nfs_subs.c revision 1.55 1 /* $NetBSD: nfs_subs.c,v 1.55 1998/02/19 00:54:13 thorpej 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_uvm.h"
43
44 /*
45 * These functions support the macros and help fiddle mbuf chains for
46 * the nfs op functions. They do things like create the rpc header and
47 * copy data between mbuf chains and uio lists.
48 */
49 #include <sys/param.h>
50 #include <sys/proc.h>
51 #include <sys/systm.h>
52 #include <sys/kernel.h>
53 #include <sys/mount.h>
54 #include <sys/vnode.h>
55 #include <sys/namei.h>
56 #include <sys/mbuf.h>
57 #include <sys/socket.h>
58 #include <sys/stat.h>
59 #include <sys/malloc.h>
60 #include <sys/time.h>
61 #include <sys/dirent.h>
62
63 #include <vm/vm.h>
64
65 #if defined(UVM)
66 #include <uvm/uvm_extern.h>
67 #endif
68
69 #include <nfs/rpcv2.h>
70 #include <nfs/nfsproto.h>
71 #include <nfs/nfsnode.h>
72 #include <nfs/nfs.h>
73 #include <nfs/xdr_subs.h>
74 #include <nfs/nfsm_subs.h>
75 #include <nfs/nfsmount.h>
76 #include <nfs/nqnfs.h>
77 #include <nfs/nfsrtt.h>
78 #include <nfs/nfs_var.h>
79
80 #include <miscfs/specfs/specdev.h>
81
82 #include <vm/vm.h>
83
84 #include <netinet/in.h>
85 #ifdef ISO
86 #include <netiso/iso.h>
87 #endif
88
89 /*
90 * Data items converted to xdr at startup, since they are constant
91 * This is kinda hokey, but may save a little time doing byte swaps
92 */
93 u_int32_t nfs_xdrneg1;
94 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
95 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
96 rpc_auth_kerb;
97 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
98
99 /* And other global data */
100 static u_int32_t nfs_xid = 0;
101 nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON,
102 NFCHR, NFNON };
103 nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK,
104 NFFIFO, NFNON };
105 enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON };
106 enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO };
107 int nfs_ticks;
108 extern struct nfs_public nfs_pub;
109
110 /* NFS client/server stats. */
111 struct nfsstats nfsstats;
112
113 /*
114 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
115 */
116 int nfsv3_procid[NFS_NPROCS] = {
117 NFSPROC_NULL,
118 NFSPROC_GETATTR,
119 NFSPROC_SETATTR,
120 NFSPROC_NOOP,
121 NFSPROC_LOOKUP,
122 NFSPROC_READLINK,
123 NFSPROC_READ,
124 NFSPROC_NOOP,
125 NFSPROC_WRITE,
126 NFSPROC_CREATE,
127 NFSPROC_REMOVE,
128 NFSPROC_RENAME,
129 NFSPROC_LINK,
130 NFSPROC_SYMLINK,
131 NFSPROC_MKDIR,
132 NFSPROC_RMDIR,
133 NFSPROC_READDIR,
134 NFSPROC_FSSTAT,
135 NFSPROC_NOOP,
136 NFSPROC_NOOP,
137 NFSPROC_NOOP,
138 NFSPROC_NOOP,
139 NFSPROC_NOOP,
140 NFSPROC_NOOP,
141 NFSPROC_NOOP,
142 NFSPROC_NOOP
143 };
144
145 /*
146 * and the reverse mapping from generic to Version 2 procedure numbers
147 */
148 int nfsv2_procid[NFS_NPROCS] = {
149 NFSV2PROC_NULL,
150 NFSV2PROC_GETATTR,
151 NFSV2PROC_SETATTR,
152 NFSV2PROC_LOOKUP,
153 NFSV2PROC_NOOP,
154 NFSV2PROC_READLINK,
155 NFSV2PROC_READ,
156 NFSV2PROC_WRITE,
157 NFSV2PROC_CREATE,
158 NFSV2PROC_MKDIR,
159 NFSV2PROC_SYMLINK,
160 NFSV2PROC_CREATE,
161 NFSV2PROC_REMOVE,
162 NFSV2PROC_RMDIR,
163 NFSV2PROC_RENAME,
164 NFSV2PROC_LINK,
165 NFSV2PROC_READDIR,
166 NFSV2PROC_NOOP,
167 NFSV2PROC_STATFS,
168 NFSV2PROC_NOOP,
169 NFSV2PROC_NOOP,
170 NFSV2PROC_NOOP,
171 NFSV2PROC_NOOP,
172 NFSV2PROC_NOOP,
173 NFSV2PROC_NOOP,
174 NFSV2PROC_NOOP,
175 };
176
177 /*
178 * Maps errno values to nfs error numbers.
179 * Use NFSERR_IO as the catch all for ones not specifically defined in
180 * RFC 1094.
181 */
182 static u_char nfsrv_v2errmap[ELAST] = {
183 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO,
184 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
185 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO,
186 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR,
187 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
188 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS,
189 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
190 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
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_NAMETOL, NFSERR_IO, NFSERR_IO,
196 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE,
197 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
198 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
199 NFSERR_IO, NFSERR_IO,
200 };
201
202 /*
203 * Maps errno values to nfs error numbers.
204 * Although it is not obvious whether or not NFS clients really care if
205 * a returned error value is in the specified list for the procedure, the
206 * safest thing to do is filter them appropriately. For Version 2, the
207 * X/Open XNFS document is the only specification that defines error values
208 * for each RPC (The RFC simply lists all possible error values for all RPCs),
209 * so I have decided to not do this for Version 2.
210 * The first entry is the default error return and the rest are the valid
211 * errors for that RPC in increasing numeric order.
212 */
213 static short nfsv3err_null[] = {
214 0,
215 0,
216 };
217
218 static short nfsv3err_getattr[] = {
219 NFSERR_IO,
220 NFSERR_IO,
221 NFSERR_STALE,
222 NFSERR_BADHANDLE,
223 NFSERR_SERVERFAULT,
224 0,
225 };
226
227 static short nfsv3err_setattr[] = {
228 NFSERR_IO,
229 NFSERR_PERM,
230 NFSERR_IO,
231 NFSERR_ACCES,
232 NFSERR_INVAL,
233 NFSERR_NOSPC,
234 NFSERR_ROFS,
235 NFSERR_DQUOT,
236 NFSERR_STALE,
237 NFSERR_BADHANDLE,
238 NFSERR_NOT_SYNC,
239 NFSERR_SERVERFAULT,
240 0,
241 };
242
243 static short nfsv3err_lookup[] = {
244 NFSERR_IO,
245 NFSERR_NOENT,
246 NFSERR_IO,
247 NFSERR_ACCES,
248 NFSERR_NOTDIR,
249 NFSERR_NAMETOL,
250 NFSERR_STALE,
251 NFSERR_BADHANDLE,
252 NFSERR_SERVERFAULT,
253 0,
254 };
255
256 static short nfsv3err_access[] = {
257 NFSERR_IO,
258 NFSERR_IO,
259 NFSERR_STALE,
260 NFSERR_BADHANDLE,
261 NFSERR_SERVERFAULT,
262 0,
263 };
264
265 static short nfsv3err_readlink[] = {
266 NFSERR_IO,
267 NFSERR_IO,
268 NFSERR_ACCES,
269 NFSERR_INVAL,
270 NFSERR_STALE,
271 NFSERR_BADHANDLE,
272 NFSERR_NOTSUPP,
273 NFSERR_SERVERFAULT,
274 0,
275 };
276
277 static short nfsv3err_read[] = {
278 NFSERR_IO,
279 NFSERR_IO,
280 NFSERR_NXIO,
281 NFSERR_ACCES,
282 NFSERR_INVAL,
283 NFSERR_STALE,
284 NFSERR_BADHANDLE,
285 NFSERR_SERVERFAULT,
286 0,
287 };
288
289 static short nfsv3err_write[] = {
290 NFSERR_IO,
291 NFSERR_IO,
292 NFSERR_ACCES,
293 NFSERR_INVAL,
294 NFSERR_FBIG,
295 NFSERR_NOSPC,
296 NFSERR_ROFS,
297 NFSERR_DQUOT,
298 NFSERR_STALE,
299 NFSERR_BADHANDLE,
300 NFSERR_SERVERFAULT,
301 0,
302 };
303
304 static short nfsv3err_create[] = {
305 NFSERR_IO,
306 NFSERR_IO,
307 NFSERR_ACCES,
308 NFSERR_EXIST,
309 NFSERR_NOTDIR,
310 NFSERR_NOSPC,
311 NFSERR_ROFS,
312 NFSERR_NAMETOL,
313 NFSERR_DQUOT,
314 NFSERR_STALE,
315 NFSERR_BADHANDLE,
316 NFSERR_NOTSUPP,
317 NFSERR_SERVERFAULT,
318 0,
319 };
320
321 static short nfsv3err_mkdir[] = {
322 NFSERR_IO,
323 NFSERR_IO,
324 NFSERR_ACCES,
325 NFSERR_EXIST,
326 NFSERR_NOTDIR,
327 NFSERR_NOSPC,
328 NFSERR_ROFS,
329 NFSERR_NAMETOL,
330 NFSERR_DQUOT,
331 NFSERR_STALE,
332 NFSERR_BADHANDLE,
333 NFSERR_NOTSUPP,
334 NFSERR_SERVERFAULT,
335 0,
336 };
337
338 static short nfsv3err_symlink[] = {
339 NFSERR_IO,
340 NFSERR_IO,
341 NFSERR_ACCES,
342 NFSERR_EXIST,
343 NFSERR_NOTDIR,
344 NFSERR_NOSPC,
345 NFSERR_ROFS,
346 NFSERR_NAMETOL,
347 NFSERR_DQUOT,
348 NFSERR_STALE,
349 NFSERR_BADHANDLE,
350 NFSERR_NOTSUPP,
351 NFSERR_SERVERFAULT,
352 0,
353 };
354
355 static short nfsv3err_mknod[] = {
356 NFSERR_IO,
357 NFSERR_IO,
358 NFSERR_ACCES,
359 NFSERR_EXIST,
360 NFSERR_NOTDIR,
361 NFSERR_NOSPC,
362 NFSERR_ROFS,
363 NFSERR_NAMETOL,
364 NFSERR_DQUOT,
365 NFSERR_STALE,
366 NFSERR_BADHANDLE,
367 NFSERR_NOTSUPP,
368 NFSERR_SERVERFAULT,
369 NFSERR_BADTYPE,
370 0,
371 };
372
373 static short nfsv3err_remove[] = {
374 NFSERR_IO,
375 NFSERR_NOENT,
376 NFSERR_IO,
377 NFSERR_ACCES,
378 NFSERR_NOTDIR,
379 NFSERR_ROFS,
380 NFSERR_NAMETOL,
381 NFSERR_STALE,
382 NFSERR_BADHANDLE,
383 NFSERR_SERVERFAULT,
384 0,
385 };
386
387 static short nfsv3err_rmdir[] = {
388 NFSERR_IO,
389 NFSERR_NOENT,
390 NFSERR_IO,
391 NFSERR_ACCES,
392 NFSERR_EXIST,
393 NFSERR_NOTDIR,
394 NFSERR_INVAL,
395 NFSERR_ROFS,
396 NFSERR_NAMETOL,
397 NFSERR_NOTEMPTY,
398 NFSERR_STALE,
399 NFSERR_BADHANDLE,
400 NFSERR_NOTSUPP,
401 NFSERR_SERVERFAULT,
402 0,
403 };
404
405 static short nfsv3err_rename[] = {
406 NFSERR_IO,
407 NFSERR_NOENT,
408 NFSERR_IO,
409 NFSERR_ACCES,
410 NFSERR_EXIST,
411 NFSERR_XDEV,
412 NFSERR_NOTDIR,
413 NFSERR_ISDIR,
414 NFSERR_INVAL,
415 NFSERR_NOSPC,
416 NFSERR_ROFS,
417 NFSERR_MLINK,
418 NFSERR_NAMETOL,
419 NFSERR_NOTEMPTY,
420 NFSERR_DQUOT,
421 NFSERR_STALE,
422 NFSERR_BADHANDLE,
423 NFSERR_NOTSUPP,
424 NFSERR_SERVERFAULT,
425 0,
426 };
427
428 static short nfsv3err_link[] = {
429 NFSERR_IO,
430 NFSERR_IO,
431 NFSERR_ACCES,
432 NFSERR_EXIST,
433 NFSERR_XDEV,
434 NFSERR_NOTDIR,
435 NFSERR_INVAL,
436 NFSERR_NOSPC,
437 NFSERR_ROFS,
438 NFSERR_MLINK,
439 NFSERR_NAMETOL,
440 NFSERR_DQUOT,
441 NFSERR_STALE,
442 NFSERR_BADHANDLE,
443 NFSERR_NOTSUPP,
444 NFSERR_SERVERFAULT,
445 0,
446 };
447
448 static short nfsv3err_readdir[] = {
449 NFSERR_IO,
450 NFSERR_IO,
451 NFSERR_ACCES,
452 NFSERR_NOTDIR,
453 NFSERR_STALE,
454 NFSERR_BADHANDLE,
455 NFSERR_BAD_COOKIE,
456 NFSERR_TOOSMALL,
457 NFSERR_SERVERFAULT,
458 0,
459 };
460
461 static short nfsv3err_readdirplus[] = {
462 NFSERR_IO,
463 NFSERR_IO,
464 NFSERR_ACCES,
465 NFSERR_NOTDIR,
466 NFSERR_STALE,
467 NFSERR_BADHANDLE,
468 NFSERR_BAD_COOKIE,
469 NFSERR_NOTSUPP,
470 NFSERR_TOOSMALL,
471 NFSERR_SERVERFAULT,
472 0,
473 };
474
475 static short nfsv3err_fsstat[] = {
476 NFSERR_IO,
477 NFSERR_IO,
478 NFSERR_STALE,
479 NFSERR_BADHANDLE,
480 NFSERR_SERVERFAULT,
481 0,
482 };
483
484 static short nfsv3err_fsinfo[] = {
485 NFSERR_STALE,
486 NFSERR_STALE,
487 NFSERR_BADHANDLE,
488 NFSERR_SERVERFAULT,
489 0,
490 };
491
492 static short nfsv3err_pathconf[] = {
493 NFSERR_STALE,
494 NFSERR_STALE,
495 NFSERR_BADHANDLE,
496 NFSERR_SERVERFAULT,
497 0,
498 };
499
500 static short nfsv3err_commit[] = {
501 NFSERR_IO,
502 NFSERR_IO,
503 NFSERR_STALE,
504 NFSERR_BADHANDLE,
505 NFSERR_SERVERFAULT,
506 0,
507 };
508
509 static short *nfsrv_v3errmap[] = {
510 nfsv3err_null,
511 nfsv3err_getattr,
512 nfsv3err_setattr,
513 nfsv3err_lookup,
514 nfsv3err_access,
515 nfsv3err_readlink,
516 nfsv3err_read,
517 nfsv3err_write,
518 nfsv3err_create,
519 nfsv3err_mkdir,
520 nfsv3err_symlink,
521 nfsv3err_mknod,
522 nfsv3err_remove,
523 nfsv3err_rmdir,
524 nfsv3err_rename,
525 nfsv3err_link,
526 nfsv3err_readdir,
527 nfsv3err_readdirplus,
528 nfsv3err_fsstat,
529 nfsv3err_fsinfo,
530 nfsv3err_pathconf,
531 nfsv3err_commit,
532 };
533
534 extern struct nfsrtt nfsrtt;
535 extern time_t nqnfsstarttime;
536 extern int nqsrv_clockskew;
537 extern int nqsrv_writeslack;
538 extern int nqsrv_maxlease;
539 extern int nqnfs_piggy[NFS_NPROCS];
540 extern nfstype nfsv2_type[9];
541 extern nfstype nfsv3_type[9];
542 extern struct nfsnodehashhead *nfsnodehashtbl;
543 extern u_long nfsnodehash;
544
545 LIST_HEAD(nfsnodehashhead, nfsnode);
546 u_long nfsdirhashmask;
547
548 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
549
550 /*
551 * Create the header for an rpc request packet
552 * The hsiz is the size of the rest of the nfs request header.
553 * (just used to decide if a cluster is a good idea)
554 */
555 struct mbuf *
556 nfsm_reqh(vp, procid, hsiz, bposp)
557 struct vnode *vp;
558 u_long procid;
559 int hsiz;
560 caddr_t *bposp;
561 {
562 register struct mbuf *mb;
563 register u_int32_t *tl;
564 register caddr_t bpos;
565 struct mbuf *mb2;
566 struct nfsmount *nmp;
567 int nqflag;
568
569 MGET(mb, M_WAIT, MT_DATA);
570 if (hsiz >= MINCLSIZE)
571 MCLGET(mb, M_WAIT);
572 mb->m_len = 0;
573 bpos = mtod(mb, caddr_t);
574
575 /*
576 * For NQNFS, add lease request.
577 */
578 if (vp) {
579 nmp = VFSTONFS(vp->v_mount);
580 if (nmp->nm_flag & NFSMNT_NQNFS) {
581 nqflag = NQNFS_NEEDLEASE(vp, procid);
582 if (nqflag) {
583 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
584 *tl++ = txdr_unsigned(nqflag);
585 *tl = txdr_unsigned(nmp->nm_leaseterm);
586 } else {
587 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
588 *tl = 0;
589 }
590 }
591 }
592 /* Finally, return values */
593 *bposp = bpos;
594 return (mb);
595 }
596
597 /*
598 * Build the RPC header and fill in the authorization info.
599 * The authorization string argument is only used when the credentials
600 * come from outside of the kernel.
601 * Returns the head of the mbuf list.
602 */
603 struct mbuf *
604 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
605 verf_str, mrest, mrest_len, mbp, xidp)
606 register struct ucred *cr;
607 int nmflag;
608 int procid;
609 int auth_type;
610 int auth_len;
611 char *auth_str;
612 int verf_len;
613 char *verf_str;
614 struct mbuf *mrest;
615 int mrest_len;
616 struct mbuf **mbp;
617 u_int32_t *xidp;
618 {
619 register struct mbuf *mb;
620 register u_int32_t *tl;
621 register caddr_t bpos;
622 register int i;
623 struct mbuf *mreq, *mb2;
624 int siz, grpsiz, authsiz;
625 struct timeval tv;
626 static u_int32_t base;
627
628 authsiz = nfsm_rndup(auth_len);
629 MGETHDR(mb, M_WAIT, MT_DATA);
630 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
631 MCLGET(mb, M_WAIT);
632 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
633 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
634 } else {
635 MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
636 }
637 mb->m_len = 0;
638 mreq = mb;
639 bpos = mtod(mb, caddr_t);
640
641 /*
642 * First the RPC header.
643 */
644 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
645
646 /*
647 * derive initial xid from system time
648 * XXX time is invalid if root not yet mounted
649 */
650 if (!base && (rootvp)) {
651 microtime(&tv);
652 base = tv.tv_sec << 12;
653 nfs_xid = base;
654 }
655 /*
656 * Skip zero xid if it should ever happen.
657 */
658 if (++nfs_xid == 0)
659 nfs_xid++;
660
661 *tl++ = *xidp = txdr_unsigned(nfs_xid);
662 *tl++ = rpc_call;
663 *tl++ = rpc_vers;
664 if (nmflag & NFSMNT_NQNFS) {
665 *tl++ = txdr_unsigned(NQNFS_PROG);
666 *tl++ = txdr_unsigned(NQNFS_VER3);
667 } else {
668 *tl++ = txdr_unsigned(NFS_PROG);
669 if (nmflag & NFSMNT_NFSV3)
670 *tl++ = txdr_unsigned(NFS_VER3);
671 else
672 *tl++ = txdr_unsigned(NFS_VER2);
673 }
674 if (nmflag & NFSMNT_NFSV3)
675 *tl++ = txdr_unsigned(procid);
676 else
677 *tl++ = txdr_unsigned(nfsv2_procid[procid]);
678
679 /*
680 * And then the authorization cred.
681 */
682 *tl++ = txdr_unsigned(auth_type);
683 *tl = txdr_unsigned(authsiz);
684 switch (auth_type) {
685 case RPCAUTH_UNIX:
686 nfsm_build(tl, u_int32_t *, auth_len);
687 *tl++ = 0; /* stamp ?? */
688 *tl++ = 0; /* NULL hostname */
689 *tl++ = txdr_unsigned(cr->cr_uid);
690 *tl++ = txdr_unsigned(cr->cr_gid);
691 grpsiz = (auth_len >> 2) - 5;
692 *tl++ = txdr_unsigned(grpsiz);
693 for (i = 0; i < grpsiz; i++)
694 *tl++ = txdr_unsigned(cr->cr_groups[i]);
695 break;
696 case RPCAUTH_KERB4:
697 siz = auth_len;
698 while (siz > 0) {
699 if (M_TRAILINGSPACE(mb) == 0) {
700 MGET(mb2, M_WAIT, MT_DATA);
701 if (siz >= MINCLSIZE)
702 MCLGET(mb2, M_WAIT);
703 mb->m_next = mb2;
704 mb = mb2;
705 mb->m_len = 0;
706 bpos = mtod(mb, caddr_t);
707 }
708 i = min(siz, M_TRAILINGSPACE(mb));
709 bcopy(auth_str, bpos, i);
710 mb->m_len += i;
711 auth_str += i;
712 bpos += i;
713 siz -= i;
714 }
715 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
716 for (i = 0; i < siz; i++)
717 *bpos++ = '\0';
718 mb->m_len += siz;
719 }
720 break;
721 };
722
723 /*
724 * And the verifier...
725 */
726 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
727 if (verf_str) {
728 *tl++ = txdr_unsigned(RPCAUTH_KERB4);
729 *tl = txdr_unsigned(verf_len);
730 siz = verf_len;
731 while (siz > 0) {
732 if (M_TRAILINGSPACE(mb) == 0) {
733 MGET(mb2, M_WAIT, MT_DATA);
734 if (siz >= MINCLSIZE)
735 MCLGET(mb2, M_WAIT);
736 mb->m_next = mb2;
737 mb = mb2;
738 mb->m_len = 0;
739 bpos = mtod(mb, caddr_t);
740 }
741 i = min(siz, M_TRAILINGSPACE(mb));
742 bcopy(verf_str, bpos, i);
743 mb->m_len += i;
744 verf_str += i;
745 bpos += i;
746 siz -= i;
747 }
748 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
749 for (i = 0; i < siz; i++)
750 *bpos++ = '\0';
751 mb->m_len += siz;
752 }
753 } else {
754 *tl++ = txdr_unsigned(RPCAUTH_NULL);
755 *tl = 0;
756 }
757 mb->m_next = mrest;
758 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
759 mreq->m_pkthdr.rcvif = (struct ifnet *)0;
760 *mbp = mb;
761 return (mreq);
762 }
763
764 /*
765 * copies mbuf chain to the uio scatter/gather list
766 */
767 int
768 nfsm_mbuftouio(mrep, uiop, siz, dpos)
769 struct mbuf **mrep;
770 register struct uio *uiop;
771 int siz;
772 caddr_t *dpos;
773 {
774 register char *mbufcp, *uiocp;
775 register int xfer, left, len;
776 register struct mbuf *mp;
777 long uiosiz, rem;
778 int error = 0;
779
780 mp = *mrep;
781 mbufcp = *dpos;
782 len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
783 rem = nfsm_rndup(siz)-siz;
784 while (siz > 0) {
785 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
786 return (EFBIG);
787 left = uiop->uio_iov->iov_len;
788 uiocp = uiop->uio_iov->iov_base;
789 if (left > siz)
790 left = siz;
791 uiosiz = left;
792 while (left > 0) {
793 while (len == 0) {
794 mp = mp->m_next;
795 if (mp == NULL)
796 return (EBADRPC);
797 mbufcp = mtod(mp, caddr_t);
798 len = mp->m_len;
799 }
800 xfer = (left > len) ? len : left;
801 #ifdef notdef
802 /* Not Yet.. */
803 if (uiop->uio_iov->iov_op != NULL)
804 (*(uiop->uio_iov->iov_op))
805 (mbufcp, uiocp, xfer);
806 else
807 #endif
808 if (uiop->uio_segflg == UIO_SYSSPACE)
809 bcopy(mbufcp, uiocp, xfer);
810 else
811 copyout(mbufcp, uiocp, xfer);
812 left -= xfer;
813 len -= xfer;
814 mbufcp += xfer;
815 uiocp += xfer;
816 uiop->uio_offset += xfer;
817 uiop->uio_resid -= xfer;
818 }
819 if (uiop->uio_iov->iov_len <= siz) {
820 uiop->uio_iovcnt--;
821 uiop->uio_iov++;
822 } else {
823 uiop->uio_iov->iov_base += uiosiz;
824 uiop->uio_iov->iov_len -= uiosiz;
825 }
826 siz -= uiosiz;
827 }
828 *dpos = mbufcp;
829 *mrep = mp;
830 if (rem > 0) {
831 if (len < rem)
832 error = nfs_adv(mrep, dpos, rem, len);
833 else
834 *dpos += rem;
835 }
836 return (error);
837 }
838
839 /*
840 * copies a uio scatter/gather list to an mbuf chain.
841 * NOTE: can ony handle iovcnt == 1
842 */
843 int
844 nfsm_uiotombuf(uiop, mq, siz, bpos)
845 register struct uio *uiop;
846 struct mbuf **mq;
847 int siz;
848 caddr_t *bpos;
849 {
850 register char *uiocp;
851 register struct mbuf *mp, *mp2;
852 register int xfer, left, mlen;
853 int uiosiz, clflg, rem;
854 char *cp;
855
856 #ifdef DIAGNOSTIC
857 if (uiop->uio_iovcnt != 1)
858 panic("nfsm_uiotombuf: iovcnt != 1");
859 #endif
860
861 if (siz > MLEN) /* or should it >= MCLBYTES ?? */
862 clflg = 1;
863 else
864 clflg = 0;
865 rem = nfsm_rndup(siz)-siz;
866 mp = mp2 = *mq;
867 while (siz > 0) {
868 left = uiop->uio_iov->iov_len;
869 uiocp = uiop->uio_iov->iov_base;
870 if (left > siz)
871 left = siz;
872 uiosiz = left;
873 while (left > 0) {
874 mlen = M_TRAILINGSPACE(mp);
875 if (mlen == 0) {
876 MGET(mp, M_WAIT, MT_DATA);
877 if (clflg)
878 MCLGET(mp, M_WAIT);
879 mp->m_len = 0;
880 mp2->m_next = mp;
881 mp2 = mp;
882 mlen = M_TRAILINGSPACE(mp);
883 }
884 xfer = (left > mlen) ? mlen : left;
885 #ifdef notdef
886 /* Not Yet.. */
887 if (uiop->uio_iov->iov_op != NULL)
888 (*(uiop->uio_iov->iov_op))
889 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
890 else
891 #endif
892 if (uiop->uio_segflg == UIO_SYSSPACE)
893 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
894 else
895 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
896 mp->m_len += xfer;
897 left -= xfer;
898 uiocp += xfer;
899 uiop->uio_offset += xfer;
900 uiop->uio_resid -= xfer;
901 }
902 uiop->uio_iov->iov_base += uiosiz;
903 uiop->uio_iov->iov_len -= uiosiz;
904 siz -= uiosiz;
905 }
906 if (rem > 0) {
907 if (rem > M_TRAILINGSPACE(mp)) {
908 MGET(mp, M_WAIT, MT_DATA);
909 mp->m_len = 0;
910 mp2->m_next = mp;
911 }
912 cp = mtod(mp, caddr_t)+mp->m_len;
913 for (left = 0; left < rem; left++)
914 *cp++ = '\0';
915 mp->m_len += rem;
916 *bpos = cp;
917 } else
918 *bpos = mtod(mp, caddr_t)+mp->m_len;
919 *mq = mp;
920 return (0);
921 }
922
923 /*
924 * Get at least "siz" bytes of correctly aligned data.
925 * When called the mbuf pointers are not necessarily correct,
926 * dsosp points to what ought to be in m_data and left contains
927 * what ought to be in m_len.
928 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
929 * cases. (The macros use the vars. dpos and dpos2)
930 */
931 int
932 nfsm_disct(mdp, dposp, siz, left, cp2)
933 struct mbuf **mdp;
934 caddr_t *dposp;
935 int siz;
936 int left;
937 caddr_t *cp2;
938 {
939 register struct mbuf *m1, *m2;
940 struct mbuf *havebuf = NULL;
941 caddr_t src = *dposp;
942 caddr_t dst;
943 int len;
944
945 #ifdef DEBUG
946 if (left < 0)
947 panic("nfsm_disct: left < 0");
948 #endif
949 m1 = *mdp;
950 /*
951 * Skip through the mbuf chain looking for an mbuf with
952 * some data. If the first mbuf found has enough data
953 * and it is correctly aligned return it.
954 */
955 while (left == 0) {
956 havebuf = m1;
957 *mdp = m1 = m1->m_next;
958 if (m1 == NULL)
959 return (EBADRPC);
960 src = mtod(m1, caddr_t);
961 left = m1->m_len;
962 /*
963 * If we start a new mbuf and it is big enough
964 * and correctly aligned just return it, don't
965 * do any pull up.
966 */
967 if (left >= siz && nfsm_aligned(src)) {
968 *cp2 = src;
969 *dposp = src + siz;
970 return (0);
971 }
972 }
973 if (m1->m_flags & M_EXT) {
974 if (havebuf) {
975 /* If the first mbuf with data has external data
976 * and there is a previous empty mbuf use it
977 * to move the data into.
978 */
979 m2 = m1;
980 *mdp = m1 = havebuf;
981 if (m1->m_flags & M_EXT) {
982 MEXTREMOVE(m1);
983 }
984 } else {
985 /*
986 * If the first mbuf has a external data
987 * and there is no previous empty mbuf
988 * allocate a new mbuf and move the external
989 * data to the new mbuf. Also make the first
990 * mbuf look empty.
991 */
992 m2 = m_get(M_WAIT, MT_DATA);
993 m2->m_ext = m1->m_ext;
994 m2->m_data = src;
995 m2->m_len = left;
996 MCLADDREFERENCE(m1, m2);
997 MEXTREMOVE(m1);
998 m2->m_next = m1->m_next;
999 m1->m_next = m2;
1000 }
1001 m1->m_len = 0;
1002 dst = m1->m_dat;
1003 } else {
1004 /*
1005 * If the first mbuf has no external data
1006 * move the data to the front of the mbuf.
1007 */
1008 if ((dst = m1->m_dat) != src)
1009 ovbcopy(src, dst, left);
1010 dst += left;
1011 m1->m_len = left;
1012 m2 = m1->m_next;
1013 }
1014 m1->m_flags &= ~M_PKTHDR;
1015 *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */
1016 *dposp = mtod(m1, caddr_t) + siz;
1017 /*
1018 * Loop through mbufs pulling data up into first mbuf until
1019 * the first mbuf is full or there is no more data to
1020 * pullup.
1021 */
1022 while ((len = (MLEN - m1->m_len)) != 0 && m2) {
1023 if ((len = min(len, m2->m_len)) != 0)
1024 bcopy(m2->m_data, dst, len);
1025 m1->m_len += len;
1026 dst += len;
1027 m2->m_data += len;
1028 m2->m_len -= len;
1029 m2 = m2->m_next;
1030 }
1031 if (m1->m_len < siz)
1032 return (EBADRPC);
1033 return (0);
1034 }
1035
1036 /*
1037 * Advance the position in the mbuf chain.
1038 */
1039 int
1040 nfs_adv(mdp, dposp, offs, left)
1041 struct mbuf **mdp;
1042 caddr_t *dposp;
1043 int offs;
1044 int left;
1045 {
1046 register struct mbuf *m;
1047 register int s;
1048
1049 m = *mdp;
1050 s = left;
1051 while (s < offs) {
1052 offs -= s;
1053 m = m->m_next;
1054 if (m == NULL)
1055 return (EBADRPC);
1056 s = m->m_len;
1057 }
1058 *mdp = m;
1059 *dposp = mtod(m, caddr_t)+offs;
1060 return (0);
1061 }
1062
1063 /*
1064 * Copy a string into mbufs for the hard cases...
1065 */
1066 int
1067 nfsm_strtmbuf(mb, bpos, cp, siz)
1068 struct mbuf **mb;
1069 char **bpos;
1070 const char *cp;
1071 long siz;
1072 {
1073 register struct mbuf *m1 = NULL, *m2;
1074 long left, xfer, len, tlen;
1075 u_int32_t *tl;
1076 int putsize;
1077
1078 putsize = 1;
1079 m2 = *mb;
1080 left = M_TRAILINGSPACE(m2);
1081 if (left > 0) {
1082 tl = ((u_int32_t *)(*bpos));
1083 *tl++ = txdr_unsigned(siz);
1084 putsize = 0;
1085 left -= NFSX_UNSIGNED;
1086 m2->m_len += NFSX_UNSIGNED;
1087 if (left > 0) {
1088 bcopy(cp, (caddr_t) tl, left);
1089 siz -= left;
1090 cp += left;
1091 m2->m_len += left;
1092 left = 0;
1093 }
1094 }
1095 /* Loop around adding mbufs */
1096 while (siz > 0) {
1097 MGET(m1, M_WAIT, MT_DATA);
1098 if (siz > MLEN)
1099 MCLGET(m1, M_WAIT);
1100 m1->m_len = NFSMSIZ(m1);
1101 m2->m_next = m1;
1102 m2 = m1;
1103 tl = mtod(m1, u_int32_t *);
1104 tlen = 0;
1105 if (putsize) {
1106 *tl++ = txdr_unsigned(siz);
1107 m1->m_len -= NFSX_UNSIGNED;
1108 tlen = NFSX_UNSIGNED;
1109 putsize = 0;
1110 }
1111 if (siz < m1->m_len) {
1112 len = nfsm_rndup(siz);
1113 xfer = siz;
1114 if (xfer < len)
1115 *(tl+(xfer>>2)) = 0;
1116 } else {
1117 xfer = len = m1->m_len;
1118 }
1119 bcopy(cp, (caddr_t) tl, xfer);
1120 m1->m_len = len+tlen;
1121 siz -= xfer;
1122 cp += xfer;
1123 }
1124 *mb = m1;
1125 *bpos = mtod(m1, caddr_t)+m1->m_len;
1126 return (0);
1127 }
1128
1129 /*
1130 * Directory caching routines. They work as follows:
1131 * - a cache is maintained per VDIR nfsnode.
1132 * - for each offset cookie that is exported to userspace, and can
1133 * thus be thrown back at us as an offset to VOP_READDIR, store
1134 * information in the cache.
1135 * - cached are:
1136 * - cookie itself
1137 * - blocknumber (essentially just a search key in the buffer cache)
1138 * - entry number in block.
1139 * - offset cookie of block in which this entry is stored
1140 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used.
1141 * - entries are looked up in a hash table
1142 * - also maintained is an LRU list of entries, used to determine
1143 * which ones to delete if the cache grows too large.
1144 * - if 32 <-> 64 translation mode is requested for a filesystem,
1145 * the cache also functions as a translation table
1146 * - in the translation case, invalidating the cache does not mean
1147 * flushing it, but just marking entries as invalid, except for
1148 * the <64bit cookie, 32bitcookie> pair which is still valid, to
1149 * still be able to use the cache as a translation table.
1150 * - 32 bit cookies are uniquely created by combining the hash table
1151 * entry value, and one generation count per hash table entry,
1152 * incremented each time an entry is appended to the chain.
1153 * - the cache is invalidated each time a direcory is modified
1154 * - sanity checks are also done; if an entry in a block turns
1155 * out not to have a matching cookie, the cache is invalidated
1156 * and a new block starting from the wanted offset is fetched from
1157 * the server.
1158 * - directory entries as read from the server are extended to contain
1159 * the 64bit and, optionally, the 32bit cookies, for sanity checking
1160 * the cache and exporting them to userspace through the cookie
1161 * argument to VOP_READDIR.
1162 */
1163
1164 u_long
1165 nfs_dirhash(off)
1166 off_t off;
1167 {
1168 int i;
1169 char *cp = (char *)&off;
1170 u_long sum = 0L;
1171
1172 for (i = 0 ; i < sizeof (off); i++)
1173 sum += *cp++;
1174
1175 return sum;
1176 }
1177
1178 void
1179 nfs_initdircache(vp)
1180 struct vnode *vp;
1181 {
1182 struct nfsnode *np = VTONFS(vp);
1183 struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1184
1185 np->n_dircachesize = 0;
1186 np->n_dblkno = 1;
1187 np->n_dircache =
1188 hashinit(NFS_DIRHASHSIZ, M_NFSDIROFF, M_WAITOK, &nfsdirhashmask);
1189 TAILQ_INIT(&np->n_dirchain);
1190 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
1191 MALLOC(np->n_dirgens, unsigned *,
1192 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF,
1193 M_WAITOK);
1194 bzero((caddr_t)np->n_dirgens,
1195 NFS_DIRHASHSIZ * sizeof (unsigned));
1196 }
1197 }
1198
1199 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0};
1200
1201 struct nfsdircache *
1202 nfs_searchdircache(vp, off, do32, hashent)
1203 struct vnode *vp;
1204 off_t off;
1205 int do32;
1206 int *hashent;
1207 {
1208 struct nfsdirhashhead *ndhp;
1209 struct nfsdircache *ndp = NULL;
1210 struct nfsnode *np = VTONFS(vp);
1211 unsigned ent;
1212
1213 /*
1214 * Zero is always a valid cookie.
1215 */
1216 if (off == 0)
1217 return &dzero;
1218
1219 /*
1220 * We use a 32bit cookie as search key, directly reconstruct
1221 * the hashentry. Else use the hashfunction.
1222 */
1223 if (do32) {
1224 ent = (u_int32_t)off >> 24;
1225 if (ent >= NFS_DIRHASHSIZ)
1226 return NULL;
1227 ndhp = &np->n_dircache[ent];
1228 } else {
1229 ndhp = NFSDIRHASH(np, off);
1230 }
1231
1232 if (hashent)
1233 *hashent = (int)(ndhp - np->n_dircache);
1234 if (do32) {
1235 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) {
1236 if (ndp->dc_cookie32 == (u_int32_t)off) {
1237 /*
1238 * An invalidated entry will become the
1239 * start of a new block fetched from
1240 * the server.
1241 */
1242 if (ndp->dc_blkno == -1) {
1243 ndp->dc_blkcookie = ndp->dc_cookie;
1244 ndp->dc_blkno = np->n_dblkno++;
1245 ndp->dc_entry = 0;
1246 }
1247 break;
1248 }
1249 }
1250 } else {
1251 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next)
1252 if (ndp->dc_cookie == off)
1253 break;
1254 }
1255 return ndp;
1256 }
1257
1258
1259 struct nfsdircache *
1260 nfs_enterdircache(vp, off, blkoff, en, blkno)
1261 struct vnode *vp;
1262 off_t off, blkoff;
1263 daddr_t blkno;
1264 int en;
1265 {
1266 struct nfsnode *np = VTONFS(vp);
1267 struct nfsdirhashhead *ndhp;
1268 struct nfsdircache *ndp = NULL, *first;
1269 struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1270 int hashent, gen, overwrite;
1271
1272 if (!np->n_dircache)
1273 /*
1274 * XXX would like to do this in nfs_nget but vtype
1275 * isn't known at that time.
1276 */
1277 nfs_initdircache(vp);
1278
1279 /*
1280 * XXX refuse entries for offset 0. amd(8) erroneously sets
1281 * cookie 0 for the '.' entry, making this necessary. This
1282 * isn't so bad, as 0 is a special case anyway.
1283 */
1284 if (off == 0)
1285 return &dzero;
1286
1287 ndp = nfs_searchdircache(vp, off, 0, &hashent);
1288
1289 if (ndp && ndp->dc_blkno != -1) {
1290 /*
1291 * Overwriting an old entry. Check if it's the same.
1292 * If so, just return. If not, remove the old entry.
1293 */
1294 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en)
1295 return ndp;
1296 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
1297 LIST_REMOVE(ndp, dc_hash);
1298 FREE(ndp, M_NFSDIROFF);
1299 ndp = 0;
1300 }
1301
1302 ndhp = &np->n_dircache[hashent];
1303
1304 if (!ndp) {
1305 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF,
1306 M_WAITOK);
1307 overwrite = 0;
1308 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
1309 /*
1310 * We're allocating a new entry, so bump the
1311 * generation number.
1312 */
1313 gen = ++np->n_dirgens[hashent];
1314 if (gen == 0) {
1315 np->n_dirgens[hashent]++;
1316 gen++;
1317 }
1318 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff);
1319 }
1320 } else
1321 overwrite = 1;
1322
1323 /*
1324 * If the entry number is 0, we are at the start of a new block, so
1325 * allocate a new blocknumber.
1326 */
1327 if (en == 0)
1328 ndp->dc_blkno = np->n_dblkno++;
1329 else
1330 ndp->dc_blkno = blkno;
1331
1332 ndp->dc_cookie = off;
1333 ndp->dc_blkcookie = blkoff;
1334 ndp->dc_entry = en;
1335
1336 if (overwrite)
1337 return ndp;
1338
1339 /*
1340 * If the maximum directory cookie cache size has been reached
1341 * for this node, take one off the front. The idea is that
1342 * directories are typically read front-to-back once, so that
1343 * the oldest entries can be thrown away without much performance
1344 * loss.
1345 */
1346 if (np->n_dircachesize == NFS_MAXDIRCACHE) {
1347 first = np->n_dirchain.tqh_first;
1348 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain);
1349 LIST_REMOVE(first, dc_hash);
1350 FREE(first, M_NFSDIROFF);
1351 } else
1352 np->n_dircachesize++;
1353
1354 LIST_INSERT_HEAD(ndhp, ndp, dc_hash);
1355 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain);
1356 return ndp;
1357 }
1358
1359 void
1360 nfs_invaldircache(vp, forcefree)
1361 struct vnode *vp;
1362 int forcefree;
1363 {
1364 struct nfsnode *np = VTONFS(vp);
1365 struct nfsdircache *ndp = NULL;
1366 struct nfsmount *nmp = VFSTONFS(vp->v_mount);
1367
1368 #ifdef DIAGNOSTIC
1369 if (vp->v_type != VDIR)
1370 panic("nfs: invaldircache: not dir");
1371 #endif
1372
1373 if (!np->n_dircache)
1374 return;
1375
1376 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) {
1377 while ((ndp = np->n_dirchain.tqh_first)) {
1378 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
1379 LIST_REMOVE(ndp, dc_hash);
1380 FREE(ndp, M_NFSDIROFF);
1381 }
1382 np->n_dircachesize = 0;
1383 if (forcefree && np->n_dirgens) {
1384 FREE(np->n_dirgens, M_NFSDIROFF);
1385 }
1386 } else {
1387 for (ndp = np->n_dirchain.tqh_first; ndp;
1388 ndp = ndp->dc_chain.tqe_next)
1389 ndp->dc_blkno = -1;
1390 }
1391
1392 np->n_dblkno = 1;
1393 }
1394
1395 /*
1396 * Called once before VFS init to initialize shared and
1397 * server-specific data structures.
1398 */
1399 void
1400 nfs_init()
1401 {
1402
1403 #if !defined(alpha) && defined(DIAGNOSTIC)
1404 /*
1405 * Check to see if major data structures haven't bloated.
1406 */
1407 if (sizeof (struct nfsnode) > NFS_NODEALLOC) {
1408 printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC);
1409 printf("Try reducing NFS_SMALLFH\n");
1410 }
1411 if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) {
1412 printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC);
1413 printf("Try reducing NFS_UIDHASHSIZ\n");
1414 }
1415 if (sizeof (struct nfsuid) > NFS_UIDALLOC) {
1416 printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC);
1417 printf("Try unionizing the nu_nickname and nu_flag fields\n");
1418 }
1419 #endif
1420
1421 nfsrtt.pos = 0;
1422 rpc_vers = txdr_unsigned(RPC_VER2);
1423 rpc_call = txdr_unsigned(RPC_CALL);
1424 rpc_reply = txdr_unsigned(RPC_REPLY);
1425 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1426 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1427 rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1428 rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1429 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1430 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1431 nfs_prog = txdr_unsigned(NFS_PROG);
1432 nqnfs_prog = txdr_unsigned(NQNFS_PROG);
1433 nfs_true = txdr_unsigned(TRUE);
1434 nfs_false = txdr_unsigned(FALSE);
1435 nfs_xdrneg1 = txdr_unsigned(-1);
1436 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1437 if (nfs_ticks < 1)
1438 nfs_ticks = 1;
1439 #ifdef NFSSERVER
1440 nfsrv_init(0); /* Init server data structures */
1441 nfsrv_initcache(); /* Init the server request cache */
1442 #endif /* NFSSERVER */
1443
1444 /*
1445 * Initialize the nqnfs data structures.
1446 */
1447 if (nqnfsstarttime == 0) {
1448 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
1449 + nqsrv_clockskew + nqsrv_writeslack;
1450 NQLOADNOVRAM(nqnfsstarttime);
1451 CIRCLEQ_INIT(&nqtimerhead);
1452 nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, M_WAITOK, &nqfhhash);
1453 }
1454
1455 /*
1456 * Initialize reply list and start timer
1457 */
1458 TAILQ_INIT(&nfs_reqq);
1459 nfs_timer(NULL);
1460 }
1461
1462 #ifdef NFS
1463 /*
1464 * Called once at VFS init to initialize client-specific data structures.
1465 */
1466 void
1467 nfs_vfs_init()
1468 {
1469 register int i;
1470
1471 /* Ensure async daemons disabled */
1472 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1473 nfs_iodwant[i] = (struct proc *)0;
1474 nfs_iodmount[i] = (struct nfsmount *)0;
1475 }
1476 nfs_nhinit(); /* Init the nfsnode table */
1477 }
1478
1479 /*
1480 * Attribute cache routines.
1481 * nfs_loadattrcache() - loads or updates the cache contents from attributes
1482 * that are on the mbuf list
1483 * nfs_getattrcache() - returns valid attributes if found in cache, returns
1484 * error otherwise
1485 */
1486
1487 /*
1488 * Load the attribute cache (that lives in the nfsnode entry) with
1489 * the values on the mbuf list and
1490 * Iff vap not NULL
1491 * copy the attributes to *vaper
1492 */
1493 int
1494 nfsm_loadattrcache(vpp, mdp, dposp, vaper)
1495 struct vnode **vpp;
1496 struct mbuf **mdp;
1497 caddr_t *dposp;
1498 struct vattr *vaper;
1499 {
1500 register int32_t t1;
1501 caddr_t cp2;
1502 int error = 0;
1503 struct mbuf *md;
1504 int v3 = NFS_ISV3(*vpp);
1505
1506 md = *mdp;
1507 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1508 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2);
1509 if (error)
1510 return (error);
1511 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper);
1512 }
1513
1514 int
1515 nfs_loadattrcache(vpp, fp, vaper)
1516 struct vnode **vpp;
1517 struct nfs_fattr *fp;
1518 struct vattr *vaper;
1519 {
1520 register struct vnode *vp = *vpp;
1521 register struct vattr *vap;
1522 int v3 = NFS_ISV3(vp);
1523 enum vtype vtyp;
1524 u_short vmode;
1525 struct timespec mtime;
1526 struct vnode *nvp;
1527 int32_t rdev;
1528 register struct nfsnode *np;
1529 extern int (**spec_nfsv2nodeop_p) __P((void *));
1530
1531 if (v3) {
1532 vtyp = nfsv3tov_type(fp->fa_type);
1533 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1534 rdev = makedev(fxdr_unsigned(u_char, fp->fa3_rdev.specdata1),
1535 fxdr_unsigned(u_char, fp->fa3_rdev.specdata2));
1536 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1537 } else {
1538 vtyp = nfsv2tov_type(fp->fa_type);
1539 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1540 if (vtyp == VNON || vtyp == VREG)
1541 vtyp = IFTOVT(vmode);
1542 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1543 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1544
1545 /*
1546 * Really ugly NFSv2 kludge.
1547 */
1548 if (vtyp == VCHR && rdev == 0xffffffff)
1549 vtyp = VFIFO;
1550 }
1551
1552 /*
1553 * If v_type == VNON it is a new node, so fill in the v_type,
1554 * n_mtime fields. Check to see if it represents a special
1555 * device, and if so, check for a possible alias. Once the
1556 * correct vnode has been obtained, fill in the rest of the
1557 * information.
1558 */
1559 np = VTONFS(vp);
1560 if (vp->v_type != vtyp) {
1561 vp->v_type = vtyp;
1562 if (vp->v_type == VFIFO) {
1563 #ifndef FIFO
1564 return (EOPNOTSUPP);
1565 #else
1566 extern int (**fifo_nfsv2nodeop_p) __P((void *));
1567 vp->v_op = fifo_nfsv2nodeop_p;
1568 #endif /* FIFO */
1569 }
1570 if (vp->v_type == VCHR || vp->v_type == VBLK) {
1571 vp->v_op = spec_nfsv2nodeop_p;
1572 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount);
1573 if (nvp) {
1574 /*
1575 * Discard unneeded vnode, but save its nfsnode.
1576 * Since the nfsnode does not have a lock, its
1577 * vnode lock has to be carried over.
1578 */
1579 #ifdef Lite2_integrated
1580 nvp->v_vnlock = vp->v_vnlock;
1581 vp->v_vnlock = NULL;
1582 #endif
1583 nvp->v_data = vp->v_data;
1584 vp->v_data = NULL;
1585 vp->v_op = spec_vnodeop_p;
1586 vrele(vp);
1587 vgone(vp);
1588 /*
1589 * Reinitialize aliased node.
1590 */
1591 np->n_vnode = nvp;
1592 *vpp = vp = nvp;
1593 }
1594 }
1595 np->n_mtime = mtime.tv_sec;
1596 }
1597 vap = np->n_vattr;
1598 vap->va_type = vtyp;
1599 vap->va_mode = vmode & ALLPERMS;
1600 vap->va_rdev = (dev_t)rdev;
1601 vap->va_mtime = mtime;
1602 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1603 if (v3) {
1604 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1605 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1606 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1607 fxdr_hyper(&fp->fa3_size, &vap->va_size);
1608 if (vtyp == VDIR)
1609 vap->va_blocksize = NFS_DIRFRAGSIZ;
1610 else
1611 vap->va_blocksize = NFS_FABLKSIZE;
1612 fxdr_hyper(&fp->fa3_used, &vap->va_bytes);
1613 vap->va_fileid = fxdr_unsigned(int32_t,
1614 fp->fa3_fileid.nfsuquad[1]);
1615 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1616 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1617 vap->va_flags = 0;
1618 vap->va_filerev = 0;
1619 } else {
1620 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1621 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1622 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1623 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1624 if (vtyp == VDIR)
1625 vap->va_blocksize = NFS_DIRFRAGSIZ;
1626 else
1627 vap->va_blocksize =
1628 fxdr_unsigned(int32_t, fp->fa2_blocksize);
1629 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks)
1630 * NFS_FABLKSIZE;
1631 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1632 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1633 vap->va_flags = 0;
1634 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1635 fp->fa2_ctime.nfsv2_sec);
1636 vap->va_ctime.tv_nsec = 0;
1637 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1638 vap->va_filerev = 0;
1639 }
1640 if (vap->va_size != np->n_size) {
1641 if (vap->va_type == VREG) {
1642 if (np->n_flag & NMODIFIED) {
1643 if (vap->va_size < np->n_size)
1644 vap->va_size = np->n_size;
1645 else
1646 np->n_size = vap->va_size;
1647 } else
1648 np->n_size = vap->va_size;
1649 #if defined(UVM)
1650 uvm_vnp_setsize(vp, np->n_size);
1651 #else
1652 vnode_pager_setsize(vp, np->n_size);
1653 #endif
1654 } else
1655 np->n_size = vap->va_size;
1656 }
1657 np->n_attrstamp = time.tv_sec;
1658 if (vaper != NULL) {
1659 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1660 if (np->n_flag & NCHG) {
1661 if (np->n_flag & NACC)
1662 vaper->va_atime = np->n_atim;
1663 if (np->n_flag & NUPD)
1664 vaper->va_mtime = np->n_mtim;
1665 }
1666 }
1667 return (0);
1668 }
1669
1670 /*
1671 * Check the time stamp
1672 * If the cache is valid, copy contents to *vap and return 0
1673 * otherwise return an error
1674 */
1675 int
1676 nfs_getattrcache(vp, vaper)
1677 register struct vnode *vp;
1678 struct vattr *vaper;
1679 {
1680 register struct nfsnode *np = VTONFS(vp);
1681 register struct vattr *vap;
1682
1683 if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) {
1684 nfsstats.attrcache_misses++;
1685 return (ENOENT);
1686 }
1687 nfsstats.attrcache_hits++;
1688 vap = np->n_vattr;
1689 if (vap->va_size != np->n_size) {
1690 if (vap->va_type == VREG) {
1691 if (np->n_flag & NMODIFIED) {
1692 if (vap->va_size < np->n_size)
1693 vap->va_size = np->n_size;
1694 else
1695 np->n_size = vap->va_size;
1696 } else
1697 np->n_size = vap->va_size;
1698 #if defined(UVM)
1699 uvm_vnp_setsize(vp, np->n_size);
1700 #else
1701 vnode_pager_setsize(vp, np->n_size);
1702 #endif
1703 } else
1704 np->n_size = vap->va_size;
1705 }
1706 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1707 if (np->n_flag & NCHG) {
1708 if (np->n_flag & NACC)
1709 vaper->va_atime = np->n_atim;
1710 if (np->n_flag & NUPD)
1711 vaper->va_mtime = np->n_mtim;
1712 }
1713 return (0);
1714 }
1715
1716 /*
1717 * Heuristic to see if the server XDR encodes directory cookies or not.
1718 * it is not supposed to, but a lot of servers may do this. Also, since
1719 * most/all servers will implement V2 as well, it is expected that they
1720 * may return just 32 bits worth of cookie information, so we need to
1721 * find out in which 32 bits this information is available. We do this
1722 * to avoid trouble with emulated binaries that can't handle 64 bit
1723 * directory offsets.
1724 */
1725
1726 void
1727 nfs_cookieheuristic(vp, flagp, p, cred)
1728 struct vnode *vp;
1729 int *flagp;
1730 struct proc *p;
1731 struct ucred *cred;
1732 {
1733 struct uio auio;
1734 struct iovec aiov;
1735 caddr_t buf, cp;
1736 struct dirent *dp;
1737 off_t *cookies, *cop;
1738 int error, eof, nc, len;
1739
1740 nc = NFS_DIRFRAGSIZ / 16;
1741 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK);
1742 MALLOC(cookies, off_t *, nc * sizeof (off_t), M_TEMP, M_WAITOK);
1743
1744 aiov.iov_base = buf;
1745 aiov.iov_len = NFS_DIRFRAGSIZ;
1746 auio.uio_iov = &aiov;
1747 auio.uio_iovcnt = 1;
1748 auio.uio_rw = UIO_READ;
1749 auio.uio_segflg = UIO_SYSSPACE;
1750 auio.uio_procp = p;
1751 auio.uio_resid = NFS_DIRFRAGSIZ;
1752 auio.uio_offset = 0;
1753
1754 error = VOP_READDIR(vp, &auio, cred, &eof, cookies, nc);
1755
1756 len = NFS_DIRFRAGSIZ - auio.uio_resid;
1757 if (error || len == 0) {
1758 FREE(buf, M_TEMP);
1759 FREE(cookies, M_TEMP);
1760 return;
1761 }
1762
1763 /*
1764 * Find the first valid entry and look at its offset cookie.
1765 */
1766
1767 cp = buf;
1768 for (cop = cookies; len > 0; len -= dp->d_reclen) {
1769 dp = (struct dirent *)cp;
1770 if (dp->d_fileno != 0 && len >= dp->d_reclen) {
1771 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) {
1772 *flagp |= NFSMNT_SWAPCOOKIE;
1773 nfs_invaldircache(vp, 0);
1774 nfs_vinvalbuf(vp, 0, cred, p, 1);
1775 }
1776 break;
1777 }
1778 cop++;
1779 cp += dp->d_reclen;
1780 }
1781
1782 FREE(buf, M_TEMP);
1783 FREE(cookies, M_TEMP);
1784 }
1785 #endif /* NFS */
1786
1787 /*
1788 * Set up nameidata for a lookup() call and do it.
1789 *
1790 * If pubflag is set, this call is done for a lookup operation on the
1791 * public filehandle. In that case we allow crossing mountpoints and
1792 * absolute pathnames. However, the caller is expected to check that
1793 * the lookup result is within the public fs, and deny access if
1794 * it is not.
1795 */
1796 int
1797 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
1798 register struct nameidata *ndp;
1799 fhandle_t *fhp;
1800 int len;
1801 struct nfssvc_sock *slp;
1802 struct mbuf *nam;
1803 struct mbuf **mdp;
1804 caddr_t *dposp;
1805 struct vnode **retdirp;
1806 struct proc *p;
1807 int kerbflag, pubflag;
1808 {
1809 register int i, rem;
1810 register struct mbuf *md;
1811 register char *fromcp, *tocp, *cp;
1812 struct iovec aiov;
1813 struct uio auio;
1814 struct vnode *dp;
1815 int error, rdonly, linklen;
1816 struct componentname *cnp = &ndp->ni_cnd;
1817
1818 *retdirp = (struct vnode *)0;
1819 MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK);
1820 /*
1821 * Copy the name from the mbuf list to ndp->ni_pnbuf
1822 * and set the various ndp fields appropriately.
1823 */
1824 fromcp = *dposp;
1825 tocp = cnp->cn_pnbuf;
1826 md = *mdp;
1827 rem = mtod(md, caddr_t) + md->m_len - fromcp;
1828 for (i = 0; i < len; i++) {
1829 while (rem == 0) {
1830 md = md->m_next;
1831 if (md == NULL) {
1832 error = EBADRPC;
1833 goto out;
1834 }
1835 fromcp = mtod(md, caddr_t);
1836 rem = md->m_len;
1837 }
1838 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1839 error = EACCES;
1840 goto out;
1841 }
1842 *tocp++ = *fromcp++;
1843 rem--;
1844 }
1845 *tocp = '\0';
1846 *mdp = md;
1847 *dposp = fromcp;
1848 len = nfsm_rndup(len)-len;
1849 if (len > 0) {
1850 if (rem >= len)
1851 *dposp += len;
1852 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1853 goto out;
1854 }
1855
1856 /*
1857 * Extract and set starting directory.
1858 */
1859 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
1860 nam, &rdonly, kerbflag, pubflag);
1861 if (error)
1862 goto out;
1863 if (dp->v_type != VDIR) {
1864 vrele(dp);
1865 error = ENOTDIR;
1866 goto out;
1867 }
1868
1869 if (rdonly)
1870 cnp->cn_flags |= RDONLY;
1871
1872 *retdirp = dp;
1873
1874 if (pubflag) {
1875 /*
1876 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1877 * and the 'native path' indicator.
1878 */
1879 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1880 fromcp = cnp->cn_pnbuf;
1881 tocp = cp;
1882 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1883 switch ((unsigned char)*fromcp) {
1884 case WEBNFS_NATIVE_CHAR:
1885 /*
1886 * 'Native' path for us is the same
1887 * as a path according to the NFS spec,
1888 * just skip the escape char.
1889 */
1890 fromcp++;
1891 break;
1892 /*
1893 * More may be added in the future, range 0x80-0xff
1894 */
1895 default:
1896 error = EIO;
1897 FREE(cp, M_NAMEI);
1898 goto out;
1899 }
1900 }
1901 /*
1902 * Translate the '%' escapes, URL-style.
1903 */
1904 while (*fromcp != '\0') {
1905 if (*fromcp == WEBNFS_ESC_CHAR) {
1906 if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1907 fromcp++;
1908 *tocp++ = HEXSTRTOI(fromcp);
1909 fromcp += 2;
1910 continue;
1911 } else {
1912 error = ENOENT;
1913 FREE(cp, M_NAMEI);
1914 goto out;
1915 }
1916 } else
1917 *tocp++ = *fromcp++;
1918 }
1919 *tocp = '\0';
1920 FREE(cnp->cn_pnbuf, M_NAMEI);
1921 cnp->cn_pnbuf = cp;
1922 }
1923
1924 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
1925 ndp->ni_segflg = UIO_SYSSPACE;
1926
1927 if (pubflag) {
1928 ndp->ni_rootdir = rootvnode;
1929 ndp->ni_loopcnt = 0;
1930 if (cnp->cn_pnbuf[0] == '/')
1931 dp = rootvnode;
1932 } else {
1933 cnp->cn_flags |= NOCROSSMOUNT;
1934 }
1935
1936 cnp->cn_proc = p;
1937 VREF(dp);
1938
1939 for (;;) {
1940 cnp->cn_nameptr = cnp->cn_pnbuf;
1941 ndp->ni_startdir = dp;
1942 /*
1943 * And call lookup() to do the real work
1944 */
1945 error = lookup(ndp);
1946 if (error)
1947 break;
1948 /*
1949 * Check for encountering a symbolic link
1950 */
1951 if ((cnp->cn_flags & ISSYMLINK) == 0) {
1952 if (cnp->cn_flags & (SAVENAME | SAVESTART)) {
1953 cnp->cn_flags |= HASBUF;
1954 return (0);
1955 }
1956 break;
1957 } else {
1958 if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1)
1959 VOP_UNLOCK(ndp->ni_dvp);
1960 if (!pubflag) {
1961 vrele(ndp->ni_dvp);
1962 vput(ndp->ni_vp);
1963 ndp->ni_vp = NULL;
1964 error = EINVAL;
1965 break;
1966 }
1967
1968 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
1969 error = ELOOP;
1970 break;
1971 }
1972 if (ndp->ni_pathlen > 1)
1973 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
1974 else
1975 cp = cnp->cn_pnbuf;
1976 aiov.iov_base = cp;
1977 aiov.iov_len = MAXPATHLEN;
1978 auio.uio_iov = &aiov;
1979 auio.uio_iovcnt = 1;
1980 auio.uio_offset = 0;
1981 auio.uio_rw = UIO_READ;
1982 auio.uio_segflg = UIO_SYSSPACE;
1983 auio.uio_procp = (struct proc *)0;
1984 auio.uio_resid = MAXPATHLEN;
1985 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
1986 if (error) {
1987 badlink:
1988 if (ndp->ni_pathlen > 1)
1989 FREE(cp, M_NAMEI);
1990 break;
1991 }
1992 linklen = MAXPATHLEN - auio.uio_resid;
1993 if (linklen == 0) {
1994 error = ENOENT;
1995 goto badlink;
1996 }
1997 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
1998 error = ENAMETOOLONG;
1999 goto badlink;
2000 }
2001 if (ndp->ni_pathlen > 1) {
2002 bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
2003 FREE(cnp->cn_pnbuf, M_NAMEI);
2004 cnp->cn_pnbuf = cp;
2005 } else
2006 cnp->cn_pnbuf[linklen] = '\0';
2007 ndp->ni_pathlen += linklen;
2008 vput(ndp->ni_vp);
2009 dp = ndp->ni_dvp;
2010 /*
2011 * Check if root directory should replace current directory.
2012 */
2013 if (cnp->cn_pnbuf[0] == '/') {
2014 vrele(dp);
2015 dp = ndp->ni_rootdir;
2016 VREF(dp);
2017 }
2018 }
2019 }
2020 out:
2021 FREE(cnp->cn_pnbuf, M_NAMEI);
2022 return (error);
2023 }
2024
2025 /*
2026 * A fiddled version of m_adj() that ensures null fill to a long
2027 * boundary and only trims off the back end
2028 */
2029 void
2030 nfsm_adj(mp, len, nul)
2031 struct mbuf *mp;
2032 register int len;
2033 int nul;
2034 {
2035 register struct mbuf *m;
2036 register int count, i;
2037 register char *cp;
2038
2039 /*
2040 * Trim from tail. Scan the mbuf chain,
2041 * calculating its length and finding the last mbuf.
2042 * If the adjustment only affects this mbuf, then just
2043 * adjust and return. Otherwise, rescan and truncate
2044 * after the remaining size.
2045 */
2046 count = 0;
2047 m = mp;
2048 for (;;) {
2049 count += m->m_len;
2050 if (m->m_next == (struct mbuf *)0)
2051 break;
2052 m = m->m_next;
2053 }
2054 if (m->m_len > len) {
2055 m->m_len -= len;
2056 if (nul > 0) {
2057 cp = mtod(m, caddr_t)+m->m_len-nul;
2058 for (i = 0; i < nul; i++)
2059 *cp++ = '\0';
2060 }
2061 return;
2062 }
2063 count -= len;
2064 if (count < 0)
2065 count = 0;
2066 /*
2067 * Correct length for chain is "count".
2068 * Find the mbuf with last data, adjust its length,
2069 * and toss data from remaining mbufs on chain.
2070 */
2071 for (m = mp; m; m = m->m_next) {
2072 if (m->m_len >= count) {
2073 m->m_len = count;
2074 if (nul > 0) {
2075 cp = mtod(m, caddr_t)+m->m_len-nul;
2076 for (i = 0; i < nul; i++)
2077 *cp++ = '\0';
2078 }
2079 break;
2080 }
2081 count -= m->m_len;
2082 }
2083 for (m = m->m_next;m;m = m->m_next)
2084 m->m_len = 0;
2085 }
2086
2087 /*
2088 * Make these functions instead of macros, so that the kernel text size
2089 * doesn't get too big...
2090 */
2091 void
2092 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
2093 struct nfsrv_descript *nfsd;
2094 int before_ret;
2095 register struct vattr *before_vap;
2096 int after_ret;
2097 struct vattr *after_vap;
2098 struct mbuf **mbp;
2099 char **bposp;
2100 {
2101 register struct mbuf *mb = *mbp, *mb2;
2102 register char *bpos = *bposp;
2103 register u_int32_t *tl;
2104
2105 if (before_ret) {
2106 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
2107 *tl = nfs_false;
2108 } else {
2109 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
2110 *tl++ = nfs_true;
2111 txdr_hyper(&(before_vap->va_size), tl);
2112 tl += 2;
2113 txdr_nfsv3time(&(before_vap->va_mtime), tl);
2114 tl += 2;
2115 txdr_nfsv3time(&(before_vap->va_ctime), tl);
2116 }
2117 *bposp = bpos;
2118 *mbp = mb;
2119 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
2120 }
2121
2122 void
2123 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
2124 struct nfsrv_descript *nfsd;
2125 int after_ret;
2126 struct vattr *after_vap;
2127 struct mbuf **mbp;
2128 char **bposp;
2129 {
2130 register struct mbuf *mb = *mbp, *mb2;
2131 register char *bpos = *bposp;
2132 register u_int32_t *tl;
2133 register struct nfs_fattr *fp;
2134
2135 if (after_ret) {
2136 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
2137 *tl = nfs_false;
2138 } else {
2139 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
2140 *tl++ = nfs_true;
2141 fp = (struct nfs_fattr *)tl;
2142 nfsm_srvfattr(nfsd, after_vap, fp);
2143 }
2144 *mbp = mb;
2145 *bposp = bpos;
2146 }
2147
2148 void
2149 nfsm_srvfattr(nfsd, vap, fp)
2150 register struct nfsrv_descript *nfsd;
2151 register struct vattr *vap;
2152 register struct nfs_fattr *fp;
2153 {
2154
2155 fp->fa_nlink = txdr_unsigned(vap->va_nlink);
2156 fp->fa_uid = txdr_unsigned(vap->va_uid);
2157 fp->fa_gid = txdr_unsigned(vap->va_gid);
2158 if (nfsd->nd_flag & ND_NFSV3) {
2159 fp->fa_type = vtonfsv3_type(vap->va_type);
2160 fp->fa_mode = vtonfsv3_mode(vap->va_mode);
2161 txdr_hyper(&vap->va_size, &fp->fa3_size);
2162 txdr_hyper(&vap->va_bytes, &fp->fa3_used);
2163 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev));
2164 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev));
2165 fp->fa3_fsid.nfsuquad[0] = 0;
2166 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
2167 fp->fa3_fileid.nfsuquad[0] = 0;
2168 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
2169 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
2170 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
2171 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
2172 } else {
2173 fp->fa_type = vtonfsv2_type(vap->va_type);
2174 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
2175 fp->fa2_size = txdr_unsigned(vap->va_size);
2176 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
2177 if (vap->va_type == VFIFO)
2178 fp->fa2_rdev = 0xffffffff;
2179 else
2180 fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
2181 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
2182 fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
2183 fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
2184 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
2185 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
2186 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
2187 }
2188 }
2189
2190 /*
2191 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
2192 * - look up fsid in mount list (if not found ret error)
2193 * - get vp and export rights by calling VFS_FHTOVP()
2194 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
2195 * - if not lockflag unlock it with VOP_UNLOCK()
2196 */
2197 int
2198 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
2199 fhandle_t *fhp;
2200 int lockflag;
2201 struct vnode **vpp;
2202 struct ucred *cred;
2203 struct nfssvc_sock *slp;
2204 struct mbuf *nam;
2205 int *rdonlyp;
2206 int kerbflag;
2207 {
2208 #ifdef Lite2_integrated
2209 struct proc *p = curproc; /* XXX */
2210 #endif
2211 register struct mount *mp;
2212 register int i;
2213 struct ucred *credanon;
2214 int error, exflags;
2215 struct sockaddr_in *saddr;
2216
2217 *vpp = (struct vnode *)0;
2218
2219 if (nfs_ispublicfh(fhp)) {
2220 if (!pubflag || !nfs_pub.np_valid)
2221 return (ESTALE);
2222 fhp = &nfs_pub.np_handle;
2223 }
2224
2225 #ifdef Lite2_integrated
2226 mp = vfs_getvfs(&fhp->fh_fsid);
2227 #else
2228 mp = getvfs(&fhp->fh_fsid);
2229 #endif
2230 if (!mp)
2231 return (ESTALE);
2232 error = VFS_FHTOVP(mp, &fhp->fh_fid, nam, vpp, &exflags, &credanon);
2233 if (error)
2234 return (error);
2235
2236 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
2237 saddr = mtod(nam, struct sockaddr_in *);
2238 if (saddr->sin_family == AF_INET &&
2239 ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
2240 vput(*vpp);
2241 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
2242 }
2243 }
2244 /*
2245 * Check/setup credentials.
2246 */
2247 if (exflags & MNT_EXKERB) {
2248 if (!kerbflag) {
2249 vput(*vpp);
2250 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
2251 }
2252 } else if (kerbflag) {
2253 vput(*vpp);
2254 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
2255 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
2256 cred->cr_uid = credanon->cr_uid;
2257 cred->cr_gid = credanon->cr_gid;
2258 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
2259 cred->cr_groups[i] = credanon->cr_groups[i];
2260 cred->cr_ngroups = i;
2261 }
2262 if (exflags & MNT_EXRDONLY)
2263 *rdonlyp = 1;
2264 else
2265 *rdonlyp = 0;
2266 if (!lockflag)
2267 #ifdef Lite2_integrated
2268 VOP_UNLOCK(*vpp, 0, p);
2269 #else
2270 VOP_UNLOCK(*vpp);
2271 #endif
2272 return (0);
2273 }
2274
2275 /*
2276 * WebNFS: check if a filehandle is a public filehandle. For v3, this
2277 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
2278 * transformed this to all zeroes in both cases, so check for it.
2279 */
2280 int
2281 nfs_ispublicfh(fhp)
2282 fhandle_t *fhp;
2283 {
2284 char *cp = (char *)fhp;
2285 int i;
2286
2287 for (i = 0; i < NFSX_V3FH; i++)
2288 if (*cp++ != 0)
2289 return (FALSE);
2290 return (TRUE);
2291 }
2292
2293 /*
2294 * This function compares two net addresses by family and returns TRUE
2295 * if they are the same host.
2296 * If there is any doubt, return FALSE.
2297 * The AF_INET family is handled as a special case so that address mbufs
2298 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
2299 */
2300 int
2301 netaddr_match(family, haddr, nam)
2302 int family;
2303 union nethostaddr *haddr;
2304 struct mbuf *nam;
2305 {
2306 register struct sockaddr_in *inetaddr;
2307
2308 switch (family) {
2309 case AF_INET:
2310 inetaddr = mtod(nam, struct sockaddr_in *);
2311 if (inetaddr->sin_family == AF_INET &&
2312 inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
2313 return (1);
2314 break;
2315 #ifdef ISO
2316 case AF_ISO:
2317 {
2318 register struct sockaddr_iso *isoaddr1, *isoaddr2;
2319
2320 isoaddr1 = mtod(nam, struct sockaddr_iso *);
2321 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *);
2322 if (isoaddr1->siso_family == AF_ISO &&
2323 isoaddr1->siso_nlen > 0 &&
2324 isoaddr1->siso_nlen == isoaddr2->siso_nlen &&
2325 SAME_ISOADDR(isoaddr1, isoaddr2))
2326 return (1);
2327 break;
2328 }
2329 #endif /* ISO */
2330 default:
2331 break;
2332 };
2333 return (0);
2334 }
2335
2336
2337 /*
2338 * The write verifier has changed (probably due to a server reboot), so all
2339 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2340 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2341 * flag. Once done the new write verifier can be set for the mount point.
2342 */
2343 void
2344 nfs_clearcommit(mp)
2345 struct mount *mp;
2346 {
2347 register struct vnode *vp, *nvp;
2348 register struct buf *bp, *nbp;
2349 int s;
2350
2351 s = splbio();
2352 loop:
2353 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) {
2354 if (vp->v_mount != mp) /* Paranoia */
2355 goto loop;
2356 nvp = vp->v_mntvnodes.le_next;
2357 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) {
2358 nbp = bp->b_vnbufs.le_next;
2359 if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT))
2360 == (B_DELWRI | B_NEEDCOMMIT))
2361 bp->b_flags &= ~B_NEEDCOMMIT;
2362 }
2363 }
2364 splx(s);
2365 }
2366
2367 /*
2368 * Map errnos to NFS error numbers. For Version 3 also filter out error
2369 * numbers not specified for the associated procedure.
2370 */
2371 int
2372 nfsrv_errmap(nd, err)
2373 struct nfsrv_descript *nd;
2374 register int err;
2375 {
2376 register short *defaulterrp, *errp;
2377
2378 if (nd->nd_flag & ND_NFSV3) {
2379 if (nd->nd_procnum <= NFSPROC_COMMIT) {
2380 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2381 while (*++errp) {
2382 if (*errp == err)
2383 return (err);
2384 else if (*errp > err)
2385 break;
2386 }
2387 return ((int)*defaulterrp);
2388 } else
2389 return (err & 0xffff);
2390 }
2391 if (err <= ELAST)
2392 return ((int)nfsrv_v2errmap[err - 1]);
2393 return (NFSERR_IO);
2394 }
2395
2396 /*
2397 * Sort the group list in increasing numerical order.
2398 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2399 * that used to be here.)
2400 */
2401 void
2402 nfsrvw_sort(list, num)
2403 register gid_t *list;
2404 register int num;
2405 {
2406 register int i, j;
2407 gid_t v;
2408
2409 /* Insertion sort. */
2410 for (i = 1; i < num; i++) {
2411 v = list[i];
2412 /* find correct slot for value v, moving others up */
2413 for (j = i; --j >= 0 && v < list[j];)
2414 list[j + 1] = list[j];
2415 list[j + 1] = v;
2416 }
2417 }
2418
2419 /*
2420 * copy credentials making sure that the result can be compared with bcmp().
2421 */
2422 void
2423 nfsrv_setcred(incred, outcred)
2424 register struct ucred *incred, *outcred;
2425 {
2426 register int i;
2427
2428 bzero((caddr_t)outcred, sizeof (struct ucred));
2429 outcred->cr_ref = 1;
2430 outcred->cr_uid = incred->cr_uid;
2431 outcred->cr_gid = incred->cr_gid;
2432 outcred->cr_ngroups = incred->cr_ngroups;
2433 for (i = 0; i < incred->cr_ngroups; i++)
2434 outcred->cr_groups[i] = incred->cr_groups[i];
2435 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2436 }
2437