xref: /dflybsd-src/sys/vfs/nfs/nfs_subs.c (revision 23c32883e759b0ea42fdaff39e661bd1a12e3b9f)
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 1.25 2005/03/13 22:17:40 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/nlookup.h>
54 #include <sys/namei.h>
55 #include <sys/mbuf.h>
56 #include <sys/socket.h>
57 #include <sys/stat.h>
58 #include <sys/malloc.h>
59 #include <sys/sysent.h>
60 #include <sys/syscall.h>
61 #include <sys/conf.h>
62 
63 #include <vm/vm.h>
64 #include <vm/vm_object.h>
65 #include <vm/vm_extern.h>
66 #include <vm/vm_zone.h>
67 
68 #include <sys/buf2.h>
69 
70 #include "rpcv2.h"
71 #include "nfsproto.h"
72 #include "nfs.h"
73 #include "nfsmount.h"
74 #include "nfsnode.h"
75 #include "xdr_subs.h"
76 #include "nfsm_subs.h"
77 #include "nqnfs.h"
78 #include "nfsrtt.h"
79 
80 #include <netinet/in.h>
81 
82 /*
83  * Data items converted to xdr at startup, since they are constant
84  * This is kinda hokey, but may save a little time doing byte swaps
85  */
86 u_int32_t nfs_xdrneg1;
87 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
88 	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
89 	rpc_auth_kerb;
90 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
91 
92 /* And other global data */
93 static u_int32_t nfs_xid = 0;
94 static enum vtype nv2tov_type[8]= {
95 	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
96 };
97 enum vtype nv3tov_type[8]= {
98 	VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
99 };
100 
101 int nfs_ticks;
102 int nfs_pbuf_freecnt = -1;	/* start out unlimited */
103 
104 struct nfs_reqq nfs_reqq;
105 struct nfssvc_sockhead nfssvc_sockhead;
106 int nfssvc_sockhead_flag;
107 struct nfsd_head nfsd_head;
108 int nfsd_head_flag;
109 struct nfs_bufq nfs_bufq;
110 struct nqtimerhead nqtimerhead;
111 struct nqfhhashhead *nqfhhashtbl;
112 u_long nqfhhash;
113 
114 static void (*nfs_prev_lease_updatetime) (int);
115 static int nfs_prev_nfssvc_sy_narg;
116 static sy_call_t *nfs_prev_nfssvc_sy_call;
117 
118 #ifndef NFS_NOSERVER
119 
120 static int (*nfs_prev_vop_lease_check)(struct vop_lease_args *);
121 
122 /*
123  * Mapping of old NFS Version 2 RPC numbers to generic numbers.
124  */
125 int nfsv3_procid[NFS_NPROCS] = {
126 	NFSPROC_NULL,
127 	NFSPROC_GETATTR,
128 	NFSPROC_SETATTR,
129 	NFSPROC_NOOP,
130 	NFSPROC_LOOKUP,
131 	NFSPROC_READLINK,
132 	NFSPROC_READ,
133 	NFSPROC_NOOP,
134 	NFSPROC_WRITE,
135 	NFSPROC_CREATE,
136 	NFSPROC_REMOVE,
137 	NFSPROC_RENAME,
138 	NFSPROC_LINK,
139 	NFSPROC_SYMLINK,
140 	NFSPROC_MKDIR,
141 	NFSPROC_RMDIR,
142 	NFSPROC_READDIR,
143 	NFSPROC_FSSTAT,
144 	NFSPROC_NOOP,
145 	NFSPROC_NOOP,
146 	NFSPROC_NOOP,
147 	NFSPROC_NOOP,
148 	NFSPROC_NOOP,
149 	NFSPROC_NOOP,
150 	NFSPROC_NOOP,
151 	NFSPROC_NOOP
152 };
153 
154 #endif /* NFS_NOSERVER */
155 /*
156  * and the reverse mapping from generic to Version 2 procedure numbers
157  */
158 int nfsv2_procid[NFS_NPROCS] = {
159 	NFSV2PROC_NULL,
160 	NFSV2PROC_GETATTR,
161 	NFSV2PROC_SETATTR,
162 	NFSV2PROC_LOOKUP,
163 	NFSV2PROC_NOOP,
164 	NFSV2PROC_READLINK,
165 	NFSV2PROC_READ,
166 	NFSV2PROC_WRITE,
167 	NFSV2PROC_CREATE,
168 	NFSV2PROC_MKDIR,
169 	NFSV2PROC_SYMLINK,
170 	NFSV2PROC_CREATE,
171 	NFSV2PROC_REMOVE,
172 	NFSV2PROC_RMDIR,
173 	NFSV2PROC_RENAME,
174 	NFSV2PROC_LINK,
175 	NFSV2PROC_READDIR,
176 	NFSV2PROC_NOOP,
177 	NFSV2PROC_STATFS,
178 	NFSV2PROC_NOOP,
179 	NFSV2PROC_NOOP,
180 	NFSV2PROC_NOOP,
181 	NFSV2PROC_NOOP,
182 	NFSV2PROC_NOOP,
183 	NFSV2PROC_NOOP,
184 	NFSV2PROC_NOOP,
185 };
186 
187 #ifndef NFS_NOSERVER
188 /*
189  * Maps errno values to nfs error numbers.
190  * Use NFSERR_IO as the catch all for ones not specifically defined in
191  * RFC 1094.
192  */
193 static u_char nfsrv_v2errmap[ELAST] = {
194   NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
195   NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
196   NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
197   NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
198   NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
199   NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
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_IO,	NFSERR_IO,	NFSERR_IO,
206   NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
207   NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
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,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
211   NFSERR_IO /* << Last is 86 */
212 };
213 
214 /*
215  * Maps errno values to nfs error numbers.
216  * Although it is not obvious whether or not NFS clients really care if
217  * a returned error value is in the specified list for the procedure, the
218  * safest thing to do is filter them appropriately. For Version 2, the
219  * X/Open XNFS document is the only specification that defines error values
220  * for each RPC (The RFC simply lists all possible error values for all RPCs),
221  * so I have decided to not do this for Version 2.
222  * The first entry is the default error return and the rest are the valid
223  * errors for that RPC in increasing numeric order.
224  */
225 static short nfsv3err_null[] = {
226 	0,
227 	0,
228 };
229 
230 static short nfsv3err_getattr[] = {
231 	NFSERR_IO,
232 	NFSERR_IO,
233 	NFSERR_STALE,
234 	NFSERR_BADHANDLE,
235 	NFSERR_SERVERFAULT,
236 	0,
237 };
238 
239 static short nfsv3err_setattr[] = {
240 	NFSERR_IO,
241 	NFSERR_PERM,
242 	NFSERR_IO,
243 	NFSERR_ACCES,
244 	NFSERR_INVAL,
245 	NFSERR_NOSPC,
246 	NFSERR_ROFS,
247 	NFSERR_DQUOT,
248 	NFSERR_STALE,
249 	NFSERR_BADHANDLE,
250 	NFSERR_NOT_SYNC,
251 	NFSERR_SERVERFAULT,
252 	0,
253 };
254 
255 static short nfsv3err_lookup[] = {
256 	NFSERR_IO,
257 	NFSERR_NOENT,
258 	NFSERR_IO,
259 	NFSERR_ACCES,
260 	NFSERR_NOTDIR,
261 	NFSERR_NAMETOL,
262 	NFSERR_STALE,
263 	NFSERR_BADHANDLE,
264 	NFSERR_SERVERFAULT,
265 	0,
266 };
267 
268 static short nfsv3err_access[] = {
269 	NFSERR_IO,
270 	NFSERR_IO,
271 	NFSERR_STALE,
272 	NFSERR_BADHANDLE,
273 	NFSERR_SERVERFAULT,
274 	0,
275 };
276 
277 static short nfsv3err_readlink[] = {
278 	NFSERR_IO,
279 	NFSERR_IO,
280 	NFSERR_ACCES,
281 	NFSERR_INVAL,
282 	NFSERR_STALE,
283 	NFSERR_BADHANDLE,
284 	NFSERR_NOTSUPP,
285 	NFSERR_SERVERFAULT,
286 	0,
287 };
288 
289 static short nfsv3err_read[] = {
290 	NFSERR_IO,
291 	NFSERR_IO,
292 	NFSERR_NXIO,
293 	NFSERR_ACCES,
294 	NFSERR_INVAL,
295 	NFSERR_STALE,
296 	NFSERR_BADHANDLE,
297 	NFSERR_SERVERFAULT,
298 	0,
299 };
300 
301 static short nfsv3err_write[] = {
302 	NFSERR_IO,
303 	NFSERR_IO,
304 	NFSERR_ACCES,
305 	NFSERR_INVAL,
306 	NFSERR_FBIG,
307 	NFSERR_NOSPC,
308 	NFSERR_ROFS,
309 	NFSERR_DQUOT,
310 	NFSERR_STALE,
311 	NFSERR_BADHANDLE,
312 	NFSERR_SERVERFAULT,
313 	0,
314 };
315 
316 static short nfsv3err_create[] = {
317 	NFSERR_IO,
318 	NFSERR_IO,
319 	NFSERR_ACCES,
320 	NFSERR_EXIST,
321 	NFSERR_NOTDIR,
322 	NFSERR_NOSPC,
323 	NFSERR_ROFS,
324 	NFSERR_NAMETOL,
325 	NFSERR_DQUOT,
326 	NFSERR_STALE,
327 	NFSERR_BADHANDLE,
328 	NFSERR_NOTSUPP,
329 	NFSERR_SERVERFAULT,
330 	0,
331 };
332 
333 static short nfsv3err_mkdir[] = {
334 	NFSERR_IO,
335 	NFSERR_IO,
336 	NFSERR_ACCES,
337 	NFSERR_EXIST,
338 	NFSERR_NOTDIR,
339 	NFSERR_NOSPC,
340 	NFSERR_ROFS,
341 	NFSERR_NAMETOL,
342 	NFSERR_DQUOT,
343 	NFSERR_STALE,
344 	NFSERR_BADHANDLE,
345 	NFSERR_NOTSUPP,
346 	NFSERR_SERVERFAULT,
347 	0,
348 };
349 
350 static short nfsv3err_symlink[] = {
351 	NFSERR_IO,
352 	NFSERR_IO,
353 	NFSERR_ACCES,
354 	NFSERR_EXIST,
355 	NFSERR_NOTDIR,
356 	NFSERR_NOSPC,
357 	NFSERR_ROFS,
358 	NFSERR_NAMETOL,
359 	NFSERR_DQUOT,
360 	NFSERR_STALE,
361 	NFSERR_BADHANDLE,
362 	NFSERR_NOTSUPP,
363 	NFSERR_SERVERFAULT,
364 	0,
365 };
366 
367 static short nfsv3err_mknod[] = {
368 	NFSERR_IO,
369 	NFSERR_IO,
370 	NFSERR_ACCES,
371 	NFSERR_EXIST,
372 	NFSERR_NOTDIR,
373 	NFSERR_NOSPC,
374 	NFSERR_ROFS,
375 	NFSERR_NAMETOL,
376 	NFSERR_DQUOT,
377 	NFSERR_STALE,
378 	NFSERR_BADHANDLE,
379 	NFSERR_NOTSUPP,
380 	NFSERR_SERVERFAULT,
381 	NFSERR_BADTYPE,
382 	0,
383 };
384 
385 static short nfsv3err_remove[] = {
386 	NFSERR_IO,
387 	NFSERR_NOENT,
388 	NFSERR_IO,
389 	NFSERR_ACCES,
390 	NFSERR_NOTDIR,
391 	NFSERR_ROFS,
392 	NFSERR_NAMETOL,
393 	NFSERR_STALE,
394 	NFSERR_BADHANDLE,
395 	NFSERR_SERVERFAULT,
396 	0,
397 };
398 
399 static short nfsv3err_rmdir[] = {
400 	NFSERR_IO,
401 	NFSERR_NOENT,
402 	NFSERR_IO,
403 	NFSERR_ACCES,
404 	NFSERR_EXIST,
405 	NFSERR_NOTDIR,
406 	NFSERR_INVAL,
407 	NFSERR_ROFS,
408 	NFSERR_NAMETOL,
409 	NFSERR_NOTEMPTY,
410 	NFSERR_STALE,
411 	NFSERR_BADHANDLE,
412 	NFSERR_NOTSUPP,
413 	NFSERR_SERVERFAULT,
414 	0,
415 };
416 
417 static short nfsv3err_rename[] = {
418 	NFSERR_IO,
419 	NFSERR_NOENT,
420 	NFSERR_IO,
421 	NFSERR_ACCES,
422 	NFSERR_EXIST,
423 	NFSERR_XDEV,
424 	NFSERR_NOTDIR,
425 	NFSERR_ISDIR,
426 	NFSERR_INVAL,
427 	NFSERR_NOSPC,
428 	NFSERR_ROFS,
429 	NFSERR_MLINK,
430 	NFSERR_NAMETOL,
431 	NFSERR_NOTEMPTY,
432 	NFSERR_DQUOT,
433 	NFSERR_STALE,
434 	NFSERR_BADHANDLE,
435 	NFSERR_NOTSUPP,
436 	NFSERR_SERVERFAULT,
437 	0,
438 };
439 
440 static short nfsv3err_link[] = {
441 	NFSERR_IO,
442 	NFSERR_IO,
443 	NFSERR_ACCES,
444 	NFSERR_EXIST,
445 	NFSERR_XDEV,
446 	NFSERR_NOTDIR,
447 	NFSERR_INVAL,
448 	NFSERR_NOSPC,
449 	NFSERR_ROFS,
450 	NFSERR_MLINK,
451 	NFSERR_NAMETOL,
452 	NFSERR_DQUOT,
453 	NFSERR_STALE,
454 	NFSERR_BADHANDLE,
455 	NFSERR_NOTSUPP,
456 	NFSERR_SERVERFAULT,
457 	0,
458 };
459 
460 static short nfsv3err_readdir[] = {
461 	NFSERR_IO,
462 	NFSERR_IO,
463 	NFSERR_ACCES,
464 	NFSERR_NOTDIR,
465 	NFSERR_STALE,
466 	NFSERR_BADHANDLE,
467 	NFSERR_BAD_COOKIE,
468 	NFSERR_TOOSMALL,
469 	NFSERR_SERVERFAULT,
470 	0,
471 };
472 
473 static short nfsv3err_readdirplus[] = {
474 	NFSERR_IO,
475 	NFSERR_IO,
476 	NFSERR_ACCES,
477 	NFSERR_NOTDIR,
478 	NFSERR_STALE,
479 	NFSERR_BADHANDLE,
480 	NFSERR_BAD_COOKIE,
481 	NFSERR_NOTSUPP,
482 	NFSERR_TOOSMALL,
483 	NFSERR_SERVERFAULT,
484 	0,
485 };
486 
487 static short nfsv3err_fsstat[] = {
488 	NFSERR_IO,
489 	NFSERR_IO,
490 	NFSERR_STALE,
491 	NFSERR_BADHANDLE,
492 	NFSERR_SERVERFAULT,
493 	0,
494 };
495 
496 static short nfsv3err_fsinfo[] = {
497 	NFSERR_STALE,
498 	NFSERR_STALE,
499 	NFSERR_BADHANDLE,
500 	NFSERR_SERVERFAULT,
501 	0,
502 };
503 
504 static short nfsv3err_pathconf[] = {
505 	NFSERR_STALE,
506 	NFSERR_STALE,
507 	NFSERR_BADHANDLE,
508 	NFSERR_SERVERFAULT,
509 	0,
510 };
511 
512 static short nfsv3err_commit[] = {
513 	NFSERR_IO,
514 	NFSERR_IO,
515 	NFSERR_STALE,
516 	NFSERR_BADHANDLE,
517 	NFSERR_SERVERFAULT,
518 	0,
519 };
520 
521 static short *nfsrv_v3errmap[] = {
522 	nfsv3err_null,
523 	nfsv3err_getattr,
524 	nfsv3err_setattr,
525 	nfsv3err_lookup,
526 	nfsv3err_access,
527 	nfsv3err_readlink,
528 	nfsv3err_read,
529 	nfsv3err_write,
530 	nfsv3err_create,
531 	nfsv3err_mkdir,
532 	nfsv3err_symlink,
533 	nfsv3err_mknod,
534 	nfsv3err_remove,
535 	nfsv3err_rmdir,
536 	nfsv3err_rename,
537 	nfsv3err_link,
538 	nfsv3err_readdir,
539 	nfsv3err_readdirplus,
540 	nfsv3err_fsstat,
541 	nfsv3err_fsinfo,
542 	nfsv3err_pathconf,
543 	nfsv3err_commit,
544 };
545 
546 #endif /* NFS_NOSERVER */
547 
548 extern struct nfsrtt nfsrtt;
549 extern time_t nqnfsstarttime;
550 extern int nqsrv_clockskew;
551 extern int nqsrv_writeslack;
552 extern int nqsrv_maxlease;
553 extern struct nfsstats nfsstats;
554 extern int nqnfs_piggy[NFS_NPROCS];
555 extern nfstype nfsv2_type[9];
556 extern nfstype nfsv3_type[9];
557 extern struct nfsnodehashhead *nfsnodehashtbl;
558 extern u_long nfsnodehash;
559 
560 struct nfssvc_args;
561 extern int nfssvc(struct proc *, struct nfssvc_args *, int *);
562 
563 LIST_HEAD(nfsnodehashhead, nfsnode);
564 
565 u_quad_t
566 nfs_curusec(void)
567 {
568 	struct timeval tv;
569 
570 	getmicrotime(&tv);
571 	return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
572 }
573 
574 /*
575  * Create the header for an rpc request packet
576  * The hsiz is the size of the rest of the nfs request header.
577  * (just used to decide if a cluster is a good idea)
578  */
579 struct mbuf *
580 nfsm_reqh(struct vnode *vp, u_long procid, int hsiz, caddr_t *bposp)
581 {
582 	struct mbuf *mb;
583 	u_int32_t *tl;
584 	caddr_t bpos;
585 	struct mbuf *mb2;
586 	struct nfsmount *nmp;
587 	int nqflag;
588 
589 	MGET(mb, MB_WAIT, MT_DATA);
590 	if (hsiz >= MINCLSIZE)
591 		MCLGET(mb, MB_WAIT);
592 	mb->m_len = 0;
593 	bpos = mtod(mb, caddr_t);
594 
595 	/*
596 	 * For NQNFS, add lease request.
597 	 */
598 	if (vp) {
599 		nmp = VFSTONFS(vp->v_mount);
600 		if (nmp->nm_flag & NFSMNT_NQNFS) {
601 			nqflag = NQNFS_NEEDLEASE(vp, procid);
602 			if (nqflag) {
603 				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
604 				*tl++ = txdr_unsigned(nqflag);
605 				*tl = txdr_unsigned(nmp->nm_leaseterm);
606 			} else {
607 				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
608 				*tl = 0;
609 			}
610 		}
611 	}
612 	/* Finally, return values */
613 	*bposp = bpos;
614 	return (mb);
615 }
616 
617 /*
618  * Build the RPC header and fill in the authorization info.
619  * The authorization string argument is only used when the credentials
620  * come from outside of the kernel.
621  * Returns the head of the mbuf list.
622  */
623 struct mbuf *
624 nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type,
625 	     int auth_len, char *auth_str, int verf_len, char *verf_str,
626 	     struct mbuf *mrest, int mrest_len, struct mbuf **mbp,
627 	     u_int32_t *xidp)
628 {
629 	struct mbuf *mb;
630 	u_int32_t *tl;
631 	caddr_t bpos;
632 	int i;
633 	struct mbuf *mreq, *mb2;
634 	int siz, grpsiz, authsiz;
635 
636 	authsiz = nfsm_rndup(auth_len);
637 	MGETHDR(mb, MB_WAIT, MT_DATA);
638 	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
639 		MCLGET(mb, MB_WAIT);
640 	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
641 		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
642 	} else {
643 		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
644 	}
645 	mb->m_len = 0;
646 	mreq = mb;
647 	bpos = mtod(mb, caddr_t);
648 
649 	/*
650 	 * First the RPC header.
651 	 */
652 	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
653 
654 	/* Get a pretty random xid to start with */
655 	if (!nfs_xid)
656 		nfs_xid = random();
657 	/*
658 	 * Skip zero xid if it should ever happen.
659 	 */
660 	if (++nfs_xid == 0)
661 		nfs_xid++;
662 
663 	*tl++ = *xidp = txdr_unsigned(nfs_xid);
664 	*tl++ = rpc_call;
665 	*tl++ = rpc_vers;
666 	if (nmflag & NFSMNT_NQNFS) {
667 		*tl++ = txdr_unsigned(NQNFS_PROG);
668 		*tl++ = txdr_unsigned(NQNFS_VER3);
669 	} else {
670 		*tl++ = txdr_unsigned(NFS_PROG);
671 		if (nmflag & NFSMNT_NFSV3)
672 			*tl++ = txdr_unsigned(NFS_VER3);
673 		else
674 			*tl++ = txdr_unsigned(NFS_VER2);
675 	}
676 	if (nmflag & NFSMNT_NFSV3)
677 		*tl++ = txdr_unsigned(procid);
678 	else
679 		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
680 
681 	/*
682 	 * And then the authorization cred.
683 	 */
684 	*tl++ = txdr_unsigned(auth_type);
685 	*tl = txdr_unsigned(authsiz);
686 	switch (auth_type) {
687 	case RPCAUTH_UNIX:
688 		nfsm_build(tl, u_int32_t *, auth_len);
689 		*tl++ = 0;		/* stamp ?? */
690 		*tl++ = 0;		/* NULL hostname */
691 		*tl++ = txdr_unsigned(cr->cr_uid);
692 		*tl++ = txdr_unsigned(cr->cr_groups[0]);
693 		grpsiz = (auth_len >> 2) - 5;
694 		*tl++ = txdr_unsigned(grpsiz);
695 		for (i = 1; i <= grpsiz; i++)
696 			*tl++ = txdr_unsigned(cr->cr_groups[i]);
697 		break;
698 	case RPCAUTH_KERB4:
699 		siz = auth_len;
700 		while (siz > 0) {
701 			if (M_TRAILINGSPACE(mb) == 0) {
702 				MGET(mb2, MB_WAIT, MT_DATA);
703 				if (siz >= MINCLSIZE)
704 					MCLGET(mb2, MB_WAIT);
705 				mb->m_next = mb2;
706 				mb = mb2;
707 				mb->m_len = 0;
708 				bpos = mtod(mb, caddr_t);
709 			}
710 			i = min(siz, M_TRAILINGSPACE(mb));
711 			bcopy(auth_str, bpos, i);
712 			mb->m_len += i;
713 			auth_str += i;
714 			bpos += i;
715 			siz -= i;
716 		}
717 		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
718 			for (i = 0; i < siz; i++)
719 				*bpos++ = '\0';
720 			mb->m_len += siz;
721 		}
722 		break;
723 	};
724 
725 	/*
726 	 * And the verifier...
727 	 */
728 	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
729 	if (verf_str) {
730 		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
731 		*tl = txdr_unsigned(verf_len);
732 		siz = verf_len;
733 		while (siz > 0) {
734 			if (M_TRAILINGSPACE(mb) == 0) {
735 				MGET(mb2, MB_WAIT, MT_DATA);
736 				if (siz >= MINCLSIZE)
737 					MCLGET(mb2, MB_WAIT);
738 				mb->m_next = mb2;
739 				mb = mb2;
740 				mb->m_len = 0;
741 				bpos = mtod(mb, caddr_t);
742 			}
743 			i = min(siz, M_TRAILINGSPACE(mb));
744 			bcopy(verf_str, bpos, i);
745 			mb->m_len += i;
746 			verf_str += i;
747 			bpos += i;
748 			siz -= i;
749 		}
750 		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
751 			for (i = 0; i < siz; i++)
752 				*bpos++ = '\0';
753 			mb->m_len += siz;
754 		}
755 	} else {
756 		*tl++ = txdr_unsigned(RPCAUTH_NULL);
757 		*tl = 0;
758 	}
759 	mb->m_next = mrest;
760 	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
761 	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
762 	*mbp = mb;
763 	return (mreq);
764 }
765 
766 /*
767  * copies mbuf chain to the uio scatter/gather list
768  */
769 int
770 nfsm_mbuftouio(struct mbuf **mrep, struct uio *uiop, int siz, caddr_t *dpos)
771 {
772 	char *mbufcp, *uiocp;
773 	int xfer, left, len;
774 	struct mbuf *mp;
775 	long uiosiz, rem;
776 	int error = 0;
777 
778 	mp = *mrep;
779 	mbufcp = *dpos;
780 	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
781 	rem = nfsm_rndup(siz)-siz;
782 	while (siz > 0) {
783 		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
784 			return (EFBIG);
785 		left = uiop->uio_iov->iov_len;
786 		uiocp = uiop->uio_iov->iov_base;
787 		if (left > siz)
788 			left = siz;
789 		uiosiz = left;
790 		while (left > 0) {
791 			while (len == 0) {
792 				mp = mp->m_next;
793 				if (mp == NULL)
794 					return (EBADRPC);
795 				mbufcp = mtod(mp, caddr_t);
796 				len = mp->m_len;
797 			}
798 			xfer = (left > len) ? len : left;
799 #ifdef notdef
800 			/* Not Yet.. */
801 			if (uiop->uio_iov->iov_op != NULL)
802 				(*(uiop->uio_iov->iov_op))
803 				(mbufcp, uiocp, xfer);
804 			else
805 #endif
806 			if (uiop->uio_segflg == UIO_SYSSPACE)
807 				bcopy(mbufcp, uiocp, xfer);
808 			else
809 				copyout(mbufcp, uiocp, xfer);
810 			left -= xfer;
811 			len -= xfer;
812 			mbufcp += xfer;
813 			uiocp += xfer;
814 			uiop->uio_offset += xfer;
815 			uiop->uio_resid -= xfer;
816 		}
817 		if (uiop->uio_iov->iov_len <= siz) {
818 			uiop->uio_iovcnt--;
819 			uiop->uio_iov++;
820 		} else {
821 			uiop->uio_iov->iov_base += uiosiz;
822 			uiop->uio_iov->iov_len -= uiosiz;
823 		}
824 		siz -= uiosiz;
825 	}
826 	*dpos = mbufcp;
827 	*mrep = mp;
828 	if (rem > 0) {
829 		if (len < rem)
830 			error = nfs_adv(mrep, dpos, rem, len);
831 		else
832 			*dpos += rem;
833 	}
834 	return (error);
835 }
836 
837 /*
838  * copies a uio scatter/gather list to an mbuf chain.
839  * NOTE: can ony handle iovcnt == 1
840  */
841 int
842 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
843 {
844 	char *uiocp;
845 	struct mbuf *mp, *mp2;
846 	int xfer, left, mlen;
847 	int uiosiz, clflg, rem;
848 	char *cp;
849 
850 #ifdef DIAGNOSTIC
851 	if (uiop->uio_iovcnt != 1)
852 		panic("nfsm_uiotombuf: iovcnt != 1");
853 #endif
854 
855 	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
856 		clflg = 1;
857 	else
858 		clflg = 0;
859 	rem = nfsm_rndup(siz)-siz;
860 	mp = mp2 = *mq;
861 	while (siz > 0) {
862 		left = uiop->uio_iov->iov_len;
863 		uiocp = uiop->uio_iov->iov_base;
864 		if (left > siz)
865 			left = siz;
866 		uiosiz = left;
867 		while (left > 0) {
868 			mlen = M_TRAILINGSPACE(mp);
869 			if (mlen == 0) {
870 				MGET(mp, MB_WAIT, MT_DATA);
871 				if (clflg)
872 					MCLGET(mp, MB_WAIT);
873 				mp->m_len = 0;
874 				mp2->m_next = mp;
875 				mp2 = mp;
876 				mlen = M_TRAILINGSPACE(mp);
877 			}
878 			xfer = (left > mlen) ? mlen : left;
879 #ifdef notdef
880 			/* Not Yet.. */
881 			if (uiop->uio_iov->iov_op != NULL)
882 				(*(uiop->uio_iov->iov_op))
883 				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
884 			else
885 #endif
886 			if (uiop->uio_segflg == UIO_SYSSPACE)
887 				bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
888 			else
889 				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
890 			mp->m_len += xfer;
891 			left -= xfer;
892 			uiocp += xfer;
893 			uiop->uio_offset += xfer;
894 			uiop->uio_resid -= xfer;
895 		}
896 		uiop->uio_iov->iov_base += uiosiz;
897 		uiop->uio_iov->iov_len -= uiosiz;
898 		siz -= uiosiz;
899 	}
900 	if (rem > 0) {
901 		if (rem > M_TRAILINGSPACE(mp)) {
902 			MGET(mp, MB_WAIT, MT_DATA);
903 			mp->m_len = 0;
904 			mp2->m_next = mp;
905 		}
906 		cp = mtod(mp, caddr_t)+mp->m_len;
907 		for (left = 0; left < rem; left++)
908 			*cp++ = '\0';
909 		mp->m_len += rem;
910 		*bpos = cp;
911 	} else
912 		*bpos = mtod(mp, caddr_t)+mp->m_len;
913 	*mq = mp;
914 	return (0);
915 }
916 
917 /*
918  * Help break down an mbuf chain by setting the first siz bytes contiguous
919  * pointed to by returned val.
920  * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
921  * cases. (The macros use the vars. dpos and dpos2)
922  */
923 int
924 nfsm_disct(struct mbuf **mdp, caddr_t *dposp, int siz, int left, caddr_t *cp2)
925 {
926 	struct mbuf *mp, *mp2;
927 	int siz2, xfer;
928 	caddr_t p;
929 
930 	mp = *mdp;
931 	while (left == 0) {
932 		*mdp = mp = mp->m_next;
933 		if (mp == NULL)
934 			return (EBADRPC);
935 		left = mp->m_len;
936 		*dposp = mtod(mp, caddr_t);
937 	}
938 	if (left >= siz) {
939 		*cp2 = *dposp;
940 		*dposp += siz;
941 	} else if (mp->m_next == NULL) {
942 		return (EBADRPC);
943 	} else if (siz > MHLEN) {
944 		panic("nfs S too big");
945 	} else {
946 		MGET(mp2, MB_WAIT, MT_DATA);
947 		mp2->m_next = mp->m_next;
948 		mp->m_next = mp2;
949 		mp->m_len -= left;
950 		mp = mp2;
951 		*cp2 = p = mtod(mp, caddr_t);
952 		bcopy(*dposp, p, left);		/* Copy what was left */
953 		siz2 = siz-left;
954 		p += left;
955 		mp2 = mp->m_next;
956 		/* Loop around copying up the siz2 bytes */
957 		while (siz2 > 0) {
958 			if (mp2 == NULL)
959 				return (EBADRPC);
960 			xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
961 			if (xfer > 0) {
962 				bcopy(mtod(mp2, caddr_t), p, xfer);
963 				NFSMADV(mp2, xfer);
964 				mp2->m_len -= xfer;
965 				p += xfer;
966 				siz2 -= xfer;
967 			}
968 			if (siz2 > 0)
969 				mp2 = mp2->m_next;
970 		}
971 		mp->m_len = siz;
972 		*mdp = mp2;
973 		*dposp = mtod(mp2, caddr_t);
974 	}
975 	return (0);
976 }
977 
978 /*
979  * Advance the position in the mbuf chain.
980  */
981 int
982 nfs_adv(struct mbuf **mdp, caddr_t *dposp, int offs, int left)
983 {
984 	struct mbuf *m;
985 	int s;
986 
987 	m = *mdp;
988 	s = left;
989 	while (s < offs) {
990 		offs -= s;
991 		m = m->m_next;
992 		if (m == NULL)
993 			return (EBADRPC);
994 		s = m->m_len;
995 	}
996 	*mdp = m;
997 	*dposp = mtod(m, caddr_t)+offs;
998 	return (0);
999 }
1000 
1001 /*
1002  * Copy a string into mbufs for the hard cases...
1003  */
1004 int
1005 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
1006 {
1007 	struct mbuf *m1 = NULL, *m2;
1008 	long left, xfer, len, tlen;
1009 	u_int32_t *tl;
1010 	int putsize;
1011 
1012 	putsize = 1;
1013 	m2 = *mb;
1014 	left = M_TRAILINGSPACE(m2);
1015 	if (left > 0) {
1016 		tl = ((u_int32_t *)(*bpos));
1017 		*tl++ = txdr_unsigned(siz);
1018 		putsize = 0;
1019 		left -= NFSX_UNSIGNED;
1020 		m2->m_len += NFSX_UNSIGNED;
1021 		if (left > 0) {
1022 			bcopy(cp, (caddr_t) tl, left);
1023 			siz -= left;
1024 			cp += left;
1025 			m2->m_len += left;
1026 			left = 0;
1027 		}
1028 	}
1029 	/* Loop around adding mbufs */
1030 	while (siz > 0) {
1031 		MGET(m1, MB_WAIT, MT_DATA);
1032 		if (siz > MLEN)
1033 			MCLGET(m1, MB_WAIT);
1034 		m1->m_len = NFSMSIZ(m1);
1035 		m2->m_next = m1;
1036 		m2 = m1;
1037 		tl = mtod(m1, u_int32_t *);
1038 		tlen = 0;
1039 		if (putsize) {
1040 			*tl++ = txdr_unsigned(siz);
1041 			m1->m_len -= NFSX_UNSIGNED;
1042 			tlen = NFSX_UNSIGNED;
1043 			putsize = 0;
1044 		}
1045 		if (siz < m1->m_len) {
1046 			len = nfsm_rndup(siz);
1047 			xfer = siz;
1048 			if (xfer < len)
1049 				*(tl+(xfer>>2)) = 0;
1050 		} else {
1051 			xfer = len = m1->m_len;
1052 		}
1053 		bcopy(cp, (caddr_t) tl, xfer);
1054 		m1->m_len = len+tlen;
1055 		siz -= xfer;
1056 		cp += xfer;
1057 	}
1058 	*mb = m1;
1059 	*bpos = mtod(m1, caddr_t)+m1->m_len;
1060 	return (0);
1061 }
1062 
1063 /*
1064  * Called once to initialize data structures...
1065  */
1066 int
1067 nfs_init(struct vfsconf *vfsp)
1068 {
1069 	int i;
1070 
1071 	callout_init(&nfs_timer_handle);
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 	callout_stop(&nfs_timer_handle);
1146 	nfs_mount_type = -1;
1147 #ifndef NFS_NOSERVER
1148 	default_vnode_vops->vop_lease = nfs_prev_vop_lease_check;
1149 #endif
1150 	lease_updatetime = nfs_prev_lease_updatetime;
1151 	sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1152 	sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1153 	return (0);
1154 }
1155 
1156 /*
1157  * Attribute cache routines.
1158  * nfs_loadattrcache() - loads or updates the cache contents from attributes
1159  *	that are on the mbuf list
1160  * nfs_getattrcache() - returns valid attributes if found in cache, returns
1161  *	error otherwise
1162  */
1163 
1164 /*
1165  * Load the attribute cache (that lives in the nfsnode entry) with
1166  * the values on the mbuf list and
1167  * Iff vap not NULL
1168  *    copy the attributes to *vaper
1169  */
1170 int
1171 nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
1172 		  struct vattr *vaper, int dontshrink)
1173 {
1174 	struct vnode *vp = *vpp;
1175 	struct vattr *vap;
1176 	struct nfs_fattr *fp;
1177 	struct nfsnode *np;
1178 	int32_t t1;
1179 	caddr_t cp2;
1180 	int error = 0;
1181 	udev_t rdev;
1182 	struct mbuf *md;
1183 	enum vtype vtyp;
1184 	u_short vmode;
1185 	struct timespec mtime;
1186 	int v3 = NFS_ISV3(vp);
1187 
1188 	md = *mdp;
1189 	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1190 	if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1191 		return (error);
1192 	fp = (struct nfs_fattr *)cp2;
1193 	if (v3) {
1194 		vtyp = nfsv3tov_type(fp->fa_type);
1195 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1196 		rdev = makeudev(fxdr_unsigned(int, fp->fa3_rdev.specdata1),
1197 			fxdr_unsigned(int, fp->fa3_rdev.specdata2));
1198 		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1199 	} else {
1200 		vtyp = nfsv2tov_type(fp->fa_type);
1201 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
1202 		/*
1203 		 * XXX
1204 		 *
1205 		 * The duplicate information returned in fa_type and fa_mode
1206 		 * is an ambiguity in the NFS version 2 protocol.
1207 		 *
1208 		 * VREG should be taken literally as a regular file.  If a
1209 		 * server intents to return some type information differently
1210 		 * in the upper bits of the mode field (e.g. for sockets, or
1211 		 * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
1212 		 * leave the examination of the mode bits even in the VREG
1213 		 * case to avoid breakage for bogus servers, but we make sure
1214 		 * that there are actually type bits set in the upper part of
1215 		 * fa_mode (and failing that, trust the va_type field).
1216 		 *
1217 		 * NFSv3 cleared the issue, and requires fa_mode to not
1218 		 * contain any type information (while also introduing sockets
1219 		 * and FIFOs for fa_type).
1220 		 */
1221 		if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1222 			vtyp = IFTOVT(vmode);
1223 		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1224 		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1225 
1226 		/*
1227 		 * Really ugly NFSv2 kludge.
1228 		 */
1229 		if (vtyp == VCHR && rdev == (udev_t)0xffffffff)
1230 			vtyp = VFIFO;
1231 	}
1232 
1233 	/*
1234 	 * If v_type == VNON it is a new node, so fill in the v_type,
1235 	 * n_mtime fields. Check to see if it represents a special
1236 	 * device, and if so, check for a possible alias. Once the
1237 	 * correct vnode has been obtained, fill in the rest of the
1238 	 * information.
1239 	 */
1240 	np = VTONFS(vp);
1241 	if (vp->v_type != vtyp) {
1242 		vp->v_type = vtyp;
1243 		if (vp->v_type == VFIFO) {
1244 			vp->v_ops = &vp->v_mount->mnt_vn_fifo_ops;
1245 		} else if (vp->v_type == VCHR || vp->v_type == VBLK) {
1246 			vp->v_ops = &vp->v_mount->mnt_vn_spec_ops;
1247 			addaliasu(vp, rdev);
1248 		} else {
1249 			vp->v_ops = &vp->v_mount->mnt_vn_use_ops;
1250 		}
1251 		np->n_mtime = mtime.tv_sec;
1252 	} else if ((np->n_flag & NMODIFIED) == 0) {
1253 		/*
1254 		 * If we haven't modified the file locally update our notion
1255 		 * of the last-modified time based on the server's
1256 		 * information.
1257 		 */
1258 		np->n_mtime = mtime.tv_sec;
1259 	}
1260 	vap = &np->n_vattr;
1261 	vap->va_type = vtyp;
1262 	vap->va_mode = (vmode & 07777);
1263 	vap->va_rdev = rdev;
1264 	vap->va_mtime = mtime;
1265 	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1266 	if (v3) {
1267 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1268 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1269 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1270 		vap->va_size = fxdr_hyper(&fp->fa3_size);
1271 		vap->va_blocksize = NFS_FABLKSIZE;
1272 		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1273 		vap->va_fileid = fxdr_unsigned(int32_t,
1274 		    fp->fa3_fileid.nfsuquad[1]);
1275 		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1276 		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1277 		vap->va_flags = 0;
1278 		vap->va_filerev = 0;
1279 	} else {
1280 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1281 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1282 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1283 		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1284 		vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1285 		vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1286 		    * NFS_FABLKSIZE;
1287 		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1288 		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1289 		vap->va_flags = 0;
1290 		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1291 		    fp->fa2_ctime.nfsv2_sec);
1292 		vap->va_ctime.tv_nsec = 0;
1293 		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1294 		vap->va_filerev = 0;
1295 	}
1296 	np->n_attrstamp = time_second;
1297 	if (vap->va_size != np->n_size) {
1298 		if (vap->va_type == VREG) {
1299 			if (dontshrink && vap->va_size < np->n_size) {
1300 				/*
1301 				 * We've been told not to shrink the file;
1302 				 * zero np->n_attrstamp to indicate that
1303 				 * the attributes are stale.
1304 				 */
1305 				vap->va_size = np->n_size;
1306 				np->n_attrstamp = 0;
1307 			} else if (np->n_flag & NMODIFIED) {
1308 				/*
1309 				 * We've modified the file: Use the larger
1310 				 * of our size, and the server's size.
1311 				 */
1312 				if (vap->va_size < np->n_size) {
1313 					vap->va_size = np->n_size;
1314 				} else {
1315 					np->n_size = vap->va_size;
1316 					np->n_flag |= NSIZECHANGED;
1317 				}
1318 			} else {
1319 				np->n_size = vap->va_size;
1320 				np->n_flag |= NSIZECHANGED;
1321 			}
1322 			vnode_pager_setsize(vp, np->n_size);
1323 		} else {
1324 			np->n_size = vap->va_size;
1325 		}
1326 	}
1327 	if (vaper != NULL) {
1328 		bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1329 		if (np->n_flag & NCHG) {
1330 			if (np->n_flag & NACC)
1331 				vaper->va_atime = np->n_atim;
1332 			if (np->n_flag & NUPD)
1333 				vaper->va_mtime = np->n_mtim;
1334 		}
1335 	}
1336 	return (0);
1337 }
1338 
1339 #ifdef NFS_ACDEBUG
1340 #include <sys/sysctl.h>
1341 SYSCTL_DECL(_vfs_nfs);
1342 static int nfs_acdebug;
1343 SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1344 #endif
1345 
1346 /*
1347  * Check the time stamp
1348  * If the cache is valid, copy contents to *vap and return 0
1349  * otherwise return an error
1350  */
1351 int
1352 nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
1353 {
1354 	struct nfsnode *np;
1355 	struct vattr *vap;
1356 	struct nfsmount *nmp;
1357 	int timeo;
1358 
1359 	np = VTONFS(vp);
1360 	vap = &np->n_vattr;
1361 	nmp = VFSTONFS(vp->v_mount);
1362 
1363 	/*
1364 	 * Dynamic timeout based on how recently the file was modified.
1365 	 */
1366 	timeo = (get_approximate_time_t() - np->n_mtime) / 10;
1367 
1368 #ifdef NFS_ACDEBUG
1369 	if (nfs_acdebug>1)
1370 		printf("nfs_getattrcache: initial timeo = %d\n", timeo);
1371 #endif
1372 
1373 	if (vap->va_type == VDIR) {
1374 		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acdirmin)
1375 			timeo = nmp->nm_acdirmin;
1376 		else if (timeo > nmp->nm_acdirmax)
1377 			timeo = nmp->nm_acdirmax;
1378 	} else {
1379 		if ((np->n_flag & NMODIFIED) || timeo < nmp->nm_acregmin)
1380 			timeo = nmp->nm_acregmin;
1381 		else if (timeo > nmp->nm_acregmax)
1382 			timeo = nmp->nm_acregmax;
1383 	}
1384 
1385 #ifdef NFS_ACDEBUG
1386 	if (nfs_acdebug > 2)
1387 		printf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1388 			nmp->nm_acregmin, nmp->nm_acregmax,
1389 			nmp->nm_acdirmin, nmp->nm_acdirmax);
1390 
1391 	if (nfs_acdebug)
1392 		printf("nfs_getattrcache: age = %d; final timeo = %d\n",
1393 			(int)(time_second - np->n_attrstamp), timeo);
1394 #endif
1395 
1396 	if (np->n_attrstamp == 0 || (time_second - np->n_attrstamp) >= timeo) {
1397 		nfsstats.attrcache_misses++;
1398 		return (ENOENT);
1399 	}
1400 	nfsstats.attrcache_hits++;
1401 	if (vap->va_size != np->n_size) {
1402 		if (vap->va_type == VREG) {
1403 			if (np->n_flag & NMODIFIED) {
1404 				if (vap->va_size < np->n_size)
1405 					vap->va_size = np->n_size;
1406 				else
1407 					np->n_size = vap->va_size;
1408 			} else {
1409 				np->n_size = vap->va_size;
1410 			}
1411 			vnode_pager_setsize(vp, np->n_size);
1412 		} else {
1413 			np->n_size = vap->va_size;
1414 		}
1415 	}
1416 	bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1417 	if (np->n_flag & NCHG) {
1418 		if (np->n_flag & NACC)
1419 			vaper->va_atime = np->n_atim;
1420 		if (np->n_flag & NUPD)
1421 			vaper->va_mtime = np->n_mtim;
1422 	}
1423 	return (0);
1424 }
1425 
1426 #ifndef NFS_NOSERVER
1427 
1428 /*
1429  * Set up nameidata for a lookup() call and do it.
1430  *
1431  * If pubflag is set, this call is done for a lookup operation on the
1432  * public filehandle. In that case we allow crossing mountpoints and
1433  * absolute pathnames. However, the caller is expected to check that
1434  * the lookup result is within the public fs, and deny access if
1435  * it is not.
1436  *
1437  * dirp may be set whether an error is returned or not, and must be
1438  * released by the caller.
1439  *
1440  * On return nd->nl_ncp usually points to the target ncp, which may represent
1441  * a negative hit.
1442  *
1443  * NOTE: the caller must call nlookup_done(nd) unconditionally on return
1444  * to cleanup.
1445  */
1446 int
1447 nfs_namei(struct nlookupdata *nd, struct ucred *cred, int nameiop,
1448 	struct vnode **dvpp, struct vnode **vpp,
1449 	fhandle_t *fhp, int len,
1450 	struct nfssvc_sock *slp, struct sockaddr *nam, struct mbuf **mdp,
1451 	caddr_t *dposp, struct vnode **dirpp, struct thread *td,
1452 	int kerbflag, int pubflag)
1453 {
1454 	int i, rem;
1455 	int flags;
1456 	struct mbuf *md;
1457 	char *fromcp, *tocp, *cp;
1458 	char *namebuf;
1459 	struct namecache *ncp;
1460 	struct vnode *dp;
1461 	int error, rdonly;
1462 
1463 	namebuf = zalloc(namei_zone);
1464 	flags = 0;
1465 	*dirpp = NULL;
1466 
1467 	/*
1468 	 * Copy the name from the mbuf list to namebuf.
1469 	 */
1470 	fromcp = *dposp;
1471 	tocp = namebuf;
1472 	md = *mdp;
1473 	rem = mtod(md, caddr_t) + md->m_len - fromcp;
1474 	for (i = 0; i < len; i++) {
1475 		while (rem == 0) {
1476 			md = md->m_next;
1477 			if (md == NULL) {
1478 				error = EBADRPC;
1479 				goto out;
1480 			}
1481 			fromcp = mtod(md, caddr_t);
1482 			rem = md->m_len;
1483 		}
1484 		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1485 			error = EACCES;
1486 			goto out;
1487 		}
1488 		*tocp++ = *fromcp++;
1489 		rem--;
1490 	}
1491 	*tocp = '\0';
1492 	*mdp = md;
1493 	*dposp = fromcp;
1494 	len = nfsm_rndup(len)-len;
1495 	if (len > 0) {
1496 		if (rem >= len)
1497 			*dposp += len;
1498 		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1499 			goto out;
1500 	}
1501 
1502 	/*
1503 	 * Extract and set starting directory.  The returned dp is refd
1504 	 * but not locked.
1505 	 */
1506 	error = nfsrv_fhtovp(fhp, FALSE, &dp, cred, slp,
1507 				nam, &rdonly, kerbflag, pubflag);
1508 	if (error)
1509 		goto out;
1510 	if (dp->v_type != VDIR) {
1511 		vrele(dp);
1512 		error = ENOTDIR;
1513 		goto out;
1514 	}
1515 
1516 	/*
1517 	 * Set return directory.  Reference to dp is implicitly transfered
1518 	 * to the returned pointer.  This must be set before we potentially
1519 	 * goto out below.
1520 	 */
1521 	*dirpp = dp;
1522 
1523 	if (pubflag) {
1524 		/*
1525 		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1526 		 * and the 'native path' indicator.
1527 		 */
1528 		cp = zalloc(namei_zone);
1529 		fromcp = namebuf;
1530 		tocp = cp;
1531 		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1532 			switch ((unsigned char)*fromcp) {
1533 			case WEBNFS_NATIVE_CHAR:
1534 				/*
1535 				 * 'Native' path for us is the same
1536 				 * as a path according to the NFS spec,
1537 				 * just skip the escape char.
1538 				 */
1539 				fromcp++;
1540 				break;
1541 			/*
1542 			 * More may be added in the future, range 0x80-0xff
1543 			 */
1544 			default:
1545 				error = EIO;
1546 				zfree(namei_zone, cp);
1547 				goto out;
1548 			}
1549 		}
1550 		/*
1551 		 * Translate the '%' escapes, URL-style.
1552 		 */
1553 		while (*fromcp != '\0') {
1554 			if (*fromcp == WEBNFS_ESC_CHAR) {
1555 				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1556 					fromcp++;
1557 					*tocp++ = HEXSTRTOI(fromcp);
1558 					fromcp += 2;
1559 					continue;
1560 				} else {
1561 					error = ENOENT;
1562 					zfree(namei_zone, cp);
1563 					goto out;
1564 				}
1565 			} else
1566 				*tocp++ = *fromcp++;
1567 		}
1568 		*tocp = '\0';
1569 		zfree(namei_zone, namebuf);
1570 		namebuf = cp;
1571 	}
1572 
1573 	/*
1574 	 * Setup for search.  We need to get a start directory from dp.  Note
1575 	 * that dp is ref'd, but we no longer 'own' the ref (*dirpp owns it).
1576 	 */
1577 	if (pubflag == 0) {
1578 		flags |= NLC_NFS_NOSOFTLINKTRAV;
1579 		flags |= NLC_NOCROSSMOUNT;
1580 	}
1581 	if (rdonly)
1582 		flags |= NLC_NFS_RDONLY;
1583 	if (nameiop == NAMEI_CREATE || nameiop == NAMEI_RENAME)
1584 		flags |= NLC_CREATE;
1585 
1586 	/*
1587 	 * We need a starting ncp from the directory vnode dp.  dp must not
1588 	 * be locked.  The returned ncp will be refd but not locked.
1589 	 *
1590 	 * If no suitable ncp is found we instruct cache_fromdvp() to create
1591 	 * one.  If this fails the directory has probably been removed while
1592 	 * the target was chdir'd into it and any further lookup will fail.
1593 	 */
1594 	if ((ncp = cache_fromdvp(dp, cred, 1)) == NULL) {
1595 		error = EINVAL;
1596 		goto out;
1597 	}
1598 	nlookup_init_raw(nd, namebuf, UIO_SYSSPACE, flags, cred, ncp);
1599 	cache_drop(ncp);
1600 
1601 	/*
1602 	 * Ok, do the lookup.
1603 	 */
1604 	error = nlookup(nd);
1605 
1606 	/*
1607 	 * If no error occured return the requested dvpp and vpp.  If
1608 	 * NLC_CREATE was specified nd->nl_ncp may represent a negative
1609 	 * cache hit in which case we do not attempt to obtain the vp.
1610 	 */
1611 	if (error == 0) {
1612 		ncp = nd->nl_ncp;
1613 		if (dvpp) {
1614 			if (ncp->nc_parent &&
1615 			    ncp->nc_parent->nc_mount == ncp->nc_mount) {
1616 				error = cache_vget(ncp->nc_parent, nd->nl_cred,
1617 						LK_EXCLUSIVE, dvpp);
1618 			} else {
1619 				error = ENXIO;
1620 			}
1621 		}
1622 		if (vpp && ncp->nc_vp) {
1623 			error = cache_vget(ncp, nd->nl_cred, LK_EXCLUSIVE, vpp);
1624 		}
1625 		if (error) {
1626 			if (dvpp && *dvpp) {
1627 				vput(*dvpp);
1628 				*dvpp = NULL;
1629 			}
1630 			if (vpp && *vpp) {
1631 				vput(*vpp);
1632 				*vpp = NULL;
1633 			}
1634 		}
1635 	}
1636 
1637 	/*
1638 	 * Finish up.
1639 	 */
1640 out:
1641 	zfree(namei_zone, namebuf);
1642 	return (error);
1643 }
1644 
1645 /*
1646  * A fiddled version of m_adj() that ensures null fill to a long
1647  * boundary and only trims off the back end
1648  */
1649 void
1650 nfsm_adj(struct mbuf *mp, int len, int nul)
1651 {
1652 	struct mbuf *m;
1653 	int count, i;
1654 	char *cp;
1655 
1656 	/*
1657 	 * Trim from tail.  Scan the mbuf chain,
1658 	 * calculating its length and finding the last mbuf.
1659 	 * If the adjustment only affects this mbuf, then just
1660 	 * adjust and return.  Otherwise, rescan and truncate
1661 	 * after the remaining size.
1662 	 */
1663 	count = 0;
1664 	m = mp;
1665 	for (;;) {
1666 		count += m->m_len;
1667 		if (m->m_next == (struct mbuf *)0)
1668 			break;
1669 		m = m->m_next;
1670 	}
1671 	if (m->m_len > len) {
1672 		m->m_len -= len;
1673 		if (nul > 0) {
1674 			cp = mtod(m, caddr_t)+m->m_len-nul;
1675 			for (i = 0; i < nul; i++)
1676 				*cp++ = '\0';
1677 		}
1678 		return;
1679 	}
1680 	count -= len;
1681 	if (count < 0)
1682 		count = 0;
1683 	/*
1684 	 * Correct length for chain is "count".
1685 	 * Find the mbuf with last data, adjust its length,
1686 	 * and toss data from remaining mbufs on chain.
1687 	 */
1688 	for (m = mp; m; m = m->m_next) {
1689 		if (m->m_len >= count) {
1690 			m->m_len = count;
1691 			if (nul > 0) {
1692 				cp = mtod(m, caddr_t)+m->m_len-nul;
1693 				for (i = 0; i < nul; i++)
1694 					*cp++ = '\0';
1695 			}
1696 			break;
1697 		}
1698 		count -= m->m_len;
1699 	}
1700 	for (m = m->m_next;m;m = m->m_next)
1701 		m->m_len = 0;
1702 }
1703 
1704 /*
1705  * Make these functions instead of macros, so that the kernel text size
1706  * doesn't get too big...
1707  */
1708 void
1709 nfsm_srvwcc(struct nfsrv_descript *nfsd, int before_ret,
1710 	    struct vattr *before_vap, int after_ret, struct vattr *after_vap,
1711 	    struct mbuf **mbp, char **bposp)
1712 {
1713 	struct mbuf *mb = *mbp, *mb2;
1714 	char *bpos = *bposp;
1715 	u_int32_t *tl;
1716 
1717 	if (before_ret) {
1718 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1719 		*tl = nfs_false;
1720 	} else {
1721 		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1722 		*tl++ = nfs_true;
1723 		txdr_hyper(before_vap->va_size, tl);
1724 		tl += 2;
1725 		txdr_nfsv3time(&(before_vap->va_mtime), tl);
1726 		tl += 2;
1727 		txdr_nfsv3time(&(before_vap->va_ctime), tl);
1728 	}
1729 	*bposp = bpos;
1730 	*mbp = mb;
1731 	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1732 }
1733 
1734 void
1735 nfsm_srvpostopattr(struct nfsrv_descript *nfsd, int after_ret,
1736 		   struct vattr *after_vap, struct mbuf **mbp, char **bposp)
1737 {
1738 	struct mbuf *mb = *mbp, *mb2;
1739 	char *bpos = *bposp;
1740 	u_int32_t *tl;
1741 	struct nfs_fattr *fp;
1742 
1743 	if (after_ret) {
1744 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1745 		*tl = nfs_false;
1746 	} else {
1747 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1748 		*tl++ = nfs_true;
1749 		fp = (struct nfs_fattr *)tl;
1750 		nfsm_srvfattr(nfsd, after_vap, fp);
1751 	}
1752 	*mbp = mb;
1753 	*bposp = bpos;
1754 }
1755 
1756 void
1757 nfsm_srvfattr(struct nfsrv_descript *nfsd, struct vattr *vap,
1758 	      struct nfs_fattr *fp)
1759 {
1760 
1761 	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1762 	fp->fa_uid = txdr_unsigned(vap->va_uid);
1763 	fp->fa_gid = txdr_unsigned(vap->va_gid);
1764 	if (nfsd->nd_flag & ND_NFSV3) {
1765 		fp->fa_type = vtonfsv3_type(vap->va_type);
1766 		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1767 		txdr_hyper(vap->va_size, &fp->fa3_size);
1768 		txdr_hyper(vap->va_bytes, &fp->fa3_used);
1769 		fp->fa3_rdev.specdata1 = txdr_unsigned(umajor(vap->va_rdev));
1770 		fp->fa3_rdev.specdata2 = txdr_unsigned(uminor(vap->va_rdev));
1771 		fp->fa3_fsid.nfsuquad[0] = 0;
1772 		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1773 		fp->fa3_fileid.nfsuquad[0] = 0;
1774 		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
1775 		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1776 		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1777 		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1778 	} else {
1779 		fp->fa_type = vtonfsv2_type(vap->va_type);
1780 		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1781 		fp->fa2_size = txdr_unsigned(vap->va_size);
1782 		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1783 		if (vap->va_type == VFIFO)
1784 			fp->fa2_rdev = 0xffffffff;
1785 		else
1786 			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
1787 		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1788 		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1789 		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1790 		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1791 		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1792 		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1793 	}
1794 }
1795 
1796 /*
1797  * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1798  * 	- look up fsid in mount list (if not found ret error)
1799  *	- get vp and export rights by calling VFS_FHTOVP()
1800  *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1801  *	- if not lockflag unlock it with VOP_UNLOCK()
1802  */
1803 int
1804 nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
1805 	     struct ucred *cred, struct nfssvc_sock *slp, struct sockaddr *nam,
1806 	     int *rdonlyp, int kerbflag, int pubflag)
1807 {
1808 	struct thread *td = curthread; /* XXX */
1809 	struct mount *mp;
1810 	int i;
1811 	struct ucred *credanon;
1812 	int error, exflags;
1813 #ifdef MNT_EXNORESPORT		/* XXX needs mountd and /etc/exports help yet */
1814 	struct sockaddr_int *saddr;
1815 #endif
1816 
1817 	*vpp = (struct vnode *)0;
1818 
1819 	if (nfs_ispublicfh(fhp)) {
1820 		if (!pubflag || !nfs_pub.np_valid)
1821 			return (ESTALE);
1822 		fhp = &nfs_pub.np_handle;
1823 	}
1824 
1825 	mp = vfs_getvfs(&fhp->fh_fsid);
1826 	if (!mp)
1827 		return (ESTALE);
1828 	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1829 	if (error)
1830 		return (error);
1831 	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1832 	if (error)
1833 		return (error);
1834 #ifdef MNT_EXNORESPORT
1835 	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1836 		saddr = (struct sockaddr_in *)nam;
1837 		if (saddr->sin_family == AF_INET &&
1838 		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1839 			vput(*vpp);
1840 			*vpp = NULL;
1841 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1842 		}
1843 	}
1844 #endif
1845 	/*
1846 	 * Check/setup credentials.
1847 	 */
1848 	if (exflags & MNT_EXKERB) {
1849 		if (!kerbflag) {
1850 			vput(*vpp);
1851 			*vpp = NULL;
1852 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1853 		}
1854 	} else if (kerbflag) {
1855 		vput(*vpp);
1856 		*vpp = NULL;
1857 		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1858 	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1859 		cred->cr_uid = credanon->cr_uid;
1860 		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1861 			cred->cr_groups[i] = credanon->cr_groups[i];
1862 		cred->cr_ngroups = i;
1863 	}
1864 	if (exflags & MNT_EXRDONLY)
1865 		*rdonlyp = 1;
1866 	else
1867 		*rdonlyp = 0;
1868 
1869 	nfsrv_object_create(*vpp);
1870 
1871 	if (!lockflag)
1872 		VOP_UNLOCK(*vpp, 0, td);
1873 	return (0);
1874 }
1875 
1876 
1877 /*
1878  * WebNFS: check if a filehandle is a public filehandle. For v3, this
1879  * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
1880  * transformed this to all zeroes in both cases, so check for it.
1881  */
1882 int
1883 nfs_ispublicfh(fhandle_t *fhp)
1884 {
1885 	char *cp = (char *)fhp;
1886 	int i;
1887 
1888 	for (i = 0; i < NFSX_V3FH; i++)
1889 		if (*cp++ != 0)
1890 			return (FALSE);
1891 	return (TRUE);
1892 }
1893 
1894 #endif /* NFS_NOSERVER */
1895 /*
1896  * This function compares two net addresses by family and returns TRUE
1897  * if they are the same host.
1898  * If there is any doubt, return FALSE.
1899  * The AF_INET family is handled as a special case so that address mbufs
1900  * don't need to be saved to store "struct in_addr", which is only 4 bytes.
1901  */
1902 int
1903 netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
1904 {
1905 	struct sockaddr_in *inetaddr;
1906 
1907 	switch (family) {
1908 	case AF_INET:
1909 		inetaddr = (struct sockaddr_in *)nam;
1910 		if (inetaddr->sin_family == AF_INET &&
1911 		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
1912 			return (1);
1913 		break;
1914 	default:
1915 		break;
1916 	};
1917 	return (0);
1918 }
1919 
1920 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
1921 /*
1922  * This function finds the directory cookie that corresponds to the
1923  * logical byte offset given.
1924  */
1925 nfsuint64 *
1926 nfs_getcookie(struct nfsnode *np, off_t off, int add)
1927 {
1928 	struct nfsdmap *dp, *dp2;
1929 	int pos;
1930 
1931 	pos = (uoff_t)off / NFS_DIRBLKSIZ;
1932 	if (pos == 0 || off < 0) {
1933 #ifdef DIAGNOSTIC
1934 		if (add)
1935 			panic("nfs getcookie add at <= 0");
1936 #endif
1937 		return (&nfs_nullcookie);
1938 	}
1939 	pos--;
1940 	dp = np->n_cookies.lh_first;
1941 	if (!dp) {
1942 		if (add) {
1943 			MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
1944 				M_NFSDIROFF, M_WAITOK);
1945 			dp->ndm_eocookie = 0;
1946 			LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
1947 		} else
1948 			return ((nfsuint64 *)0);
1949 	}
1950 	while (pos >= NFSNUMCOOKIES) {
1951 		pos -= NFSNUMCOOKIES;
1952 		if (dp->ndm_list.le_next) {
1953 			if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
1954 				pos >= dp->ndm_eocookie)
1955 				return ((nfsuint64 *)0);
1956 			dp = dp->ndm_list.le_next;
1957 		} else if (add) {
1958 			MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
1959 				M_NFSDIROFF, M_WAITOK);
1960 			dp2->ndm_eocookie = 0;
1961 			LIST_INSERT_AFTER(dp, dp2, ndm_list);
1962 			dp = dp2;
1963 		} else
1964 			return ((nfsuint64 *)0);
1965 	}
1966 	if (pos >= dp->ndm_eocookie) {
1967 		if (add)
1968 			dp->ndm_eocookie = pos + 1;
1969 		else
1970 			return ((nfsuint64 *)0);
1971 	}
1972 	return (&dp->ndm_cookies[pos]);
1973 }
1974 
1975 /*
1976  * Invalidate cached directory information, except for the actual directory
1977  * blocks (which are invalidated separately).
1978  * Done mainly to avoid the use of stale offset cookies.
1979  */
1980 void
1981 nfs_invaldir(struct vnode *vp)
1982 {
1983 	struct nfsnode *np = VTONFS(vp);
1984 
1985 #ifdef DIAGNOSTIC
1986 	if (vp->v_type != VDIR)
1987 		panic("nfs: invaldir not dir");
1988 #endif
1989 	np->n_direofoffset = 0;
1990 	np->n_cookieverf.nfsuquad[0] = 0;
1991 	np->n_cookieverf.nfsuquad[1] = 0;
1992 	if (np->n_cookies.lh_first)
1993 		np->n_cookies.lh_first->ndm_eocookie = 0;
1994 }
1995 
1996 /*
1997  * The write verifier has changed (probably due to a server reboot), so all
1998  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
1999  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2000  * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
2001  * mount point.
2002  *
2003  * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
2004  * writes are not clusterable.
2005  */
2006 void
2007 nfs_clearcommit(struct mount *mp)
2008 {
2009 	struct vnode *vp, *nvp;
2010 	struct buf *bp, *nbp;
2011 	lwkt_tokref ilock;
2012 	int s;
2013 
2014 	lwkt_gettoken(&ilock, &mntvnode_token);
2015 	s = splbio();
2016 	for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp; vp = nvp) {
2017 		nvp = TAILQ_NEXT(vp, v_nmntvnodes);	/* ZZZ */
2018 		if (vp->v_flag & VPLACEMARKER)
2019 			continue;
2020 		for (bp = TAILQ_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2021 			nbp = TAILQ_NEXT(bp, b_vnbufs);
2022 			if (BUF_REFCNT(bp) == 0 &&
2023 			    (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2024 			     == (B_DELWRI | B_NEEDCOMMIT)) {
2025 				bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
2026 			}
2027 		}
2028 	}
2029 	splx(s);
2030 	lwkt_reltoken(&ilock);
2031 }
2032 
2033 #ifndef NFS_NOSERVER
2034 /*
2035  * Map errnos to NFS error numbers. For Version 3 also filter out error
2036  * numbers not specified for the associated procedure.
2037  */
2038 int
2039 nfsrv_errmap(struct nfsrv_descript *nd, int err)
2040 {
2041 	short *defaulterrp, *errp;
2042 
2043 	if (nd->nd_flag & ND_NFSV3) {
2044 	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
2045 		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2046 		while (*++errp) {
2047 			if (*errp == err)
2048 				return (err);
2049 			else if (*errp > err)
2050 				break;
2051 		}
2052 		return ((int)*defaulterrp);
2053 	    } else
2054 		return (err & 0xffff);
2055 	}
2056 	if (err <= ELAST)
2057 		return ((int)nfsrv_v2errmap[err - 1]);
2058 	return (NFSERR_IO);
2059 }
2060 
2061 int
2062 nfsrv_object_create(struct vnode *vp)
2063 {
2064 	struct thread *td = curthread;
2065 
2066 	if (vp == NULL || vp->v_type != VREG)
2067 		return (1);
2068 	return (vfs_object_create(vp, td));
2069 }
2070 
2071 /*
2072  * Sort the group list in increasing numerical order.
2073  * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2074  *  that used to be here.)
2075  */
2076 void
2077 nfsrvw_sort(gid_t *list, int num)
2078 {
2079 	int i, j;
2080 	gid_t v;
2081 
2082 	/* Insertion sort. */
2083 	for (i = 1; i < num; i++) {
2084 		v = list[i];
2085 		/* find correct slot for value v, moving others up */
2086 		for (j = i; --j >= 0 && v < list[j];)
2087 			list[j + 1] = list[j];
2088 		list[j + 1] = v;
2089 	}
2090 }
2091 
2092 /*
2093  * copy credentials making sure that the result can be compared with bcmp().
2094  */
2095 void
2096 nfsrv_setcred(struct ucred *incred, struct ucred *outcred)
2097 {
2098 	int i;
2099 
2100 	bzero((caddr_t)outcred, sizeof (struct ucred));
2101 	outcred->cr_ref = 1;
2102 	outcred->cr_uid = incred->cr_uid;
2103 	outcred->cr_ngroups = incred->cr_ngroups;
2104 	for (i = 0; i < incred->cr_ngroups; i++)
2105 		outcred->cr_groups[i] = incred->cr_groups[i];
2106 	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2107 }
2108 #endif /* NFS_NOSERVER */
2109