vput(nvp);
3050 
3051 			/*
3052 			 * If either the dircount or maxcount will be
3053 			 * exceeded, get out now. Both of these lengths
3054 			 * are calculated conservatively, including all
3055 			 * XDR overheads.
3056 			 */
3057 			len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
3058 				NFSX_V3POSTOPATTR);
3059 			dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
3060 			if (len > cnt || dirlen > fullsiz) {
3061 				eofflag = 0;
3062 				break;
3063 			}
3064 
3065 			/*
3066 			 * Build the directory record xdr from
3067 			 * the dirent entry.
3068 			 */
3069 			fp = (struct nfs_fattr *)&fl.fl_fattr;
3070 			nfsm_srvfillattr(vap, fp);
3071 			fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
3072 			fl.fl_fhok = nfs_true;
3073 			fl.fl_postopok = nfs_true;
3074 			txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
3075 
3076 			nfsm_clget;
3077 			*tl = nfs_true;
3078 			bp += NFSX_UNSIGNED;
3079 			nfsm_clget;
3080 			*tl = txdr_unsigned(dp->d_fileno >> 32);
3081 			bp += NFSX_UNSIGNED;
3082 			nfsm_clget;
3083 			*tl = txdr_unsigned(dp->d_fileno);
3084 			bp += NFSX_UNSIGNED;
3085 			nfsm_clget;
3086 			*tl = txdr_unsigned(nlen);
3087 			bp += NFSX_UNSIGNED;
3088 
3089 			/* And loop around copying the name */
3090 			xfer = nlen;
3091 			cp = dp->d_name;
3092 			while (xfer > 0) {
3093 				nfsm_clget;
3094 				if ((bp + xfer) > be)
3095 					tsiz = be - bp;
3096 				else
3097 					tsiz = xfer;
3098 				memcpy(bp, cp, tsiz);
3099 				bp += tsiz;
3100 				xfer -= tsiz;
3101 				if (xfer > 0)
3102 					cp += tsiz;
3103 			}
3104 			/* And null pad to an int32_t boundary */
3105 			for (i = 0; i < rem; i++)
3106 				*bp++ = '\0';
3107 
3108 			/*
3109 			 * Now copy the flrep structure out.
3110 			 */
3111 			xfer = sizeof (struct flrep);
3112 			cp = (caddr_t)&fl;
3113 			while (xfer > 0) {
3114 				nfsm_clget;
3115 				if ((bp + xfer) > be)
3116 					tsiz = be - bp;
3117 				else
3118 					tsiz = xfer;
3119 				memcpy(bp, cp, tsiz);
3120 				bp += tsiz;
3121 				xfer -= tsiz;
3122 				if (xfer > 0)
3123 					cp += tsiz;
3124 			}
3125 		}
3126 invalid:
3127 		cpos += dp->d_reclen;
3128 		dp = (struct dirent *)cpos;
3129 		cookiep++;
3130 		ncookies--;
3131 	}
3132 	vrele(vp);
3133 	nfsm_clget;
3134 	*tl = nfs_false;
3135 	bp += NFSX_UNSIGNED;
3136 	nfsm_clget;
3137 	if (eofflag)
3138 		*tl = nfs_true;
3139 	else
3140 		*tl = nfs_false;
3141 	bp += NFSX_UNSIGNED;
3142 	if (mp != mb) {
3143 		if (bp < be)
3144 			mp->m_len = bp - mtod(mp, caddr_t);
3145 	} else
3146 		mp->m_len += bp - bpos;
3147 	free((caddr_t)cookies, M_TEMP);
3148 	free((caddr_t)rbuf, M_TEMP);
3149 	nfsm_srvdone;
3150 }
3151 
3152 /*
3153  * nfs commit service
3154  */
3155 int
3156 nfsrv_commit(nfsd, slp, lwp, mrq)
3157 	struct nfsrv_descript *nfsd;
3158 	struct nfssvc_sock *slp;
3159 	struct lwp *lwp;
3160 	struct mbuf **mrq;
3161 {
3162 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3163 	struct mbuf *nam = nfsd->nd_nam;
3164 	caddr_t dpos = nfsd->nd_dpos;
3165 	kauth_cred_t cred = nfsd->nd_cr;
3166 	struct vattr bfor, aft;
3167 	struct vnode *vp;
3168 	nfsfh_t nfh;
3169 	fhandle_t *fhp;
3170 	u_int32_t *tl;
3171 	int32_t t1;
3172 	caddr_t bpos;
3173 	int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache = 0;
3174 	uint32_t cnt;
3175 	char *cp2;
3176 	struct mbuf *mb, *mreq;
3177 	u_quad_t frev, off, end;
3178 	struct mount *mp = NULL;
3179 
3180 	fhp = &nfh.fh_generic;
3181 	nfsm_srvmtofh(fhp);
3182 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
3183 		return (ESTALE);
3184 	vn_start_write(NULL, &mp, V_WAIT);
3185 	nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3186 
3187 	off = fxdr_hyper(tl);
3188 	tl += 2;
3189 	cnt = fxdr_unsigned(uint32_t, *tl);
3190 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3191 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3192 	if (error) {
3193 		nfsm_reply(2 * NFSX_UNSIGNED);
3194 		nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3195 		vn_finished_write(mp, 0);
3196 		return (0);
3197 	}
3198 	for_ret = VOP_GETATTR(vp, &bfor, cred, lwp);
3199 	end = (cnt > 0) ? off + cnt : vp->v_size;
3200 	if (end < off || end > vp->v_size)
3201 		end = vp->v_size;
3202 	if (off < vp->v_size)
3203 		error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, lwp);
3204 	/* else error == 0, from nfsrv_fhtovp() */
3205 	aft_ret = VOP_GETATTR(vp, &aft, cred, lwp);
3206 	vput(vp);
3207 	nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3208 	nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3209 	if (!error) {
3210 		nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3211 		*tl++ = txdr_unsigned(boottime.tv_sec);
3212 		*tl = txdr_unsigned(boottime.tv_usec);
3213 	} else {
3214 		vn_finished_write(mp, 0);
3215 		return (0);
3216 	}
3217 	vn_finished_write(mp, 0);
3218 	nfsm_srvdone;
3219 }
3220 
3221 /*
3222  * nfs statfs service
3223  */
3224 int
3225 nfsrv_statfs(nfsd, slp, lwp, mrq)
3226 	struct nfsrv_descript *nfsd;
3227 	struct nfssvc_sock *slp;
3228 	struct lwp *lwp;
3229 	struct mbuf **mrq;
3230 {
3231 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3232 	struct mbuf *nam = nfsd->nd_nam;
3233 	caddr_t dpos = nfsd->nd_dpos;
3234 	kauth_cred_t cred = nfsd->nd_cr;
3235 	struct statvfs *sf = NULL;
3236 	struct nfs_statfs *sfp;
3237 	u_int32_t *tl;
3238 	int32_t t1;
3239 	caddr_t bpos;
3240 	int error = 0, rdonly, cache = 0, getret = 1;
3241 	int v3 = (nfsd->nd_flag & ND_NFSV3);
3242 	char *cp2;
3243 	struct mbuf *mb, *mreq;
3244 	struct vnode *vp;
3245 	struct vattr at;
3246 	nfsfh_t nfh;
3247 	fhandle_t *fhp;
3248 	u_quad_t frev, tval;
3249 
3250 	fhp = &nfh.fh_generic;
3251 	nfsm_srvmtofh(fhp);
3252 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3253 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3254 	if (error) {
3255 		nfsm_reply(NFSX_UNSIGNED);
3256 		nfsm_srvpostop_attr(getret, &at);
3257 		return (0);
3258 	}
3259 	sf = malloc(sizeof(*sf), M_TEMP, M_WAITOK);
3260 	error = VFS_STATVFS(vp->v_mount, sf, lwp);
3261 	getret = VOP_GETATTR(vp, &at, cred, lwp);
3262 	vput(vp);
3263 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3264 	if (v3)
3265 		nfsm_srvpostop_attr(getret, &at);
3266 	if (error) {
3267 		free(sf, M_TEMP);
3268 		return (0);
3269 	}
3270 	nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3271 	if (v3) {
3272 		tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize);
3273 		txdr_hyper(tval, &sfp->sf_tbytes);
3274 		tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize);
3275 		txdr_hyper(tval, &sfp->sf_fbytes);
3276 		tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize);
3277 		txdr_hyper(tval, &sfp->sf_abytes);
3278 		tval = (u_quad_t)sf->f_files;
3279 		txdr_hyper(tval, &sfp->sf_tfiles);
3280 		tval = (u_quad_t)sf->f_ffree;
3281 		txdr_hyper(tval, &sfp->sf_ffiles);
3282 		txdr_hyper(tval, &sfp->sf_afiles);
3283 		sfp->sf_invarsec = 0;
3284 	} else {
3285 		sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3286 		sfp->sf_bsize = txdr_unsigned(sf->f_frsize);
3287 		sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3288 		sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3289 		sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3290 	}
3291 nfsmout:
3292 	if (sf)
3293 	    free(sf, M_TEMP);
3294 	return error;
3295 }
3296 
3297 /*
3298  * nfs fsinfo service
3299  */
3300 int
3301 nfsrv_fsinfo(nfsd, slp, lwp, mrq)
3302 	struct nfsrv_descript *nfsd;
3303 	struct nfssvc_sock *slp;
3304 	struct lwp *lwp;
3305 	struct mbuf **mrq;
3306 {
3307 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3308 	struct mbuf *nam = nfsd->nd_nam;
3309 	caddr_t dpos = nfsd->nd_dpos;
3310 	kauth_cred_t cred = nfsd->nd_cr;
3311 	u_int32_t *tl;
3312 	struct nfsv3_fsinfo *sip;
3313 	int32_t t1;
3314 	caddr_t bpos;
3315 	int error = 0, rdonly, cache = 0, getret = 1;
3316 	uint32_t maxdata;
3317 	char *cp2;
3318 	struct mbuf *mb, *mreq;
3319 	struct vnode *vp;
3320 	struct vattr at;
3321 	nfsfh_t nfh;
3322 	fhandle_t *fhp;
3323 	u_quad_t frev, maxfsize;
3324 	struct statvfs *sb;
3325 
3326 	fhp = &nfh.fh_generic;
3327 	nfsm_srvmtofh(fhp);
3328 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3329 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3330 	if (error) {
3331 		nfsm_reply(NFSX_UNSIGNED);
3332 		nfsm_srvpostop_attr(getret, &at);
3333 		return (0);
3334 	}
3335 
3336 	/* XXX Try to make a guess on the max file size. */
3337 	sb = malloc(sizeof(*sb), M_TEMP, M_WAITOK);
3338 	VFS_STATVFS(vp->v_mount, sb, (struct lwp *)0);
3339 	maxfsize = (u_quad_t)0x80000000 * sb->f_frsize - 1;
3340 	free(sb, M_TEMP);
3341 
3342 	getret = VOP_GETATTR(vp, &at, cred, lwp);
3343 	vput(vp);
3344 	nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3345 	nfsm_srvpostop_attr(getret, &at);
3346 	nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3347 
3348 	/*
3349 	 * XXX
3350 	 * There should be file system VFS OP(s) to get this information.
3351 	 * For now, assume ufs.
3352 	 */
3353 	if (slp->ns_so->so_type == SOCK_DGRAM)
3354 		maxdata = NFS_MAXDGRAMDATA;
3355 	else
3356 		maxdata = NFS_MAXDATA;
3357 	sip->fs_rtmax = txdr_unsigned(maxdata);
3358 	sip->fs_rtpref = txdr_unsigned(maxdata);
3359 	sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3360 	sip->fs_wtmax = txdr_unsigned(maxdata);
3361 	sip->fs_wtpref = txdr_unsigned(maxdata);
3362 	sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3363 	sip->fs_dtpref = txdr_unsigned(maxdata);
3364 	txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3365 	sip->fs_timedelta.nfsv3_sec = 0;
3366 	sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3367 	sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3368 		NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3369 		NFSV3FSINFO_CANSETTIME);
3370 	nfsm_srvdone;
3371 }
3372 
3373 /*
3374  * nfs pathconf service
3375  */
3376 int
3377 nfsrv_pathconf(nfsd, slp, lwp, mrq)
3378 	struct nfsrv_descript *nfsd;
3379 	struct nfssvc_sock *slp;
3380 	struct lwp *lwp;
3381 	struct mbuf **mrq;
3382 {
3383 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3384 	struct mbuf *nam = nfsd->nd_nam;
3385 	caddr_t dpos = nfsd->nd_dpos;
3386 	kauth_cred_t cred = nfsd->nd_cr;
3387 	u_int32_t *tl;
3388 	struct nfsv3_pathconf *pc;
3389 	int32_t t1;
3390 	caddr_t bpos;
3391 	int error = 0, rdonly, cache = 0, getret = 1;
3392 	register_t linkmax, namemax, chownres, notrunc;
3393 	char *cp2;
3394 	struct mbuf *mb, *mreq;
3395 	struct vnode *vp;
3396 	struct vattr at;
3397 	nfsfh_t nfh;
3398 	fhandle_t *fhp;
3399 	u_quad_t frev;
3400 
3401 	fhp = &nfh.fh_generic;
3402 	nfsm_srvmtofh(fhp);
3403 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3404 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3405 	if (error) {
3406 		nfsm_reply(NFSX_UNSIGNED);
3407 		nfsm_srvpostop_attr(getret, &at);
3408 		return (0);
3409 	}
3410 	error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3411 	if (!error)
3412 		error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3413 	if (!error)
3414 		error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3415 	if (!error)
3416 		error = VOP_PATHCONF(vp, _PC_NO_TRUNC, &notrunc);
3417 	getret = VOP_GETATTR(vp, &at, cred, lwp);
3418 	vput(vp);
3419 	nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3420 	nfsm_srvpostop_attr(getret, &at);
3421 	if (error)
3422 		return (0);
3423 	nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3424 
3425 	pc->pc_linkmax = txdr_unsigned(linkmax);
3426 	pc->pc_namemax = txdr_unsigned(namemax);
3427 	pc->pc_notrunc = txdr_unsigned(notrunc);
3428 	pc->pc_chownrestricted = txdr_unsigned(chownres);
3429 
3430 	/*
3431 	 * These should probably be supported by VOP_PATHCONF(), but
3432 	 * until msdosfs is exportable (why would you want to?), the
3433 	 * Unix defaults should be ok.
3434 	 */
3435 	pc->pc_caseinsensitive = nfs_false;
3436 	pc->pc_casepreserving = nfs_true;
3437 	nfsm_srvdone;
3438 }
3439 
3440 /*
3441  * Null operation, used by clients to ping server
3442  */
3443 /* ARGSUSED */
3444 int
3445 nfsrv_null(nfsd, slp, lwp, mrq)
3446 	struct nfsrv_descript *nfsd;
3447 	struct nfssvc_sock *slp;
3448 	struct lwp *lwp;
3449 	struct mbuf **mrq;
3450 {
3451 	struct mbuf *mrep = nfsd->nd_mrep;
3452 	caddr_t bpos;
3453 	int error = NFSERR_RETVOID, cache = 0;
3454 	struct mbuf *mb, *mreq;
3455 	u_quad_t frev;
3456 
3457 	nfsm_reply(0);
3458 	return (0);
3459 }
3460 
3461 /*
3462  * No operation, used for obsolete procedures
3463  */
3464 /* ARGSUSED */
3465 int
3466 nfsrv_noop(nfsd, slp, lwp, mrq)
3467 	struct nfsrv_descript *nfsd;
3468 	struct nfssvc_sock *slp;
3469 	struct lwp *lwp;
3470 	struct mbuf **mrq;
3471 {
3472 	struct mbuf *mrep = nfsd->nd_mrep;
3473 	caddr_t bpos;
3474 	int error, cache = 0;
3475 	struct mbuf *mb, *mreq;
3476 	u_quad_t frev;
3477 
3478 	if (nfsd->nd_repstat)
3479 		error = nfsd->nd_repstat;
3480 	else
3481 		error = EPROCUNAVAIL;
3482 	nfsm_reply(0);
3483 	return (0);
3484 }
3485 
3486 /*
3487  * Perform access checking for vnodes obtained from file handles that would
3488  * refer to files already opened by a Unix client. You cannot just use
3489  * vn_writechk() and VOP_ACCESS() for two reasons.
3490  * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3491  * 2 - The owner is to be given access irrespective of mode bits for some
3492  *     operations, so that processes that chmod after opening a file don't
3493  *     break. I don't like this because it opens a security hole, but since
3494  *     the nfs server opens a security hole the size of a barn door anyhow,
3495  *     what the heck.
3496  *
3497  * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3498  * will return EPERM instead of EACCESS. EPERM is always an error.
3499  */
3500 int
3501 nfsrv_access(vp, flags, cred, rdonly, lwp, override)
3502 	struct vnode *vp;
3503 	int flags;
3504 	kauth_cred_t cred;
3505 	int rdonly;
3506 	struct lwp *lwp;
3507 {
3508 	struct vattr vattr;
3509 	int error;
3510 	if (flags & VWRITE) {
3511 		/* Just vn_writechk() changed to check rdonly */
3512 		/*
3513 		 * Disallow write attempts on read-only file systems;
3514 		 * unless the file is a socket or a block or character
3515 		 * device resident on the file system.
3516 		 */
3517 		if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3518 			switch (vp->v_type) {
3519 			case VREG:
3520 			case VDIR:
3521 			case VLNK:
3522 				return (EROFS);
3523 			default:
3524 				break;
3525 			}
3526 		}
3527 
3528 		/*
3529 		 * If the vnode is in use as a process's text,
3530 		 * we can't allow writing.
3531 		 */
3532 		if (vp->v_flag & VTEXT)
3533 			return (ETXTBSY);
3534 	}
3535 	error = VOP_GETATTR(vp, &vattr, cred, lwp);
3536 	if (error)
3537 		return (error);
3538 	error = VOP_ACCESS(vp, flags, cred, lwp);
3539 	/*
3540 	 * Allow certain operations for the owner (reads and writes
3541 	 * on files that are already open).
3542 	 */
3543 	if (override && error == EACCES && kauth_cred_geteuid(cred) == vattr.va_uid)
3544 		error = 0;
3545 	return error;
3546 }
3547