xref: /netbsd-src/sys/nfs/nfs_serv.c (revision 5b84b3983f71fd20a534cfa5d1556623a8aaa717)
1 /*	$NetBSD: nfs_serv.c,v 1.97 2005/09/06 09:36:28 jmmv 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 1.97 2005/09/06 09:36:28 jmmv 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 <ufs/ufs/dir.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, procp, mrq)
100 	struct nfsrv_descript *nfsd;
101 	struct nfssvc_sock *slp;
102 	struct proc *procp;
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 	struct ucred *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, procp, 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, procp, 0) == 0)
141 			outmode |= testmode;
142 		if ((inmode & NFSV3ACCESS_EXECUTE) &&
143 		    nfsrv_access(vp, VEXEC, cred, rdonly, procp, 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, procp, 0) == 0)
150 			outmode |= testmode;
151 		if ((inmode & NFSV3ACCESS_LOOKUP) &&
152 		    nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0) == 0)
153 			outmode |= NFSV3ACCESS_LOOKUP;
154 	}
155 	getret = VOP_GETATTR(vp, &va, cred, procp);
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, procp, mrq)
169 	struct nfsrv_descript *nfsd;
170 	struct nfssvc_sock *slp;
171 	struct proc *procp;
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 	struct ucred *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;
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, procp);
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, procp, mrq)
215 	struct nfsrv_descript *nfsd;
216 	struct nfssvc_sock *slp;
217 	struct proc *procp;
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 	struct ucred *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, 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 	fhp = &nfh.fh_generic;
242 	nfsm_srvmtofh(fhp);
243 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
244 		return (ESTALE);
245 	vn_start_write(NULL, &mp, V_WAIT);
246 	VATTR_NULL(&va);
247 	if (v3) {
248 		nfsm_srvsattr(&va);
249 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
250 		gcheck = fxdr_unsigned(int, *tl);
251 		if (gcheck) {
252 			nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
253 			fxdr_nfsv3time(tl, &guard);
254 		}
255 	} else {
256 		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
257 		/*
258 		 * Nah nah nah nah na nah
259 		 * There is a bug in the Sun client that puts 0xffff in the mode
260 		 * field of sattr when it should put in 0xffffffff. The u_short
261 		 * doesn't sign extend.
262 		 * --> check the low order 2 bytes for 0xffff
263 		 */
264 		if ((fxdr_unsigned(int, sp->sa_mode) & 0xffff) != 0xffff)
265 			va.va_mode = nfstov_mode(sp->sa_mode);
266 		if (sp->sa_uid != nfs_xdrneg1)
267 			va.va_uid = fxdr_unsigned(uid_t, sp->sa_uid);
268 		if (sp->sa_gid != nfs_xdrneg1)
269 			va.va_gid = fxdr_unsigned(gid_t, sp->sa_gid);
270 		if (sp->sa_size != nfs_xdrneg1)
271 			va.va_size = fxdr_unsigned(u_quad_t, sp->sa_size);
272 		if (sp->sa_atime.nfsv2_sec != nfs_xdrneg1) {
273 #ifdef notyet
274 			fxdr_nfsv2time(&sp->sa_atime, &va.va_atime);
275 #else
276 			va.va_atime.tv_sec =
277 				fxdr_unsigned(u_int32_t,sp->sa_atime.nfsv2_sec);
278 			va.va_atime.tv_nsec = 0;
279 #endif
280 		}
281 		if (sp->sa_mtime.nfsv2_sec != nfs_xdrneg1)
282 			fxdr_nfsv2time(&sp->sa_mtime, &va.va_mtime);
283 
284 	}
285 
286 	/*
287 	 * Now that we have all the fields, lets do it.
288 	 */
289 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam, &rdonly,
290 	    (nfsd->nd_flag & ND_KERBAUTH), FALSE);
291 	if (error) {
292 		nfsm_reply(2 * NFSX_UNSIGNED);
293 		nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
294 		vn_finished_write(mp, 0);
295 		return (0);
296 	}
297 	nqsrv_getl(vp, ND_WRITE);
298 	if (v3) {
299 		error = preat_ret = VOP_GETATTR(vp, &preat, cred, procp);
300 		if (!error && gcheck &&
301 			(preat.va_ctime.tv_sec != guard.tv_sec ||
302 			 preat.va_ctime.tv_nsec != guard.tv_nsec))
303 			error = NFSERR_NOT_SYNC;
304 		if (error) {
305 			vput(vp);
306 			nfsm_reply(NFSX_WCCDATA(v3));
307 			nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
308 			vn_finished_write(mp, 0);
309 			return (0);
310 		}
311 	}
312 
313 	/*
314 	 * If the size is being changed write acces is required, otherwise
315 	 * just check for a read only file system.
316 	 */
317 	if (va.va_size == ((u_quad_t)((quad_t) -1))) {
318 		if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
319 			error = EROFS;
320 			goto out;
321 		}
322 	} else {
323 		if (vp->v_type == VDIR) {
324 			error = EISDIR;
325 			goto out;
326 		} else if ((error = nfsrv_access(vp, VWRITE, cred, rdonly,
327 			procp, 0)) != 0)
328 			goto out;
329 	}
330 	error = VOP_SETATTR(vp, &va, cred, procp);
331 	postat_ret = VOP_GETATTR(vp, &va, cred, procp);
332 	if (!error)
333 		error = postat_ret;
334 out:
335 	vput(vp);
336 	vn_finished_write(mp, 0);
337 	nfsm_reply(NFSX_WCCORFATTR(v3));
338 	if (v3) {
339 		nfsm_srvwcc_data(preat_ret, &preat, postat_ret, &va);
340 		return (0);
341 	} else {
342 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
343 		nfsm_srvfillattr(&va, fp);
344 	}
345 	nfsm_srvdone;
346 }
347 
348 /*
349  * nfs lookup rpc
350  */
351 int
352 nfsrv_lookup(nfsd, slp, procp, mrq)
353 	struct nfsrv_descript *nfsd;
354 	struct nfssvc_sock *slp;
355 	struct proc *procp;
356 	struct mbuf **mrq;
357 {
358 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
359 	struct mbuf *nam = nfsd->nd_nam;
360 	caddr_t dpos = nfsd->nd_dpos;
361 	struct ucred *cred = &nfsd->nd_cr;
362 	struct nfs_fattr *fp;
363 	struct nameidata nd, ind, *ndp = &nd;
364 	struct vnode *vp, *dirp;
365 	nfsfh_t nfh;
366 	fhandle_t *fhp;
367 	caddr_t cp;
368 	u_int32_t *tl;
369 	int32_t t1;
370 	caddr_t bpos;
371 	int error = 0, cache, dirattr_ret = 1;
372 	uint32_t len;
373 	int v3 = (nfsd->nd_flag & ND_NFSV3), pubflag;
374 	char *cp2;
375 	struct mbuf *mb, *mreq;
376 	struct vattr va, dirattr;
377 	u_quad_t frev;
378 
379 	fhp = &nfh.fh_generic;
380 	nfsm_srvmtofh(fhp);
381 	nfsm_srvnamesiz(len);
382 
383 	pubflag = nfs_ispublicfh(fhp);
384 
385 	nd.ni_cnd.cn_cred = cred;
386 	nd.ni_cnd.cn_nameiop = LOOKUP;
387 	nd.ni_cnd.cn_flags = LOCKLEAF | SAVESTART;
388 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
389 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), pubflag);
390 
391 	if (!error && pubflag) {
392 		if (nd.ni_vp->v_type == VDIR && nfs_pub.np_index != NULL) {
393 			/*
394 			 * Setup call to lookup() to see if we can find
395 			 * the index file. Arguably, this doesn't belong
396 			 * in a kernel.. Ugh.
397 			 */
398 			ind = nd;
399 			VOP_UNLOCK(nd.ni_vp, 0);
400 			ind.ni_pathlen = strlen(nfs_pub.np_index);
401 			ind.ni_cnd.cn_nameptr = ind.ni_cnd.cn_pnbuf =
402 			    nfs_pub.np_index;
403 			ind.ni_startdir = nd.ni_vp;
404 			VREF(ind.ni_startdir);
405 			error = lookup(&ind);
406 			if (!error) {
407 				/*
408 				 * Found an index file. Get rid of
409 				 * the old references.
410 				 */
411 				if (dirp)
412 					vrele(dirp);
413 				dirp = nd.ni_vp;
414 				vrele(nd.ni_startdir);
415 				ndp = &ind;
416 			} else
417 				error = 0;
418 		}
419 		/*
420 		 * If the public filehandle was used, check that this lookup
421 		 * didn't result in a filehandle outside the publicly exported
422 		 * filesystem.
423 		 */
424 
425 		if (!error && ndp->ni_vp->v_mount != nfs_pub.np_mount) {
426 			vput(nd.ni_vp);
427 			error = EPERM;
428 		}
429 	}
430 
431 	if (dirp) {
432 		if (v3)
433 			dirattr_ret = VOP_GETATTR(dirp, &dirattr, cred,
434 				procp);
435 		vrele(dirp);
436 	}
437 
438 	if (error) {
439 		nfsm_reply(NFSX_POSTOPATTR(v3));
440 		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
441 		return (0);
442 	}
443 
444 	nqsrv_getl(ndp->ni_startdir, ND_READ);
445 	vrele(ndp->ni_startdir);
446 	PNBUF_PUT(nd.ni_cnd.cn_pnbuf);
447 	vp = ndp->ni_vp;
448 	memset((caddr_t)fhp, 0, sizeof(nfh));
449 	fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
450 	error = VFS_VPTOFH(vp, &fhp->fh_fid);
451 	if (!error)
452 		error = VOP_GETATTR(vp, &va, cred, procp);
453 	vput(vp);
454 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPORFATTR(v3) + NFSX_POSTOPATTR(v3));
455 	if (error) {
456 		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
457 		return (0);
458 	}
459 	nfsm_srvfhtom(fhp, v3);
460 	if (v3) {
461 		nfsm_srvpostop_attr(0, &va);
462 		nfsm_srvpostop_attr(dirattr_ret, &dirattr);
463 	} else {
464 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
465 		nfsm_srvfillattr(&va, fp);
466 	}
467 	nfsm_srvdone;
468 }
469 
470 /*
471  * nfs readlink service
472  */
473 int
474 nfsrv_readlink(nfsd, slp, procp, mrq)
475 	struct nfsrv_descript *nfsd;
476 	struct nfssvc_sock *slp;
477 	struct proc *procp;
478 	struct mbuf **mrq;
479 {
480 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
481 	struct mbuf *nam = nfsd->nd_nam;
482 	caddr_t dpos = nfsd->nd_dpos;
483 	struct ucred *cred = &nfsd->nd_cr;
484 	struct iovec iv[(NFS_MAXPATHLEN+MLEN-1)/MLEN];
485 	struct iovec *ivp = iv;
486 	struct mbuf *mp;
487 	u_int32_t *tl;
488 	int32_t t1;
489 	caddr_t bpos;
490 	int error = 0, rdonly, cache, i, padlen, getret;
491 	uint32_t len;
492 	int v3 = (nfsd->nd_flag & ND_NFSV3);
493 	char *cp2;
494 	struct mbuf *mb, *mp2 = NULL, *mp3 = NULL, *mreq;
495 	struct vnode *vp;
496 	struct vattr attr;
497 	nfsfh_t nfh;
498 	fhandle_t *fhp;
499 	struct uio io, *uiop = &io;
500 	u_quad_t frev;
501 
502 	fhp = &nfh.fh_generic;
503 	nfsm_srvmtofh(fhp);
504 	len = 0;
505 	i = 0;
506 	while (len < NFS_MAXPATHLEN) {
507 		mp = m_get(M_WAIT, MT_DATA);
508 		MCLAIM(mp, &nfs_mowner);
509 		m_clget(mp, M_WAIT);
510 		mp->m_len = NFSMSIZ(mp);
511 		if (len == 0)
512 			mp3 = mp2 = mp;
513 		else {
514 			mp2->m_next = mp;
515 			mp2 = mp;
516 		}
517 		if ((len+mp->m_len) > NFS_MAXPATHLEN) {
518 			mp->m_len = NFS_MAXPATHLEN-len;
519 			len = NFS_MAXPATHLEN;
520 		} else
521 			len += mp->m_len;
522 		ivp->iov_base = mtod(mp, caddr_t);
523 		ivp->iov_len = mp->m_len;
524 		i++;
525 		ivp++;
526 	}
527 	uiop->uio_iov = iv;
528 	uiop->uio_iovcnt = i;
529 	uiop->uio_offset = 0;
530 	uiop->uio_resid = len;
531 	uiop->uio_rw = UIO_READ;
532 	uiop->uio_segflg = UIO_SYSSPACE;
533 	uiop->uio_procp = NULL;
534 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
535 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
536 	if (error) {
537 		m_freem(mp3);
538 		nfsm_reply(2 * NFSX_UNSIGNED);
539 		nfsm_srvpostop_attr(1, (struct vattr *)0);
540 		return (0);
541 	}
542 	if (vp->v_type != VLNK) {
543 		if (v3)
544 			error = EINVAL;
545 		else
546 			error = ENXIO;
547 		goto out;
548 	}
549 	nqsrv_getl(vp, ND_READ);
550 	error = VOP_READLINK(vp, uiop, cred);
551 out:
552 	getret = VOP_GETATTR(vp, &attr, cred, procp);
553 	vput(vp);
554 	if (error)
555 		m_freem(mp3);
556 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_UNSIGNED);
557 	if (v3) {
558 		nfsm_srvpostop_attr(getret, &attr);
559 		if (error)
560 			return (0);
561 	}
562 	len -= uiop->uio_resid;
563 	padlen = nfsm_padlen(len);
564 	if (uiop->uio_resid || padlen)
565 		nfs_zeropad(mp3, uiop->uio_resid, padlen);
566 	nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
567 	*tl = txdr_unsigned(len);
568 	mb->m_next = mp3;
569 	nfsm_srvdone;
570 }
571 
572 /*
573  * nfs read service
574  */
575 int
576 nfsrv_read(nfsd, slp, procp, mrq)
577 	struct nfsrv_descript *nfsd;
578 	struct nfssvc_sock *slp;
579 	struct proc *procp;
580 	struct mbuf **mrq;
581 {
582 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
583 	struct mbuf *nam = nfsd->nd_nam;
584 	caddr_t dpos = nfsd->nd_dpos;
585 	struct ucred *cred = &nfsd->nd_cr;
586 	struct mbuf *m;
587 	struct nfs_fattr *fp;
588 	u_int32_t *tl;
589 	int32_t t1;
590 	int i;
591 	caddr_t bpos;
592 	int error = 0, rdonly, cache, getret;
593 	int v3 = (nfsd->nd_flag & ND_NFSV3);
594 	uint32_t reqlen, len, cnt, left;
595 	int padlen;
596 	char *cp2;
597 	struct mbuf *mb, *mreq;
598 	struct vnode *vp;
599 	nfsfh_t nfh;
600 	fhandle_t *fhp;
601 	struct uio io, *uiop = &io;
602 	struct vattr va;
603 	off_t off;
604 	u_quad_t frev;
605 
606 	fhp = &nfh.fh_generic;
607 	nfsm_srvmtofh(fhp);
608 	if (v3) {
609 		nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
610 		off = fxdr_hyper(tl);
611 	} else {
612 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
613 		off = (off_t)fxdr_unsigned(u_int32_t, *tl);
614 	}
615 	nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
616 	reqlen = fxdr_unsigned(uint32_t, *tl);
617 	reqlen = MIN(reqlen, NFS_SRVMAXDATA(nfsd));
618 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
619 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
620 	if (error) {
621 		nfsm_reply(2 * NFSX_UNSIGNED);
622 		nfsm_srvpostop_attr(1, (struct vattr *)0);
623 		return (0);
624 	}
625 	if (vp->v_type != VREG) {
626 		if (v3)
627 			error = EINVAL;
628 		else
629 			error = (vp->v_type == VDIR) ? EISDIR : EACCES;
630 	}
631 	if (!error) {
632 	    nqsrv_getl(vp, ND_READ);
633 	    if ((error = nfsrv_access(vp, VREAD, cred, rdonly, procp, 1)) != 0)
634 		error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 1);
635 	}
636 	getret = VOP_GETATTR(vp, &va, cred, procp);
637 	if (!error)
638 		error = getret;
639 	if (error) {
640 		vput(vp);
641 		nfsm_reply(NFSX_POSTOPATTR(v3));
642 		nfsm_srvpostop_attr(getret, &va);
643 		return (0);
644 	}
645 	if (off >= va.va_size)
646 		cnt = 0;
647 	else if ((off + reqlen) > va.va_size)
648 		cnt = va.va_size - off;
649 	else
650 		cnt = reqlen;
651 	nfsm_reply(NFSX_POSTOPORFATTR(v3) + 3 * NFSX_UNSIGNED+nfsm_rndup(cnt));
652 	if (v3) {
653 		nfsm_build(tl, u_int32_t *, NFSX_V3FATTR + 4 * NFSX_UNSIGNED);
654 		*tl++ = nfs_true;
655 		fp = (struct nfs_fattr *)tl;
656 		tl += (NFSX_V3FATTR / sizeof (u_int32_t));
657 	} else {
658 		nfsm_build(tl, u_int32_t *, NFSX_V2FATTR + NFSX_UNSIGNED);
659 		fp = (struct nfs_fattr *)tl;
660 		tl += (NFSX_V2FATTR / sizeof (u_int32_t));
661 	}
662 	len = left = cnt;
663 	if (cnt > 0) {
664 		if (nfsd_use_loan) {
665 			struct vm_page **pgpp;
666 			voff_t pgoff = trunc_page(off);
667 			int npages;
668 			vaddr_t lva;
669 
670 			npages = (round_page(off + cnt) - pgoff) >> PAGE_SHIFT;
671 			KASSERT(npages <= M_EXT_MAXPAGES); /* XXX */
672 
673 			/* allocate kva for mbuf data */
674 			lva = sokvaalloc(npages << PAGE_SHIFT, slp->ns_so);
675 			if (lva == 0) {
676 				/* fall back to VOP_READ */
677 				goto loan_fail;
678 			}
679 
680 			/* allocate mbuf */
681 			m = m_get(M_WAIT, MT_DATA);
682 			MCLAIM(m, &nfs_mowner);
683 			pgpp = m->m_ext.ext_pgs;
684 
685 			/* loan pages */
686 			error = uvm_loanuobjpages(&vp->v_uobj, pgoff, npages,
687 			    pgpp);
688 			if (error) {
689 				sokvafree(lva, npages << PAGE_SHIFT);
690 				m_free(m);
691 				if (error == EBUSY)
692 					goto loan_fail;
693 				goto read_error;
694 			}
695 
696 			/* associate kva to mbuf */
697 			MEXTADD(m, (void *)(lva + ((vaddr_t)off & PAGE_MASK)),
698 			    cnt, M_MBUF, soloanfree, slp->ns_so);
699 			m->m_flags |= M_EXT_PAGES | M_EXT_ROMAP;
700 			m->m_len = cnt;
701 
702 			/* map pages */
703 			for (i = 0; i < npages; i++) {
704 				pmap_kenter_pa(lva, VM_PAGE_TO_PHYS(pgpp[i]),
705 				    VM_PROT_READ);
706 				lva += PAGE_SIZE;
707 			}
708 
709 			pmap_update(pmap_kernel());
710 
711 			mb->m_next = m;
712 			mb = m;
713 			error = 0;
714 			uiop->uio_resid = 0;
715 		} else {
716 			struct iovec *iv;
717 			struct iovec *iv2;
718 			struct mbuf *m2;
719 			int siz;
720 loan_fail:
721 			/*
722 			 * Generate the mbuf list with the uio_iov ref. to it.
723 			 */
724 			i = 0;
725 			m = m2 = mb;
726 			while (left > 0) {
727 				siz = min(M_TRAILINGSPACE(m), left);
728 				if (siz > 0) {
729 					left -= siz;
730 					i++;
731 				}
732 				if (left > 0) {
733 					m = m_get(M_WAIT, MT_DATA);
734 					MCLAIM(m, &nfs_mowner);
735 					m_clget(m, M_WAIT);
736 					m->m_len = 0;
737 					m2->m_next = m;
738 					m2 = m;
739 				}
740 			}
741 			iv = malloc(i * sizeof(struct iovec), M_TEMP, M_WAITOK);
742 			uiop->uio_iov = iv2 = iv;
743 			m = mb;
744 			left = cnt;
745 			i = 0;
746 			while (left > 0) {
747 				if (m == NULL)
748 					panic("nfsrv_read iov");
749 				siz = min(M_TRAILINGSPACE(m), left);
750 				if (siz > 0) {
751 					iv->iov_base = mtod(m, caddr_t) +
752 					    m->m_len;
753 					iv->iov_len = siz;
754 					m->m_len += siz;
755 					left -= siz;
756 					iv++;
757 					i++;
758 				}
759 				m = m->m_next;
760 			}
761 			uiop->uio_iovcnt = i;
762 			uiop->uio_offset = off;
763 			uiop->uio_resid = cnt;
764 			uiop->uio_rw = UIO_READ;
765 			uiop->uio_segflg = UIO_SYSSPACE;
766 			error = VOP_READ(vp, uiop, IO_NODELOCKED, cred);
767 			free((caddr_t)iv2, M_TEMP);
768 		}
769 read_error:
770 		if (error || (getret = VOP_GETATTR(vp, &va, cred, procp)) != 0){
771 			if (!error)
772 				error = getret;
773 			m_freem(mreq);
774 			vput(vp);
775 			nfsm_reply(NFSX_POSTOPATTR(v3));
776 			nfsm_srvpostop_attr(getret, &va);
777 			return (0);
778 		}
779 	} else {
780 		uiop->uio_resid = 0;
781 	}
782 	vput(vp);
783 	nfsm_srvfillattr(&va, fp);
784 	len -= uiop->uio_resid;
785 	padlen = nfsm_padlen(len);
786 	if (uiop->uio_resid || padlen)
787 		nfs_zeropad(mb, uiop->uio_resid, padlen);
788 	if (v3) {
789 		/* count */
790 		*tl++ = txdr_unsigned(len);
791 		/* eof */
792 		if (off + len >= va.va_size)
793 			*tl++ = nfs_true;
794 		else
795 			*tl++ = nfs_false;
796 	}
797 	*tl = txdr_unsigned(len);
798 	nfsm_srvdone;
799 }
800 
801 /*
802  * nfs write service
803  */
804 int
805 nfsrv_write(nfsd, slp, procp, mrq)
806 	struct nfsrv_descript *nfsd;
807 	struct nfssvc_sock *slp;
808 	struct proc *procp;
809 	struct mbuf **mrq;
810 {
811 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
812 	struct mbuf *nam = nfsd->nd_nam;
813 	caddr_t dpos = nfsd->nd_dpos;
814 	struct ucred *cred = &nfsd->nd_cr;
815 	struct iovec *ivp;
816 	int i, cnt;
817 	struct mbuf *mp;
818 	struct nfs_fattr *fp;
819 	struct iovec *iv;
820 	struct vattr va, forat;
821 	u_int32_t *tl;
822 	int32_t t1;
823 	caddr_t bpos;
824 	int error = 0, rdonly, cache, len, forat_ret = 1;
825 	int ioflags, aftat_ret = 1, retlen, zeroing, adjust;
826 	int stable = NFSV3WRITE_FILESYNC;
827 	int v3 = (nfsd->nd_flag & ND_NFSV3);
828 	char *cp2;
829 	struct mbuf *mb, *mreq;
830 	struct vnode *vp;
831 	nfsfh_t nfh;
832 	fhandle_t *fhp;
833 	struct uio io, *uiop = &io;
834 	off_t off;
835 	u_quad_t frev;
836 	struct mount *mntp = NULL;
837 
838 	if (mrep == NULL) {
839 		*mrq = NULL;
840 		return (0);
841 	}
842 	fhp = &nfh.fh_generic;
843 	nfsm_srvmtofh(fhp);
844 	if ((mntp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
845 		return (ESTALE);
846 	vn_start_write(NULL, &mntp, V_WAIT);
847 	if (v3) {
848 		nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
849 		off = fxdr_hyper(tl);
850 		tl += 3;
851 		stable = fxdr_unsigned(int, *tl++);
852 	} else {
853 		nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
854 		off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
855 		tl += 2;
856 	}
857 	retlen = len = fxdr_unsigned(int32_t, *tl);
858 	cnt = i = 0;
859 
860 	/*
861 	 * For NFS Version 2, it is not obvious what a write of zero length
862 	 * should do, but I might as well be consistent with Version 3,
863 	 * which is to return ok so long as there are no permission problems.
864 	 */
865 	if (len > 0) {
866 		zeroing = 1;
867 		mp = mrep;
868 		while (mp) {
869 			if (mp == md) {
870 				zeroing = 0;
871 				adjust = dpos - mtod(mp, caddr_t);
872 				mp->m_len -= adjust;
873 				if (mp->m_len > 0 && adjust > 0)
874 					NFSMADV(mp, adjust);
875 			}
876 			if (zeroing)
877 				mp->m_len = 0;
878 			else if (mp->m_len > 0) {
879 				i += mp->m_len;
880 				if (i > len) {
881 					mp->m_len -= (i - len);
882 					zeroing	= 1;
883 				}
884 				if (mp->m_len > 0)
885 					cnt++;
886 			}
887 			mp = mp->m_next;
888 		}
889 	}
890 	if (len > NFS_MAXDATA || len < 0 || i < len) {
891 		error = EIO;
892 		nfsm_reply(2 * NFSX_UNSIGNED);
893 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
894 		vn_finished_write(mntp, 0);
895 		return (0);
896 	}
897 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
898 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
899 	if (error) {
900 		nfsm_reply(2 * NFSX_UNSIGNED);
901 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
902 		vn_finished_write(mntp, 0);
903 		return (0);
904 	}
905 	if (v3)
906 		forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
907 	if (vp->v_type != VREG) {
908 		if (v3)
909 			error = EINVAL;
910 		else
911 			error = (vp->v_type == VDIR) ? EISDIR : EACCES;
912 	}
913 	if (!error) {
914 		nqsrv_getl(vp, ND_WRITE);
915 		error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
916 	}
917 	if (error) {
918 		vput(vp);
919 		nfsm_reply(NFSX_WCCDATA(v3));
920 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
921 		vn_finished_write(mntp, 0);
922 		return (0);
923 	}
924 
925 	if (len > 0) {
926 		ivp = malloc(cnt * sizeof (struct iovec), M_TEMP, M_WAITOK);
927 		uiop->uio_iov = iv = ivp;
928 		uiop->uio_iovcnt = cnt;
929 		mp = mrep;
930 		while (mp) {
931 			if (mp->m_len > 0) {
932 				ivp->iov_base = mtod(mp, caddr_t);
933 				ivp->iov_len = mp->m_len;
934 				ivp++;
935 			}
936 			mp = mp->m_next;
937 		}
938 
939 		/*
940 		 * XXX
941 		 * The IO_METASYNC flag indicates that all metadata (and not
942 		 * just enough to ensure data integrity) must be written to
943 		 * stable storage synchronously.
944 		 * (IO_METASYNC is not yet implemented in 4.4BSD-Lite.)
945 		 */
946 		if (stable == NFSV3WRITE_UNSTABLE)
947 			ioflags = IO_NODELOCKED;
948 		else if (stable == NFSV3WRITE_DATASYNC)
949 			ioflags = (IO_SYNC | IO_NODELOCKED);
950 		else
951 			ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
952 		uiop->uio_resid = len;
953 		uiop->uio_rw = UIO_WRITE;
954 		uiop->uio_segflg = UIO_SYSSPACE;
955 		uiop->uio_procp = NULL;
956 		uiop->uio_offset = off;
957 		error = VOP_WRITE(vp, uiop, ioflags, cred);
958 		nfsstats.srvvop_writes++;
959 		free(iv, M_TEMP);
960 	}
961 	aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
962 	vput(vp);
963 	vn_finished_write(mntp, 0);
964 	if (!error)
965 		error = aftat_ret;
966 	nfsm_reply(NFSX_PREOPATTR(v3) + NFSX_POSTOPORFATTR(v3) +
967 		2 * NFSX_UNSIGNED + NFSX_WRITEVERF(v3));
968 	if (v3) {
969 		nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
970 		if (error)
971 			return (0);
972 		nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
973 		*tl++ = txdr_unsigned(retlen);
974 		if (stable == NFSV3WRITE_UNSTABLE)
975 			*tl++ = txdr_unsigned(stable);
976 		else
977 			*tl++ = txdr_unsigned(NFSV3WRITE_FILESYNC);
978 		/*
979 		 * Actually, there is no need to txdr these fields,
980 		 * but it may make the values more human readable,
981 		 * for debugging purposes.
982 		 */
983 		*tl++ = txdr_unsigned(boottime.tv_sec);
984 		*tl = txdr_unsigned(boottime.tv_usec);
985 	} else {
986 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
987 		nfsm_srvfillattr(&va, fp);
988 	}
989 	nfsm_srvdone;
990 }
991 
992 /*
993  * NFS write service with write gathering support. Called when
994  * nfsrvw_procrastinate > 0.
995  * See: Chet Juszczak, "Improving the Write Performance of an NFS Server",
996  * in Proc. of the Winter 1994 Usenix Conference, pg. 247-259, San Franscisco,
997  * Jan. 1994.
998  */
999 int
1000 nfsrv_writegather(ndp, slp, procp, mrq)
1001 	struct nfsrv_descript **ndp;
1002 	struct nfssvc_sock *slp;
1003 	struct proc *procp;
1004 	struct mbuf **mrq;
1005 {
1006 	struct iovec *ivp;
1007 	struct mbuf *mp;
1008 	struct nfsrv_descript *wp, *nfsd, *owp, *swp;
1009 	struct nfs_fattr *fp;
1010 	int i = 0;
1011 	struct iovec *iov;
1012 	struct nfsrvw_delayhash *wpp;
1013 	struct ucred *cred;
1014 	struct vattr va, forat;
1015 	u_int32_t *tl;
1016 	int32_t t1;
1017 	caddr_t bpos, dpos;
1018 	int error = 0, rdonly, cache, len = 0, forat_ret = 1;
1019 	int ioflags, aftat_ret = 1, s, adjust, v3, zeroing;
1020 	char *cp2;
1021 	struct mbuf *mb, *mreq, *mrep, *md;
1022 	struct vnode *vp;
1023 	struct uio io, *uiop = &io;
1024 	u_quad_t frev, cur_usec;
1025 	struct mount *mntp = NULL;
1026 
1027 	*mrq = NULL;
1028 	if (*ndp) {
1029 	    nfsd = *ndp;
1030 	    *ndp = NULL;
1031 	    mrep = nfsd->nd_mrep;
1032 	    md = nfsd->nd_md;
1033 	    dpos = nfsd->nd_dpos;
1034 	    cred = &nfsd->nd_cr;
1035 	    v3 = (nfsd->nd_flag & ND_NFSV3);
1036 	    LIST_INIT(&nfsd->nd_coalesce);
1037 	    nfsd->nd_mreq = NULL;
1038 	    nfsd->nd_stable = NFSV3WRITE_FILESYNC;
1039 	    cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1040 	    nfsd->nd_time = cur_usec + nfsrvw_procrastinate;
1041 
1042 	    /*
1043 	     * Now, get the write header..
1044 	     */
1045 	    nfsm_srvmtofh(&nfsd->nd_fh);
1046 	    if (v3) {
1047 		nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
1048 		nfsd->nd_off = fxdr_hyper(tl);
1049 		tl += 3;
1050 		nfsd->nd_stable = fxdr_unsigned(int, *tl++);
1051 	    } else {
1052 		nfsm_dissect(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1053 		nfsd->nd_off = (off_t)fxdr_unsigned(u_int32_t, *++tl);
1054 		tl += 2;
1055 	    }
1056 	    len = fxdr_unsigned(int32_t, *tl);
1057 	    nfsd->nd_len = len;
1058 	    nfsd->nd_eoff = nfsd->nd_off + len;
1059 
1060 	    /*
1061 	     * Trim the header out of the mbuf list and trim off any trailing
1062 	     * junk so that the mbuf list has only the write data.
1063 	     */
1064 	    zeroing = 1;
1065 	    i = 0;
1066 	    mp = mrep;
1067 	    while (mp) {
1068 		if (mp == md) {
1069 		    zeroing = 0;
1070 		    adjust = dpos - mtod(mp, caddr_t);
1071 		    mp->m_len -= adjust;
1072 		    if (mp->m_len > 0 && adjust > 0)
1073 			NFSMADV(mp, adjust);
1074 		}
1075 		if (zeroing)
1076 		    mp->m_len = 0;
1077 		else {
1078 		    i += mp->m_len;
1079 		    if (i > len) {
1080 			mp->m_len -= (i - len);
1081 			zeroing = 1;
1082 		    }
1083 		}
1084 		mp = mp->m_next;
1085 	    }
1086 	    if (len > NFS_MAXDATA || len < 0  || i < len) {
1087 nfsmout:
1088 		m_freem(mrep);
1089 		error = EIO;
1090 		nfsm_writereply(2 * NFSX_UNSIGNED, v3);
1091 		if (v3)
1092 		    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1093 		nfsd->nd_mreq = mreq;
1094 		nfsd->nd_mrep = NULL;
1095 		nfsd->nd_time = 0;
1096 	    }
1097 
1098 	    /*
1099 	     * Add this entry to the hash and time queues.
1100 	     */
1101 	    s = splsoftclock();
1102 	    owp = NULL;
1103 	    wp = LIST_FIRST(&slp->ns_tq);
1104 	    while (wp && wp->nd_time < nfsd->nd_time) {
1105 		owp = wp;
1106 		wp = LIST_NEXT(wp, nd_tq);
1107 	    }
1108 	    if (owp) {
1109 		LIST_INSERT_AFTER(owp, nfsd, nd_tq);
1110 	    } else {
1111 		LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1112 	    }
1113 	    if (nfsd->nd_mrep) {
1114 		wpp = NWDELAYHASH(slp, nfsd->nd_fh.fh_fid.fid_data);
1115 		owp = NULL;
1116 		wp = LIST_FIRST(wpp);
1117 		while (wp &&
1118 		    memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1119 		    owp = wp;
1120 		    wp = LIST_NEXT(wp, nd_hash);
1121 		}
1122 		while (wp && wp->nd_off < nfsd->nd_off &&
1123 		    !memcmp((caddr_t)&nfsd->nd_fh, (caddr_t)&wp->nd_fh, NFSX_V3FH)) {
1124 		    owp = wp;
1125 		    wp = LIST_NEXT(wp, nd_hash);
1126 		}
1127 		if (owp) {
1128 		    LIST_INSERT_AFTER(owp, nfsd, nd_hash);
1129 
1130 		    /*
1131 		     * Search the hash list for overlapping entries and
1132 		     * coalesce.
1133 		     */
1134 		    for(; nfsd && NFSW_CONTIG(owp, nfsd); nfsd = wp) {
1135 			wp = LIST_NEXT(nfsd, nd_hash);
1136 			if (NFSW_SAMECRED(owp, nfsd))
1137 			    nfsrvw_coalesce(owp, nfsd);
1138 		    }
1139 		} else {
1140 		    LIST_INSERT_HEAD(wpp, nfsd, nd_hash);
1141 		}
1142 	    }
1143 	    splx(s);
1144 	}
1145 
1146 	/*
1147 	 * Now, do VOP_WRITE()s for any one(s) that need to be done now
1148 	 * and generate the associated reply mbuf list(s).
1149 	 */
1150 loop1:
1151 	cur_usec = (u_quad_t)time.tv_sec * 1000000 + (u_quad_t)time.tv_usec;
1152 	s = splsoftclock();
1153 	for (nfsd = LIST_FIRST(&slp->ns_tq); nfsd; nfsd = owp) {
1154 		owp = LIST_NEXT(nfsd, nd_tq);
1155 		if (nfsd->nd_time > cur_usec)
1156 		    break;
1157 		if (nfsd->nd_mreq)
1158 		    continue;
1159 		LIST_REMOVE(nfsd, nd_tq);
1160 		LIST_REMOVE(nfsd, nd_hash);
1161 		splx(s);
1162 		mrep = nfsd->nd_mrep;
1163 		nfsd->nd_mrep = NULL;
1164 		cred = &nfsd->nd_cr;
1165 		v3 = (nfsd->nd_flag & ND_NFSV3);
1166 		forat_ret = aftat_ret = 1;
1167 		error = nfsrv_fhtovp(&nfsd->nd_fh, 1, &vp, cred, slp,
1168 		    nfsd->nd_nam, &rdonly, (nfsd->nd_flag & ND_KERBAUTH),
1169 		    FALSE);
1170 		if (!error) {
1171 		    if (v3)
1172 			forat_ret = VOP_GETATTR(vp, &forat, cred, procp);
1173 		    if (vp->v_type != VREG) {
1174 			if (v3)
1175 			    error = EINVAL;
1176 			else
1177 			    error = (vp->v_type == VDIR) ? EISDIR : EACCES;
1178 		    }
1179 		} else
1180 		    vp = NULL;
1181 		if (!error) {
1182 		    nqsrv_getl(vp, ND_WRITE);
1183 		    error = nfsrv_access(vp, VWRITE, cred, rdonly, procp, 1);
1184 		}
1185 
1186 		if (nfsd->nd_stable == NFSV3WRITE_UNSTABLE)
1187 		    ioflags = IO_NODELOCKED;
1188 		else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC)
1189 		    ioflags = (IO_SYNC | IO_NODELOCKED);
1190 		else
1191 		    ioflags = (IO_METASYNC | IO_SYNC | IO_NODELOCKED);
1192 		uiop->uio_rw = UIO_WRITE;
1193 		uiop->uio_segflg = UIO_SYSSPACE;
1194 		uiop->uio_procp = NULL;
1195 		uiop->uio_offset = nfsd->nd_off;
1196 		uiop->uio_resid = nfsd->nd_eoff - nfsd->nd_off;
1197 		if (uiop->uio_resid > 0) {
1198 		    mp = mrep;
1199 		    i = 0;
1200 		    while (mp) {
1201 			if (mp->m_len > 0)
1202 			    i++;
1203 			mp = mp->m_next;
1204 		    }
1205 		    uiop->uio_iovcnt = i;
1206 		    iov = malloc(i * sizeof (struct iovec), M_TEMP, M_WAITOK);
1207 		    uiop->uio_iov = ivp = iov;
1208 		    mp = mrep;
1209 		    while (mp) {
1210 			if (mp->m_len > 0) {
1211 			    ivp->iov_base = mtod(mp, caddr_t);
1212 			    ivp->iov_len = mp->m_len;
1213 			    ivp++;
1214 			}
1215 			mp = mp->m_next;
1216 		    }
1217 		    if (!error) {
1218 			if (vn_start_write(vp, &mntp, V_NOWAIT) != 0) {
1219 			    VOP_UNLOCK(vp, 0);
1220 			    vn_start_write(NULL, &mntp, V_WAIT);
1221 			    vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
1222 			}
1223 			if (!error) {
1224 			    error = VOP_WRITE(vp, uiop, ioflags, cred);
1225 			    nfsstats.srvvop_writes++;
1226 			    vn_finished_write(mntp, 0);
1227 			}
1228 		    }
1229 		    free((caddr_t)iov, M_TEMP);
1230 		}
1231 		m_freem(mrep);
1232 		if (vp) {
1233 		    aftat_ret = VOP_GETATTR(vp, &va, cred, procp);
1234 		    vput(vp);
1235 		}
1236 
1237 		/*
1238 		 * Loop around generating replies for all write rpcs that have
1239 		 * now been completed.
1240 		 */
1241 		swp = nfsd;
1242 		do {
1243 		    if (error) {
1244 			nfsm_writereply(NFSX_WCCDATA(v3), v3);
1245 			if (v3) {
1246 			    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1247 			}
1248 		    } else {
1249 			nfsm_writereply(NFSX_PREOPATTR(v3) +
1250 			    NFSX_POSTOPORFATTR(v3) + 2 * NFSX_UNSIGNED +
1251 			    NFSX_WRITEVERF(v3), v3);
1252 			if (v3) {
1253 			    nfsm_srvwcc_data(forat_ret, &forat, aftat_ret, &va);
1254 			    nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
1255 			    *tl++ = txdr_unsigned(nfsd->nd_len);
1256 			    *tl++ = txdr_unsigned(swp->nd_stable);
1257 			    /*
1258 			     * Actually, there is no need to txdr these fields,
1259 			     * but it may make the values more human readable,
1260 			     * for debugging purposes.
1261 			     */
1262 			    *tl++ = txdr_unsigned(boottime.tv_sec);
1263 			    *tl = txdr_unsigned(boottime.tv_usec);
1264 			} else {
1265 			    nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1266 			    nfsm_srvfillattr(&va, fp);
1267 			}
1268 		    }
1269 		    nfsd->nd_mreq = mreq;
1270 		    if (nfsd->nd_mrep)
1271 			panic("nfsrv_write: nd_mrep not free");
1272 
1273 		    /*
1274 		     * Done. Put it at the head of the timer queue so that
1275 		     * the final phase can return the reply.
1276 		     */
1277 		    s = splsoftclock();
1278 		    if (nfsd != swp) {
1279 			nfsd->nd_time = 0;
1280 			LIST_INSERT_HEAD(&slp->ns_tq, nfsd, nd_tq);
1281 		    }
1282 		    nfsd = LIST_FIRST(&swp->nd_coalesce);
1283 		    if (nfsd) {
1284 			LIST_REMOVE(nfsd, nd_tq);
1285 		    }
1286 		    splx(s);
1287 		} while (nfsd);
1288 		s = splsoftclock();
1289 		swp->nd_time = 0;
1290 		LIST_INSERT_HEAD(&slp->ns_tq, swp, nd_tq);
1291 		splx(s);
1292 		goto loop1;
1293 	}
1294 	splx(s);
1295 
1296 	/*
1297 	 * Search for a reply to return.
1298 	 */
1299 	s = splsoftclock();
1300 	LIST_FOREACH(nfsd, &slp->ns_tq, nd_tq) {
1301 		if (nfsd->nd_mreq) {
1302 		    LIST_REMOVE(nfsd, nd_tq);
1303 		    *mrq = nfsd->nd_mreq;
1304 		    *ndp = nfsd;
1305 		    break;
1306 		}
1307 	}
1308 	splx(s);
1309 	return (0);
1310 }
1311 
1312 /*
1313  * Coalesce the write request nfsd into owp. To do this we must:
1314  * - remove nfsd from the queues
1315  * - merge nfsd->nd_mrep into owp->nd_mrep
1316  * - update the nd_eoff and nd_stable for owp
1317  * - put nfsd on owp's nd_coalesce list
1318  * NB: Must be called at splsoftclock().
1319  */
1320 void
1321 nfsrvw_coalesce(owp, nfsd)
1322         struct nfsrv_descript *owp;
1323         struct nfsrv_descript *nfsd;
1324 {
1325         int overlap;
1326         struct mbuf *mp;
1327 	struct nfsrv_descript *m;
1328 
1329         LIST_REMOVE(nfsd, nd_hash);
1330         LIST_REMOVE(nfsd, nd_tq);
1331         if (owp->nd_eoff < nfsd->nd_eoff) {
1332             overlap = owp->nd_eoff - nfsd->nd_off;
1333             if (overlap < 0)
1334                 panic("nfsrv_coalesce: bad off");
1335             if (overlap > 0)
1336                 m_adj(nfsd->nd_mrep, overlap);
1337             mp = owp->nd_mrep;
1338             while (mp->m_next)
1339                 mp = mp->m_next;
1340             mp->m_next = nfsd->nd_mrep;
1341             owp->nd_eoff = nfsd->nd_eoff;
1342         } else
1343             m_freem(nfsd->nd_mrep);
1344         nfsd->nd_mrep = NULL;
1345         if (nfsd->nd_stable == NFSV3WRITE_FILESYNC)
1346             owp->nd_stable = NFSV3WRITE_FILESYNC;
1347         else if (nfsd->nd_stable == NFSV3WRITE_DATASYNC &&
1348             owp->nd_stable == NFSV3WRITE_UNSTABLE)
1349             owp->nd_stable = NFSV3WRITE_DATASYNC;
1350         LIST_INSERT_HEAD(&owp->nd_coalesce, nfsd, nd_tq);
1351  	/*
1352  	 * nfsd might hold coalesce elements! Move them to owp.
1353  	 * Otherwise, requests may be lost and clients will be stuck.
1354  	 */
1355 	while ((m = LIST_FIRST(&nfsd->nd_coalesce)) != NULL) {
1356 		LIST_REMOVE(m, nd_tq);
1357 		LIST_INSERT_HEAD(&owp->nd_coalesce, m, nd_tq);
1358 	}
1359 }
1360 
1361 /*
1362  * nfs create service
1363  * now does a truncate to 0 length via. setattr if it already exists
1364  */
1365 int
1366 nfsrv_create(nfsd, slp, procp, mrq)
1367 	struct nfsrv_descript *nfsd;
1368 	struct nfssvc_sock *slp;
1369 	struct proc *procp;
1370 	struct mbuf **mrq;
1371 {
1372 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1373 	struct mbuf *nam = nfsd->nd_nam;
1374 	caddr_t dpos = nfsd->nd_dpos;
1375 	struct ucred *cred = &nfsd->nd_cr;
1376 	struct nfs_fattr *fp;
1377 	struct vattr va, dirfor, diraft;
1378 	struct nfsv2_sattr *sp;
1379 	u_int32_t *tl;
1380 	struct nameidata nd;
1381 	caddr_t cp;
1382 	int32_t t1;
1383 	caddr_t bpos;
1384 	int error = 0, cache, len, tsize, dirfor_ret = 1, diraft_ret = 1;
1385 	int rdev = 0;
1386 	int v3 = (nfsd->nd_flag & ND_NFSV3), how, exclusive_flag = 0;
1387 	char *cp2;
1388 	struct mbuf *mb, *mreq;
1389 	struct vnode *vp = NULL, *dirp = NULL;
1390 	nfsfh_t nfh;
1391 	fhandle_t *fhp;
1392 	u_quad_t frev, tempsize;
1393 	u_char cverf[NFSX_V3CREATEVERF];
1394 	struct mount *mp = NULL;
1395 
1396 	nd.ni_cnd.cn_nameiop = 0;
1397 	fhp = &nfh.fh_generic;
1398 	nfsm_srvmtofh(fhp);
1399 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1400 		return (ESTALE);
1401 	vn_start_write(NULL, &mp, V_WAIT);
1402 	nfsm_srvnamesiz(len);
1403 	nd.ni_cnd.cn_cred = cred;
1404 	nd.ni_cnd.cn_nameiop = CREATE;
1405 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1406 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1407 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1408 	if (dirp) {
1409 		if (v3)
1410 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1411 				procp);
1412 		else {
1413 			vrele(dirp);
1414 			dirp = (struct vnode *)0;
1415 		}
1416 	}
1417 	if (error) {
1418 		nfsm_reply(NFSX_WCCDATA(v3));
1419 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1420 		if (dirp)
1421 			vrele(dirp);
1422 		vn_finished_write(mp, 0);
1423 		return (0);
1424 	}
1425 	VATTR_NULL(&va);
1426 	if (v3) {
1427 		va.va_mode = 0;
1428 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1429 		how = fxdr_unsigned(int, *tl);
1430 		switch (how) {
1431 		case NFSV3CREATE_GUARDED:
1432 			if (nd.ni_vp) {
1433 				error = EEXIST;
1434 				break;
1435 			}
1436 		case NFSV3CREATE_UNCHECKED:
1437 			nfsm_srvsattr(&va);
1438 			break;
1439 		case NFSV3CREATE_EXCLUSIVE:
1440 			nfsm_dissect(cp, caddr_t, NFSX_V3CREATEVERF);
1441 			memcpy(cverf, cp, NFSX_V3CREATEVERF);
1442 			exclusive_flag = 1;
1443 			break;
1444 		};
1445 		va.va_type = VREG;
1446 	} else {
1447 		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
1448 		va.va_type = IFTOVT(fxdr_unsigned(u_int32_t, sp->sa_mode));
1449 		if (va.va_type == VNON)
1450 			va.va_type = VREG;
1451 		va.va_mode = nfstov_mode(sp->sa_mode);
1452 		switch (va.va_type) {
1453 		case VREG:
1454 			tsize = fxdr_unsigned(int32_t, sp->sa_size);
1455 			if (tsize != -1)
1456 				va.va_size = (u_quad_t)tsize;
1457 			break;
1458 		case VCHR:
1459 		case VBLK:
1460 		case VFIFO:
1461 			rdev = fxdr_unsigned(int32_t, sp->sa_size);
1462 			break;
1463 		default:
1464 			break;
1465 		};
1466 	}
1467 
1468 	/*
1469 	 * Iff doesn't exist, create it
1470 	 * otherwise just truncate to 0 length
1471 	 *   should I set the mode too ??
1472 	 */
1473 	if (nd.ni_vp == NULL) {
1474 		if (va.va_type == VREG || va.va_type == VSOCK) {
1475 			nqsrv_getl(nd.ni_dvp, ND_WRITE);
1476 			error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1477 			if (!error) {
1478 				if (exclusive_flag) {
1479 					exclusive_flag = 0;
1480 					VATTR_NULL(&va);
1481 					/*
1482 					 * XXX
1483 					 * assuming NFSX_V3CREATEVERF
1484 					 * == sizeof(nfstime3)
1485 					 */
1486 					fxdr_nfsv3time(cverf, &va.va_atime);
1487 					error = VOP_SETATTR(nd.ni_vp, &va, cred,
1488 						procp);
1489 				}
1490 			}
1491 		} else if (va.va_type == VCHR || va.va_type == VBLK ||
1492 			va.va_type == VFIFO) {
1493 			if (va.va_type == VCHR && rdev == 0xffffffff)
1494 				va.va_type = VFIFO;
1495 			if (va.va_type != VFIFO &&
1496 			    (error = suser(cred, (u_short *)0))) {
1497 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1498 				vput(nd.ni_dvp);
1499 				nfsm_reply(0);
1500 				vn_finished_write(mp, 0);
1501 				return (error);
1502 			} else
1503 				va.va_rdev = (dev_t)rdev;
1504 			nqsrv_getl(nd.ni_dvp, ND_WRITE);
1505 			error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd,
1506 			    &va);
1507 			if (error) {
1508 				nfsm_reply(0);
1509 			}
1510 			if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1511 				vrele(nd.ni_dvp);
1512 				vput(nd.ni_vp);
1513 				VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1514 				error = EINVAL;
1515 				nfsm_reply(0);
1516 			}
1517 		} else {
1518 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1519 			vput(nd.ni_dvp);
1520 			error = ENXIO;
1521 		}
1522 		vp = nd.ni_vp;
1523 	} else {
1524 		vp = nd.ni_vp;
1525 		if (nd.ni_dvp == vp)
1526 			vrele(nd.ni_dvp);
1527 		else
1528 			vput(nd.ni_dvp);
1529 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1530 		if (!error && va.va_size != -1) {
1531 			error = nfsrv_access(vp, VWRITE, cred,
1532 			    (nd.ni_cnd.cn_flags & RDONLY), procp, 0);
1533 			if (!error) {
1534 				nqsrv_getl(vp, ND_WRITE);
1535 				tempsize = va.va_size;
1536 				VATTR_NULL(&va);
1537 				va.va_size = tempsize;
1538 				error = VOP_SETATTR(vp, &va, cred,
1539 					 procp);
1540 			}
1541 		}
1542 		if (error)
1543 			vput(vp);
1544 	}
1545 	if (!error) {
1546 		memset((caddr_t)fhp, 0, sizeof(nfh));
1547 		fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1548 		error = VFS_VPTOFH(vp, &fhp->fh_fid);
1549 		if (!error)
1550 			error = VOP_GETATTR(vp, &va, cred, procp);
1551 		vput(vp);
1552 	}
1553 	if (v3) {
1554 		if (exclusive_flag && !error) {
1555 			/*
1556 			 * XXX assuming NFSX_V3CREATEVERF == sizeof(nfstime3)
1557 			 */
1558 			char oldverf[NFSX_V3CREATEVERF];
1559 
1560 			txdr_nfsv3time(&va.va_atime, oldverf);
1561 			if (memcmp(cverf, oldverf, NFSX_V3CREATEVERF))
1562 				error = EEXIST;
1563 		}
1564 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1565 		vrele(dirp);
1566 	}
1567 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_FATTR(v3) + NFSX_WCCDATA(v3));
1568 	if (v3) {
1569 		if (!error) {
1570 			nfsm_srvpostop_fh(fhp);
1571 			nfsm_srvpostop_attr(0, &va);
1572 		}
1573 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1574 	} else {
1575 		nfsm_srvfhtom(fhp, v3);
1576 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
1577 		nfsm_srvfillattr(&va, fp);
1578 	}
1579 	vn_finished_write(mp, 0);
1580 	return (0);
1581 nfsmout:
1582 	if (dirp)
1583 		vrele(dirp);
1584 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1585 	if (nd.ni_dvp == nd.ni_vp)
1586 		vrele(nd.ni_dvp);
1587 	else
1588 		vput(nd.ni_dvp);
1589 	if (nd.ni_vp)
1590 		vput(nd.ni_vp);
1591 	vn_finished_write(mp, 0);
1592 	return (error);
1593 }
1594 
1595 /*
1596  * nfs v3 mknod service
1597  */
1598 int
1599 nfsrv_mknod(nfsd, slp, procp, mrq)
1600 	struct nfsrv_descript *nfsd;
1601 	struct nfssvc_sock *slp;
1602 	struct proc *procp;
1603 	struct mbuf **mrq;
1604 {
1605 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1606 	struct mbuf *nam = nfsd->nd_nam;
1607 	caddr_t dpos = nfsd->nd_dpos;
1608 	struct ucred *cred = &nfsd->nd_cr;
1609 	struct vattr va, dirfor, diraft;
1610 	u_int32_t *tl;
1611 	struct nameidata nd;
1612 	int32_t t1;
1613 	caddr_t bpos;
1614 	int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1615 	u_int32_t major, minor;
1616 	enum vtype vtyp;
1617 	char *cp2;
1618 	struct mbuf *mb, *mreq;
1619 	struct vnode *vp, *dirp = (struct vnode *)0;
1620 	nfsfh_t nfh;
1621 	fhandle_t *fhp;
1622 	u_quad_t frev;
1623 	struct mount *mp = NULL;
1624 
1625 	nd.ni_cnd.cn_nameiop = 0;
1626 	fhp = &nfh.fh_generic;
1627 	nfsm_srvmtofh(fhp);
1628 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1629 		return (ESTALE);
1630 	vn_start_write(NULL, &mp, V_WAIT);
1631 	nfsm_srvnamesiz(len);
1632 	nd.ni_cnd.cn_cred = cred;
1633 	nd.ni_cnd.cn_nameiop = CREATE;
1634 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1635 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1636 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1637 	if (dirp)
1638 		dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred, procp);
1639 	if (error) {
1640 		nfsm_reply(NFSX_WCCDATA(1));
1641 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1642 		if (dirp)
1643 			vrele(dirp);
1644 		vn_finished_write(mp, 0);
1645 		return (0);
1646 	}
1647 	nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
1648 	vtyp = nfsv3tov_type(*tl);
1649 	if (vtyp != VCHR && vtyp != VBLK && vtyp != VSOCK && vtyp != VFIFO) {
1650 		error = NFSERR_BADTYPE;
1651 		goto abort;
1652 	}
1653 	VATTR_NULL(&va);
1654 	va.va_mode = 0;
1655 	nfsm_srvsattr(&va);
1656 	if (vtyp == VCHR || vtyp == VBLK) {
1657 		dev_t rdev;
1658 
1659 		nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
1660 		major = fxdr_unsigned(u_int32_t, *tl++);
1661 		minor = fxdr_unsigned(u_int32_t, *tl);
1662 		rdev = makedev(major, minor);
1663 		if (major(rdev) != major || minor(rdev) != minor) {
1664 			error = EINVAL;
1665 			goto abort;
1666 		}
1667 		va.va_rdev = rdev;
1668 	}
1669 
1670 	/*
1671 	 * Iff doesn't exist, create it.
1672 	 */
1673 	if (nd.ni_vp) {
1674 		error = EEXIST;
1675 abort:
1676 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1677 		if (nd.ni_dvp == nd.ni_vp)
1678 			vrele(nd.ni_dvp);
1679 		else
1680 			vput(nd.ni_dvp);
1681 		if (nd.ni_vp)
1682 			vput(nd.ni_vp);
1683 		goto out;
1684 	}
1685 	va.va_type = vtyp;
1686 	if (vtyp == VSOCK) {
1687 		nqsrv_getl(nd.ni_dvp, ND_WRITE);
1688 		error = VOP_CREATE(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1689 	} else {
1690 		if (va.va_type != VFIFO &&
1691 		    (error = suser(cred, (u_short *)0))) {
1692 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1693 			vput(nd.ni_dvp);
1694 			goto out;
1695 		}
1696 		nqsrv_getl(nd.ni_dvp, ND_WRITE);
1697 		error = VOP_MKNOD(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
1698 		if (error) {
1699 			goto out;
1700 		}
1701 		if (error)
1702 			goto out;
1703 		if (nd.ni_cnd.cn_flags & ISSYMLINK) {
1704 			vput(nd.ni_vp);
1705 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1706 			error = EINVAL;
1707 		}
1708 	}
1709 out:
1710 	vp = nd.ni_vp;
1711 	if (!error) {
1712 		memset((caddr_t)fhp, 0, sizeof(nfh));
1713 		fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
1714 		error = VFS_VPTOFH(vp, &fhp->fh_fid);
1715 		if (!error)
1716 			error = VOP_GETATTR(vp, &va, cred, procp);
1717 		vput(vp);
1718 	}
1719 	diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1720 	vrele(dirp);
1721 	nfsm_reply(NFSX_SRVFH(1) + NFSX_POSTOPATTR(1) + NFSX_WCCDATA(1));
1722 	if (!error) {
1723 		nfsm_srvpostop_fh(fhp);
1724 		nfsm_srvpostop_attr(0, &va);
1725 	}
1726 	nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1727 	vn_finished_write(mp, 0);
1728 	return (0);
1729 nfsmout:
1730 	if (dirp)
1731 		vrele(dirp);
1732 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1733 	if (nd.ni_dvp == nd.ni_vp)
1734 		vrele(nd.ni_dvp);
1735 	else
1736 		vput(nd.ni_dvp);
1737 	if (nd.ni_vp)
1738 		vput(nd.ni_vp);
1739 	vn_finished_write(mp, 0);
1740 	return (error);
1741 }
1742 
1743 /*
1744  * nfs remove service
1745  */
1746 int
1747 nfsrv_remove(nfsd, slp, procp, mrq)
1748 	struct nfsrv_descript *nfsd;
1749 	struct nfssvc_sock *slp;
1750 	struct proc *procp;
1751 	struct mbuf **mrq;
1752 {
1753 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1754 	struct mbuf *nam = nfsd->nd_nam;
1755 	caddr_t dpos = nfsd->nd_dpos;
1756 	struct ucred *cred = &nfsd->nd_cr;
1757 	struct nameidata nd;
1758 	u_int32_t *tl;
1759 	int32_t t1;
1760 	caddr_t bpos;
1761 	int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
1762 	int v3 = (nfsd->nd_flag & ND_NFSV3);
1763 	char *cp2;
1764 	struct mbuf *mb, *mreq;
1765 	struct vnode *vp, *dirp;
1766 	struct vattr dirfor, diraft;
1767 	nfsfh_t nfh;
1768 	fhandle_t *fhp;
1769 	u_quad_t frev;
1770 	struct mount *mp = NULL;
1771 
1772 #ifndef nolint
1773 	vp = (struct vnode *)0;
1774 #endif
1775 	fhp = &nfh.fh_generic;
1776 	nfsm_srvmtofh(fhp);
1777 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
1778 		return (ESTALE);
1779 	vn_start_write(NULL, &mp, V_WAIT);
1780 	nfsm_srvnamesiz(len);
1781 	nd.ni_cnd.cn_cred = cred;
1782 	nd.ni_cnd.cn_nameiop = DELETE;
1783 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
1784 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
1785 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1786 	if (dirp) {
1787 		if (v3)
1788 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
1789 				procp);
1790 		else
1791 			vrele(dirp);
1792 	}
1793 	if (!error) {
1794 		vp = nd.ni_vp;
1795 		if (vp->v_type == VDIR &&
1796 		    (error = suser(cred, (u_short *)0)) != 0)
1797 			goto out;
1798 		/*
1799 		 * The root of a mounted filesystem cannot be deleted.
1800 		 */
1801 		if (vp->v_flag & VROOT) {
1802 			error = EBUSY;
1803 			goto out;
1804 		}
1805 out:
1806 		if (!error) {
1807 			nqsrv_getl(nd.ni_dvp, ND_WRITE);
1808 			nqsrv_getl(vp, ND_WRITE);
1809 			error = VOP_REMOVE(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
1810 		} else {
1811 			VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
1812 			if (nd.ni_dvp == vp)
1813 				vrele(nd.ni_dvp);
1814 			else
1815 				vput(nd.ni_dvp);
1816 			vput(vp);
1817 		}
1818 	}
1819 	if (dirp && v3) {
1820 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
1821 		vrele(dirp);
1822 	}
1823 	nfsm_reply(NFSX_WCCDATA(v3));
1824 	if (v3) {
1825 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
1826 		vn_finished_write(mp, 0);
1827 		return (0);
1828 	}
1829 	vn_finished_write(mp, 0);
1830 	nfsm_srvdone;
1831 }
1832 
1833 /*
1834  * nfs rename service
1835  */
1836 int
1837 nfsrv_rename(nfsd, slp, procp, mrq)
1838 	struct nfsrv_descript *nfsd;
1839 	struct nfssvc_sock *slp;
1840 	struct proc *procp;
1841 	struct mbuf **mrq;
1842 {
1843 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
1844 	struct mbuf *nam = nfsd->nd_nam;
1845 	caddr_t dpos = nfsd->nd_dpos;
1846 	struct ucred *cred = &nfsd->nd_cr;
1847 	u_int32_t *tl;
1848 	int32_t t1;
1849 	caddr_t bpos;
1850 	int error = 0, cache, fdirfor_ret = 1, fdiraft_ret = 1;
1851 	uint32_t len, len2;
1852 	int tdirfor_ret = 1, tdiraft_ret = 1;
1853 	int v3 = (nfsd->nd_flag & ND_NFSV3);
1854 	char *cp2;
1855 	struct mbuf *mb, *mreq;
1856 	struct nameidata fromnd, tond;
1857 	struct vnode *fvp, *tvp, *tdvp, *fdirp = (struct vnode *)0;
1858 	struct vnode *tdirp = (struct vnode *)0;
1859 	struct vattr fdirfor, fdiraft, tdirfor, tdiraft;
1860 	nfsfh_t fnfh, tnfh;
1861 	fhandle_t *ffhp, *tfhp;
1862 	u_quad_t frev;
1863 	uid_t saved_uid;
1864 	struct mount *mp = NULL;
1865 
1866 #ifndef nolint
1867 	fvp = (struct vnode *)0;
1868 #endif
1869 	ffhp = &fnfh.fh_generic;
1870 	tfhp = &tnfh.fh_generic;
1871 	fromnd.ni_cnd.cn_nameiop = 0;
1872 	tond.ni_cnd.cn_nameiop = 0;
1873 	nfsm_srvmtofh(ffhp);
1874 	if ((mp = vfs_getvfs(&ffhp->fh_fsid)) == NULL)
1875 		return (ESTALE);
1876 	vn_start_write(NULL, &mp, V_WAIT);
1877 	nfsm_srvnamesiz(len);
1878 	/*
1879 	 * Remember our original uid so that we can reset cr_uid before
1880 	 * the second nfs_namei() call, in case it is remapped.
1881 	 */
1882 	saved_uid = cred->cr_uid;
1883 	fromnd.ni_cnd.cn_cred = cred;
1884 	fromnd.ni_cnd.cn_nameiop = DELETE;
1885 	fromnd.ni_cnd.cn_flags = WANTPARENT | SAVESTART;
1886 	error = nfs_namei(&fromnd, ffhp, len, slp, nam, &md,
1887 		&dpos, &fdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1888 	if (fdirp) {
1889 		if (v3)
1890 			fdirfor_ret = VOP_GETATTR(fdirp, &fdirfor, cred,
1891 				procp);
1892 		else {
1893 			vrele(fdirp);
1894 			fdirp = (struct vnode *)0;
1895 		}
1896 	}
1897 	if (error) {
1898 		nfsm_reply(2 * NFSX_WCCDATA(v3));
1899 		nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
1900 		nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
1901 		if (fdirp)
1902 			vrele(fdirp);
1903 		vn_finished_write(mp, 0);
1904 		return (0);
1905 	}
1906 	fvp = fromnd.ni_vp;
1907 	nfsm_srvmtofh(tfhp);
1908 	if (v3) {
1909 		nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
1910 		len2 = fxdr_unsigned(uint32_t, *tl);
1911 		/* len2 will be checked by nfs_namei */
1912 	}
1913 	else {
1914 		/* NFSv2 */
1915 		nfsm_strsiz(len2, NFS_MAXNAMLEN);
1916 	}
1917 	cred->cr_uid = saved_uid;
1918 	tond.ni_cnd.cn_cred = cred;
1919 	tond.ni_cnd.cn_nameiop = RENAME;
1920 	tond.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF | NOCACHE | SAVESTART;
1921 	error = nfs_namei(&tond, tfhp, len2, slp, nam, &md,
1922 		&dpos, &tdirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
1923 	if (tdirp) {
1924 		if (v3)
1925 			tdirfor_ret = VOP_GETATTR(tdirp, &tdirfor, cred,
1926 				procp);
1927 		else {
1928 			vrele(tdirp);
1929 			tdirp = (struct vnode *)0;
1930 		}
1931 	}
1932 	if (error) {
1933 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
1934 		vrele(fromnd.ni_dvp);
1935 		vrele(fvp);
1936 		goto out1;
1937 	}
1938 	tdvp = tond.ni_dvp;
1939 	tvp = tond.ni_vp;
1940 	if (tvp != NULL) {
1941 		if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
1942 			if (v3)
1943 				error = EEXIST;
1944 			else
1945 				error = EISDIR;
1946 			goto out;
1947 		} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
1948 			if (v3)
1949 				error = EEXIST;
1950 			else
1951 				error = ENOTDIR;
1952 			goto out;
1953 		}
1954 		if (tvp->v_type == VDIR && tvp->v_mountedhere) {
1955 			if (v3)
1956 				error = EXDEV;
1957 			else
1958 				error = ENOTEMPTY;
1959 			goto out;
1960 		}
1961 	}
1962 	if (fvp->v_type == VDIR && fvp->v_mountedhere) {
1963 		if (v3)
1964 			error = EXDEV;
1965 		else
1966 			error = ENOTEMPTY;
1967 		goto out;
1968 	}
1969 	if (fvp->v_mount != tdvp->v_mount) {
1970 		if (v3)
1971 			error = EXDEV;
1972 		else
1973 			error = ENOTEMPTY;
1974 		goto out;
1975 	}
1976 	if (fvp == tdvp) {
1977 		if (v3)
1978 			error = EINVAL;
1979 		else
1980 			error = ENOTEMPTY;
1981 	}
1982 	/*
1983 	 * If source is the same as the destination (that is the
1984 	 * same vnode with the same name in the same directory),
1985 	 * then there is nothing to do.
1986 	 */
1987 	if (fvp == tvp && fromnd.ni_dvp == tdvp &&
1988 	    fromnd.ni_cnd.cn_namelen == tond.ni_cnd.cn_namelen &&
1989 	    !memcmp(fromnd.ni_cnd.cn_nameptr, tond.ni_cnd.cn_nameptr,
1990 	      fromnd.ni_cnd.cn_namelen))
1991 		error = -1;
1992 out:
1993 	if (!error) {
1994 		nqsrv_getl(fromnd.ni_dvp, ND_WRITE);
1995 		nqsrv_getl(tdvp, ND_WRITE);
1996 		if (tvp) {
1997 			nqsrv_getl(tvp, ND_WRITE);
1998 		}
1999 		error = VOP_RENAME(fromnd.ni_dvp, fromnd.ni_vp, &fromnd.ni_cnd,
2000 				   tond.ni_dvp, tond.ni_vp, &tond.ni_cnd);
2001 	} else {
2002 		VOP_ABORTOP(tond.ni_dvp, &tond.ni_cnd);
2003 		if (tdvp == tvp)
2004 			vrele(tdvp);
2005 		else
2006 			vput(tdvp);
2007 		if (tvp)
2008 			vput(tvp);
2009 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2010 		vrele(fromnd.ni_dvp);
2011 		vrele(fvp);
2012 		if (error == -1)
2013 			error = 0;
2014 	}
2015 	vrele(tond.ni_startdir);
2016 	PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2017 out1:
2018 	if (fdirp) {
2019 		fdiraft_ret = VOP_GETATTR(fdirp, &fdiraft, cred, procp);
2020 		vrele(fdirp);
2021 	}
2022 	if (tdirp) {
2023 		tdiraft_ret = VOP_GETATTR(tdirp, &tdiraft, cred, procp);
2024 		vrele(tdirp);
2025 	}
2026 	vrele(fromnd.ni_startdir);
2027 	PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2028 	nfsm_reply(2 * NFSX_WCCDATA(v3));
2029 	if (v3) {
2030 		nfsm_srvwcc_data(fdirfor_ret, &fdirfor, fdiraft_ret, &fdiraft);
2031 		nfsm_srvwcc_data(tdirfor_ret, &tdirfor, tdiraft_ret, &tdiraft);
2032 	}
2033 	vn_finished_write(mp, 0);
2034 	return (0);
2035 
2036 nfsmout:
2037 	if (fdirp)
2038 		vrele(fdirp);
2039 	if (tdirp)
2040 		vrele(tdirp);
2041 	if (tond.ni_cnd.cn_nameiop) {
2042 		vrele(tond.ni_startdir);
2043 		PNBUF_PUT(tond.ni_cnd.cn_pnbuf);
2044 	}
2045 	if (fromnd.ni_cnd.cn_nameiop) {
2046 		vrele(fromnd.ni_startdir);
2047 		PNBUF_PUT(fromnd.ni_cnd.cn_pnbuf);
2048 		VOP_ABORTOP(fromnd.ni_dvp, &fromnd.ni_cnd);
2049 		vrele(fromnd.ni_dvp);
2050 		vrele(fvp);
2051 	}
2052 	return (error);
2053 }
2054 
2055 /*
2056  * nfs link service
2057  */
2058 int
2059 nfsrv_link(nfsd, slp, procp, mrq)
2060 	struct nfsrv_descript *nfsd;
2061 	struct nfssvc_sock *slp;
2062 	struct proc *procp;
2063 	struct mbuf **mrq;
2064 {
2065 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2066 	struct mbuf *nam = nfsd->nd_nam;
2067 	caddr_t dpos = nfsd->nd_dpos;
2068 	struct ucred *cred = &nfsd->nd_cr;
2069 	struct nameidata nd;
2070 	u_int32_t *tl;
2071 	int32_t t1;
2072 	caddr_t bpos;
2073 	int error = 0, rdonly, cache, len, dirfor_ret = 1, diraft_ret = 1;
2074 	int getret = 1, v3 = (nfsd->nd_flag & ND_NFSV3);
2075 	char *cp2;
2076 	struct mbuf *mb, *mreq;
2077 	struct vnode *vp, *xp, *dirp = (struct vnode *)0;
2078 	struct vattr dirfor, diraft, at;
2079 	nfsfh_t nfh, dnfh;
2080 	fhandle_t *fhp, *dfhp;
2081 	u_quad_t frev;
2082 	struct mount *mp = NULL;
2083 
2084 	fhp = &nfh.fh_generic;
2085 	dfhp = &dnfh.fh_generic;
2086 	nfsm_srvmtofh(fhp);
2087 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2088 		return (ESTALE);
2089 	vn_start_write(NULL, &mp, V_WAIT);
2090 	nfsm_srvmtofh(dfhp);
2091 	nfsm_srvnamesiz(len);
2092 	error = nfsrv_fhtovp(fhp, FALSE, &vp, cred, slp, nam,
2093 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2094 	if (error) {
2095 		nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2096 		nfsm_srvpostop_attr(getret, &at);
2097 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2098 		vn_finished_write(mp, 0);
2099 		return (0);
2100 	}
2101 	if (vp->v_type == VDIR && (error = suser(cred, (u_short *)0)) != 0)
2102 		goto out1;
2103 	nd.ni_cnd.cn_cred = cred;
2104 	nd.ni_cnd.cn_nameiop = CREATE;
2105 	nd.ni_cnd.cn_flags = LOCKPARENT;
2106 	error = nfs_namei(&nd, dfhp, len, slp, nam, &md, &dpos,
2107 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2108 	if (dirp) {
2109 		if (v3)
2110 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2111 				procp);
2112 		else {
2113 			vrele(dirp);
2114 			dirp = (struct vnode *)0;
2115 		}
2116 	}
2117 	if (error)
2118 		goto out1;
2119 	xp = nd.ni_vp;
2120 	if (xp != NULL) {
2121 		error = EEXIST;
2122 		goto out;
2123 	}
2124 	xp = nd.ni_dvp;
2125 	if (vp->v_mount != xp->v_mount)
2126 		error = EXDEV;
2127 out:
2128 	if (!error) {
2129 		nqsrv_getl(vp, ND_WRITE);
2130 		nqsrv_getl(xp, ND_WRITE);
2131 		error = VOP_LINK(nd.ni_dvp, vp, &nd.ni_cnd);
2132 	} else {
2133 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2134 		if (nd.ni_dvp == nd.ni_vp)
2135 			vrele(nd.ni_dvp);
2136 		else
2137 			vput(nd.ni_dvp);
2138 		if (nd.ni_vp)
2139 			vrele(nd.ni_vp);
2140 	}
2141 out1:
2142 	if (v3)
2143 		getret = VOP_GETATTR(vp, &at, cred, procp);
2144 	if (dirp) {
2145 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2146 		vrele(dirp);
2147 	}
2148 	vrele(vp);
2149 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2150 	if (v3) {
2151 		nfsm_srvpostop_attr(getret, &at);
2152 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2153 		vn_finished_write(mp, 0);
2154 		return (0);
2155 	}
2156 	vn_finished_write(mp, 0);
2157 	nfsm_srvdone;
2158 }
2159 
2160 /*
2161  * nfs symbolic link service
2162  */
2163 int
2164 nfsrv_symlink(nfsd, slp, procp, mrq)
2165 	struct nfsrv_descript *nfsd;
2166 	struct nfssvc_sock *slp;
2167 	struct proc *procp;
2168 	struct mbuf **mrq;
2169 {
2170 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2171 	struct mbuf *nam = nfsd->nd_nam;
2172 	caddr_t dpos = nfsd->nd_dpos;
2173 	struct ucred *cred = &nfsd->nd_cr;
2174 	struct vattr va, dirfor, diraft;
2175 	struct nameidata nd;
2176 	u_int32_t *tl;
2177 	int32_t t1;
2178 	struct nfsv2_sattr *sp;
2179 	char *bpos, *pathcp = NULL, *cp2;
2180 	struct uio io;
2181 	struct iovec iv;
2182 	int error = 0, cache, dirfor_ret = 1, diraft_ret = 1;
2183 	uint32_t len, len2;
2184 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2185 	struct mbuf *mb, *mreq;
2186 	struct vnode *dirp = (struct vnode *)0;
2187 	nfsfh_t nfh;
2188 	fhandle_t *fhp;
2189 	u_quad_t frev;
2190 	struct mount *mp = NULL;
2191 
2192 	nd.ni_cnd.cn_nameiop = 0;
2193 	fhp = &nfh.fh_generic;
2194 	nfsm_srvmtofh(fhp);
2195 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2196 		return (ESTALE);
2197 	vn_start_write(NULL, &mp, V_WAIT);
2198 	nfsm_srvnamesiz(len);
2199 	nd.ni_cnd.cn_cred = cred;
2200 	nd.ni_cnd.cn_nameiop = CREATE;
2201 	nd.ni_cnd.cn_flags = LOCKPARENT;
2202 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2203 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2204 	if (dirp) {
2205 		if (v3)
2206 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2207 				procp);
2208 		else {
2209 			vrele(dirp);
2210 			dirp = (struct vnode *)0;
2211 		}
2212 	}
2213 	if (error)
2214 		goto out;
2215 	VATTR_NULL(&va);
2216 	va.va_type = VLNK;
2217 	if (v3) {
2218 		va.va_mode = 0;
2219 		nfsm_srvsattr(&va);
2220 		nfsm_dissect(tl, uint32_t *, NFSX_UNSIGNED);
2221 		len2 = fxdr_unsigned(uint32_t, *tl);
2222 		if (len2 > PATH_MAX) {
2223 			/* XXX should check _PC_NO_TRUNC */
2224 			error = ENAMETOOLONG;
2225 			goto abortop;
2226 		}
2227 	}
2228 	else {
2229 		/* NFSv2 */
2230 		nfsm_strsiz(len2, NFS_MAXPATHLEN);
2231 	}
2232 	pathcp = malloc(len2 + 1, M_TEMP, M_WAITOK);
2233 	iv.iov_base = pathcp;
2234 	iv.iov_len = len2;
2235 	io.uio_resid = len2;
2236 	io.uio_offset = 0;
2237 	io.uio_iov = &iv;
2238 	io.uio_iovcnt = 1;
2239 	io.uio_segflg = UIO_SYSSPACE;
2240 	io.uio_rw = UIO_READ;
2241 	io.uio_procp = NULL;
2242 	nfsm_mtouio(&io, len2);
2243 	if (!v3) {
2244 		nfsm_dissect(sp, struct nfsv2_sattr *, NFSX_V2SATTR);
2245 		va.va_mode = fxdr_unsigned(u_int16_t, sp->sa_mode);
2246 	}
2247 	*(pathcp + len2) = '\0';
2248 	if (nd.ni_vp) {
2249 		error = EEXIST;
2250 abortop:
2251 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2252 		if (nd.ni_dvp == nd.ni_vp)
2253 			vrele(nd.ni_dvp);
2254 		else
2255 			vput(nd.ni_dvp);
2256 		if (nd.ni_vp)
2257 			vrele(nd.ni_vp);
2258 		goto out;
2259 	}
2260 	nqsrv_getl(nd.ni_dvp, ND_WRITE);
2261 	error = VOP_SYMLINK(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va, pathcp);
2262 	if (!error) {
2263 	    if (v3) {
2264 		memset((caddr_t)fhp, 0, sizeof(nfh));
2265 		fhp->fh_fsid = nd.ni_vp->v_mount->mnt_stat.f_fsidx;
2266 		error = VFS_VPTOFH(nd.ni_vp, &fhp->fh_fid);
2267 		if (!error)
2268 		    error = VOP_GETATTR(nd.ni_vp, &va, cred,
2269 					procp);
2270 		vput(nd.ni_vp);
2271 	    } else {
2272 		vput(nd.ni_vp);
2273 	    }
2274 	}
2275 out:
2276 	if (pathcp)
2277 		free(pathcp, M_TEMP);
2278 	if (dirp) {
2279 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2280 		vrele(dirp);
2281 	}
2282 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2283 	if (v3) {
2284 		if (!error) {
2285 			nfsm_srvpostop_fh(fhp);
2286 			nfsm_srvpostop_attr(0, &va);
2287 		}
2288 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2289 	}
2290 	vn_finished_write(mp, 0);
2291 	return (0);
2292 nfsmout:
2293 	if (dirp)
2294 		vrele(dirp);
2295 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2296 	if (nd.ni_dvp == nd.ni_vp)
2297 		vrele(nd.ni_dvp);
2298 	else
2299 		vput(nd.ni_dvp);
2300 	if (nd.ni_vp)
2301 		vrele(nd.ni_vp);
2302 	if (pathcp)
2303 		free(pathcp, M_TEMP);
2304 	vn_finished_write(mp, 0);
2305 	return (error);
2306 }
2307 
2308 /*
2309  * nfs mkdir service
2310  */
2311 int
2312 nfsrv_mkdir(nfsd, slp, procp, mrq)
2313 	struct nfsrv_descript *nfsd;
2314 	struct nfssvc_sock *slp;
2315 	struct proc *procp;
2316 	struct mbuf **mrq;
2317 {
2318 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2319 	struct mbuf *nam = nfsd->nd_nam;
2320 	caddr_t dpos = nfsd->nd_dpos;
2321 	struct ucred *cred = &nfsd->nd_cr;
2322 	struct vattr va, dirfor, diraft;
2323 	struct nfs_fattr *fp;
2324 	struct nameidata nd;
2325 	caddr_t cp;
2326 	u_int32_t *tl;
2327 	int32_t t1;
2328 	caddr_t bpos;
2329 	int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2330 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2331 	char *cp2;
2332 	struct mbuf *mb, *mreq;
2333 	struct vnode *vp, *dirp = (struct vnode *)0;
2334 	nfsfh_t nfh;
2335 	fhandle_t *fhp;
2336 	u_quad_t frev;
2337 	struct mount *mp = NULL;
2338 
2339 	fhp = &nfh.fh_generic;
2340 	nfsm_srvmtofh(fhp);
2341 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2342 		return (ESTALE);
2343 	vn_start_write(NULL, &mp, V_WAIT);
2344 	nfsm_srvnamesiz(len);
2345 	nd.ni_cnd.cn_cred = cred;
2346 	nd.ni_cnd.cn_nameiop = CREATE;
2347 	nd.ni_cnd.cn_flags = LOCKPARENT;
2348 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2349 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2350 	if (dirp) {
2351 		if (v3)
2352 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2353 				procp);
2354 		else {
2355 			vrele(dirp);
2356 			dirp = (struct vnode *)0;
2357 		}
2358 	}
2359 	if (error) {
2360 		nfsm_reply(NFSX_WCCDATA(v3));
2361 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2362 		if (dirp)
2363 			vrele(dirp);
2364 		vn_finished_write(mp, 0);
2365 		return (0);
2366 	}
2367 	VATTR_NULL(&va);
2368 	if (v3) {
2369 		va.va_mode = 0;
2370 		nfsm_srvsattr(&va);
2371 	} else {
2372 		nfsm_dissect(tl, u_int32_t *, NFSX_UNSIGNED);
2373 		va.va_mode = nfstov_mode(*tl++);
2374 	}
2375 	va.va_type = VDIR;
2376 	vp = nd.ni_vp;
2377 	if (vp != NULL) {
2378 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2379 		if (nd.ni_dvp == vp)
2380 			vrele(nd.ni_dvp);
2381 		else
2382 			vput(nd.ni_dvp);
2383 		vrele(vp);
2384 		error = EEXIST;
2385 		goto out;
2386 	}
2387 	nqsrv_getl(nd.ni_dvp, ND_WRITE);
2388 	error = VOP_MKDIR(nd.ni_dvp, &nd.ni_vp, &nd.ni_cnd, &va);
2389 	if (!error) {
2390 		vp = nd.ni_vp;
2391 		memset((caddr_t)fhp, 0, sizeof(nfh));
2392 		fhp->fh_fsid = vp->v_mount->mnt_stat.f_fsidx;
2393 		error = VFS_VPTOFH(vp, &fhp->fh_fid);
2394 		if (!error)
2395 			error = VOP_GETATTR(vp, &va, cred, procp);
2396 		vput(vp);
2397 	}
2398 out:
2399 	if (dirp) {
2400 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2401 		vrele(dirp);
2402 	}
2403 	nfsm_reply(NFSX_SRVFH(v3) + NFSX_POSTOPATTR(v3) + NFSX_WCCDATA(v3));
2404 	if (v3) {
2405 		if (!error) {
2406 			nfsm_srvpostop_fh(fhp);
2407 			nfsm_srvpostop_attr(0, &va);
2408 		}
2409 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2410 	} else {
2411 		nfsm_srvfhtom(fhp, v3);
2412 		nfsm_build(fp, struct nfs_fattr *, NFSX_V2FATTR);
2413 		nfsm_srvfillattr(&va, fp);
2414 	}
2415 	vn_finished_write(mp, 0);
2416 	return (0);
2417 nfsmout:
2418 	if (dirp)
2419 		vrele(dirp);
2420 	VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2421 	if (nd.ni_dvp == nd.ni_vp)
2422 		vrele(nd.ni_dvp);
2423 	else
2424 		vput(nd.ni_dvp);
2425 	if (nd.ni_vp)
2426 		vrele(nd.ni_vp);
2427 	vn_finished_write(mp, 0);
2428 	return (error);
2429 }
2430 
2431 /*
2432  * nfs rmdir service
2433  */
2434 int
2435 nfsrv_rmdir(nfsd, slp, procp, mrq)
2436 	struct nfsrv_descript *nfsd;
2437 	struct nfssvc_sock *slp;
2438 	struct proc *procp;
2439 	struct mbuf **mrq;
2440 {
2441 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2442 	struct mbuf *nam = nfsd->nd_nam;
2443 	caddr_t dpos = nfsd->nd_dpos;
2444 	struct ucred *cred = &nfsd->nd_cr;
2445 	u_int32_t *tl;
2446 	int32_t t1;
2447 	caddr_t bpos;
2448 	int error = 0, cache, len, dirfor_ret = 1, diraft_ret = 1;
2449 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2450 	char *cp2;
2451 	struct mbuf *mb, *mreq;
2452 	struct vnode *vp, *dirp = (struct vnode *)0;
2453 	struct vattr dirfor, diraft;
2454 	nfsfh_t nfh;
2455 	fhandle_t *fhp;
2456 	struct nameidata nd;
2457 	u_quad_t frev;
2458 	struct mount *mp = NULL;
2459 
2460 	fhp = &nfh.fh_generic;
2461 	nfsm_srvmtofh(fhp);
2462 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
2463 		return (ESTALE);
2464 	vn_start_write(NULL, &mp, V_WAIT);
2465 	nfsm_srvnamesiz(len);
2466 	nd.ni_cnd.cn_cred = cred;
2467 	nd.ni_cnd.cn_nameiop = DELETE;
2468 	nd.ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF;
2469 	error = nfs_namei(&nd, fhp, len, slp, nam, &md, &dpos,
2470 		&dirp, procp, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2471 	if (dirp) {
2472 		if (v3)
2473 			dirfor_ret = VOP_GETATTR(dirp, &dirfor, cred,
2474 				procp);
2475 		else {
2476 			vrele(dirp);
2477 			dirp = (struct vnode *)0;
2478 		}
2479 	}
2480 	if (error) {
2481 		nfsm_reply(NFSX_WCCDATA(v3));
2482 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2483 		if (dirp)
2484 			vrele(dirp);
2485 		vn_finished_write(mp, 0);
2486 		return (0);
2487 	}
2488 	vp = nd.ni_vp;
2489 	if (vp->v_type != VDIR) {
2490 		error = ENOTDIR;
2491 		goto out;
2492 	}
2493 	/*
2494 	 * No rmdir "." please.
2495 	 */
2496 	if (nd.ni_dvp == vp) {
2497 		error = EINVAL;
2498 		goto out;
2499 	}
2500 	/*
2501 	 * The root of a mounted filesystem cannot be deleted.
2502 	 */
2503 	if (vp->v_flag & VROOT)
2504 		error = EBUSY;
2505 out:
2506 	if (!error) {
2507 		nqsrv_getl(nd.ni_dvp, ND_WRITE);
2508 		nqsrv_getl(vp, ND_WRITE);
2509 		error = VOP_RMDIR(nd.ni_dvp, nd.ni_vp, &nd.ni_cnd);
2510 	} else {
2511 		VOP_ABORTOP(nd.ni_dvp, &nd.ni_cnd);
2512 		if (nd.ni_dvp == nd.ni_vp)
2513 			vrele(nd.ni_dvp);
2514 		else
2515 			vput(nd.ni_dvp);
2516 		vput(vp);
2517 	}
2518 	if (dirp) {
2519 		diraft_ret = VOP_GETATTR(dirp, &diraft, cred, procp);
2520 		vrele(dirp);
2521 	}
2522 	nfsm_reply(NFSX_WCCDATA(v3));
2523 	if (v3) {
2524 		nfsm_srvwcc_data(dirfor_ret, &dirfor, diraft_ret, &diraft);
2525 		vn_finished_write(mp, 0);
2526 		return (0);
2527 	}
2528 	vn_finished_write(mp, 0);
2529 	nfsm_srvdone;
2530 }
2531 
2532 /*
2533  * nfs readdir service
2534  * - mallocs what it thinks is enough to read
2535  *	count rounded up to a multiple of NFS_DIRBLKSIZ <= NFS_MAXREADDIR
2536  * - calls VOP_READDIR()
2537  * - loops around building the reply
2538  *	if the output generated exceeds count break out of loop
2539  *	The nfsm_clget macro is used here so that the reply will be packed
2540  *	tightly in mbuf clusters.
2541  * - it only knows that it has encountered eof when the VOP_READDIR()
2542  *	reads nothing
2543  * - as such one readdir rpc will return eof false although you are there
2544  *	and then the next will return eof
2545  * - it trims out records with d_fileno == 0
2546  *	this doesn't matter for Unix clients, but they might confuse clients
2547  *	for other os'.
2548  * - it trims out records with d_type == DT_WHT
2549  *	these cannot be seen through NFS (unless we extend the protocol)
2550  * NB: It is tempting to set eof to true if the VOP_READDIR() reads less
2551  *	than requested, but this may not apply to all filesystems. For
2552  *	example, client NFS does not { although it is never remote mounted
2553  *	anyhow }
2554  *     The alternate call nfsrv_readdirplus() does lookups as well.
2555  * PS: The NFS protocol spec. does not clarify what the "count" byte
2556  *	argument is a count of.. just name strings and file id's or the
2557  *	entire reply rpc or ...
2558  *	I tried just file name and id sizes and it confused the Sun client,
2559  *	so I am using the full rpc size now. The "paranoia.." comment refers
2560  *	to including the status longwords that are not a part of the dir.
2561  *	"entry" structures, but are in the rpc.
2562  */
2563 struct flrep {
2564 	nfsuint64 fl_off;
2565 	u_int32_t fl_postopok;
2566 	u_int32_t fl_fattr[NFSX_V3FATTR / sizeof (u_int32_t)];
2567 	u_int32_t fl_fhok;
2568 	u_int32_t fl_fhsize;
2569 	u_int32_t fl_nfh[NFSX_V3FH / sizeof (u_int32_t)];
2570 };
2571 
2572 int
2573 nfsrv_readdir(nfsd, slp, procp, mrq)
2574 	struct nfsrv_descript *nfsd;
2575 	struct nfssvc_sock *slp;
2576 	struct proc *procp;
2577 	struct mbuf **mrq;
2578 {
2579 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2580 	struct mbuf *nam = nfsd->nd_nam;
2581 	caddr_t dpos = nfsd->nd_dpos;
2582 	struct ucred *cred = &nfsd->nd_cr;
2583 	char *bp, *be;
2584 	struct mbuf *mp;
2585 	struct dirent *dp;
2586 	caddr_t cp;
2587 	u_int32_t *tl;
2588 	int32_t t1;
2589 	caddr_t bpos;
2590 	struct mbuf *mb, *mreq, *mp2;
2591 	char *cpos, *cend, *cp2, *rbuf;
2592 	struct vnode *vp;
2593 	struct vattr at;
2594 	nfsfh_t nfh;
2595 	fhandle_t *fhp;
2596 	struct uio io;
2597 	struct iovec iv;
2598 	int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2599 	int siz, cnt, fullsiz, eofflag, rdonly, cache, ncookies;
2600 	int v3 = (nfsd->nd_flag & ND_NFSV3);
2601 	u_quad_t frev, off, toff, verf;
2602 	off_t *cookies = NULL, *cookiep;
2603 	nfsuint64 jar;
2604 
2605 	fhp = &nfh.fh_generic;
2606 	nfsm_srvmtofh(fhp);
2607 	if (v3) {
2608 		nfsm_dissect(tl, u_int32_t *, 5 * NFSX_UNSIGNED);
2609 		toff = fxdr_hyper(tl);
2610 		tl += 2;
2611 		verf = fxdr_hyper(tl);
2612 		tl += 2;
2613 	} else {
2614 		nfsm_dissect(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2615 		toff = fxdr_unsigned(u_quad_t, *tl++);
2616 	}
2617 	off = toff;
2618 	cnt = fxdr_unsigned(int, *tl);
2619 	siz = ((cnt + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2620 	xfer = NFS_SRVMAXDATA(nfsd);
2621 	if (siz > xfer)
2622 		siz = xfer;
2623 	fullsiz = siz;
2624 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2625 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2626 	if (!error && vp->v_type != VDIR) {
2627 		error = ENOTDIR;
2628 		vput(vp);
2629 	}
2630 	if (error) {
2631 		nfsm_reply(NFSX_UNSIGNED);
2632 		nfsm_srvpostop_attr(getret, &at);
2633 		return (0);
2634 	}
2635 	nqsrv_getl(vp, ND_READ);
2636 	if (v3) {
2637 		error = getret = VOP_GETATTR(vp, &at, cred, procp);
2638 #ifdef NFS3_STRICTVERF
2639 		/*
2640 		 * XXX This check is too strict for Solaris 2.5 clients.
2641 		 */
2642 		if (!error && toff && verf != at.va_filerev)
2643 			error = NFSERR_BAD_COOKIE;
2644 #endif
2645 	}
2646 	if (!error)
2647 		error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2648 	if (error) {
2649 		vput(vp);
2650 		nfsm_reply(NFSX_POSTOPATTR(v3));
2651 		nfsm_srvpostop_attr(getret, &at);
2652 		return (0);
2653 	}
2654 	VOP_UNLOCK(vp, 0);
2655 	rbuf = malloc(siz, M_TEMP, M_WAITOK);
2656 again:
2657 	iv.iov_base = rbuf;
2658 	iv.iov_len = fullsiz;
2659 	io.uio_iov = &iv;
2660 	io.uio_iovcnt = 1;
2661 	io.uio_offset = (off_t)off;
2662 	io.uio_resid = fullsiz;
2663 	io.uio_segflg = UIO_SYSSPACE;
2664 	io.uio_rw = UIO_READ;
2665 	io.uio_procp = NULL;
2666 	eofflag = 0;
2667 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2668 
2669 	error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2670 
2671 	off = (off_t)io.uio_offset;
2672 	if (!cookies && !error)
2673 		error = NFSERR_PERM;
2674 	if (v3) {
2675 		getret = VOP_GETATTR(vp, &at, cred, procp);
2676 		if (!error)
2677 			error = getret;
2678 	}
2679 
2680 	VOP_UNLOCK(vp, 0);
2681 	if (error) {
2682 		vrele(vp);
2683 		free((caddr_t)rbuf, M_TEMP);
2684 		if (cookies)
2685 			free((caddr_t)cookies, M_TEMP);
2686 		nfsm_reply(NFSX_POSTOPATTR(v3));
2687 		nfsm_srvpostop_attr(getret, &at);
2688 		return (0);
2689 	}
2690 	if (io.uio_resid) {
2691 		siz -= io.uio_resid;
2692 
2693 		/*
2694 		 * If nothing read, return eof
2695 		 * rpc reply
2696 		 */
2697 		if (siz == 0) {
2698 			vrele(vp);
2699 			nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) +
2700 				2 * NFSX_UNSIGNED);
2701 			if (v3) {
2702 				nfsm_srvpostop_attr(getret, &at);
2703 				nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2704 				txdr_hyper(at.va_filerev, tl);
2705 				tl += 2;
2706 			} else
2707 				nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2708 			*tl++ = nfs_false;
2709 			*tl = nfs_true;
2710 			free((caddr_t)rbuf, M_TEMP);
2711 			free((caddr_t)cookies, M_TEMP);
2712 			return (0);
2713 		}
2714 	}
2715 
2716 	/*
2717 	 * Check for degenerate cases of nothing useful read.
2718 	 * If so go try again
2719 	 */
2720 	cpos = rbuf;
2721 	cend = rbuf + siz;
2722 	dp = (struct dirent *)cpos;
2723 	cookiep = cookies;
2724 
2725 	while (cpos < cend && ncookies > 0 &&
2726 		(dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2727 		cpos += dp->d_reclen;
2728 		dp = (struct dirent *)cpos;
2729 		cookiep++;
2730 		ncookies--;
2731 	}
2732 	if (cpos >= cend || ncookies == 0) {
2733 		toff = off;
2734 		siz = fullsiz;
2735 		goto again;
2736 	}
2737 
2738 	len = 3 * NFSX_UNSIGNED;	/* paranoia, probably can be 0 */
2739 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_COOKIEVERF(v3) + siz);
2740 	if (v3) {
2741 		nfsm_srvpostop_attr(getret, &at);
2742 		nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
2743 		txdr_hyper(at.va_filerev, tl);
2744 	}
2745 	mp = mp2 = mb;
2746 	bp = bpos;
2747 	be = bp + M_TRAILINGSPACE(mp);
2748 
2749 	/* Loop through the records and build reply */
2750 	while (cpos < cend && ncookies > 0) {
2751 		if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
2752 			nlen = dp->d_namlen;
2753 			rem = nfsm_rndup(nlen)-nlen;
2754 			len += (4 * NFSX_UNSIGNED + nlen + rem);
2755 			if (v3)
2756 				len += 2 * NFSX_UNSIGNED;
2757 			if (len > cnt) {
2758 				eofflag = 0;
2759 				break;
2760 			}
2761 			/*
2762 			 * Build the directory record xdr from
2763 			 * the dirent entry.
2764 			 */
2765 			nfsm_clget;
2766 			*tl = nfs_true;
2767 			bp += NFSX_UNSIGNED;
2768 			if (v3) {
2769 				nfsm_clget;
2770 				*tl = txdr_unsigned(dp->d_fileno >> 32);
2771 				bp += NFSX_UNSIGNED;
2772 			}
2773 			nfsm_clget;
2774 			*tl = txdr_unsigned(dp->d_fileno);
2775 			bp += NFSX_UNSIGNED;
2776 			nfsm_clget;
2777 			*tl = txdr_unsigned(nlen);
2778 			bp += NFSX_UNSIGNED;
2779 
2780 			/* And loop around copying the name */
2781 			xfer = nlen;
2782 			cp = dp->d_name;
2783 			while (xfer > 0) {
2784 				nfsm_clget;
2785 				if ((bp+xfer) > be)
2786 					tsiz = be-bp;
2787 				else
2788 					tsiz = xfer;
2789 				memcpy(bp, cp, tsiz);
2790 				bp += tsiz;
2791 				xfer -= tsiz;
2792 				if (xfer > 0)
2793 					cp += tsiz;
2794 			}
2795 			/* And null pad to an int32_t boundary */
2796 			for (i = 0; i < rem; i++)
2797 				*bp++ = '\0';
2798 			nfsm_clget;
2799 
2800 			/* Finish off the record */
2801 			txdr_hyper(*cookiep, &jar);
2802 			if (v3) {
2803 				*tl = jar.nfsuquad[0];
2804 				bp += NFSX_UNSIGNED;
2805 				nfsm_clget;
2806 			}
2807 			*tl = jar.nfsuquad[1];
2808 			bp += NFSX_UNSIGNED;
2809 		}
2810 		cpos += dp->d_reclen;
2811 		dp = (struct dirent *)cpos;
2812 		cookiep++;
2813 		ncookies--;
2814 	}
2815 	vrele(vp);
2816 	nfsm_clget;
2817 	*tl = nfs_false;
2818 	bp += NFSX_UNSIGNED;
2819 	nfsm_clget;
2820 	if (eofflag)
2821 		*tl = nfs_true;
2822 	else
2823 		*tl = nfs_false;
2824 	bp += NFSX_UNSIGNED;
2825 	if (mp != mb) {
2826 		if (bp < be)
2827 			mp->m_len = bp - mtod(mp, caddr_t);
2828 	} else
2829 		mp->m_len += bp - bpos;
2830 	free((caddr_t)rbuf, M_TEMP);
2831 	free((caddr_t)cookies, M_TEMP);
2832 	nfsm_srvdone;
2833 }
2834 
2835 int
2836 nfsrv_readdirplus(nfsd, slp, procp, mrq)
2837 	struct nfsrv_descript *nfsd;
2838 	struct nfssvc_sock *slp;
2839 	struct proc *procp;
2840 	struct mbuf **mrq;
2841 {
2842 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
2843 	struct mbuf *nam = nfsd->nd_nam;
2844 	caddr_t dpos = nfsd->nd_dpos;
2845 	struct ucred *cred = &nfsd->nd_cr;
2846 	char *bp, *be;
2847 	struct mbuf *mp;
2848 	struct dirent *dp;
2849 	caddr_t cp;
2850 	u_int32_t *tl;
2851 	int32_t t1;
2852 	caddr_t bpos;
2853 	struct mbuf *mb, *mreq, *mp2;
2854 	char *cpos, *cend, *cp2, *rbuf;
2855 	struct vnode *vp, *nvp;
2856 	struct flrep fl;
2857 	nfsfh_t nfh;
2858 	fhandle_t *fhp, *nfhp = (fhandle_t *)fl.fl_nfh;
2859 	struct uio io;
2860 	struct iovec iv;
2861 	struct vattr va, at, *vap = &va;
2862 	struct nfs_fattr *fp;
2863 	int len, nlen, rem, xfer, tsiz, i, error = 0, getret = 1;
2864 	int siz, cnt, fullsiz, eofflag, rdonly, cache, dirlen, ncookies;
2865 	u_quad_t frev, off, toff, verf;
2866 	off_t *cookies = NULL, *cookiep;
2867 
2868 	fhp = &nfh.fh_generic;
2869 	nfsm_srvmtofh(fhp);
2870 	nfsm_dissect(tl, u_int32_t *, 6 * NFSX_UNSIGNED);
2871 	toff = fxdr_hyper(tl);
2872 	tl += 2;
2873 	verf = fxdr_hyper(tl);
2874 	tl += 2;
2875 	siz = fxdr_unsigned(int, *tl++);
2876 	cnt = fxdr_unsigned(int, *tl);
2877 	off = toff;
2878 	siz = ((siz + DIRBLKSIZ - 1) & ~(DIRBLKSIZ - 1));
2879 	xfer = NFS_SRVMAXDATA(nfsd);
2880 	if (siz > xfer)
2881 		siz = xfer;
2882 	fullsiz = siz;
2883 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
2884 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
2885 	if (!error && vp->v_type != VDIR) {
2886 		error = ENOTDIR;
2887 		vput(vp);
2888 	}
2889 	if (error) {
2890 		nfsm_reply(NFSX_UNSIGNED);
2891 		nfsm_srvpostop_attr(getret, &at);
2892 		return (0);
2893 	}
2894 	error = getret = VOP_GETATTR(vp, &at, cred, procp);
2895 #ifdef NFS3_STRICTVERF
2896 	/*
2897 	 * XXX This check is too strict for Solaris 2.5 clients.
2898 	 */
2899 	if (!error && toff && verf != at.va_filerev)
2900 		error = NFSERR_BAD_COOKIE;
2901 #endif
2902 	if (!error) {
2903 		nqsrv_getl(vp, ND_READ);
2904 		error = nfsrv_access(vp, VEXEC, cred, rdonly, procp, 0);
2905 	}
2906 	if (error) {
2907 		vput(vp);
2908 		nfsm_reply(NFSX_V3POSTOPATTR);
2909 		nfsm_srvpostop_attr(getret, &at);
2910 		return (0);
2911 	}
2912 	VOP_UNLOCK(vp, 0);
2913 
2914 	rbuf = malloc(siz, M_TEMP, M_WAITOK);
2915 again:
2916 	iv.iov_base = rbuf;
2917 	iv.iov_len = fullsiz;
2918 	io.uio_iov = &iv;
2919 	io.uio_iovcnt = 1;
2920 	io.uio_offset = (off_t)off;
2921 	io.uio_resid = fullsiz;
2922 	io.uio_segflg = UIO_SYSSPACE;
2923 	io.uio_rw = UIO_READ;
2924 	io.uio_procp = NULL;
2925 	eofflag = 0;
2926 
2927 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
2928 
2929 	error = VOP_READDIR(vp, &io, cred, &eofflag, &cookies, &ncookies);
2930 
2931 	off = (u_quad_t)io.uio_offset;
2932 	getret = VOP_GETATTR(vp, &at, cred, procp);
2933 
2934 	VOP_UNLOCK(vp, 0);
2935 
2936 	/*
2937 	 * If the VGET operation doesn't work for this filesystem,
2938 	 * we can't support readdirplus. Returning NOTSUPP should
2939 	 * make clients fall back to plain readdir.
2940 	 * There's no need to check for VPTOFH as well, we wouldn't
2941 	 * even be here otherwise.
2942 	 */
2943 	if (!getret) {
2944 		if ((getret = VFS_VGET(vp->v_mount, at.va_fileid, &nvp)))
2945 			getret = (getret == EOPNOTSUPP) ?
2946 				NFSERR_NOTSUPP : NFSERR_IO;
2947 		else
2948 			vput(nvp);
2949 	}
2950 
2951 	if (!cookies && !error)
2952 		error = NFSERR_PERM;
2953 	if (!error)
2954 		error = getret;
2955 	if (error) {
2956 		vrele(vp);
2957 		if (cookies)
2958 			free((caddr_t)cookies, M_TEMP);
2959 		free((caddr_t)rbuf, M_TEMP);
2960 		nfsm_reply(NFSX_V3POSTOPATTR);
2961 		nfsm_srvpostop_attr(getret, &at);
2962 		return (0);
2963 	}
2964 	if (io.uio_resid) {
2965 		siz -= io.uio_resid;
2966 
2967 		/*
2968 		 * If nothing read, return eof
2969 		 * rpc reply
2970 		 */
2971 		if (siz == 0) {
2972 			vrele(vp);
2973 			nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF +
2974 				2 * NFSX_UNSIGNED);
2975 			nfsm_srvpostop_attr(getret, &at);
2976 			nfsm_build(tl, u_int32_t *, 4 * NFSX_UNSIGNED);
2977 			txdr_hyper(at.va_filerev, tl);
2978 			tl += 2;
2979 			*tl++ = nfs_false;
2980 			*tl = nfs_true;
2981 			free((caddr_t)cookies, M_TEMP);
2982 			free((caddr_t)rbuf, M_TEMP);
2983 			return (0);
2984 		}
2985 	}
2986 
2987 	/*
2988 	 * Check for degenerate cases of nothing useful read.
2989 	 * If so go try again
2990 	 */
2991 	cpos = rbuf;
2992 	cend = rbuf + siz;
2993 	dp = (struct dirent *)cpos;
2994 	cookiep = cookies;
2995 
2996 	while (cpos < cend && ncookies > 0 &&
2997 		(dp->d_fileno == 0 || dp->d_type == DT_WHT)) {
2998 		cpos += dp->d_reclen;
2999 		dp = (struct dirent *)cpos;
3000 		cookiep++;
3001 		ncookies--;
3002 	}
3003 	if (cpos >= cend || ncookies == 0) {
3004 		toff = off;
3005 		siz = fullsiz;
3006 		goto again;
3007 	}
3008 
3009 	dirlen = len = NFSX_V3POSTOPATTR + NFSX_V3COOKIEVERF + 2 * NFSX_UNSIGNED;
3010 	nfsm_reply(cnt);
3011 	nfsm_srvpostop_attr(getret, &at);
3012 	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
3013 	txdr_hyper(at.va_filerev, tl);
3014 	mp = mp2 = mb;
3015 	bp = bpos;
3016 	be = bp + M_TRAILINGSPACE(mp);
3017 
3018 	/* Loop through the records and build reply */
3019 	while (cpos < cend && ncookies > 0) {
3020 		if (dp->d_fileno != 0 && dp->d_type != DT_WHT) {
3021 			nlen = dp->d_namlen;
3022 			rem = nfsm_rndup(nlen)-nlen;
3023 
3024 			/*
3025 			 * For readdir_and_lookup get the vnode using
3026 			 * the file number.
3027 			 */
3028 			if (VFS_VGET(vp->v_mount, dp->d_fileno, &nvp))
3029 				goto invalid;
3030 			memset((caddr_t)nfhp, 0, NFSX_V3FH);
3031 			nfhp->fh_fsid =
3032 				nvp->v_mount->mnt_stat.f_fsidx;
3033 			if (VFS_VPTOFH(nvp, &nfhp->fh_fid)) {
3034 				vput(nvp);
3035 				goto invalid;
3036 			}
3037 			if (VOP_GETATTR(nvp, vap, cred, procp)) {
3038 				vput(nvp);
3039 				goto invalid;
3040 			}
3041 			vput(nvp);
3042 
3043 			/*
3044 			 * If either the dircount or maxcount will be
3045 			 * exceeded, get out now. Both of these lengths
3046 			 * are calculated conservatively, including all
3047 			 * XDR overheads.
3048 			 */
3049 			len += (8 * NFSX_UNSIGNED + nlen + rem + NFSX_V3FH +
3050 				NFSX_V3POSTOPATTR);
3051 			dirlen += (6 * NFSX_UNSIGNED + nlen + rem);
3052 			if (len > cnt || dirlen > fullsiz) {
3053 				eofflag = 0;
3054 				break;
3055 			}
3056 
3057 			/*
3058 			 * Build the directory record xdr from
3059 			 * the dirent entry.
3060 			 */
3061 			fp = (struct nfs_fattr *)&fl.fl_fattr;
3062 			nfsm_srvfillattr(vap, fp);
3063 			fl.fl_fhsize = txdr_unsigned(NFSX_V3FH);
3064 			fl.fl_fhok = nfs_true;
3065 			fl.fl_postopok = nfs_true;
3066 			txdr_hyper(*cookiep, fl.fl_off.nfsuquad);
3067 
3068 			nfsm_clget;
3069 			*tl = nfs_true;
3070 			bp += NFSX_UNSIGNED;
3071 			nfsm_clget;
3072 			*tl = txdr_unsigned(dp->d_fileno >> 32);
3073 			bp += NFSX_UNSIGNED;
3074 			nfsm_clget;
3075 			*tl = txdr_unsigned(dp->d_fileno);
3076 			bp += NFSX_UNSIGNED;
3077 			nfsm_clget;
3078 			*tl = txdr_unsigned(nlen);
3079 			bp += NFSX_UNSIGNED;
3080 
3081 			/* And loop around copying the name */
3082 			xfer = nlen;
3083 			cp = dp->d_name;
3084 			while (xfer > 0) {
3085 				nfsm_clget;
3086 				if ((bp + xfer) > be)
3087 					tsiz = be - bp;
3088 				else
3089 					tsiz = xfer;
3090 				memcpy(bp, cp, tsiz);
3091 				bp += tsiz;
3092 				xfer -= tsiz;
3093 				if (xfer > 0)
3094 					cp += tsiz;
3095 			}
3096 			/* And null pad to an int32_t boundary */
3097 			for (i = 0; i < rem; i++)
3098 				*bp++ = '\0';
3099 
3100 			/*
3101 			 * Now copy the flrep structure out.
3102 			 */
3103 			xfer = sizeof (struct flrep);
3104 			cp = (caddr_t)&fl;
3105 			while (xfer > 0) {
3106 				nfsm_clget;
3107 				if ((bp + xfer) > be)
3108 					tsiz = be - bp;
3109 				else
3110 					tsiz = xfer;
3111 				memcpy(bp, cp, tsiz);
3112 				bp += tsiz;
3113 				xfer -= tsiz;
3114 				if (xfer > 0)
3115 					cp += tsiz;
3116 			}
3117 		}
3118 invalid:
3119 		cpos += dp->d_reclen;
3120 		dp = (struct dirent *)cpos;
3121 		cookiep++;
3122 		ncookies--;
3123 	}
3124 	vrele(vp);
3125 	nfsm_clget;
3126 	*tl = nfs_false;
3127 	bp += NFSX_UNSIGNED;
3128 	nfsm_clget;
3129 	if (eofflag)
3130 		*tl = nfs_true;
3131 	else
3132 		*tl = nfs_false;
3133 	bp += NFSX_UNSIGNED;
3134 	if (mp != mb) {
3135 		if (bp < be)
3136 			mp->m_len = bp - mtod(mp, caddr_t);
3137 	} else
3138 		mp->m_len += bp - bpos;
3139 	free((caddr_t)cookies, M_TEMP);
3140 	free((caddr_t)rbuf, M_TEMP);
3141 	nfsm_srvdone;
3142 }
3143 
3144 /*
3145  * nfs commit service
3146  */
3147 int
3148 nfsrv_commit(nfsd, slp, procp, mrq)
3149 	struct nfsrv_descript *nfsd;
3150 	struct nfssvc_sock *slp;
3151 	struct proc *procp;
3152 	struct mbuf **mrq;
3153 {
3154 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3155 	struct mbuf *nam = nfsd->nd_nam;
3156 	caddr_t dpos = nfsd->nd_dpos;
3157 	struct ucred *cred = &nfsd->nd_cr;
3158 	struct vattr bfor, aft;
3159 	struct vnode *vp;
3160 	nfsfh_t nfh;
3161 	fhandle_t *fhp;
3162 	u_int32_t *tl;
3163 	int32_t t1;
3164 	caddr_t bpos;
3165 	int error = 0, rdonly, for_ret = 1, aft_ret = 1, cache;
3166 	uint32_t cnt;
3167 	char *cp2;
3168 	struct mbuf *mb, *mreq;
3169 	u_quad_t frev, off, end;
3170 	struct mount *mp = NULL;
3171 
3172 #ifndef nolint
3173 	cache = 0;
3174 #endif
3175 	fhp = &nfh.fh_generic;
3176 	nfsm_srvmtofh(fhp);
3177 	if ((mp = vfs_getvfs(&fhp->fh_fsid)) == NULL)
3178 		return (ESTALE);
3179 	vn_start_write(NULL, &mp, V_WAIT);
3180 	nfsm_dissect(tl, u_int32_t *, 3 * NFSX_UNSIGNED);
3181 
3182 	off = fxdr_hyper(tl);
3183 	tl += 2;
3184 	cnt = fxdr_unsigned(uint32_t, *tl);
3185 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3186 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3187 	if (error) {
3188 		nfsm_reply(2 * NFSX_UNSIGNED);
3189 		nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3190 		vn_finished_write(mp, 0);
3191 		return (0);
3192 	}
3193 	for_ret = VOP_GETATTR(vp, &bfor, cred, procp);
3194 	end = (cnt > 0) ? off + cnt : vp->v_size;
3195 	if (end < off || end > vp->v_size)
3196 		end = vp->v_size;
3197 	if (off < vp->v_size)
3198 		error = VOP_FSYNC(vp, cred, FSYNC_WAIT, off, end, procp);
3199 	/* else error == 0, from nfsrv_fhtovp() */
3200 	aft_ret = VOP_GETATTR(vp, &aft, cred, procp);
3201 	vput(vp);
3202 	nfsm_reply(NFSX_V3WCCDATA + NFSX_V3WRITEVERF);
3203 	nfsm_srvwcc_data(for_ret, &bfor, aft_ret, &aft);
3204 	if (!error) {
3205 		nfsm_build(tl, u_int32_t *, NFSX_V3WRITEVERF);
3206 		*tl++ = txdr_unsigned(boottime.tv_sec);
3207 		*tl = txdr_unsigned(boottime.tv_usec);
3208 	} else {
3209 		vn_finished_write(mp, 0);
3210 		return (0);
3211 	}
3212 	vn_finished_write(mp, 0);
3213 	nfsm_srvdone;
3214 }
3215 
3216 /*
3217  * nfs statfs service
3218  */
3219 int
3220 nfsrv_statfs(nfsd, slp, procp, mrq)
3221 	struct nfsrv_descript *nfsd;
3222 	struct nfssvc_sock *slp;
3223 	struct proc *procp;
3224 	struct mbuf **mrq;
3225 {
3226 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3227 	struct mbuf *nam = nfsd->nd_nam;
3228 	caddr_t dpos = nfsd->nd_dpos;
3229 	struct ucred *cred = &nfsd->nd_cr;
3230 	struct statvfs *sf;
3231 	struct nfs_statfs *sfp;
3232 	u_int32_t *tl;
3233 	int32_t t1;
3234 	caddr_t bpos;
3235 	int error = 0, rdonly, cache, getret = 1;
3236 	int v3 = (nfsd->nd_flag & ND_NFSV3);
3237 	char *cp2;
3238 	struct mbuf *mb, *mreq;
3239 	struct vnode *vp;
3240 	struct vattr at;
3241 	nfsfh_t nfh;
3242 	fhandle_t *fhp;
3243 	struct statvfs statvfs;
3244 	u_quad_t frev, tval;
3245 
3246 #ifndef nolint
3247 	cache = 0;
3248 #endif
3249 	fhp = &nfh.fh_generic;
3250 	nfsm_srvmtofh(fhp);
3251 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3252 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3253 	if (error) {
3254 		nfsm_reply(NFSX_UNSIGNED);
3255 		nfsm_srvpostop_attr(getret, &at);
3256 		return (0);
3257 	}
3258 	sf = &statvfs;
3259 	error = VFS_STATVFS(vp->v_mount, sf, procp);
3260 	getret = VOP_GETATTR(vp, &at, cred, procp);
3261 	vput(vp);
3262 	nfsm_reply(NFSX_POSTOPATTR(v3) + NFSX_STATFS(v3));
3263 	if (v3)
3264 		nfsm_srvpostop_attr(getret, &at);
3265 	if (error)
3266 		return (0);
3267 	nfsm_build(sfp, struct nfs_statfs *, NFSX_STATFS(v3));
3268 	if (v3) {
3269 		tval = (u_quad_t)((quad_t)sf->f_blocks * (quad_t)sf->f_frsize);
3270 		txdr_hyper(tval, &sfp->sf_tbytes);
3271 		tval = (u_quad_t)((quad_t)sf->f_bfree * (quad_t)sf->f_frsize);
3272 		txdr_hyper(tval, &sfp->sf_fbytes);
3273 		tval = (u_quad_t)((quad_t)sf->f_bavail * (quad_t)sf->f_frsize);
3274 		txdr_hyper(tval, &sfp->sf_abytes);
3275 		tval = (u_quad_t)sf->f_files;
3276 		txdr_hyper(tval, &sfp->sf_tfiles);
3277 		tval = (u_quad_t)sf->f_ffree;
3278 		txdr_hyper(tval, &sfp->sf_ffiles);
3279 		txdr_hyper(tval, &sfp->sf_afiles);
3280 		sfp->sf_invarsec = 0;
3281 	} else {
3282 		sfp->sf_tsize = txdr_unsigned(NFS_MAXDGRAMDATA);
3283 		sfp->sf_bsize = txdr_unsigned(sf->f_frsize);
3284 		sfp->sf_blocks = txdr_unsigned(sf->f_blocks);
3285 		sfp->sf_bfree = txdr_unsigned(sf->f_bfree);
3286 		sfp->sf_bavail = txdr_unsigned(sf->f_bavail);
3287 	}
3288 	nfsm_srvdone;
3289 }
3290 
3291 /*
3292  * nfs fsinfo service
3293  */
3294 int
3295 nfsrv_fsinfo(nfsd, slp, procp, mrq)
3296 	struct nfsrv_descript *nfsd;
3297 	struct nfssvc_sock *slp;
3298 	struct proc *procp;
3299 	struct mbuf **mrq;
3300 {
3301 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3302 	struct mbuf *nam = nfsd->nd_nam;
3303 	caddr_t dpos = nfsd->nd_dpos;
3304 	struct ucred *cred = &nfsd->nd_cr;
3305 	u_int32_t *tl;
3306 	struct nfsv3_fsinfo *sip;
3307 	int32_t t1;
3308 	caddr_t bpos;
3309 	int error = 0, rdonly, cache, getret = 1;
3310 	uint32_t maxdata;
3311 	char *cp2;
3312 	struct mbuf *mb, *mreq;
3313 	struct vnode *vp;
3314 	struct vattr at;
3315 	nfsfh_t nfh;
3316 	fhandle_t *fhp;
3317 	u_quad_t frev, maxfsize;
3318 	struct statvfs sb;
3319 
3320 #ifndef nolint
3321 	cache = 0;
3322 #endif
3323 	fhp = &nfh.fh_generic;
3324 	nfsm_srvmtofh(fhp);
3325 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3326 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3327 	if (error) {
3328 		nfsm_reply(NFSX_UNSIGNED);
3329 		nfsm_srvpostop_attr(getret, &at);
3330 		return (0);
3331 	}
3332 
3333 	/* XXX Try to make a guess on the max file size. */
3334 	VFS_STATVFS(vp->v_mount, &sb, (struct proc *)0);
3335 	maxfsize = (u_quad_t)0x80000000 * sb.f_frsize - 1;
3336 
3337 	getret = VOP_GETATTR(vp, &at, cred, procp);
3338 	vput(vp);
3339 	nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3FSINFO);
3340 	nfsm_srvpostop_attr(getret, &at);
3341 	nfsm_build(sip, struct nfsv3_fsinfo *, NFSX_V3FSINFO);
3342 
3343 	/*
3344 	 * XXX
3345 	 * There should be file system VFS OP(s) to get this information.
3346 	 * For now, assume ufs.
3347 	 */
3348 	if (slp->ns_so->so_type == SOCK_DGRAM)
3349 		maxdata = NFS_MAXDGRAMDATA;
3350 	else
3351 		maxdata = NFS_MAXDATA;
3352 	sip->fs_rtmax = txdr_unsigned(maxdata);
3353 	sip->fs_rtpref = txdr_unsigned(maxdata);
3354 	sip->fs_rtmult = txdr_unsigned(NFS_FABLKSIZE);
3355 	sip->fs_wtmax = txdr_unsigned(maxdata);
3356 	sip->fs_wtpref = txdr_unsigned(maxdata);
3357 	sip->fs_wtmult = txdr_unsigned(NFS_FABLKSIZE);
3358 	sip->fs_dtpref = txdr_unsigned(maxdata);
3359 	txdr_hyper(maxfsize, &sip->fs_maxfilesize);
3360 	sip->fs_timedelta.nfsv3_sec = 0;
3361 	sip->fs_timedelta.nfsv3_nsec = txdr_unsigned(1);
3362 	sip->fs_properties = txdr_unsigned(NFSV3FSINFO_LINK |
3363 		NFSV3FSINFO_SYMLINK | NFSV3FSINFO_HOMOGENEOUS |
3364 		NFSV3FSINFO_CANSETTIME);
3365 	nfsm_srvdone;
3366 }
3367 
3368 /*
3369  * nfs pathconf service
3370  */
3371 int
3372 nfsrv_pathconf(nfsd, slp, procp, mrq)
3373 	struct nfsrv_descript *nfsd;
3374 	struct nfssvc_sock *slp;
3375 	struct proc *procp;
3376 	struct mbuf **mrq;
3377 {
3378 	struct mbuf *mrep = nfsd->nd_mrep, *md = nfsd->nd_md;
3379 	struct mbuf *nam = nfsd->nd_nam;
3380 	caddr_t dpos = nfsd->nd_dpos;
3381 	struct ucred *cred = &nfsd->nd_cr;
3382 	u_int32_t *tl;
3383 	struct nfsv3_pathconf *pc;
3384 	int32_t t1;
3385 	caddr_t bpos;
3386 	int error = 0, rdonly, cache, getret = 1;
3387 	register_t linkmax, namemax, chownres, notrunc;
3388 	char *cp2;
3389 	struct mbuf *mb, *mreq;
3390 	struct vnode *vp;
3391 	struct vattr at;
3392 	nfsfh_t nfh;
3393 	fhandle_t *fhp;
3394 	u_quad_t frev;
3395 
3396 #ifndef nolint
3397 	cache = 0;
3398 #endif
3399 	fhp = &nfh.fh_generic;
3400 	nfsm_srvmtofh(fhp);
3401 	error = nfsrv_fhtovp(fhp, 1, &vp, cred, slp, nam,
3402 		 &rdonly, (nfsd->nd_flag & ND_KERBAUTH), FALSE);
3403 	if (error) {
3404 		nfsm_reply(NFSX_UNSIGNED);
3405 		nfsm_srvpostop_attr(getret, &at);
3406 		return (0);
3407 	}
3408 	error = VOP_PATHCONF(vp, _PC_LINK_MAX, &linkmax);
3409 	if (!error)
3410 		error = VOP_PATHCONF(vp, _PC_NAME_MAX, &namemax);
3411 	if (!error)
3412 		error = VOP_PATHCONF(vp, _PC_CHOWN_RESTRICTED, &chownres);
3413 	if (!error)
3414 		error = VOP_PATHCONF(vp, _PC_NO_TRUNC, &notrunc);
3415 	getret = VOP_GETATTR(vp, &at, cred, procp);
3416 	vput(vp);
3417 	nfsm_reply(NFSX_V3POSTOPATTR + NFSX_V3PATHCONF);
3418 	nfsm_srvpostop_attr(getret, &at);
3419 	if (error)
3420 		return (0);
3421 	nfsm_build(pc, struct nfsv3_pathconf *, NFSX_V3PATHCONF);
3422 
3423 	pc->pc_linkmax = txdr_unsigned(linkmax);
3424 	pc->pc_namemax = txdr_unsigned(namemax);
3425 	pc->pc_notrunc = txdr_unsigned(notrunc);
3426 	pc->pc_chownrestricted = txdr_unsigned(chownres);
3427 
3428 	/*
3429 	 * These should probably be supported by VOP_PATHCONF(), but
3430 	 * until msdosfs is exportable (why would you want to?), the
3431 	 * Unix defaults should be ok.
3432 	 */
3433 	pc->pc_caseinsensitive = nfs_false;
3434 	pc->pc_casepreserving = nfs_true;
3435 	nfsm_srvdone;
3436 }
3437 
3438 /*
3439  * Null operation, used by clients to ping server
3440  */
3441 /* ARGSUSED */
3442 int
3443 nfsrv_null(nfsd, slp, procp, mrq)
3444 	struct nfsrv_descript *nfsd;
3445 	struct nfssvc_sock *slp;
3446 	struct proc *procp;
3447 	struct mbuf **mrq;
3448 {
3449 	struct mbuf *mrep = nfsd->nd_mrep;
3450 	caddr_t bpos;
3451 	int error = NFSERR_RETVOID, cache = 0;
3452 	struct mbuf *mb, *mreq;
3453 	u_quad_t frev;
3454 
3455 	nfsm_reply(0);
3456 	return (0);
3457 }
3458 
3459 /*
3460  * No operation, used for obsolete procedures
3461  */
3462 /* ARGSUSED */
3463 int
3464 nfsrv_noop(nfsd, slp, procp, mrq)
3465 	struct nfsrv_descript *nfsd;
3466 	struct nfssvc_sock *slp;
3467 	struct proc *procp;
3468 	struct mbuf **mrq;
3469 {
3470 	struct mbuf *mrep = nfsd->nd_mrep;
3471 	caddr_t bpos;
3472 	int error, cache = 0;
3473 	struct mbuf *mb, *mreq;
3474 	u_quad_t frev;
3475 
3476 	if (nfsd->nd_repstat)
3477 		error = nfsd->nd_repstat;
3478 	else
3479 		error = EPROCUNAVAIL;
3480 	nfsm_reply(0);
3481 	return (0);
3482 }
3483 
3484 /*
3485  * Perform access checking for vnodes obtained from file handles that would
3486  * refer to files already opened by a Unix client. You cannot just use
3487  * vn_writechk() and VOP_ACCESS() for two reasons.
3488  * 1 - You must check for exported rdonly as well as MNT_RDONLY for the write case
3489  * 2 - The owner is to be given access irrespective of mode bits for some
3490  *     operations, so that processes that chmod after opening a file don't
3491  *     break. I don't like this because it opens a security hole, but since
3492  *     the nfs server opens a security hole the size of a barn door anyhow,
3493  *     what the heck.
3494  *
3495  * The exception to rule 2 is EPERM. If a file is IMMUTABLE, VOP_ACCESS()
3496  * will return EPERM instead of EACCESS. EPERM is always an error.
3497  */
3498 int
3499 nfsrv_access(vp, flags, cred, rdonly, p, override)
3500 	struct vnode *vp;
3501 	int flags;
3502 	struct ucred *cred;
3503 	int rdonly;
3504 	struct proc *p;
3505 {
3506 	struct vattr vattr;
3507 	int error;
3508 	if (flags & VWRITE) {
3509 		/* Just vn_writechk() changed to check rdonly */
3510 		/*
3511 		 * Disallow write attempts on read-only file systems;
3512 		 * unless the file is a socket or a block or character
3513 		 * device resident on the file system.
3514 		 */
3515 		if (rdonly || (vp->v_mount->mnt_flag & MNT_RDONLY)) {
3516 			switch (vp->v_type) {
3517 			case VREG:
3518 			case VDIR:
3519 			case VLNK:
3520 				return (EROFS);
3521 			default:
3522 				break;
3523 			}
3524 		}
3525 
3526 		/*
3527 		 * If the vnode is in use as a process's text,
3528 		 * we can't allow writing.
3529 		 */
3530 		if (vp->v_flag & VTEXT)
3531 			return (ETXTBSY);
3532 	}
3533 	error = VOP_GETATTR(vp, &vattr, cred, p);
3534 	if (error)
3535 		return (error);
3536 	error = VOP_ACCESS(vp, flags, cred, p);
3537 	/*
3538 	 * Allow certain operations for the owner (reads and writes
3539 	 * on files that are already open).
3540 	 */
3541 	if (override && error == EACCES && cred->cr_uid == vattr.va_uid)
3542 		error = 0;
3543 	return error;
3544 }
3545