nfs_subs.c revision 1.64 1 /* $NetBSD: nfs_subs.c,v 1.64 1999/02/26 23:44:47 wrstuden Exp $ */
2
3 /*
4 * Copyright (c) 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * This code is derived from software contributed to Berkeley by
8 * Rick Macklem at The University of Guelph.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
39 */
40
41 #include "fs_nfs.h"
42 #include "opt_nfsserver.h"
43 #include "opt_iso.h"
44 #include "opt_uvm.h"
45
46 /*
47 * These functions support the macros and help fiddle mbuf chains for
48 * the nfs op functions. They do things like create the rpc header and
49 * copy data between mbuf chains and uio lists.
50 */
51 #include <sys/param.h>
52 #include <sys/proc.h>
53 #include <sys/systm.h>
54 #include <sys/kernel.h>
55 #include <sys/mount.h>
56 #include <sys/vnode.h>
57 #include <sys/namei.h>
58 #include <sys/mbuf.h>
59 #include <sys/socket.h>
60 #include <sys/stat.h>
61 #include <sys/malloc.h>
62 #include <sys/time.h>
63 #include <sys/dirent.h>
64
65 #include <vm/vm.h>
66
67 #if defined(UVM)
68 #include <uvm/uvm_extern.h>
69 #endif
70
71 #include <nfs/rpcv2.h>
72 #include <nfs/nfsproto.h>
73 #include <nfs/nfsnode.h>
74 #include <nfs/nfs.h>
75 #include <nfs/xdr_subs.h>
76 #include <nfs/nfsm_subs.h>
77 #include <nfs/nfsmount.h>
78 #include <nfs/nqnfs.h>
79 #include <nfs/nfsrtt.h>
80 #include <nfs/nfs_var.h>
81
82 #include <miscfs/specfs/specdev.h>
83
84 #include <vm/vm.h>
85
86 #include <netinet/in.h>
87 #ifdef ISO
88 #include <netiso/iso.h>
89 #endif
90
91 /*
92 * Data items converted to xdr at startup, since they are constant
93 * This is kinda hokey, but may save a little time doing byte swaps
94 */
95 u_int32_t nfs_xdrneg1;
96 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
97 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
98 rpc_auth_kerb;
99 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
100
101 /* And other global data */
102 static u_int32_t nfs_xid = 0;
103 nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON,
104 NFCHR, NFNON };
105 nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK,
106 NFFIFO, NFNON };
107 enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON };
108 enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO };
109 int nfs_ticks;
110 extern struct nfs_public nfs_pub;
111
112 /* NFS client/server stats. */
113 struct nfsstats nfsstats;
114
115 /*
116 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
117 */
118 int nfsv3_procid[NFS_NPROCS] = {
119 NFSPROC_NULL,
120 NFSPROC_GETATTR,
121 NFSPROC_SETATTR,
122 NFSPROC_NOOP,
123 NFSPROC_LOOKUP,
124 NFSPROC_READLINK,
125 NFSPROC_READ,
126 NFSPROC_NOOP,
127 NFSPROC_WRITE,
128 NFSPROC_CREATE,
129 NFSPROC_REMOVE,
130 NFSPROC_RENAME,
131 NFSPROC_LINK,
132 NFSPROC_SYMLINK,
133 NFSPROC_MKDIR,
134 NFSPROC_RMDIR,
135 NFSPROC_READDIR,
136 NFSPROC_FSSTAT,
137 NFSPROC_NOOP,
138 NFSPROC_NOOP,
139 NFSPROC_NOOP,
140 NFSPROC_NOOP,
141 NFSPROC_NOOP,
142 NFSPROC_NOOP,
143 NFSPROC_NOOP,
144 NFSPROC_NOOP
145 };
146
147 /*
148 * and the reverse mapping from generic to Version 2 procedure numbers
149 */
150 int nfsv2_procid[NFS_NPROCS] = {
151 NFSV2PROC_NULL,
152 NFSV2PROC_GETATTR,
153 NFSV2PROC_SETATTR,
154 NFSV2PROC_LOOKUP,
155 NFSV2PROC_NOOP,
156 NFSV2PROC_READLINK,
157 NFSV2PROC_READ,
158 NFSV2PROC_WRITE,
159 NFSV2PROC_CREATE,
160 NFSV2PROC_MKDIR,
161 NFSV2PROC_SYMLINK,
162 NFSV2PROC_CREATE,
163 NFSV2PROC_REMOVE,
164 NFSV2PROC_RMDIR,
165 NFSV2PROC_RENAME,
166 NFSV2PROC_LINK,
167 NFSV2PROC_READDIR,
168 NFSV2PROC_NOOP,
169 NFSV2PROC_STATFS,
170 NFSV2PROC_NOOP,
171 NFSV2PROC_NOOP,
172 NFSV2PROC_NOOP,
173 NFSV2PROC_NOOP,
174 NFSV2PROC_NOOP,
175 NFSV2PROC_NOOP,
176 NFSV2PROC_NOOP,
177 };
178
179 /*
180 * Maps errno values to nfs error numbers.
181 * Use NFSERR_IO as the catch all for ones not specifically defined in
182 * RFC 1094.
183 */
184 static u_char nfsrv_v2errmap[ELAST] = {
185 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO,
186 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
187 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO,
188 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR,
189 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
190 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS,
191 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
192 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
193 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
194 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
195 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
196 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
197 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO,
198 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE,
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,
202 };
203
204 /*
205 * Maps errno values to nfs error numbers.
206 * Although it is not obvious whether or not NFS clients really care if
207 * a returned error value is in the specified list for the procedure, the
208 * safest thing to do is filter them appropriately. For Version 2, the
209 * X/Open XNFS document is the only specification that defines error values
210 * for each RPC (The RFC simply lists all possible error values for all RPCs),
211 * so I have decided to not do this for Version 2.
212 * The first entry is the default error return and the rest are the valid
213 * errors for that RPC in increasing numeric order.
214 */
215 static short nfsv3err_null[] = {
216 0,
217 0,
218 };
219
220 static short nfsv3err_getattr[] = {
221 NFSERR_IO,
222 NFSERR_IO,
223 NFSERR_STALE,
224 NFSERR_BADHANDLE,
225 NFSERR_SERVERFAULT,
226 0,
227 };
228
229 static short nfsv3err_setattr[] = {
230 NFSERR_IO,
231 NFSERR_PERM,
232 NFSERR_IO,
233 NFSERR_ACCES,
234 NFSERR_INVAL,
235 NFSERR_NOSPC,
236 NFSERR_ROFS,
237 NFSERR_DQUOT,
238 NFSERR_STALE,
239 NFSERR_BADHANDLE,
240 NFSERR_NOT_SYNC,
241 NFSERR_SERVERFAULT,
242 0,
243 };
244
245 static short nfsv3err_lookup[] = {
246 NFSERR_IO,
247 NFSERR_NOENT,
248 NFSERR_IO,
249 NFSERR_ACCES,
250 NFSERR_NOTDIR,
251 NFSERR_NAMETOL,
252 NFSERR_STALE,
253 NFSERR_BADHANDLE,
254 NFSERR_SERVERFAULT,
255 0,
256 };
257
258 static short nfsv3err_access[] = {
259 NFSERR_IO,
260 NFSERR_IO,
261 NFSERR_STALE,
262 NFSERR_BADHANDLE,
263 NFSERR_SERVERFAULT,
264 0,
265 };
266
267 static short nfsv3err_readlink[] = {
268 NFSERR_IO,
269 NFSERR_IO,
270 NFSERR_ACCES,
271 NFSERR_INVAL,
272 NFSERR_STALE,
273 NFSERR_BADHANDLE,
274 NFSERR_NOTSUPP,
275 NFSERR_SERVERFAULT,
276 0,
277 };
278
279 static short nfsv3err_read[] = {
280 NFSERR_IO,
281 NFSERR_IO,
282 NFSERR_NXIO,
283 NFSERR_ACCES,
284 NFSERR_INVAL,
285 NFSERR_STALE,
286 NFSERR_BADHANDLE,
287 NFSERR_SERVERFAULT,
288 0,
289 };
290
291 static short nfsv3err_write[] = {
292 NFSERR_IO,
293 NFSERR_IO,
294 NFSERR_ACCES,
295 NFSERR_INVAL,
296 NFSERR_FBIG,
297 NFSERR_NOSPC,
298 NFSERR_ROFS,
299 NFSERR_DQUOT,
300 NFSERR_STALE,
301 NFSERR_BADHANDLE,
302 NFSERR_SERVERFAULT,
303 0,
304 };
305
306 static short nfsv3err_create[] = {
307 NFSERR_IO,
308 NFSERR_IO,
309 NFSERR_ACCES,
310 NFSERR_EXIST,
311 NFSERR_NOTDIR,
312 NFSERR_NOSPC,
313 NFSERR_ROFS,
314 NFSERR_NAMETOL,
315 NFSERR_DQUOT,
316 NFSERR_STALE,
317 NFSERR_BADHANDLE,
318 NFSERR_NOTSUPP,
319 NFSERR_SERVERFAULT,
320 0,
321 };
322
323 static short nfsv3err_mkdir[] = {
324 NFSERR_IO,
325 NFSERR_IO,
326 NFSERR_ACCES,
327 NFSERR_EXIST,
328 NFSERR_NOTDIR,
329 NFSERR_NOSPC,
330 NFSERR_ROFS,
331 NFSERR_NAMETOL,
332 NFSERR_DQUOT,
333 NFSERR_STALE,
334 NFSERR_BADHANDLE,
335 NFSERR_NOTSUPP,
336 NFSERR_SERVERFAULT,
337 0,
338 };
339
340 static short nfsv3err_symlink[] = {
341 NFSERR_IO,
342 NFSERR_IO,
343 NFSERR_ACCES,
344 NFSERR_EXIST,
345 NFSERR_NOTDIR,
346 NFSERR_NOSPC,
347 NFSERR_ROFS,
348 NFSERR_NAMETOL,
349 NFSERR_DQUOT,
350 NFSERR_STALE,
351 NFSERR_BADHANDLE,
352 NFSERR_NOTSUPP,
353 NFSERR_SERVERFAULT,
354 0,
355 };
356
357 static short nfsv3err_mknod[] = {
358 NFSERR_IO,
359 NFSERR_IO,
360 NFSERR_ACCES,
361 NFSERR_EXIST,
362 NFSERR_NOTDIR,
363 NFSERR_NOSPC,
364 NFSERR_ROFS,
365 NFSERR_NAMETOL,
366 NFSERR_DQUOT,
367 NFSERR_STALE,
368 NFSERR_BADHANDLE,
369 NFSERR_NOTSUPP,
370 NFSERR_SERVERFAULT,
371 NFSERR_BADTYPE,
372 0,
373 };
374
375 static short nfsv3err_remove[] = {
376 NFSERR_IO,
377 NFSERR_NOENT,
378 NFSERR_IO,
379 NFSERR_ACCES,
380 NFSERR_NOTDIR,
381 NFSERR_ROFS,
382 NFSERR_NAMETOL,
383 NFSERR_STALE,
384 NFSERR_BADHANDLE,
385 NFSERR_SERVERFAULT,
386 0,
387 };
388
389 static short nfsv3err_rmdir[] = {
390 NFSERR_IO,
391 NFSERR_NOENT,
392 NFSERR_IO,
393 NFSERR_ACCES,
394 NFSERR_EXIST,
395 NFSERR_NOTDIR,
396 NFSERR_INVAL,
397 NFSERR_ROFS,
398 NFSERR_NAMETOL,
399 NFSERR_NOTEMPTY,
400 NFSERR_STALE,
401 NFSERR_BADHANDLE,
402 NFSERR_NOTSUPP,
403 NFSERR_SERVERFAULT,
404 0,
405 };
406
407 static short nfsv3err_rename[] = {
408 NFSERR_IO,
409 NFSERR_NOENT,
410 NFSERR_IO,
411 NFSERR_ACCES,
412 NFSERR_EXIST,
413 NFSERR_XDEV,
414 NFSERR_NOTDIR,
415 NFSERR_ISDIR,
416 NFSERR_INVAL,
417 NFSERR_NOSPC,
418 NFSERR_ROFS,
419 NFSERR_MLINK,
420 NFSERR_NAMETOL,
421 NFSERR_NOTEMPTY,
422 NFSERR_DQUOT,
423 NFSERR_STALE,
424 NFSERR_BADHANDLE,
425 NFSERR_NOTSUPP,
426 NFSERR_SERVERFAULT,
427 0,
428 };
429
430 static short nfsv3err_link[] = {
431 NFSERR_IO,
432 NFSERR_IO,
433 NFSERR_ACCES,
434 NFSERR_EXIST,
435 NFSERR_XDEV,
436 NFSERR_NOTDIR,
437 NFSERR_INVAL,
438 NFSERR_NOSPC,
439 NFSERR_ROFS,
440 NFSERR_MLINK,
441 NFSERR_NAMETOL,
442 NFSERR_DQUOT,
443 NFSERR_STALE,
444 NFSERR_BADHANDLE,
445 NFSERR_NOTSUPP,
446 NFSERR_SERVERFAULT,
447 0,
448 };
449
450 static short nfsv3err_readdir[] = {
451 NFSERR_IO,
452 NFSERR_IO,
453 NFSERR_ACCES,
454 NFSERR_NOTDIR,
455 NFSERR_STALE,
456 NFSERR_BADHANDLE,
457 NFSERR_BAD_COOKIE,
458 NFSERR_TOOSMALL,
459 NFSERR_SERVERFAULT,
460 0,
461 };
462
463 static short nfsv3err_readdirplus[] = {
464 NFSERR_IO,
465 NFSERR_IO,
466 NFSERR_ACCES,
467 NFSERR_NOTDIR,
468 NFSERR_STALE,
469 NFSERR_BADHANDLE,
470 NFSERR_BAD_COOKIE,
471 NFSERR_NOTSUPP,
472 NFSERR_TOOSMALL,
473 NFSERR_SERVERFAULT,
474 0,
475 };
476
477 static short nfsv3err_fsstat[] = {
478 NFSERR_IO,
479 NFSERR_IO,
480 NFSERR_STALE,
481 NFSERR_BADHANDLE,
482 NFSERR_SERVERFAULT,
483 0,
484 };
485
486 static short nfsv3err_fsinfo[] = {
487 NFSERR_STALE,
488 NFSERR_STALE,
489 NFSERR_BADHANDLE,
490 NFSERR_SERVERFAULT,
491 0,
492 };
493
494 static short nfsv3err_pathconf[] = {
495 NFSERR_STALE,
496 NFSERR_STALE,
497 NFSERR_BADHANDLE,
498 NFSERR_SERVERFAULT,
499 0,
500 };
501
502 static short nfsv3err_commit[] = {
503 NFSERR_IO,
504 NFSERR_IO,
505 NFSERR_STALE,
506 NFSERR_BADHANDLE,
507 NFSERR_SERVERFAULT,
508 0,
509 };
510
511 static short *nfsrv_v3errmap[] = {
512 nfsv3err_null,
513 nfsv3err_getattr,
514 nfsv3err_setattr,
515 nfsv3err_lookup,
516 nfsv3err_access,
517 nfsv3err_readlink,
518 nfsv3err_read,
519 nfsv3err_write,
520 nfsv3err_create,
521 nfsv3err_mkdir,
522 nfsv3err_symlink,
523 nfsv3err_mknod,
524 nfsv3err_remove,
525 nfsv3err_rmdir,
526 nfsv3err_rename,
527 nfsv3err_link,
528 nfsv3err_readdir,
529 nfsv3err_readdirplus,
530 nfsv3err_fsstat,
531 nfsv3err_fsinfo,
532 nfsv3err_pathconf,
533 nfsv3err_commit,
534 };
535
536 extern struct nfsrtt nfsrtt;
537 extern time_t nqnfsstarttime;
538 extern int nqsrv_clockskew;
539 extern int nqsrv_writeslack;
540 extern int nqsrv_maxlease;
541 extern int nqnfs_piggy[NFS_NPROCS];
542 extern nfstype nfsv2_type[9];
543 extern nfstype nfsv3_type[9];
544 extern struct nfsnodehashhead *nfsnodehashtbl;
545 extern u_long nfsnodehash;
546
547 LIST_HEAD(nfsnodehashhead, nfsnode);
548 u_long nfsdirhashmask;
549
550 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
551
552 /*
553 * Create the header for an rpc request packet
554 * The hsiz is the size of the rest of the nfs request header.
555 * (just used to decide if a cluster is a good idea)
556 */
557 struct mbuf *
558 nfsm_reqh(vp, procid, hsiz, bposp)
559 struct vnode *vp;
560 u_long procid;
561 int hsiz;
562 caddr_t *bposp;
563 {
564 register struct mbuf *mb;
565 register u_int32_t *tl;
566 register caddr_t bpos;
567 struct mbuf *mb2;
568 struct nfsmount *nmp;
569 int nqflag;
570
571 MGET(mb, M_WAIT, MT_DATA);
572 if (hsiz >= MINCLSIZE)
573 MCLGET(mb, M_WAIT);
574 mb->m_len = 0;
575 bpos = mtod(mb, caddr_t);
576
577 /*
578 * For NQNFS, add lease request.
579 */
580 if (vp) {
581 nmp = VFSTONFS(vp->v_mount);
582 if (nmp->nm_flag & NFSMNT_NQNFS) {
583 nqflag = NQNFS_NEEDLEASE(vp, procid);
584 if (nqflag) {
585 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
586 *tl++ = txdr_unsigned(nqflag);
587 *tl = txdr_unsigned(nmp->nm_leaseterm);
588 } else {
589 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
590 *tl = 0;
591 }
592 }
593 }
594 /* Finally, return values */
595 *bposp = bpos;
596 return (mb);
597 }
598
599 /*
600 * Build the RPC header and fill in the authorization info.
601 * The authorization string argument is only used when the credentials
602 * come from outside of the kernel.
603 * Returns the head of the mbuf list.
604 */
605 struct mbuf *
606 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
607 verf_str, mrest, mrest_len, mbp, xidp)
608 register struct ucred *cr;
609 int nmflag;
610 int procid;
611 int auth_type;
612 int auth_len;
613 char *auth_str;
614 int verf_len;
615 char *verf_str;
616 struct mbuf *mrest;
617 int mrest_len;
618 struct mbuf **mbp;
619 u_int32_t *xidp;
620 {
621 register struct mbuf *mb;
622 register u_int32_t *tl;
623 register caddr_t bpos;
624 register int i;
625 struct mbuf *mreq, *mb2;
626 int siz, grpsiz, authsiz;
627 struct timeval tv;
628 static u_int32_t base;
629
630 authsiz = nfsm_rndup(auth_len);
631 MGETHDR(mb, M_WAIT, MT_DATA);
632 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
633 MCLGET(mb, M_WAIT);
634 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
635 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
636 } else {
637 MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
638 }
639 mb->m_len = 0;
640 mreq = mb;
641 bpos = mtod(mb, caddr_t);
642
643 /*
644 * First the RPC header.
645 */
646 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
647
648 /*
649 * derive initial xid from system time
650 * XXX time is invalid if root not yet mounted
651 */
652 if (!base && (rootvp)) {
653 microtime(&tv);
654 base = tv.tv_sec << 12;
655 nfs_xid = base;
656 }
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_gid);
693 grpsiz = (auth_len >> 2) - 5;
694 *tl++ = txdr_unsigned(grpsiz);
695 for (i = 0; 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, M_WAIT, MT_DATA);
703 if (siz >= MINCLSIZE)
704 MCLGET(mb2, M_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 memcpy(bpos, auth_str, 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, M_WAIT, MT_DATA);
736 if (siz >= MINCLSIZE)
737 MCLGET(mb2, M_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 memcpy(bpos, verf_str, 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(mrep, uiop, siz, dpos)
771 struct mbuf **mrep;
772 register struct uio *uiop;
773 int siz;
774 caddr_t *dpos;
775 {
776 register char *mbufcp, *uiocp;
777 register int xfer, left, len;
778 register struct mbuf *mp;
779 long uiosiz, rem;
780 int error = 0;
781
782 mp = *mrep;
783 mbufcp = *dpos;
784 len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
785 rem = nfsm_rndup(siz)-siz;
786 while (siz > 0) {
787 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
788 return (EFBIG);
789 left = uiop->uio_iov->iov_len;
790 uiocp = uiop->uio_iov->iov_base;
791 if (left > siz)
792 left = siz;
793 uiosiz = left;
794 while (left > 0) {
795 while (len == 0) {
796 mp = mp->m_next;
797 if (mp == NULL)
798 return (EBADRPC);
799 mbufcp = mtod(mp, caddr_t);
800 len = mp->m_len;
801 }
802 xfer = (left > len) ? len : left;
803 #ifdef notdef
804 /* Not Yet.. */
805 if (uiop->uio_iov->iov_op != NULL)
806 (*(uiop->uio_iov->iov_op))
807 (mbufcp, uiocp, xfer);
808 else
809 #endif
810 if (uiop->uio_segflg == UIO_SYSSPACE)
811 memcpy(uiocp, mbufcp, xfer);
812 else
813 copyout(mbufcp, uiocp, xfer);
814 left -= xfer;
815 len -= xfer;
816 mbufcp += xfer;
817 uiocp += xfer;
818 uiop->uio_offset += xfer;
819 uiop->uio_resid -= xfer;
820 }
821 if (uiop->uio_iov->iov_len <= siz) {
822 uiop->uio_iovcnt--;
823 uiop->uio_iov++;
824 } else {
825 (caddr_t)uiop->uio_iov->iov_base += uiosiz;
826 uiop->uio_iov->iov_len -= uiosiz;
827 }
828 siz -= uiosiz;
829 }
830 *dpos = mbufcp;
831 *mrep = mp;
832 if (rem > 0) {
833 if (len < rem)
834 error = nfs_adv(mrep, dpos, rem, len);
835 else
836 *dpos += rem;
837 }
838 return (error);
839 }
840
841 /*
842 * copies a uio scatter/gather list to an mbuf chain.
843 * NOTE: can ony handle iovcnt == 1
844 */
845 int
846 nfsm_uiotombuf(uiop, mq, siz, bpos)
847 register struct uio *uiop;
848 struct mbuf **mq;
849 int siz;
850 caddr_t *bpos;
851 {
852 register char *uiocp;
853 register struct mbuf *mp, *mp2;
854 register int xfer, left, mlen;
855 int uiosiz, clflg, rem;
856 char *cp;
857
858 #ifdef DIAGNOSTIC
859 if (uiop->uio_iovcnt != 1)
860 panic("nfsm_uiotombuf: iovcnt != 1");
861 #endif
862
863 if (siz > MLEN) /* or should it >= MCLBYTES ?? */
864 clflg = 1;
865 else
866 clflg = 0;
867 rem = nfsm_rndup(siz)-siz;
868 mp = mp2 = *mq;
869 while (siz > 0) {
870 left = uiop->uio_iov->iov_len;
871 uiocp = uiop->uio_iov->iov_base;
872 if (left > siz)
873 left = siz;
874 uiosiz = left;
875 while (left > 0) {
876 mlen = M_TRAILINGSPACE(mp);
877 if (mlen == 0) {
878 MGET(mp, M_WAIT, MT_DATA);
879 if (clflg)
880 MCLGET(mp, M_WAIT);
881 mp->m_len = 0;
882 mp2->m_next = mp;
883 mp2 = mp;
884 mlen = M_TRAILINGSPACE(mp);
885 }
886 xfer = (left > mlen) ? mlen : left;
887 #ifdef notdef
888 /* Not Yet.. */
889 if (uiop->uio_iov->iov_op != NULL)
890 (*(uiop->uio_iov->iov_op))
891 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
892 else
893 #endif
894 if (uiop->uio_segflg == UIO_SYSSPACE)
895 memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer);
896 else
897 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
898 mp->m_len += xfer;
899 left -= xfer;
900 uiocp += xfer;
901 uiop->uio_offset += xfer;
902 uiop->uio_resid -= xfer;
903 }
904 (caddr_t)uiop->uio_iov->iov_base += uiosiz;
905 uiop->uio_iov->iov_len -= uiosiz;
906 siz -= uiosiz;
907 }
908 if (rem > 0) {
909 if (rem > M_TRAILINGSPACE(mp)) {
910 MGET(mp, M_WAIT, MT_DATA);
911 mp->m_len = 0;
912 mp2->m_next = mp;
913 }
914 cp = mtod(mp, caddr_t)+mp->m_len;
915 for (left = 0; left < rem; left++)
916 *cp++ = '\0';
917 mp->m_len += rem;
918 *bpos = cp;
919 } else
920 *bpos = mtod(mp, caddr_t)+mp->m_len;
921 *mq = mp;
922 return (0);
923 }
924
925 /*
926 * Get at least "siz" bytes of correctly aligned data.
927 * When called the mbuf pointers are not necessarily correct,
928 * dsosp points to what ought to be in m_data and left contains
929 * what ought to be in m_len.
930 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
931 * cases. (The macros use the vars. dpos and dpos2)
932 */
933 int
934 nfsm_disct(mdp, dposp, siz, left, cp2)
935 struct mbuf **mdp;
936 caddr_t *dposp;
937 int siz;
938 int left;
939 caddr_t *cp2;
940 {
941 register struct mbuf *m1, *m2;
942 struct mbuf *havebuf = NULL;
943 caddr_t src = *dposp;
944 caddr_t dst;
945 int len;
946
947 #ifdef DEBUG
948 if (left < 0)
949 panic("nfsm_disct: left < 0");
950 #endif
951 m1 = *mdp;
952 /*
953 * Skip through the mbuf chain looking for an mbuf with
954 * some data. If the first mbuf found has enough data
955 * and it is correctly aligned return it.
956 */
957 while (left == 0) {
958 havebuf = m1;
959 *mdp = m1 = m1->m_next;
960 if (m1 == NULL)
961 return (EBADRPC);
962 src = mtod(m1, caddr_t);
963 left = m1->m_len;
964 /*
965 * If we start a new mbuf and it is big enough
966 * and correctly aligned just return it, don't
967 * do any pull up.
968 */
969 if (left >= siz && nfsm_aligned(src)) {
970 *cp2 = src;
971 *dposp = src + siz;
972 return (0);
973 }
974 }
975 if (m1->m_flags & M_EXT) {
976 if (havebuf) {
977 /* If the first mbuf with data has external data
978 * and there is a previous empty mbuf use it
979 * to move the data into.
980 */
981 m2 = m1;
982 *mdp = m1 = havebuf;
983 if (m1->m_flags & M_EXT) {
984 MEXTREMOVE(m1);
985 }
986 } else {
987 /*
988 * If the first mbuf has a external data
989 * and there is no previous empty mbuf
990 * allocate a new mbuf and move the external
991 * data to the new mbuf. Also make the first
992 * mbuf look empty.
993 */
994 m2 = m_get(M_WAIT, MT_DATA);
995 m2->m_ext = m1->m_ext;
996 m2->m_data = src;
997 m2->m_len = left;
998 MCLADDREFERENCE(m1, m2);
999 MEXTREMOVE(m1);
1000 m2->m_next = m1->m_next;
1001 m1->m_next = m2;
1002 }
1003 m1->m_len = 0;
1004 dst = m1->m_dat;
1005 } else {
1006 /*
1007 * If the first mbuf has no external data
1008 * move the data to the front of the mbuf.
1009 */
1010 if ((dst = m1->m_dat) != src)
1011 memmove(dst, src, left);
1012 dst += left;
1013 m1->m_len = left;
1014 m2 = m1->m_next;
1015 }
1016 m1->m_flags &= ~M_PKTHDR;
1017 *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */
1018 *dposp = mtod(m1, caddr_t) + siz;
1019 /*
1020 * Loop through mbufs pulling data up into first mbuf until
1021 * the first mbuf is full or there is no more data to
1022 * pullup.
1023 */
1024 while ((len = (MLEN - m1->m_len)) != 0 && m2) {
1025 if ((len = min(len, m2->m_len)) != 0)
1026 memcpy(dst, m2->m_data, len);
1027 m1->m_len += len;
1028 dst += len;
1029 m2->m_data += len;
1030 m2->m_len -= len;
1031 m2 = m2->m_next;
1032 }
1033 if (m1->m_len < siz)
1034 return (EBADRPC);
1035 return (0);
1036 }
1037
1038 /*
1039 * Advance the position in the mbuf chain.
1040 */
1041 int
1042 nfs_adv(mdp, dposp, offs, left)
1043 struct mbuf **mdp;
1044 caddr_t *dposp;
1045 int offs;
1046 int left;
1047 {
1048 register struct mbuf *m;
1049 register int s;
1050
1051 m = *mdp;
1052 s = left;
1053 while (s < offs) {
1054 offs -= s;
1055 m = m->m_next;
1056 if (m == NULL)
1057 return (EBADRPC);
1058 s = m->m_len;
1059 }
1060 *mdp = m;
1061 *dposp = mtod(m, caddr_t)+offs;
1062 return (0);
1063 }
1064
1065 /*
1066 * Copy a string into mbufs for the hard cases...
1067 */
1068 int
1069 nfsm_strtmbuf(mb, bpos, cp, siz)
1070 struct mbuf **mb;
1071 char **bpos;
1072 const char *cp;
1073 long siz;
1074 {
1075 register struct mbuf *m1 = NULL, *m2;
1076 long left, xfer, len, tlen;
1077 u_int32_t *tl;
1078 int putsize;
1079
1080 putsize = 1;
1081 m2 = *mb;
1082 left = M_TRAILINGSPACE(m2);
1083 if (left > 0) {
1084 tl = ((u_int32_t *)(*bpos));
1085 *tl++ = txdr_unsigned(siz);
1086 putsize = 0;
1087 left -= NFSX_UNSIGNED;
1088 m2->m_len += NFSX_UNSIGNED;
1089 if (left > 0) {
1090 memcpy((caddr_t) tl, cp, left);
1091 siz -= left;
1092 cp += left;
1093 m2->m_len += left;
1094 left = 0;
1095 }
1096 }
1097 /* Loop around adding mbufs */
1098 while (siz > 0) {
1099 MGET(m1, M_WAIT, MT_DATA);
1100 if (siz > MLEN)
1101 MCLGET(m1, M_WAIT);
1102 m1->m_len = NFSMSIZ(m1);
1103 m2->m_next = m1;
1104 m2 = m1;
1105 tl = mtod(m1, u_int32_t *);
1106 tlen = 0;
1107 if (putsize) {
1108 *tl++ = txdr_unsigned(siz);
1109 m1->m_len -= NFSX_UNSIGNED;
1110 tlen = NFSX_UNSIGNED;
1111 putsize = 0;
1112 }
1113 if (siz < m1->m_len) {
1114 len = nfsm_rndup(siz);
1115 xfer = siz;
1116 if (xfer < len)
1117 *(tl+(xfer>>2)) = 0;
1118 } else {
1119 xfer = len = m1->m_len;
1120 }
1121 memcpy((caddr_t) tl, cp, xfer);
1122 m1->m_len = len+tlen;
1123 siz -= xfer;
1124 cp += xfer;
1125 }
1126 *mb = m1;
1127 *bpos = mtod(m1, caddr_t)+m1->m_len;
1128 return (0);
1129 }
1130
1131 /*
1132 * Directory caching routines. They work as follows:
1133 * - a cache is maintained per VDIR nfsnode.
1134 * - for each offset cookie that is exported to userspace, and can
1135 * thus be thrown back at us as an offset to VOP_READDIR, store
1136 * information in the cache.
1137 * - cached are:
1138 * - cookie itself
1139 * - blocknumber (essentially just a search key in the buffer cache)
1140 * - entry number in block.
1141 * - offset cookie of block in which this entry is stored
1142 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used.
1143 * - entries are looked up in a hash table
1144 * - also maintained is an LRU list of entries, used to determine
1145 * which ones to delete if the cache grows too large.
1146 * - if 32 <-> 64 translation mode is requested for a filesystem,
1147 * the cache also functions as a translation table
1148 * - in the translation case, invalidating the cache does not mean
1149 * flushing it, but just marking entries as invalid, except for
1150 * the <64bit cookie, 32bitcookie> pair which is still valid, to
1151 * still be able to use the cache as a translation table.
1152 * - 32 bit cookies are uniquely created by combining the hash table
1153 * entry value, and one generation count per hash table entry,
1154 * incremented each time an entry is appended to the chain.
1155 * - the cache is invalidated each time a direcory is modified
1156 * - sanity checks are also done; if an entry in a block turns
1157 * out not to have a matching cookie, the cache is invalidated
1158 * and a new block starting from the wanted offset is fetched from
1159 * the server.
1160 * - directory entries as read from the server are extended to contain
1161 * the 64bit and, optionally, the 32bit cookies, for sanity checking
1162 * the cache and exporting them to userspace through the cookie
1163 * argument to VOP_READDIR.
1164 */
1165
1166 u_long
1167 nfs_dirhash(off)
1168 off_t off;
1169 {
1170 int i;
1171 char *cp = (char *)&off;
1172 u_long sum = 0L;
1173
1174 for (i = 0 ; i < sizeof (off); i++)
1175 sum += *cp++;
1176
1177 return sum;
1178 }
1179
1180 void
1181 nfs_initdircache(vp)
1182 struct vnode *vp;
1183 {
1184 struct nfsnode *np = VTONFS(vp);
1185 struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1186
1187 np->n_dircachesize = 0;
1188 np->n_dblkno = 1;
1189 np->n_dircache =
1190 hashinit(NFS_DIRHASHSIZ, M_NFSDIROFF, M_WAITOK, &nfsdirhashmask);
1191 TAILQ_INIT(&np->n_dirchain);
1192 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
1193 MALLOC(np->n_dirgens, unsigned *,
1194 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF,
1195 M_WAITOK);
1196 memset((caddr_t)np->n_dirgens, 0,
1197 NFS_DIRHASHSIZ * sizeof (unsigned));
1198 }
1199 }
1200
1201 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0};
1202
1203 struct nfsdircache *
1204 nfs_searchdircache(vp, off, do32, hashent)
1205 struct vnode *vp;
1206 off_t off;
1207 int do32;
1208 int *hashent;
1209 {
1210 struct nfsdirhashhead *ndhp;
1211 struct nfsdircache *ndp = NULL;
1212 struct nfsnode *np = VTONFS(vp);
1213 unsigned ent;
1214
1215 /*
1216 * Zero is always a valid cookie.
1217 */
1218 if (off == 0)
1219 return &dzero;
1220
1221 /*
1222 * We use a 32bit cookie as search key, directly reconstruct
1223 * the hashentry. Else use the hashfunction.
1224 */
1225 if (do32) {
1226 ent = (u_int32_t)off >> 24;
1227 if (ent >= NFS_DIRHASHSIZ)
1228 return NULL;
1229 ndhp = &np->n_dircache[ent];
1230 } else {
1231 ndhp = NFSDIRHASH(np, off);
1232 }
1233
1234 if (hashent)
1235 *hashent = (int)(ndhp - np->n_dircache);
1236 if (do32) {
1237 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) {
1238 if (ndp->dc_cookie32 == (u_int32_t)off) {
1239 /*
1240 * An invalidated entry will become the
1241 * start of a new block fetched from
1242 * the server.
1243 */
1244 if (ndp->dc_blkno == -1) {
1245 ndp->dc_blkcookie = ndp->dc_cookie;
1246 ndp->dc_blkno = np->n_dblkno++;
1247 ndp->dc_entry = 0;
1248 }
1249 break;
1250 }
1251 }
1252 } else {
1253 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next)
1254 if (ndp->dc_cookie == off)
1255 break;
1256 }
1257 return ndp;
1258 }
1259
1260
1261 struct nfsdircache *
1262 nfs_enterdircache(vp, off, blkoff, en, blkno)
1263 struct vnode *vp;
1264 off_t off, blkoff;
1265 daddr_t blkno;
1266 int en;
1267 {
1268 struct nfsnode *np = VTONFS(vp);
1269 struct nfsdirhashhead *ndhp;
1270 struct nfsdircache *ndp = NULL, *first;
1271 struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1272 int hashent, gen, overwrite;
1273
1274 if (!np->n_dircache)
1275 /*
1276 * XXX would like to do this in nfs_nget but vtype
1277 * isn't known at that time.
1278 */
1279 nfs_initdircache(vp);
1280
1281 /*
1282 * XXX refuse entries for offset 0. amd(8) erroneously sets
1283 * cookie 0 for the '.' entry, making this necessary. This
1284 * isn't so bad, as 0 is a special case anyway.
1285 */
1286 if (off == 0)
1287 return &dzero;
1288
1289 ndp = nfs_searchdircache(vp, off, 0, &hashent);
1290
1291 if (ndp && ndp->dc_blkno != -1) {
1292 /*
1293 * Overwriting an old entry. Check if it's the same.
1294 * If so, just return. If not, remove the old entry.
1295 */
1296 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en)
1297 return ndp;
1298 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
1299 LIST_REMOVE(ndp, dc_hash);
1300 FREE(ndp, M_NFSDIROFF);
1301 ndp = 0;
1302 }
1303
1304 ndhp = &np->n_dircache[hashent];
1305
1306 if (!ndp) {
1307 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF,
1308 M_WAITOK);
1309 overwrite = 0;
1310 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
1311 /*
1312 * We're allocating a new entry, so bump the
1313 * generation number.
1314 */
1315 gen = ++np->n_dirgens[hashent];
1316 if (gen == 0) {
1317 np->n_dirgens[hashent]++;
1318 gen++;
1319 }
1320 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff);
1321 }
1322 } else
1323 overwrite = 1;
1324
1325 /*
1326 * If the entry number is 0, we are at the start of a new block, so
1327 * allocate a new blocknumber.
1328 */
1329 if (en == 0)
1330 ndp->dc_blkno = np->n_dblkno++;
1331 else
1332 ndp->dc_blkno = blkno;
1333
1334 ndp->dc_cookie = off;
1335 ndp->dc_blkcookie = blkoff;
1336 ndp->dc_entry = en;
1337
1338 if (overwrite)
1339 return ndp;
1340
1341 /*
1342 * If the maximum directory cookie cache size has been reached
1343 * for this node, take one off the front. The idea is that
1344 * directories are typically read front-to-back once, so that
1345 * the oldest entries can be thrown away without much performance
1346 * loss.
1347 */
1348 if (np->n_dircachesize == NFS_MAXDIRCACHE) {
1349 first = np->n_dirchain.tqh_first;
1350 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain);
1351 LIST_REMOVE(first, dc_hash);
1352 FREE(first, M_NFSDIROFF);
1353 } else
1354 np->n_dircachesize++;
1355
1356 LIST_INSERT_HEAD(ndhp, ndp, dc_hash);
1357 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain);
1358 return ndp;
1359 }
1360
1361 void
1362 nfs_invaldircache(vp, forcefree)
1363 struct vnode *vp;
1364 int forcefree;
1365 {
1366 struct nfsnode *np = VTONFS(vp);
1367 struct nfsdircache *ndp = NULL;
1368 struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1369
1370 #ifdef DIAGNOSTIC
1371 if (vp->v_type != VDIR)
1372 panic("nfs: invaldircache: not dir");
1373 #endif
1374
1375 if (!np->n_dircache)
1376 return;
1377
1378 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) {
1379 while ((ndp = np->n_dirchain.tqh_first)) {
1380 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
1381 LIST_REMOVE(ndp, dc_hash);
1382 FREE(ndp, M_NFSDIROFF);
1383 }
1384 np->n_dircachesize = 0;
1385 if (forcefree && np->n_dirgens) {
1386 FREE(np->n_dirgens, M_NFSDIROFF);
1387 }
1388 } else {
1389 for (ndp = np->n_dirchain.tqh_first; ndp;
1390 ndp = ndp->dc_chain.tqe_next)
1391 ndp->dc_blkno = -1;
1392 }
1393
1394 np->n_dblkno = 1;
1395 }
1396
1397 /*
1398 * Called once before VFS init to initialize shared and
1399 * server-specific data structures.
1400 */
1401 void
1402 nfs_init()
1403 {
1404
1405 #if !defined(alpha) && defined(DIAGNOSTIC)
1406 /*
1407 * Check to see if major data structures haven't bloated.
1408 */
1409 if (sizeof (struct nfsnode) > NFS_NODEALLOC) {
1410 printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC);
1411 printf("Try reducing NFS_SMALLFH\n");
1412 }
1413 if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) {
1414 printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC);
1415 printf("Try reducing NFS_UIDHASHSIZ\n");
1416 }
1417 if (sizeof (struct nfsuid) > NFS_UIDALLOC) {
1418 printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC);
1419 printf("Try unionizing the nu_nickname and nu_flag fields\n");
1420 }
1421 #endif
1422
1423 nfsrtt.pos = 0;
1424 rpc_vers = txdr_unsigned(RPC_VER2);
1425 rpc_call = txdr_unsigned(RPC_CALL);
1426 rpc_reply = txdr_unsigned(RPC_REPLY);
1427 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1428 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1429 rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1430 rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1431 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1432 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1433 nfs_prog = txdr_unsigned(NFS_PROG);
1434 nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1435 nfs_true = txdr_unsigned(TRUE);
1436 nfs_false = txdr_unsigned(FALSE);
1437 nfs_xdrneg1 = txdr_unsigned(-1);
1438 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1439 if (nfs_ticks < 1)
1440 nfs_ticks = 1;
1441 #ifdef NFSSERVER
1442 nfsrv_init(0); /* Init server data structures */
1443 nfsrv_initcache(); /* Init the server request cache */
1444 #endif /* NFSSERVER */
1445
1446 /*
1447 * Initialize the nqnfs data structures.
1448 */
1449 if (nqnfsstarttime == 0) {
1450 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1451 + nqsrv_clockskew + nqsrv_writeslack;
1452 NQLOADNOVRAM(nqnfsstarttime);
1453 CIRCLEQ_INIT(&nqtimerhead);
1454 nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, M_WAITOK, &nqfhhash);
1455 }
1456
1457 /*
1458 * Initialize reply list and start timer
1459 */
1460 TAILQ_INIT(&nfs_reqq);
1461 nfs_timer(NULL);
1462 }
1463
1464 #ifdef NFS
1465 /*
1466 * Called once at VFS init to initialize client-specific data structures.
1467 */
1468 void
1469 nfs_vfs_init()
1470 {
1471 register int i;
1472
1473 /* Ensure async daemons disabled */
1474 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1475 nfs_iodwant[i] = (struct proc *)0;
1476 nfs_iodmount[i] = (struct nfsmount *)0;
1477 }
1478 nfs_nhinit(); /* Init the nfsnode table */
1479 }
1480
1481 /*
1482 * Attribute cache routines.
1483 * nfs_loadattrcache() - loads or updates the cache contents from attributes
1484 * that are on the mbuf list
1485 * nfs_getattrcache() - returns valid attributes if found in cache, returns
1486 * error otherwise
1487 */
1488
1489 /*
1490 * Load the attribute cache (that lives in the nfsnode entry) with
1491 * the values on the mbuf list and
1492 * Iff vap not NULL
1493 * copy the attributes to *vaper
1494 */
1495 int
1496 nfsm_loadattrcache(vpp, mdp, dposp, vaper)
1497 struct vnode **vpp;
1498 struct mbuf **mdp;
1499 caddr_t *dposp;
1500 struct vattr *vaper;
1501 {
1502 register int32_t t1;
1503 caddr_t cp2;
1504 int error = 0;
1505 struct mbuf *md;
1506 int v3 = NFS_ISV3(*vpp);
1507
1508 md = *mdp;
1509 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1510 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2);
1511 if (error)
1512 return (error);
1513 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper);
1514 }
1515
1516 int
1517 nfs_loadattrcache(vpp, fp, vaper)
1518 struct vnode **vpp;
1519 struct nfs_fattr *fp;
1520 struct vattr *vaper;
1521 {
1522 register struct vnode *vp = *vpp;
1523 register struct vattr *vap;
1524 int v3 = NFS_ISV3(vp);
1525 enum vtype vtyp;
1526 u_short vmode;
1527 struct timespec mtime;
1528 struct vnode *nvp;
1529 int32_t rdev;
1530 register struct nfsnode *np;
1531 extern int (**spec_nfsv2nodeop_p) __P((void *));
1532
1533 if (v3) {
1534 vtyp = nfsv3tov_type(fp->fa_type);
1535 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1536 rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1),
1537 fxdr_unsigned(u_char, fp->fa3_rdev.specdata2));
1538 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1539 } else {
1540 vtyp = nfsv2tov_type(fp->fa_type);
1541 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1542 if (vtyp == VNON || vtyp == VREG)
1543 vtyp = IFTOVT(vmode);
1544 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1545 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1546
1547 /*
1548 * Really ugly NFSv2 kludge.
1549 */
1550 if (vtyp == VCHR && rdev == 0xffffffff)
1551 vtyp = VFIFO;
1552 }
1553
1554 /*
1555 * If v_type == VNON it is a new node, so fill in the v_type,
1556 * n_mtime fields. Check to see if it represents a special
1557 * device, and if so, check for a possible alias. Once the
1558 * correct vnode has been obtained, fill in the rest of the
1559 * information.
1560 */
1561 np = VTONFS(vp);
1562 if (vp->v_type != vtyp) {
1563 vp->v_type = vtyp;
1564 if (vp->v_type == VFIFO) {
1565 extern int (**fifo_nfsv2nodeop_p) __P((void *));
1566 vp->v_op = fifo_nfsv2nodeop_p;
1567 }
1568 if (vp->v_type == VCHR || vp->v_type == VBLK) {
1569 vp->v_op = spec_nfsv2nodeop_p;
1570 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount);
1571 if (nvp) {
1572 /*
1573 * Discard unneeded vnode, but save its nfsnode.
1574 * Since the nfsnode does not have a lock, its
1575 * vnode lock has to be carried over.
1576 */
1577 nvp->v_vnlock = vp->v_vnlock;
1578 vp->v_vnlock = NULL;
1579 nvp->v_data = vp->v_data;
1580 vp->v_data = NULL;
1581 vp->v_op = spec_vnodeop_p;
1582 vrele(vp);
1583 vgone(vp);
1584 /*
1585 * Reinitialize aliased node.
1586 */
1587 np->n_vnode = nvp;
1588 *vpp = vp = nvp;
1589 }
1590 }
1591 np->n_mtime = mtime.tv_sec;
1592 }
1593 vap = np->n_vattr;
1594 vap->va_type = vtyp;
1595 vap->va_mode = vmode & ALLPERMS;
1596 vap->va_rdev = (dev_t)rdev;
1597 vap->va_mtime = mtime;
1598 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1599 if (v3) {
1600 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1601 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1602 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1603 fxdr_hyper(&fp->fa3_size, &vap->va_size);
1604 if (vtyp == VDIR)
1605 vap->va_blocksize = NFS_DIRFRAGSIZ;
1606 else
1607 vap->va_blocksize = NFS_FABLKSIZE;
1608 fxdr_hyper(&fp->fa3_used, &vap->va_bytes);
1609 vap->va_fileid = fxdr_unsigned(int32_t,
1610 fp->fa3_fileid.nfsuquad[1]);
1611 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1612 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1613 vap->va_flags = 0;
1614 vap->va_filerev = 0;
1615 } else {
1616 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1617 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1618 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1619 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1620 if (vtyp == VDIR)
1621 vap->va_blocksize = NFS_DIRFRAGSIZ;
1622 else
1623 vap->va_blocksize =
1624 fxdr_unsigned(int32_t, fp->fa2_blocksize);
1625 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks)
1626 * NFS_FABLKSIZE;
1627 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1628 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1629 vap->va_flags = 0;
1630 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1631 fp->fa2_ctime.nfsv2_sec);
1632 vap->va_ctime.tv_nsec = 0;
1633 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1634 vap->va_filerev = 0;
1635 }
1636 if (vap->va_size != np->n_size) {
1637 if (vap->va_type == VREG) {
1638 if (np->n_flag & NMODIFIED) {
1639 if (vap->va_size < np->n_size)
1640 vap->va_size = np->n_size;
1641 else
1642 np->n_size = vap->va_size;
1643 } else
1644 np->n_size = vap->va_size;
1645 #if defined(UVM)
1646 uvm_vnp_setsize(vp, np->n_size);
1647 #else
1648 vnode_pager_setsize(vp, np->n_size);
1649 #endif
1650 } else
1651 np->n_size = vap->va_size;
1652 }
1653 np->n_attrstamp = time.tv_sec;
1654 if (vaper != NULL) {
1655 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap));
1656 if (np->n_flag & NCHG) {
1657 if (np->n_flag & NACC)
1658 vaper->va_atime = np->n_atim;
1659 if (np->n_flag & NUPD)
1660 vaper->va_mtime = np->n_mtim;
1661 }
1662 }
1663 return (0);
1664 }
1665
1666 /*
1667 * Check the time stamp
1668 * If the cache is valid, copy contents to *vap and return 0
1669 * otherwise return an error
1670 */
1671 int
1672 nfs_getattrcache(vp, vaper)
1673 register struct vnode *vp;
1674 struct vattr *vaper;
1675 {
1676 register struct nfsnode *np = VTONFS(vp);
1677 register struct vattr *vap;
1678
1679 if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) {
1680 nfsstats.attrcache_misses++;
1681 return (ENOENT);
1682 }
1683 nfsstats.attrcache_hits++;
1684 vap = np->n_vattr;
1685 if (vap->va_size != np->n_size) {
1686 if (vap->va_type == VREG) {
1687 if (np->n_flag & NMODIFIED) {
1688 if (vap->va_size < np->n_size)
1689 vap->va_size = np->n_size;
1690 else
1691 np->n_size = vap->va_size;
1692 } else
1693 np->n_size = vap->va_size;
1694 #if defined(UVM)
1695 uvm_vnp_setsize(vp, np->n_size);
1696 #else
1697 vnode_pager_setsize(vp, np->n_size);
1698 #endif
1699 } else
1700 np->n_size = vap->va_size;
1701 }
1702 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr));
1703 if (np->n_flag & NCHG) {
1704 if (np->n_flag & NACC)
1705 vaper->va_atime = np->n_atim;
1706 if (np->n_flag & NUPD)
1707 vaper->va_mtime = np->n_mtim;
1708 }
1709 return (0);
1710 }
1711
1712 /*
1713 * Heuristic to see if the server XDR encodes directory cookies or not.
1714 * it is not supposed to, but a lot of servers may do this. Also, since
1715 * most/all servers will implement V2 as well, it is expected that they
1716 * may return just 32 bits worth of cookie information, so we need to
1717 * find out in which 32 bits this information is available. We do this
1718 * to avoid trouble with emulated binaries that can't handle 64 bit
1719 * directory offsets.
1720 */
1721
1722 void
1723 nfs_cookieheuristic(vp, flagp, p, cred)
1724 struct vnode *vp;
1725 int *flagp;
1726 struct proc *p;
1727 struct ucred *cred;
1728 {
1729 struct uio auio;
1730 struct iovec aiov;
1731 caddr_t buf, cp;
1732 struct dirent *dp;
1733 off_t *cookies = NULL, *cop;
1734 int error, eof, nc, len;
1735
1736 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK);
1737
1738 aiov.iov_base = buf;
1739 aiov.iov_len = NFS_DIRFRAGSIZ;
1740 auio.uio_iov = &aiov;
1741 auio.uio_iovcnt = 1;
1742 auio.uio_rw = UIO_READ;
1743 auio.uio_segflg = UIO_SYSSPACE;
1744 auio.uio_procp = p;
1745 auio.uio_resid = NFS_DIRFRAGSIZ;
1746 auio.uio_offset = 0;
1747
1748 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc);
1749
1750 len = NFS_DIRFRAGSIZ - auio.uio_resid;
1751 if (error || len == 0) {
1752 FREE(buf, M_TEMP);
1753 if (cookies)
1754 FREE(cookies, M_TEMP);
1755 return;
1756 }
1757
1758 /*
1759 * Find the first valid entry and look at its offset cookie.
1760 */
1761
1762 cp = buf;
1763 for (cop = cookies; len > 0; len -= dp->d_reclen) {
1764 dp = (struct dirent *)cp;
1765 if (dp->d_fileno != 0 && len >= dp->d_reclen) {
1766 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) {
1767 *flagp |= NFSMNT_SWAPCOOKIE;
1768 nfs_invaldircache(vp, 0);
1769 nfs_vinvalbuf(vp, 0, cred, p, 1);
1770 }
1771 break;
1772 }
1773 cop++;
1774 cp += dp->d_reclen;
1775 }
1776
1777 FREE(buf, M_TEMP);
1778 FREE(cookies, M_TEMP);
1779 }
1780 #endif /* NFS */
1781
1782 /*
1783 * Set up nameidata for a lookup() call and do it.
1784 *
1785 * If pubflag is set, this call is done for a lookup operation on the
1786 * public filehandle. In that case we allow crossing mountpoints and
1787 * absolute pathnames. However, the caller is expected to check that
1788 * the lookup result is within the public fs, and deny access if
1789 * it is not.
1790 */
1791 int
1792 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
1793 register struct nameidata *ndp;
1794 fhandle_t *fhp;
1795 int len;
1796 struct nfssvc_sock *slp;
1797 struct mbuf *nam;
1798 struct mbuf **mdp;
1799 caddr_t *dposp;
1800 struct vnode **retdirp;
1801 struct proc *p;
1802 int kerbflag, pubflag;
1803 {
1804 register int i, rem;
1805 register struct mbuf *md;
1806 register char *fromcp, *tocp, *cp;
1807 struct iovec aiov;
1808 struct uio auio;
1809 struct vnode *dp;
1810 int error, rdonly, linklen;
1811 struct componentname *cnp = &ndp->ni_cnd;
1812
1813 *retdirp = (struct vnode *)0;
1814 MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK);
1815 /*
1816 * Copy the name from the mbuf list to ndp->ni_pnbuf
1817 * and set the various ndp fields appropriately.
1818 */
1819 fromcp = *dposp;
1820 tocp = cnp->cn_pnbuf;
1821 md = *mdp;
1822 rem = mtod(md, caddr_t) + md->m_len - fromcp;
1823 for (i = 0; i < len; i++) {
1824 while (rem == 0) {
1825 md = md->m_next;
1826 if (md == NULL) {
1827 error = EBADRPC;
1828 goto out;
1829 }
1830 fromcp = mtod(md, caddr_t);
1831 rem = md->m_len;
1832 }
1833 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1834 error = EACCES;
1835 goto out;
1836 }
1837 *tocp++ = *fromcp++;
1838 rem--;
1839 }
1840 *tocp = '\0';
1841 *mdp = md;
1842 *dposp = fromcp;
1843 len = nfsm_rndup(len)-len;
1844 if (len > 0) {
1845 if (rem >= len)
1846 *dposp += len;
1847 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1848 goto out;
1849 }
1850
1851 /*
1852 * Extract and set starting directory.
1853 */
1854 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
1855 nam, &rdonly, kerbflag, pubflag);
1856 if (error)
1857 goto out;
1858 if (dp->v_type != VDIR) {
1859 vrele(dp);
1860 error = ENOTDIR;
1861 goto out;
1862 }
1863
1864 if (rdonly)
1865 cnp->cn_flags |= RDONLY;
1866
1867 *retdirp = dp;
1868
1869 if (pubflag) {
1870 /*
1871 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1872 * and the 'native path' indicator.
1873 */
1874 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1875 fromcp = cnp->cn_pnbuf;
1876 tocp = cp;
1877 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1878 switch ((unsigned char)*fromcp) {
1879 case WEBNFS_NATIVE_CHAR:
1880 /*
1881 * 'Native' path for us is the same
1882 * as a path according to the NFS spec,
1883 * just skip the escape char.
1884 */
1885 fromcp++;
1886 break;
1887 /*
1888 * More may be added in the future, range 0x80-0xff
1889 */
1890 default:
1891 error = EIO;
1892 FREE(cp, M_NAMEI);
1893 goto out;
1894 }
1895 }
1896 /*
1897 * Translate the '%' escapes, URL-style.
1898 */
1899 while (*fromcp != '\0') {
1900 if (*fromcp == WEBNFS_ESC_CHAR) {
1901 if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1902 fromcp++;
1903 *tocp++ = HEXSTRTOI(fromcp);
1904 fromcp += 2;
1905 continue;
1906 } else {
1907 error = ENOENT;
1908 FREE(cp, M_NAMEI);
1909 goto out;
1910 }
1911 } else
1912 *tocp++ = *fromcp++;
1913 }
1914 *tocp = '\0';
1915 FREE(cnp->cn_pnbuf, M_NAMEI);
1916 cnp->cn_pnbuf = cp;
1917 }
1918
1919 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
1920 ndp->ni_segflg = UIO_SYSSPACE;
1921
1922 if (pubflag) {
1923 ndp->ni_rootdir = rootvnode;
1924 ndp->ni_loopcnt = 0;
1925 if (cnp->cn_pnbuf[0] == '/')
1926 dp = rootvnode;
1927 } else {
1928 cnp->cn_flags |= NOCROSSMOUNT;
1929 }
1930
1931 cnp->cn_proc = p;
1932 VREF(dp);
1933
1934 for (;;) {
1935 cnp->cn_nameptr = cnp->cn_pnbuf;
1936 ndp->ni_startdir = dp;
1937 /*
1938 * And call lookup() to do the real work
1939 */
1940 error = lookup(ndp);
1941 if (error)
1942 break;
1943 /*
1944 * Check for encountering a symbolic link
1945 */
1946 if ((cnp->cn_flags & ISSYMLINK) == 0) {
1947 if (cnp->cn_flags & (SAVENAME | SAVESTART)) {
1948 cnp->cn_flags |= HASBUF;
1949 return (0);
1950 }
1951 break;
1952 } else {
1953 if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
1954 VOP_UNLOCK(ndp->ni_dvp, 0);
1955 if (!pubflag) {
1956 vrele(ndp->ni_dvp);
1957 vput(ndp->ni_vp);
1958 ndp->ni_vp = NULL;
1959 error = EINVAL;
1960 break;
1961 }
1962
1963 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
1964 error = ELOOP;
1965 break;
1966 }
1967 if (ndp->ni_pathlen > 1)
1968 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1969 else
1970 cp = cnp->cn_pnbuf;
1971 aiov.iov_base = cp;
1972 aiov.iov_len = MAXPATHLEN;
1973 auio.uio_iov = &aiov;
1974 auio.uio_iovcnt = 1;
1975 auio.uio_offset = 0;
1976 auio.uio_rw = UIO_READ;
1977 auio.uio_segflg = UIO_SYSSPACE;
1978 auio.uio_procp = (struct proc *)0;
1979 auio.uio_resid = MAXPATHLEN;
1980 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
1981 if (error) {
1982 badlink:
1983 if (ndp->ni_pathlen > 1)
1984 FREE(cp, M_NAMEI);
1985 break;
1986 }
1987 linklen = MAXPATHLEN - auio.uio_resid;
1988 if (linklen == 0) {
1989 error = ENOENT;
1990 goto badlink;
1991 }
1992 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
1993 error = ENAMETOOLONG;
1994 goto badlink;
1995 }
1996 if (ndp->ni_pathlen > 1) {
1997 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen);
1998 FREE(cnp->cn_pnbuf, M_NAMEI);
1999 cnp->cn_pnbuf = cp;
2000 } else
2001 cnp->cn_pnbuf[linklen] = '\0';
2002 ndp->ni_pathlen += linklen;
2003 vput(ndp->ni_vp);
2004 dp = ndp->ni_dvp;
2005 /*
2006 * Check if root directory should replace current directory.
2007 */
2008 if (cnp->cn_pnbuf[0] == '/') {
2009 vrele(dp);
2010 dp = ndp->ni_rootdir;
2011 VREF(dp);
2012 }
2013 }
2014 }
2015 out:
2016 FREE(cnp->cn_pnbuf, M_NAMEI);
2017 return (error);
2018 }
2019
2020 /*
2021 * A fiddled version of m_adj() that ensures null fill to a long
2022 * boundary and only trims off the back end
2023 */
2024 void
2025 nfsm_adj(mp, len, nul)
2026 struct mbuf *mp;
2027 register int len;
2028 int nul;
2029 {
2030 register struct mbuf *m;
2031 register int count, i;
2032 register char *cp;
2033
2034 /*
2035 * Trim from tail. Scan the mbuf chain,
2036 * calculating its length and finding the last mbuf.
2037 * If the adjustment only affects this mbuf, then just
2038 * adjust and return. Otherwise, rescan and truncate
2039 * after the remaining size.
2040 */
2041 count = 0;
2042 m = mp;
2043 for (;;) {
2044 count += m->m_len;
2045 if (m->m_next == (struct mbuf *)0)
2046 break;
2047 m = m->m_next;
2048 }
2049 if (m->m_len > len) {
2050 m->m_len -= len;
2051 if (nul > 0) {
2052 cp = mtod(m, caddr_t)+m->m_len-nul;
2053 for (i = 0; i < nul; i++)
2054 *cp++ = '\0';
2055 }
2056 return;
2057 }
2058 count -= len;
2059 if (count < 0)
2060 count = 0;
2061 /*
2062 * Correct length for chain is "count".
2063 * Find the mbuf with last data, adjust its length,
2064 * and toss data from remaining mbufs on chain.
2065 */
2066 for (m = mp; m; m = m->m_next) {
2067 if (m->m_len >= count) {
2068 m->m_len = count;
2069 if (nul > 0) {
2070 cp = mtod(m, caddr_t)+m->m_len-nul;
2071 for (i = 0; i < nul; i++)
2072 *cp++ = '\0';
2073 }
2074 break;
2075 }
2076 count -= m->m_len;
2077 }
2078 for (m = m->m_next;m;m = m->m_next)
2079 m->m_len = 0;
2080 }
2081
2082 /*
2083 * Make these functions instead of macros, so that the kernel text size
2084 * doesn't get too big...
2085 */
2086 void
2087 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
2088 struct nfsrv_descript *nfsd;
2089 int before_ret;
2090 register struct vattr *before_vap;
2091 int after_ret;
2092 struct vattr *after_vap;
2093 struct mbuf **mbp;
2094 char **bposp;
2095 {
2096 register struct mbuf *mb = *mbp, *mb2;
2097 register char *bpos = *bposp;
2098 register u_int32_t *tl;
2099
2100 if (before_ret) {
2101 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
2102 *tl = nfs_false;
2103 } else {
2104 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
2105 *tl++ = nfs_true;
2106 txdr_hyper(&(before_vap->va_size), tl);
2107 tl += 2;
2108 txdr_nfsv3time(&(before_vap->va_mtime), tl);
2109 tl += 2;
2110 txdr_nfsv3time(&(before_vap->va_ctime), tl);
2111 }
2112 *bposp = bpos;
2113 *mbp = mb;
2114 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
2115 }
2116
2117 void
2118 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
2119 struct nfsrv_descript *nfsd;
2120 int after_ret;
2121 struct vattr *after_vap;
2122 struct mbuf **mbp;
2123 char **bposp;
2124 {
2125 register struct mbuf *mb = *mbp, *mb2;
2126 register char *bpos = *bposp;
2127 register u_int32_t *tl;
2128 register struct nfs_fattr *fp;
2129
2130 if (after_ret) {
2131 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
2132 *tl = nfs_false;
2133 } else {
2134 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
2135 *tl++ = nfs_true;
2136 fp = (struct nfs_fattr *)tl;
2137 nfsm_srvfattr(nfsd, after_vap, fp);
2138 }
2139 *mbp = mb;
2140 *bposp = bpos;
2141 }
2142
2143 void
2144 nfsm_srvfattr(nfsd, vap, fp)
2145 register struct nfsrv_descript *nfsd;
2146 register struct vattr *vap;
2147 register struct nfs_fattr *fp;
2148 {
2149
2150 fp->fa_nlink = txdr_unsigned(vap->va_nlink);
2151 fp->fa_uid = txdr_unsigned(vap->va_uid);
2152 fp->fa_gid = txdr_unsigned(vap->va_gid);
2153 if (nfsd->nd_flag & ND_NFSV3) {
2154 fp->fa_type = vtonfsv3_type(vap->va_type);
2155 fp->fa_mode = vtonfsv3_mode(vap->va_mode);
2156 txdr_hyper(&vap->va_size, &fp->fa3_size);
2157 txdr_hyper(&vap->va_bytes, &fp->fa3_used);
2158 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev));
2159 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev));
2160 fp->fa3_fsid.nfsuquad[0] = 0;
2161 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
2162 fp->fa3_fileid.nfsuquad[0] = 0;
2163 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
2164 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
2165 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
2166 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
2167 } else {
2168 fp->fa_type = vtonfsv2_type(vap->va_type);
2169 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
2170 fp->fa2_size = txdr_unsigned(vap->va_size);
2171 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
2172 if (vap->va_type == VFIFO)
2173 fp->fa2_rdev = 0xffffffff;
2174 else
2175 fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
2176 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
2177 fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
2178 fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
2179 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
2180 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
2181 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
2182 }
2183 }
2184
2185 /*
2186 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
2187 * - look up fsid in mount list (if not found ret error)
2188 * - get vp and export rights by calling VFS_FHTOVP()
2189 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
2190 * - if not lockflag unlock it with VOP_UNLOCK()
2191 */
2192 int
2193 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
2194 fhandle_t *fhp;
2195 int lockflag;
2196 struct vnode **vpp;
2197 struct ucred *cred;
2198 struct nfssvc_sock *slp;
2199 struct mbuf *nam;
2200 int *rdonlyp;
2201 int kerbflag;
2202 {
2203 register struct mount *mp;
2204 register int i;
2205 struct ucred *credanon;
2206 int error, exflags;
2207 struct sockaddr_in *saddr;
2208
2209 *vpp = (struct vnode *)0;
2210
2211 if (nfs_ispublicfh(fhp)) {
2212 if (!pubflag || !nfs_pub.np_valid)
2213 return (ESTALE);
2214 fhp = &nfs_pub.np_handle;
2215 }
2216
2217 mp = vfs_getvfs(&fhp->fh_fsid);
2218 if (!mp)
2219 return (ESTALE);
2220 error = VFS_CHKEXP(mp, nam, &exflags, &credanon);
2221 if (error)
2222 return (error);
2223 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
2224 if (error)
2225 return (error);
2226
2227 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
2228 saddr = mtod(nam, struct sockaddr_in *);
2229 if (saddr->sin_family == AF_INET &&
2230 ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
2231 vput(*vpp);
2232 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
2233 }
2234 }
2235 /*
2236 * Check/setup credentials.
2237 */
2238 if (exflags & MNT_EXKERB) {
2239 if (!kerbflag) {
2240 vput(*vpp);
2241 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
2242 }
2243 } else if (kerbflag) {
2244 vput(*vpp);
2245 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
2246 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
2247 cred->cr_uid = credanon->cr_uid;
2248 cred->cr_gid = credanon->cr_gid;
2249 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
2250 cred->cr_groups[i] = credanon->cr_groups[i];
2251 cred->cr_ngroups = i;
2252 }
2253 if (exflags & MNT_EXRDONLY)
2254 *rdonlyp = 1;
2255 else
2256 *rdonlyp = 0;
2257 if (!lockflag)
2258 VOP_UNLOCK(*vpp, 0);
2259 return (0);
2260 }
2261
2262 /*
2263 * WebNFS: check if a filehandle is a public filehandle. For v3, this
2264 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
2265 * transformed this to all zeroes in both cases, so check for it.
2266 */
2267 int
2268 nfs_ispublicfh(fhp)
2269 fhandle_t *fhp;
2270 {
2271 char *cp = (char *)fhp;
2272 int i;
2273
2274 for (i = 0; i < NFSX_V3FH; i++)
2275 if (*cp++ != 0)
2276 return (FALSE);
2277 return (TRUE);
2278 }
2279
2280 /*
2281 * This function compares two net addresses by family and returns TRUE
2282 * if they are the same host.
2283 * If there is any doubt, return FALSE.
2284 * The AF_INET family is handled as a special case so that address mbufs
2285 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
2286 */
2287 int
2288 netaddr_match(family, haddr, nam)
2289 int family;
2290 union nethostaddr *haddr;
2291 struct mbuf *nam;
2292 {
2293 register struct sockaddr_in *inetaddr;
2294
2295 switch (family) {
2296 case AF_INET:
2297 inetaddr = mtod(nam, struct sockaddr_in *);
2298 if (inetaddr->sin_family == AF_INET &&
2299 inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
2300 return (1);
2301 break;
2302 #ifdef ISO
2303 case AF_ISO:
2304 {
2305 register struct sockaddr_iso *isoaddr1, *isoaddr2;
2306
2307 isoaddr1 = mtod(nam, struct sockaddr_iso *);
2308 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *);
2309 if (isoaddr1->siso_family == AF_ISO &&
2310 isoaddr1->siso_nlen > 0 &&
2311 isoaddr1->siso_nlen == isoaddr2->siso_nlen &&
2312 SAME_ISOADDR(isoaddr1, isoaddr2))
2313 return (1);
2314 break;
2315 }
2316 #endif /* ISO */
2317 default:
2318 break;
2319 };
2320 return (0);
2321 }
2322
2323
2324 /*
2325 * The write verifier has changed (probably due to a server reboot), so all
2326 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2327 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2328 * flag. Once done the new write verifier can be set for the mount point.
2329 */
2330 void
2331 nfs_clearcommit(mp)
2332 struct mount *mp;
2333 {
2334 register struct vnode *vp, *nvp;
2335 register struct buf *bp, *nbp;
2336 int s;
2337
2338 s = splbio();
2339 loop:
2340 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
2341 if (vp->v_mount != mp) /* Paranoia */
2342 goto loop;
2343 nvp = vp->v_mntvnodes.le_next;
2344 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) {
2345 nbp = bp->b_vnbufs.le_next;
2346 if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT))
2347 == (B_DELWRI | B_NEEDCOMMIT))
2348 bp->b_flags &= ~B_NEEDCOMMIT;
2349 }
2350 }
2351 splx(s);
2352 }
2353
2354 /*
2355 * Map errnos to NFS error numbers. For Version 3 also filter out error
2356 * numbers not specified for the associated procedure.
2357 */
2358 int
2359 nfsrv_errmap(nd, err)
2360 struct nfsrv_descript *nd;
2361 register int err;
2362 {
2363 register short *defaulterrp, *errp;
2364
2365 if (nd->nd_flag & ND_NFSV3) {
2366 if (nd->nd_procnum <= NFSPROC_COMMIT) {
2367 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2368 while (*++errp) {
2369 if (*errp == err)
2370 return (err);
2371 else if (*errp > err)
2372 break;
2373 }
2374 return ((int)*defaulterrp);
2375 } else
2376 return (err & 0xffff);
2377 }
2378 if (err <= ELAST)
2379 return ((int)nfsrv_v2errmap[err - 1]);
2380 return (NFSERR_IO);
2381 }
2382
2383 /*
2384 * Sort the group list in increasing numerical order.
2385 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2386 * that used to be here.)
2387 */
2388 void
2389 nfsrvw_sort(list, num)
2390 register gid_t *list;
2391 register int num;
2392 {
2393 register int i, j;
2394 gid_t v;
2395
2396 /* Insertion sort. */
2397 for (i = 1; i < num; i++) {
2398 v = list[i];
2399 /* find correct slot for value v, moving others up */
2400 for (j = i; --j >= 0 && v < list[j];)
2401 list[j + 1] = list[j];
2402 list[j + 1] = v;
2403 }
2404 }
2405
2406 /*
2407 * copy credentials making sure that the result can be compared with memcmp().
2408 */
2409 void
2410 nfsrv_setcred(incred, outcred)
2411 register struct ucred *incred, *outcred;
2412 {
2413 register int i;
2414
2415 memset((caddr_t)outcred, 0, sizeof (struct ucred));
2416 outcred->cr_ref = 1;
2417 outcred->cr_uid = incred->cr_uid;
2418 outcred->cr_gid = incred->cr_gid;
2419 outcred->cr_ngroups = incred->cr_ngroups;
2420 for (i = 0; i < incred->cr_ngroups; i++)
2421 outcred->cr_groups[i] = incred->cr_groups[i];
2422 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2423 }
2424