Home | History | Annotate | Line # | Download | only in nfs
nfs_subs.c revision 1.103
      1 /*	$NetBSD: nfs_subs.c,v 1.103 2002/03/17 22:22:40 christos 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 /*
     42  * Copyright 2000 Wasabi Systems, Inc.
     43  * All rights reserved.
     44  *
     45  * Written by Frank van der Linden for Wasabi Systems, Inc.
     46  *
     47  * Redistribution and use in source and binary forms, with or without
     48  * modification, are permitted provided that the following conditions
     49  * are met:
     50  * 1. Redistributions of source code must retain the above copyright
     51  *    notice, this list of conditions and the following disclaimer.
     52  * 2. Redistributions in binary form must reproduce the above copyright
     53  *    notice, this list of conditions and the following disclaimer in the
     54  *    documentation and/or other materials provided with the distribution.
     55  * 3. All advertising materials mentioning features or use of this software
     56  *    must display the following acknowledgement:
     57  *      This product includes software developed for the NetBSD Project by
     58  *      Wasabi Systems, Inc.
     59  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
     60  *    or promote products derived from this software without specific prior
     61  *    written permission.
     62  *
     63  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
     64  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     65  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     66  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
     67  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     68  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     69  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     70  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     71  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     72  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     73  * POSSIBILITY OF SUCH DAMAGE.
     74  */
     75 
     76 #include <sys/cdefs.h>
     77 __KERNEL_RCSID(0, "$NetBSD: nfs_subs.c,v 1.103 2002/03/17 22:22:40 christos Exp $");
     78 
     79 #include "fs_nfs.h"
     80 #include "opt_nfs.h"
     81 #include "opt_nfsserver.h"
     82 #include "opt_iso.h"
     83 #include "opt_inet.h"
     84 
     85 /*
     86  * These functions support the macros and help fiddle mbuf chains for
     87  * the nfs op functions. They do things like create the rpc header and
     88  * copy data between mbuf chains and uio lists.
     89  */
     90 #include <sys/param.h>
     91 #include <sys/proc.h>
     92 #include <sys/systm.h>
     93 #include <sys/kernel.h>
     94 #include <sys/mount.h>
     95 #include <sys/vnode.h>
     96 #include <sys/namei.h>
     97 #include <sys/mbuf.h>
     98 #include <sys/socket.h>
     99 #include <sys/stat.h>
    100 #include <sys/malloc.h>
    101 #include <sys/filedesc.h>
    102 #include <sys/time.h>
    103 #include <sys/dirent.h>
    104 
    105 #include <uvm/uvm_extern.h>
    106 
    107 #include <nfs/rpcv2.h>
    108 #include <nfs/nfsproto.h>
    109 #include <nfs/nfsnode.h>
    110 #include <nfs/nfs.h>
    111 #include <nfs/xdr_subs.h>
    112 #include <nfs/nfsm_subs.h>
    113 #include <nfs/nfsmount.h>
    114 #include <nfs/nqnfs.h>
    115 #include <nfs/nfsrtt.h>
    116 #include <nfs/nfs_var.h>
    117 
    118 #include <miscfs/specfs/specdev.h>
    119 
    120 #include <netinet/in.h>
    121 #ifdef ISO
    122 #include <netiso/iso.h>
    123 #endif
    124 
    125 /*
    126  * Data items converted to xdr at startup, since they are constant
    127  * This is kinda hokey, but may save a little time doing byte swaps
    128  */
    129 u_int32_t nfs_xdrneg1;
    130 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
    131 	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
    132 	rpc_auth_kerb;
    133 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
    134 
    135 /* And other global data */
    136 static u_int32_t nfs_xid = 0;
    137 const nfstype nfsv2_type[9] =
    138 	{ NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON };
    139 const nfstype nfsv3_type[9] =
    140 	{ NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON };
    141 const enum vtype nv2tov_type[8] =
    142 	{ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON };
    143 const enum vtype nv3tov_type[8] =
    144 	{ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO };
    145 int nfs_ticks;
    146 int nfs_commitsize;
    147 
    148 /* NFS client/server stats. */
    149 struct nfsstats nfsstats;
    150 
    151 /*
    152  * Mapping of old NFS Version 2 RPC numbers to generic numbers.
    153  */
    154 const int nfsv3_procid[NFS_NPROCS] = {
    155 	NFSPROC_NULL,
    156 	NFSPROC_GETATTR,
    157 	NFSPROC_SETATTR,
    158 	NFSPROC_NOOP,
    159 	NFSPROC_LOOKUP,
    160 	NFSPROC_READLINK,
    161 	NFSPROC_READ,
    162 	NFSPROC_NOOP,
    163 	NFSPROC_WRITE,
    164 	NFSPROC_CREATE,
    165 	NFSPROC_REMOVE,
    166 	NFSPROC_RENAME,
    167 	NFSPROC_LINK,
    168 	NFSPROC_SYMLINK,
    169 	NFSPROC_MKDIR,
    170 	NFSPROC_RMDIR,
    171 	NFSPROC_READDIR,
    172 	NFSPROC_FSSTAT,
    173 	NFSPROC_NOOP,
    174 	NFSPROC_NOOP,
    175 	NFSPROC_NOOP,
    176 	NFSPROC_NOOP,
    177 	NFSPROC_NOOP,
    178 	NFSPROC_NOOP,
    179 	NFSPROC_NOOP,
    180 	NFSPROC_NOOP
    181 };
    182 
    183 /*
    184  * and the reverse mapping from generic to Version 2 procedure numbers
    185  */
    186 const int nfsv2_procid[NFS_NPROCS] = {
    187 	NFSV2PROC_NULL,
    188 	NFSV2PROC_GETATTR,
    189 	NFSV2PROC_SETATTR,
    190 	NFSV2PROC_LOOKUP,
    191 	NFSV2PROC_NOOP,
    192 	NFSV2PROC_READLINK,
    193 	NFSV2PROC_READ,
    194 	NFSV2PROC_WRITE,
    195 	NFSV2PROC_CREATE,
    196 	NFSV2PROC_MKDIR,
    197 	NFSV2PROC_SYMLINK,
    198 	NFSV2PROC_CREATE,
    199 	NFSV2PROC_REMOVE,
    200 	NFSV2PROC_RMDIR,
    201 	NFSV2PROC_RENAME,
    202 	NFSV2PROC_LINK,
    203 	NFSV2PROC_READDIR,
    204 	NFSV2PROC_NOOP,
    205 	NFSV2PROC_STATFS,
    206 	NFSV2PROC_NOOP,
    207 	NFSV2PROC_NOOP,
    208 	NFSV2PROC_NOOP,
    209 	NFSV2PROC_NOOP,
    210 	NFSV2PROC_NOOP,
    211 	NFSV2PROC_NOOP,
    212 	NFSV2PROC_NOOP,
    213 };
    214 
    215 /*
    216  * Maps errno values to nfs error numbers.
    217  * Use NFSERR_IO as the catch all for ones not specifically defined in
    218  * RFC 1094.
    219  */
    220 static const u_char nfsrv_v2errmap[ELAST] = {
    221   NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    222   NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    223   NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
    224   NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
    225   NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    226   NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
    227   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    228   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    229   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    230   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    231   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    232   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    233   NFSERR_IO,	NFSERR_IO,	NFSERR_NAMETOL,	NFSERR_IO,	NFSERR_IO,
    234   NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
    235   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    236   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    237   NFSERR_IO,	NFSERR_IO,
    238 };
    239 
    240 /*
    241  * Maps errno values to nfs error numbers.
    242  * Although it is not obvious whether or not NFS clients really care if
    243  * a returned error value is in the specified list for the procedure, the
    244  * safest thing to do is filter them appropriately. For Version 2, the
    245  * X/Open XNFS document is the only specification that defines error values
    246  * for each RPC (The RFC simply lists all possible error values for all RPCs),
    247  * so I have decided to not do this for Version 2.
    248  * The first entry is the default error return and the rest are the valid
    249  * errors for that RPC in increasing numeric order.
    250  */
    251 static const short nfsv3err_null[] = {
    252 	0,
    253 	0,
    254 };
    255 
    256 static const short nfsv3err_getattr[] = {
    257 	NFSERR_IO,
    258 	NFSERR_IO,
    259 	NFSERR_STALE,
    260 	NFSERR_BADHANDLE,
    261 	NFSERR_SERVERFAULT,
    262 	0,
    263 };
    264 
    265 static const short nfsv3err_setattr[] = {
    266 	NFSERR_IO,
    267 	NFSERR_PERM,
    268 	NFSERR_IO,
    269 	NFSERR_ACCES,
    270 	NFSERR_INVAL,
    271 	NFSERR_NOSPC,
    272 	NFSERR_ROFS,
    273 	NFSERR_DQUOT,
    274 	NFSERR_STALE,
    275 	NFSERR_BADHANDLE,
    276 	NFSERR_NOT_SYNC,
    277 	NFSERR_SERVERFAULT,
    278 	0,
    279 };
    280 
    281 static const short nfsv3err_lookup[] = {
    282 	NFSERR_IO,
    283 	NFSERR_NOENT,
    284 	NFSERR_IO,
    285 	NFSERR_ACCES,
    286 	NFSERR_NOTDIR,
    287 	NFSERR_NAMETOL,
    288 	NFSERR_STALE,
    289 	NFSERR_BADHANDLE,
    290 	NFSERR_SERVERFAULT,
    291 	0,
    292 };
    293 
    294 static const short nfsv3err_access[] = {
    295 	NFSERR_IO,
    296 	NFSERR_IO,
    297 	NFSERR_STALE,
    298 	NFSERR_BADHANDLE,
    299 	NFSERR_SERVERFAULT,
    300 	0,
    301 };
    302 
    303 static const short nfsv3err_readlink[] = {
    304 	NFSERR_IO,
    305 	NFSERR_IO,
    306 	NFSERR_ACCES,
    307 	NFSERR_INVAL,
    308 	NFSERR_STALE,
    309 	NFSERR_BADHANDLE,
    310 	NFSERR_NOTSUPP,
    311 	NFSERR_SERVERFAULT,
    312 	0,
    313 };
    314 
    315 static const short nfsv3err_read[] = {
    316 	NFSERR_IO,
    317 	NFSERR_IO,
    318 	NFSERR_NXIO,
    319 	NFSERR_ACCES,
    320 	NFSERR_INVAL,
    321 	NFSERR_STALE,
    322 	NFSERR_BADHANDLE,
    323 	NFSERR_SERVERFAULT,
    324 	NFSERR_JUKEBOX,
    325 	0,
    326 };
    327 
    328 static const short nfsv3err_write[] = {
    329 	NFSERR_IO,
    330 	NFSERR_IO,
    331 	NFSERR_ACCES,
    332 	NFSERR_INVAL,
    333 	NFSERR_FBIG,
    334 	NFSERR_NOSPC,
    335 	NFSERR_ROFS,
    336 	NFSERR_DQUOT,
    337 	NFSERR_STALE,
    338 	NFSERR_BADHANDLE,
    339 	NFSERR_SERVERFAULT,
    340 	NFSERR_JUKEBOX,
    341 	0,
    342 };
    343 
    344 static const short nfsv3err_create[] = {
    345 	NFSERR_IO,
    346 	NFSERR_IO,
    347 	NFSERR_ACCES,
    348 	NFSERR_EXIST,
    349 	NFSERR_NOTDIR,
    350 	NFSERR_NOSPC,
    351 	NFSERR_ROFS,
    352 	NFSERR_NAMETOL,
    353 	NFSERR_DQUOT,
    354 	NFSERR_STALE,
    355 	NFSERR_BADHANDLE,
    356 	NFSERR_NOTSUPP,
    357 	NFSERR_SERVERFAULT,
    358 	0,
    359 };
    360 
    361 static const short nfsv3err_mkdir[] = {
    362 	NFSERR_IO,
    363 	NFSERR_IO,
    364 	NFSERR_ACCES,
    365 	NFSERR_EXIST,
    366 	NFSERR_NOTDIR,
    367 	NFSERR_NOSPC,
    368 	NFSERR_ROFS,
    369 	NFSERR_NAMETOL,
    370 	NFSERR_DQUOT,
    371 	NFSERR_STALE,
    372 	NFSERR_BADHANDLE,
    373 	NFSERR_NOTSUPP,
    374 	NFSERR_SERVERFAULT,
    375 	0,
    376 };
    377 
    378 static const short nfsv3err_symlink[] = {
    379 	NFSERR_IO,
    380 	NFSERR_IO,
    381 	NFSERR_ACCES,
    382 	NFSERR_EXIST,
    383 	NFSERR_NOTDIR,
    384 	NFSERR_NOSPC,
    385 	NFSERR_ROFS,
    386 	NFSERR_NAMETOL,
    387 	NFSERR_DQUOT,
    388 	NFSERR_STALE,
    389 	NFSERR_BADHANDLE,
    390 	NFSERR_NOTSUPP,
    391 	NFSERR_SERVERFAULT,
    392 	0,
    393 };
    394 
    395 static const short nfsv3err_mknod[] = {
    396 	NFSERR_IO,
    397 	NFSERR_IO,
    398 	NFSERR_ACCES,
    399 	NFSERR_EXIST,
    400 	NFSERR_NOTDIR,
    401 	NFSERR_NOSPC,
    402 	NFSERR_ROFS,
    403 	NFSERR_NAMETOL,
    404 	NFSERR_DQUOT,
    405 	NFSERR_STALE,
    406 	NFSERR_BADHANDLE,
    407 	NFSERR_NOTSUPP,
    408 	NFSERR_SERVERFAULT,
    409 	NFSERR_BADTYPE,
    410 	0,
    411 };
    412 
    413 static const short nfsv3err_remove[] = {
    414 	NFSERR_IO,
    415 	NFSERR_NOENT,
    416 	NFSERR_IO,
    417 	NFSERR_ACCES,
    418 	NFSERR_NOTDIR,
    419 	NFSERR_ROFS,
    420 	NFSERR_NAMETOL,
    421 	NFSERR_STALE,
    422 	NFSERR_BADHANDLE,
    423 	NFSERR_SERVERFAULT,
    424 	0,
    425 };
    426 
    427 static const short nfsv3err_rmdir[] = {
    428 	NFSERR_IO,
    429 	NFSERR_NOENT,
    430 	NFSERR_IO,
    431 	NFSERR_ACCES,
    432 	NFSERR_EXIST,
    433 	NFSERR_NOTDIR,
    434 	NFSERR_INVAL,
    435 	NFSERR_ROFS,
    436 	NFSERR_NAMETOL,
    437 	NFSERR_NOTEMPTY,
    438 	NFSERR_STALE,
    439 	NFSERR_BADHANDLE,
    440 	NFSERR_NOTSUPP,
    441 	NFSERR_SERVERFAULT,
    442 	0,
    443 };
    444 
    445 static const short nfsv3err_rename[] = {
    446 	NFSERR_IO,
    447 	NFSERR_NOENT,
    448 	NFSERR_IO,
    449 	NFSERR_ACCES,
    450 	NFSERR_EXIST,
    451 	NFSERR_XDEV,
    452 	NFSERR_NOTDIR,
    453 	NFSERR_ISDIR,
    454 	NFSERR_INVAL,
    455 	NFSERR_NOSPC,
    456 	NFSERR_ROFS,
    457 	NFSERR_MLINK,
    458 	NFSERR_NAMETOL,
    459 	NFSERR_NOTEMPTY,
    460 	NFSERR_DQUOT,
    461 	NFSERR_STALE,
    462 	NFSERR_BADHANDLE,
    463 	NFSERR_NOTSUPP,
    464 	NFSERR_SERVERFAULT,
    465 	0,
    466 };
    467 
    468 static const short nfsv3err_link[] = {
    469 	NFSERR_IO,
    470 	NFSERR_IO,
    471 	NFSERR_ACCES,
    472 	NFSERR_EXIST,
    473 	NFSERR_XDEV,
    474 	NFSERR_NOTDIR,
    475 	NFSERR_INVAL,
    476 	NFSERR_NOSPC,
    477 	NFSERR_ROFS,
    478 	NFSERR_MLINK,
    479 	NFSERR_NAMETOL,
    480 	NFSERR_DQUOT,
    481 	NFSERR_STALE,
    482 	NFSERR_BADHANDLE,
    483 	NFSERR_NOTSUPP,
    484 	NFSERR_SERVERFAULT,
    485 	0,
    486 };
    487 
    488 static const short nfsv3err_readdir[] = {
    489 	NFSERR_IO,
    490 	NFSERR_IO,
    491 	NFSERR_ACCES,
    492 	NFSERR_NOTDIR,
    493 	NFSERR_STALE,
    494 	NFSERR_BADHANDLE,
    495 	NFSERR_BAD_COOKIE,
    496 	NFSERR_TOOSMALL,
    497 	NFSERR_SERVERFAULT,
    498 	0,
    499 };
    500 
    501 static const short nfsv3err_readdirplus[] = {
    502 	NFSERR_IO,
    503 	NFSERR_IO,
    504 	NFSERR_ACCES,
    505 	NFSERR_NOTDIR,
    506 	NFSERR_STALE,
    507 	NFSERR_BADHANDLE,
    508 	NFSERR_BAD_COOKIE,
    509 	NFSERR_NOTSUPP,
    510 	NFSERR_TOOSMALL,
    511 	NFSERR_SERVERFAULT,
    512 	0,
    513 };
    514 
    515 static const short nfsv3err_fsstat[] = {
    516 	NFSERR_IO,
    517 	NFSERR_IO,
    518 	NFSERR_STALE,
    519 	NFSERR_BADHANDLE,
    520 	NFSERR_SERVERFAULT,
    521 	0,
    522 };
    523 
    524 static const short nfsv3err_fsinfo[] = {
    525 	NFSERR_STALE,
    526 	NFSERR_STALE,
    527 	NFSERR_BADHANDLE,
    528 	NFSERR_SERVERFAULT,
    529 	0,
    530 };
    531 
    532 static const short nfsv3err_pathconf[] = {
    533 	NFSERR_STALE,
    534 	NFSERR_STALE,
    535 	NFSERR_BADHANDLE,
    536 	NFSERR_SERVERFAULT,
    537 	0,
    538 };
    539 
    540 static const short nfsv3err_commit[] = {
    541 	NFSERR_IO,
    542 	NFSERR_IO,
    543 	NFSERR_STALE,
    544 	NFSERR_BADHANDLE,
    545 	NFSERR_SERVERFAULT,
    546 	0,
    547 };
    548 
    549 static const short * const nfsrv_v3errmap[] = {
    550 	nfsv3err_null,
    551 	nfsv3err_getattr,
    552 	nfsv3err_setattr,
    553 	nfsv3err_lookup,
    554 	nfsv3err_access,
    555 	nfsv3err_readlink,
    556 	nfsv3err_read,
    557 	nfsv3err_write,
    558 	nfsv3err_create,
    559 	nfsv3err_mkdir,
    560 	nfsv3err_symlink,
    561 	nfsv3err_mknod,
    562 	nfsv3err_remove,
    563 	nfsv3err_rmdir,
    564 	nfsv3err_rename,
    565 	nfsv3err_link,
    566 	nfsv3err_readdir,
    567 	nfsv3err_readdirplus,
    568 	nfsv3err_fsstat,
    569 	nfsv3err_fsinfo,
    570 	nfsv3err_pathconf,
    571 	nfsv3err_commit,
    572 };
    573 
    574 extern struct nfsrtt nfsrtt;
    575 extern time_t nqnfsstarttime;
    576 extern int nqsrv_clockskew;
    577 extern int nqsrv_writeslack;
    578 extern int nqsrv_maxlease;
    579 extern const int nqnfs_piggy[NFS_NPROCS];
    580 extern struct nfsnodehashhead *nfsnodehashtbl;
    581 extern u_long nfsnodehash;
    582 
    583 LIST_HEAD(nfsnodehashhead, nfsnode);
    584 u_long nfsdirhashmask;
    585 
    586 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
    587 
    588 /*
    589  * Create the header for an rpc request packet
    590  * The hsiz is the size of the rest of the nfs request header.
    591  * (just used to decide if a cluster is a good idea)
    592  */
    593 struct mbuf *
    594 nfsm_reqh(vp, procid, hsiz, bposp)
    595 	struct vnode *vp;
    596 	u_long procid;
    597 	int hsiz;
    598 	caddr_t *bposp;
    599 {
    600 	struct mbuf *mb;
    601 	caddr_t bpos;
    602 	struct nfsmount *nmp;
    603 #ifndef NFS_V2_ONLY
    604 	u_int32_t *tl;
    605 	struct mbuf *mb2;
    606 	int nqflag;
    607 #endif
    608 
    609 	MGET(mb, M_WAIT, MT_DATA);
    610 	if (hsiz >= MINCLSIZE)
    611 		MCLGET(mb, M_WAIT);
    612 	mb->m_len = 0;
    613 	bpos = mtod(mb, caddr_t);
    614 
    615 	/*
    616 	 * For NQNFS, add lease request.
    617 	 */
    618 	if (vp) {
    619 		nmp = VFSTONFS(vp->v_mount);
    620 #ifndef NFS_V2_ONLY
    621 		if (nmp->nm_flag & NFSMNT_NQNFS) {
    622 			nqflag = NQNFS_NEEDLEASE(vp, procid);
    623 			if (nqflag) {
    624 				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
    625 				*tl++ = txdr_unsigned(nqflag);
    626 				*tl = txdr_unsigned(nmp->nm_leaseterm);
    627 			} else {
    628 				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
    629 				*tl = 0;
    630 			}
    631 		}
    632 #endif
    633 	}
    634 	/* Finally, return values */
    635 	*bposp = bpos;
    636 	return (mb);
    637 }
    638 
    639 /*
    640  * Build the RPC header and fill in the authorization info.
    641  * The authorization string argument is only used when the credentials
    642  * come from outside of the kernel.
    643  * Returns the head of the mbuf list.
    644  */
    645 struct mbuf *
    646 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
    647 	verf_str, mrest, mrest_len, mbp, xidp)
    648 	struct ucred *cr;
    649 	int nmflag;
    650 	int procid;
    651 	int auth_type;
    652 	int auth_len;
    653 	char *auth_str;
    654 	int verf_len;
    655 	char *verf_str;
    656 	struct mbuf *mrest;
    657 	int mrest_len;
    658 	struct mbuf **mbp;
    659 	u_int32_t *xidp;
    660 {
    661 	struct mbuf *mb;
    662 	u_int32_t *tl;
    663 	caddr_t bpos;
    664 	int i;
    665 	struct mbuf *mreq, *mb2;
    666 	int siz, grpsiz, authsiz;
    667 	struct timeval tv;
    668 	static u_int32_t base;
    669 
    670 	authsiz = nfsm_rndup(auth_len);
    671 	MGETHDR(mb, M_WAIT, MT_DATA);
    672 	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
    673 		MCLGET(mb, M_WAIT);
    674 	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
    675 		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
    676 	} else {
    677 		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
    678 	}
    679 	mb->m_len = 0;
    680 	mreq = mb;
    681 	bpos = mtod(mb, caddr_t);
    682 
    683 	/*
    684 	 * First the RPC header.
    685 	 */
    686 	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
    687 
    688 	/*
    689 	 * derive initial xid from system time
    690 	 * XXX time is invalid if root not yet mounted
    691 	 */
    692 	if (!base && (rootvp)) {
    693 		microtime(&tv);
    694 		base = tv.tv_sec << 12;
    695 		nfs_xid = base;
    696 	}
    697 	/*
    698 	 * Skip zero xid if it should ever happen.
    699 	 */
    700 	if (++nfs_xid == 0)
    701 		nfs_xid++;
    702 
    703 	*tl++ = *xidp = txdr_unsigned(nfs_xid);
    704 	*tl++ = rpc_call;
    705 	*tl++ = rpc_vers;
    706 	if (nmflag & NFSMNT_NQNFS) {
    707 		*tl++ = txdr_unsigned(NQNFS_PROG);
    708 		*tl++ = txdr_unsigned(NQNFS_VER3);
    709 	} else {
    710 		*tl++ = txdr_unsigned(NFS_PROG);
    711 		if (nmflag & NFSMNT_NFSV3)
    712 			*tl++ = txdr_unsigned(NFS_VER3);
    713 		else
    714 			*tl++ = txdr_unsigned(NFS_VER2);
    715 	}
    716 	if (nmflag & NFSMNT_NFSV3)
    717 		*tl++ = txdr_unsigned(procid);
    718 	else
    719 		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
    720 
    721 	/*
    722 	 * And then the authorization cred.
    723 	 */
    724 	*tl++ = txdr_unsigned(auth_type);
    725 	*tl = txdr_unsigned(authsiz);
    726 	switch (auth_type) {
    727 	case RPCAUTH_UNIX:
    728 		nfsm_build(tl, u_int32_t *, auth_len);
    729 		*tl++ = 0;		/* stamp ?? */
    730 		*tl++ = 0;		/* NULL hostname */
    731 		*tl++ = txdr_unsigned(cr->cr_uid);
    732 		*tl++ = txdr_unsigned(cr->cr_gid);
    733 		grpsiz = (auth_len >> 2) - 5;
    734 		*tl++ = txdr_unsigned(grpsiz);
    735 		for (i = 0; i < grpsiz; i++)
    736 			*tl++ = txdr_unsigned(cr->cr_groups[i]);
    737 		break;
    738 	case RPCAUTH_KERB4:
    739 		siz = auth_len;
    740 		while (siz > 0) {
    741 			if (M_TRAILINGSPACE(mb) == 0) {
    742 				MGET(mb2, M_WAIT, MT_DATA);
    743 				if (siz >= MINCLSIZE)
    744 					MCLGET(mb2, M_WAIT);
    745 				mb->m_next = mb2;
    746 				mb = mb2;
    747 				mb->m_len = 0;
    748 				bpos = mtod(mb, caddr_t);
    749 			}
    750 			i = min(siz, M_TRAILINGSPACE(mb));
    751 			memcpy(bpos, auth_str, i);
    752 			mb->m_len += i;
    753 			auth_str += i;
    754 			bpos += i;
    755 			siz -= i;
    756 		}
    757 		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
    758 			for (i = 0; i < siz; i++)
    759 				*bpos++ = '\0';
    760 			mb->m_len += siz;
    761 		}
    762 		break;
    763 	};
    764 
    765 	/*
    766 	 * And the verifier...
    767 	 */
    768 	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
    769 	if (verf_str) {
    770 		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
    771 		*tl = txdr_unsigned(verf_len);
    772 		siz = verf_len;
    773 		while (siz > 0) {
    774 			if (M_TRAILINGSPACE(mb) == 0) {
    775 				MGET(mb2, M_WAIT, MT_DATA);
    776 				if (siz >= MINCLSIZE)
    777 					MCLGET(mb2, M_WAIT);
    778 				mb->m_next = mb2;
    779 				mb = mb2;
    780 				mb->m_len = 0;
    781 				bpos = mtod(mb, caddr_t);
    782 			}
    783 			i = min(siz, M_TRAILINGSPACE(mb));
    784 			memcpy(bpos, verf_str, i);
    785 			mb->m_len += i;
    786 			verf_str += i;
    787 			bpos += i;
    788 			siz -= i;
    789 		}
    790 		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
    791 			for (i = 0; i < siz; i++)
    792 				*bpos++ = '\0';
    793 			mb->m_len += siz;
    794 		}
    795 	} else {
    796 		*tl++ = txdr_unsigned(RPCAUTH_NULL);
    797 		*tl = 0;
    798 	}
    799 	mb->m_next = mrest;
    800 	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
    801 	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
    802 	*mbp = mb;
    803 	return (mreq);
    804 }
    805 
    806 /*
    807  * copies mbuf chain to the uio scatter/gather list
    808  */
    809 int
    810 nfsm_mbuftouio(mrep, uiop, siz, dpos)
    811 	struct mbuf **mrep;
    812 	struct uio *uiop;
    813 	int siz;
    814 	caddr_t *dpos;
    815 {
    816 	char *mbufcp, *uiocp;
    817 	int xfer, left, len;
    818 	struct mbuf *mp;
    819 	long uiosiz, rem;
    820 	int error = 0;
    821 
    822 	mp = *mrep;
    823 	mbufcp = *dpos;
    824 	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
    825 	rem = nfsm_rndup(siz)-siz;
    826 	while (siz > 0) {
    827 		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
    828 			return (EFBIG);
    829 		left = uiop->uio_iov->iov_len;
    830 		uiocp = uiop->uio_iov->iov_base;
    831 		if (left > siz)
    832 			left = siz;
    833 		uiosiz = left;
    834 		while (left > 0) {
    835 			while (len == 0) {
    836 				mp = mp->m_next;
    837 				if (mp == NULL)
    838 					return (EBADRPC);
    839 				mbufcp = mtod(mp, caddr_t);
    840 				len = mp->m_len;
    841 			}
    842 			xfer = (left > len) ? len : left;
    843 #ifdef notdef
    844 			/* Not Yet.. */
    845 			if (uiop->uio_iov->iov_op != NULL)
    846 				(*(uiop->uio_iov->iov_op))
    847 				(mbufcp, uiocp, xfer);
    848 			else
    849 #endif
    850 			if (uiop->uio_segflg == UIO_SYSSPACE)
    851 				memcpy(uiocp, mbufcp, xfer);
    852 			else
    853 				copyout(mbufcp, uiocp, xfer);
    854 			left -= xfer;
    855 			len -= xfer;
    856 			mbufcp += xfer;
    857 			uiocp += xfer;
    858 			uiop->uio_offset += xfer;
    859 			uiop->uio_resid -= xfer;
    860 		}
    861 		if (uiop->uio_iov->iov_len <= siz) {
    862 			uiop->uio_iovcnt--;
    863 			uiop->uio_iov++;
    864 		} else {
    865 			uiop->uio_iov->iov_base =
    866 			    (caddr_t)uiop->uio_iov->iov_base + uiosiz;
    867 			uiop->uio_iov->iov_len -= uiosiz;
    868 		}
    869 		siz -= uiosiz;
    870 	}
    871 	*dpos = mbufcp;
    872 	*mrep = mp;
    873 	if (rem > 0) {
    874 		if (len < rem)
    875 			error = nfs_adv(mrep, dpos, rem, len);
    876 		else
    877 			*dpos += rem;
    878 	}
    879 	return (error);
    880 }
    881 
    882 /*
    883  * copies a uio scatter/gather list to an mbuf chain.
    884  * NOTE: can ony handle iovcnt == 1
    885  */
    886 int
    887 nfsm_uiotombuf(uiop, mq, siz, bpos)
    888 	struct uio *uiop;
    889 	struct mbuf **mq;
    890 	int siz;
    891 	caddr_t *bpos;
    892 {
    893 	char *uiocp;
    894 	struct mbuf *mp, *mp2;
    895 	int xfer, left, mlen;
    896 	int uiosiz, clflg, rem;
    897 	char *cp;
    898 
    899 #ifdef DIAGNOSTIC
    900 	if (uiop->uio_iovcnt != 1)
    901 		panic("nfsm_uiotombuf: iovcnt != 1");
    902 #endif
    903 
    904 	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
    905 		clflg = 1;
    906 	else
    907 		clflg = 0;
    908 	rem = nfsm_rndup(siz)-siz;
    909 	mp = mp2 = *mq;
    910 	while (siz > 0) {
    911 		left = uiop->uio_iov->iov_len;
    912 		uiocp = uiop->uio_iov->iov_base;
    913 		if (left > siz)
    914 			left = siz;
    915 		uiosiz = left;
    916 		while (left > 0) {
    917 			mlen = M_TRAILINGSPACE(mp);
    918 			if (mlen == 0) {
    919 				MGET(mp, M_WAIT, MT_DATA);
    920 				if (clflg)
    921 					MCLGET(mp, M_WAIT);
    922 				mp->m_len = 0;
    923 				mp2->m_next = mp;
    924 				mp2 = mp;
    925 				mlen = M_TRAILINGSPACE(mp);
    926 			}
    927 			xfer = (left > mlen) ? mlen : left;
    928 #ifdef notdef
    929 			/* Not Yet.. */
    930 			if (uiop->uio_iov->iov_op != NULL)
    931 				(*(uiop->uio_iov->iov_op))
    932 				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
    933 			else
    934 #endif
    935 			if (uiop->uio_segflg == UIO_SYSSPACE)
    936 				memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer);
    937 			else
    938 				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
    939 			mp->m_len += xfer;
    940 			left -= xfer;
    941 			uiocp += xfer;
    942 			uiop->uio_offset += xfer;
    943 			uiop->uio_resid -= xfer;
    944 		}
    945 		uiop->uio_iov->iov_base = (caddr_t)uiop->uio_iov->iov_base +
    946 		    uiosiz;
    947 		uiop->uio_iov->iov_len -= uiosiz;
    948 		siz -= uiosiz;
    949 	}
    950 	if (rem > 0) {
    951 		if (rem > M_TRAILINGSPACE(mp)) {
    952 			MGET(mp, M_WAIT, MT_DATA);
    953 			mp->m_len = 0;
    954 			mp2->m_next = mp;
    955 		}
    956 		cp = mtod(mp, caddr_t)+mp->m_len;
    957 		for (left = 0; left < rem; left++)
    958 			*cp++ = '\0';
    959 		mp->m_len += rem;
    960 		*bpos = cp;
    961 	} else
    962 		*bpos = mtod(mp, caddr_t)+mp->m_len;
    963 	*mq = mp;
    964 	return (0);
    965 }
    966 
    967 /*
    968  * Get at least "siz" bytes of correctly aligned data.
    969  * When called the mbuf pointers are not necessarily correct,
    970  * dsosp points to what ought to be in m_data and left contains
    971  * what ought to be in m_len.
    972  * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
    973  * cases. (The macros use the vars. dpos and dpos2)
    974  */
    975 int
    976 nfsm_disct(mdp, dposp, siz, left, cp2)
    977 	struct mbuf **mdp;
    978 	caddr_t *dposp;
    979 	int siz;
    980 	int left;
    981 	caddr_t *cp2;
    982 {
    983 	struct mbuf *m1, *m2;
    984 	struct mbuf *havebuf = NULL;
    985 	caddr_t src = *dposp;
    986 	caddr_t dst;
    987 	int len;
    988 
    989 #ifdef DEBUG
    990 	if (left < 0)
    991 		panic("nfsm_disct: left < 0");
    992 #endif
    993 	m1 = *mdp;
    994 	/*
    995 	 * Skip through the mbuf chain looking for an mbuf with
    996 	 * some data. If the first mbuf found has enough data
    997 	 * and it is correctly aligned return it.
    998 	 */
    999 	while (left == 0) {
   1000 		havebuf = m1;
   1001 		*mdp = m1 = m1->m_next;
   1002 		if (m1 == NULL)
   1003 			return (EBADRPC);
   1004 		src = mtod(m1, caddr_t);
   1005 		left = m1->m_len;
   1006 		/*
   1007 		 * If we start a new mbuf and it is big enough
   1008 		 * and correctly aligned just return it, don't
   1009 		 * do any pull up.
   1010 		 */
   1011 		if (left >= siz && nfsm_aligned(src)) {
   1012 			*cp2 = src;
   1013 			*dposp = src + siz;
   1014 			return (0);
   1015 		}
   1016 	}
   1017 	if (m1->m_flags & M_EXT) {
   1018 		if (havebuf) {
   1019 			/* If the first mbuf with data has external data
   1020 			 * and there is a previous empty mbuf use it
   1021 			 * to move the data into.
   1022 			 */
   1023 			m2 = m1;
   1024 			*mdp = m1 = havebuf;
   1025 			if (m1->m_flags & M_EXT) {
   1026 				MEXTREMOVE(m1);
   1027 			}
   1028 		} else {
   1029 			/*
   1030 			 * If the first mbuf has a external data
   1031 			 * and there is no previous empty mbuf
   1032 			 * allocate a new mbuf and move the external
   1033 			 * data to the new mbuf. Also make the first
   1034 			 * mbuf look empty.
   1035 			 */
   1036 			m2 = m_get(M_WAIT, MT_DATA);
   1037 			m2->m_ext = m1->m_ext;
   1038 			m2->m_data = src;
   1039 			m2->m_len = left;
   1040 			MCLADDREFERENCE(m1, m2);
   1041 			MEXTREMOVE(m1);
   1042 			m2->m_next = m1->m_next;
   1043 			m1->m_next = m2;
   1044 		}
   1045 		m1->m_len = 0;
   1046 		dst = m1->m_dat;
   1047 	} else {
   1048 		/*
   1049 		 * If the first mbuf has no external data
   1050 		 * move the data to the front of the mbuf.
   1051 		 */
   1052 		if ((dst = m1->m_dat) != src)
   1053 			memmove(dst, src, left);
   1054 		dst += left;
   1055 		m1->m_len = left;
   1056 		m2 = m1->m_next;
   1057 	}
   1058 	m1->m_flags &= ~M_PKTHDR;
   1059 	*cp2 = m1->m_data = m1->m_dat;   /* data is at beginning of buffer */
   1060 	*dposp = mtod(m1, caddr_t) + siz;
   1061 	/*
   1062 	 * Loop through mbufs pulling data up into first mbuf until
   1063 	 * the first mbuf is full or there is no more data to
   1064 	 * pullup.
   1065 	 */
   1066 	while ((len = (MLEN - m1->m_len)) != 0 && m2) {
   1067 		if ((len = min(len, m2->m_len)) != 0)
   1068 			memcpy(dst, m2->m_data, len);
   1069 		m1->m_len += len;
   1070 		dst += len;
   1071 		m2->m_data += len;
   1072 		m2->m_len -= len;
   1073 		m2 = m2->m_next;
   1074 	}
   1075 	if (m1->m_len < siz)
   1076 		return (EBADRPC);
   1077 	return (0);
   1078 }
   1079 
   1080 /*
   1081  * Advance the position in the mbuf chain.
   1082  */
   1083 int
   1084 nfs_adv(mdp, dposp, offs, left)
   1085 	struct mbuf **mdp;
   1086 	caddr_t *dposp;
   1087 	int offs;
   1088 	int left;
   1089 {
   1090 	struct mbuf *m;
   1091 	int s;
   1092 
   1093 	m = *mdp;
   1094 	s = left;
   1095 	while (s < offs) {
   1096 		offs -= s;
   1097 		m = m->m_next;
   1098 		if (m == NULL)
   1099 			return (EBADRPC);
   1100 		s = m->m_len;
   1101 	}
   1102 	*mdp = m;
   1103 	*dposp = mtod(m, caddr_t)+offs;
   1104 	return (0);
   1105 }
   1106 
   1107 /*
   1108  * Copy a string into mbufs for the hard cases...
   1109  */
   1110 int
   1111 nfsm_strtmbuf(mb, bpos, cp, siz)
   1112 	struct mbuf **mb;
   1113 	char **bpos;
   1114 	const char *cp;
   1115 	long siz;
   1116 {
   1117 	struct mbuf *m1 = NULL, *m2;
   1118 	long left, xfer, len, tlen;
   1119 	u_int32_t *tl;
   1120 	int putsize;
   1121 
   1122 	putsize = 1;
   1123 	m2 = *mb;
   1124 	left = M_TRAILINGSPACE(m2);
   1125 	if (left > 0) {
   1126 		tl = ((u_int32_t *)(*bpos));
   1127 		*tl++ = txdr_unsigned(siz);
   1128 		putsize = 0;
   1129 		left -= NFSX_UNSIGNED;
   1130 		m2->m_len += NFSX_UNSIGNED;
   1131 		if (left > 0) {
   1132 			memcpy((caddr_t) tl, cp, left);
   1133 			siz -= left;
   1134 			cp += left;
   1135 			m2->m_len += left;
   1136 			left = 0;
   1137 		}
   1138 	}
   1139 	/* Loop around adding mbufs */
   1140 	while (siz > 0) {
   1141 		MGET(m1, M_WAIT, MT_DATA);
   1142 		if (siz > MLEN)
   1143 			MCLGET(m1, M_WAIT);
   1144 		m1->m_len = NFSMSIZ(m1);
   1145 		m2->m_next = m1;
   1146 		m2 = m1;
   1147 		tl = mtod(m1, u_int32_t *);
   1148 		tlen = 0;
   1149 		if (putsize) {
   1150 			*tl++ = txdr_unsigned(siz);
   1151 			m1->m_len -= NFSX_UNSIGNED;
   1152 			tlen = NFSX_UNSIGNED;
   1153 			putsize = 0;
   1154 		}
   1155 		if (siz < m1->m_len) {
   1156 			len = nfsm_rndup(siz);
   1157 			xfer = siz;
   1158 			if (xfer < len)
   1159 				*(tl+(xfer>>2)) = 0;
   1160 		} else {
   1161 			xfer = len = m1->m_len;
   1162 		}
   1163 		memcpy((caddr_t) tl, cp, xfer);
   1164 		m1->m_len = len+tlen;
   1165 		siz -= xfer;
   1166 		cp += xfer;
   1167 	}
   1168 	*mb = m1;
   1169 	*bpos = mtod(m1, caddr_t)+m1->m_len;
   1170 	return (0);
   1171 }
   1172 
   1173 /*
   1174  * Directory caching routines. They work as follows:
   1175  * - a cache is maintained per VDIR nfsnode.
   1176  * - for each offset cookie that is exported to userspace, and can
   1177  *   thus be thrown back at us as an offset to VOP_READDIR, store
   1178  *   information in the cache.
   1179  * - cached are:
   1180  *   - cookie itself
   1181  *   - blocknumber (essentially just a search key in the buffer cache)
   1182  *   - entry number in block.
   1183  *   - offset cookie of block in which this entry is stored
   1184  *   - 32 bit cookie if NFSMNT_XLATECOOKIE is used.
   1185  * - entries are looked up in a hash table
   1186  * - also maintained is an LRU list of entries, used to determine
   1187  *   which ones to delete if the cache grows too large.
   1188  * - if 32 <-> 64 translation mode is requested for a filesystem,
   1189  *   the cache also functions as a translation table
   1190  * - in the translation case, invalidating the cache does not mean
   1191  *   flushing it, but just marking entries as invalid, except for
   1192  *   the <64bit cookie, 32bitcookie> pair which is still valid, to
   1193  *   still be able to use the cache as a translation table.
   1194  * - 32 bit cookies are uniquely created by combining the hash table
   1195  *   entry value, and one generation count per hash table entry,
   1196  *   incremented each time an entry is appended to the chain.
   1197  * - the cache is invalidated each time a direcory is modified
   1198  * - sanity checks are also done; if an entry in a block turns
   1199  *   out not to have a matching cookie, the cache is invalidated
   1200  *   and a new block starting from the wanted offset is fetched from
   1201  *   the server.
   1202  * - directory entries as read from the server are extended to contain
   1203  *   the 64bit and, optionally, the 32bit cookies, for sanity checking
   1204  *   the cache and exporting them to userspace through the cookie
   1205  *   argument to VOP_READDIR.
   1206  */
   1207 
   1208 u_long
   1209 nfs_dirhash(off)
   1210 	off_t off;
   1211 {
   1212 	int i;
   1213 	char *cp = (char *)&off;
   1214 	u_long sum = 0L;
   1215 
   1216 	for (i = 0 ; i < sizeof (off); i++)
   1217 		sum += *cp++;
   1218 
   1219 	return sum;
   1220 }
   1221 
   1222 void
   1223 nfs_initdircache(vp)
   1224 	struct vnode *vp;
   1225 {
   1226 	struct nfsnode *np = VTONFS(vp);
   1227 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1228 
   1229 	np->n_dircachesize = 0;
   1230 	np->n_dblkno = 1;
   1231 	np->n_dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, M_NFSDIROFF,
   1232 	    M_WAITOK, &nfsdirhashmask);
   1233 	TAILQ_INIT(&np->n_dirchain);
   1234 	if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
   1235 		MALLOC(np->n_dirgens, unsigned *,
   1236 		    NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF,
   1237 		    M_WAITOK);
   1238 		memset((caddr_t)np->n_dirgens, 0,
   1239 		    NFS_DIRHASHSIZ * sizeof (unsigned));
   1240 	}
   1241 }
   1242 
   1243 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0};
   1244 
   1245 struct nfsdircache *
   1246 nfs_searchdircache(vp, off, do32, hashent)
   1247 	struct vnode *vp;
   1248 	off_t off;
   1249 	int do32;
   1250 	int *hashent;
   1251 {
   1252 	struct nfsdirhashhead *ndhp;
   1253 	struct nfsdircache *ndp = NULL;
   1254 	struct nfsnode *np = VTONFS(vp);
   1255 	unsigned ent;
   1256 
   1257 	/*
   1258 	 * Zero is always a valid cookie.
   1259 	 */
   1260 	if (off == 0)
   1261 		return &dzero;
   1262 
   1263 	/*
   1264 	 * We use a 32bit cookie as search key, directly reconstruct
   1265 	 * the hashentry. Else use the hashfunction.
   1266 	 */
   1267 	if (do32) {
   1268 		ent = (u_int32_t)off >> 24;
   1269 		if (ent >= NFS_DIRHASHSIZ)
   1270 			return NULL;
   1271 		ndhp = &np->n_dircache[ent];
   1272 	} else {
   1273 		ndhp = NFSDIRHASH(np, off);
   1274 	}
   1275 
   1276 	if (hashent)
   1277 		*hashent = (int)(ndhp - np->n_dircache);
   1278 	if (do32) {
   1279 		for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) {
   1280 			if (ndp->dc_cookie32 == (u_int32_t)off) {
   1281 				/*
   1282 				 * An invalidated entry will become the
   1283 				 * start of a new block fetched from
   1284 				 * the server.
   1285 				 */
   1286 				if (ndp->dc_blkno == -1) {
   1287 					ndp->dc_blkcookie = ndp->dc_cookie;
   1288 					ndp->dc_blkno = np->n_dblkno++;
   1289 					ndp->dc_entry = 0;
   1290 				}
   1291 				break;
   1292 			}
   1293 		}
   1294 	} else {
   1295 		for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next)
   1296 			if (ndp->dc_cookie == off)
   1297 				break;
   1298 	}
   1299 	return ndp;
   1300 }
   1301 
   1302 
   1303 struct nfsdircache *
   1304 nfs_enterdircache(vp, off, blkoff, en, blkno)
   1305 	struct vnode *vp;
   1306 	off_t off, blkoff;
   1307 	int en;
   1308 	daddr_t blkno;
   1309 {
   1310 	struct nfsnode *np = VTONFS(vp);
   1311 	struct nfsdirhashhead *ndhp;
   1312 	struct nfsdircache *ndp = NULL, *first;
   1313 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1314 	int hashent, gen, overwrite;
   1315 
   1316 	if (!np->n_dircache)
   1317 		/*
   1318 		 * XXX would like to do this in nfs_nget but vtype
   1319 		 * isn't known at that time.
   1320 		 */
   1321 		nfs_initdircache(vp);
   1322 
   1323 	/*
   1324 	 * XXX refuse entries for offset 0. amd(8) erroneously sets
   1325 	 * cookie 0 for the '.' entry, making this necessary. This
   1326 	 * isn't so bad, as 0 is a special case anyway.
   1327 	 */
   1328 	if (off == 0)
   1329 		return &dzero;
   1330 
   1331 	ndp = nfs_searchdircache(vp, off, 0, &hashent);
   1332 
   1333 	if (ndp && ndp->dc_blkno != -1) {
   1334 		/*
   1335 		 * Overwriting an old entry. Check if it's the same.
   1336 		 * If so, just return. If not, remove the old entry.
   1337 		 */
   1338 		if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en)
   1339 			return ndp;
   1340 		TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
   1341 		LIST_REMOVE(ndp, dc_hash);
   1342 		FREE(ndp, M_NFSDIROFF);
   1343 		ndp = 0;
   1344 	}
   1345 
   1346 	ndhp = &np->n_dircache[hashent];
   1347 
   1348 	if (!ndp) {
   1349 		MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF,
   1350 		    M_WAITOK);
   1351 		overwrite = 0;
   1352 		if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
   1353 			/*
   1354 			 * We're allocating a new entry, so bump the
   1355 			 * generation number.
   1356 			 */
   1357 			gen = ++np->n_dirgens[hashent];
   1358 			if (gen == 0) {
   1359 				np->n_dirgens[hashent]++;
   1360 				gen++;
   1361 			}
   1362 			ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff);
   1363 		}
   1364 	} else
   1365 		overwrite = 1;
   1366 
   1367 	/*
   1368 	 * If the entry number is 0, we are at the start of a new block, so
   1369 	 * allocate a new blocknumber.
   1370 	 */
   1371 	if (en == 0)
   1372 		ndp->dc_blkno = np->n_dblkno++;
   1373 	else
   1374 		ndp->dc_blkno = blkno;
   1375 
   1376 	ndp->dc_cookie = off;
   1377 	ndp->dc_blkcookie = blkoff;
   1378 	ndp->dc_entry = en;
   1379 
   1380 	if (overwrite)
   1381 		return ndp;
   1382 
   1383 	/*
   1384 	 * If the maximum directory cookie cache size has been reached
   1385 	 * for this node, take one off the front. The idea is that
   1386 	 * directories are typically read front-to-back once, so that
   1387 	 * the oldest entries can be thrown away without much performance
   1388 	 * loss.
   1389 	 */
   1390 	if (np->n_dircachesize == NFS_MAXDIRCACHE) {
   1391 		first = np->n_dirchain.tqh_first;
   1392 		TAILQ_REMOVE(&np->n_dirchain, first, dc_chain);
   1393 		LIST_REMOVE(first, dc_hash);
   1394 		FREE(first, M_NFSDIROFF);
   1395 	} else
   1396 		np->n_dircachesize++;
   1397 
   1398 	LIST_INSERT_HEAD(ndhp, ndp, dc_hash);
   1399 	TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain);
   1400 	return ndp;
   1401 }
   1402 
   1403 void
   1404 nfs_invaldircache(vp, forcefree)
   1405 	struct vnode *vp;
   1406 	int forcefree;
   1407 {
   1408 	struct nfsnode *np = VTONFS(vp);
   1409 	struct nfsdircache *ndp = NULL;
   1410 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1411 
   1412 #ifdef DIAGNOSTIC
   1413 	if (vp->v_type != VDIR)
   1414 		panic("nfs: invaldircache: not dir");
   1415 #endif
   1416 
   1417 	if (!np->n_dircache)
   1418 		return;
   1419 
   1420 	if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) {
   1421 		while ((ndp = np->n_dirchain.tqh_first)) {
   1422 			TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
   1423 			LIST_REMOVE(ndp, dc_hash);
   1424 			FREE(ndp, M_NFSDIROFF);
   1425 		}
   1426 		np->n_dircachesize = 0;
   1427 		if (forcefree && np->n_dirgens) {
   1428 			FREE(np->n_dirgens, M_NFSDIROFF);
   1429 		}
   1430 	} else {
   1431 		for (ndp = np->n_dirchain.tqh_first; ndp;
   1432 		    ndp = ndp->dc_chain.tqe_next)
   1433 			ndp->dc_blkno = -1;
   1434 	}
   1435 
   1436 	np->n_dblkno = 1;
   1437 }
   1438 
   1439 /*
   1440  * Called once before VFS init to initialize shared and
   1441  * server-specific data structures.
   1442  */
   1443 void
   1444 nfs_init()
   1445 {
   1446 	nfsrtt.pos = 0;
   1447 	rpc_vers = txdr_unsigned(RPC_VER2);
   1448 	rpc_call = txdr_unsigned(RPC_CALL);
   1449 	rpc_reply = txdr_unsigned(RPC_REPLY);
   1450 	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
   1451 	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
   1452 	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
   1453 	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
   1454 	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
   1455 	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
   1456 	nfs_prog = txdr_unsigned(NFS_PROG);
   1457 	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
   1458 	nfs_true = txdr_unsigned(TRUE);
   1459 	nfs_false = txdr_unsigned(FALSE);
   1460 	nfs_xdrneg1 = txdr_unsigned(-1);
   1461 	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
   1462 	if (nfs_ticks < 1)
   1463 		nfs_ticks = 1;
   1464 #ifdef NFSSERVER
   1465 	nfsrv_init(0);			/* Init server data structures */
   1466 	nfsrv_initcache();		/* Init the server request cache */
   1467 #endif /* NFSSERVER */
   1468 
   1469 #if defined(NFSSERVER) || !defined(NFS_V2_ONLY)
   1470 	/*
   1471 	 * Initialize the nqnfs data structures.
   1472 	 */
   1473 	if (nqnfsstarttime == 0) {
   1474 		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
   1475 			+ nqsrv_clockskew + nqsrv_writeslack;
   1476 		NQLOADNOVRAM(nqnfsstarttime);
   1477 		CIRCLEQ_INIT(&nqtimerhead);
   1478 		nqfhhashtbl = hashinit(NQLCHSZ, HASH_LIST, M_NQLEASE,
   1479 		    M_WAITOK, &nqfhhash);
   1480 	}
   1481 #endif
   1482 
   1483 	exithook_establish(nfs_exit, NULL);
   1484 
   1485 	/*
   1486 	 * Initialize reply list and start timer
   1487 	 */
   1488 	TAILQ_INIT(&nfs_reqq);
   1489 	nfs_timer(NULL);
   1490 }
   1491 
   1492 #ifdef NFS
   1493 /*
   1494  * Called once at VFS init to initialize client-specific data structures.
   1495  */
   1496 void
   1497 nfs_vfs_init()
   1498 {
   1499 	nfs_nhinit();			/* Init the nfsnode table */
   1500 	nfs_commitsize = uvmexp.npages << (PAGE_SHIFT - 4);
   1501 }
   1502 
   1503 void
   1504 nfs_vfs_reinit()
   1505 {
   1506 	nfs_nhreinit();
   1507 }
   1508 
   1509 void
   1510 nfs_vfs_done()
   1511 {
   1512 	nfs_nhdone();
   1513 }
   1514 
   1515 /*
   1516  * Attribute cache routines.
   1517  * nfs_loadattrcache() - loads or updates the cache contents from attributes
   1518  *	that are on the mbuf list
   1519  * nfs_getattrcache() - returns valid attributes if found in cache, returns
   1520  *	error otherwise
   1521  */
   1522 
   1523 /*
   1524  * Load the attribute cache (that lives in the nfsnode entry) with
   1525  * the values on the mbuf list and
   1526  * Iff vap not NULL
   1527  *    copy the attributes to *vaper
   1528  */
   1529 int
   1530 nfsm_loadattrcache(vpp, mdp, dposp, vaper)
   1531 	struct vnode **vpp;
   1532 	struct mbuf **mdp;
   1533 	caddr_t *dposp;
   1534 	struct vattr *vaper;
   1535 {
   1536 	int32_t t1;
   1537 	caddr_t cp2;
   1538 	int error = 0;
   1539 	struct mbuf *md;
   1540 	int v3 = NFS_ISV3(*vpp);
   1541 
   1542 	md = *mdp;
   1543 	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
   1544 	error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2);
   1545 	if (error)
   1546 		return (error);
   1547 	return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper);
   1548 }
   1549 
   1550 int
   1551 nfs_loadattrcache(vpp, fp, vaper)
   1552 	struct vnode **vpp;
   1553 	struct nfs_fattr *fp;
   1554 	struct vattr *vaper;
   1555 {
   1556 	struct vnode *vp = *vpp;
   1557 	struct vattr *vap;
   1558 	int v3 = NFS_ISV3(vp);
   1559 	enum vtype vtyp;
   1560 	u_short vmode;
   1561 	struct timespec mtime;
   1562 	struct vnode *nvp;
   1563 	int32_t rdev;
   1564 	struct nfsnode *np;
   1565 	extern int (**spec_nfsv2nodeop_p) __P((void *));
   1566 	uid_t uid;
   1567 	gid_t gid;
   1568 
   1569 	if (v3) {
   1570 		vtyp = nfsv3tov_type(fp->fa_type);
   1571 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
   1572 		rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1),
   1573 			fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2));
   1574 		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
   1575 	} else {
   1576 		vtyp = nfsv2tov_type(fp->fa_type);
   1577 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
   1578 		if (vtyp == VNON || vtyp == VREG)
   1579 			vtyp = IFTOVT(vmode);
   1580 		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
   1581 		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
   1582 
   1583 		/*
   1584 		 * Really ugly NFSv2 kludge.
   1585 		 */
   1586 		if (vtyp == VCHR && rdev == 0xffffffff)
   1587 			vtyp = VFIFO;
   1588 	}
   1589 
   1590 	vmode &= ALLPERMS;
   1591 
   1592 	/*
   1593 	 * If v_type == VNON it is a new node, so fill in the v_type,
   1594 	 * n_mtime fields. Check to see if it represents a special
   1595 	 * device, and if so, check for a possible alias. Once the
   1596 	 * correct vnode has been obtained, fill in the rest of the
   1597 	 * information.
   1598 	 */
   1599 	np = VTONFS(vp);
   1600 	if (vp->v_type == VNON) {
   1601 		vp->v_type = vtyp;
   1602 		if (vp->v_type == VFIFO) {
   1603 			extern int (**fifo_nfsv2nodeop_p) __P((void *));
   1604 			vp->v_op = fifo_nfsv2nodeop_p;
   1605 		}
   1606 		if (vp->v_type == VCHR || vp->v_type == VBLK) {
   1607 			vp->v_op = spec_nfsv2nodeop_p;
   1608 			nvp = checkalias(vp, (dev_t)rdev, vp->v_mount);
   1609 			if (nvp) {
   1610 				/*
   1611 				 * Discard unneeded vnode, but save its nfsnode.
   1612 				 * Since the nfsnode does not have a lock, its
   1613 				 * vnode lock has to be carried over.
   1614 				 */
   1615 				/*
   1616 				 * XXX is the old node sure to be locked here?
   1617 				 */
   1618 				KASSERT(lockstatus(&vp->v_lock) ==
   1619 				    LK_EXCLUSIVE);
   1620 				nvp->v_data = vp->v_data;
   1621 				vp->v_data = NULL;
   1622 				VOP_UNLOCK(vp, 0);
   1623 				vp->v_op = spec_vnodeop_p;
   1624 				vrele(vp);
   1625 				vgone(vp);
   1626 				lockmgr(&nvp->v_lock, LK_EXCLUSIVE,
   1627 				    &nvp->v_interlock);
   1628 				/*
   1629 				 * Reinitialize aliased node.
   1630 				 */
   1631 				np->n_vnode = nvp;
   1632 				*vpp = vp = nvp;
   1633 			}
   1634 		}
   1635 		np->n_mtime = mtime.tv_sec;
   1636 	}
   1637 	uid = fxdr_unsigned(uid_t, fp->fa_uid);
   1638 	gid = fxdr_unsigned(gid_t, fp->fa_gid);
   1639 	vap = np->n_vattr;
   1640 
   1641 	/*
   1642 	 * Invalidate access cache if uid, gid or mode changed.
   1643 	 */
   1644 	if (np->n_accstamp != -1 &&
   1645 	    (gid != vap->va_gid || uid != vap->va_uid || vmode != vap->va_mode))
   1646 		np->n_accstamp = -1;
   1647 
   1648 	vap->va_type = vtyp;
   1649 	vap->va_mode = vmode;
   1650 	vap->va_rdev = (dev_t)rdev;
   1651 	vap->va_mtime = mtime;
   1652 	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
   1653 	switch (vtyp) {
   1654 	case VDIR:
   1655 		vap->va_blocksize = NFS_DIRFRAGSIZ;
   1656 		break;
   1657 	case VBLK:
   1658 		vap->va_blocksize = BLKDEV_IOSIZE;
   1659 		break;
   1660 	case VCHR:
   1661 		vap->va_blocksize = MAXBSIZE;
   1662 		break;
   1663 	default:
   1664 		vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize :
   1665 		    fxdr_unsigned(int32_t, fp->fa2_blocksize);
   1666 		break;
   1667 	}
   1668 	if (v3) {
   1669 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
   1670 		vap->va_uid = uid;
   1671 		vap->va_gid = gid;
   1672 		vap->va_size = fxdr_hyper(&fp->fa3_size);
   1673 		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
   1674 		vap->va_fileid = fxdr_unsigned(int32_t,
   1675 		    fp->fa3_fileid.nfsuquad[1]);
   1676 		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
   1677 		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
   1678 		vap->va_flags = 0;
   1679 		vap->va_filerev = 0;
   1680 	} else {
   1681 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
   1682 		vap->va_uid = uid;
   1683 		vap->va_gid = gid;
   1684 		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
   1685 		vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks)
   1686 		    * NFS_FABLKSIZE;
   1687 		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
   1688 		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
   1689 		vap->va_flags = 0;
   1690 		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
   1691 		    fp->fa2_ctime.nfsv2_sec);
   1692 		vap->va_ctime.tv_nsec = 0;
   1693 		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
   1694 		vap->va_filerev = 0;
   1695 	}
   1696 	if (vap->va_size != np->n_size) {
   1697 		if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) {
   1698 			vap->va_size = np->n_size;
   1699 		} else {
   1700 			np->n_size = vap->va_size;
   1701 			if (vap->va_type == VREG) {
   1702 				uvm_vnp_setsize(vp, np->n_size);
   1703 			}
   1704 		}
   1705 	}
   1706 	np->n_attrstamp = time.tv_sec;
   1707 	if (vaper != NULL) {
   1708 		memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap));
   1709 		if (np->n_flag & NCHG) {
   1710 			if (np->n_flag & NACC)
   1711 				vaper->va_atime = np->n_atim;
   1712 			if (np->n_flag & NUPD)
   1713 				vaper->va_mtime = np->n_mtim;
   1714 		}
   1715 	}
   1716 	return (0);
   1717 }
   1718 
   1719 /*
   1720  * Check the time stamp
   1721  * If the cache is valid, copy contents to *vap and return 0
   1722  * otherwise return an error
   1723  */
   1724 int
   1725 nfs_getattrcache(vp, vaper)
   1726 	struct vnode *vp;
   1727 	struct vattr *vaper;
   1728 {
   1729 	struct nfsnode *np = VTONFS(vp);
   1730 	struct vattr *vap;
   1731 
   1732 	if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) {
   1733 		nfsstats.attrcache_misses++;
   1734 		return (ENOENT);
   1735 	}
   1736 	nfsstats.attrcache_hits++;
   1737 	vap = np->n_vattr;
   1738 	if (vap->va_size != np->n_size) {
   1739 		if (vap->va_type == VREG) {
   1740 			if (np->n_flag & NMODIFIED) {
   1741 				if (vap->va_size < np->n_size)
   1742 					vap->va_size = np->n_size;
   1743 				else
   1744 					np->n_size = vap->va_size;
   1745 			} else
   1746 				np->n_size = vap->va_size;
   1747 			uvm_vnp_setsize(vp, np->n_size);
   1748 		} else
   1749 			np->n_size = vap->va_size;
   1750 	}
   1751 	memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr));
   1752 	if (np->n_flag & NCHG) {
   1753 		if (np->n_flag & NACC)
   1754 			vaper->va_atime = np->n_atim;
   1755 		if (np->n_flag & NUPD)
   1756 			vaper->va_mtime = np->n_mtim;
   1757 	}
   1758 	return (0);
   1759 }
   1760 
   1761 /*
   1762  * Heuristic to see if the server XDR encodes directory cookies or not.
   1763  * it is not supposed to, but a lot of servers may do this. Also, since
   1764  * most/all servers will implement V2 as well, it is expected that they
   1765  * may return just 32 bits worth of cookie information, so we need to
   1766  * find out in which 32 bits this information is available. We do this
   1767  * to avoid trouble with emulated binaries that can't handle 64 bit
   1768  * directory offsets.
   1769  */
   1770 
   1771 void
   1772 nfs_cookieheuristic(vp, flagp, p, cred)
   1773 	struct vnode *vp;
   1774 	int *flagp;
   1775 	struct proc *p;
   1776 	struct ucred *cred;
   1777 {
   1778 	struct uio auio;
   1779 	struct iovec aiov;
   1780 	caddr_t buf, cp;
   1781 	struct dirent *dp;
   1782 	off_t *cookies = NULL, *cop;
   1783 	int error, eof, nc, len;
   1784 
   1785 	MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK);
   1786 
   1787 	aiov.iov_base = buf;
   1788 	aiov.iov_len = NFS_DIRFRAGSIZ;
   1789 	auio.uio_iov = &aiov;
   1790 	auio.uio_iovcnt = 1;
   1791 	auio.uio_rw = UIO_READ;
   1792 	auio.uio_segflg = UIO_SYSSPACE;
   1793 	auio.uio_procp = p;
   1794 	auio.uio_resid = NFS_DIRFRAGSIZ;
   1795 	auio.uio_offset = 0;
   1796 
   1797 	error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc);
   1798 
   1799 	len = NFS_DIRFRAGSIZ - auio.uio_resid;
   1800 	if (error || len == 0) {
   1801 		FREE(buf, M_TEMP);
   1802 		if (cookies)
   1803 			free(cookies, M_TEMP);
   1804 		return;
   1805 	}
   1806 
   1807 	/*
   1808 	 * Find the first valid entry and look at its offset cookie.
   1809 	 */
   1810 
   1811 	cp = buf;
   1812 	for (cop = cookies; len > 0; len -= dp->d_reclen) {
   1813 		dp = (struct dirent *)cp;
   1814 		if (dp->d_fileno != 0 && len >= dp->d_reclen) {
   1815 			if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) {
   1816 				*flagp |= NFSMNT_SWAPCOOKIE;
   1817 				nfs_invaldircache(vp, 0);
   1818 				nfs_vinvalbuf(vp, 0, cred, p, 1);
   1819 			}
   1820 			break;
   1821 		}
   1822 		cop++;
   1823 		cp += dp->d_reclen;
   1824 	}
   1825 
   1826 	FREE(buf, M_TEMP);
   1827 	free(cookies, M_TEMP);
   1828 }
   1829 #endif /* NFS */
   1830 
   1831 /*
   1832  * Set up nameidata for a lookup() call and do it.
   1833  *
   1834  * If pubflag is set, this call is done for a lookup operation on the
   1835  * public filehandle. In that case we allow crossing mountpoints and
   1836  * absolute pathnames. However, the caller is expected to check that
   1837  * the lookup result is within the public fs, and deny access if
   1838  * it is not.
   1839  */
   1840 int
   1841 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
   1842 	struct nameidata *ndp;
   1843 	fhandle_t *fhp;
   1844 	int len;
   1845 	struct nfssvc_sock *slp;
   1846 	struct mbuf *nam;
   1847 	struct mbuf **mdp;
   1848 	caddr_t *dposp;
   1849 	struct vnode **retdirp;
   1850 	struct proc *p;
   1851 	int kerbflag, pubflag;
   1852 {
   1853 	int i, rem;
   1854 	struct mbuf *md;
   1855 	char *fromcp, *tocp, *cp;
   1856 	struct iovec aiov;
   1857 	struct uio auio;
   1858 	struct vnode *dp;
   1859 	int error, rdonly, linklen;
   1860 	struct componentname *cnp = &ndp->ni_cnd;
   1861 
   1862 	*retdirp = (struct vnode *)0;
   1863 
   1864 	if ((len + 1) > MAXPATHLEN)
   1865 		return (ENAMETOOLONG);
   1866 	cnp->cn_pnbuf = PNBUF_GET();
   1867 
   1868 	/*
   1869 	 * Copy the name from the mbuf list to ndp->ni_pnbuf
   1870 	 * and set the various ndp fields appropriately.
   1871 	 */
   1872 	fromcp = *dposp;
   1873 	tocp = cnp->cn_pnbuf;
   1874 	md = *mdp;
   1875 	rem = mtod(md, caddr_t) + md->m_len - fromcp;
   1876 	for (i = 0; i < len; i++) {
   1877 		while (rem == 0) {
   1878 			md = md->m_next;
   1879 			if (md == NULL) {
   1880 				error = EBADRPC;
   1881 				goto out;
   1882 			}
   1883 			fromcp = mtod(md, caddr_t);
   1884 			rem = md->m_len;
   1885 		}
   1886 		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
   1887 			error = EACCES;
   1888 			goto out;
   1889 		}
   1890 		*tocp++ = *fromcp++;
   1891 		rem--;
   1892 	}
   1893 	*tocp = '\0';
   1894 	*mdp = md;
   1895 	*dposp = fromcp;
   1896 	len = nfsm_rndup(len)-len;
   1897 	if (len > 0) {
   1898 		if (rem >= len)
   1899 			*dposp += len;
   1900 		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
   1901 			goto out;
   1902 	}
   1903 
   1904 	/*
   1905 	 * Extract and set starting directory.
   1906 	 */
   1907 	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
   1908 	    nam, &rdonly, kerbflag, pubflag);
   1909 	if (error)
   1910 		goto out;
   1911 	if (dp->v_type != VDIR) {
   1912 		vrele(dp);
   1913 		error = ENOTDIR;
   1914 		goto out;
   1915 	}
   1916 
   1917 	if (rdonly)
   1918 		cnp->cn_flags |= RDONLY;
   1919 
   1920 	*retdirp = dp;
   1921 
   1922 	if (pubflag) {
   1923 		/*
   1924 		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
   1925 		 * and the 'native path' indicator.
   1926 		 */
   1927 		cp = PNBUF_GET();
   1928 		fromcp = cnp->cn_pnbuf;
   1929 		tocp = cp;
   1930 		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
   1931 			switch ((unsigned char)*fromcp) {
   1932 			case WEBNFS_NATIVE_CHAR:
   1933 				/*
   1934 				 * 'Native' path for us is the same
   1935 				 * as a path according to the NFS spec,
   1936 				 * just skip the escape char.
   1937 				 */
   1938 				fromcp++;
   1939 				break;
   1940 			/*
   1941 			 * More may be added in the future, range 0x80-0xff
   1942 			 */
   1943 			default:
   1944 				error = EIO;
   1945 				FREE(cp, M_NAMEI);
   1946 				goto out;
   1947 			}
   1948 		}
   1949 		/*
   1950 		 * Translate the '%' escapes, URL-style.
   1951 		 */
   1952 		while (*fromcp != '\0') {
   1953 			if (*fromcp == WEBNFS_ESC_CHAR) {
   1954 				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
   1955 					fromcp++;
   1956 					*tocp++ = HEXSTRTOI(fromcp);
   1957 					fromcp += 2;
   1958 					continue;
   1959 				} else {
   1960 					error = ENOENT;
   1961 					FREE(cp, M_NAMEI);
   1962 					goto out;
   1963 				}
   1964 			} else
   1965 				*tocp++ = *fromcp++;
   1966 		}
   1967 		*tocp = '\0';
   1968 		PNBUF_PUT(cnp->cn_pnbuf);
   1969 		cnp->cn_pnbuf = cp;
   1970 	}
   1971 
   1972 	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
   1973 	ndp->ni_segflg = UIO_SYSSPACE;
   1974 	ndp->ni_rootdir = rootvnode;
   1975 
   1976 	if (pubflag) {
   1977 		ndp->ni_loopcnt = 0;
   1978 		if (cnp->cn_pnbuf[0] == '/')
   1979 			dp = rootvnode;
   1980 	} else {
   1981 		cnp->cn_flags |= NOCROSSMOUNT;
   1982 	}
   1983 
   1984 	cnp->cn_proc = p;
   1985 	VREF(dp);
   1986 
   1987     for (;;) {
   1988 	cnp->cn_nameptr = cnp->cn_pnbuf;
   1989 	ndp->ni_startdir = dp;
   1990 	/*
   1991 	 * And call lookup() to do the real work
   1992 	 */
   1993 	error = lookup(ndp);
   1994 	if (error) {
   1995 		PNBUF_PUT(cnp->cn_pnbuf);
   1996 		return (error);
   1997 	}
   1998 	/*
   1999 	 * Check for encountering a symbolic link
   2000 	 */
   2001 	if ((cnp->cn_flags & ISSYMLINK) == 0) {
   2002 		if (cnp->cn_flags & (SAVENAME | SAVESTART))
   2003 			cnp->cn_flags |= HASBUF;
   2004 		else
   2005 			PNBUF_PUT(cnp->cn_pnbuf);
   2006 		return (0);
   2007 	} else {
   2008 		if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN))
   2009 			VOP_UNLOCK(ndp->ni_dvp, 0);
   2010 		if (!pubflag) {
   2011 			error = EINVAL;
   2012 			break;
   2013 		}
   2014 
   2015 		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
   2016 			error = ELOOP;
   2017 			break;
   2018 		}
   2019 		if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) {
   2020 			error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred,
   2021 			    cnp->cn_proc);
   2022 			if (error != 0)
   2023 				break;
   2024 		}
   2025 		if (ndp->ni_pathlen > 1)
   2026 			cp = PNBUF_GET();
   2027 		else
   2028 			cp = cnp->cn_pnbuf;
   2029 		aiov.iov_base = cp;
   2030 		aiov.iov_len = MAXPATHLEN;
   2031 		auio.uio_iov = &aiov;
   2032 		auio.uio_iovcnt = 1;
   2033 		auio.uio_offset = 0;
   2034 		auio.uio_rw = UIO_READ;
   2035 		auio.uio_segflg = UIO_SYSSPACE;
   2036 		auio.uio_procp = (struct proc *)0;
   2037 		auio.uio_resid = MAXPATHLEN;
   2038 		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
   2039 		if (error) {
   2040 		badlink:
   2041 			if (ndp->ni_pathlen > 1)
   2042 				PNBUF_PUT(cp);
   2043 			break;
   2044 		}
   2045 		linklen = MAXPATHLEN - auio.uio_resid;
   2046 		if (linklen == 0) {
   2047 			error = ENOENT;
   2048 			goto badlink;
   2049 		}
   2050 		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
   2051 			error = ENAMETOOLONG;
   2052 			goto badlink;
   2053 		}
   2054 		if (ndp->ni_pathlen > 1) {
   2055 			memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen);
   2056 			PNBUF_PUT(cnp->cn_pnbuf);
   2057 			cnp->cn_pnbuf = cp;
   2058 		} else
   2059 			cnp->cn_pnbuf[linklen] = '\0';
   2060 		ndp->ni_pathlen += linklen;
   2061 		vput(ndp->ni_vp);
   2062 		dp = ndp->ni_dvp;
   2063 		/*
   2064 		 * Check if root directory should replace current directory.
   2065 		 */
   2066 		if (cnp->cn_pnbuf[0] == '/') {
   2067 			vrele(dp);
   2068 			dp = ndp->ni_rootdir;
   2069 			VREF(dp);
   2070 		}
   2071 	}
   2072    }
   2073 	vrele(ndp->ni_dvp);
   2074 	vput(ndp->ni_vp);
   2075 	ndp->ni_vp = NULL;
   2076 out:
   2077 	PNBUF_PUT(cnp->cn_pnbuf);
   2078 	return (error);
   2079 }
   2080 
   2081 /*
   2082  * A fiddled version of m_adj() that ensures null fill to a long
   2083  * boundary and only trims off the back end
   2084  */
   2085 void
   2086 nfsm_adj(mp, len, nul)
   2087 	struct mbuf *mp;
   2088 	int len;
   2089 	int nul;
   2090 {
   2091 	struct mbuf *m;
   2092 	int count, i;
   2093 	char *cp;
   2094 
   2095 	/*
   2096 	 * Trim from tail.  Scan the mbuf chain,
   2097 	 * calculating its length and finding the last mbuf.
   2098 	 * If the adjustment only affects this mbuf, then just
   2099 	 * adjust and return.  Otherwise, rescan and truncate
   2100 	 * after the remaining size.
   2101 	 */
   2102 	count = 0;
   2103 	m = mp;
   2104 	for (;;) {
   2105 		count += m->m_len;
   2106 		if (m->m_next == (struct mbuf *)0)
   2107 			break;
   2108 		m = m->m_next;
   2109 	}
   2110 	if (m->m_len > len) {
   2111 		m->m_len -= len;
   2112 		if (nul > 0) {
   2113 			cp = mtod(m, caddr_t)+m->m_len-nul;
   2114 			for (i = 0; i < nul; i++)
   2115 				*cp++ = '\0';
   2116 		}
   2117 		return;
   2118 	}
   2119 	count -= len;
   2120 	if (count < 0)
   2121 		count = 0;
   2122 	/*
   2123 	 * Correct length for chain is "count".
   2124 	 * Find the mbuf with last data, adjust its length,
   2125 	 * and toss data from remaining mbufs on chain.
   2126 	 */
   2127 	for (m = mp; m; m = m->m_next) {
   2128 		if (m->m_len >= count) {
   2129 			m->m_len = count;
   2130 			if (nul > 0) {
   2131 				cp = mtod(m, caddr_t)+m->m_len-nul;
   2132 				for (i = 0; i < nul; i++)
   2133 					*cp++ = '\0';
   2134 			}
   2135 			break;
   2136 		}
   2137 		count -= m->m_len;
   2138 	}
   2139 	for (m = m->m_next;m;m = m->m_next)
   2140 		m->m_len = 0;
   2141 }
   2142 
   2143 /*
   2144  * Make these functions instead of macros, so that the kernel text size
   2145  * doesn't get too big...
   2146  */
   2147 void
   2148 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
   2149 	struct nfsrv_descript *nfsd;
   2150 	int before_ret;
   2151 	struct vattr *before_vap;
   2152 	int after_ret;
   2153 	struct vattr *after_vap;
   2154 	struct mbuf **mbp;
   2155 	char **bposp;
   2156 {
   2157 	struct mbuf *mb = *mbp, *mb2;
   2158 	char *bpos = *bposp;
   2159 	u_int32_t *tl;
   2160 
   2161 	if (before_ret) {
   2162 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
   2163 		*tl = nfs_false;
   2164 	} else {
   2165 		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
   2166 		*tl++ = nfs_true;
   2167 		txdr_hyper(before_vap->va_size, tl);
   2168 		tl += 2;
   2169 		txdr_nfsv3time(&(before_vap->va_mtime), tl);
   2170 		tl += 2;
   2171 		txdr_nfsv3time(&(before_vap->va_ctime), tl);
   2172 	}
   2173 	*bposp = bpos;
   2174 	*mbp = mb;
   2175 	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
   2176 }
   2177 
   2178 void
   2179 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
   2180 	struct nfsrv_descript *nfsd;
   2181 	int after_ret;
   2182 	struct vattr *after_vap;
   2183 	struct mbuf **mbp;
   2184 	char **bposp;
   2185 {
   2186 	struct mbuf *mb = *mbp, *mb2;
   2187 	char *bpos = *bposp;
   2188 	u_int32_t *tl;
   2189 	struct nfs_fattr *fp;
   2190 
   2191 	if (after_ret) {
   2192 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
   2193 		*tl = nfs_false;
   2194 	} else {
   2195 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
   2196 		*tl++ = nfs_true;
   2197 		fp = (struct nfs_fattr *)tl;
   2198 		nfsm_srvfattr(nfsd, after_vap, fp);
   2199 	}
   2200 	*mbp = mb;
   2201 	*bposp = bpos;
   2202 }
   2203 
   2204 void
   2205 nfsm_srvfattr(nfsd, vap, fp)
   2206 	struct nfsrv_descript *nfsd;
   2207 	struct vattr *vap;
   2208 	struct nfs_fattr *fp;
   2209 {
   2210 
   2211 	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
   2212 	fp->fa_uid = txdr_unsigned(vap->va_uid);
   2213 	fp->fa_gid = txdr_unsigned(vap->va_gid);
   2214 	if (nfsd->nd_flag & ND_NFSV3) {
   2215 		fp->fa_type = vtonfsv3_type(vap->va_type);
   2216 		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
   2217 		txdr_hyper(vap->va_size, &fp->fa3_size);
   2218 		txdr_hyper(vap->va_bytes, &fp->fa3_used);
   2219 		fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev));
   2220 		fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev));
   2221 		fp->fa3_fsid.nfsuquad[0] = 0;
   2222 		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
   2223 		fp->fa3_fileid.nfsuquad[0] = 0;
   2224 		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
   2225 		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
   2226 		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
   2227 		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
   2228 	} else {
   2229 		fp->fa_type = vtonfsv2_type(vap->va_type);
   2230 		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
   2231 		fp->fa2_size = txdr_unsigned(vap->va_size);
   2232 		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
   2233 		if (vap->va_type == VFIFO)
   2234 			fp->fa2_rdev = 0xffffffff;
   2235 		else
   2236 			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
   2237 		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
   2238 		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
   2239 		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
   2240 		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
   2241 		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
   2242 		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
   2243 	}
   2244 }
   2245 
   2246 /*
   2247  * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
   2248  * 	- look up fsid in mount list (if not found ret error)
   2249  *	- get vp and export rights by calling VFS_FHTOVP()
   2250  *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
   2251  *	- if not lockflag unlock it with VOP_UNLOCK()
   2252  */
   2253 int
   2254 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
   2255 	fhandle_t *fhp;
   2256 	int lockflag;
   2257 	struct vnode **vpp;
   2258 	struct ucred *cred;
   2259 	struct nfssvc_sock *slp;
   2260 	struct mbuf *nam;
   2261 	int *rdonlyp;
   2262 	int kerbflag;
   2263 {
   2264 	struct mount *mp;
   2265 	int i;
   2266 	struct ucred *credanon;
   2267 	int error, exflags;
   2268 	struct sockaddr_in *saddr;
   2269 
   2270 	*vpp = (struct vnode *)0;
   2271 
   2272 	if (nfs_ispublicfh(fhp)) {
   2273 		if (!pubflag || !nfs_pub.np_valid)
   2274 			return (ESTALE);
   2275 		fhp = &nfs_pub.np_handle;
   2276 	}
   2277 
   2278 	mp = vfs_getvfs(&fhp->fh_fsid);
   2279 	if (!mp)
   2280 		return (ESTALE);
   2281 	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
   2282 	if (error)
   2283 		return (error);
   2284 	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
   2285 	if (error)
   2286 		return (error);
   2287 
   2288 	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
   2289 		saddr = mtod(nam, struct sockaddr_in *);
   2290 		if ((saddr->sin_family == AF_INET) &&
   2291 		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
   2292 			vput(*vpp);
   2293 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2294 		}
   2295 #ifdef INET6
   2296 		if ((saddr->sin_family == AF_INET6) &&
   2297 		    ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) {
   2298 			vput(*vpp);
   2299 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2300 		}
   2301 #endif
   2302 	}
   2303 	/*
   2304 	 * Check/setup credentials.
   2305 	 */
   2306 	if (exflags & MNT_EXKERB) {
   2307 		if (!kerbflag) {
   2308 			vput(*vpp);
   2309 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2310 		}
   2311 	} else if (kerbflag) {
   2312 		vput(*vpp);
   2313 		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2314 	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
   2315 		cred->cr_uid = credanon->cr_uid;
   2316 		cred->cr_gid = credanon->cr_gid;
   2317 		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
   2318 			cred->cr_groups[i] = credanon->cr_groups[i];
   2319 		cred->cr_ngroups = i;
   2320 	}
   2321 	if (exflags & MNT_EXRDONLY)
   2322 		*rdonlyp = 1;
   2323 	else
   2324 		*rdonlyp = 0;
   2325 	if (!lockflag)
   2326 		VOP_UNLOCK(*vpp, 0);
   2327 	return (0);
   2328 }
   2329 
   2330 /*
   2331  * WebNFS: check if a filehandle is a public filehandle. For v3, this
   2332  * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
   2333  * transformed this to all zeroes in both cases, so check for it.
   2334  */
   2335 int
   2336 nfs_ispublicfh(fhp)
   2337 	fhandle_t *fhp;
   2338 {
   2339 	char *cp = (char *)fhp;
   2340 	int i;
   2341 
   2342 	for (i = 0; i < NFSX_V3FH; i++)
   2343 		if (*cp++ != 0)
   2344 			return (FALSE);
   2345 	return (TRUE);
   2346 }
   2347 
   2348 /*
   2349  * This function compares two net addresses by family and returns TRUE
   2350  * if they are the same host.
   2351  * If there is any doubt, return FALSE.
   2352  * The AF_INET family is handled as a special case so that address mbufs
   2353  * don't need to be saved to store "struct in_addr", which is only 4 bytes.
   2354  */
   2355 int
   2356 netaddr_match(family, haddr, nam)
   2357 	int family;
   2358 	union nethostaddr *haddr;
   2359 	struct mbuf *nam;
   2360 {
   2361 	struct sockaddr_in *inetaddr;
   2362 
   2363 	switch (family) {
   2364 	case AF_INET:
   2365 		inetaddr = mtod(nam, struct sockaddr_in *);
   2366 		if (inetaddr->sin_family == AF_INET &&
   2367 		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
   2368 			return (1);
   2369 		break;
   2370 #ifdef INET6
   2371 	case AF_INET6:
   2372 	    {
   2373 		struct sockaddr_in6 *sin6_1, *sin6_2;
   2374 
   2375 		sin6_1 = mtod(nam, struct sockaddr_in6 *);
   2376 		sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *);
   2377 		if (sin6_1->sin6_family == AF_INET6 &&
   2378 		    IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr))
   2379 			return 1;
   2380 	    }
   2381 #endif
   2382 #ifdef ISO
   2383 	case AF_ISO:
   2384 	    {
   2385 		struct sockaddr_iso *isoaddr1, *isoaddr2;
   2386 
   2387 		isoaddr1 = mtod(nam, struct sockaddr_iso *);
   2388 		isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *);
   2389 		if (isoaddr1->siso_family == AF_ISO &&
   2390 		    isoaddr1->siso_nlen > 0 &&
   2391 		    isoaddr1->siso_nlen == isoaddr2->siso_nlen &&
   2392 		    SAME_ISOADDR(isoaddr1, isoaddr2))
   2393 			return (1);
   2394 		break;
   2395 	    }
   2396 #endif	/* ISO */
   2397 	default:
   2398 		break;
   2399 	};
   2400 	return (0);
   2401 }
   2402 
   2403 /*
   2404  * The write verifier has changed (probably due to a server reboot), so all
   2405  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
   2406  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
   2407  * flag. Once done the new write verifier can be set for the mount point.
   2408  */
   2409 void
   2410 nfs_clearcommit(mp)
   2411 	struct mount *mp;
   2412 {
   2413 	struct vnode *vp;
   2414 	struct nfsnode *np;
   2415 	struct vm_page *pg;
   2416 	int s;
   2417 
   2418 	s = splbio();
   2419 	LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) {
   2420 		KASSERT(vp->v_mount == mp);
   2421 		if (vp->v_type == VNON)
   2422 			continue;
   2423 		np = VTONFS(vp);
   2424 		np->n_pushlo = np->n_pushhi = np->n_pushedlo =
   2425 		    np->n_pushedhi = 0;
   2426 		np->n_commitflags &=
   2427 		    ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID);
   2428 		simple_lock(&vp->v_uobj.vmobjlock);
   2429 		TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) {
   2430 			pg->flags &= ~PG_NEEDCOMMIT;
   2431 		}
   2432 		simple_unlock(&vp->v_uobj.vmobjlock);
   2433 	}
   2434 	splx(s);
   2435 }
   2436 
   2437 void
   2438 nfs_merge_commit_ranges(vp)
   2439 	struct vnode *vp;
   2440 {
   2441 	struct nfsnode *np = VTONFS(vp);
   2442 
   2443 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) {
   2444 		np->n_pushedlo = np->n_pushlo;
   2445 		np->n_pushedhi = np->n_pushhi;
   2446 		np->n_commitflags |= NFS_COMMIT_PUSHED_VALID;
   2447 	} else {
   2448 		if (np->n_pushlo < np->n_pushedlo)
   2449 			np->n_pushedlo = np->n_pushlo;
   2450 		if (np->n_pushhi > np->n_pushedhi)
   2451 			np->n_pushedhi = np->n_pushhi;
   2452 	}
   2453 
   2454 	np->n_pushlo = np->n_pushhi = 0;
   2455 	np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID;
   2456 
   2457 #ifdef fvdl_debug
   2458 	printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2459 	    (unsigned)np->n_pushedhi);
   2460 #endif
   2461 }
   2462 
   2463 int
   2464 nfs_in_committed_range(vp, off, len)
   2465 	struct vnode *vp;
   2466 	off_t off, len;
   2467 {
   2468 	struct nfsnode *np = VTONFS(vp);
   2469 	off_t lo, hi;
   2470 
   2471 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID))
   2472 		return 0;
   2473 	lo = off;
   2474 	hi = lo + len;
   2475 
   2476 	return (lo >= np->n_pushedlo && hi <= np->n_pushedhi);
   2477 }
   2478 
   2479 int
   2480 nfs_in_tobecommitted_range(vp, off, len)
   2481 	struct vnode *vp;
   2482 	off_t off, len;
   2483 {
   2484 	struct nfsnode *np = VTONFS(vp);
   2485 	off_t lo, hi;
   2486 
   2487 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID))
   2488 		return 0;
   2489 	lo = off;
   2490 	hi = lo + len;
   2491 
   2492 	return (lo >= np->n_pushlo && hi <= np->n_pushhi);
   2493 }
   2494 
   2495 void
   2496 nfs_add_committed_range(vp, off, len)
   2497 	struct vnode *vp;
   2498 	off_t off, len;
   2499 {
   2500 	struct nfsnode *np = VTONFS(vp);
   2501 	off_t lo, hi;
   2502 
   2503 	lo = off;
   2504 	hi = lo + len;
   2505 
   2506 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) {
   2507 		np->n_pushedlo = lo;
   2508 		np->n_pushedhi = hi;
   2509 		np->n_commitflags |= NFS_COMMIT_PUSHED_VALID;
   2510 	} else {
   2511 		if (hi > np->n_pushedhi)
   2512 			np->n_pushedhi = hi;
   2513 		if (lo < np->n_pushedlo)
   2514 			np->n_pushedlo = lo;
   2515 	}
   2516 #ifdef fvdl_debug
   2517 	printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2518 	    (unsigned)np->n_pushedhi);
   2519 #endif
   2520 }
   2521 
   2522 void
   2523 nfs_del_committed_range(vp, off, len)
   2524 	struct vnode *vp;
   2525 	off_t off, len;
   2526 {
   2527 	struct nfsnode *np = VTONFS(vp);
   2528 	off_t lo, hi;
   2529 
   2530 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID))
   2531 		return;
   2532 
   2533 	lo = off;
   2534 	hi = lo + len;
   2535 
   2536 	if (lo > np->n_pushedhi || hi < np->n_pushedlo)
   2537 		return;
   2538 	if (lo <= np->n_pushedlo)
   2539 		np->n_pushedlo = hi;
   2540 	else if (hi >= np->n_pushedhi)
   2541 		np->n_pushedhi = lo;
   2542 	else {
   2543 		/*
   2544 		 * XXX There's only one range. If the deleted range
   2545 		 * is in the middle, pick the largest of the
   2546 		 * contiguous ranges that it leaves.
   2547 		 */
   2548 		if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi))
   2549 			np->n_pushedhi = lo;
   2550 		else
   2551 			np->n_pushedlo = hi;
   2552 	}
   2553 #ifdef fvdl_debug
   2554 	printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2555 	    (unsigned)np->n_pushedhi);
   2556 #endif
   2557 }
   2558 
   2559 void
   2560 nfs_add_tobecommitted_range(vp, off, len)
   2561 	struct vnode *vp;
   2562 	off_t off, len;
   2563 {
   2564 	struct nfsnode *np = VTONFS(vp);
   2565 	off_t lo, hi;
   2566 
   2567 	lo = off;
   2568 	hi = lo + len;
   2569 
   2570 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) {
   2571 		np->n_pushlo = lo;
   2572 		np->n_pushhi = hi;
   2573 		np->n_commitflags |= NFS_COMMIT_PUSH_VALID;
   2574 	} else {
   2575 		if (lo < np->n_pushlo)
   2576 			np->n_pushlo = lo;
   2577 		if (hi > np->n_pushhi)
   2578 			np->n_pushhi = hi;
   2579 	}
   2580 #ifdef fvdl_debug
   2581 	printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo,
   2582 	    (unsigned)np->n_pushhi);
   2583 #endif
   2584 }
   2585 
   2586 void
   2587 nfs_del_tobecommitted_range(vp, off, len)
   2588 	struct vnode *vp;
   2589 	off_t off, len;
   2590 {
   2591 	struct nfsnode *np = VTONFS(vp);
   2592 	off_t lo, hi;
   2593 
   2594 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID))
   2595 		return;
   2596 
   2597 	lo = off;
   2598 	hi = lo + len;
   2599 
   2600 	if (lo > np->n_pushhi || hi < np->n_pushlo)
   2601 		return;
   2602 
   2603 	if (lo <= np->n_pushlo)
   2604 		np->n_pushlo = hi;
   2605 	else if (hi >= np->n_pushhi)
   2606 		np->n_pushhi = lo;
   2607 	else {
   2608 		/*
   2609 		 * XXX There's only one range. If the deleted range
   2610 		 * is in the middle, pick the largest of the
   2611 		 * contiguous ranges that it leaves.
   2612 		 */
   2613 		if ((np->n_pushlo - lo) > (hi - np->n_pushhi))
   2614 			np->n_pushhi = lo;
   2615 		else
   2616 			np->n_pushlo = hi;
   2617 	}
   2618 #ifdef fvdl_debug
   2619 	printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo,
   2620 	    (unsigned)np->n_pushhi);
   2621 #endif
   2622 }
   2623 
   2624 /*
   2625  * Map errnos to NFS error numbers. For Version 3 also filter out error
   2626  * numbers not specified for the associated procedure.
   2627  */
   2628 int
   2629 nfsrv_errmap(nd, err)
   2630 	struct nfsrv_descript *nd;
   2631 	int err;
   2632 {
   2633 	const short *defaulterrp, *errp;
   2634 
   2635 	if (nd->nd_flag & ND_NFSV3) {
   2636 	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
   2637 		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
   2638 		while (*++errp) {
   2639 			if (*errp == err)
   2640 				return (err);
   2641 			else if (*errp > err)
   2642 				break;
   2643 		}
   2644 		return ((int)*defaulterrp);
   2645 	    } else
   2646 		return (err & 0xffff);
   2647 	}
   2648 	if (err <= ELAST)
   2649 		return ((int)nfsrv_v2errmap[err - 1]);
   2650 	return (NFSERR_IO);
   2651 }
   2652 
   2653 /*
   2654  * Sort the group list in increasing numerical order.
   2655  * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
   2656  *  that used to be here.)
   2657  */
   2658 void
   2659 nfsrvw_sort(list, num)
   2660         gid_t *list;
   2661         int num;
   2662 {
   2663 	int i, j;
   2664 	gid_t v;
   2665 
   2666 	/* Insertion sort. */
   2667 	for (i = 1; i < num; i++) {
   2668 		v = list[i];
   2669 		/* find correct slot for value v, moving others up */
   2670 		for (j = i; --j >= 0 && v < list[j];)
   2671 			list[j + 1] = list[j];
   2672 		list[j + 1] = v;
   2673 	}
   2674 }
   2675 
   2676 /*
   2677  * copy credentials making sure that the result can be compared with memcmp().
   2678  */
   2679 void
   2680 nfsrv_setcred(incred, outcred)
   2681 	struct ucred *incred, *outcred;
   2682 {
   2683 	int i;
   2684 
   2685 	memset((caddr_t)outcred, 0, sizeof (struct ucred));
   2686 	outcred->cr_ref = 1;
   2687 	outcred->cr_uid = incred->cr_uid;
   2688 	outcred->cr_gid = incred->cr_gid;
   2689 	outcred->cr_ngroups = incred->cr_ngroups;
   2690 	for (i = 0; i < incred->cr_ngroups; i++)
   2691 		outcred->cr_groups[i] = incred->cr_groups[i];
   2692 	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
   2693 }
   2694