NFS - Fix numerous issues with NFS root mounts, primarily for vkernels
[dragonfly.git] / sys / vfs / nfs / nfs_subs.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Rick Macklem at The University of Guelph.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95
79e5012e 37 * $FreeBSD: /repoman/r/ncvs/src/sys/nfsclient/nfs_subs.c,v 1.128 2004/04/14 23:23:55 peadar Exp $
67863d04 38 * $DragonFly: src/sys/vfs/nfs/nfs_subs.c,v 1.48 2008/09/17 21:44:24 dillon Exp $
984263bc
MD
39 */
40
41/*
42 * These functions support the macros and help fiddle mbuf chains for
43 * the nfs op functions. They do things like create the rpc header and
44 * copy data between mbuf chains and uio lists.
45 */
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/kernel.h>
49#include <sys/buf.h>
50#include <sys/proc.h>
51#include <sys/mount.h>
52#include <sys/vnode.h>
fad57d0e 53#include <sys/nlookup.h>
984263bc
MD
54#include <sys/namei.h>
55#include <sys/mbuf.h>
56#include <sys/socket.h>
57#include <sys/stat.h>
58#include <sys/malloc.h>
59#include <sys/sysent.h>
60#include <sys/syscall.h>
61#include <sys/conf.h>
70aac194 62#include <sys/objcache.h>
984263bc
MD
63
64#include <vm/vm.h>
65#include <vm/vm_object.h>
66#include <vm/vm_extern.h>
67#include <vm/vm_zone.h>
68
3020e3be
MD
69#include <sys/buf2.h>
70
1f2de5d4
MD
71#include "rpcv2.h"
72#include "nfsproto.h"
73#include "nfs.h"
c1cf1e59 74#include "nfsmount.h"
1f2de5d4
MD
75#include "nfsnode.h"
76#include "xdr_subs.h"
77#include "nfsm_subs.h"
1f2de5d4 78#include "nfsrtt.h"
984263bc
MD
79
80#include <netinet/in.h>
81
82/*
83 * Data items converted to xdr at startup, since they are constant
84 * This is kinda hokey, but may save a little time doing byte swaps
85 */
86u_int32_t nfs_xdrneg1;
e97453f3
MD
87u_int32_t rpc_reply, rpc_msgdenied, rpc_mismatch, rpc_vers;
88u_int32_t rpc_auth_unix, rpc_msgaccepted, rpc_call, rpc_autherr;
89u_int32_t rpc_auth_kerb;
e07fef60 90u_int32_t nfs_prog, nfs_true, nfs_false;
984263bc
MD
91
92/* And other global data */
984263bc
MD
93static enum vtype nv2tov_type[8]= {
94 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON
95};
96enum vtype nv3tov_type[8]= {
97 VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
98};
99
100int nfs_ticks;
e97453f3
MD
101
102static int nfs_pbuf_freecnt = -1; /* start out unlimited */
984263bc 103
49433307 104struct nfsmount_head nfs_mountq = TAILQ_HEAD_INITIALIZER(nfs_mountq);
984263bc
MD
105struct nfssvc_sockhead nfssvc_sockhead;
106int nfssvc_sockhead_flag;
107struct nfsd_head nfsd_head;
108int nfsd_head_flag;
109struct nfs_bufq nfs_bufq;
984263bc
MD
110struct nqfhhashhead *nqfhhashtbl;
111u_long nqfhhash;
112
984263bc
MD
113static int nfs_prev_nfssvc_sy_narg;
114static sy_call_t *nfs_prev_nfssvc_sy_call;
115
116#ifndef NFS_NOSERVER
117
984263bc
MD
118/*
119 * Mapping of old NFS Version 2 RPC numbers to generic numbers.
120 */
121int nfsv3_procid[NFS_NPROCS] = {
122 NFSPROC_NULL,
123 NFSPROC_GETATTR,
124 NFSPROC_SETATTR,
125 NFSPROC_NOOP,
126 NFSPROC_LOOKUP,
127 NFSPROC_READLINK,
128 NFSPROC_READ,
129 NFSPROC_NOOP,
130 NFSPROC_WRITE,
131 NFSPROC_CREATE,
132 NFSPROC_REMOVE,
133 NFSPROC_RENAME,
134 NFSPROC_LINK,
135 NFSPROC_SYMLINK,
136 NFSPROC_MKDIR,
137 NFSPROC_RMDIR,
138 NFSPROC_READDIR,
139 NFSPROC_FSSTAT,
140 NFSPROC_NOOP,
141 NFSPROC_NOOP,
142 NFSPROC_NOOP,
143 NFSPROC_NOOP,
144 NFSPROC_NOOP,
145 NFSPROC_NOOP,
146 NFSPROC_NOOP,
147 NFSPROC_NOOP
148};
149
150#endif /* NFS_NOSERVER */
151/*
152 * and the reverse mapping from generic to Version 2 procedure numbers
153 */
154int nfsv2_procid[NFS_NPROCS] = {
155 NFSV2PROC_NULL,
156 NFSV2PROC_GETATTR,
157 NFSV2PROC_SETATTR,
158 NFSV2PROC_LOOKUP,
159 NFSV2PROC_NOOP,
160 NFSV2PROC_READLINK,
161 NFSV2PROC_READ,
162 NFSV2PROC_WRITE,
163 NFSV2PROC_CREATE,
164 NFSV2PROC_MKDIR,
165 NFSV2PROC_SYMLINK,
166 NFSV2PROC_CREATE,
167 NFSV2PROC_REMOVE,
168 NFSV2PROC_RMDIR,
169 NFSV2PROC_RENAME,
170 NFSV2PROC_LINK,
171 NFSV2PROC_READDIR,
172 NFSV2PROC_NOOP,
173 NFSV2PROC_STATFS,
174 NFSV2PROC_NOOP,
175 NFSV2PROC_NOOP,
176 NFSV2PROC_NOOP,
177 NFSV2PROC_NOOP,
178 NFSV2PROC_NOOP,
179 NFSV2PROC_NOOP,
180 NFSV2PROC_NOOP,
181};
182
183#ifndef NFS_NOSERVER
184/*
185 * Maps errno values to nfs error numbers.
186 * Use NFSERR_IO as the catch all for ones not specifically defined in
187 * RFC 1094.
188 */
189static u_char nfsrv_v2errmap[ELAST] = {
190 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO,
191 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
192 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO,
193 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR,
194 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
195 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS,
196 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
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, NFSERR_IO, NFSERR_IO, NFSERR_IO,
200 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
201 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
202 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO,
203 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE,
204 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
205 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
206 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO,
207 NFSERR_IO /* << Last is 86 */
208};
209
210/*
211 * Maps errno values to nfs error numbers.
212 * Although it is not obvious whether or not NFS clients really care if
213 * a returned error value is in the specified list for the procedure, the
214 * safest thing to do is filter them appropriately. For Version 2, the
215 * X/Open XNFS document is the only specification that defines error values
216 * for each RPC (The RFC simply lists all possible error values for all RPCs),
217 * so I have decided to not do this for Version 2.
218 * The first entry is the default error return and the rest are the valid
219 * errors for that RPC in increasing numeric order.
220 */
221static short nfsv3err_null[] = {
222 0,
223 0,
224};
225
226static short nfsv3err_getattr[] = {
227 NFSERR_IO,
228 NFSERR_IO,
229 NFSERR_STALE,
230 NFSERR_BADHANDLE,
231 NFSERR_SERVERFAULT,
232 0,
233};
234
235static short nfsv3err_setattr[] = {
236 NFSERR_IO,
237 NFSERR_PERM,
238 NFSERR_IO,
239 NFSERR_ACCES,
240 NFSERR_INVAL,
241 NFSERR_NOSPC,
242 NFSERR_ROFS,
243 NFSERR_DQUOT,
244 NFSERR_STALE,
245 NFSERR_BADHANDLE,
246 NFSERR_NOT_SYNC,
247 NFSERR_SERVERFAULT,
248 0,
249};
250
251static short nfsv3err_lookup[] = {
252 NFSERR_IO,
253 NFSERR_NOENT,
254 NFSERR_IO,
255 NFSERR_ACCES,
256 NFSERR_NOTDIR,
257 NFSERR_NAMETOL,
258 NFSERR_STALE,
259 NFSERR_BADHANDLE,
260 NFSERR_SERVERFAULT,
261 0,
262};
263
264static short nfsv3err_access[] = {
265 NFSERR_IO,
266 NFSERR_IO,
267 NFSERR_STALE,
268 NFSERR_BADHANDLE,
269 NFSERR_SERVERFAULT,
270 0,
271};
272
273static short nfsv3err_readlink[] = {
274 NFSERR_IO,
275 NFSERR_IO,
276 NFSERR_ACCES,
277 NFSERR_INVAL,
278 NFSERR_STALE,
279 NFSERR_BADHANDLE,
280 NFSERR_NOTSUPP,
281 NFSERR_SERVERFAULT,
282 0,
283};
284
285static short nfsv3err_read[] = {
286 NFSERR_IO,
287 NFSERR_IO,
288 NFSERR_NXIO,
289 NFSERR_ACCES,
290 NFSERR_INVAL,
291 NFSERR_STALE,
292 NFSERR_BADHANDLE,
293 NFSERR_SERVERFAULT,
294 0,
295};
296
297static short nfsv3err_write[] = {
298 NFSERR_IO,
299 NFSERR_IO,
300 NFSERR_ACCES,
301 NFSERR_INVAL,
302 NFSERR_FBIG,
303 NFSERR_NOSPC,
304 NFSERR_ROFS,
305 NFSERR_DQUOT,
306 NFSERR_STALE,
307 NFSERR_BADHANDLE,
308 NFSERR_SERVERFAULT,
309 0,
310};
311
312static short nfsv3err_create[] = {
313 NFSERR_IO,
314 NFSERR_IO,
315 NFSERR_ACCES,
316 NFSERR_EXIST,
317 NFSERR_NOTDIR,
318 NFSERR_NOSPC,
319 NFSERR_ROFS,
320 NFSERR_NAMETOL,
321 NFSERR_DQUOT,
322 NFSERR_STALE,
323 NFSERR_BADHANDLE,
324 NFSERR_NOTSUPP,
325 NFSERR_SERVERFAULT,
326 0,
327};
328
329static short nfsv3err_mkdir[] = {
330 NFSERR_IO,
331 NFSERR_IO,
332 NFSERR_ACCES,
333 NFSERR_EXIST,
334 NFSERR_NOTDIR,
335 NFSERR_NOSPC,
336 NFSERR_ROFS,
337 NFSERR_NAMETOL,
338 NFSERR_DQUOT,
339 NFSERR_STALE,
340 NFSERR_BADHANDLE,
341 NFSERR_NOTSUPP,
342 NFSERR_SERVERFAULT,
343 0,
344};
345
346static short nfsv3err_symlink[] = {
347 NFSERR_IO,
348 NFSERR_IO,
349 NFSERR_ACCES,
350 NFSERR_EXIST,
351 NFSERR_NOTDIR,
352 NFSERR_NOSPC,
353 NFSERR_ROFS,
354 NFSERR_NAMETOL,
355 NFSERR_DQUOT,
356 NFSERR_STALE,
357 NFSERR_BADHANDLE,
358 NFSERR_NOTSUPP,
359 NFSERR_SERVERFAULT,
360 0,
361};
362
363static short nfsv3err_mknod[] = {
364 NFSERR_IO,
365 NFSERR_IO,
366 NFSERR_ACCES,
367 NFSERR_EXIST,
368 NFSERR_NOTDIR,
369 NFSERR_NOSPC,
370 NFSERR_ROFS,
371 NFSERR_NAMETOL,
372 NFSERR_DQUOT,
373 NFSERR_STALE,
374 NFSERR_BADHANDLE,
375 NFSERR_NOTSUPP,
376 NFSERR_SERVERFAULT,
377 NFSERR_BADTYPE,
378 0,
379};
380
381static short nfsv3err_remove[] = {
382 NFSERR_IO,
383 NFSERR_NOENT,
384 NFSERR_IO,
385 NFSERR_ACCES,
386 NFSERR_NOTDIR,
387 NFSERR_ROFS,
388 NFSERR_NAMETOL,
389 NFSERR_STALE,
390 NFSERR_BADHANDLE,
391 NFSERR_SERVERFAULT,
392 0,
393};
394
395static short nfsv3err_rmdir[] = {
396 NFSERR_IO,
397 NFSERR_NOENT,
398 NFSERR_IO,
399 NFSERR_ACCES,
400 NFSERR_EXIST,
401 NFSERR_NOTDIR,
402 NFSERR_INVAL,
403 NFSERR_ROFS,
404 NFSERR_NAMETOL,
405 NFSERR_NOTEMPTY,
406 NFSERR_STALE,
407 NFSERR_BADHANDLE,
408 NFSERR_NOTSUPP,
409 NFSERR_SERVERFAULT,
410 0,
411};
412
413static short nfsv3err_rename[] = {
414 NFSERR_IO,
415 NFSERR_NOENT,
416 NFSERR_IO,
417 NFSERR_ACCES,
418 NFSERR_EXIST,
419 NFSERR_XDEV,
420 NFSERR_NOTDIR,
421 NFSERR_ISDIR,
422 NFSERR_INVAL,
423 NFSERR_NOSPC,
424 NFSERR_ROFS,
425 NFSERR_MLINK,
426 NFSERR_NAMETOL,
427 NFSERR_NOTEMPTY,
428 NFSERR_DQUOT,
429 NFSERR_STALE,
430 NFSERR_BADHANDLE,
431 NFSERR_NOTSUPP,
432 NFSERR_SERVERFAULT,
433 0,
434};
435
436static short nfsv3err_link[] = {
437 NFSERR_IO,
438 NFSERR_IO,
439 NFSERR_ACCES,
440 NFSERR_EXIST,
441 NFSERR_XDEV,
442 NFSERR_NOTDIR,
443 NFSERR_INVAL,
444 NFSERR_NOSPC,
445 NFSERR_ROFS,
446 NFSERR_MLINK,
447 NFSERR_NAMETOL,
448 NFSERR_DQUOT,
449 NFSERR_STALE,
450 NFSERR_BADHANDLE,
451 NFSERR_NOTSUPP,
452 NFSERR_SERVERFAULT,
453 0,
454};
455
456static short nfsv3err_readdir[] = {
457 NFSERR_IO,
458 NFSERR_IO,
459 NFSERR_ACCES,
460 NFSERR_NOTDIR,
461 NFSERR_STALE,
462 NFSERR_BADHANDLE,
463 NFSERR_BAD_COOKIE,
464 NFSERR_TOOSMALL,
465 NFSERR_SERVERFAULT,
466 0,
467};
468
469static short nfsv3err_readdirplus[] = {
470 NFSERR_IO,
471 NFSERR_IO,
472 NFSERR_ACCES,
473 NFSERR_NOTDIR,
474 NFSERR_STALE,
475 NFSERR_BADHANDLE,
476 NFSERR_BAD_COOKIE,
477 NFSERR_NOTSUPP,
478 NFSERR_TOOSMALL,
479 NFSERR_SERVERFAULT,
480 0,
481};
482
483static short nfsv3err_fsstat[] = {
484 NFSERR_IO,
485 NFSERR_IO,
486 NFSERR_STALE,
487 NFSERR_BADHANDLE,
488 NFSERR_SERVERFAULT,
489 0,
490};
491
492static short nfsv3err_fsinfo[] = {
493 NFSERR_STALE,
494 NFSERR_STALE,
495 NFSERR_BADHANDLE,
496 NFSERR_SERVERFAULT,
497 0,
498};
499
500static short nfsv3err_pathconf[] = {
501 NFSERR_STALE,
502 NFSERR_STALE,
503 NFSERR_BADHANDLE,
504 NFSERR_SERVERFAULT,
505 0,
506};
507
508static short nfsv3err_commit[] = {
509 NFSERR_IO,
510 NFSERR_IO,
511 NFSERR_STALE,
512 NFSERR_BADHANDLE,
513 NFSERR_SERVERFAULT,
514 0,
515};
516
517static short *nfsrv_v3errmap[] = {
518 nfsv3err_null,
519 nfsv3err_getattr,
520 nfsv3err_setattr,
521 nfsv3err_lookup,
522 nfsv3err_access,
523 nfsv3err_readlink,
524 nfsv3err_read,
525 nfsv3err_write,
526 nfsv3err_create,
527 nfsv3err_mkdir,
528 nfsv3err_symlink,
529 nfsv3err_mknod,
530 nfsv3err_remove,
531 nfsv3err_rmdir,
532 nfsv3err_rename,
533 nfsv3err_link,
534 nfsv3err_readdir,
535 nfsv3err_readdirplus,
536 nfsv3err_fsstat,
537 nfsv3err_fsinfo,
538 nfsv3err_pathconf,
539 nfsv3err_commit,
540};
541
542#endif /* NFS_NOSERVER */
543
984263bc 544struct nfssvc_args;
753fd850 545extern int sys_nfssvc(struct proc *, struct nfssvc_args *, int *);
984263bc 546
2cb16efc
MD
547/*
548 * This needs to return a monotonically increasing or close to monotonically
549 * increasing result, otherwise the write gathering queues won't work
550 * properly.
551 */
984263bc 552u_quad_t
e851b29e 553nfs_curusec(void)
984263bc
MD
554{
555 struct timeval tv;
556
2cb16efc 557 getmicrouptime(&tv);
984263bc
MD
558 return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
559}
560
561/*
984263bc
MD
562 * Called once to initialize data structures...
563 */
564int
e851b29e 565nfs_init(struct vfsconf *vfsp)
984263bc 566{
f786cc86 567 callout_init(&nfs_timer_handle);
984263bc
MD
568 nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
569
570 nfs_mount_type = vfsp->vfc_typenum;
571 nfsrtt.pos = 0;
572 rpc_vers = txdr_unsigned(RPC_VER2);
573 rpc_call = txdr_unsigned(RPC_CALL);
574 rpc_reply = txdr_unsigned(RPC_REPLY);
575 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
576 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
577 rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
578 rpc_autherr = txdr_unsigned(RPC_AUTHERR);
579 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
580 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
581 nfs_prog = txdr_unsigned(NFS_PROG);
984263bc
MD
582 nfs_true = txdr_unsigned(TRUE);
583 nfs_false = txdr_unsigned(FALSE);
584 nfs_xdrneg1 = txdr_unsigned(-1);
585 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
586 if (nfs_ticks < 1)
587 nfs_ticks = 1;
984263bc
MD
588 nfs_nhinit(); /* Init the nfsnode table */
589#ifndef NFS_NOSERVER
590 nfsrv_init(0); /* Init server data structures */
591 nfsrv_initcache(); /* Init the server request cache */
592#endif
593
594 /*
b9a7a2bd
MD
595 * Mainly for vkernel operation. If memory is severely limited
596 */
597 if (nfs_maxasyncbio > nmbclusters * MCLBYTES / NFS_MAXDATA / 3)
598 nfs_maxasyncbio = nmbclusters * MCLBYTES / NFS_MAXDATA / 3;
599 if (nfs_maxasyncbio < 4)
600 nfs_maxasyncbio = 4;
601
602 /*
984263bc
MD
603 * Initialize reply list and start timer
604 */
984263bc
MD
605 nfs_timer(0);
606
984263bc
MD
607 nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
608 sysent[SYS_nfssvc].sy_narg = 2;
609 nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
753fd850 610 sysent[SYS_nfssvc].sy_call = (sy_call_t *)sys_nfssvc;
984263bc
MD
611
612 nfs_pbuf_freecnt = nswbuf / 2 + 1;
613
614 return (0);
615}
616
617int
e851b29e 618nfs_uninit(struct vfsconf *vfsp)
984263bc 619{
f786cc86 620 callout_stop(&nfs_timer_handle);
984263bc 621 nfs_mount_type = -1;
984263bc
MD
622 sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
623 sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
624 return (0);
625}
626
627/*
628 * Attribute cache routines.
629 * nfs_loadattrcache() - loads or updates the cache contents from attributes
630 * that are on the mbuf list
631 * nfs_getattrcache() - returns valid attributes if found in cache, returns
632 * error otherwise
633 */
634
635/*
636 * Load the attribute cache (that lives in the nfsnode entry) with
5a9187cb
MD
637 * the values on the mbuf list. Load *vaper with the attributes. vaper
638 * may be NULL.
639 *
640 * As a side effect n_mtime, which we use to determine if the file was
641 * modified by some other host, is set to the attribute timestamp and
642 * NRMODIFIED is set if the two values differ.
643 *
644 * WARNING: the mtime loaded into vaper does not necessarily represent
645 * n_mtime or n_attr.mtime due to NACC and NUPD.
984263bc
MD
646 */
647int
42edf14f 648nfs_loadattrcache(struct vnode *vp, struct mbuf **mdp, caddr_t *dposp,
5a9187cb 649 struct vattr *vaper, int lattr_flags)
984263bc 650{
40393ded
RG
651 struct vattr *vap;
652 struct nfs_fattr *fp;
653 struct nfsnode *np;
654 int32_t t1;
984263bc 655 caddr_t cp2;
e4c9c0c8 656 int error = 0;
0e9b9130 657 int rmajor, rminor;
e4c9c0c8 658 udev_t rdev;
984263bc
MD
659 struct mbuf *md;
660 enum vtype vtyp;
661 u_short vmode;
662 struct timespec mtime;
663 int v3 = NFS_ISV3(vp);
664
665 md = *mdp;
666 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
667 if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
668 return (error);
669 fp = (struct nfs_fattr *)cp2;
670 if (v3) {
671 vtyp = nfsv3tov_type(fp->fa_type);
672 vmode = fxdr_unsigned(u_short, fp->fa_mode);
0e9b9130
MD
673 rmajor = (int)fxdr_unsigned(int, fp->fa3_rdev.specdata1);
674 rminor = (int)fxdr_unsigned(int, fp->fa3_rdev.specdata2);
984263bc
MD
675 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
676 } else {
677 vtyp = nfsv2tov_type(fp->fa_type);
678 vmode = fxdr_unsigned(u_short, fp->fa_mode);
679 /*
680 * XXX
681 *
682 * The duplicate information returned in fa_type and fa_mode
683 * is an ambiguity in the NFS version 2 protocol.
684 *
685 * VREG should be taken literally as a regular file. If a
686 * server intents to return some type information differently
687 * in the upper bits of the mode field (e.g. for sockets, or
688 * FIFOs), NFSv2 mandates fa_type to be VNON. Anyway, we
689 * leave the examination of the mode bits even in the VREG
690 * case to avoid breakage for bogus servers, but we make sure
691 * that there are actually type bits set in the upper part of
692 * fa_mode (and failing that, trust the va_type field).
693 *
694 * NFSv3 cleared the issue, and requires fa_mode to not
695 * contain any type information (while also introduing sockets
696 * and FIFOs for fa_type).
697 */
698 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
699 vtyp = IFTOVT(vmode);
700 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
0e9b9130
MD
701 rmajor = umajor(rdev);
702 rminor = uminor(rdev);
984263bc
MD
703 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
704
705 /*
706 * Really ugly NFSv2 kludge.
707 */
e4c9c0c8 708 if (vtyp == VCHR && rdev == (udev_t)0xffffffff)
984263bc
MD
709 vtyp = VFIFO;
710 }
711
712 /*
713 * If v_type == VNON it is a new node, so fill in the v_type,
714 * n_mtime fields. Check to see if it represents a special
715 * device, and if so, check for a possible alias. Once the
716 * correct vnode has been obtained, fill in the rest of the
717 * information.
718 */
719 np = VTONFS(vp);
720 if (vp->v_type != vtyp) {
1c843a13 721 nfs_setvtype(vp, vtyp);
984263bc 722 if (vp->v_type == VFIFO) {
6ddb7618 723 vp->v_ops = &vp->v_mount->mnt_vn_fifo_ops;
0961aa92 724 } else if (vp->v_type == VCHR || vp->v_type == VBLK) {
6ddb7618 725 vp->v_ops = &vp->v_mount->mnt_vn_spec_ops;
0e9b9130 726 addaliasu(vp, rmajor, rminor);
0961aa92 727 } else {
6ddb7618 728 vp->v_ops = &vp->v_mount->mnt_vn_use_ops;
984263bc
MD
729 }
730 np->n_mtime = mtime.tv_sec;
5a9187cb 731 } else if (np->n_mtime != mtime.tv_sec) {
9793c819 732 /*
5a9187cb
MD
733 * If we haven't modified the file locally and the server
734 * timestamp does not match, then the server probably
735 * modified the file. We must flag this condition so
736 * the proper syncnronization can be done. We do not
737 * try to synchronize the state here because that
738 * could lead to an endless recursion.
739 *
740 * XXX loadattrcache can be set during the reply to a write,
741 * before the write timestamp is properly processed. To
742 * avoid unconditionally setting the rmodified bit (which
743 * has the effect of flushing the cache), we only do this
744 * check if the lmodified bit is not set.
9793c819
MD
745 */
746 np->n_mtime = mtime.tv_sec;
5a9187cb
MD
747 if ((lattr_flags & NFS_LATTR_NOMTIMECHECK) == 0)
748 np->n_flag |= NRMODIFIED;
984263bc
MD
749 }
750 vap = &np->n_vattr;
751 vap->va_type = vtyp;
752 vap->va_mode = (vmode & 07777);
0e9b9130
MD
753 vap->va_rmajor = rmajor;
754 vap->va_rminor = rminor;
984263bc
MD
755 vap->va_mtime = mtime;
756 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
757 if (v3) {
758 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
759 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
760 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
761 vap->va_size = fxdr_hyper(&fp->fa3_size);
762 vap->va_blocksize = NFS_FABLKSIZE;
763 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
50626622 764 vap->va_fileid = fxdr_hyper(&fp->fa3_fileid);
984263bc
MD
765 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
766 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
767 vap->va_flags = 0;
768 vap->va_filerev = 0;
769 } else {
770 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
771 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
772 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
773 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
774 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
775 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
776 * NFS_FABLKSIZE;
777 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
778 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
779 vap->va_flags = 0;
780 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
781 fp->fa2_ctime.nfsv2_sec);
782 vap->va_ctime.tv_nsec = 0;
783 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
784 vap->va_filerev = 0;
785 }
786 np->n_attrstamp = time_second;
787 if (vap->va_size != np->n_size) {
788 if (vap->va_type == VREG) {
5a9187cb
MD
789 if ((lattr_flags & NFS_LATTR_NOSHRINK) &&
790 vap->va_size < np->n_size) {
984263bc
MD
791 /*
792 * We've been told not to shrink the file;
793 * zero np->n_attrstamp to indicate that
794 * the attributes are stale.
5a9187cb
MD
795 *
796 * This occurs primarily due to recursive
797 * NFS ops that are executed during periods
798 * where we cannot safely reduce the size of
799 * the file.
800 *
801 * Additionally, write rpcs are broken down
802 * into buffers and np->n_size is
803 * pre-extended. Setting NRMODIFIED here
804 * can result in n_size getting reset to a
805 * lower value, which is NOT what we want.
806 * XXX this needs to be cleaned up a lot
807 * more.
984263bc
MD
808 */
809 vap->va_size = np->n_size;
810 np->n_attrstamp = 0;
5a9187cb
MD
811 if ((np->n_flag & NLMODIFIED) == 0)
812 np->n_flag |= NRMODIFIED;
813 } else if (np->n_flag & NLMODIFIED) {
79e5012e
MD
814 /*
815 * We've modified the file: Use the larger
5a9187cb
MD
816 * of our size, and the server's size. At
817 * this point the cache coherency is all
818 * shot to hell. To try to handle multiple
819 * clients appending to the file at the same
820 * time mark that the server has changed
821 * the file if the server's notion of the
822 * file size is larger then our notion.
823 *
824 * XXX this needs work.
79e5012e
MD
825 */
826 if (vap->va_size < np->n_size) {
984263bc 827 vap->va_size = np->n_size;
79e5012e 828 } else {
984263bc 829 np->n_size = vap->va_size;
5a9187cb 830 np->n_flag |= NRMODIFIED;
79e5012e 831 }
984263bc 832 } else {
5a9187cb
MD
833 /*
834 * Someone changed the file's size on the
835 * server and there are no local changes
836 * to get in the way, set the size and mark
837 * it.
838 */
984263bc 839 np->n_size = vap->va_size;
5a9187cb 840 np->n_flag |= NRMODIFIED;
984263bc
MD
841 }
842 vnode_pager_setsize(vp, np->n_size);
843 } else {
844 np->n_size = vap->va_size;
845 }
846 }
847 if (vaper != NULL) {
848 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
849 if (np->n_flag & NCHG) {
850 if (np->n_flag & NACC)
851 vaper->va_atime = np->n_atim;
852 if (np->n_flag & NUPD)
853 vaper->va_mtime = np->n_mtim;
854 }
855 }
856 return (0);
857}
858
859#ifdef NFS_ACDEBUG
860#include <sys/sysctl.h>
861SYSCTL_DECL(_vfs_nfs);
862static int nfs_acdebug;
863SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
864#endif
865
866/*
867 * Check the time stamp
868 * If the cache is valid, copy contents to *vap and return 0
869 * otherwise return an error
870 */
871int
e851b29e 872nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
984263bc 873{
40393ded
RG
874 struct nfsnode *np;
875 struct vattr *vap;
984263bc
MD
876 struct nfsmount *nmp;
877 int timeo;
878
879 np = VTONFS(vp);
880 vap = &np->n_vattr;
881 nmp = VFSTONFS(vp->v_mount);
fad57d0e
MD
882
883 /*
884 * Dynamic timeout based on how recently the file was modified.
5a9187cb 885 * n_mtime is always valid.
fad57d0e 886 */
97100839 887 timeo = (get_approximate_time_t() - np->n_mtime) / 60;
984263bc
MD
888
889#ifdef NFS_ACDEBUG
890 if (nfs_acdebug>1)
086c1d7e 891 kprintf("nfs_getattrcache: initial timeo = %d\n", timeo);
984263bc
MD
892#endif
893
894 if (vap->va_type == VDIR) {
5a9187cb 895 if ((np->n_flag & NLMODIFIED) || timeo < nmp->nm_acdirmin)
984263bc
MD
896 timeo = nmp->nm_acdirmin;
897 else if (timeo > nmp->nm_acdirmax)
898 timeo = nmp->nm_acdirmax;
899 } else {
5a9187cb 900 if ((np->n_flag & NLMODIFIED) || timeo < nmp->nm_acregmin)
984263bc
MD
901 timeo = nmp->nm_acregmin;
902 else if (timeo > nmp->nm_acregmax)
903 timeo = nmp->nm_acregmax;
904 }
905
906#ifdef NFS_ACDEBUG
907 if (nfs_acdebug > 2)
086c1d7e 908 kprintf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
984263bc
MD
909 nmp->nm_acregmin, nmp->nm_acregmax,
910 nmp->nm_acdirmin, nmp->nm_acdirmax);
911
912 if (nfs_acdebug)
086c1d7e 913 kprintf("nfs_getattrcache: age = %d; final timeo = %d\n",
9793c819 914 (int)(time_second - np->n_attrstamp), timeo);
984263bc
MD
915#endif
916
9793c819 917 if (np->n_attrstamp == 0 || (time_second - np->n_attrstamp) >= timeo) {
984263bc
MD
918 nfsstats.attrcache_misses++;
919 return (ENOENT);
920 }
921 nfsstats.attrcache_hits++;
5a9187cb
MD
922
923 /*
924 * Our attribute cache can be stale due to modifications made on
925 * this host. XXX this is a bad hack. We need a more deterministic
926 * means of finding out which np fields are valid verses attr cache
927 * fields. We really should update the vattr info on the fly when
928 * making local changes.
929 */
984263bc
MD
930 if (vap->va_size != np->n_size) {
931 if (vap->va_type == VREG) {
5a9187cb 932 if (np->n_flag & NLMODIFIED) {
984263bc
MD
933 if (vap->va_size < np->n_size)
934 vap->va_size = np->n_size;
935 else
936 np->n_size = vap->va_size;
937 } else {
938 np->n_size = vap->va_size;
939 }
940 vnode_pager_setsize(vp, np->n_size);
941 } else {
942 np->n_size = vap->va_size;
943 }
944 }
945 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
946 if (np->n_flag & NCHG) {
947 if (np->n_flag & NACC)
948 vaper->va_atime = np->n_atim;
949 if (np->n_flag & NUPD)
950 vaper->va_mtime = np->n_mtim;
951 }
952 return (0);
953}
954
955#ifndef NFS_NOSERVER
fad57d0e 956
984263bc
MD
957/*
958 * Set up nameidata for a lookup() call and do it.
959 *
960 * If pubflag is set, this call is done for a lookup operation on the
961 * public filehandle. In that case we allow crossing mountpoints and
962 * absolute pathnames. However, the caller is expected to check that
963 * the lookup result is within the public fs, and deny access if
964 * it is not.
965 *
984263bc
MD
966 * dirp may be set whether an error is returned or not, and must be
967 * released by the caller.
fad57d0e 968 *
28623bf9 969 * On return nd->nl_nch usually points to the target ncp, which may represent
fad57d0e
MD
970 * a negative hit.
971 *
972 * NOTE: the caller must call nlookup_done(nd) unconditionally on return
973 * to cleanup.
984263bc
MD
974 */
975int
3a907475 976nfs_namei(struct nlookupdata *nd, struct ucred *cred, int nflags,
fad57d0e
MD
977 struct vnode **dvpp, struct vnode **vpp,
978 fhandle_t *fhp, int len,
979 struct nfssvc_sock *slp, struct sockaddr *nam, struct mbuf **mdp,
980 caddr_t *dposp, struct vnode **dirpp, struct thread *td,
981 int kerbflag, int pubflag)
984263bc 982{
40393ded
RG
983 int i, rem;
984 struct mbuf *md;
985 char *fromcp, *tocp, *cp;
fad57d0e 986 char *namebuf;
28623bf9 987 struct nchandle nch;
984263bc 988 struct vnode *dp;
67863d04 989 struct mount *mp;
fad57d0e 990 int error, rdonly;
984263bc 991
70aac194 992 namebuf = objcache_get(namei_oc, M_WAITOK);
fad57d0e 993 *dirpp = NULL;
984263bc
MD
994
995 /*
fad57d0e 996 * Copy the name from the mbuf list to namebuf.
984263bc
MD
997 */
998 fromcp = *dposp;
fad57d0e 999 tocp = namebuf;
984263bc
MD
1000 md = *mdp;
1001 rem = mtod(md, caddr_t) + md->m_len - fromcp;
1002 for (i = 0; i < len; i++) {
1003 while (rem == 0) {
1004 md = md->m_next;
1005 if (md == NULL) {
1006 error = EBADRPC;
1007 goto out;
1008 }
1009 fromcp = mtod(md, caddr_t);
1010 rem = md->m_len;
1011 }
1012 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1013 error = EACCES;
1014 goto out;
1015 }
1016 *tocp++ = *fromcp++;
1017 rem--;
1018 }
1019 *tocp = '\0';
1020 *mdp = md;
1021 *dposp = fromcp;
1022 len = nfsm_rndup(len)-len;
1023 if (len > 0) {
1024 if (rem >= len)
1025 *dposp += len;
1026 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1027 goto out;
1028 }
1029
1030 /*
fad57d0e
MD
1031 * Extract and set starting directory. The returned dp is refd
1032 * but not locked.
984263bc 1033 */
67863d04 1034 error = nfsrv_fhtovp(fhp, FALSE, &mp, &dp, cred, slp,
fad57d0e 1035 nam, &rdonly, kerbflag, pubflag);
984263bc
MD
1036 if (error)
1037 goto out;
1038 if (dp->v_type != VDIR) {
1039 vrele(dp);
1040 error = ENOTDIR;
1041 goto out;
1042 }
1043
984263bc
MD
1044 /*
1045 * Set return directory. Reference to dp is implicitly transfered
fad57d0e
MD
1046 * to the returned pointer. This must be set before we potentially
1047 * goto out below.
984263bc 1048 */
fad57d0e 1049 *dirpp = dp;
984263bc 1050
d64fd6d3
MD
1051 /*
1052 * read-only - NLC_DELETE, NLC_RENAME_DST are disallowed. NLC_CREATE
1053 * is passed through to nlookup() and will be disallowed
1054 * if the file does not already exist.
1055 */
1056 if (rdonly) {
1057 nflags |= NLC_NFS_RDONLY;
1058 if (nflags & (NLC_DELETE | NLC_RENAME_DST)) {
1059 error = EROFS;
1060 goto out;
1061 }
1062 }
1063
1064 /*
1065 * Oh joy. For WebNFS, handle those pesky '%' escapes,
1066 * and the 'native path' indicator.
1067 */
984263bc 1068 if (pubflag) {
70aac194 1069 cp = objcache_get(namei_oc, M_WAITOK);
fad57d0e 1070 fromcp = namebuf;
984263bc
MD
1071 tocp = cp;
1072 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1073 switch ((unsigned char)*fromcp) {
1074 case WEBNFS_NATIVE_CHAR:
1075 /*
1076 * 'Native' path for us is the same
1077 * as a path according to the NFS spec,
1078 * just skip the escape char.
1079 */
1080 fromcp++;
1081 break;
1082 /*
1083 * More may be added in the future, range 0x80-0xff
1084 */
1085 default:
1086 error = EIO;
70aac194 1087 objcache_put(namei_oc, cp);
984263bc
MD
1088 goto out;
1089 }
1090 }
1091 /*
1092 * Translate the '%' escapes, URL-style.
1093 */
1094 while (*fromcp != '\0') {
1095 if (*fromcp == WEBNFS_ESC_CHAR) {
1096 if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1097 fromcp++;
1098 *tocp++ = HEXSTRTOI(fromcp);
1099 fromcp += 2;
1100 continue;
1101 } else {
1102 error = ENOENT;
70aac194 1103 objcache_put(namei_oc, cp);
984263bc
MD
1104 goto out;
1105 }
1106 } else
1107 *tocp++ = *fromcp++;
1108 }
1109 *tocp = '\0';
70aac194 1110 objcache_put(namei_oc, namebuf);
fad57d0e 1111 namebuf = cp;
984263bc
MD
1112 }
1113
fad57d0e
MD
1114 /*
1115 * Setup for search. We need to get a start directory from dp. Note
1116 * that dp is ref'd, but we no longer 'own' the ref (*dirpp owns it).
1117 */
1118 if (pubflag == 0) {
3a907475
MD
1119 nflags |= NLC_NFS_NOSOFTLINKTRAV;
1120 nflags |= NLC_NOCROSSMOUNT;
984263bc
MD
1121 }
1122
1123 /*
fad57d0e
MD
1124 * We need a starting ncp from the directory vnode dp. dp must not
1125 * be locked. The returned ncp will be refd but not locked.
1126 *
1127 * If no suitable ncp is found we instruct cache_fromdvp() to create
1128 * one. If this fails the directory has probably been removed while
1129 * the target was chdir'd into it and any further lookup will fail.
984263bc 1130 */
28623bf9 1131 if ((error = cache_fromdvp(dp, cred, 1, &nch)) != 0)
fad57d0e 1132 goto out;
3a907475 1133 nlookup_init_raw(nd, namebuf, UIO_SYSSPACE, nflags, cred, &nch);
28623bf9 1134 cache_drop(&nch);
984263bc 1135
fad57d0e
MD
1136 /*
1137 * Ok, do the lookup.
1138 */
1139 error = nlookup(nd);
984263bc 1140
fad57d0e
MD
1141 /*
1142 * If no error occured return the requested dvpp and vpp. If
28623bf9 1143 * NLC_CREATE was specified nd->nl_nch may represent a negative
fad57d0e
MD
1144 * cache hit in which case we do not attempt to obtain the vp.
1145 */
1146 if (error == 0) {
fad57d0e 1147 if (dvpp) {
443472eb 1148 if (nd->nl_nch.ncp->nc_parent) {
28623bf9
MD
1149 nch = nd->nl_nch;
1150 nch.ncp = nch.ncp->nc_parent;
1151 error = cache_vget(&nch, nd->nl_cred,
1152 LK_EXCLUSIVE, dvpp);
fad57d0e
MD
1153 } else {
1154 error = ENXIO;
1155 }
984263bc 1156 }
28623bf9
MD
1157 if (vpp && nd->nl_nch.ncp->nc_vp) {
1158 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, vpp);
984263bc 1159 }
984263bc 1160 if (error) {
fad57d0e
MD
1161 if (dvpp && *dvpp) {
1162 vput(*dvpp);
1163 *dvpp = NULL;
1164 }
1165 if (vpp && *vpp) {
1166 vput(*vpp);
1167 *vpp = NULL;
1168 }
984263bc 1169 }
984263bc
MD
1170 }
1171
1172 /*
fad57d0e 1173 * Finish up.
984263bc
MD
1174 */
1175out:
70aac194 1176 objcache_put(namei_oc, namebuf);
984263bc
MD
1177 return (error);
1178}
1179
1180/*
984263bc
MD
1181 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1182 * - look up fsid in mount list (if not found ret error)
1183 * - get vp and export rights by calling VFS_FHTOVP()
1184 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
a11aaa81 1185 * - if not lockflag unlock it with vn_unlock()
984263bc
MD
1186 */
1187int
67863d04
MD
1188nfsrv_fhtovp(fhandle_t *fhp, int lockflag,
1189 struct mount **mpp, struct vnode **vpp,
e851b29e
CP
1190 struct ucred *cred, struct nfssvc_sock *slp, struct sockaddr *nam,
1191 int *rdonlyp, int kerbflag, int pubflag)
984263bc 1192{
40393ded
RG
1193 struct mount *mp;
1194 int i;
984263bc
MD
1195 struct ucred *credanon;
1196 int error, exflags;
1197#ifdef MNT_EXNORESPORT /* XXX needs mountd and /etc/exports help yet */
1198 struct sockaddr_int *saddr;
1199#endif
1200
67863d04
MD
1201 *vpp = NULL;
1202 *mpp = NULL;
984263bc
MD
1203
1204 if (nfs_ispublicfh(fhp)) {
1205 if (!pubflag || !nfs_pub.np_valid)
1206 return (ESTALE);
1207 fhp = &nfs_pub.np_handle;
1208 }
1209
67863d04
MD
1210 mp = *mpp = vfs_getvfs(&fhp->fh_fsid);
1211 if (mp == NULL)
984263bc
MD
1212 return (ESTALE);
1213 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1214 if (error)
1215 return (error);
67863d04 1216 error = VFS_FHTOVP(mp, NULL, &fhp->fh_fid, vpp);
984263bc
MD
1217 if (error)
1218 return (error);
1219#ifdef MNT_EXNORESPORT
1220 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1221 saddr = (struct sockaddr_in *)nam;
1222 if (saddr->sin_family == AF_INET &&
1223 ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1224 vput(*vpp);
1225 *vpp = NULL;
1226 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1227 }
1228 }
1229#endif
1230 /*
1231 * Check/setup credentials.
1232 */
1233 if (exflags & MNT_EXKERB) {
1234 if (!kerbflag) {
1235 vput(*vpp);
1236 *vpp = NULL;
1237 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1238 }
1239 } else if (kerbflag) {
1240 vput(*vpp);
1241 *vpp = NULL;
1242 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1243 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1244 cred->cr_uid = credanon->cr_uid;
1245 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1246 cred->cr_groups[i] = credanon->cr_groups[i];
1247 cred->cr_ngroups = i;
1248 }
1249 if (exflags & MNT_EXRDONLY)
1250 *rdonlyp = 1;
1251 else
1252 *rdonlyp = 0;
1253
984263bc 1254 if (!lockflag)
a11aaa81 1255 vn_unlock(*vpp);
984263bc
MD
1256 return (0);
1257}
1258
984263bc
MD
1259/*
1260 * WebNFS: check if a filehandle is a public filehandle. For v3, this
1261 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
1262 * transformed this to all zeroes in both cases, so check for it.
1263 */
1264int
e851b29e 1265nfs_ispublicfh(fhandle_t *fhp)
984263bc
MD
1266{
1267 char *cp = (char *)fhp;
1268 int i;
1269
1270 for (i = 0; i < NFSX_V3FH; i++)
1271 if (*cp++ != 0)
1272 return (FALSE);
1273 return (TRUE);
1274}
1275
1276#endif /* NFS_NOSERVER */
1277/*
1278 * This function compares two net addresses by family and returns TRUE
1279 * if they are the same host.
1280 * If there is any doubt, return FALSE.
1281 * The AF_INET family is handled as a special case so that address mbufs
1282 * don't need to be saved to store "struct in_addr", which is only 4 bytes.
1283 */
1284int
e851b29e 1285netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
984263bc 1286{
40393ded 1287 struct sockaddr_in *inetaddr;
984263bc
MD
1288
1289 switch (family) {
1290 case AF_INET:
1291 inetaddr = (struct sockaddr_in *)nam;
1292 if (inetaddr->sin_family == AF_INET &&
1293 inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
1294 return (1);
1295 break;
1296 default:
1297 break;
1298 };
1299 return (0);
1300}
1301
1302static nfsuint64 nfs_nullcookie = { { 0, 0 } };
1303/*
1304 * This function finds the directory cookie that corresponds to the
1305 * logical byte offset given.
1306 */
1307nfsuint64 *
e851b29e 1308nfs_getcookie(struct nfsnode *np, off_t off, int add)
984263bc 1309{
40393ded
RG
1310 struct nfsdmap *dp, *dp2;
1311 int pos;
984263bc
MD
1312
1313 pos = (uoff_t)off / NFS_DIRBLKSIZ;
1314 if (pos == 0 || off < 0) {
1315#ifdef DIAGNOSTIC
1316 if (add)
1317 panic("nfs getcookie add at <= 0");
1318#endif
1319 return (&nfs_nullcookie);
1320 }
1321 pos--;
1322 dp = np->n_cookies.lh_first;
1323 if (!dp) {
1324 if (add) {
1325 MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
1326 M_NFSDIROFF, M_WAITOK);
1327 dp->ndm_eocookie = 0;
1328 LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
1329 } else
60233e58 1330 return (NULL);
984263bc
MD
1331 }
1332 while (pos >= NFSNUMCOOKIES) {
1333 pos -= NFSNUMCOOKIES;
1334 if (dp->ndm_list.le_next) {
1335 if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
1336 pos >= dp->ndm_eocookie)
60233e58 1337 return (NULL);
984263bc
MD
1338 dp = dp->ndm_list.le_next;
1339 } else if (add) {
1340 MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
1341 M_NFSDIROFF, M_WAITOK);
1342 dp2->ndm_eocookie = 0;
1343 LIST_INSERT_AFTER(dp, dp2, ndm_list);
1344 dp = dp2;
1345 } else
60233e58 1346 return (NULL);
984263bc
MD
1347 }
1348 if (pos >= dp->ndm_eocookie) {
1349 if (add)
1350 dp->ndm_eocookie = pos + 1;
1351 else
60233e58 1352 return (NULL);
984263bc
MD
1353 }
1354 return (&dp->ndm_cookies[pos]);
1355}
1356
1357/*
1358 * Invalidate cached directory information, except for the actual directory
1359 * blocks (which are invalidated separately).
1360 * Done mainly to avoid the use of stale offset cookies.
1361 */
1362void
e851b29e 1363nfs_invaldir(struct vnode *vp)
984263bc 1364{
40393ded 1365 struct nfsnode *np = VTONFS(vp);
984263bc
MD
1366
1367#ifdef DIAGNOSTIC
1368 if (vp->v_type != VDIR)
1369 panic("nfs: invaldir not dir");
1370#endif
1371 np->n_direofoffset = 0;
1372 np->n_cookieverf.nfsuquad[0] = 0;
1373 np->n_cookieverf.nfsuquad[1] = 0;
1374 if (np->n_cookies.lh_first)
1375 np->n_cookies.lh_first->ndm_eocookie = 0;
1376}
1377
1378/*
1c843a13
MD
1379 * Set the v_type field for an NFS client's vnode and initialize for
1380 * buffer cache operations if necessary.
1381 */
1382void
1383nfs_setvtype(struct vnode *vp, enum vtype vtyp)
1384{
1385 vp->v_type = vtyp;
1386
1387 switch(vtyp) {
1388 case VREG:
1389 case VDIR:
1390 case VLNK:
1391 vinitvmio(vp, 0); /* needs VMIO, size not yet known */
1392 break;
1393 default:
1394 break;
1395 }
1396}
1397
1398/*
984263bc
MD
1399 * The write verifier has changed (probably due to a server reboot), so all
1400 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
1401 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
1402 * and B_CLUSTEROK flags. Once done the new write verifier can be set for the
1403 * mount point.
1404 *
1405 * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
1406 * writes are not clusterable.
1407 */
6bae6177
MD
1408
1409static int nfs_clearcommit_bp(struct buf *bp, void *data __unused);
0202303b
MD
1410static int nfs_clearcommit_callback(struct mount *mp, struct vnode *vp,
1411 void *data __unused);
6bae6177 1412
984263bc 1413void
e851b29e 1414nfs_clearcommit(struct mount *mp)
984263bc 1415{
0202303b
MD
1416 vmntvnodescan(mp, VMSC_NOWAIT, nfs_clearcommit_callback, NULL, NULL);
1417}
1418
1419static int
1420nfs_clearcommit_callback(struct mount *mp, struct vnode *vp,
1421 void *data __unused)
1422{
1423 lwkt_tokref vlock;
1424
1425 vhold(vp);
1426 lwkt_gettoken(&vlock, &vp->v_token);
1427 RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
1428 nfs_clearcommit_bp, NULL);
1429 lwkt_reltoken(&vlock);
1430 vdrop(vp);
1431 return(0);
984263bc
MD
1432}
1433
6bae6177
MD
1434static int
1435nfs_clearcommit_bp(struct buf *bp, void *data __unused)
1436{
1437 if (BUF_REFCNT(bp) == 0 &&
1438 (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
1439 == (B_DELWRI | B_NEEDCOMMIT)) {
1440 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
1441 }
1442 return(0);
1443}
1444
984263bc
MD
1445#ifndef NFS_NOSERVER
1446/*
1447 * Map errnos to NFS error numbers. For Version 3 also filter out error
1448 * numbers not specified for the associated procedure.
1449 */
1450int
e851b29e 1451nfsrv_errmap(struct nfsrv_descript *nd, int err)
984263bc 1452{
40393ded 1453 short *defaulterrp, *errp;
984263bc
MD
1454
1455 if (nd->nd_flag & ND_NFSV3) {
1456 if (nd->nd_procnum <= NFSPROC_COMMIT) {
1457 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
1458 while (*++errp) {
1459 if (*errp == err)
1460 return (err);
1461 else if (*errp > err)
1462 break;
1463 }
1464 return ((int)*defaulterrp);
1465 } else
1466 return (err & 0xffff);
1467 }
1468 if (err <= ELAST)
1469 return ((int)nfsrv_v2errmap[err - 1]);
1470 return (NFSERR_IO);
1471}
1472
984263bc
MD
1473/*
1474 * Sort the group list in increasing numerical order.
1475 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
1476 * that used to be here.)
1477 */
1478void
e851b29e 1479nfsrvw_sort(gid_t *list, int num)
984263bc 1480{
40393ded 1481 int i, j;
984263bc
MD
1482 gid_t v;
1483
1484 /* Insertion sort. */
1485 for (i = 1; i < num; i++) {
1486 v = list[i];
1487 /* find correct slot for value v, moving others up */
1488 for (j = i; --j >= 0 && v < list[j];)
1489 list[j + 1] = list[j];
1490 list[j + 1] = v;
1491 }
1492}
1493
1494/*
1495 * copy credentials making sure that the result can be compared with bcmp().
1496 */
1497void
e851b29e 1498nfsrv_setcred(struct ucred *incred, struct ucred *outcred)
984263bc 1499{
40393ded 1500 int i;
984263bc
MD
1501
1502 bzero((caddr_t)outcred, sizeof (struct ucred));
1503 outcred->cr_ref = 1;
1504 outcred->cr_uid = incred->cr_uid;
1505 outcred->cr_ngroups = incred->cr_ngroups;
1506 for (i = 0; i < incred->cr_ngroups; i++)
1507 outcred->cr_groups[i] = incred->cr_groups[i];
1508 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
1509}
1510#endif /* NFS_NOSERVER */