Home | History | Annotate | Line # | Download | only in nfs
nfs_subs.c revision 1.98
      1 /*	$NetBSD: nfs_subs.c,v 1.98 2001/09/27 21:44:56 fvdl 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 "fs_nfs.h"
     77 #include "opt_nfs.h"
     78 #include "opt_nfsserver.h"
     79 #include "opt_iso.h"
     80 #include "opt_inet.h"
     81 
     82 /*
     83  * These functions support the macros and help fiddle mbuf chains for
     84  * the nfs op functions. They do things like create the rpc header and
     85  * copy data between mbuf chains and uio lists.
     86  */
     87 #include <sys/param.h>
     88 #include <sys/proc.h>
     89 #include <sys/systm.h>
     90 #include <sys/kernel.h>
     91 #include <sys/mount.h>
     92 #include <sys/vnode.h>
     93 #include <sys/namei.h>
     94 #include <sys/mbuf.h>
     95 #include <sys/socket.h>
     96 #include <sys/stat.h>
     97 #include <sys/malloc.h>
     98 #include <sys/filedesc.h>
     99 #include <sys/time.h>
    100 #include <sys/dirent.h>
    101 
    102 #include <uvm/uvm_extern.h>
    103 
    104 #include <nfs/rpcv2.h>
    105 #include <nfs/nfsproto.h>
    106 #include <nfs/nfsnode.h>
    107 #include <nfs/nfs.h>
    108 #include <nfs/xdr_subs.h>
    109 #include <nfs/nfsm_subs.h>
    110 #include <nfs/nfsmount.h>
    111 #include <nfs/nqnfs.h>
    112 #include <nfs/nfsrtt.h>
    113 #include <nfs/nfs_var.h>
    114 
    115 #include <miscfs/specfs/specdev.h>
    116 
    117 #include <netinet/in.h>
    118 #ifdef ISO
    119 #include <netiso/iso.h>
    120 #endif
    121 
    122 /*
    123  * Data items converted to xdr at startup, since they are constant
    124  * This is kinda hokey, but may save a little time doing byte swaps
    125  */
    126 u_int32_t nfs_xdrneg1;
    127 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
    128 	rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
    129 	rpc_auth_kerb;
    130 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false;
    131 
    132 /* And other global data */
    133 static u_int32_t nfs_xid = 0;
    134 const nfstype nfsv2_type[9] =
    135 	{ NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON };
    136 const nfstype nfsv3_type[9] =
    137 	{ NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON };
    138 const enum vtype nv2tov_type[8] =
    139 	{ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON };
    140 const enum vtype nv3tov_type[8] =
    141 	{ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO };
    142 int nfs_ticks;
    143 
    144 /* NFS client/server stats. */
    145 struct nfsstats nfsstats;
    146 
    147 /*
    148  * Mapping of old NFS Version 2 RPC numbers to generic numbers.
    149  */
    150 const int nfsv3_procid[NFS_NPROCS] = {
    151 	NFSPROC_NULL,
    152 	NFSPROC_GETATTR,
    153 	NFSPROC_SETATTR,
    154 	NFSPROC_NOOP,
    155 	NFSPROC_LOOKUP,
    156 	NFSPROC_READLINK,
    157 	NFSPROC_READ,
    158 	NFSPROC_NOOP,
    159 	NFSPROC_WRITE,
    160 	NFSPROC_CREATE,
    161 	NFSPROC_REMOVE,
    162 	NFSPROC_RENAME,
    163 	NFSPROC_LINK,
    164 	NFSPROC_SYMLINK,
    165 	NFSPROC_MKDIR,
    166 	NFSPROC_RMDIR,
    167 	NFSPROC_READDIR,
    168 	NFSPROC_FSSTAT,
    169 	NFSPROC_NOOP,
    170 	NFSPROC_NOOP,
    171 	NFSPROC_NOOP,
    172 	NFSPROC_NOOP,
    173 	NFSPROC_NOOP,
    174 	NFSPROC_NOOP,
    175 	NFSPROC_NOOP,
    176 	NFSPROC_NOOP
    177 };
    178 
    179 /*
    180  * and the reverse mapping from generic to Version 2 procedure numbers
    181  */
    182 const int nfsv2_procid[NFS_NPROCS] = {
    183 	NFSV2PROC_NULL,
    184 	NFSV2PROC_GETATTR,
    185 	NFSV2PROC_SETATTR,
    186 	NFSV2PROC_LOOKUP,
    187 	NFSV2PROC_NOOP,
    188 	NFSV2PROC_READLINK,
    189 	NFSV2PROC_READ,
    190 	NFSV2PROC_WRITE,
    191 	NFSV2PROC_CREATE,
    192 	NFSV2PROC_MKDIR,
    193 	NFSV2PROC_SYMLINK,
    194 	NFSV2PROC_CREATE,
    195 	NFSV2PROC_REMOVE,
    196 	NFSV2PROC_RMDIR,
    197 	NFSV2PROC_RENAME,
    198 	NFSV2PROC_LINK,
    199 	NFSV2PROC_READDIR,
    200 	NFSV2PROC_NOOP,
    201 	NFSV2PROC_STATFS,
    202 	NFSV2PROC_NOOP,
    203 	NFSV2PROC_NOOP,
    204 	NFSV2PROC_NOOP,
    205 	NFSV2PROC_NOOP,
    206 	NFSV2PROC_NOOP,
    207 	NFSV2PROC_NOOP,
    208 	NFSV2PROC_NOOP,
    209 };
    210 
    211 /*
    212  * Maps errno values to nfs error numbers.
    213  * Use NFSERR_IO as the catch all for ones not specifically defined in
    214  * RFC 1094.
    215  */
    216 static const u_char nfsrv_v2errmap[ELAST] = {
    217   NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    218   NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    219   NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
    220   NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
    221   NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    222   NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
    223   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    224   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    225   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    226   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    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_NAMETOL,	NFSERR_IO,	NFSERR_IO,
    230   NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
    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,
    234 };
    235 
    236 /*
    237  * Maps errno values to nfs error numbers.
    238  * Although it is not obvious whether or not NFS clients really care if
    239  * a returned error value is in the specified list for the procedure, the
    240  * safest thing to do is filter them appropriately. For Version 2, the
    241  * X/Open XNFS document is the only specification that defines error values
    242  * for each RPC (The RFC simply lists all possible error values for all RPCs),
    243  * so I have decided to not do this for Version 2.
    244  * The first entry is the default error return and the rest are the valid
    245  * errors for that RPC in increasing numeric order.
    246  */
    247 static const short nfsv3err_null[] = {
    248 	0,
    249 	0,
    250 };
    251 
    252 static const short nfsv3err_getattr[] = {
    253 	NFSERR_IO,
    254 	NFSERR_IO,
    255 	NFSERR_STALE,
    256 	NFSERR_BADHANDLE,
    257 	NFSERR_SERVERFAULT,
    258 	0,
    259 };
    260 
    261 static const short nfsv3err_setattr[] = {
    262 	NFSERR_IO,
    263 	NFSERR_PERM,
    264 	NFSERR_IO,
    265 	NFSERR_ACCES,
    266 	NFSERR_INVAL,
    267 	NFSERR_NOSPC,
    268 	NFSERR_ROFS,
    269 	NFSERR_DQUOT,
    270 	NFSERR_STALE,
    271 	NFSERR_BADHANDLE,
    272 	NFSERR_NOT_SYNC,
    273 	NFSERR_SERVERFAULT,
    274 	0,
    275 };
    276 
    277 static const short nfsv3err_lookup[] = {
    278 	NFSERR_IO,
    279 	NFSERR_NOENT,
    280 	NFSERR_IO,
    281 	NFSERR_ACCES,
    282 	NFSERR_NOTDIR,
    283 	NFSERR_NAMETOL,
    284 	NFSERR_STALE,
    285 	NFSERR_BADHANDLE,
    286 	NFSERR_SERVERFAULT,
    287 	0,
    288 };
    289 
    290 static const short nfsv3err_access[] = {
    291 	NFSERR_IO,
    292 	NFSERR_IO,
    293 	NFSERR_STALE,
    294 	NFSERR_BADHANDLE,
    295 	NFSERR_SERVERFAULT,
    296 	0,
    297 };
    298 
    299 static const short nfsv3err_readlink[] = {
    300 	NFSERR_IO,
    301 	NFSERR_IO,
    302 	NFSERR_ACCES,
    303 	NFSERR_INVAL,
    304 	NFSERR_STALE,
    305 	NFSERR_BADHANDLE,
    306 	NFSERR_NOTSUPP,
    307 	NFSERR_SERVERFAULT,
    308 	0,
    309 };
    310 
    311 static const short nfsv3err_read[] = {
    312 	NFSERR_IO,
    313 	NFSERR_IO,
    314 	NFSERR_NXIO,
    315 	NFSERR_ACCES,
    316 	NFSERR_INVAL,
    317 	NFSERR_STALE,
    318 	NFSERR_BADHANDLE,
    319 	NFSERR_SERVERFAULT,
    320 	NFSERR_JUKEBOX,
    321 	0,
    322 };
    323 
    324 static const short nfsv3err_write[] = {
    325 	NFSERR_IO,
    326 	NFSERR_IO,
    327 	NFSERR_ACCES,
    328 	NFSERR_INVAL,
    329 	NFSERR_FBIG,
    330 	NFSERR_NOSPC,
    331 	NFSERR_ROFS,
    332 	NFSERR_DQUOT,
    333 	NFSERR_STALE,
    334 	NFSERR_BADHANDLE,
    335 	NFSERR_SERVERFAULT,
    336 	NFSERR_JUKEBOX,
    337 	0,
    338 };
    339 
    340 static const short nfsv3err_create[] = {
    341 	NFSERR_IO,
    342 	NFSERR_IO,
    343 	NFSERR_ACCES,
    344 	NFSERR_EXIST,
    345 	NFSERR_NOTDIR,
    346 	NFSERR_NOSPC,
    347 	NFSERR_ROFS,
    348 	NFSERR_NAMETOL,
    349 	NFSERR_DQUOT,
    350 	NFSERR_STALE,
    351 	NFSERR_BADHANDLE,
    352 	NFSERR_NOTSUPP,
    353 	NFSERR_SERVERFAULT,
    354 	0,
    355 };
    356 
    357 static const short nfsv3err_mkdir[] = {
    358 	NFSERR_IO,
    359 	NFSERR_IO,
    360 	NFSERR_ACCES,
    361 	NFSERR_EXIST,
    362 	NFSERR_NOTDIR,
    363 	NFSERR_NOSPC,
    364 	NFSERR_ROFS,
    365 	NFSERR_NAMETOL,
    366 	NFSERR_DQUOT,
    367 	NFSERR_STALE,
    368 	NFSERR_BADHANDLE,
    369 	NFSERR_NOTSUPP,
    370 	NFSERR_SERVERFAULT,
    371 	0,
    372 };
    373 
    374 static const short nfsv3err_symlink[] = {
    375 	NFSERR_IO,
    376 	NFSERR_IO,
    377 	NFSERR_ACCES,
    378 	NFSERR_EXIST,
    379 	NFSERR_NOTDIR,
    380 	NFSERR_NOSPC,
    381 	NFSERR_ROFS,
    382 	NFSERR_NAMETOL,
    383 	NFSERR_DQUOT,
    384 	NFSERR_STALE,
    385 	NFSERR_BADHANDLE,
    386 	NFSERR_NOTSUPP,
    387 	NFSERR_SERVERFAULT,
    388 	0,
    389 };
    390 
    391 static const short nfsv3err_mknod[] = {
    392 	NFSERR_IO,
    393 	NFSERR_IO,
    394 	NFSERR_ACCES,
    395 	NFSERR_EXIST,
    396 	NFSERR_NOTDIR,
    397 	NFSERR_NOSPC,
    398 	NFSERR_ROFS,
    399 	NFSERR_NAMETOL,
    400 	NFSERR_DQUOT,
    401 	NFSERR_STALE,
    402 	NFSERR_BADHANDLE,
    403 	NFSERR_NOTSUPP,
    404 	NFSERR_SERVERFAULT,
    405 	NFSERR_BADTYPE,
    406 	0,
    407 };
    408 
    409 static const short nfsv3err_remove[] = {
    410 	NFSERR_IO,
    411 	NFSERR_NOENT,
    412 	NFSERR_IO,
    413 	NFSERR_ACCES,
    414 	NFSERR_NOTDIR,
    415 	NFSERR_ROFS,
    416 	NFSERR_NAMETOL,
    417 	NFSERR_STALE,
    418 	NFSERR_BADHANDLE,
    419 	NFSERR_SERVERFAULT,
    420 	0,
    421 };
    422 
    423 static const short nfsv3err_rmdir[] = {
    424 	NFSERR_IO,
    425 	NFSERR_NOENT,
    426 	NFSERR_IO,
    427 	NFSERR_ACCES,
    428 	NFSERR_EXIST,
    429 	NFSERR_NOTDIR,
    430 	NFSERR_INVAL,
    431 	NFSERR_ROFS,
    432 	NFSERR_NAMETOL,
    433 	NFSERR_NOTEMPTY,
    434 	NFSERR_STALE,
    435 	NFSERR_BADHANDLE,
    436 	NFSERR_NOTSUPP,
    437 	NFSERR_SERVERFAULT,
    438 	0,
    439 };
    440 
    441 static const short nfsv3err_rename[] = {
    442 	NFSERR_IO,
    443 	NFSERR_NOENT,
    444 	NFSERR_IO,
    445 	NFSERR_ACCES,
    446 	NFSERR_EXIST,
    447 	NFSERR_XDEV,
    448 	NFSERR_NOTDIR,
    449 	NFSERR_ISDIR,
    450 	NFSERR_INVAL,
    451 	NFSERR_NOSPC,
    452 	NFSERR_ROFS,
    453 	NFSERR_MLINK,
    454 	NFSERR_NAMETOL,
    455 	NFSERR_NOTEMPTY,
    456 	NFSERR_DQUOT,
    457 	NFSERR_STALE,
    458 	NFSERR_BADHANDLE,
    459 	NFSERR_NOTSUPP,
    460 	NFSERR_SERVERFAULT,
    461 	0,
    462 };
    463 
    464 static const short nfsv3err_link[] = {
    465 	NFSERR_IO,
    466 	NFSERR_IO,
    467 	NFSERR_ACCES,
    468 	NFSERR_EXIST,
    469 	NFSERR_XDEV,
    470 	NFSERR_NOTDIR,
    471 	NFSERR_INVAL,
    472 	NFSERR_NOSPC,
    473 	NFSERR_ROFS,
    474 	NFSERR_MLINK,
    475 	NFSERR_NAMETOL,
    476 	NFSERR_DQUOT,
    477 	NFSERR_STALE,
    478 	NFSERR_BADHANDLE,
    479 	NFSERR_NOTSUPP,
    480 	NFSERR_SERVERFAULT,
    481 	0,
    482 };
    483 
    484 static const short nfsv3err_readdir[] = {
    485 	NFSERR_IO,
    486 	NFSERR_IO,
    487 	NFSERR_ACCES,
    488 	NFSERR_NOTDIR,
    489 	NFSERR_STALE,
    490 	NFSERR_BADHANDLE,
    491 	NFSERR_BAD_COOKIE,
    492 	NFSERR_TOOSMALL,
    493 	NFSERR_SERVERFAULT,
    494 	0,
    495 };
    496 
    497 static const short nfsv3err_readdirplus[] = {
    498 	NFSERR_IO,
    499 	NFSERR_IO,
    500 	NFSERR_ACCES,
    501 	NFSERR_NOTDIR,
    502 	NFSERR_STALE,
    503 	NFSERR_BADHANDLE,
    504 	NFSERR_BAD_COOKIE,
    505 	NFSERR_NOTSUPP,
    506 	NFSERR_TOOSMALL,
    507 	NFSERR_SERVERFAULT,
    508 	0,
    509 };
    510 
    511 static const short nfsv3err_fsstat[] = {
    512 	NFSERR_IO,
    513 	NFSERR_IO,
    514 	NFSERR_STALE,
    515 	NFSERR_BADHANDLE,
    516 	NFSERR_SERVERFAULT,
    517 	0,
    518 };
    519 
    520 static const short nfsv3err_fsinfo[] = {
    521 	NFSERR_STALE,
    522 	NFSERR_STALE,
    523 	NFSERR_BADHANDLE,
    524 	NFSERR_SERVERFAULT,
    525 	0,
    526 };
    527 
    528 static const short nfsv3err_pathconf[] = {
    529 	NFSERR_STALE,
    530 	NFSERR_STALE,
    531 	NFSERR_BADHANDLE,
    532 	NFSERR_SERVERFAULT,
    533 	0,
    534 };
    535 
    536 static const short nfsv3err_commit[] = {
    537 	NFSERR_IO,
    538 	NFSERR_IO,
    539 	NFSERR_STALE,
    540 	NFSERR_BADHANDLE,
    541 	NFSERR_SERVERFAULT,
    542 	0,
    543 };
    544 
    545 static const short * const nfsrv_v3errmap[] = {
    546 	nfsv3err_null,
    547 	nfsv3err_getattr,
    548 	nfsv3err_setattr,
    549 	nfsv3err_lookup,
    550 	nfsv3err_access,
    551 	nfsv3err_readlink,
    552 	nfsv3err_read,
    553 	nfsv3err_write,
    554 	nfsv3err_create,
    555 	nfsv3err_mkdir,
    556 	nfsv3err_symlink,
    557 	nfsv3err_mknod,
    558 	nfsv3err_remove,
    559 	nfsv3err_rmdir,
    560 	nfsv3err_rename,
    561 	nfsv3err_link,
    562 	nfsv3err_readdir,
    563 	nfsv3err_readdirplus,
    564 	nfsv3err_fsstat,
    565 	nfsv3err_fsinfo,
    566 	nfsv3err_pathconf,
    567 	nfsv3err_commit,
    568 };
    569 
    570 extern struct nfsrtt nfsrtt;
    571 extern time_t nqnfsstarttime;
    572 extern int nqsrv_clockskew;
    573 extern int nqsrv_writeslack;
    574 extern int nqsrv_maxlease;
    575 extern const int nqnfs_piggy[NFS_NPROCS];
    576 extern struct nfsnodehashhead *nfsnodehashtbl;
    577 extern u_long nfsnodehash;
    578 
    579 LIST_HEAD(nfsnodehashhead, nfsnode);
    580 u_long nfsdirhashmask;
    581 
    582 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
    583 
    584 /*
    585  * Create the header for an rpc request packet
    586  * The hsiz is the size of the rest of the nfs request header.
    587  * (just used to decide if a cluster is a good idea)
    588  */
    589 struct mbuf *
    590 nfsm_reqh(vp, procid, hsiz, bposp)
    591 	struct vnode *vp;
    592 	u_long procid;
    593 	int hsiz;
    594 	caddr_t *bposp;
    595 {
    596 	struct mbuf *mb;
    597 	caddr_t bpos;
    598 	struct nfsmount *nmp;
    599 #ifndef NFS_V2_ONLY
    600 	u_int32_t *tl;
    601 	struct mbuf *mb2;
    602 	int nqflag;
    603 #endif
    604 
    605 	MGET(mb, M_WAIT, MT_DATA);
    606 	if (hsiz >= MINCLSIZE)
    607 		MCLGET(mb, M_WAIT);
    608 	mb->m_len = 0;
    609 	bpos = mtod(mb, caddr_t);
    610 
    611 	/*
    612 	 * For NQNFS, add lease request.
    613 	 */
    614 	if (vp) {
    615 		nmp = VFSTONFS(vp->v_mount);
    616 #ifndef NFS_V2_ONLY
    617 		if (nmp->nm_flag & NFSMNT_NQNFS) {
    618 			nqflag = NQNFS_NEEDLEASE(vp, procid);
    619 			if (nqflag) {
    620 				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
    621 				*tl++ = txdr_unsigned(nqflag);
    622 				*tl = txdr_unsigned(nmp->nm_leaseterm);
    623 			} else {
    624 				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
    625 				*tl = 0;
    626 			}
    627 		}
    628 #endif
    629 	}
    630 	/* Finally, return values */
    631 	*bposp = bpos;
    632 	return (mb);
    633 }
    634 
    635 /*
    636  * Build the RPC header and fill in the authorization info.
    637  * The authorization string argument is only used when the credentials
    638  * come from outside of the kernel.
    639  * Returns the head of the mbuf list.
    640  */
    641 struct mbuf *
    642 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
    643 	verf_str, mrest, mrest_len, mbp, xidp)
    644 	struct ucred *cr;
    645 	int nmflag;
    646 	int procid;
    647 	int auth_type;
    648 	int auth_len;
    649 	char *auth_str;
    650 	int verf_len;
    651 	char *verf_str;
    652 	struct mbuf *mrest;
    653 	int mrest_len;
    654 	struct mbuf **mbp;
    655 	u_int32_t *xidp;
    656 {
    657 	struct mbuf *mb;
    658 	u_int32_t *tl;
    659 	caddr_t bpos;
    660 	int i;
    661 	struct mbuf *mreq, *mb2;
    662 	int siz, grpsiz, authsiz;
    663 	struct timeval tv;
    664 	static u_int32_t base;
    665 
    666 	authsiz = nfsm_rndup(auth_len);
    667 	MGETHDR(mb, M_WAIT, MT_DATA);
    668 	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
    669 		MCLGET(mb, M_WAIT);
    670 	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
    671 		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
    672 	} else {
    673 		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
    674 	}
    675 	mb->m_len = 0;
    676 	mreq = mb;
    677 	bpos = mtod(mb, caddr_t);
    678 
    679 	/*
    680 	 * First the RPC header.
    681 	 */
    682 	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
    683 
    684 	/*
    685 	 * derive initial xid from system time
    686 	 * XXX time is invalid if root not yet mounted
    687 	 */
    688 	if (!base && (rootvp)) {
    689 		microtime(&tv);
    690 		base = tv.tv_sec << 12;
    691 		nfs_xid = base;
    692 	}
    693 	/*
    694 	 * Skip zero xid if it should ever happen.
    695 	 */
    696 	if (++nfs_xid == 0)
    697 		nfs_xid++;
    698 
    699 	*tl++ = *xidp = txdr_unsigned(nfs_xid);
    700 	*tl++ = rpc_call;
    701 	*tl++ = rpc_vers;
    702 	if (nmflag & NFSMNT_NQNFS) {
    703 		*tl++ = txdr_unsigned(NQNFS_PROG);
    704 		*tl++ = txdr_unsigned(NQNFS_VER3);
    705 	} else {
    706 		*tl++ = txdr_unsigned(NFS_PROG);
    707 		if (nmflag & NFSMNT_NFSV3)
    708 			*tl++ = txdr_unsigned(NFS_VER3);
    709 		else
    710 			*tl++ = txdr_unsigned(NFS_VER2);
    711 	}
    712 	if (nmflag & NFSMNT_NFSV3)
    713 		*tl++ = txdr_unsigned(procid);
    714 	else
    715 		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
    716 
    717 	/*
    718 	 * And then the authorization cred.
    719 	 */
    720 	*tl++ = txdr_unsigned(auth_type);
    721 	*tl = txdr_unsigned(authsiz);
    722 	switch (auth_type) {
    723 	case RPCAUTH_UNIX:
    724 		nfsm_build(tl, u_int32_t *, auth_len);
    725 		*tl++ = 0;		/* stamp ?? */
    726 		*tl++ = 0;		/* NULL hostname */
    727 		*tl++ = txdr_unsigned(cr->cr_uid);
    728 		*tl++ = txdr_unsigned(cr->cr_gid);
    729 		grpsiz = (auth_len >> 2) - 5;
    730 		*tl++ = txdr_unsigned(grpsiz);
    731 		for (i = 0; i < grpsiz; i++)
    732 			*tl++ = txdr_unsigned(cr->cr_groups[i]);
    733 		break;
    734 	case RPCAUTH_KERB4:
    735 		siz = auth_len;
    736 		while (siz > 0) {
    737 			if (M_TRAILINGSPACE(mb) == 0) {
    738 				MGET(mb2, M_WAIT, MT_DATA);
    739 				if (siz >= MINCLSIZE)
    740 					MCLGET(mb2, M_WAIT);
    741 				mb->m_next = mb2;
    742 				mb = mb2;
    743 				mb->m_len = 0;
    744 				bpos = mtod(mb, caddr_t);
    745 			}
    746 			i = min(siz, M_TRAILINGSPACE(mb));
    747 			memcpy(bpos, auth_str, i);
    748 			mb->m_len += i;
    749 			auth_str += i;
    750 			bpos += i;
    751 			siz -= i;
    752 		}
    753 		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
    754 			for (i = 0; i < siz; i++)
    755 				*bpos++ = '\0';
    756 			mb->m_len += siz;
    757 		}
    758 		break;
    759 	};
    760 
    761 	/*
    762 	 * And the verifier...
    763 	 */
    764 	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
    765 	if (verf_str) {
    766 		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
    767 		*tl = txdr_unsigned(verf_len);
    768 		siz = verf_len;
    769 		while (siz > 0) {
    770 			if (M_TRAILINGSPACE(mb) == 0) {
    771 				MGET(mb2, M_WAIT, MT_DATA);
    772 				if (siz >= MINCLSIZE)
    773 					MCLGET(mb2, M_WAIT);
    774 				mb->m_next = mb2;
    775 				mb = mb2;
    776 				mb->m_len = 0;
    777 				bpos = mtod(mb, caddr_t);
    778 			}
    779 			i = min(siz, M_TRAILINGSPACE(mb));
    780 			memcpy(bpos, verf_str, i);
    781 			mb->m_len += i;
    782 			verf_str += i;
    783 			bpos += i;
    784 			siz -= i;
    785 		}
    786 		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
    787 			for (i = 0; i < siz; i++)
    788 				*bpos++ = '\0';
    789 			mb->m_len += siz;
    790 		}
    791 	} else {
    792 		*tl++ = txdr_unsigned(RPCAUTH_NULL);
    793 		*tl = 0;
    794 	}
    795 	mb->m_next = mrest;
    796 	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
    797 	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
    798 	*mbp = mb;
    799 	return (mreq);
    800 }
    801 
    802 /*
    803  * copies mbuf chain to the uio scatter/gather list
    804  */
    805 int
    806 nfsm_mbuftouio(mrep, uiop, siz, dpos)
    807 	struct mbuf **mrep;
    808 	struct uio *uiop;
    809 	int siz;
    810 	caddr_t *dpos;
    811 {
    812 	char *mbufcp, *uiocp;
    813 	int xfer, left, len;
    814 	struct mbuf *mp;
    815 	long uiosiz, rem;
    816 	int error = 0;
    817 
    818 	mp = *mrep;
    819 	mbufcp = *dpos;
    820 	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
    821 	rem = nfsm_rndup(siz)-siz;
    822 	while (siz > 0) {
    823 		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
    824 			return (EFBIG);
    825 		left = uiop->uio_iov->iov_len;
    826 		uiocp = uiop->uio_iov->iov_base;
    827 		if (left > siz)
    828 			left = siz;
    829 		uiosiz = left;
    830 		while (left > 0) {
    831 			while (len == 0) {
    832 				mp = mp->m_next;
    833 				if (mp == NULL)
    834 					return (EBADRPC);
    835 				mbufcp = mtod(mp, caddr_t);
    836 				len = mp->m_len;
    837 			}
    838 			xfer = (left > len) ? len : left;
    839 #ifdef notdef
    840 			/* Not Yet.. */
    841 			if (uiop->uio_iov->iov_op != NULL)
    842 				(*(uiop->uio_iov->iov_op))
    843 				(mbufcp, uiocp, xfer);
    844 			else
    845 #endif
    846 			if (uiop->uio_segflg == UIO_SYSSPACE)
    847 				memcpy(uiocp, mbufcp, xfer);
    848 			else
    849 				copyout(mbufcp, uiocp, xfer);
    850 			left -= xfer;
    851 			len -= xfer;
    852 			mbufcp += xfer;
    853 			uiocp += xfer;
    854 			uiop->uio_offset += xfer;
    855 			uiop->uio_resid -= xfer;
    856 		}
    857 		if (uiop->uio_iov->iov_len <= siz) {
    858 			uiop->uio_iovcnt--;
    859 			uiop->uio_iov++;
    860 		} else {
    861 			uiop->uio_iov->iov_base =
    862 			    (caddr_t)uiop->uio_iov->iov_base + uiosiz;
    863 			uiop->uio_iov->iov_len -= uiosiz;
    864 		}
    865 		siz -= uiosiz;
    866 	}
    867 	*dpos = mbufcp;
    868 	*mrep = mp;
    869 	if (rem > 0) {
    870 		if (len < rem)
    871 			error = nfs_adv(mrep, dpos, rem, len);
    872 		else
    873 			*dpos += rem;
    874 	}
    875 	return (error);
    876 }
    877 
    878 /*
    879  * copies a uio scatter/gather list to an mbuf chain.
    880  * NOTE: can ony handle iovcnt == 1
    881  */
    882 int
    883 nfsm_uiotombuf(uiop, mq, siz, bpos)
    884 	struct uio *uiop;
    885 	struct mbuf **mq;
    886 	int siz;
    887 	caddr_t *bpos;
    888 {
    889 	char *uiocp;
    890 	struct mbuf *mp, *mp2;
    891 	int xfer, left, mlen;
    892 	int uiosiz, clflg, rem;
    893 	char *cp;
    894 
    895 #ifdef DIAGNOSTIC
    896 	if (uiop->uio_iovcnt != 1)
    897 		panic("nfsm_uiotombuf: iovcnt != 1");
    898 #endif
    899 
    900 	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
    901 		clflg = 1;
    902 	else
    903 		clflg = 0;
    904 	rem = nfsm_rndup(siz)-siz;
    905 	mp = mp2 = *mq;
    906 	while (siz > 0) {
    907 		left = uiop->uio_iov->iov_len;
    908 		uiocp = uiop->uio_iov->iov_base;
    909 		if (left > siz)
    910 			left = siz;
    911 		uiosiz = left;
    912 		while (left > 0) {
    913 			mlen = M_TRAILINGSPACE(mp);
    914 			if (mlen == 0) {
    915 				MGET(mp, M_WAIT, MT_DATA);
    916 				if (clflg)
    917 					MCLGET(mp, M_WAIT);
    918 				mp->m_len = 0;
    919 				mp2->m_next = mp;
    920 				mp2 = mp;
    921 				mlen = M_TRAILINGSPACE(mp);
    922 			}
    923 			xfer = (left > mlen) ? mlen : left;
    924 #ifdef notdef
    925 			/* Not Yet.. */
    926 			if (uiop->uio_iov->iov_op != NULL)
    927 				(*(uiop->uio_iov->iov_op))
    928 				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
    929 			else
    930 #endif
    931 			if (uiop->uio_segflg == UIO_SYSSPACE)
    932 				memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer);
    933 			else
    934 				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
    935 			mp->m_len += xfer;
    936 			left -= xfer;
    937 			uiocp += xfer;
    938 			uiop->uio_offset += xfer;
    939 			uiop->uio_resid -= xfer;
    940 		}
    941 		uiop->uio_iov->iov_base = (caddr_t)uiop->uio_iov->iov_base +
    942 		    uiosiz;
    943 		uiop->uio_iov->iov_len -= uiosiz;
    944 		siz -= uiosiz;
    945 	}
    946 	if (rem > 0) {
    947 		if (rem > M_TRAILINGSPACE(mp)) {
    948 			MGET(mp, M_WAIT, MT_DATA);
    949 			mp->m_len = 0;
    950 			mp2->m_next = mp;
    951 		}
    952 		cp = mtod(mp, caddr_t)+mp->m_len;
    953 		for (left = 0; left < rem; left++)
    954 			*cp++ = '\0';
    955 		mp->m_len += rem;
    956 		*bpos = cp;
    957 	} else
    958 		*bpos = mtod(mp, caddr_t)+mp->m_len;
    959 	*mq = mp;
    960 	return (0);
    961 }
    962 
    963 /*
    964  * Get at least "siz" bytes of correctly aligned data.
    965  * When called the mbuf pointers are not necessarily correct,
    966  * dsosp points to what ought to be in m_data and left contains
    967  * what ought to be in m_len.
    968  * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
    969  * cases. (The macros use the vars. dpos and dpos2)
    970  */
    971 int
    972 nfsm_disct(mdp, dposp, siz, left, cp2)
    973 	struct mbuf **mdp;
    974 	caddr_t *dposp;
    975 	int siz;
    976 	int left;
    977 	caddr_t *cp2;
    978 {
    979 	struct mbuf *m1, *m2;
    980 	struct mbuf *havebuf = NULL;
    981 	caddr_t src = *dposp;
    982 	caddr_t dst;
    983 	int len;
    984 
    985 #ifdef DEBUG
    986 	if (left < 0)
    987 		panic("nfsm_disct: left < 0");
    988 #endif
    989 	m1 = *mdp;
    990 	/*
    991 	 * Skip through the mbuf chain looking for an mbuf with
    992 	 * some data. If the first mbuf found has enough data
    993 	 * and it is correctly aligned return it.
    994 	 */
    995 	while (left == 0) {
    996 		havebuf = m1;
    997 		*mdp = m1 = m1->m_next;
    998 		if (m1 == NULL)
    999 			return (EBADRPC);
   1000 		src = mtod(m1, caddr_t);
   1001 		left = m1->m_len;
   1002 		/*
   1003 		 * If we start a new mbuf and it is big enough
   1004 		 * and correctly aligned just return it, don't
   1005 		 * do any pull up.
   1006 		 */
   1007 		if (left >= siz && nfsm_aligned(src)) {
   1008 			*cp2 = src;
   1009 			*dposp = src + siz;
   1010 			return (0);
   1011 		}
   1012 	}
   1013 	if (m1->m_flags & M_EXT) {
   1014 		if (havebuf) {
   1015 			/* If the first mbuf with data has external data
   1016 			 * and there is a previous empty mbuf use it
   1017 			 * to move the data into.
   1018 			 */
   1019 			m2 = m1;
   1020 			*mdp = m1 = havebuf;
   1021 			if (m1->m_flags & M_EXT) {
   1022 				MEXTREMOVE(m1);
   1023 			}
   1024 		} else {
   1025 			/*
   1026 			 * If the first mbuf has a external data
   1027 			 * and there is no previous empty mbuf
   1028 			 * allocate a new mbuf and move the external
   1029 			 * data to the new mbuf. Also make the first
   1030 			 * mbuf look empty.
   1031 			 */
   1032 			m2 = m_get(M_WAIT, MT_DATA);
   1033 			m2->m_ext = m1->m_ext;
   1034 			m2->m_data = src;
   1035 			m2->m_len = left;
   1036 			MCLADDREFERENCE(m1, m2);
   1037 			MEXTREMOVE(m1);
   1038 			m2->m_next = m1->m_next;
   1039 			m1->m_next = m2;
   1040 		}
   1041 		m1->m_len = 0;
   1042 		dst = m1->m_dat;
   1043 	} else {
   1044 		/*
   1045 		 * If the first mbuf has no external data
   1046 		 * move the data to the front of the mbuf.
   1047 		 */
   1048 		if ((dst = m1->m_dat) != src)
   1049 			memmove(dst, src, left);
   1050 		dst += left;
   1051 		m1->m_len = left;
   1052 		m2 = m1->m_next;
   1053 	}
   1054 	m1->m_flags &= ~M_PKTHDR;
   1055 	*cp2 = m1->m_data = m1->m_dat;   /* data is at beginning of buffer */
   1056 	*dposp = mtod(m1, caddr_t) + siz;
   1057 	/*
   1058 	 * Loop through mbufs pulling data up into first mbuf until
   1059 	 * the first mbuf is full or there is no more data to
   1060 	 * pullup.
   1061 	 */
   1062 	while ((len = (MLEN - m1->m_len)) != 0 && m2) {
   1063 		if ((len = min(len, m2->m_len)) != 0)
   1064 			memcpy(dst, m2->m_data, len);
   1065 		m1->m_len += len;
   1066 		dst += len;
   1067 		m2->m_data += len;
   1068 		m2->m_len -= len;
   1069 		m2 = m2->m_next;
   1070 	}
   1071 	if (m1->m_len < siz)
   1072 		return (EBADRPC);
   1073 	return (0);
   1074 }
   1075 
   1076 /*
   1077  * Advance the position in the mbuf chain.
   1078  */
   1079 int
   1080 nfs_adv(mdp, dposp, offs, left)
   1081 	struct mbuf **mdp;
   1082 	caddr_t *dposp;
   1083 	int offs;
   1084 	int left;
   1085 {
   1086 	struct mbuf *m;
   1087 	int s;
   1088 
   1089 	m = *mdp;
   1090 	s = left;
   1091 	while (s < offs) {
   1092 		offs -= s;
   1093 		m = m->m_next;
   1094 		if (m == NULL)
   1095 			return (EBADRPC);
   1096 		s = m->m_len;
   1097 	}
   1098 	*mdp = m;
   1099 	*dposp = mtod(m, caddr_t)+offs;
   1100 	return (0);
   1101 }
   1102 
   1103 /*
   1104  * Copy a string into mbufs for the hard cases...
   1105  */
   1106 int
   1107 nfsm_strtmbuf(mb, bpos, cp, siz)
   1108 	struct mbuf **mb;
   1109 	char **bpos;
   1110 	const char *cp;
   1111 	long siz;
   1112 {
   1113 	struct mbuf *m1 = NULL, *m2;
   1114 	long left, xfer, len, tlen;
   1115 	u_int32_t *tl;
   1116 	int putsize;
   1117 
   1118 	putsize = 1;
   1119 	m2 = *mb;
   1120 	left = M_TRAILINGSPACE(m2);
   1121 	if (left > 0) {
   1122 		tl = ((u_int32_t *)(*bpos));
   1123 		*tl++ = txdr_unsigned(siz);
   1124 		putsize = 0;
   1125 		left -= NFSX_UNSIGNED;
   1126 		m2->m_len += NFSX_UNSIGNED;
   1127 		if (left > 0) {
   1128 			memcpy((caddr_t) tl, cp, left);
   1129 			siz -= left;
   1130 			cp += left;
   1131 			m2->m_len += left;
   1132 			left = 0;
   1133 		}
   1134 	}
   1135 	/* Loop around adding mbufs */
   1136 	while (siz > 0) {
   1137 		MGET(m1, M_WAIT, MT_DATA);
   1138 		if (siz > MLEN)
   1139 			MCLGET(m1, M_WAIT);
   1140 		m1->m_len = NFSMSIZ(m1);
   1141 		m2->m_next = m1;
   1142 		m2 = m1;
   1143 		tl = mtod(m1, u_int32_t *);
   1144 		tlen = 0;
   1145 		if (putsize) {
   1146 			*tl++ = txdr_unsigned(siz);
   1147 			m1->m_len -= NFSX_UNSIGNED;
   1148 			tlen = NFSX_UNSIGNED;
   1149 			putsize = 0;
   1150 		}
   1151 		if (siz < m1->m_len) {
   1152 			len = nfsm_rndup(siz);
   1153 			xfer = siz;
   1154 			if (xfer < len)
   1155 				*(tl+(xfer>>2)) = 0;
   1156 		} else {
   1157 			xfer = len = m1->m_len;
   1158 		}
   1159 		memcpy((caddr_t) tl, cp, xfer);
   1160 		m1->m_len = len+tlen;
   1161 		siz -= xfer;
   1162 		cp += xfer;
   1163 	}
   1164 	*mb = m1;
   1165 	*bpos = mtod(m1, caddr_t)+m1->m_len;
   1166 	return (0);
   1167 }
   1168 
   1169 /*
   1170  * Directory caching routines. They work as follows:
   1171  * - a cache is maintained per VDIR nfsnode.
   1172  * - for each offset cookie that is exported to userspace, and can
   1173  *   thus be thrown back at us as an offset to VOP_READDIR, store
   1174  *   information in the cache.
   1175  * - cached are:
   1176  *   - cookie itself
   1177  *   - blocknumber (essentially just a search key in the buffer cache)
   1178  *   - entry number in block.
   1179  *   - offset cookie of block in which this entry is stored
   1180  *   - 32 bit cookie if NFSMNT_XLATECOOKIE is used.
   1181  * - entries are looked up in a hash table
   1182  * - also maintained is an LRU list of entries, used to determine
   1183  *   which ones to delete if the cache grows too large.
   1184  * - if 32 <-> 64 translation mode is requested for a filesystem,
   1185  *   the cache also functions as a translation table
   1186  * - in the translation case, invalidating the cache does not mean
   1187  *   flushing it, but just marking entries as invalid, except for
   1188  *   the <64bit cookie, 32bitcookie> pair which is still valid, to
   1189  *   still be able to use the cache as a translation table.
   1190  * - 32 bit cookies are uniquely created by combining the hash table
   1191  *   entry value, and one generation count per hash table entry,
   1192  *   incremented each time an entry is appended to the chain.
   1193  * - the cache is invalidated each time a direcory is modified
   1194  * - sanity checks are also done; if an entry in a block turns
   1195  *   out not to have a matching cookie, the cache is invalidated
   1196  *   and a new block starting from the wanted offset is fetched from
   1197  *   the server.
   1198  * - directory entries as read from the server are extended to contain
   1199  *   the 64bit and, optionally, the 32bit cookies, for sanity checking
   1200  *   the cache and exporting them to userspace through the cookie
   1201  *   argument to VOP_READDIR.
   1202  */
   1203 
   1204 u_long
   1205 nfs_dirhash(off)
   1206 	off_t off;
   1207 {
   1208 	int i;
   1209 	char *cp = (char *)&off;
   1210 	u_long sum = 0L;
   1211 
   1212 	for (i = 0 ; i < sizeof (off); i++)
   1213 		sum += *cp++;
   1214 
   1215 	return sum;
   1216 }
   1217 
   1218 void
   1219 nfs_initdircache(vp)
   1220 	struct vnode *vp;
   1221 {
   1222 	struct nfsnode *np = VTONFS(vp);
   1223 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1224 
   1225 	np->n_dircachesize = 0;
   1226 	np->n_dblkno = 1;
   1227 	np->n_dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, M_NFSDIROFF,
   1228 	    M_WAITOK, &nfsdirhashmask);
   1229 	TAILQ_INIT(&np->n_dirchain);
   1230 	if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
   1231 		MALLOC(np->n_dirgens, unsigned *,
   1232 		    NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF,
   1233 		    M_WAITOK);
   1234 		memset((caddr_t)np->n_dirgens, 0,
   1235 		    NFS_DIRHASHSIZ * sizeof (unsigned));
   1236 	}
   1237 }
   1238 
   1239 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0};
   1240 
   1241 struct nfsdircache *
   1242 nfs_searchdircache(vp, off, do32, hashent)
   1243 	struct vnode *vp;
   1244 	off_t off;
   1245 	int do32;
   1246 	int *hashent;
   1247 {
   1248 	struct nfsdirhashhead *ndhp;
   1249 	struct nfsdircache *ndp = NULL;
   1250 	struct nfsnode *np = VTONFS(vp);
   1251 	unsigned ent;
   1252 
   1253 	/*
   1254 	 * Zero is always a valid cookie.
   1255 	 */
   1256 	if (off == 0)
   1257 		return &dzero;
   1258 
   1259 	/*
   1260 	 * We use a 32bit cookie as search key, directly reconstruct
   1261 	 * the hashentry. Else use the hashfunction.
   1262 	 */
   1263 	if (do32) {
   1264 		ent = (u_int32_t)off >> 24;
   1265 		if (ent >= NFS_DIRHASHSIZ)
   1266 			return NULL;
   1267 		ndhp = &np->n_dircache[ent];
   1268 	} else {
   1269 		ndhp = NFSDIRHASH(np, off);
   1270 	}
   1271 
   1272 	if (hashent)
   1273 		*hashent = (int)(ndhp - np->n_dircache);
   1274 	if (do32) {
   1275 		for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) {
   1276 			if (ndp->dc_cookie32 == (u_int32_t)off) {
   1277 				/*
   1278 				 * An invalidated entry will become the
   1279 				 * start of a new block fetched from
   1280 				 * the server.
   1281 				 */
   1282 				if (ndp->dc_blkno == -1) {
   1283 					ndp->dc_blkcookie = ndp->dc_cookie;
   1284 					ndp->dc_blkno = np->n_dblkno++;
   1285 					ndp->dc_entry = 0;
   1286 				}
   1287 				break;
   1288 			}
   1289 		}
   1290 	} else {
   1291 		for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next)
   1292 			if (ndp->dc_cookie == off)
   1293 				break;
   1294 	}
   1295 	return ndp;
   1296 }
   1297 
   1298 
   1299 struct nfsdircache *
   1300 nfs_enterdircache(vp, off, blkoff, en, blkno)
   1301 	struct vnode *vp;
   1302 	off_t off, blkoff;
   1303 	daddr_t blkno;
   1304 	int en;
   1305 {
   1306 	struct nfsnode *np = VTONFS(vp);
   1307 	struct nfsdirhashhead *ndhp;
   1308 	struct nfsdircache *ndp = NULL, *first;
   1309 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1310 	int hashent, gen, overwrite;
   1311 
   1312 	if (!np->n_dircache)
   1313 		/*
   1314 		 * XXX would like to do this in nfs_nget but vtype
   1315 		 * isn't known at that time.
   1316 		 */
   1317 		nfs_initdircache(vp);
   1318 
   1319 	/*
   1320 	 * XXX refuse entries for offset 0. amd(8) erroneously sets
   1321 	 * cookie 0 for the '.' entry, making this necessary. This
   1322 	 * isn't so bad, as 0 is a special case anyway.
   1323 	 */
   1324 	if (off == 0)
   1325 		return &dzero;
   1326 
   1327 	ndp = nfs_searchdircache(vp, off, 0, &hashent);
   1328 
   1329 	if (ndp && ndp->dc_blkno != -1) {
   1330 		/*
   1331 		 * Overwriting an old entry. Check if it's the same.
   1332 		 * If so, just return. If not, remove the old entry.
   1333 		 */
   1334 		if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en)
   1335 			return ndp;
   1336 		TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
   1337 		LIST_REMOVE(ndp, dc_hash);
   1338 		FREE(ndp, M_NFSDIROFF);
   1339 		ndp = 0;
   1340 	}
   1341 
   1342 	ndhp = &np->n_dircache[hashent];
   1343 
   1344 	if (!ndp) {
   1345 		MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF,
   1346 		    M_WAITOK);
   1347 		overwrite = 0;
   1348 		if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
   1349 			/*
   1350 			 * We're allocating a new entry, so bump the
   1351 			 * generation number.
   1352 			 */
   1353 			gen = ++np->n_dirgens[hashent];
   1354 			if (gen == 0) {
   1355 				np->n_dirgens[hashent]++;
   1356 				gen++;
   1357 			}
   1358 			ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff);
   1359 		}
   1360 	} else
   1361 		overwrite = 1;
   1362 
   1363 	/*
   1364 	 * If the entry number is 0, we are at the start of a new block, so
   1365 	 * allocate a new blocknumber.
   1366 	 */
   1367 	if (en == 0)
   1368 		ndp->dc_blkno = np->n_dblkno++;
   1369 	else
   1370 		ndp->dc_blkno = blkno;
   1371 
   1372 	ndp->dc_cookie = off;
   1373 	ndp->dc_blkcookie = blkoff;
   1374 	ndp->dc_entry = en;
   1375 
   1376 	if (overwrite)
   1377 		return ndp;
   1378 
   1379 	/*
   1380 	 * If the maximum directory cookie cache size has been reached
   1381 	 * for this node, take one off the front. The idea is that
   1382 	 * directories are typically read front-to-back once, so that
   1383 	 * the oldest entries can be thrown away without much performance
   1384 	 * loss.
   1385 	 */
   1386 	if (np->n_dircachesize == NFS_MAXDIRCACHE) {
   1387 		first = np->n_dirchain.tqh_first;
   1388 		TAILQ_REMOVE(&np->n_dirchain, first, dc_chain);
   1389 		LIST_REMOVE(first, dc_hash);
   1390 		FREE(first, M_NFSDIROFF);
   1391 	} else
   1392 		np->n_dircachesize++;
   1393 
   1394 	LIST_INSERT_HEAD(ndhp, ndp, dc_hash);
   1395 	TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain);
   1396 	return ndp;
   1397 }
   1398 
   1399 void
   1400 nfs_invaldircache(vp, forcefree)
   1401 	struct vnode *vp;
   1402 	int forcefree;
   1403 {
   1404 	struct nfsnode *np = VTONFS(vp);
   1405 	struct nfsdircache *ndp = NULL;
   1406 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1407 
   1408 #ifdef DIAGNOSTIC
   1409 	if (vp->v_type != VDIR)
   1410 		panic("nfs: invaldircache: not dir");
   1411 #endif
   1412 
   1413 	if (!np->n_dircache)
   1414 		return;
   1415 
   1416 	if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) {
   1417 		while ((ndp = np->n_dirchain.tqh_first)) {
   1418 			TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
   1419 			LIST_REMOVE(ndp, dc_hash);
   1420 			FREE(ndp, M_NFSDIROFF);
   1421 		}
   1422 		np->n_dircachesize = 0;
   1423 		if (forcefree && np->n_dirgens) {
   1424 			FREE(np->n_dirgens, M_NFSDIROFF);
   1425 		}
   1426 	} else {
   1427 		for (ndp = np->n_dirchain.tqh_first; ndp;
   1428 		    ndp = ndp->dc_chain.tqe_next)
   1429 			ndp->dc_blkno = -1;
   1430 	}
   1431 
   1432 	np->n_dblkno = 1;
   1433 }
   1434 
   1435 /*
   1436  * Called once before VFS init to initialize shared and
   1437  * server-specific data structures.
   1438  */
   1439 void
   1440 nfs_init()
   1441 {
   1442 	nfsrtt.pos = 0;
   1443 	rpc_vers = txdr_unsigned(RPC_VER2);
   1444 	rpc_call = txdr_unsigned(RPC_CALL);
   1445 	rpc_reply = txdr_unsigned(RPC_REPLY);
   1446 	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
   1447 	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
   1448 	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
   1449 	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
   1450 	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
   1451 	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
   1452 	nfs_prog = txdr_unsigned(NFS_PROG);
   1453 	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
   1454 	nfs_true = txdr_unsigned(TRUE);
   1455 	nfs_false = txdr_unsigned(FALSE);
   1456 	nfs_xdrneg1 = txdr_unsigned(-1);
   1457 	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
   1458 	if (nfs_ticks < 1)
   1459 		nfs_ticks = 1;
   1460 #ifdef NFSSERVER
   1461 	nfsrv_init(0);			/* Init server data structures */
   1462 	nfsrv_initcache();		/* Init the server request cache */
   1463 #endif /* NFSSERVER */
   1464 
   1465 #if defined(NFSSERVER) || !defined(NFS_V2_ONLY)
   1466 	/*
   1467 	 * Initialize the nqnfs data structures.
   1468 	 */
   1469 	if (nqnfsstarttime == 0) {
   1470 		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
   1471 			+ nqsrv_clockskew + nqsrv_writeslack;
   1472 		NQLOADNOVRAM(nqnfsstarttime);
   1473 		CIRCLEQ_INIT(&nqtimerhead);
   1474 		nqfhhashtbl = hashinit(NQLCHSZ, HASH_LIST, M_NQLEASE,
   1475 		    M_WAITOK, &nqfhhash);
   1476 	}
   1477 #endif
   1478 
   1479 	/*
   1480 	 * Initialize reply list and start timer
   1481 	 */
   1482 	TAILQ_INIT(&nfs_reqq);
   1483 	nfs_timer(NULL);
   1484 }
   1485 
   1486 #ifdef NFS
   1487 /*
   1488  * Called once at VFS init to initialize client-specific data structures.
   1489  */
   1490 void
   1491 nfs_vfs_init()
   1492 {
   1493 	nfs_nhinit();			/* Init the nfsnode table */
   1494 }
   1495 
   1496 void
   1497 nfs_vfs_reinit()
   1498 {
   1499 	nfs_nhreinit();
   1500 }
   1501 
   1502 void
   1503 nfs_vfs_done()
   1504 {
   1505 	nfs_nhdone();
   1506 }
   1507 
   1508 /*
   1509  * Attribute cache routines.
   1510  * nfs_loadattrcache() - loads or updates the cache contents from attributes
   1511  *	that are on the mbuf list
   1512  * nfs_getattrcache() - returns valid attributes if found in cache, returns
   1513  *	error otherwise
   1514  */
   1515 
   1516 /*
   1517  * Load the attribute cache (that lives in the nfsnode entry) with
   1518  * the values on the mbuf list and
   1519  * Iff vap not NULL
   1520  *    copy the attributes to *vaper
   1521  */
   1522 int
   1523 nfsm_loadattrcache(vpp, mdp, dposp, vaper)
   1524 	struct vnode **vpp;
   1525 	struct mbuf **mdp;
   1526 	caddr_t *dposp;
   1527 	struct vattr *vaper;
   1528 {
   1529 	int32_t t1;
   1530 	caddr_t cp2;
   1531 	int error = 0;
   1532 	struct mbuf *md;
   1533 	int v3 = NFS_ISV3(*vpp);
   1534 
   1535 	md = *mdp;
   1536 	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
   1537 	error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2);
   1538 	if (error)
   1539 		return (error);
   1540 	return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper);
   1541 }
   1542 
   1543 int
   1544 nfs_loadattrcache(vpp, fp, vaper)
   1545 	struct vnode **vpp;
   1546 	struct nfs_fattr *fp;
   1547 	struct vattr *vaper;
   1548 {
   1549 	struct vnode *vp = *vpp;
   1550 	struct vattr *vap;
   1551 	int v3 = NFS_ISV3(vp);
   1552 	enum vtype vtyp;
   1553 	u_short vmode;
   1554 	struct timespec mtime;
   1555 	struct vnode *nvp;
   1556 	int32_t rdev;
   1557 	struct nfsnode *np;
   1558 	extern int (**spec_nfsv2nodeop_p) __P((void *));
   1559 
   1560 	if (v3) {
   1561 		vtyp = nfsv3tov_type(fp->fa_type);
   1562 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
   1563 		rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1),
   1564 			fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2));
   1565 		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
   1566 	} else {
   1567 		vtyp = nfsv2tov_type(fp->fa_type);
   1568 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
   1569 		if (vtyp == VNON || vtyp == VREG)
   1570 			vtyp = IFTOVT(vmode);
   1571 		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
   1572 		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
   1573 
   1574 		/*
   1575 		 * Really ugly NFSv2 kludge.
   1576 		 */
   1577 		if (vtyp == VCHR && rdev == 0xffffffff)
   1578 			vtyp = VFIFO;
   1579 	}
   1580 
   1581 	/*
   1582 	 * If v_type == VNON it is a new node, so fill in the v_type,
   1583 	 * n_mtime fields. Check to see if it represents a special
   1584 	 * device, and if so, check for a possible alias. Once the
   1585 	 * correct vnode has been obtained, fill in the rest of the
   1586 	 * information.
   1587 	 */
   1588 	np = VTONFS(vp);
   1589 	if (vp->v_type == VNON) {
   1590 		vp->v_type = vtyp;
   1591 		if (vp->v_type == VFIFO) {
   1592 			extern int (**fifo_nfsv2nodeop_p) __P((void *));
   1593 			vp->v_op = fifo_nfsv2nodeop_p;
   1594 		}
   1595 		if (vp->v_type == VCHR || vp->v_type == VBLK) {
   1596 			vp->v_op = spec_nfsv2nodeop_p;
   1597 			nvp = checkalias(vp, (dev_t)rdev, vp->v_mount);
   1598 			if (nvp) {
   1599 				/*
   1600 				 * Discard unneeded vnode, but save its nfsnode.
   1601 				 * Since the nfsnode does not have a lock, its
   1602 				 * vnode lock has to be carried over.
   1603 				 */
   1604 				/*
   1605 				 * XXX is the old node sure to be locked here?
   1606 				 */
   1607 				KASSERT(lockstatus(&vp->v_lock) ==
   1608 				    LK_EXCLUSIVE);
   1609 				nvp->v_data = vp->v_data;
   1610 				vp->v_data = NULL;
   1611 				VOP_UNLOCK(vp, 0);
   1612 				vp->v_op = spec_vnodeop_p;
   1613 				vrele(vp);
   1614 				vgone(vp);
   1615 				lockmgr(&nvp->v_lock, LK_EXCLUSIVE,
   1616 				    &nvp->v_interlock);
   1617 				/*
   1618 				 * Reinitialize aliased node.
   1619 				 */
   1620 				np->n_vnode = nvp;
   1621 				*vpp = vp = nvp;
   1622 			}
   1623 		}
   1624 		np->n_mtime = mtime.tv_sec;
   1625 	}
   1626 	vap = np->n_vattr;
   1627 	vap->va_type = vtyp;
   1628 	vap->va_mode = vmode & ALLPERMS;
   1629 	vap->va_rdev = (dev_t)rdev;
   1630 	vap->va_mtime = mtime;
   1631 	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
   1632 	switch (vtyp) {
   1633 	case VDIR:
   1634 		vap->va_blocksize = NFS_DIRFRAGSIZ;
   1635 		break;
   1636 	case VBLK:
   1637 		vap->va_blocksize = BLKDEV_IOSIZE;
   1638 		break;
   1639 	case VCHR:
   1640 		vap->va_blocksize = MAXBSIZE;
   1641 		break;
   1642 	default:
   1643 		vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize :
   1644 		    fxdr_unsigned(int32_t, fp->fa2_blocksize);
   1645 		break;
   1646 	}
   1647 	if (v3) {
   1648 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
   1649 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
   1650 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
   1651 		vap->va_size = fxdr_hyper(&fp->fa3_size);
   1652 		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
   1653 		vap->va_fileid = fxdr_unsigned(int32_t,
   1654 		    fp->fa3_fileid.nfsuquad[1]);
   1655 		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
   1656 		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
   1657 		vap->va_flags = 0;
   1658 		vap->va_filerev = 0;
   1659 	} else {
   1660 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
   1661 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
   1662 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
   1663 		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
   1664 		vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks)
   1665 		    * NFS_FABLKSIZE;
   1666 		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
   1667 		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
   1668 		vap->va_flags = 0;
   1669 		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
   1670 		    fp->fa2_ctime.nfsv2_sec);
   1671 		vap->va_ctime.tv_nsec = 0;
   1672 		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
   1673 		vap->va_filerev = 0;
   1674 	}
   1675 	if (vap->va_size != np->n_size) {
   1676 		if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) {
   1677 			vap->va_size = np->n_size;
   1678 		} else {
   1679 			np->n_size = vap->va_size;
   1680 			if (vap->va_type == VREG) {
   1681 				uvm_vnp_setsize(vp, np->n_size);
   1682 			}
   1683 		}
   1684 	}
   1685 	np->n_attrstamp = time.tv_sec;
   1686 	if (vaper != NULL) {
   1687 		memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap));
   1688 		if (np->n_flag & NCHG) {
   1689 			if (np->n_flag & NACC)
   1690 				vaper->va_atime = np->n_atim;
   1691 			if (np->n_flag & NUPD)
   1692 				vaper->va_mtime = np->n_mtim;
   1693 		}
   1694 	}
   1695 	return (0);
   1696 }
   1697 
   1698 /*
   1699  * Check the time stamp
   1700  * If the cache is valid, copy contents to *vap and return 0
   1701  * otherwise return an error
   1702  */
   1703 int
   1704 nfs_getattrcache(vp, vaper)
   1705 	struct vnode *vp;
   1706 	struct vattr *vaper;
   1707 {
   1708 	struct nfsnode *np = VTONFS(vp);
   1709 	struct vattr *vap;
   1710 
   1711 	if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) {
   1712 		nfsstats.attrcache_misses++;
   1713 		return (ENOENT);
   1714 	}
   1715 	nfsstats.attrcache_hits++;
   1716 	vap = np->n_vattr;
   1717 	if (vap->va_size != np->n_size) {
   1718 		if (vap->va_type == VREG) {
   1719 			if (np->n_flag & NMODIFIED) {
   1720 				if (vap->va_size < np->n_size)
   1721 					vap->va_size = np->n_size;
   1722 				else
   1723 					np->n_size = vap->va_size;
   1724 			} else
   1725 				np->n_size = vap->va_size;
   1726 			uvm_vnp_setsize(vp, np->n_size);
   1727 		} else
   1728 			np->n_size = vap->va_size;
   1729 	}
   1730 	memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr));
   1731 	if (np->n_flag & NCHG) {
   1732 		if (np->n_flag & NACC)
   1733 			vaper->va_atime = np->n_atim;
   1734 		if (np->n_flag & NUPD)
   1735 			vaper->va_mtime = np->n_mtim;
   1736 	}
   1737 	return (0);
   1738 }
   1739 
   1740 /*
   1741  * Heuristic to see if the server XDR encodes directory cookies or not.
   1742  * it is not supposed to, but a lot of servers may do this. Also, since
   1743  * most/all servers will implement V2 as well, it is expected that they
   1744  * may return just 32 bits worth of cookie information, so we need to
   1745  * find out in which 32 bits this information is available. We do this
   1746  * to avoid trouble with emulated binaries that can't handle 64 bit
   1747  * directory offsets.
   1748  */
   1749 
   1750 void
   1751 nfs_cookieheuristic(vp, flagp, p, cred)
   1752 	struct vnode *vp;
   1753 	int *flagp;
   1754 	struct proc *p;
   1755 	struct ucred *cred;
   1756 {
   1757 	struct uio auio;
   1758 	struct iovec aiov;
   1759 	caddr_t buf, cp;
   1760 	struct dirent *dp;
   1761 	off_t *cookies = NULL, *cop;
   1762 	int error, eof, nc, len;
   1763 
   1764 	MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK);
   1765 
   1766 	aiov.iov_base = buf;
   1767 	aiov.iov_len = NFS_DIRFRAGSIZ;
   1768 	auio.uio_iov = &aiov;
   1769 	auio.uio_iovcnt = 1;
   1770 	auio.uio_rw = UIO_READ;
   1771 	auio.uio_segflg = UIO_SYSSPACE;
   1772 	auio.uio_procp = p;
   1773 	auio.uio_resid = NFS_DIRFRAGSIZ;
   1774 	auio.uio_offset = 0;
   1775 
   1776 	error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc);
   1777 
   1778 	len = NFS_DIRFRAGSIZ - auio.uio_resid;
   1779 	if (error || len == 0) {
   1780 		FREE(buf, M_TEMP);
   1781 		if (cookies)
   1782 			free(cookies, M_TEMP);
   1783 		return;
   1784 	}
   1785 
   1786 	/*
   1787 	 * Find the first valid entry and look at its offset cookie.
   1788 	 */
   1789 
   1790 	cp = buf;
   1791 	for (cop = cookies; len > 0; len -= dp->d_reclen) {
   1792 		dp = (struct dirent *)cp;
   1793 		if (dp->d_fileno != 0 && len >= dp->d_reclen) {
   1794 			if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) {
   1795 				*flagp |= NFSMNT_SWAPCOOKIE;
   1796 				nfs_invaldircache(vp, 0);
   1797 				nfs_vinvalbuf(vp, 0, cred, p, 1);
   1798 			}
   1799 			break;
   1800 		}
   1801 		cop++;
   1802 		cp += dp->d_reclen;
   1803 	}
   1804 
   1805 	FREE(buf, M_TEMP);
   1806 	free(cookies, M_TEMP);
   1807 }
   1808 #endif /* NFS */
   1809 
   1810 /*
   1811  * Set up nameidata for a lookup() call and do it.
   1812  *
   1813  * If pubflag is set, this call is done for a lookup operation on the
   1814  * public filehandle. In that case we allow crossing mountpoints and
   1815  * absolute pathnames. However, the caller is expected to check that
   1816  * the lookup result is within the public fs, and deny access if
   1817  * it is not.
   1818  */
   1819 int
   1820 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
   1821 	struct nameidata *ndp;
   1822 	fhandle_t *fhp;
   1823 	int len;
   1824 	struct nfssvc_sock *slp;
   1825 	struct mbuf *nam;
   1826 	struct mbuf **mdp;
   1827 	caddr_t *dposp;
   1828 	struct vnode **retdirp;
   1829 	struct proc *p;
   1830 	int kerbflag, pubflag;
   1831 {
   1832 	int i, rem;
   1833 	struct mbuf *md;
   1834 	char *fromcp, *tocp, *cp;
   1835 	struct iovec aiov;
   1836 	struct uio auio;
   1837 	struct vnode *dp;
   1838 	int error, rdonly, linklen;
   1839 	struct componentname *cnp = &ndp->ni_cnd;
   1840 
   1841 	*retdirp = (struct vnode *)0;
   1842 
   1843 	if ((len + 1) > MAXPATHLEN)
   1844 		return (ENAMETOOLONG);
   1845 	cnp->cn_pnbuf = PNBUF_GET();
   1846 
   1847 	/*
   1848 	 * Copy the name from the mbuf list to ndp->ni_pnbuf
   1849 	 * and set the various ndp fields appropriately.
   1850 	 */
   1851 	fromcp = *dposp;
   1852 	tocp = cnp->cn_pnbuf;
   1853 	md = *mdp;
   1854 	rem = mtod(md, caddr_t) + md->m_len - fromcp;
   1855 	for (i = 0; i < len; i++) {
   1856 		while (rem == 0) {
   1857 			md = md->m_next;
   1858 			if (md == NULL) {
   1859 				error = EBADRPC;
   1860 				goto out;
   1861 			}
   1862 			fromcp = mtod(md, caddr_t);
   1863 			rem = md->m_len;
   1864 		}
   1865 		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
   1866 			error = EACCES;
   1867 			goto out;
   1868 		}
   1869 		*tocp++ = *fromcp++;
   1870 		rem--;
   1871 	}
   1872 	*tocp = '\0';
   1873 	*mdp = md;
   1874 	*dposp = fromcp;
   1875 	len = nfsm_rndup(len)-len;
   1876 	if (len > 0) {
   1877 		if (rem >= len)
   1878 			*dposp += len;
   1879 		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
   1880 			goto out;
   1881 	}
   1882 
   1883 	/*
   1884 	 * Extract and set starting directory.
   1885 	 */
   1886 	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
   1887 	    nam, &rdonly, kerbflag, pubflag);
   1888 	if (error)
   1889 		goto out;
   1890 	if (dp->v_type != VDIR) {
   1891 		vrele(dp);
   1892 		error = ENOTDIR;
   1893 		goto out;
   1894 	}
   1895 
   1896 	if (rdonly)
   1897 		cnp->cn_flags |= RDONLY;
   1898 
   1899 	*retdirp = dp;
   1900 
   1901 	if (pubflag) {
   1902 		/*
   1903 		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
   1904 		 * and the 'native path' indicator.
   1905 		 */
   1906 		cp = PNBUF_GET();
   1907 		fromcp = cnp->cn_pnbuf;
   1908 		tocp = cp;
   1909 		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
   1910 			switch ((unsigned char)*fromcp) {
   1911 			case WEBNFS_NATIVE_CHAR:
   1912 				/*
   1913 				 * 'Native' path for us is the same
   1914 				 * as a path according to the NFS spec,
   1915 				 * just skip the escape char.
   1916 				 */
   1917 				fromcp++;
   1918 				break;
   1919 			/*
   1920 			 * More may be added in the future, range 0x80-0xff
   1921 			 */
   1922 			default:
   1923 				error = EIO;
   1924 				FREE(cp, M_NAMEI);
   1925 				goto out;
   1926 			}
   1927 		}
   1928 		/*
   1929 		 * Translate the '%' escapes, URL-style.
   1930 		 */
   1931 		while (*fromcp != '\0') {
   1932 			if (*fromcp == WEBNFS_ESC_CHAR) {
   1933 				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
   1934 					fromcp++;
   1935 					*tocp++ = HEXSTRTOI(fromcp);
   1936 					fromcp += 2;
   1937 					continue;
   1938 				} else {
   1939 					error = ENOENT;
   1940 					FREE(cp, M_NAMEI);
   1941 					goto out;
   1942 				}
   1943 			} else
   1944 				*tocp++ = *fromcp++;
   1945 		}
   1946 		*tocp = '\0';
   1947 		PNBUF_PUT(cnp->cn_pnbuf);
   1948 		cnp->cn_pnbuf = cp;
   1949 	}
   1950 
   1951 	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
   1952 	ndp->ni_segflg = UIO_SYSSPACE;
   1953 	ndp->ni_rootdir = rootvnode;
   1954 
   1955 	if (pubflag) {
   1956 		ndp->ni_loopcnt = 0;
   1957 		if (cnp->cn_pnbuf[0] == '/')
   1958 			dp = rootvnode;
   1959 	} else {
   1960 		cnp->cn_flags |= NOCROSSMOUNT;
   1961 	}
   1962 
   1963 	cnp->cn_proc = p;
   1964 	VREF(dp);
   1965 
   1966     for (;;) {
   1967 	cnp->cn_nameptr = cnp->cn_pnbuf;
   1968 	ndp->ni_startdir = dp;
   1969 	/*
   1970 	 * And call lookup() to do the real work
   1971 	 */
   1972 	error = lookup(ndp);
   1973 	if (error) {
   1974 		PNBUF_PUT(cnp->cn_pnbuf);
   1975 		return (error);
   1976 	}
   1977 	/*
   1978 	 * Check for encountering a symbolic link
   1979 	 */
   1980 	if ((cnp->cn_flags & ISSYMLINK) == 0) {
   1981 		if (cnp->cn_flags & (SAVENAME | SAVESTART))
   1982 			cnp->cn_flags |= HASBUF;
   1983 		else
   1984 			PNBUF_PUT(cnp->cn_pnbuf);
   1985 		return (0);
   1986 	} else {
   1987 		if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN))
   1988 			VOP_UNLOCK(ndp->ni_dvp, 0);
   1989 		if (!pubflag) {
   1990 			error = EINVAL;
   1991 			break;
   1992 		}
   1993 
   1994 		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
   1995 			error = ELOOP;
   1996 			break;
   1997 		}
   1998 		if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) {
   1999 			error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred,
   2000 			    cnp->cn_proc);
   2001 			if (error != 0)
   2002 				break;
   2003 		}
   2004 		if (ndp->ni_pathlen > 1)
   2005 			cp = PNBUF_GET();
   2006 		else
   2007 			cp = cnp->cn_pnbuf;
   2008 		aiov.iov_base = cp;
   2009 		aiov.iov_len = MAXPATHLEN;
   2010 		auio.uio_iov = &aiov;
   2011 		auio.uio_iovcnt = 1;
   2012 		auio.uio_offset = 0;
   2013 		auio.uio_rw = UIO_READ;
   2014 		auio.uio_segflg = UIO_SYSSPACE;
   2015 		auio.uio_procp = (struct proc *)0;
   2016 		auio.uio_resid = MAXPATHLEN;
   2017 		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
   2018 		if (error) {
   2019 		badlink:
   2020 			if (ndp->ni_pathlen > 1)
   2021 				PNBUF_PUT(cp);
   2022 			break;
   2023 		}
   2024 		linklen = MAXPATHLEN - auio.uio_resid;
   2025 		if (linklen == 0) {
   2026 			error = ENOENT;
   2027 			goto badlink;
   2028 		}
   2029 		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
   2030 			error = ENAMETOOLONG;
   2031 			goto badlink;
   2032 		}
   2033 		if (ndp->ni_pathlen > 1) {
   2034 			memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen);
   2035 			PNBUF_PUT(cnp->cn_pnbuf);
   2036 			cnp->cn_pnbuf = cp;
   2037 		} else
   2038 			cnp->cn_pnbuf[linklen] = '\0';
   2039 		ndp->ni_pathlen += linklen;
   2040 		vput(ndp->ni_vp);
   2041 		dp = ndp->ni_dvp;
   2042 		/*
   2043 		 * Check if root directory should replace current directory.
   2044 		 */
   2045 		if (cnp->cn_pnbuf[0] == '/') {
   2046 			vrele(dp);
   2047 			dp = ndp->ni_rootdir;
   2048 			VREF(dp);
   2049 		}
   2050 	}
   2051    }
   2052 	vrele(ndp->ni_dvp);
   2053 	vput(ndp->ni_vp);
   2054 	ndp->ni_vp = NULL;
   2055 out:
   2056 	PNBUF_PUT(cnp->cn_pnbuf);
   2057 	return (error);
   2058 }
   2059 
   2060 /*
   2061  * A fiddled version of m_adj() that ensures null fill to a long
   2062  * boundary and only trims off the back end
   2063  */
   2064 void
   2065 nfsm_adj(mp, len, nul)
   2066 	struct mbuf *mp;
   2067 	int len;
   2068 	int nul;
   2069 {
   2070 	struct mbuf *m;
   2071 	int count, i;
   2072 	char *cp;
   2073 
   2074 	/*
   2075 	 * Trim from tail.  Scan the mbuf chain,
   2076 	 * calculating its length and finding the last mbuf.
   2077 	 * If the adjustment only affects this mbuf, then just
   2078 	 * adjust and return.  Otherwise, rescan and truncate
   2079 	 * after the remaining size.
   2080 	 */
   2081 	count = 0;
   2082 	m = mp;
   2083 	for (;;) {
   2084 		count += m->m_len;
   2085 		if (m->m_next == (struct mbuf *)0)
   2086 			break;
   2087 		m = m->m_next;
   2088 	}
   2089 	if (m->m_len > len) {
   2090 		m->m_len -= len;
   2091 		if (nul > 0) {
   2092 			cp = mtod(m, caddr_t)+m->m_len-nul;
   2093 			for (i = 0; i < nul; i++)
   2094 				*cp++ = '\0';
   2095 		}
   2096 		return;
   2097 	}
   2098 	count -= len;
   2099 	if (count < 0)
   2100 		count = 0;
   2101 	/*
   2102 	 * Correct length for chain is "count".
   2103 	 * Find the mbuf with last data, adjust its length,
   2104 	 * and toss data from remaining mbufs on chain.
   2105 	 */
   2106 	for (m = mp; m; m = m->m_next) {
   2107 		if (m->m_len >= count) {
   2108 			m->m_len = count;
   2109 			if (nul > 0) {
   2110 				cp = mtod(m, caddr_t)+m->m_len-nul;
   2111 				for (i = 0; i < nul; i++)
   2112 					*cp++ = '\0';
   2113 			}
   2114 			break;
   2115 		}
   2116 		count -= m->m_len;
   2117 	}
   2118 	for (m = m->m_next;m;m = m->m_next)
   2119 		m->m_len = 0;
   2120 }
   2121 
   2122 /*
   2123  * Make these functions instead of macros, so that the kernel text size
   2124  * doesn't get too big...
   2125  */
   2126 void
   2127 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
   2128 	struct nfsrv_descript *nfsd;
   2129 	int before_ret;
   2130 	struct vattr *before_vap;
   2131 	int after_ret;
   2132 	struct vattr *after_vap;
   2133 	struct mbuf **mbp;
   2134 	char **bposp;
   2135 {
   2136 	struct mbuf *mb = *mbp, *mb2;
   2137 	char *bpos = *bposp;
   2138 	u_int32_t *tl;
   2139 
   2140 	if (before_ret) {
   2141 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
   2142 		*tl = nfs_false;
   2143 	} else {
   2144 		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
   2145 		*tl++ = nfs_true;
   2146 		txdr_hyper(before_vap->va_size, tl);
   2147 		tl += 2;
   2148 		txdr_nfsv3time(&(before_vap->va_mtime), tl);
   2149 		tl += 2;
   2150 		txdr_nfsv3time(&(before_vap->va_ctime), tl);
   2151 	}
   2152 	*bposp = bpos;
   2153 	*mbp = mb;
   2154 	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
   2155 }
   2156 
   2157 void
   2158 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
   2159 	struct nfsrv_descript *nfsd;
   2160 	int after_ret;
   2161 	struct vattr *after_vap;
   2162 	struct mbuf **mbp;
   2163 	char **bposp;
   2164 {
   2165 	struct mbuf *mb = *mbp, *mb2;
   2166 	char *bpos = *bposp;
   2167 	u_int32_t *tl;
   2168 	struct nfs_fattr *fp;
   2169 
   2170 	if (after_ret) {
   2171 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
   2172 		*tl = nfs_false;
   2173 	} else {
   2174 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
   2175 		*tl++ = nfs_true;
   2176 		fp = (struct nfs_fattr *)tl;
   2177 		nfsm_srvfattr(nfsd, after_vap, fp);
   2178 	}
   2179 	*mbp = mb;
   2180 	*bposp = bpos;
   2181 }
   2182 
   2183 void
   2184 nfsm_srvfattr(nfsd, vap, fp)
   2185 	struct nfsrv_descript *nfsd;
   2186 	struct vattr *vap;
   2187 	struct nfs_fattr *fp;
   2188 {
   2189 
   2190 	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
   2191 	fp->fa_uid = txdr_unsigned(vap->va_uid);
   2192 	fp->fa_gid = txdr_unsigned(vap->va_gid);
   2193 	if (nfsd->nd_flag & ND_NFSV3) {
   2194 		fp->fa_type = vtonfsv3_type(vap->va_type);
   2195 		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
   2196 		txdr_hyper(vap->va_size, &fp->fa3_size);
   2197 		txdr_hyper(vap->va_bytes, &fp->fa3_used);
   2198 		fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev));
   2199 		fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev));
   2200 		fp->fa3_fsid.nfsuquad[0] = 0;
   2201 		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
   2202 		fp->fa3_fileid.nfsuquad[0] = 0;
   2203 		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
   2204 		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
   2205 		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
   2206 		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
   2207 	} else {
   2208 		fp->fa_type = vtonfsv2_type(vap->va_type);
   2209 		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
   2210 		fp->fa2_size = txdr_unsigned(vap->va_size);
   2211 		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
   2212 		if (vap->va_type == VFIFO)
   2213 			fp->fa2_rdev = 0xffffffff;
   2214 		else
   2215 			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
   2216 		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
   2217 		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
   2218 		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
   2219 		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
   2220 		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
   2221 		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
   2222 	}
   2223 }
   2224 
   2225 /*
   2226  * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
   2227  * 	- look up fsid in mount list (if not found ret error)
   2228  *	- get vp and export rights by calling VFS_FHTOVP()
   2229  *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
   2230  *	- if not lockflag unlock it with VOP_UNLOCK()
   2231  */
   2232 int
   2233 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
   2234 	fhandle_t *fhp;
   2235 	int lockflag;
   2236 	struct vnode **vpp;
   2237 	struct ucred *cred;
   2238 	struct nfssvc_sock *slp;
   2239 	struct mbuf *nam;
   2240 	int *rdonlyp;
   2241 	int kerbflag;
   2242 {
   2243 	struct mount *mp;
   2244 	int i;
   2245 	struct ucred *credanon;
   2246 	int error, exflags;
   2247 	struct sockaddr_in *saddr;
   2248 
   2249 	*vpp = (struct vnode *)0;
   2250 
   2251 	if (nfs_ispublicfh(fhp)) {
   2252 		if (!pubflag || !nfs_pub.np_valid)
   2253 			return (ESTALE);
   2254 		fhp = &nfs_pub.np_handle;
   2255 	}
   2256 
   2257 	mp = vfs_getvfs(&fhp->fh_fsid);
   2258 	if (!mp)
   2259 		return (ESTALE);
   2260 	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
   2261 	if (error)
   2262 		return (error);
   2263 	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
   2264 	if (error)
   2265 		return (error);
   2266 
   2267 	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
   2268 		saddr = mtod(nam, struct sockaddr_in *);
   2269 		if ((saddr->sin_family == AF_INET) &&
   2270 		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
   2271 			vput(*vpp);
   2272 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2273 		}
   2274 #ifdef INET6
   2275 		if ((saddr->sin_family == AF_INET6) &&
   2276 		    ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) {
   2277 			vput(*vpp);
   2278 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2279 		}
   2280 #endif
   2281 	}
   2282 	/*
   2283 	 * Check/setup credentials.
   2284 	 */
   2285 	if (exflags & MNT_EXKERB) {
   2286 		if (!kerbflag) {
   2287 			vput(*vpp);
   2288 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2289 		}
   2290 	} else if (kerbflag) {
   2291 		vput(*vpp);
   2292 		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2293 	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
   2294 		cred->cr_uid = credanon->cr_uid;
   2295 		cred->cr_gid = credanon->cr_gid;
   2296 		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
   2297 			cred->cr_groups[i] = credanon->cr_groups[i];
   2298 		cred->cr_ngroups = i;
   2299 	}
   2300 	if (exflags & MNT_EXRDONLY)
   2301 		*rdonlyp = 1;
   2302 	else
   2303 		*rdonlyp = 0;
   2304 	if (!lockflag)
   2305 		VOP_UNLOCK(*vpp, 0);
   2306 	return (0);
   2307 }
   2308 
   2309 /*
   2310  * WebNFS: check if a filehandle is a public filehandle. For v3, this
   2311  * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
   2312  * transformed this to all zeroes in both cases, so check for it.
   2313  */
   2314 int
   2315 nfs_ispublicfh(fhp)
   2316 	fhandle_t *fhp;
   2317 {
   2318 	char *cp = (char *)fhp;
   2319 	int i;
   2320 
   2321 	for (i = 0; i < NFSX_V3FH; i++)
   2322 		if (*cp++ != 0)
   2323 			return (FALSE);
   2324 	return (TRUE);
   2325 }
   2326 
   2327 /*
   2328  * This function compares two net addresses by family and returns TRUE
   2329  * if they are the same host.
   2330  * If there is any doubt, return FALSE.
   2331  * The AF_INET family is handled as a special case so that address mbufs
   2332  * don't need to be saved to store "struct in_addr", which is only 4 bytes.
   2333  */
   2334 int
   2335 netaddr_match(family, haddr, nam)
   2336 	int family;
   2337 	union nethostaddr *haddr;
   2338 	struct mbuf *nam;
   2339 {
   2340 	struct sockaddr_in *inetaddr;
   2341 
   2342 	switch (family) {
   2343 	case AF_INET:
   2344 		inetaddr = mtod(nam, struct sockaddr_in *);
   2345 		if (inetaddr->sin_family == AF_INET &&
   2346 		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
   2347 			return (1);
   2348 		break;
   2349 #ifdef INET6
   2350 	case AF_INET6:
   2351 	    {
   2352 		struct sockaddr_in6 *sin6_1, *sin6_2;
   2353 
   2354 		sin6_1 = mtod(nam, struct sockaddr_in6 *);
   2355 		sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *);
   2356 		if (sin6_1->sin6_family == AF_INET6 &&
   2357 		    IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr))
   2358 			return 1;
   2359 	    }
   2360 #endif
   2361 #ifdef ISO
   2362 	case AF_ISO:
   2363 	    {
   2364 		struct sockaddr_iso *isoaddr1, *isoaddr2;
   2365 
   2366 		isoaddr1 = mtod(nam, struct sockaddr_iso *);
   2367 		isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *);
   2368 		if (isoaddr1->siso_family == AF_ISO &&
   2369 		    isoaddr1->siso_nlen > 0 &&
   2370 		    isoaddr1->siso_nlen == isoaddr2->siso_nlen &&
   2371 		    SAME_ISOADDR(isoaddr1, isoaddr2))
   2372 			return (1);
   2373 		break;
   2374 	    }
   2375 #endif	/* ISO */
   2376 	default:
   2377 		break;
   2378 	};
   2379 	return (0);
   2380 }
   2381 
   2382 /*
   2383  * The write verifier has changed (probably due to a server reboot), so all
   2384  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
   2385  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
   2386  * flag. Once done the new write verifier can be set for the mount point.
   2387  */
   2388 void
   2389 nfs_clearcommit(mp)
   2390 	struct mount *mp;
   2391 {
   2392 	struct vnode *vp;
   2393 	struct nfsnode *np;
   2394 	struct vm_page *pg;
   2395 	int s;
   2396 
   2397 	s = splbio();
   2398 	LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) {
   2399 		KASSERT(vp->v_mount == mp);
   2400 		if (vp->v_type == VNON)
   2401 			continue;
   2402 		np = VTONFS(vp);
   2403 		np->n_pushlo = np->n_pushhi = np->n_pushedlo =
   2404 		    np->n_pushedhi = 0;
   2405 		np->n_commitflags &=
   2406 		    ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID);
   2407 		simple_lock(&vp->v_uobj.vmobjlock);
   2408 		TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) {
   2409 			pg->flags &= ~PG_NEEDCOMMIT;
   2410 		}
   2411 		simple_unlock(&vp->v_uobj.vmobjlock);
   2412 	}
   2413 	splx(s);
   2414 }
   2415 
   2416 void
   2417 nfs_merge_commit_ranges(vp)
   2418 	struct vnode *vp;
   2419 {
   2420 	struct nfsnode *np = VTONFS(vp);
   2421 
   2422 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) {
   2423 		np->n_pushedlo = np->n_pushlo;
   2424 		np->n_pushedhi = np->n_pushhi;
   2425 		np->n_commitflags |= NFS_COMMIT_PUSHED_VALID;
   2426 	} else {
   2427 		if (np->n_pushlo < np->n_pushedlo)
   2428 			np->n_pushedlo = np->n_pushlo;
   2429 		if (np->n_pushhi > np->n_pushedhi)
   2430 			np->n_pushedhi = np->n_pushhi;
   2431 	}
   2432 
   2433 	np->n_pushlo = np->n_pushhi = 0;
   2434 	np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID;
   2435 
   2436 #ifdef fvdl_debug
   2437 	printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2438 	    (unsigned)np->n_pushedhi);
   2439 #endif
   2440 }
   2441 
   2442 int
   2443 nfs_in_committed_range(vp, off, len)
   2444 	struct vnode *vp;
   2445 	off_t off, len;
   2446 {
   2447 	struct nfsnode *np = VTONFS(vp);
   2448 	off_t lo, hi;
   2449 
   2450 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID))
   2451 		return 0;
   2452 	lo = off;
   2453 	hi = lo + len;
   2454 
   2455 	return (lo >= np->n_pushedlo && hi <= np->n_pushedhi);
   2456 }
   2457 
   2458 int
   2459 nfs_in_tobecommitted_range(vp, off, len)
   2460 	struct vnode *vp;
   2461 	off_t off, len;
   2462 {
   2463 	struct nfsnode *np = VTONFS(vp);
   2464 	off_t lo, hi;
   2465 
   2466 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID))
   2467 		return 0;
   2468 	lo = off;
   2469 	hi = lo + len;
   2470 
   2471 	return (lo >= np->n_pushlo && hi <= np->n_pushhi);
   2472 }
   2473 
   2474 void
   2475 nfs_add_committed_range(vp, off, len)
   2476 	struct vnode *vp;
   2477 	off_t off, len;
   2478 {
   2479 	struct nfsnode *np = VTONFS(vp);
   2480 	off_t lo, hi;
   2481 
   2482 	lo = off;
   2483 	hi = lo + len;
   2484 
   2485 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) {
   2486 		np->n_pushedlo = lo;
   2487 		np->n_pushedhi = hi;
   2488 		np->n_commitflags |= NFS_COMMIT_PUSHED_VALID;
   2489 	} else {
   2490 		if (hi > np->n_pushedhi)
   2491 			np->n_pushedhi = hi;
   2492 		if (lo < np->n_pushedlo)
   2493 			np->n_pushedlo = lo;
   2494 	}
   2495 #ifdef fvdl_debug
   2496 	printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2497 	    (unsigned)np->n_pushedhi);
   2498 #endif
   2499 }
   2500 
   2501 void
   2502 nfs_del_committed_range(vp, off, len)
   2503 	struct vnode *vp;
   2504 	off_t off, len;
   2505 {
   2506 	struct nfsnode *np = VTONFS(vp);
   2507 	off_t lo, hi;
   2508 
   2509 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID))
   2510 		return;
   2511 
   2512 	lo = off;
   2513 	hi = lo + len;
   2514 
   2515 	if (lo > np->n_pushedhi || hi < np->n_pushedlo)
   2516 		return;
   2517 	if (lo <= np->n_pushedlo)
   2518 		np->n_pushedlo = hi;
   2519 	else if (hi >= np->n_pushedhi)
   2520 		np->n_pushedhi = lo;
   2521 	else {
   2522 		/*
   2523 		 * XXX There's only one range. If the deleted range
   2524 		 * is in the middle, pick the largest of the
   2525 		 * contiguous ranges that it leaves.
   2526 		 */
   2527 		if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi))
   2528 			np->n_pushedhi = lo;
   2529 		else
   2530 			np->n_pushedlo = hi;
   2531 	}
   2532 #ifdef fvdl_debug
   2533 	printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2534 	    (unsigned)np->n_pushedhi);
   2535 #endif
   2536 }
   2537 
   2538 void
   2539 nfs_add_tobecommitted_range(vp, off, len)
   2540 	struct vnode *vp;
   2541 	off_t off, len;
   2542 {
   2543 	struct nfsnode *np = VTONFS(vp);
   2544 	off_t lo, hi;
   2545 
   2546 	lo = off;
   2547 	hi = lo + len;
   2548 
   2549 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) {
   2550 		np->n_pushlo = lo;
   2551 		np->n_pushhi = hi;
   2552 		np->n_commitflags |= NFS_COMMIT_PUSH_VALID;
   2553 	} else {
   2554 		if (lo < np->n_pushlo)
   2555 			np->n_pushlo = lo;
   2556 		if (hi > np->n_pushhi)
   2557 			np->n_pushhi = hi;
   2558 	}
   2559 #ifdef fvdl_debug
   2560 	printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo,
   2561 	    (unsigned)np->n_pushhi);
   2562 #endif
   2563 }
   2564 
   2565 void
   2566 nfs_del_tobecommitted_range(vp, off, len)
   2567 	struct vnode *vp;
   2568 	off_t off, len;
   2569 {
   2570 	struct nfsnode *np = VTONFS(vp);
   2571 	off_t lo, hi;
   2572 
   2573 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID))
   2574 		return;
   2575 
   2576 	lo = off;
   2577 	hi = lo + len;
   2578 
   2579 	if (lo > np->n_pushhi || hi < np->n_pushlo)
   2580 		return;
   2581 
   2582 	if (lo <= np->n_pushlo)
   2583 		np->n_pushlo = hi;
   2584 	else if (hi >= np->n_pushhi)
   2585 		np->n_pushhi = lo;
   2586 	else {
   2587 		/*
   2588 		 * XXX There's only one range. If the deleted range
   2589 		 * is in the middle, pick the largest of the
   2590 		 * contiguous ranges that it leaves.
   2591 		 */
   2592 		if ((np->n_pushlo - lo) > (hi - np->n_pushhi))
   2593 			np->n_pushhi = lo;
   2594 		else
   2595 			np->n_pushlo = hi;
   2596 	}
   2597 #ifdef fvdl_debug
   2598 	printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo,
   2599 	    (unsigned)np->n_pushhi);
   2600 #endif
   2601 }
   2602 
   2603 /*
   2604  * Map errnos to NFS error numbers. For Version 3 also filter out error
   2605  * numbers not specified for the associated procedure.
   2606  */
   2607 int
   2608 nfsrv_errmap(nd, err)
   2609 	struct nfsrv_descript *nd;
   2610 	int err;
   2611 {
   2612 	const short *defaulterrp, *errp;
   2613 
   2614 	if (nd->nd_flag & ND_NFSV3) {
   2615 	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
   2616 		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
   2617 		while (*++errp) {
   2618 			if (*errp == err)
   2619 				return (err);
   2620 			else if (*errp > err)
   2621 				break;
   2622 		}
   2623 		return ((int)*defaulterrp);
   2624 	    } else
   2625 		return (err & 0xffff);
   2626 	}
   2627 	if (err <= ELAST)
   2628 		return ((int)nfsrv_v2errmap[err - 1]);
   2629 	return (NFSERR_IO);
   2630 }
   2631 
   2632 /*
   2633  * Sort the group list in increasing numerical order.
   2634  * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
   2635  *  that used to be here.)
   2636  */
   2637 void
   2638 nfsrvw_sort(list, num)
   2639         gid_t *list;
   2640         int num;
   2641 {
   2642 	int i, j;
   2643 	gid_t v;
   2644 
   2645 	/* Insertion sort. */
   2646 	for (i = 1; i < num; i++) {
   2647 		v = list[i];
   2648 		/* find correct slot for value v, moving others up */
   2649 		for (j = i; --j >= 0 && v < list[j];)
   2650 			list[j + 1] = list[j];
   2651 		list[j + 1] = v;
   2652 	}
   2653 }
   2654 
   2655 /*
   2656  * copy credentials making sure that the result can be compared with memcmp().
   2657  */
   2658 void
   2659 nfsrv_setcred(incred, outcred)
   2660 	struct ucred *incred, *outcred;
   2661 {
   2662 	int i;
   2663 
   2664 	memset((caddr_t)outcred, 0, sizeof (struct ucred));
   2665 	outcred->cr_ref = 1;
   2666 	outcred->cr_uid = incred->cr_uid;
   2667 	outcred->cr_gid = incred->cr_gid;
   2668 	outcred->cr_ngroups = incred->cr_ngroups;
   2669 	for (i = 0; i < incred->cr_ngroups; i++)
   2670 		outcred->cr_groups[i] = incred->cr_groups[i];
   2671 	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
   2672 }
   2673