xref: /dragonfly/sys/vfs/nfs/nfs_subs.c (revision 86fe9e07)
1 /*
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Rick Macklem at The University of Guelph.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  *	@(#)nfs_subs.c  8.8 (Berkeley) 5/22/95
37  * $FreeBSD: /repoman/r/ncvs/src/sys/nfsclient/nfs_subs.c,v 1.128 2004/04/14 23:23:55 peadar Exp $
38  * $DragonFly: src/sys/vfs/nfs/nfs_subs.c,v 1.20 2004/08/17 18:57:34 dillon Exp $
39  */
40 
41 /*
42  * These functions support the macros and help fiddle mbuf chains for
43  * the nfs op functions. They do things like create the rpc header and
44  * copy data between mbuf chains and uio lists.
45  */
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/kernel.h>
49 #include <sys/buf.h>
50 #include <sys/proc.h>
51 #include <sys/mount.h>
52 #include <sys/vnode.h>
53 #include <sys/namei.h>
54 #include <sys/mbuf.h>
55 #include <sys/socket.h>
56 #include <sys/stat.h>
57 #include <sys/malloc.h>
58 #include <sys/sysent.h>
59 #include <sys/syscall.h>
60 #include <sys/conf.h>
61 
62 #include <vm/vm.h>
63 #include <vm/vm_object.h>
64 #include <vm/vm_extern.h>
65 #include <vm/vm_zone.h>
66 
67 #include <sys/buf2.h>
68 
69 #include "rpcv2.h"
70 #include "nfsproto.h"
71 #include "nfs.h"
72 #include "nfsmount.h"
73 #include "nfsnode.h"
74 #include "xdr_subs.h"
75 #include "nfsm_subs.h"
76 #include "nqnfs.h"
77 #include "nfsrtt.h"
78 
79 #include <netinet/in.h>
80 
81 /*
82  * Data items converted to xdr at startup, since they are constant
83  * This is kinda hokey, but may save a little time doing byte swaps
84  */
85 u_int32_t nfs_xdrneg1;
86 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
87 	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
88 	rpc_auth_kerb;
89 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
90 
91 /* And other global data */
92 static u_int32_t nfs_xid = 0;
93 static enum vtype nv2tov_type[8]= {
94 	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
95 };
96 enum vtype nv3tov_type[8]= {
97 	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
98 };
99 
100 int nfs_ticks;
101 int nfs_pbuf_freecnt = -1;	/* start out unlimited */
102 
103 struct nfs_reqq nfs_reqq;
104 struct nfssvc_sockhead nfssvc_sockhead;
105 int nfssvc_sockhead_flag;
106 struct nfsd_head nfsd_head;
107 int nfsd_head_flag;
108 struct nfs_bufq nfs_bufq;
109 struct nqtimerhead nqtimerhead;
110 struct nqfhhashhead *nqfhhashtbl;
111 u_long nqfhhash;
112 
113 static void (*nfs_prev_lease_updatetime) (int);
114 static int nfs_prev_nfssvc_sy_narg;
115 static sy_call_t *nfs_prev_nfssvc_sy_call;
116 
117 #ifndef NFS_NOSERVER
118 
119 static int (*nfs_prev_vop_lease_check)(struct vop_lease_args *);
120 
121 /*
122  * Mapping of old NFS Version 2 RPC numbers to generic numbers.
123  */
124 int nfsv3_procid[NFS_NPROCS] = {
125 	NFSPROC_NULL,
126 	NFSPROC_GETATTR,
127 	NFSPROC_SETATTR,
128 	NFSPROC_NOOP,
129 	NFSPROC_LOOKUP,
130 	NFSPROC_READLINK,
131 	NFSPROC_READ,
132 	NFSPROC_NOOP,
133 	NFSPROC_WRITE,
134 	NFSPROC_CREATE,
135 	NFSPROC_REMOVE,
136 	NFSPROC_RENAME,
137 	NFSPROC_LINK,
138 	NFSPROC_SYMLINK,
139 	NFSPROC_MKDIR,
140 	NFSPROC_RMDIR,
141 	NFSPROC_READDIR,
142 	NFSPROC_FSSTAT,
143 	NFSPROC_NOOP,
144 	NFSPROC_NOOP,
145 	NFSPROC_NOOP,
146 	NFSPROC_NOOP,
147 	NFSPROC_NOOP,
148 	NFSPROC_NOOP,
149 	NFSPROC_NOOP,
150 	NFSPROC_NOOP
151 };
152 
153 #endif /* NFS_NOSERVER */
154 /*
155  * and the reverse mapping from generic to Version 2 procedure numbers
156  */
157 int nfsv2_procid[NFS_NPROCS] = {
158 	NFSV2PROC_NULL,
159 	NFSV2PROC_GETATTR,
160 	NFSV2PROC_SETATTR,
161 	NFSV2PROC_LOOKUP,
162 	NFSV2PROC_NOOP,
163 	NFSV2PROC_READLINK,
164 	NFSV2PROC_READ,
165 	NFSV2PROC_WRITE,
166 	NFSV2PROC_CREATE,
167 	NFSV2PROC_MKDIR,
168 	NFSV2PROC_SYMLINK,
169 	NFSV2PROC_CREATE,
170 	NFSV2PROC_REMOVE,
171 	NFSV2PROC_RMDIR,
172 	NFSV2PROC_RENAME,
173 	NFSV2PROC_LINK,
174 	NFSV2PROC_READDIR,
175 	NFSV2PROC_NOOP,
176 	NFSV2PROC_STATFS,
177 	NFSV2PROC_NOOP,
178 	NFSV2PROC_NOOP,
179 	NFSV2PROC_NOOP,
180 	NFSV2PROC_NOOP,
181 	NFSV2PROC_NOOP,
182 	NFSV2PROC_NOOP,
183 	NFSV2PROC_NOOP,
184 };
185 
186 #ifndef NFS_NOSERVER
187 /*
188  * Maps errno values to nfs error numbers.
189  * Use NFSERR_IO as the catch all for ones not specifically defined in
190  * RFC 1094.
191  */
192 static u_char nfsrv_v2errmap[ELAST] = {
193   NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
194   NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
195   NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
196   NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
197   NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
198   NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
199   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
200   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
201   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
202   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
203   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
204   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
205   NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
206   NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
207   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
208   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
209   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
210   NFSERR_IO /* << Last is 86 */
211 };
212 
213 /*
214  * Maps errno values to nfs error numbers.
215  * Although it is not obvious whether or not NFS clients really care if
216  * a returned error value is in the specified list for the procedure, the
217  * safest thing to do is filter them appropriately. For Version 2, the
218  * X/Open XNFS document is the only specification that defines error values
219  * for each RPC (The RFC simply lists all possible error values for all RPCs),
220  * so I have decided to not do this for Version 2.
221  * The first entry is the default error return and the rest are the valid
222  * errors for that RPC in increasing numeric order.
223  */
224 static short nfsv3err_null[] = {
225 	0,
226 	0,
227 };
228 
229 static short nfsv3err_getattr[] = {
230 	NFSERR_IO,
231 	NFSERR_IO,
232 	NFSERR_STALE,
233 	NFSERR_BADHANDLE,
234 	NFSERR_SERVERFAULT,
235 	0,
236 };
237 
238 static short nfsv3err_setattr[] = {
239 	NFSERR_IO,
240 	NFSERR_PERM,
241 	NFSERR_IO,
242 	NFSERR_ACCES,
243 	NFSERR_INVAL,
244 	NFSERR_NOSPC,
245 	NFSERR_ROFS,
246 	NFSERR_DQUOT,
247 	NFSERR_STALE,
248 	NFSERR_BADHANDLE,
249 	NFSERR_NOT_SYNC,
250 	NFSERR_SERVERFAULT,
251 	0,
252 };
253 
254 static short nfsv3err_lookup[] = {
255 	NFSERR_IO,
256 	NFSERR_NOENT,
257 	NFSERR_IO,
258 	NFSERR_ACCES,
259 	NFSERR_NOTDIR,
260 	NFSERR_NAMETOL,
261 	NFSERR_STALE,
262 	NFSERR_BADHANDLE,
263 	NFSERR_SERVERFAULT,
264 	0,
265 };
266 
267 static short nfsv3err_access[] = {
268 	NFSERR_IO,
269 	NFSERR_IO,
270 	NFSERR_STALE,
271 	NFSERR_BADHANDLE,
272 	NFSERR_SERVERFAULT,
273 	0,
274 };
275 
276 static short nfsv3err_readlink[] = {
277 	NFSERR_IO,
278 	NFSERR_IO,
279 	NFSERR_ACCES,
280 	NFSERR_INVAL,
281 	NFSERR_STALE,
282 	NFSERR_BADHANDLE,
283 	NFSERR_NOTSUPP,
284 	NFSERR_SERVERFAULT,
285 	0,
286 };
287 
288 static short nfsv3err_read[] = {
289 	NFSERR_IO,
290 	NFSERR_IO,
291 	NFSERR_NXIO,
292 	NFSERR_ACCES,
293 	NFSERR_INVAL,
294 	NFSERR_STALE,
295 	NFSERR_BADHANDLE,
296 	NFSERR_SERVERFAULT,
297 	0,
298 };
299 
300 static short nfsv3err_write[] = {
301 	NFSERR_IO,
302 	NFSERR_IO,
303 	NFSERR_ACCES,
304 	NFSERR_INVAL,
305 	NFSERR_FBIG,
306 	NFSERR_NOSPC,
307 	NFSERR_ROFS,
308 	NFSERR_DQUOT,
309 	NFSERR_STALE,
310 	NFSERR_BADHANDLE,
311 	NFSERR_SERVERFAULT,
312 	0,
313 };
314 
315 static short nfsv3err_create[] = {
316 	NFSERR_IO,
317 	NFSERR_IO,
318 	NFSERR_ACCES,
319 	NFSERR_EXIST,
320 	NFSERR_NOTDIR,
321 	NFSERR_NOSPC,
322 	NFSERR_ROFS,
323 	NFSERR_NAMETOL,
324 	NFSERR_DQUOT,
325 	NFSERR_STALE,
326 	NFSERR_BADHANDLE,
327 	NFSERR_NOTSUPP,
328 	NFSERR_SERVERFAULT,
329 	0,
330 };
331 
332 static short nfsv3err_mkdir[] = {
333 	NFSERR_IO,
334 	NFSERR_IO,
335 	NFSERR_ACCES,
336 	NFSERR_EXIST,
337 	NFSERR_NOTDIR,
338 	NFSERR_NOSPC,
339 	NFSERR_ROFS,
340 	NFSERR_NAMETOL,
341 	NFSERR_DQUOT,
342 	NFSERR_STALE,
343 	NFSERR_BADHANDLE,
344 	NFSERR_NOTSUPP,
345 	NFSERR_SERVERFAULT,
346 	0,
347 };
348 
349 static short nfsv3err_symlink[] = {
350 	NFSERR_IO,
351 	NFSERR_IO,
352 	NFSERR_ACCES,
353 	NFSERR_EXIST,
354 	NFSERR_NOTDIR,
355 	NFSERR_NOSPC,
356 	NFSERR_ROFS,
357 	NFSERR_NAMETOL,
358 	NFSERR_DQUOT,
359 	NFSERR_STALE,
360 	NFSERR_BADHANDLE,
361 	NFSERR_NOTSUPP,
362 	NFSERR_SERVERFAULT,
363 	0,
364 };
365 
366 static short nfsv3err_mknod[] = {
367 	NFSERR_IO,
368 	NFSERR_IO,
369 	NFSERR_ACCES,
370 	NFSERR_EXIST,
371 	NFSERR_NOTDIR,
372 	NFSERR_NOSPC,
373 	NFSERR_ROFS,
374 	NFSERR_NAMETOL,
375 	NFSERR_DQUOT,
376 	NFSERR_STALE,
377 	NFSERR_BADHANDLE,
378 	NFSERR_NOTSUPP,
379 	NFSERR_SERVERFAULT,
380 	NFSERR_BADTYPE,
381 	0,
382 };
383 
384 static short nfsv3err_remove[] = {
385 	NFSERR_IO,
386 	NFSERR_NOENT,
387 	NFSERR_IO,
388 	NFSERR_ACCES,
389 	NFSERR_NOTDIR,
390 	NFSERR_ROFS,
391 	NFSERR_NAMETOL,
392 	NFSERR_STALE,
393 	NFSERR_BADHANDLE,
394 	NFSERR_SERVERFAULT,
395 	0,
396 };
397 
398 static short nfsv3err_rmdir[] = {
399 	NFSERR_IO,
400 	NFSERR_NOENT,
401 	NFSERR_IO,
402 	NFSERR_ACCES,
403 	NFSERR_EXIST,
404 	NFSERR_NOTDIR,
405 	NFSERR_INVAL,
406 	NFSERR_ROFS,
407 	NFSERR_NAMETOL,
408 	NFSERR_NOTEMPTY,
409 	NFSERR_STALE,
410 	NFSERR_BADHANDLE,
411 	NFSERR_NOTSUPP,
412 	NFSERR_SERVERFAULT,
413 	0,
414 };
415 
416 static short nfsv3err_rename[] = {
417 	NFSERR_IO,
418 	NFSERR_NOENT,
419 	NFSERR_IO,
420 	NFSERR_ACCES,
421 	NFSERR_EXIST,
422 	NFSERR_XDEV,
423 	NFSERR_NOTDIR,
424 	NFSERR_ISDIR,
425 	NFSERR_INVAL,
426 	NFSERR_NOSPC,
427 	NFSERR_ROFS,
428 	NFSERR_MLINK,
429 	NFSERR_NAMETOL,
430 	NFSERR_NOTEMPTY,
431 	NFSERR_DQUOT,
432 	NFSERR_STALE,
433 	NFSERR_BADHANDLE,
434 	NFSERR_NOTSUPP,
435 	NFSERR_SERVERFAULT,
436 	0,
437 };
438 
439 static short nfsv3err_link[] = {
440 	NFSERR_IO,
441 	NFSERR_IO,
442 	NFSERR_ACCES,
443 	NFSERR_EXIST,
444 	NFSERR_XDEV,
445 	NFSERR_NOTDIR,
446 	NFSERR_INVAL,
447 	NFSERR_NOSPC,
448 	NFSERR_ROFS,
449 	NFSERR_MLINK,
450 	NFSERR_NAMETOL,
451 	NFSERR_DQUOT,
452 	NFSERR_STALE,
453 	NFSERR_BADHANDLE,
454 	NFSERR_NOTSUPP,
455 	NFSERR_SERVERFAULT,
456 	0,
457 };
458 
459 static short nfsv3err_readdir[] = {
460 	NFSERR_IO,
461 	NFSERR_IO,
462 	NFSERR_ACCES,
463 	NFSERR_NOTDIR,
464 	NFSERR_STALE,
465 	NFSERR_BADHANDLE,
466 	NFSERR_BAD_COOKIE,
467 	NFSERR_TOOSMALL,
468 	NFSERR_SERVERFAULT,
469 	0,
470 };
471 
472 static short nfsv3err_readdirplus[] = {
473 	NFSERR_IO,
474 	NFSERR_IO,
475 	NFSERR_ACCES,
476 	NFSERR_NOTDIR,
477 	NFSERR_STALE,
478 	NFSERR_BADHANDLE,
479 	NFSERR_BAD_COOKIE,
480 	NFSERR_NOTSUPP,
481 	NFSERR_TOOSMALL,
482 	NFSERR_SERVERFAULT,
483 	0,
484 };
485 
486 static short nfsv3err_fsstat[] = {
487 	NFSERR_IO,
488 	NFSERR_IO,
489 	NFSERR_STALE,
490 	NFSERR_BADHANDLE,
491 	NFSERR_SERVERFAULT,
492 	0,
493 };
494 
495 static short nfsv3err_fsinfo[] = {
496 	NFSERR_STALE,
497 	NFSERR_STALE,
498 	NFSERR_BADHANDLE,
499 	NFSERR_SERVERFAULT,
500 	0,
501 };
502 
503 static short nfsv3err_pathconf[] = {
504 	NFSERR_STALE,
505 	NFSERR_STALE,
506 	NFSERR_BADHANDLE,
507 	NFSERR_SERVERFAULT,
508 	0,
509 };
510 
511 static short nfsv3err_commit[] = {
512 	NFSERR_IO,
513 	NFSERR_IO,
514 	NFSERR_STALE,
515 	NFSERR_BADHANDLE,
516 	NFSERR_SERVERFAULT,
517 	0,
518 };
519 
520 static short *nfsrv_v3errmap[] = {
521 	nfsv3err_null,
522 	nfsv3err_getattr,
523 	nfsv3err_setattr,
524 	nfsv3err_lookup,
525 	nfsv3err_access,
526 	nfsv3err_readlink,
527 	nfsv3err_read,
528 	nfsv3err_write,
529 	nfsv3err_create,
530 	nfsv3err_mkdir,
531 	nfsv3err_symlink,
532 	nfsv3err_mknod,
533 	nfsv3err_remove,
534 	nfsv3err_rmdir,
535 	nfsv3err_rename,
536 	nfsv3err_link,
537 	nfsv3err_readdir,
538 	nfsv3err_readdirplus,
539 	nfsv3err_fsstat,
540 	nfsv3err_fsinfo,
541 	nfsv3err_pathconf,
542 	nfsv3err_commit,
543 };
544 
545 #endif /* NFS_NOSERVER */
546 
547 extern struct nfsrtt nfsrtt;
548 extern time_t nqnfsstarttime;
549 extern int nqsrv_clockskew;
550 extern int nqsrv_writeslack;
551 extern int nqsrv_maxlease;
552 extern struct nfsstats nfsstats;
553 extern int nqnfs_piggy[NFS_NPROCS];
554 extern nfstype nfsv2_type[9];
555 extern nfstype nfsv3_type[9];
556 extern struct nfsnodehashhead *nfsnodehashtbl;
557 extern u_long nfsnodehash;
558 
559 struct nfssvc_args;
560 extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
561 
562 LIST_HEAD(nfsnodehashhead, nfsnode);
563 
564 int nfs_webnamei (struct nameidata *, struct vnode *, struct proc *);
565 
566 u_quad_t
567 nfs_curusec(void)
568 {
569 	struct timeval tv;
570 
571 	getmicrotime(&tv);
572 	return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
573 }
574 
575 /*
576  * Create the header for an rpc request packet
577  * The hsiz is the size of the rest of the nfs request header.
578  * (just used to decide if a cluster is a good idea)
579  */
580 struct mbuf *
581 nfsm_reqh(struct vnode *vp, u_long procid, int hsiz, caddr_t *bposp)
582 {
583 	struct mbuf *mb;
584 	u_int32_t *tl;
585 	caddr_t bpos;
586 	struct mbuf *mb2;
587 	struct nfsmount *nmp;
588 	int nqflag;
589 
590 	MGET(mb, MB_WAIT, MT_DATA);
591 	if (hsiz >= MINCLSIZE)
592 		MCLGET(mb, MB_WAIT);
593 	mb->m_len = 0;
594 	bpos = mtod(mb, caddr_t);
595 
596 	/*
597 	 * For NQNFS, add lease request.
598 	 */
599 	if (vp) {
600 		nmp = VFSTONFS(vp->v_mount);
601 		if (nmp->nm_flag & NFSMNT_NQNFS) {
602 			nqflag = NQNFS_NEEDLEASE(vp, procid);
603 			if (nqflag) {
604 				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
605 				*tl++ = txdr_unsigned(nqflag);
606 				*tl = txdr_unsigned(nmp->nm_leaseterm);
607 			} else {
608 				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
609 				*tl = 0;
610 			}
611 		}
612 	}
613 	/* Finally, return values */
614 	*bposp = bpos;
615 	return (mb);
616 }
617 
618 /*
619  * Build the RPC header and fill in the authorization info.
620  * The authorization string argument is only used when the credentials
621  * come from outside of the kernel.
622  * Returns the head of the mbuf list.
623  */
624 struct mbuf *
625 nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type,
626 	     int auth_len, char *auth_str, int verf_len, char *verf_str,
627 	     struct mbuf *mrest, int mrest_len, struct mbuf **mbp,
628 	     u_int32_t *xidp)
629 {
630 	struct mbuf *mb;
631 	u_int32_t *tl;
632 	caddr_t bpos;
633 	int i;
634 	struct mbuf *mreq, *mb2;
635 	int siz, grpsiz, authsiz;
636 
637 	authsiz = nfsm_rndup(auth_len);
638 	MGETHDR(mb, MB_WAIT, MT_DATA);
639 	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
640 		MCLGET(mb, MB_WAIT);
641 	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
642 		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
643 	} else {
644 		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
645 	}
646 	mb->m_len = 0;
647 	mreq = mb;
648 	bpos = mtod(mb, caddr_t);
649 
650 	/*
651 	 * First the RPC header.
652 	 */
653 	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
654 
655 	/* Get a pretty random xid to start with */
656 	if (!nfs_xid)
657 		nfs_xid = random();
658 	/*
659 	 * Skip zero xid if it should ever happen.
660 	 */
661 	if (++nfs_xid == 0)
662 		nfs_xid++;
663 
664 	*tl++ = *xidp = txdr_unsigned(nfs_xid);
665 	*tl++ = rpc_call;
666 	*tl++ = rpc_vers;
667 	if (nmflag & NFSMNT_NQNFS) {
668 		*tl++ = txdr_unsigned(NQNFS_PROG);
669 		*tl++ = txdr_unsigned(NQNFS_VER3);
670 	} else {
671 		*tl++ = txdr_unsigned(NFS_PROG);
672 		if (nmflag & NFSMNT_NFSV3)
673 			*tl++ = txdr_unsigned(NFS_VER3);
674 		else
675 			*tl++ = txdr_unsigned(NFS_VER2);
676 	}
677 	if (nmflag & NFSMNT_NFSV3)
678 		*tl++ = txdr_unsigned(procid);
679 	else
680 		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
681 
682 	/*
683 	 * And then the authorization cred.
684 	 */
685 	*tl++ = txdr_unsigned(auth_type);
686 	*tl = txdr_unsigned(authsiz);
687 	switch (auth_type) {
688 	case RPCAUTH_UNIX:
689 		nfsm_build(tl, u_int32_t *, auth_len);
690 		*tl++ = 0;		/* stamp ?? */
691 		*tl++ = 0;		/* NULL hostname */
692 		*tl++ = txdr_unsigned(cr->cr_uid);
693 		*tl++ = txdr_unsigned(cr->cr_groups[0]);
694 		grpsiz = (auth_len >> 2) - 5;
695 		*tl++ = txdr_unsigned(grpsiz);
696 		for (i = 1; i <= grpsiz; i++)
697 			*tl++ = txdr_unsigned(cr->cr_groups[i]);
698 		break;
699 	case RPCAUTH_KERB4:
700 		siz = auth_len;
701 		while (siz > 0) {
702 			if (M_TRAILINGSPACE(mb) == 0) {
703 				MGET(mb2, MB_WAIT, MT_DATA);
704 				if (siz >= MINCLSIZE)
705 					MCLGET(mb2, MB_WAIT);
706 				mb->m_next = mb2;
707 				mb = mb2;
708 				mb->m_len = 0;
709 				bpos = mtod(mb, caddr_t);
710 			}
711 			i = min(siz, M_TRAILINGSPACE(mb));
712 			bcopy(auth_str, bpos, i);
713 			mb->m_len += i;
714 			auth_str += i;
715 			bpos += i;
716 			siz -= i;
717 		}
718 		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
719 			for (i = 0; i < siz; i++)
720 				*bpos++ = '\0';
721 			mb->m_len += siz;
722 		}
723 		break;
724 	};
725 
726 	/*
727 	 * And the verifier...
728 	 */
729 	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
730 	if (verf_str) {
731 		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
732 		*tl = txdr_unsigned(verf_len);
733 		siz = verf_len;
734 		while (siz > 0) {
735 			if (M_TRAILINGSPACE(mb) == 0) {
736 				MGET(mb2, MB_WAIT, MT_DATA);
737 				if (siz >= MINCLSIZE)
738 					MCLGET(mb2, MB_WAIT);
739 				mb->m_next = mb2;
740 				mb = mb2;
741 				mb->m_len = 0;
742 				bpos = mtod(mb, caddr_t);
743 			}
744 			i = min(siz, M_TRAILINGSPACE(mb));
745 			bcopy(verf_str, bpos, i);
746 			mb->m_len += i;
747 			verf_str += i;
748 			bpos += i;
749 			siz -= i;
750 		}
751 		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
752 			for (i = 0; i < siz; i++)
753 				*bpos++ = '\0';
754 			mb->m_len += siz;
755 		}
756 	} else {
757 		*tl++ = txdr_unsigned(RPCAUTH_NULL);
758 		*tl = 0;
759 	}
760 	mb->m_next = mrest;
761 	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
762 	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
763 	*mbp = mb;
764 	return (mreq);
765 }
766 
767 /*
768  * copies mbuf chain to the uio scatter/gather list
769  */
770 int
771 nfsm_mbuftouio(struct mbuf **mrep, struct uio *uiop, int siz, caddr_t *dpos)
772 {
773 	char *mbufcp, *uiocp;
774 	int xfer, left, len;
775 	struct mbuf *mp;
776 	long uiosiz, rem;
777 	int error = 0;
778 
779 	mp = *mrep;
780 	mbufcp = *dpos;
781 	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
782 	rem = nfsm_rndup(siz)-siz;
783 	while (siz > 0) {
784 		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
785 			return (EFBIG);
786 		left = uiop->uio_iov->iov_len;
787 		uiocp = uiop->uio_iov->iov_base;
788 		if (left > siz)
789 			left = siz;
790 		uiosiz = left;
791 		while (left > 0) {
792 			while (len == 0) {
793 				mp = mp->m_next;
794 				if (mp == NULL)
795 					return (EBADRPC);
796 				mbufcp = mtod(mp, caddr_t);
797 				len = mp->m_len;
798 			}
799 			xfer = (left > len) ? len : left;
800 #ifdef notdef
801 			/* Not Yet.. */
802 			if (uiop->uio_iov->iov_op != NULL)
803 				(*(uiop->uio_iov->iov_op))
804 				(mbufcp, uiocp, xfer);
805 			else
806 #endif
807 			if (uiop->uio_segflg == UIO_SYSSPACE)
808 				bcopy(mbufcp, uiocp, xfer);
809 			else
810 				copyout(mbufcp, uiocp, xfer);
811 			left -= xfer;
812 			len -= xfer;
813 			mbufcp += xfer;
814 			uiocp += xfer;
815 			uiop->uio_offset += xfer;
816 			uiop->uio_resid -= xfer;
817 		}
818 		if (uiop->uio_iov->iov_len <= siz) {
819 			uiop->uio_iovcnt--;
820 			uiop->uio_iov++;
821 		} else {
822 			uiop->uio_iov->iov_base += uiosiz;
823 			uiop->uio_iov->iov_len -= uiosiz;
824 		}
825 		siz -= uiosiz;
826 	}
827 	*dpos = mbufcp;
828 	*mrep = mp;
829 	if (rem > 0) {
830 		if (len < rem)
831 			error = nfs_adv(mrep, dpos, rem, len);
832 		else
833 			*dpos += rem;
834 	}
835 	return (error);
836 }
837 
838 /*
839  * copies a uio scatter/gather list to an mbuf chain.
840  * NOTE: can ony handle iovcnt == 1
841  */
842 int
843 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
844 {
845 	char *uiocp;
846 	struct mbuf *mp, *mp2;
847 	int xfer, left, mlen;
848 	int uiosiz, clflg, rem;
849 	char *cp;
850 
851 #ifdef DIAGNOSTIC
852 	if (uiop->uio_iovcnt != 1)
853 		panic("nfsm_uiotombuf: iovcnt != 1");
854 #endif
855 
856 	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
857 		clflg = 1;
858 	else
859 		clflg = 0;
860 	rem = nfsm_rndup(siz)-siz;
861 	mp = mp2 = *mq;
862 	while (siz > 0) {
863 		left = uiop->uio_iov->iov_len;
864 		uiocp = uiop->uio_iov->iov_base;
865 		if (left > siz)
866 			left = siz;
867 		uiosiz = left;
868 		while (left > 0) {
869 			mlen = M_TRAILINGSPACE(mp);
870 			if (mlen == 0) {
871 				MGET(mp, MB_WAIT, MT_DATA);
872 				if (clflg)
873 					MCLGET(mp, MB_WAIT);
874 				mp->m_len = 0;
875 				mp2->m_next = mp;
876 				mp2 = mp;
877 				mlen = M_TRAILINGSPACE(mp);
878 			}
879 			xfer = (left > mlen) ? mlen : left;
880 #ifdef notdef
881 			/* Not Yet.. */
882 			if (uiop->uio_iov->iov_op != NULL)
883 				(*(uiop->uio_iov->iov_op))
884 				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
885 			else
886 #endif
887 			if (uiop->uio_segflg == UIO_SYSSPACE)
888 				bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
889 			else
890 				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
891 			mp->m_len += xfer;
892 			left -= xfer;
893 			uiocp += xfer;
894 			uiop->uio_offset += xfer;
895 			uiop->uio_resid -= xfer;
896 		}
897 		uiop->uio_iov->iov_base += uiosiz;
898 		uiop->uio_iov->iov_len -= uiosiz;
899 		siz -= uiosiz;
900 	}
901 	if (rem > 0) {
902 		if (rem > M_TRAILINGSPACE(mp)) {
903 			MGET(mp, MB_WAIT, MT_DATA);
904 			mp->m_len = 0;
905 			mp2->m_next = mp;
906 		}
907 		cp = mtod(mp, caddr_t)+mp->m_len;
908 		for (left = 0; left < rem; left++)
909 			*cp++ = '\0';
910 		mp->m_len += rem;
911 		*bpos = cp;
912 	} else
913 		*bpos = mtod(mp, caddr_t)+mp->m_len;
914 	*mq = mp;
915 	return (0);
916 }
917 
918 /*
919  * Help break down an mbuf chain by setting the first siz bytes contiguous
920  * pointed to by returned val.
921  * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
922  * cases. (The macros use the vars. dpos and dpos2)
923  */
924 int
925 nfsm_disct(struct mbuf **mdp, caddr_t *dposp, int siz, int left, caddr_t *cp2)
926 {
927 	struct mbuf *mp, *mp2;
928 	int siz2, xfer;
929 	caddr_t p;
930 
931 	mp = *mdp;
932 	while (left == 0) {
933 		*mdp = mp = mp->m_next;
934 		if (mp == NULL)
935 			return (EBADRPC);
936 		left = mp->m_len;
937 		*dposp = mtod(mp, caddr_t);
938 	}
939 	if (left >= siz) {
940 		*cp2 = *dposp;
941 		*dposp += siz;
942 	} else if (mp->m_next == NULL) {
943 		return (EBADRPC);
944 	} else if (siz > MHLEN) {
945 		panic("nfs S too big");
946 	} else {
947 		MGET(mp2, MB_WAIT, MT_DATA);
948 		mp2->m_next = mp->m_next;
949 		mp->m_next = mp2;
950 		mp->m_len -= left;
951 		mp = mp2;
952 		*cp2 = p = mtod(mp, caddr_t);
953 		bcopy(*dposp, p, left);		/* Copy what was left */
954 		siz2 = siz-left;
955 		p += left;
956 		mp2 = mp->m_next;
957 		/* Loop around copying up the siz2 bytes */
958 		while (siz2 > 0) {
959 			if (mp2 == NULL)
960 				return (EBADRPC);
961 			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
962 			if (xfer > 0) {
963 				bcopy(mtod(mp2, caddr_t), p, xfer);
964 				NFSMADV(mp2, xfer);
965 				mp2->m_len -= xfer;
966 				p += xfer;
967 				siz2 -= xfer;
968 			}
969 			if (siz2 > 0)
970 				mp2 = mp2->m_next;
971 		}
972 		mp->m_len = siz;
973 		*mdp = mp2;
974 		*dposp = mtod(mp2, caddr_t);
975 	}
976 	return (0);
977 }
978 
979 /*
980  * Advance the position in the mbuf chain.
981  */
982 int
983 nfs_adv(struct mbuf **mdp, caddr_t *dposp, int offs, int left)
984 {
985 	struct mbuf *m;
986 	int s;
987 
988 	m = *mdp;
989 	s = left;
990 	while (s < offs) {
991 		offs -= s;
992 		m = m->m_next;
993 		if (m == NULL)
994 			return (EBADRPC);
995 		s = m->m_len;
996 	}
997 	*mdp = m;
998 	*dposp = mtod(m, caddr_t)+offs;
999 	return (0);
1000 }
1001 
1002 /*
1003  * Copy a string into mbufs for the hard cases...
1004  */
1005 int
1006 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
1007 {
1008 	struct mbuf *m1 = NULL, *m2;
1009 	long left, xfer, len, tlen;
1010 	u_int32_t *tl;
1011 	int putsize;
1012 
1013 	putsize = 1;
1014 	m2 = *mb;
1015 	left = M_TRAILINGSPACE(m2);
1016 	if (left > 0) {
1017 		tl = ((u_int32_t *)(*bpos));
1018 		*tl++ = txdr_unsigned(siz);
1019 		putsize = 0;
1020 		left -= NFSX_UNSIGNED;
1021 		m2->m_len += NFSX_UNSIGNED;
1022 		if (left > 0) {
1023 			bcopy(cp, (caddr_t) tl, left);
1024 			siz -= left;
1025 			cp += left;
1026 			m2->m_len += left;
1027 			left = 0;
1028 		}
1029 	}
1030 	/* Loop around adding mbufs */
1031 	while (siz > 0) {
1032 		MGET(m1, MB_WAIT, MT_DATA);
1033 		if (siz > MLEN)
1034 			MCLGET(m1, MB_WAIT);
1035 		m1->m_len = NFSMSIZ(m1);
1036 		m2->m_next = m1;
1037 		m2 = m1;
1038 		tl = mtod(m1, u_int32_t *);
1039 		tlen = 0;
1040 		if (putsize) {
1041 			*tl++ = txdr_unsigned(siz);
1042 			m1->m_len -= NFSX_UNSIGNED;
1043 			tlen = NFSX_UNSIGNED;
1044 			putsize = 0;
1045 		}
1046 		if (siz < m1->m_len) {
1047 			len = nfsm_rndup(siz);
1048 			xfer = siz;
1049 			if (xfer < len)
1050 				*(tl+(xfer>>2)) = 0;
1051 		} else {
1052 			xfer = len = m1->m_len;
1053 		}
1054 		bcopy(cp, (caddr_t) tl, xfer);
1055 		m1->m_len = len+tlen;
1056 		siz -= xfer;
1057 		cp += xfer;
1058 	}
1059 	*mb = m1;
1060 	*bpos = mtod(m1, caddr_t)+m1->m_len;
1061 	return (0);
1062 }
1063 
1064 /*
1065  * Called once to initialize data structures...
1066  */
1067 int
1068 nfs_init(struct vfsconf *vfsp)
1069 {
1070 	int i;
1071 
1072 	nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
1073 
1074 	nfs_mount_type = vfsp->vfc_typenum;
1075 	nfsrtt.pos = 0;
1076 	rpc_vers = txdr_unsigned(RPC_VER2);
1077 	rpc_call = txdr_unsigned(RPC_CALL);
1078 	rpc_reply = txdr_unsigned(RPC_REPLY);
1079 	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1080 	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1081 	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1082 	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1083 	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1084 	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1085 	nfs_prog = txdr_unsigned(NFS_PROG);
1086 	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1087 	nfs_true = txdr_unsigned(TRUE);
1088 	nfs_false = txdr_unsigned(FALSE);
1089 	nfs_xdrneg1 = txdr_unsigned(-1);
1090 	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1091 	if (nfs_ticks < 1)
1092 		nfs_ticks = 1;
1093 	/* Ensure async daemons disabled */
1094 	for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1095 		nfs_iodwant[i] = NULL;
1096 		nfs_iodmount[i] = (struct nfsmount *)0;
1097 	}
1098 	nfs_nhinit();			/* Init the nfsnode table */
1099 #ifndef NFS_NOSERVER
1100 	nfsrv_init(0);			/* Init server data structures */
1101 	nfsrv_initcache();		/* Init the server request cache */
1102 #endif
1103 
1104 	/*
1105 	 * Initialize the nqnfs server stuff.
1106 	 */
1107 	if (nqnfsstarttime == 0) {
1108 		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1109 			+ nqsrv_clockskew + nqsrv_writeslack;
1110 		NQLOADNOVRAM(nqnfsstarttime);
1111 		CIRCLEQ_INIT(&nqtimerhead);
1112 		nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, &nqfhhash);
1113 	}
1114 
1115 	/*
1116 	 * Initialize reply list and start timer
1117 	 */
1118 	TAILQ_INIT(&nfs_reqq);
1119 
1120 	nfs_timer(0);
1121 
1122 	/*
1123 	 * Set up lease_check and lease_updatetime so that other parts
1124 	 * of the system can call us, if we are loadable.
1125 	 */
1126 #ifndef NFS_NOSERVER
1127 	nfs_prev_vop_lease_check = default_vnode_vops->vop_lease;
1128 	default_vnode_vops->vop_lease = nqnfs_vop_lease_check;
1129 #endif
1130 	nfs_prev_lease_updatetime = lease_updatetime;
1131 	lease_updatetime = nfs_lease_updatetime;
1132 	nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1133 	sysent[SYS_nfssvc].sy_narg = 2;
1134 	nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
1135 	sysent[SYS_nfssvc].sy_call = (sy_call_t *)nfssvc;
1136 
1137 	nfs_pbuf_freecnt = nswbuf / 2 + 1;
1138 
1139 	return (0);
1140 }
1141 
1142 int
1143 nfs_uninit(struct vfsconf *vfsp)
1144 {
1145 
1146 	untimeout(nfs_timer, (void *)NULL, nfs_timer_handle);
1147 	nfs_mount_type = -1;
1148 #ifndef NFS_NOSERVER
1149 	default_vnode_vops->vop_lease = nfs_prev_vop_lease_check;
1150 #endif
1151 	lease_updatetime = nfs_prev_lease_updatetime;
1152 	sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1153 	sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1154 	return (0);
1155 }
1156 
1157 /*
1158  * Attribute cache routines.
1159  * nfs_loadattrcache() - loads or updates the cache contents from attributes
1160  *	that are on the mbuf list
1161  * nfs_getattrcache() - returns valid attributes if found in cache, returns
1162  *	error otherwise
1163  */
1164 
1165 /*
1166  * Load the attribute cache (that lives in the nfsnode entry) with
1167  * the values on the mbuf list and
1168  * Iff vap not NULL
1169  *    copy the attributes to *vaper
1170  */
1171 int
1172 nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
1173 		  struct vattr *vaper, int dontshrink)
1174 {
1175 	struct vnode *vp = *vpp;
1176 	struct vattr *vap;
1177 	struct nfs_fattr *fp;
1178 	struct nfsnode *np;
1179 	int32_t t1;
1180 	caddr_t cp2;
1181 	int error = 0;
1182 	udev_t rdev;
1183 	struct mbuf *md;
1184 	enum vtype vtyp;
1185 	u_short vmode;
1186 	struct timespec mtime;
1187 	int v3 = NFS_ISV3(vp);
1188 
1189 	md = *mdp;
1190 	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1191 	if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1192 		return (error);
1193 	fp = (struct nfs_fattr *)cp2;
1194 	if (v3) {
1195 		vtyp = nfsv3tov_type(fp->fa_type);
1196 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1197 		rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
1198 			fxdr_unsigned(int, fp->fa3_rdev.specdata2));
1199 		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1200 	} else {
1201 		vtyp = nfsv2tov_type(fp->fa_type);
1202 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1203 		/*
1204 		 * XXX
1205 		 *
1206 		 * The duplicate information returned in fa_type and fa_mode
1207 		 * is an ambiguity in the NFS version 2 protocol.
1208 		 *
1209 		 * VREG should be taken literally as a regular file.  If a
1210 		 * server intents to return some type information differently
1211 		 * in the upper bits of the mode field (e.g. for sockets, or
1212 		 * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
1213 		 * leave the examination of the mode bits even in the VREG
1214 		 * case to avoid breakage for bogus servers, but we make sure
1215 		 * that there are actually type bits set in the upper part of
1216 		 * fa_mode (and failing that, trust the va_type field).
1217 		 *
1218 		 * NFSv3 cleared the issue, and requires fa_mode to not
1219 		 * contain any type information (while also introduing sockets
1220 		 * and FIFOs for fa_type).
1221 		 */
1222 		if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1223 			vtyp = IFTOVT(vmode);
1224 		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1225 		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1226 
1227 		/*
1228 		 * Really ugly NFSv2 kludge.
1229 		 */
1230 		if (vtyp == VCHR && rdev == (udev_t)0xffffffff)
1231 			vtyp = VFIFO;
1232 	}
1233 
1234 	/*
1235 	 * If v_type == VNON it is a new node, so fill in the v_type,
1236 	 * n_mtime fields. Check to see if it represents a special
1237 	 * device, and if so, check for a possible alias. Once the
1238 	 * correct vnode has been obtained, fill in the rest of the
1239 	 * information.
1240 	 */
1241 	np = VTONFS(vp);
1242 	if (vp->v_type != vtyp) {
1243 		vp->v_type = vtyp;
1244 		if (vp->v_type == VFIFO) {
1245 			vp->v_ops = vp->v_mount->mnt_vn_fifo_ops;
1246 		} else if (vp->v_type == VCHR || vp->v_type == VBLK) {
1247 			vp->v_ops = vp->v_mount->mnt_vn_spec_ops;
1248 			addaliasu(vp, rdev);
1249 		} else {
1250 			vp->v_ops = vp->v_mount->mnt_vn_ops;
1251 		}
1252 		np->n_mtime = mtime.tv_sec;
1253 	}
1254 	vap = &np->n_vattr;
1255 	vap->va_type = vtyp;
1256 	vap->va_mode = (vmode & 07777);
1257 	vap->va_rdev = rdev;
1258 	vap->va_mtime = mtime;
1259 	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1260 	if (v3) {
1261 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1262 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1263 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1264 		vap->va_size = fxdr_hyper(&fp->fa3_size);
1265 		vap->va_blocksize = NFS_FABLKSIZE;
1266 		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1267 		vap->va_fileid = fxdr_unsigned(int32_t,
1268 		    fp->fa3_fileid.nfsuquad[1]);
1269 		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1270 		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1271 		vap->va_flags = 0;
1272 		vap->va_filerev = 0;
1273 	} else {
1274 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1275 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1276 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1277 		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1278 		vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1279 		vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1280 		    * NFS_FABLKSIZE;
1281 		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1282 		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1283 		vap->va_flags = 0;
1284 		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1285 		    fp->fa2_ctime.nfsv2_sec);
1286 		vap->va_ctime.tv_nsec = 0;
1287 		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1288 		vap->va_filerev = 0;
1289 	}
1290 	np->n_attrstamp = time_second;
1291 	if (vap->va_size != np->n_size) {
1292 		if (vap->va_type == VREG) {
1293 			if (dontshrink && vap->va_size < np->n_size) {
1294 				/*
1295 				 * We've been told not to shrink the file;
1296 				 * zero np->n_attrstamp to indicate that
1297 				 * the attributes are stale.
1298 				 */
1299 				vap->va_size = np->n_size;
1300 				np->n_attrstamp = 0;
1301 			} else if (np->n_flag & NMODIFIED) {
1302 				/*
1303 				 * We've modified the file: Use the larger
1304 				 * of our size, and the server's size.
1305 				 */
1306 				if (vap->va_size < np->n_size) {
1307 					vap->va_size = np->n_size;
1308 				} else {
1309 					np->n_size = vap->va_size;
1310 					np->n_flag |= NSIZECHANGED;
1311 				}
1312 			} else {
1313 				np->n_size = vap->va_size;
1314 				np->n_flag |= NSIZECHANGED;
1315 			}
1316 			vnode_pager_setsize(vp, np->n_size);
1317 		} else {
1318 			np->n_size = vap->va_size;
1319 		}
1320 	}
1321 	if (vaper != NULL) {
1322 		bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1323 		if (np->n_flag & NCHG) {
1324 			if (np->n_flag & NACC)
1325 				vaper->va_atime = np->n_atim;
1326 			if (np->n_flag & NUPD)
1327 				vaper->va_mtime = np->n_mtim;
1328 		}
1329 	}
1330 	return (0);
1331 }
1332 
1333 #ifdef NFS_ACDEBUG
1334 #include <sys/sysctl.h>
1335 SYSCTL_DECL(_vfs_nfs);
1336 static int nfs_acdebug;
1337 SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1338 #endif
1339 
1340 /*
1341  * Check the time stamp
1342  * If the cache is valid, copy contents to *vap and return 0
1343  * otherwise return an error
1344  */
1345 int
1346 nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
1347 {
1348 	struct nfsnode *np;
1349 	struct vattr *vap;
1350 	struct nfsmount *nmp;
1351 	int timeo;
1352 
1353 	np = VTONFS(vp);
1354 	vap = &np->n_vattr;
1355 	nmp = VFSTONFS(vp->v_mount);
1356 	/* XXX n_mtime doesn't seem to be updated on a miss-and-reload */
1357 	timeo = (time_second - np->n_mtime) / 10;
1358 
1359 #ifdef NFS_ACDEBUG
1360 	if (nfs_acdebug>1)
1361 		printf("nfs_getattrcache: initial timeo = %d\n", timeo);
1362 #endif
1363 
1364 	if (vap->va_type == VDIR) {
1365 		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
1366 			timeo = nmp->nm_acdirmin;
1367 		else if (timeo > nmp->nm_acdirmax)
1368 			timeo = nmp->nm_acdirmax;
1369 	} else {
1370 		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
1371 			timeo = nmp->nm_acregmin;
1372 		else if (timeo > nmp->nm_acregmax)
1373 			timeo = nmp->nm_acregmax;
1374 	}
1375 
1376 #ifdef NFS_ACDEBUG
1377 	if (nfs_acdebug > 2)
1378 		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1379 			nmp->nm_acregmin, nmp->nm_acregmax,
1380 			nmp->nm_acdirmin, nmp->nm_acdirmax);
1381 
1382 	if (nfs_acdebug)
1383 		printf("nfs_getattrcache: age = %d; final timeo = %d\n",
1384 			(time_second - np->n_attrstamp), timeo);
1385 #endif
1386 
1387 	if ((time_second - np->n_attrstamp) >= timeo) {
1388 		nfsstats.attrcache_misses++;
1389 		return (ENOENT);
1390 	}
1391 	nfsstats.attrcache_hits++;
1392 	if (vap->va_size != np->n_size) {
1393 		if (vap->va_type == VREG) {
1394 			if (np->n_flag & NMODIFIED) {
1395 				if (vap->va_size < np->n_size)
1396 					vap->va_size = np->n_size;
1397 				else
1398 					np->n_size = vap->va_size;
1399 			} else {
1400 				np->n_size = vap->va_size;
1401 			}
1402 			vnode_pager_setsize(vp, np->n_size);
1403 		} else {
1404 			np->n_size = vap->va_size;
1405 		}
1406 	}
1407 	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1408 	if (np->n_flag & NCHG) {
1409 		if (np->n_flag & NACC)
1410 			vaper->va_atime = np->n_atim;
1411 		if (np->n_flag & NUPD)
1412 			vaper->va_mtime = np->n_mtim;
1413 	}
1414 	return (0);
1415 }
1416 
1417 #ifndef NFS_NOSERVER
1418 /*
1419  * Set up nameidata for a lookup() call and do it.
1420  *
1421  * If pubflag is set, this call is done for a lookup operation on the
1422  * public filehandle. In that case we allow crossing mountpoints and
1423  * absolute pathnames. However, the caller is expected to check that
1424  * the lookup result is within the public fs, and deny access if
1425  * it is not.
1426  *
1427  * nfs_namei() clears out garbage fields that namei() might leave garbage.
1428  * This is mainly ni_vp and ni_dvp when an error occurs, and ni_dvp when no
1429  * error occurs but the parent was not requested.
1430  *
1431  * dirp may be set whether an error is returned or not, and must be
1432  * released by the caller.
1433  */
1434 int
1435 nfs_namei(struct nameidata *ndp, fhandle_t *fhp, int len,
1436 	  struct nfssvc_sock *slp, struct sockaddr *nam, struct mbuf **mdp,
1437 	  caddr_t *dposp, struct vnode **retdirp, struct thread *td,
1438 	  int kerbflag, int pubflag)
1439 {
1440 	int i, rem;
1441 	struct mbuf *md;
1442 	char *fromcp, *tocp, *cp;
1443 	struct iovec aiov;
1444 	struct uio auio;
1445 	struct vnode *dp;
1446 	int error, rdonly, linklen;
1447 	struct componentname *cnp = &ndp->ni_cnd;
1448 
1449 	*retdirp = (struct vnode *)0;
1450 	cnp->cn_pnbuf = zalloc(namei_zone);
1451 
1452 	/*
1453 	 * Copy the name from the mbuf list to ndp->ni_pnbuf
1454 	 * and set the various ndp fields appropriately.
1455 	 */
1456 	fromcp = *dposp;
1457 	tocp = cnp->cn_pnbuf;
1458 	md = *mdp;
1459 	rem = mtod(md, caddr_t) + md->m_len - fromcp;
1460 	for (i = 0; i < len; i++) {
1461 		while (rem == 0) {
1462 			md = md->m_next;
1463 			if (md == NULL) {
1464 				error = EBADRPC;
1465 				goto out;
1466 			}
1467 			fromcp = mtod(md, caddr_t);
1468 			rem = md->m_len;
1469 		}
1470 		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1471 			error = EACCES;
1472 			goto out;
1473 		}
1474 		*tocp++ = *fromcp++;
1475 		rem--;
1476 	}
1477 	*tocp = '\0';
1478 	*mdp = md;
1479 	*dposp = fromcp;
1480 	len = nfsm_rndup(len)-len;
1481 	if (len > 0) {
1482 		if (rem >= len)
1483 			*dposp += len;
1484 		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1485 			goto out;
1486 	}
1487 
1488 	/*
1489 	 * Extract and set starting directory.
1490 	 */
1491 	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
1492 	    nam, &rdonly, kerbflag, pubflag);
1493 	if (error)
1494 		goto out;
1495 	if (dp->v_type != VDIR) {
1496 		vrele(dp);
1497 		error = ENOTDIR;
1498 		goto out;
1499 	}
1500 
1501 	if (rdonly)
1502 		cnp->cn_flags |= CNP_RDONLY;
1503 
1504 	/*
1505 	 * Set return directory.  Reference to dp is implicitly transfered
1506 	 * to the returned pointer
1507 	 */
1508 	*retdirp = dp;
1509 
1510 	if (pubflag) {
1511 		/*
1512 		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1513 		 * and the 'native path' indicator.
1514 		 */
1515 		cp = zalloc(namei_zone);
1516 		fromcp = cnp->cn_pnbuf;
1517 		tocp = cp;
1518 		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1519 			switch ((unsigned char)*fromcp) {
1520 			case WEBNFS_NATIVE_CHAR:
1521 				/*
1522 				 * 'Native' path for us is the same
1523 				 * as a path according to the NFS spec,
1524 				 * just skip the escape char.
1525 				 */
1526 				fromcp++;
1527 				break;
1528 			/*
1529 			 * More may be added in the future, range 0x80-0xff
1530 			 */
1531 			default:
1532 				error = EIO;
1533 				zfree(namei_zone, cp);
1534 				goto out;
1535 			}
1536 		}
1537 		/*
1538 		 * Translate the '%' escapes, URL-style.
1539 		 */
1540 		while (*fromcp != '\0') {
1541 			if (*fromcp == WEBNFS_ESC_CHAR) {
1542 				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1543 					fromcp++;
1544 					*tocp++ = HEXSTRTOI(fromcp);
1545 					fromcp += 2;
1546 					continue;
1547 				} else {
1548 					error = ENOENT;
1549 					zfree(namei_zone, cp);
1550 					goto out;
1551 				}
1552 			} else
1553 				*tocp++ = *fromcp++;
1554 		}
1555 		*tocp = '\0';
1556 		zfree(namei_zone, cnp->cn_pnbuf);
1557 		cnp->cn_pnbuf = cp;
1558 	}
1559 
1560 	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
1561 	ndp->ni_segflg = UIO_SYSSPACE;
1562 
1563 	if (pubflag) {
1564 		ndp->ni_rootdir = rootvnode;
1565 		ndp->ni_loopcnt = 0;
1566 		if (cnp->cn_pnbuf[0] == '/')
1567 			dp = rootvnode;
1568 	} else {
1569 		cnp->cn_flags |= CNP_NOCROSSMOUNT;
1570 	}
1571 
1572 	/*
1573 	 * Initialize for scan, set ni_startdir and bump ref on dp again
1574 	 * because lookup() will dereference ni_startdir.
1575 	 */
1576 
1577 	cnp->cn_td = td;
1578 	vref(dp);
1579 	ndp->ni_startdir = dp;
1580 
1581 	for (;;) {
1582 		cnp->cn_nameptr = cnp->cn_pnbuf;
1583 		/*
1584 		 * Call lookup() to do the real work.  If an error occurs,
1585 		 * ndp->ni_vp and ni_dvp are left uninitialized or NULL and
1586 		 * we do not have to dereference anything before returning.
1587 		 * In either case ni_startdir will be dereferenced and NULLed
1588 		 * out.
1589 		 */
1590 		error = lookup(ndp);
1591 		if (error)
1592 			break;
1593 
1594 		/*
1595 		 * Check for encountering a symbolic link.  Trivial
1596 		 * termination occurs if no symlink encountered.
1597 		 * Note: zfree is safe because error is 0, so we will
1598 		 * not zfree it again when we break.
1599 		 */
1600 		if ((cnp->cn_flags & CNP_ISSYMLINK) == 0) {
1601 			nfsrv_object_create(ndp->ni_vp);
1602 			if (cnp->cn_flags & (CNP_SAVENAME | CNP_SAVESTART))
1603 				cnp->cn_flags |= CNP_HASBUF;
1604 			else
1605 				zfree(namei_zone, cnp->cn_pnbuf);
1606 			break;
1607 		}
1608 
1609 		/*
1610 		 * Validate symlink
1611 		 */
1612 		if ((cnp->cn_flags & CNP_LOCKPARENT) && ndp->ni_pathlen == 1)
1613 			VOP_UNLOCK(ndp->ni_dvp, NULL, 0, td);
1614 		if (!pubflag) {
1615 			error = EINVAL;
1616 			goto badlink2;
1617 		}
1618 
1619 		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
1620 			error = ELOOP;
1621 			goto badlink2;
1622 		}
1623 		if (ndp->ni_pathlen > 1)
1624 			cp = zalloc(namei_zone);
1625 		else
1626 			cp = cnp->cn_pnbuf;
1627 		aiov.iov_base = cp;
1628 		aiov.iov_len = MAXPATHLEN;
1629 		auio.uio_iov = &aiov;
1630 		auio.uio_iovcnt = 1;
1631 		auio.uio_offset = 0;
1632 		auio.uio_rw = UIO_READ;
1633 		auio.uio_segflg = UIO_SYSSPACE;
1634 		auio.uio_td = NULL;
1635 		auio.uio_resid = MAXPATHLEN;
1636 		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
1637 		if (error) {
1638 		badlink1:
1639 			if (ndp->ni_pathlen > 1)
1640 				zfree(namei_zone, cp);
1641 		badlink2:
1642 			vrele(ndp->ni_dvp);
1643 			vput(ndp->ni_vp);
1644 			break;
1645 		}
1646 		linklen = MAXPATHLEN - auio.uio_resid;
1647 		if (linklen == 0) {
1648 			error = ENOENT;
1649 			goto badlink1;
1650 		}
1651 		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
1652 			error = ENAMETOOLONG;
1653 			goto badlink1;
1654 		}
1655 
1656 		/*
1657 		 * Adjust or replace path
1658 		 */
1659 		if (ndp->ni_pathlen > 1) {
1660 			bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
1661 			zfree(namei_zone, cnp->cn_pnbuf);
1662 			cnp->cn_pnbuf = cp;
1663 		} else
1664 			cnp->cn_pnbuf[linklen] = '\0';
1665 		ndp->ni_pathlen += linklen;
1666 
1667 		/*
1668 		 * Cleanup refs for next loop and check if root directory
1669 		 * should replace current directory.  Normally ni_dvp
1670 		 * becomes the new base directory and is cleaned up when
1671 		 * we loop.  Explicitly null pointers after invalidation
1672 		 * to clarify operation.
1673 		 */
1674 		vput(ndp->ni_vp);
1675 		ndp->ni_vp = NULL;
1676 
1677 		if (cnp->cn_pnbuf[0] == '/') {
1678 			vrele(ndp->ni_dvp);
1679 			ndp->ni_dvp = ndp->ni_rootdir;
1680 			vref(ndp->ni_dvp);
1681 		}
1682 		ndp->ni_startdir = ndp->ni_dvp;
1683 		ndp->ni_dvp = NULL;
1684 	}
1685 
1686 	/*
1687 	 * nfs_namei() guarentees that fields will not contain garbage
1688 	 * whether an error occurs or not.  This allows the caller to track
1689 	 * cleanup state trivially.
1690 	 */
1691 out:
1692 	if (error) {
1693 		zfree(namei_zone, cnp->cn_pnbuf);
1694 		ndp->ni_vp = NULL;
1695 		ndp->ni_dvp = NULL;
1696 		ndp->ni_startdir = NULL;
1697 		cnp->cn_flags &= ~CNP_HASBUF;
1698 	} else if ((ndp->ni_cnd.cn_flags & (CNP_WANTPARENT|CNP_LOCKPARENT)) == 0) {
1699 		ndp->ni_dvp = NULL;
1700 	}
1701 	return (error);
1702 }
1703 
1704 /*
1705  * A fiddled version of m_adj() that ensures null fill to a long
1706  * boundary and only trims off the back end
1707  */
1708 void
1709 nfsm_adj(struct mbuf *mp, int len, int nul)
1710 {
1711 	struct mbuf *m;
1712 	int count, i;
1713 	char *cp;
1714 
1715 	/*
1716 	 * Trim from tail.  Scan the mbuf chain,
1717 	 * calculating its length and finding the last mbuf.
1718 	 * If the adjustment only affects this mbuf, then just
1719 	 * adjust and return.  Otherwise, rescan and truncate
1720 	 * after the remaining size.
1721 	 */
1722 	count = 0;
1723 	m = mp;
1724 	for (;;) {
1725 		count += m->m_len;
1726 		if (m->m_next == (struct mbuf *)0)
1727 			break;
1728 		m = m->m_next;
1729 	}
1730 	if (m->m_len > len) {
1731 		m->m_len -= len;
1732 		if (nul > 0) {
1733 			cp = mtod(m, caddr_t)+m->m_len-nul;
1734 			for (i = 0; i < nul; i++)
1735 				*cp++ = '\0';
1736 		}
1737 		return;
1738 	}
1739 	count -= len;
1740 	if (count < 0)
1741 		count = 0;
1742 	/*
1743 	 * Correct length for chain is "count".
1744 	 * Find the mbuf with last data, adjust its length,
1745 	 * and toss data from remaining mbufs on chain.
1746 	 */
1747 	for (m = mp; m; m = m->m_next) {
1748 		if (m->m_len >= count) {
1749 			m->m_len = count;
1750 			if (nul > 0) {
1751 				cp = mtod(m, caddr_t)+m->m_len-nul;
1752 				for (i = 0; i < nul; i++)
1753 					*cp++ = '\0';
1754 			}
1755 			break;
1756 		}
1757 		count -= m->m_len;
1758 	}
1759 	for (m = m->m_next;m;m = m->m_next)
1760 		m->m_len = 0;
1761 }
1762 
1763 /*
1764  * Make these functions instead of macros, so that the kernel text size
1765  * doesn't get too big...
1766  */
1767 void
1768 nfsm_srvwcc(struct nfsrv_descript *nfsd, int before_ret,
1769 	    struct vattr *before_vap, int after_ret, struct vattr *after_vap,
1770 	    struct mbuf **mbp, char **bposp)
1771 {
1772 	struct mbuf *mb = *mbp, *mb2;
1773 	char *bpos = *bposp;
1774 	u_int32_t *tl;
1775 
1776 	if (before_ret) {
1777 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1778 		*tl = nfs_false;
1779 	} else {
1780 		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1781 		*tl++ = nfs_true;
1782 		txdr_hyper(before_vap->va_size, tl);
1783 		tl += 2;
1784 		txdr_nfsv3time(&(before_vap->va_mtime), tl);
1785 		tl += 2;
1786 		txdr_nfsv3time(&(before_vap->va_ctime), tl);
1787 	}
1788 	*bposp = bpos;
1789 	*mbp = mb;
1790 	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1791 }
1792 
1793 void
1794 nfsm_srvpostopattr(struct nfsrv_descript *nfsd, int after_ret,
1795 		   struct vattr *after_vap, struct mbuf **mbp, char **bposp)
1796 {
1797 	struct mbuf *mb = *mbp, *mb2;
1798 	char *bpos = *bposp;
1799 	u_int32_t *tl;
1800 	struct nfs_fattr *fp;
1801 
1802 	if (after_ret) {
1803 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1804 		*tl = nfs_false;
1805 	} else {
1806 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1807 		*tl++ = nfs_true;
1808 		fp = (struct nfs_fattr *)tl;
1809 		nfsm_srvfattr(nfsd, after_vap, fp);
1810 	}
1811 	*mbp = mb;
1812 	*bposp = bpos;
1813 }
1814 
1815 void
1816 nfsm_srvfattr(struct nfsrv_descript *nfsd, struct vattr *vap,
1817 	      struct nfs_fattr *fp)
1818 {
1819 
1820 	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1821 	fp->fa_uid = txdr_unsigned(vap->va_uid);
1822 	fp->fa_gid = txdr_unsigned(vap->va_gid);
1823 	if (nfsd->nd_flag & ND_NFSV3) {
1824 		fp->fa_type = vtonfsv3_type(vap->va_type);
1825 		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1826 		txdr_hyper(vap->va_size, &fp->fa3_size);
1827 		txdr_hyper(vap->va_bytes, &fp->fa3_used);
1828 		fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
1829 		fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(vap->va_rdev));
1830 		fp->fa3_fsid.nfsuquad[0] = 0;
1831 		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1832 		fp->fa3_fileid.nfsuquad[0] = 0;
1833 		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
1834 		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1835 		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1836 		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1837 	} else {
1838 		fp->fa_type = vtonfsv2_type(vap->va_type);
1839 		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1840 		fp->fa2_size = txdr_unsigned(vap->va_size);
1841 		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1842 		if (vap->va_type == VFIFO)
1843 			fp->fa2_rdev = 0xffffffff;
1844 		else
1845 			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
1846 		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1847 		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1848 		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1849 		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1850 		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1851 		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1852 	}
1853 }
1854 
1855 /*
1856  * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1857  * 	- look up fsid in mount list (if not found ret error)
1858  *	- get vp and export rights by calling VFS_FHTOVP()
1859  *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1860  *	- if not lockflag unlock it with VOP_UNLOCK()
1861  */
1862 int
1863 nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
1864 	     struct ucred *cred, struct nfssvc_sock *slp, struct sockaddr *nam,
1865 	     int *rdonlyp, int kerbflag, int pubflag)
1866 {
1867 	struct thread *td = curthread; /* XXX */
1868 	struct mount *mp;
1869 	int i;
1870 	struct ucred *credanon;
1871 	int error, exflags;
1872 #ifdef MNT_EXNORESPORT		/* XXX needs mountd and /etc/exports help yet */
1873 	struct sockaddr_int *saddr;
1874 #endif
1875 
1876 	*vpp = (struct vnode *)0;
1877 
1878 	if (nfs_ispublicfh(fhp)) {
1879 		if (!pubflag || !nfs_pub.np_valid)
1880 			return (ESTALE);
1881 		fhp = &nfs_pub.np_handle;
1882 	}
1883 
1884 	mp = vfs_getvfs(&fhp->fh_fsid);
1885 	if (!mp)
1886 		return (ESTALE);
1887 	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1888 	if (error)
1889 		return (error);
1890 	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1891 	if (error)
1892 		return (error);
1893 #ifdef MNT_EXNORESPORT
1894 	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1895 		saddr = (struct sockaddr_in *)nam;
1896 		if (saddr->sin_family == AF_INET &&
1897 		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1898 			vput(*vpp);
1899 			*vpp = NULL;
1900 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1901 		}
1902 	}
1903 #endif
1904 	/*
1905 	 * Check/setup credentials.
1906 	 */
1907 	if (exflags & MNT_EXKERB) {
1908 		if (!kerbflag) {
1909 			vput(*vpp);
1910 			*vpp = NULL;
1911 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1912 		}
1913 	} else if (kerbflag) {
1914 		vput(*vpp);
1915 		*vpp = NULL;
1916 		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1917 	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1918 		cred->cr_uid = credanon->cr_uid;
1919 		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1920 			cred->cr_groups[i] = credanon->cr_groups[i];
1921 		cred->cr_ngroups = i;
1922 	}
1923 	if (exflags & MNT_EXRDONLY)
1924 		*rdonlyp = 1;
1925 	else
1926 		*rdonlyp = 0;
1927 
1928 	nfsrv_object_create(*vpp);
1929 
1930 	if (!lockflag)
1931 		VOP_UNLOCK(*vpp, NULL, 0, td);
1932 	return (0);
1933 }
1934 
1935 
1936 /*
1937  * WebNFS: check if a filehandle is a public filehandle. For v3, this
1938  * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
1939  * transformed this to all zeroes in both cases, so check for it.
1940  */
1941 int
1942 nfs_ispublicfh(fhandle_t *fhp)
1943 {
1944 	char *cp = (char *)fhp;
1945 	int i;
1946 
1947 	for (i = 0; i < NFSX_V3FH; i++)
1948 		if (*cp++ != 0)
1949 			return (FALSE);
1950 	return (TRUE);
1951 }
1952 
1953 #endif /* NFS_NOSERVER */
1954 /*
1955  * This function compares two net addresses by family and returns TRUE
1956  * if they are the same host.
1957  * If there is any doubt, return FALSE.
1958  * The AF_INET family is handled as a special case so that address mbufs
1959  * don't need to be saved to store "struct in_addr", which is only 4 bytes.
1960  */
1961 int
1962 netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
1963 {
1964 	struct sockaddr_in *inetaddr;
1965 
1966 	switch (family) {
1967 	case AF_INET:
1968 		inetaddr = (struct sockaddr_in *)nam;
1969 		if (inetaddr->sin_family == AF_INET &&
1970 		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
1971 			return (1);
1972 		break;
1973 	default:
1974 		break;
1975 	};
1976 	return (0);
1977 }
1978 
1979 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
1980 /*
1981  * This function finds the directory cookie that corresponds to the
1982  * logical byte offset given.
1983  */
1984 nfsuint64 *
1985 nfs_getcookie(struct nfsnode *np, off_t off, int add)
1986 {
1987 	struct nfsdmap *dp, *dp2;
1988 	int pos;
1989 
1990 	pos = (uoff_t)off / NFS_DIRBLKSIZ;
1991 	if (pos == 0 || off < 0) {
1992 #ifdef DIAGNOSTIC
1993 		if (add)
1994 			panic("nfs getcookie add at <= 0");
1995 #endif
1996 		return (&nfs_nullcookie);
1997 	}
1998 	pos--;
1999 	dp = np->n_cookies.lh_first;
2000 	if (!dp) {
2001 		if (add) {
2002 			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
2003 				M_NFSDIROFF, M_WAITOK);
2004 			dp->ndm_eocookie = 0;
2005 			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
2006 		} else
2007 			return ((nfsuint64 *)0);
2008 	}
2009 	while (pos >= NFSNUMCOOKIES) {
2010 		pos -= NFSNUMCOOKIES;
2011 		if (dp->ndm_list.le_next) {
2012 			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
2013 				pos >= dp->ndm_eocookie)
2014 				return ((nfsuint64 *)0);
2015 			dp = dp->ndm_list.le_next;
2016 		} else if (add) {
2017 			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
2018 				M_NFSDIROFF, M_WAITOK);
2019 			dp2->ndm_eocookie = 0;
2020 			LIST_INSERT_AFTER(dp, dp2, ndm_list);
2021 			dp = dp2;
2022 		} else
2023 			return ((nfsuint64 *)0);
2024 	}
2025 	if (pos >= dp->ndm_eocookie) {
2026 		if (add)
2027 			dp->ndm_eocookie = pos + 1;
2028 		else
2029 			return ((nfsuint64 *)0);
2030 	}
2031 	return (&dp->ndm_cookies[pos]);
2032 }
2033 
2034 /*
2035  * Invalidate cached directory information, except for the actual directory
2036  * blocks (which are invalidated separately).
2037  * Done mainly to avoid the use of stale offset cookies.
2038  */
2039 void
2040 nfs_invaldir(struct vnode *vp)
2041 {
2042 	struct nfsnode *np = VTONFS(vp);
2043 
2044 #ifdef DIAGNOSTIC
2045 	if (vp->v_type != VDIR)
2046 		panic("nfs: invaldir not dir");
2047 #endif
2048 	np->n_direofoffset = 0;
2049 	np->n_cookieverf.nfsuquad[0] = 0;
2050 	np->n_cookieverf.nfsuquad[1] = 0;
2051 	if (np->n_cookies.lh_first)
2052 		np->n_cookies.lh_first->ndm_eocookie = 0;
2053 }
2054 
2055 /*
2056  * The write verifier has changed (probably due to a server reboot), so all
2057  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2058  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2059  * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
2060  * mount point.
2061  *
2062  * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
2063  * writes are not clusterable.
2064  */
2065 void
2066 nfs_clearcommit(struct mount *mp)
2067 {
2068 	struct vnode *vp, *nvp;
2069 	struct buf *bp, *nbp;
2070 	lwkt_tokref ilock;
2071 	int s;
2072 
2073 	lwkt_gettoken(&ilock, &mntvnode_token);
2074 	s = splbio();
2075 	for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp; vp = nvp) {
2076 		nvp = TAILQ_NEXT(vp, v_nmntvnodes);	/* ZZZ */
2077 		if (vp->v_flag & VPLACEMARKER)
2078 			continue;
2079 		for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2080 			nbp = TAILQ_NEXT(bp, b_vnbufs);
2081 			if (BUF_REFCNT(bp) == 0 &&
2082 			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2083 			     == (B_DELWRI | B_NEEDCOMMIT)) {
2084 				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
2085 			}
2086 		}
2087 	}
2088 	splx(s);
2089 	lwkt_reltoken(&ilock);
2090 }
2091 
2092 #ifndef NFS_NOSERVER
2093 /*
2094  * Map errnos to NFS error numbers. For Version 3 also filter out error
2095  * numbers not specified for the associated procedure.
2096  */
2097 int
2098 nfsrv_errmap(struct nfsrv_descript *nd, int err)
2099 {
2100 	short *defaulterrp, *errp;
2101 
2102 	if (nd->nd_flag & ND_NFSV3) {
2103 	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
2104 		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2105 		while (*++errp) {
2106 			if (*errp == err)
2107 				return (err);
2108 			else if (*errp > err)
2109 				break;
2110 		}
2111 		return ((int)*defaulterrp);
2112 	    } else
2113 		return (err & 0xffff);
2114 	}
2115 	if (err <= ELAST)
2116 		return ((int)nfsrv_v2errmap[err - 1]);
2117 	return (NFSERR_IO);
2118 }
2119 
2120 int
2121 nfsrv_object_create(struct vnode *vp)
2122 {
2123 	struct thread *td = curthread;
2124 
2125 	if (vp == NULL || vp->v_type != VREG)
2126 		return (1);
2127 	return (vfs_object_create(vp, td));
2128 }
2129 
2130 /*
2131  * Sort the group list in increasing numerical order.
2132  * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2133  *  that used to be here.)
2134  */
2135 void
2136 nfsrvw_sort(gid_t *list, int num)
2137 {
2138 	int i, j;
2139 	gid_t v;
2140 
2141 	/* Insertion sort. */
2142 	for (i = 1; i < num; i++) {
2143 		v = list[i];
2144 		/* find correct slot for value v, moving others up */
2145 		for (j = i; --j >= 0 && v < list[j];)
2146 			list[j + 1] = list[j];
2147 		list[j + 1] = v;
2148 	}
2149 }
2150 
2151 /*
2152  * copy credentials making sure that the result can be compared with bcmp().
2153  */
2154 void
2155 nfsrv_setcred(struct ucred *incred, struct ucred *outcred)
2156 {
2157 	int i;
2158 
2159 	bzero((caddr_t)outcred, sizeof (struct ucred));
2160 	outcred->cr_ref = 1;
2161 	outcred->cr_uid = incred->cr_uid;
2162 	outcred->cr_ngroups = incred->cr_ngroups;
2163 	for (i = 0; i < incred->cr_ngroups; i++)
2164 		outcred->cr_groups[i] = incred->cr_groups[i];
2165 	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2166 }
2167 #endif /* NFS_NOSERVER */
2168