Home | History | Annotate | Line # | Download | only in nfs
nfs_subs.c revision 1.99
      1 /*	$NetBSD: nfs_subs.c,v 1.99 2001/11/10 10:59:10 lukem 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 1.99 2001/11/10 10:59:10 lukem 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 
    147 /* NFS client/server stats. */
    148 struct nfsstats nfsstats;
    149 
    150 /*
    151  * Mapping of old NFS Version 2 RPC numbers to generic numbers.
    152  */
    153 const int nfsv3_procid[NFS_NPROCS] = {
    154 	NFSPROC_NULL,
    155 	NFSPROC_GETATTR,
    156 	NFSPROC_SETATTR,
    157 	NFSPROC_NOOP,
    158 	NFSPROC_LOOKUP,
    159 	NFSPROC_READLINK,
    160 	NFSPROC_READ,
    161 	NFSPROC_NOOP,
    162 	NFSPROC_WRITE,
    163 	NFSPROC_CREATE,
    164 	NFSPROC_REMOVE,
    165 	NFSPROC_RENAME,
    166 	NFSPROC_LINK,
    167 	NFSPROC_SYMLINK,
    168 	NFSPROC_MKDIR,
    169 	NFSPROC_RMDIR,
    170 	NFSPROC_READDIR,
    171 	NFSPROC_FSSTAT,
    172 	NFSPROC_NOOP,
    173 	NFSPROC_NOOP,
    174 	NFSPROC_NOOP,
    175 	NFSPROC_NOOP,
    176 	NFSPROC_NOOP,
    177 	NFSPROC_NOOP,
    178 	NFSPROC_NOOP,
    179 	NFSPROC_NOOP
    180 };
    181 
    182 /*
    183  * and the reverse mapping from generic to Version 2 procedure numbers
    184  */
    185 const int nfsv2_procid[NFS_NPROCS] = {
    186 	NFSV2PROC_NULL,
    187 	NFSV2PROC_GETATTR,
    188 	NFSV2PROC_SETATTR,
    189 	NFSV2PROC_LOOKUP,
    190 	NFSV2PROC_NOOP,
    191 	NFSV2PROC_READLINK,
    192 	NFSV2PROC_READ,
    193 	NFSV2PROC_WRITE,
    194 	NFSV2PROC_CREATE,
    195 	NFSV2PROC_MKDIR,
    196 	NFSV2PROC_SYMLINK,
    197 	NFSV2PROC_CREATE,
    198 	NFSV2PROC_REMOVE,
    199 	NFSV2PROC_RMDIR,
    200 	NFSV2PROC_RENAME,
    201 	NFSV2PROC_LINK,
    202 	NFSV2PROC_READDIR,
    203 	NFSV2PROC_NOOP,
    204 	NFSV2PROC_STATFS,
    205 	NFSV2PROC_NOOP,
    206 	NFSV2PROC_NOOP,
    207 	NFSV2PROC_NOOP,
    208 	NFSV2PROC_NOOP,
    209 	NFSV2PROC_NOOP,
    210 	NFSV2PROC_NOOP,
    211 	NFSV2PROC_NOOP,
    212 };
    213 
    214 /*
    215  * Maps errno values to nfs error numbers.
    216  * Use NFSERR_IO as the catch all for ones not specifically defined in
    217  * RFC 1094.
    218  */
    219 static const u_char nfsrv_v2errmap[ELAST] = {
    220   NFSERR_PERM,	NFSERR_NOENT,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    221   NFSERR_NXIO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    222   NFSERR_IO,	NFSERR_IO,	NFSERR_ACCES,	NFSERR_IO,	NFSERR_IO,
    223   NFSERR_IO,	NFSERR_EXIST,	NFSERR_IO,	NFSERR_NODEV,	NFSERR_NOTDIR,
    224   NFSERR_ISDIR,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    225   NFSERR_IO,	NFSERR_FBIG,	NFSERR_NOSPC,	NFSERR_IO,	NFSERR_ROFS,
    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_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_NAMETOL,	NFSERR_IO,	NFSERR_IO,
    233   NFSERR_NOTEMPTY, NFSERR_IO,	NFSERR_IO,	NFSERR_DQUOT,	NFSERR_STALE,
    234   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    235   NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,	NFSERR_IO,
    236   NFSERR_IO,	NFSERR_IO,
    237 };
    238 
    239 /*
    240  * Maps errno values to nfs error numbers.
    241  * Although it is not obvious whether or not NFS clients really care if
    242  * a returned error value is in the specified list for the procedure, the
    243  * safest thing to do is filter them appropriately. For Version 2, the
    244  * X/Open XNFS document is the only specification that defines error values
    245  * for each RPC (The RFC simply lists all possible error values for all RPCs),
    246  * so I have decided to not do this for Version 2.
    247  * The first entry is the default error return and the rest are the valid
    248  * errors for that RPC in increasing numeric order.
    249  */
    250 static const short nfsv3err_null[] = {
    251 	0,
    252 	0,
    253 };
    254 
    255 static const short nfsv3err_getattr[] = {
    256 	NFSERR_IO,
    257 	NFSERR_IO,
    258 	NFSERR_STALE,
    259 	NFSERR_BADHANDLE,
    260 	NFSERR_SERVERFAULT,
    261 	0,
    262 };
    263 
    264 static const short nfsv3err_setattr[] = {
    265 	NFSERR_IO,
    266 	NFSERR_PERM,
    267 	NFSERR_IO,
    268 	NFSERR_ACCES,
    269 	NFSERR_INVAL,
    270 	NFSERR_NOSPC,
    271 	NFSERR_ROFS,
    272 	NFSERR_DQUOT,
    273 	NFSERR_STALE,
    274 	NFSERR_BADHANDLE,
    275 	NFSERR_NOT_SYNC,
    276 	NFSERR_SERVERFAULT,
    277 	0,
    278 };
    279 
    280 static const short nfsv3err_lookup[] = {
    281 	NFSERR_IO,
    282 	NFSERR_NOENT,
    283 	NFSERR_IO,
    284 	NFSERR_ACCES,
    285 	NFSERR_NOTDIR,
    286 	NFSERR_NAMETOL,
    287 	NFSERR_STALE,
    288 	NFSERR_BADHANDLE,
    289 	NFSERR_SERVERFAULT,
    290 	0,
    291 };
    292 
    293 static const short nfsv3err_access[] = {
    294 	NFSERR_IO,
    295 	NFSERR_IO,
    296 	NFSERR_STALE,
    297 	NFSERR_BADHANDLE,
    298 	NFSERR_SERVERFAULT,
    299 	0,
    300 };
    301 
    302 static const short nfsv3err_readlink[] = {
    303 	NFSERR_IO,
    304 	NFSERR_IO,
    305 	NFSERR_ACCES,
    306 	NFSERR_INVAL,
    307 	NFSERR_STALE,
    308 	NFSERR_BADHANDLE,
    309 	NFSERR_NOTSUPP,
    310 	NFSERR_SERVERFAULT,
    311 	0,
    312 };
    313 
    314 static const short nfsv3err_read[] = {
    315 	NFSERR_IO,
    316 	NFSERR_IO,
    317 	NFSERR_NXIO,
    318 	NFSERR_ACCES,
    319 	NFSERR_INVAL,
    320 	NFSERR_STALE,
    321 	NFSERR_BADHANDLE,
    322 	NFSERR_SERVERFAULT,
    323 	NFSERR_JUKEBOX,
    324 	0,
    325 };
    326 
    327 static const short nfsv3err_write[] = {
    328 	NFSERR_IO,
    329 	NFSERR_IO,
    330 	NFSERR_ACCES,
    331 	NFSERR_INVAL,
    332 	NFSERR_FBIG,
    333 	NFSERR_NOSPC,
    334 	NFSERR_ROFS,
    335 	NFSERR_DQUOT,
    336 	NFSERR_STALE,
    337 	NFSERR_BADHANDLE,
    338 	NFSERR_SERVERFAULT,
    339 	NFSERR_JUKEBOX,
    340 	0,
    341 };
    342 
    343 static const short nfsv3err_create[] = {
    344 	NFSERR_IO,
    345 	NFSERR_IO,
    346 	NFSERR_ACCES,
    347 	NFSERR_EXIST,
    348 	NFSERR_NOTDIR,
    349 	NFSERR_NOSPC,
    350 	NFSERR_ROFS,
    351 	NFSERR_NAMETOL,
    352 	NFSERR_DQUOT,
    353 	NFSERR_STALE,
    354 	NFSERR_BADHANDLE,
    355 	NFSERR_NOTSUPP,
    356 	NFSERR_SERVERFAULT,
    357 	0,
    358 };
    359 
    360 static const short nfsv3err_mkdir[] = {
    361 	NFSERR_IO,
    362 	NFSERR_IO,
    363 	NFSERR_ACCES,
    364 	NFSERR_EXIST,
    365 	NFSERR_NOTDIR,
    366 	NFSERR_NOSPC,
    367 	NFSERR_ROFS,
    368 	NFSERR_NAMETOL,
    369 	NFSERR_DQUOT,
    370 	NFSERR_STALE,
    371 	NFSERR_BADHANDLE,
    372 	NFSERR_NOTSUPP,
    373 	NFSERR_SERVERFAULT,
    374 	0,
    375 };
    376 
    377 static const short nfsv3err_symlink[] = {
    378 	NFSERR_IO,
    379 	NFSERR_IO,
    380 	NFSERR_ACCES,
    381 	NFSERR_EXIST,
    382 	NFSERR_NOTDIR,
    383 	NFSERR_NOSPC,
    384 	NFSERR_ROFS,
    385 	NFSERR_NAMETOL,
    386 	NFSERR_DQUOT,
    387 	NFSERR_STALE,
    388 	NFSERR_BADHANDLE,
    389 	NFSERR_NOTSUPP,
    390 	NFSERR_SERVERFAULT,
    391 	0,
    392 };
    393 
    394 static const short nfsv3err_mknod[] = {
    395 	NFSERR_IO,
    396 	NFSERR_IO,
    397 	NFSERR_ACCES,
    398 	NFSERR_EXIST,
    399 	NFSERR_NOTDIR,
    400 	NFSERR_NOSPC,
    401 	NFSERR_ROFS,
    402 	NFSERR_NAMETOL,
    403 	NFSERR_DQUOT,
    404 	NFSERR_STALE,
    405 	NFSERR_BADHANDLE,
    406 	NFSERR_NOTSUPP,
    407 	NFSERR_SERVERFAULT,
    408 	NFSERR_BADTYPE,
    409 	0,
    410 };
    411 
    412 static const short nfsv3err_remove[] = {
    413 	NFSERR_IO,
    414 	NFSERR_NOENT,
    415 	NFSERR_IO,
    416 	NFSERR_ACCES,
    417 	NFSERR_NOTDIR,
    418 	NFSERR_ROFS,
    419 	NFSERR_NAMETOL,
    420 	NFSERR_STALE,
    421 	NFSERR_BADHANDLE,
    422 	NFSERR_SERVERFAULT,
    423 	0,
    424 };
    425 
    426 static const short nfsv3err_rmdir[] = {
    427 	NFSERR_IO,
    428 	NFSERR_NOENT,
    429 	NFSERR_IO,
    430 	NFSERR_ACCES,
    431 	NFSERR_EXIST,
    432 	NFSERR_NOTDIR,
    433 	NFSERR_INVAL,
    434 	NFSERR_ROFS,
    435 	NFSERR_NAMETOL,
    436 	NFSERR_NOTEMPTY,
    437 	NFSERR_STALE,
    438 	NFSERR_BADHANDLE,
    439 	NFSERR_NOTSUPP,
    440 	NFSERR_SERVERFAULT,
    441 	0,
    442 };
    443 
    444 static const short nfsv3err_rename[] = {
    445 	NFSERR_IO,
    446 	NFSERR_NOENT,
    447 	NFSERR_IO,
    448 	NFSERR_ACCES,
    449 	NFSERR_EXIST,
    450 	NFSERR_XDEV,
    451 	NFSERR_NOTDIR,
    452 	NFSERR_ISDIR,
    453 	NFSERR_INVAL,
    454 	NFSERR_NOSPC,
    455 	NFSERR_ROFS,
    456 	NFSERR_MLINK,
    457 	NFSERR_NAMETOL,
    458 	NFSERR_NOTEMPTY,
    459 	NFSERR_DQUOT,
    460 	NFSERR_STALE,
    461 	NFSERR_BADHANDLE,
    462 	NFSERR_NOTSUPP,
    463 	NFSERR_SERVERFAULT,
    464 	0,
    465 };
    466 
    467 static const short nfsv3err_link[] = {
    468 	NFSERR_IO,
    469 	NFSERR_IO,
    470 	NFSERR_ACCES,
    471 	NFSERR_EXIST,
    472 	NFSERR_XDEV,
    473 	NFSERR_NOTDIR,
    474 	NFSERR_INVAL,
    475 	NFSERR_NOSPC,
    476 	NFSERR_ROFS,
    477 	NFSERR_MLINK,
    478 	NFSERR_NAMETOL,
    479 	NFSERR_DQUOT,
    480 	NFSERR_STALE,
    481 	NFSERR_BADHANDLE,
    482 	NFSERR_NOTSUPP,
    483 	NFSERR_SERVERFAULT,
    484 	0,
    485 };
    486 
    487 static const short nfsv3err_readdir[] = {
    488 	NFSERR_IO,
    489 	NFSERR_IO,
    490 	NFSERR_ACCES,
    491 	NFSERR_NOTDIR,
    492 	NFSERR_STALE,
    493 	NFSERR_BADHANDLE,
    494 	NFSERR_BAD_COOKIE,
    495 	NFSERR_TOOSMALL,
    496 	NFSERR_SERVERFAULT,
    497 	0,
    498 };
    499 
    500 static const short nfsv3err_readdirplus[] = {
    501 	NFSERR_IO,
    502 	NFSERR_IO,
    503 	NFSERR_ACCES,
    504 	NFSERR_NOTDIR,
    505 	NFSERR_STALE,
    506 	NFSERR_BADHANDLE,
    507 	NFSERR_BAD_COOKIE,
    508 	NFSERR_NOTSUPP,
    509 	NFSERR_TOOSMALL,
    510 	NFSERR_SERVERFAULT,
    511 	0,
    512 };
    513 
    514 static const short nfsv3err_fsstat[] = {
    515 	NFSERR_IO,
    516 	NFSERR_IO,
    517 	NFSERR_STALE,
    518 	NFSERR_BADHANDLE,
    519 	NFSERR_SERVERFAULT,
    520 	0,
    521 };
    522 
    523 static const short nfsv3err_fsinfo[] = {
    524 	NFSERR_STALE,
    525 	NFSERR_STALE,
    526 	NFSERR_BADHANDLE,
    527 	NFSERR_SERVERFAULT,
    528 	0,
    529 };
    530 
    531 static const short nfsv3err_pathconf[] = {
    532 	NFSERR_STALE,
    533 	NFSERR_STALE,
    534 	NFSERR_BADHANDLE,
    535 	NFSERR_SERVERFAULT,
    536 	0,
    537 };
    538 
    539 static const short nfsv3err_commit[] = {
    540 	NFSERR_IO,
    541 	NFSERR_IO,
    542 	NFSERR_STALE,
    543 	NFSERR_BADHANDLE,
    544 	NFSERR_SERVERFAULT,
    545 	0,
    546 };
    547 
    548 static const short * const nfsrv_v3errmap[] = {
    549 	nfsv3err_null,
    550 	nfsv3err_getattr,
    551 	nfsv3err_setattr,
    552 	nfsv3err_lookup,
    553 	nfsv3err_access,
    554 	nfsv3err_readlink,
    555 	nfsv3err_read,
    556 	nfsv3err_write,
    557 	nfsv3err_create,
    558 	nfsv3err_mkdir,
    559 	nfsv3err_symlink,
    560 	nfsv3err_mknod,
    561 	nfsv3err_remove,
    562 	nfsv3err_rmdir,
    563 	nfsv3err_rename,
    564 	nfsv3err_link,
    565 	nfsv3err_readdir,
    566 	nfsv3err_readdirplus,
    567 	nfsv3err_fsstat,
    568 	nfsv3err_fsinfo,
    569 	nfsv3err_pathconf,
    570 	nfsv3err_commit,
    571 };
    572 
    573 extern struct nfsrtt nfsrtt;
    574 extern time_t nqnfsstarttime;
    575 extern int nqsrv_clockskew;
    576 extern int nqsrv_writeslack;
    577 extern int nqsrv_maxlease;
    578 extern const int nqnfs_piggy[NFS_NPROCS];
    579 extern struct nfsnodehashhead *nfsnodehashtbl;
    580 extern u_long nfsnodehash;
    581 
    582 LIST_HEAD(nfsnodehashhead, nfsnode);
    583 u_long nfsdirhashmask;
    584 
    585 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *));
    586 
    587 /*
    588  * Create the header for an rpc request packet
    589  * The hsiz is the size of the rest of the nfs request header.
    590  * (just used to decide if a cluster is a good idea)
    591  */
    592 struct mbuf *
    593 nfsm_reqh(vp, procid, hsiz, bposp)
    594 	struct vnode *vp;
    595 	u_long procid;
    596 	int hsiz;
    597 	caddr_t *bposp;
    598 {
    599 	struct mbuf *mb;
    600 	caddr_t bpos;
    601 	struct nfsmount *nmp;
    602 #ifndef NFS_V2_ONLY
    603 	u_int32_t *tl;
    604 	struct mbuf *mb2;
    605 	int nqflag;
    606 #endif
    607 
    608 	MGET(mb, M_WAIT, MT_DATA);
    609 	if (hsiz >= MINCLSIZE)
    610 		MCLGET(mb, M_WAIT);
    611 	mb->m_len = 0;
    612 	bpos = mtod(mb, caddr_t);
    613 
    614 	/*
    615 	 * For NQNFS, add lease request.
    616 	 */
    617 	if (vp) {
    618 		nmp = VFSTONFS(vp->v_mount);
    619 #ifndef NFS_V2_ONLY
    620 		if (nmp->nm_flag & NFSMNT_NQNFS) {
    621 			nqflag = NQNFS_NEEDLEASE(vp, procid);
    622 			if (nqflag) {
    623 				nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED);
    624 				*tl++ = txdr_unsigned(nqflag);
    625 				*tl = txdr_unsigned(nmp->nm_leaseterm);
    626 			} else {
    627 				nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
    628 				*tl = 0;
    629 			}
    630 		}
    631 #endif
    632 	}
    633 	/* Finally, return values */
    634 	*bposp = bpos;
    635 	return (mb);
    636 }
    637 
    638 /*
    639  * Build the RPC header and fill in the authorization info.
    640  * The authorization string argument is only used when the credentials
    641  * come from outside of the kernel.
    642  * Returns the head of the mbuf list.
    643  */
    644 struct mbuf *
    645 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len,
    646 	verf_str, mrest, mrest_len, mbp, xidp)
    647 	struct ucred *cr;
    648 	int nmflag;
    649 	int procid;
    650 	int auth_type;
    651 	int auth_len;
    652 	char *auth_str;
    653 	int verf_len;
    654 	char *verf_str;
    655 	struct mbuf *mrest;
    656 	int mrest_len;
    657 	struct mbuf **mbp;
    658 	u_int32_t *xidp;
    659 {
    660 	struct mbuf *mb;
    661 	u_int32_t *tl;
    662 	caddr_t bpos;
    663 	int i;
    664 	struct mbuf *mreq, *mb2;
    665 	int siz, grpsiz, authsiz;
    666 	struct timeval tv;
    667 	static u_int32_t base;
    668 
    669 	authsiz = nfsm_rndup(auth_len);
    670 	MGETHDR(mb, M_WAIT, MT_DATA);
    671 	if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) {
    672 		MCLGET(mb, M_WAIT);
    673 	} else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) {
    674 		MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED);
    675 	} else {
    676 		MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
    677 	}
    678 	mb->m_len = 0;
    679 	mreq = mb;
    680 	bpos = mtod(mb, caddr_t);
    681 
    682 	/*
    683 	 * First the RPC header.
    684 	 */
    685 	nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
    686 
    687 	/*
    688 	 * derive initial xid from system time
    689 	 * XXX time is invalid if root not yet mounted
    690 	 */
    691 	if (!base && (rootvp)) {
    692 		microtime(&tv);
    693 		base = tv.tv_sec << 12;
    694 		nfs_xid = base;
    695 	}
    696 	/*
    697 	 * Skip zero xid if it should ever happen.
    698 	 */
    699 	if (++nfs_xid == 0)
    700 		nfs_xid++;
    701 
    702 	*tl++ = *xidp = txdr_unsigned(nfs_xid);
    703 	*tl++ = rpc_call;
    704 	*tl++ = rpc_vers;
    705 	if (nmflag & NFSMNT_NQNFS) {
    706 		*tl++ = txdr_unsigned(NQNFS_PROG);
    707 		*tl++ = txdr_unsigned(NQNFS_VER3);
    708 	} else {
    709 		*tl++ = txdr_unsigned(NFS_PROG);
    710 		if (nmflag & NFSMNT_NFSV3)
    711 			*tl++ = txdr_unsigned(NFS_VER3);
    712 		else
    713 			*tl++ = txdr_unsigned(NFS_VER2);
    714 	}
    715 	if (nmflag & NFSMNT_NFSV3)
    716 		*tl++ = txdr_unsigned(procid);
    717 	else
    718 		*tl++ = txdr_unsigned(nfsv2_procid[procid]);
    719 
    720 	/*
    721 	 * And then the authorization cred.
    722 	 */
    723 	*tl++ = txdr_unsigned(auth_type);
    724 	*tl = txdr_unsigned(authsiz);
    725 	switch (auth_type) {
    726 	case RPCAUTH_UNIX:
    727 		nfsm_build(tl, u_int32_t *, auth_len);
    728 		*tl++ = 0;		/* stamp ?? */
    729 		*tl++ = 0;		/* NULL hostname */
    730 		*tl++ = txdr_unsigned(cr->cr_uid);
    731 		*tl++ = txdr_unsigned(cr->cr_gid);
    732 		grpsiz = (auth_len >> 2) - 5;
    733 		*tl++ = txdr_unsigned(grpsiz);
    734 		for (i = 0; i < grpsiz; i++)
    735 			*tl++ = txdr_unsigned(cr->cr_groups[i]);
    736 		break;
    737 	case RPCAUTH_KERB4:
    738 		siz = auth_len;
    739 		while (siz > 0) {
    740 			if (M_TRAILINGSPACE(mb) == 0) {
    741 				MGET(mb2, M_WAIT, MT_DATA);
    742 				if (siz >= MINCLSIZE)
    743 					MCLGET(mb2, M_WAIT);
    744 				mb->m_next = mb2;
    745 				mb = mb2;
    746 				mb->m_len = 0;
    747 				bpos = mtod(mb, caddr_t);
    748 			}
    749 			i = min(siz, M_TRAILINGSPACE(mb));
    750 			memcpy(bpos, auth_str, i);
    751 			mb->m_len += i;
    752 			auth_str += i;
    753 			bpos += i;
    754 			siz -= i;
    755 		}
    756 		if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
    757 			for (i = 0; i < siz; i++)
    758 				*bpos++ = '\0';
    759 			mb->m_len += siz;
    760 		}
    761 		break;
    762 	};
    763 
    764 	/*
    765 	 * And the verifier...
    766 	 */
    767 	nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
    768 	if (verf_str) {
    769 		*tl++ = txdr_unsigned(RPCAUTH_KERB4);
    770 		*tl = txdr_unsigned(verf_len);
    771 		siz = verf_len;
    772 		while (siz > 0) {
    773 			if (M_TRAILINGSPACE(mb) == 0) {
    774 				MGET(mb2, M_WAIT, MT_DATA);
    775 				if (siz >= MINCLSIZE)
    776 					MCLGET(mb2, M_WAIT);
    777 				mb->m_next = mb2;
    778 				mb = mb2;
    779 				mb->m_len = 0;
    780 				bpos = mtod(mb, caddr_t);
    781 			}
    782 			i = min(siz, M_TRAILINGSPACE(mb));
    783 			memcpy(bpos, verf_str, i);
    784 			mb->m_len += i;
    785 			verf_str += i;
    786 			bpos += i;
    787 			siz -= i;
    788 		}
    789 		if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
    790 			for (i = 0; i < siz; i++)
    791 				*bpos++ = '\0';
    792 			mb->m_len += siz;
    793 		}
    794 	} else {
    795 		*tl++ = txdr_unsigned(RPCAUTH_NULL);
    796 		*tl = 0;
    797 	}
    798 	mb->m_next = mrest;
    799 	mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
    800 	mreq->m_pkthdr.rcvif = (struct ifnet *)0;
    801 	*mbp = mb;
    802 	return (mreq);
    803 }
    804 
    805 /*
    806  * copies mbuf chain to the uio scatter/gather list
    807  */
    808 int
    809 nfsm_mbuftouio(mrep, uiop, siz, dpos)
    810 	struct mbuf **mrep;
    811 	struct uio *uiop;
    812 	int siz;
    813 	caddr_t *dpos;
    814 {
    815 	char *mbufcp, *uiocp;
    816 	int xfer, left, len;
    817 	struct mbuf *mp;
    818 	long uiosiz, rem;
    819 	int error = 0;
    820 
    821 	mp = *mrep;
    822 	mbufcp = *dpos;
    823 	len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
    824 	rem = nfsm_rndup(siz)-siz;
    825 	while (siz > 0) {
    826 		if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
    827 			return (EFBIG);
    828 		left = uiop->uio_iov->iov_len;
    829 		uiocp = uiop->uio_iov->iov_base;
    830 		if (left > siz)
    831 			left = siz;
    832 		uiosiz = left;
    833 		while (left > 0) {
    834 			while (len == 0) {
    835 				mp = mp->m_next;
    836 				if (mp == NULL)
    837 					return (EBADRPC);
    838 				mbufcp = mtod(mp, caddr_t);
    839 				len = mp->m_len;
    840 			}
    841 			xfer = (left > len) ? len : left;
    842 #ifdef notdef
    843 			/* Not Yet.. */
    844 			if (uiop->uio_iov->iov_op != NULL)
    845 				(*(uiop->uio_iov->iov_op))
    846 				(mbufcp, uiocp, xfer);
    847 			else
    848 #endif
    849 			if (uiop->uio_segflg == UIO_SYSSPACE)
    850 				memcpy(uiocp, mbufcp, xfer);
    851 			else
    852 				copyout(mbufcp, uiocp, xfer);
    853 			left -= xfer;
    854 			len -= xfer;
    855 			mbufcp += xfer;
    856 			uiocp += xfer;
    857 			uiop->uio_offset += xfer;
    858 			uiop->uio_resid -= xfer;
    859 		}
    860 		if (uiop->uio_iov->iov_len <= siz) {
    861 			uiop->uio_iovcnt--;
    862 			uiop->uio_iov++;
    863 		} else {
    864 			uiop->uio_iov->iov_base =
    865 			    (caddr_t)uiop->uio_iov->iov_base + uiosiz;
    866 			uiop->uio_iov->iov_len -= uiosiz;
    867 		}
    868 		siz -= uiosiz;
    869 	}
    870 	*dpos = mbufcp;
    871 	*mrep = mp;
    872 	if (rem > 0) {
    873 		if (len < rem)
    874 			error = nfs_adv(mrep, dpos, rem, len);
    875 		else
    876 			*dpos += rem;
    877 	}
    878 	return (error);
    879 }
    880 
    881 /*
    882  * copies a uio scatter/gather list to an mbuf chain.
    883  * NOTE: can ony handle iovcnt == 1
    884  */
    885 int
    886 nfsm_uiotombuf(uiop, mq, siz, bpos)
    887 	struct uio *uiop;
    888 	struct mbuf **mq;
    889 	int siz;
    890 	caddr_t *bpos;
    891 {
    892 	char *uiocp;
    893 	struct mbuf *mp, *mp2;
    894 	int xfer, left, mlen;
    895 	int uiosiz, clflg, rem;
    896 	char *cp;
    897 
    898 #ifdef DIAGNOSTIC
    899 	if (uiop->uio_iovcnt != 1)
    900 		panic("nfsm_uiotombuf: iovcnt != 1");
    901 #endif
    902 
    903 	if (siz > MLEN)		/* or should it >= MCLBYTES ?? */
    904 		clflg = 1;
    905 	else
    906 		clflg = 0;
    907 	rem = nfsm_rndup(siz)-siz;
    908 	mp = mp2 = *mq;
    909 	while (siz > 0) {
    910 		left = uiop->uio_iov->iov_len;
    911 		uiocp = uiop->uio_iov->iov_base;
    912 		if (left > siz)
    913 			left = siz;
    914 		uiosiz = left;
    915 		while (left > 0) {
    916 			mlen = M_TRAILINGSPACE(mp);
    917 			if (mlen == 0) {
    918 				MGET(mp, M_WAIT, MT_DATA);
    919 				if (clflg)
    920 					MCLGET(mp, M_WAIT);
    921 				mp->m_len = 0;
    922 				mp2->m_next = mp;
    923 				mp2 = mp;
    924 				mlen = M_TRAILINGSPACE(mp);
    925 			}
    926 			xfer = (left > mlen) ? mlen : left;
    927 #ifdef notdef
    928 			/* Not Yet.. */
    929 			if (uiop->uio_iov->iov_op != NULL)
    930 				(*(uiop->uio_iov->iov_op))
    931 				(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
    932 			else
    933 #endif
    934 			if (uiop->uio_segflg == UIO_SYSSPACE)
    935 				memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer);
    936 			else
    937 				copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
    938 			mp->m_len += xfer;
    939 			left -= xfer;
    940 			uiocp += xfer;
    941 			uiop->uio_offset += xfer;
    942 			uiop->uio_resid -= xfer;
    943 		}
    944 		uiop->uio_iov->iov_base = (caddr_t)uiop->uio_iov->iov_base +
    945 		    uiosiz;
    946 		uiop->uio_iov->iov_len -= uiosiz;
    947 		siz -= uiosiz;
    948 	}
    949 	if (rem > 0) {
    950 		if (rem > M_TRAILINGSPACE(mp)) {
    951 			MGET(mp, M_WAIT, MT_DATA);
    952 			mp->m_len = 0;
    953 			mp2->m_next = mp;
    954 		}
    955 		cp = mtod(mp, caddr_t)+mp->m_len;
    956 		for (left = 0; left < rem; left++)
    957 			*cp++ = '\0';
    958 		mp->m_len += rem;
    959 		*bpos = cp;
    960 	} else
    961 		*bpos = mtod(mp, caddr_t)+mp->m_len;
    962 	*mq = mp;
    963 	return (0);
    964 }
    965 
    966 /*
    967  * Get at least "siz" bytes of correctly aligned data.
    968  * When called the mbuf pointers are not necessarily correct,
    969  * dsosp points to what ought to be in m_data and left contains
    970  * what ought to be in m_len.
    971  * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
    972  * cases. (The macros use the vars. dpos and dpos2)
    973  */
    974 int
    975 nfsm_disct(mdp, dposp, siz, left, cp2)
    976 	struct mbuf **mdp;
    977 	caddr_t *dposp;
    978 	int siz;
    979 	int left;
    980 	caddr_t *cp2;
    981 {
    982 	struct mbuf *m1, *m2;
    983 	struct mbuf *havebuf = NULL;
    984 	caddr_t src = *dposp;
    985 	caddr_t dst;
    986 	int len;
    987 
    988 #ifdef DEBUG
    989 	if (left < 0)
    990 		panic("nfsm_disct: left < 0");
    991 #endif
    992 	m1 = *mdp;
    993 	/*
    994 	 * Skip through the mbuf chain looking for an mbuf with
    995 	 * some data. If the first mbuf found has enough data
    996 	 * and it is correctly aligned return it.
    997 	 */
    998 	while (left == 0) {
    999 		havebuf = m1;
   1000 		*mdp = m1 = m1->m_next;
   1001 		if (m1 == NULL)
   1002 			return (EBADRPC);
   1003 		src = mtod(m1, caddr_t);
   1004 		left = m1->m_len;
   1005 		/*
   1006 		 * If we start a new mbuf and it is big enough
   1007 		 * and correctly aligned just return it, don't
   1008 		 * do any pull up.
   1009 		 */
   1010 		if (left >= siz && nfsm_aligned(src)) {
   1011 			*cp2 = src;
   1012 			*dposp = src + siz;
   1013 			return (0);
   1014 		}
   1015 	}
   1016 	if (m1->m_flags & M_EXT) {
   1017 		if (havebuf) {
   1018 			/* If the first mbuf with data has external data
   1019 			 * and there is a previous empty mbuf use it
   1020 			 * to move the data into.
   1021 			 */
   1022 			m2 = m1;
   1023 			*mdp = m1 = havebuf;
   1024 			if (m1->m_flags & M_EXT) {
   1025 				MEXTREMOVE(m1);
   1026 			}
   1027 		} else {
   1028 			/*
   1029 			 * If the first mbuf has a external data
   1030 			 * and there is no previous empty mbuf
   1031 			 * allocate a new mbuf and move the external
   1032 			 * data to the new mbuf. Also make the first
   1033 			 * mbuf look empty.
   1034 			 */
   1035 			m2 = m_get(M_WAIT, MT_DATA);
   1036 			m2->m_ext = m1->m_ext;
   1037 			m2->m_data = src;
   1038 			m2->m_len = left;
   1039 			MCLADDREFERENCE(m1, m2);
   1040 			MEXTREMOVE(m1);
   1041 			m2->m_next = m1->m_next;
   1042 			m1->m_next = m2;
   1043 		}
   1044 		m1->m_len = 0;
   1045 		dst = m1->m_dat;
   1046 	} else {
   1047 		/*
   1048 		 * If the first mbuf has no external data
   1049 		 * move the data to the front of the mbuf.
   1050 		 */
   1051 		if ((dst = m1->m_dat) != src)
   1052 			memmove(dst, src, left);
   1053 		dst += left;
   1054 		m1->m_len = left;
   1055 		m2 = m1->m_next;
   1056 	}
   1057 	m1->m_flags &= ~M_PKTHDR;
   1058 	*cp2 = m1->m_data = m1->m_dat;   /* data is at beginning of buffer */
   1059 	*dposp = mtod(m1, caddr_t) + siz;
   1060 	/*
   1061 	 * Loop through mbufs pulling data up into first mbuf until
   1062 	 * the first mbuf is full or there is no more data to
   1063 	 * pullup.
   1064 	 */
   1065 	while ((len = (MLEN - m1->m_len)) != 0 && m2) {
   1066 		if ((len = min(len, m2->m_len)) != 0)
   1067 			memcpy(dst, m2->m_data, len);
   1068 		m1->m_len += len;
   1069 		dst += len;
   1070 		m2->m_data += len;
   1071 		m2->m_len -= len;
   1072 		m2 = m2->m_next;
   1073 	}
   1074 	if (m1->m_len < siz)
   1075 		return (EBADRPC);
   1076 	return (0);
   1077 }
   1078 
   1079 /*
   1080  * Advance the position in the mbuf chain.
   1081  */
   1082 int
   1083 nfs_adv(mdp, dposp, offs, left)
   1084 	struct mbuf **mdp;
   1085 	caddr_t *dposp;
   1086 	int offs;
   1087 	int left;
   1088 {
   1089 	struct mbuf *m;
   1090 	int s;
   1091 
   1092 	m = *mdp;
   1093 	s = left;
   1094 	while (s < offs) {
   1095 		offs -= s;
   1096 		m = m->m_next;
   1097 		if (m == NULL)
   1098 			return (EBADRPC);
   1099 		s = m->m_len;
   1100 	}
   1101 	*mdp = m;
   1102 	*dposp = mtod(m, caddr_t)+offs;
   1103 	return (0);
   1104 }
   1105 
   1106 /*
   1107  * Copy a string into mbufs for the hard cases...
   1108  */
   1109 int
   1110 nfsm_strtmbuf(mb, bpos, cp, siz)
   1111 	struct mbuf **mb;
   1112 	char **bpos;
   1113 	const char *cp;
   1114 	long siz;
   1115 {
   1116 	struct mbuf *m1 = NULL, *m2;
   1117 	long left, xfer, len, tlen;
   1118 	u_int32_t *tl;
   1119 	int putsize;
   1120 
   1121 	putsize = 1;
   1122 	m2 = *mb;
   1123 	left = M_TRAILINGSPACE(m2);
   1124 	if (left > 0) {
   1125 		tl = ((u_int32_t *)(*bpos));
   1126 		*tl++ = txdr_unsigned(siz);
   1127 		putsize = 0;
   1128 		left -= NFSX_UNSIGNED;
   1129 		m2->m_len += NFSX_UNSIGNED;
   1130 		if (left > 0) {
   1131 			memcpy((caddr_t) tl, cp, left);
   1132 			siz -= left;
   1133 			cp += left;
   1134 			m2->m_len += left;
   1135 			left = 0;
   1136 		}
   1137 	}
   1138 	/* Loop around adding mbufs */
   1139 	while (siz > 0) {
   1140 		MGET(m1, M_WAIT, MT_DATA);
   1141 		if (siz > MLEN)
   1142 			MCLGET(m1, M_WAIT);
   1143 		m1->m_len = NFSMSIZ(m1);
   1144 		m2->m_next = m1;
   1145 		m2 = m1;
   1146 		tl = mtod(m1, u_int32_t *);
   1147 		tlen = 0;
   1148 		if (putsize) {
   1149 			*tl++ = txdr_unsigned(siz);
   1150 			m1->m_len -= NFSX_UNSIGNED;
   1151 			tlen = NFSX_UNSIGNED;
   1152 			putsize = 0;
   1153 		}
   1154 		if (siz < m1->m_len) {
   1155 			len = nfsm_rndup(siz);
   1156 			xfer = siz;
   1157 			if (xfer < len)
   1158 				*(tl+(xfer>>2)) = 0;
   1159 		} else {
   1160 			xfer = len = m1->m_len;
   1161 		}
   1162 		memcpy((caddr_t) tl, cp, xfer);
   1163 		m1->m_len = len+tlen;
   1164 		siz -= xfer;
   1165 		cp += xfer;
   1166 	}
   1167 	*mb = m1;
   1168 	*bpos = mtod(m1, caddr_t)+m1->m_len;
   1169 	return (0);
   1170 }
   1171 
   1172 /*
   1173  * Directory caching routines. They work as follows:
   1174  * - a cache is maintained per VDIR nfsnode.
   1175  * - for each offset cookie that is exported to userspace, and can
   1176  *   thus be thrown back at us as an offset to VOP_READDIR, store
   1177  *   information in the cache.
   1178  * - cached are:
   1179  *   - cookie itself
   1180  *   - blocknumber (essentially just a search key in the buffer cache)
   1181  *   - entry number in block.
   1182  *   - offset cookie of block in which this entry is stored
   1183  *   - 32 bit cookie if NFSMNT_XLATECOOKIE is used.
   1184  * - entries are looked up in a hash table
   1185  * - also maintained is an LRU list of entries, used to determine
   1186  *   which ones to delete if the cache grows too large.
   1187  * - if 32 <-> 64 translation mode is requested for a filesystem,
   1188  *   the cache also functions as a translation table
   1189  * - in the translation case, invalidating the cache does not mean
   1190  *   flushing it, but just marking entries as invalid, except for
   1191  *   the <64bit cookie, 32bitcookie> pair which is still valid, to
   1192  *   still be able to use the cache as a translation table.
   1193  * - 32 bit cookies are uniquely created by combining the hash table
   1194  *   entry value, and one generation count per hash table entry,
   1195  *   incremented each time an entry is appended to the chain.
   1196  * - the cache is invalidated each time a direcory is modified
   1197  * - sanity checks are also done; if an entry in a block turns
   1198  *   out not to have a matching cookie, the cache is invalidated
   1199  *   and a new block starting from the wanted offset is fetched from
   1200  *   the server.
   1201  * - directory entries as read from the server are extended to contain
   1202  *   the 64bit and, optionally, the 32bit cookies, for sanity checking
   1203  *   the cache and exporting them to userspace through the cookie
   1204  *   argument to VOP_READDIR.
   1205  */
   1206 
   1207 u_long
   1208 nfs_dirhash(off)
   1209 	off_t off;
   1210 {
   1211 	int i;
   1212 	char *cp = (char *)&off;
   1213 	u_long sum = 0L;
   1214 
   1215 	for (i = 0 ; i < sizeof (off); i++)
   1216 		sum += *cp++;
   1217 
   1218 	return sum;
   1219 }
   1220 
   1221 void
   1222 nfs_initdircache(vp)
   1223 	struct vnode *vp;
   1224 {
   1225 	struct nfsnode *np = VTONFS(vp);
   1226 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1227 
   1228 	np->n_dircachesize = 0;
   1229 	np->n_dblkno = 1;
   1230 	np->n_dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, M_NFSDIROFF,
   1231 	    M_WAITOK, &nfsdirhashmask);
   1232 	TAILQ_INIT(&np->n_dirchain);
   1233 	if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
   1234 		MALLOC(np->n_dirgens, unsigned *,
   1235 		    NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF,
   1236 		    M_WAITOK);
   1237 		memset((caddr_t)np->n_dirgens, 0,
   1238 		    NFS_DIRHASHSIZ * sizeof (unsigned));
   1239 	}
   1240 }
   1241 
   1242 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0};
   1243 
   1244 struct nfsdircache *
   1245 nfs_searchdircache(vp, off, do32, hashent)
   1246 	struct vnode *vp;
   1247 	off_t off;
   1248 	int do32;
   1249 	int *hashent;
   1250 {
   1251 	struct nfsdirhashhead *ndhp;
   1252 	struct nfsdircache *ndp = NULL;
   1253 	struct nfsnode *np = VTONFS(vp);
   1254 	unsigned ent;
   1255 
   1256 	/*
   1257 	 * Zero is always a valid cookie.
   1258 	 */
   1259 	if (off == 0)
   1260 		return &dzero;
   1261 
   1262 	/*
   1263 	 * We use a 32bit cookie as search key, directly reconstruct
   1264 	 * the hashentry. Else use the hashfunction.
   1265 	 */
   1266 	if (do32) {
   1267 		ent = (u_int32_t)off >> 24;
   1268 		if (ent >= NFS_DIRHASHSIZ)
   1269 			return NULL;
   1270 		ndhp = &np->n_dircache[ent];
   1271 	} else {
   1272 		ndhp = NFSDIRHASH(np, off);
   1273 	}
   1274 
   1275 	if (hashent)
   1276 		*hashent = (int)(ndhp - np->n_dircache);
   1277 	if (do32) {
   1278 		for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) {
   1279 			if (ndp->dc_cookie32 == (u_int32_t)off) {
   1280 				/*
   1281 				 * An invalidated entry will become the
   1282 				 * start of a new block fetched from
   1283 				 * the server.
   1284 				 */
   1285 				if (ndp->dc_blkno == -1) {
   1286 					ndp->dc_blkcookie = ndp->dc_cookie;
   1287 					ndp->dc_blkno = np->n_dblkno++;
   1288 					ndp->dc_entry = 0;
   1289 				}
   1290 				break;
   1291 			}
   1292 		}
   1293 	} else {
   1294 		for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next)
   1295 			if (ndp->dc_cookie == off)
   1296 				break;
   1297 	}
   1298 	return ndp;
   1299 }
   1300 
   1301 
   1302 struct nfsdircache *
   1303 nfs_enterdircache(vp, off, blkoff, en, blkno)
   1304 	struct vnode *vp;
   1305 	off_t off, blkoff;
   1306 	daddr_t blkno;
   1307 	int en;
   1308 {
   1309 	struct nfsnode *np = VTONFS(vp);
   1310 	struct nfsdirhashhead *ndhp;
   1311 	struct nfsdircache *ndp = NULL, *first;
   1312 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1313 	int hashent, gen, overwrite;
   1314 
   1315 	if (!np->n_dircache)
   1316 		/*
   1317 		 * XXX would like to do this in nfs_nget but vtype
   1318 		 * isn't known at that time.
   1319 		 */
   1320 		nfs_initdircache(vp);
   1321 
   1322 	/*
   1323 	 * XXX refuse entries for offset 0. amd(8) erroneously sets
   1324 	 * cookie 0 for the '.' entry, making this necessary. This
   1325 	 * isn't so bad, as 0 is a special case anyway.
   1326 	 */
   1327 	if (off == 0)
   1328 		return &dzero;
   1329 
   1330 	ndp = nfs_searchdircache(vp, off, 0, &hashent);
   1331 
   1332 	if (ndp && ndp->dc_blkno != -1) {
   1333 		/*
   1334 		 * Overwriting an old entry. Check if it's the same.
   1335 		 * If so, just return. If not, remove the old entry.
   1336 		 */
   1337 		if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en)
   1338 			return ndp;
   1339 		TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
   1340 		LIST_REMOVE(ndp, dc_hash);
   1341 		FREE(ndp, M_NFSDIROFF);
   1342 		ndp = 0;
   1343 	}
   1344 
   1345 	ndhp = &np->n_dircache[hashent];
   1346 
   1347 	if (!ndp) {
   1348 		MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF,
   1349 		    M_WAITOK);
   1350 		overwrite = 0;
   1351 		if (nmp->nm_flag & NFSMNT_XLATECOOKIE) {
   1352 			/*
   1353 			 * We're allocating a new entry, so bump the
   1354 			 * generation number.
   1355 			 */
   1356 			gen = ++np->n_dirgens[hashent];
   1357 			if (gen == 0) {
   1358 				np->n_dirgens[hashent]++;
   1359 				gen++;
   1360 			}
   1361 			ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff);
   1362 		}
   1363 	} else
   1364 		overwrite = 1;
   1365 
   1366 	/*
   1367 	 * If the entry number is 0, we are at the start of a new block, so
   1368 	 * allocate a new blocknumber.
   1369 	 */
   1370 	if (en == 0)
   1371 		ndp->dc_blkno = np->n_dblkno++;
   1372 	else
   1373 		ndp->dc_blkno = blkno;
   1374 
   1375 	ndp->dc_cookie = off;
   1376 	ndp->dc_blkcookie = blkoff;
   1377 	ndp->dc_entry = en;
   1378 
   1379 	if (overwrite)
   1380 		return ndp;
   1381 
   1382 	/*
   1383 	 * If the maximum directory cookie cache size has been reached
   1384 	 * for this node, take one off the front. The idea is that
   1385 	 * directories are typically read front-to-back once, so that
   1386 	 * the oldest entries can be thrown away without much performance
   1387 	 * loss.
   1388 	 */
   1389 	if (np->n_dircachesize == NFS_MAXDIRCACHE) {
   1390 		first = np->n_dirchain.tqh_first;
   1391 		TAILQ_REMOVE(&np->n_dirchain, first, dc_chain);
   1392 		LIST_REMOVE(first, dc_hash);
   1393 		FREE(first, M_NFSDIROFF);
   1394 	} else
   1395 		np->n_dircachesize++;
   1396 
   1397 	LIST_INSERT_HEAD(ndhp, ndp, dc_hash);
   1398 	TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain);
   1399 	return ndp;
   1400 }
   1401 
   1402 void
   1403 nfs_invaldircache(vp, forcefree)
   1404 	struct vnode *vp;
   1405 	int forcefree;
   1406 {
   1407 	struct nfsnode *np = VTONFS(vp);
   1408 	struct nfsdircache *ndp = NULL;
   1409 	struct nfsmount *nmp = VFSTONFS(vp->v_mount);
   1410 
   1411 #ifdef DIAGNOSTIC
   1412 	if (vp->v_type != VDIR)
   1413 		panic("nfs: invaldircache: not dir");
   1414 #endif
   1415 
   1416 	if (!np->n_dircache)
   1417 		return;
   1418 
   1419 	if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) {
   1420 		while ((ndp = np->n_dirchain.tqh_first)) {
   1421 			TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain);
   1422 			LIST_REMOVE(ndp, dc_hash);
   1423 			FREE(ndp, M_NFSDIROFF);
   1424 		}
   1425 		np->n_dircachesize = 0;
   1426 		if (forcefree && np->n_dirgens) {
   1427 			FREE(np->n_dirgens, M_NFSDIROFF);
   1428 		}
   1429 	} else {
   1430 		for (ndp = np->n_dirchain.tqh_first; ndp;
   1431 		    ndp = ndp->dc_chain.tqe_next)
   1432 			ndp->dc_blkno = -1;
   1433 	}
   1434 
   1435 	np->n_dblkno = 1;
   1436 }
   1437 
   1438 /*
   1439  * Called once before VFS init to initialize shared and
   1440  * server-specific data structures.
   1441  */
   1442 void
   1443 nfs_init()
   1444 {
   1445 	nfsrtt.pos = 0;
   1446 	rpc_vers = txdr_unsigned(RPC_VER2);
   1447 	rpc_call = txdr_unsigned(RPC_CALL);
   1448 	rpc_reply = txdr_unsigned(RPC_REPLY);
   1449 	rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
   1450 	rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
   1451 	rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
   1452 	rpc_autherr = txdr_unsigned(RPC_AUTHERR);
   1453 	rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
   1454 	rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
   1455 	nfs_prog = txdr_unsigned(NFS_PROG);
   1456 	nqnfs_prog = txdr_unsigned(NQNFS_PROG);
   1457 	nfs_true = txdr_unsigned(TRUE);
   1458 	nfs_false = txdr_unsigned(FALSE);
   1459 	nfs_xdrneg1 = txdr_unsigned(-1);
   1460 	nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
   1461 	if (nfs_ticks < 1)
   1462 		nfs_ticks = 1;
   1463 #ifdef NFSSERVER
   1464 	nfsrv_init(0);			/* Init server data structures */
   1465 	nfsrv_initcache();		/* Init the server request cache */
   1466 #endif /* NFSSERVER */
   1467 
   1468 #if defined(NFSSERVER) || !defined(NFS_V2_ONLY)
   1469 	/*
   1470 	 * Initialize the nqnfs data structures.
   1471 	 */
   1472 	if (nqnfsstarttime == 0) {
   1473 		nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease
   1474 			+ nqsrv_clockskew + nqsrv_writeslack;
   1475 		NQLOADNOVRAM(nqnfsstarttime);
   1476 		CIRCLEQ_INIT(&nqtimerhead);
   1477 		nqfhhashtbl = hashinit(NQLCHSZ, HASH_LIST, M_NQLEASE,
   1478 		    M_WAITOK, &nqfhhash);
   1479 	}
   1480 #endif
   1481 
   1482 	/*
   1483 	 * Initialize reply list and start timer
   1484 	 */
   1485 	TAILQ_INIT(&nfs_reqq);
   1486 	nfs_timer(NULL);
   1487 }
   1488 
   1489 #ifdef NFS
   1490 /*
   1491  * Called once at VFS init to initialize client-specific data structures.
   1492  */
   1493 void
   1494 nfs_vfs_init()
   1495 {
   1496 	nfs_nhinit();			/* Init the nfsnode table */
   1497 }
   1498 
   1499 void
   1500 nfs_vfs_reinit()
   1501 {
   1502 	nfs_nhreinit();
   1503 }
   1504 
   1505 void
   1506 nfs_vfs_done()
   1507 {
   1508 	nfs_nhdone();
   1509 }
   1510 
   1511 /*
   1512  * Attribute cache routines.
   1513  * nfs_loadattrcache() - loads or updates the cache contents from attributes
   1514  *	that are on the mbuf list
   1515  * nfs_getattrcache() - returns valid attributes if found in cache, returns
   1516  *	error otherwise
   1517  */
   1518 
   1519 /*
   1520  * Load the attribute cache (that lives in the nfsnode entry) with
   1521  * the values on the mbuf list and
   1522  * Iff vap not NULL
   1523  *    copy the attributes to *vaper
   1524  */
   1525 int
   1526 nfsm_loadattrcache(vpp, mdp, dposp, vaper)
   1527 	struct vnode **vpp;
   1528 	struct mbuf **mdp;
   1529 	caddr_t *dposp;
   1530 	struct vattr *vaper;
   1531 {
   1532 	int32_t t1;
   1533 	caddr_t cp2;
   1534 	int error = 0;
   1535 	struct mbuf *md;
   1536 	int v3 = NFS_ISV3(*vpp);
   1537 
   1538 	md = *mdp;
   1539 	t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
   1540 	error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2);
   1541 	if (error)
   1542 		return (error);
   1543 	return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper);
   1544 }
   1545 
   1546 int
   1547 nfs_loadattrcache(vpp, fp, vaper)
   1548 	struct vnode **vpp;
   1549 	struct nfs_fattr *fp;
   1550 	struct vattr *vaper;
   1551 {
   1552 	struct vnode *vp = *vpp;
   1553 	struct vattr *vap;
   1554 	int v3 = NFS_ISV3(vp);
   1555 	enum vtype vtyp;
   1556 	u_short vmode;
   1557 	struct timespec mtime;
   1558 	struct vnode *nvp;
   1559 	int32_t rdev;
   1560 	struct nfsnode *np;
   1561 	extern int (**spec_nfsv2nodeop_p) __P((void *));
   1562 
   1563 	if (v3) {
   1564 		vtyp = nfsv3tov_type(fp->fa_type);
   1565 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
   1566 		rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1),
   1567 			fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2));
   1568 		fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
   1569 	} else {
   1570 		vtyp = nfsv2tov_type(fp->fa_type);
   1571 		vmode = fxdr_unsigned(u_short, fp->fa_mode);
   1572 		if (vtyp == VNON || vtyp == VREG)
   1573 			vtyp = IFTOVT(vmode);
   1574 		rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
   1575 		fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
   1576 
   1577 		/*
   1578 		 * Really ugly NFSv2 kludge.
   1579 		 */
   1580 		if (vtyp == VCHR && rdev == 0xffffffff)
   1581 			vtyp = VFIFO;
   1582 	}
   1583 
   1584 	/*
   1585 	 * If v_type == VNON it is a new node, so fill in the v_type,
   1586 	 * n_mtime fields. Check to see if it represents a special
   1587 	 * device, and if so, check for a possible alias. Once the
   1588 	 * correct vnode has been obtained, fill in the rest of the
   1589 	 * information.
   1590 	 */
   1591 	np = VTONFS(vp);
   1592 	if (vp->v_type == VNON) {
   1593 		vp->v_type = vtyp;
   1594 		if (vp->v_type == VFIFO) {
   1595 			extern int (**fifo_nfsv2nodeop_p) __P((void *));
   1596 			vp->v_op = fifo_nfsv2nodeop_p;
   1597 		}
   1598 		if (vp->v_type == VCHR || vp->v_type == VBLK) {
   1599 			vp->v_op = spec_nfsv2nodeop_p;
   1600 			nvp = checkalias(vp, (dev_t)rdev, vp->v_mount);
   1601 			if (nvp) {
   1602 				/*
   1603 				 * Discard unneeded vnode, but save its nfsnode.
   1604 				 * Since the nfsnode does not have a lock, its
   1605 				 * vnode lock has to be carried over.
   1606 				 */
   1607 				/*
   1608 				 * XXX is the old node sure to be locked here?
   1609 				 */
   1610 				KASSERT(lockstatus(&vp->v_lock) ==
   1611 				    LK_EXCLUSIVE);
   1612 				nvp->v_data = vp->v_data;
   1613 				vp->v_data = NULL;
   1614 				VOP_UNLOCK(vp, 0);
   1615 				vp->v_op = spec_vnodeop_p;
   1616 				vrele(vp);
   1617 				vgone(vp);
   1618 				lockmgr(&nvp->v_lock, LK_EXCLUSIVE,
   1619 				    &nvp->v_interlock);
   1620 				/*
   1621 				 * Reinitialize aliased node.
   1622 				 */
   1623 				np->n_vnode = nvp;
   1624 				*vpp = vp = nvp;
   1625 			}
   1626 		}
   1627 		np->n_mtime = mtime.tv_sec;
   1628 	}
   1629 	vap = np->n_vattr;
   1630 	vap->va_type = vtyp;
   1631 	vap->va_mode = vmode & ALLPERMS;
   1632 	vap->va_rdev = (dev_t)rdev;
   1633 	vap->va_mtime = mtime;
   1634 	vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
   1635 	switch (vtyp) {
   1636 	case VDIR:
   1637 		vap->va_blocksize = NFS_DIRFRAGSIZ;
   1638 		break;
   1639 	case VBLK:
   1640 		vap->va_blocksize = BLKDEV_IOSIZE;
   1641 		break;
   1642 	case VCHR:
   1643 		vap->va_blocksize = MAXBSIZE;
   1644 		break;
   1645 	default:
   1646 		vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize :
   1647 		    fxdr_unsigned(int32_t, fp->fa2_blocksize);
   1648 		break;
   1649 	}
   1650 	if (v3) {
   1651 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
   1652 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
   1653 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
   1654 		vap->va_size = fxdr_hyper(&fp->fa3_size);
   1655 		vap->va_bytes = fxdr_hyper(&fp->fa3_used);
   1656 		vap->va_fileid = fxdr_unsigned(int32_t,
   1657 		    fp->fa3_fileid.nfsuquad[1]);
   1658 		fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
   1659 		fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
   1660 		vap->va_flags = 0;
   1661 		vap->va_filerev = 0;
   1662 	} else {
   1663 		vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
   1664 		vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
   1665 		vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
   1666 		vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
   1667 		vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks)
   1668 		    * NFS_FABLKSIZE;
   1669 		vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
   1670 		fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
   1671 		vap->va_flags = 0;
   1672 		vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
   1673 		    fp->fa2_ctime.nfsv2_sec);
   1674 		vap->va_ctime.tv_nsec = 0;
   1675 		vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
   1676 		vap->va_filerev = 0;
   1677 	}
   1678 	if (vap->va_size != np->n_size) {
   1679 		if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) {
   1680 			vap->va_size = np->n_size;
   1681 		} else {
   1682 			np->n_size = vap->va_size;
   1683 			if (vap->va_type == VREG) {
   1684 				uvm_vnp_setsize(vp, np->n_size);
   1685 			}
   1686 		}
   1687 	}
   1688 	np->n_attrstamp = time.tv_sec;
   1689 	if (vaper != NULL) {
   1690 		memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap));
   1691 		if (np->n_flag & NCHG) {
   1692 			if (np->n_flag & NACC)
   1693 				vaper->va_atime = np->n_atim;
   1694 			if (np->n_flag & NUPD)
   1695 				vaper->va_mtime = np->n_mtim;
   1696 		}
   1697 	}
   1698 	return (0);
   1699 }
   1700 
   1701 /*
   1702  * Check the time stamp
   1703  * If the cache is valid, copy contents to *vap and return 0
   1704  * otherwise return an error
   1705  */
   1706 int
   1707 nfs_getattrcache(vp, vaper)
   1708 	struct vnode *vp;
   1709 	struct vattr *vaper;
   1710 {
   1711 	struct nfsnode *np = VTONFS(vp);
   1712 	struct vattr *vap;
   1713 
   1714 	if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) {
   1715 		nfsstats.attrcache_misses++;
   1716 		return (ENOENT);
   1717 	}
   1718 	nfsstats.attrcache_hits++;
   1719 	vap = np->n_vattr;
   1720 	if (vap->va_size != np->n_size) {
   1721 		if (vap->va_type == VREG) {
   1722 			if (np->n_flag & NMODIFIED) {
   1723 				if (vap->va_size < np->n_size)
   1724 					vap->va_size = np->n_size;
   1725 				else
   1726 					np->n_size = vap->va_size;
   1727 			} else
   1728 				np->n_size = vap->va_size;
   1729 			uvm_vnp_setsize(vp, np->n_size);
   1730 		} else
   1731 			np->n_size = vap->va_size;
   1732 	}
   1733 	memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr));
   1734 	if (np->n_flag & NCHG) {
   1735 		if (np->n_flag & NACC)
   1736 			vaper->va_atime = np->n_atim;
   1737 		if (np->n_flag & NUPD)
   1738 			vaper->va_mtime = np->n_mtim;
   1739 	}
   1740 	return (0);
   1741 }
   1742 
   1743 /*
   1744  * Heuristic to see if the server XDR encodes directory cookies or not.
   1745  * it is not supposed to, but a lot of servers may do this. Also, since
   1746  * most/all servers will implement V2 as well, it is expected that they
   1747  * may return just 32 bits worth of cookie information, so we need to
   1748  * find out in which 32 bits this information is available. We do this
   1749  * to avoid trouble with emulated binaries that can't handle 64 bit
   1750  * directory offsets.
   1751  */
   1752 
   1753 void
   1754 nfs_cookieheuristic(vp, flagp, p, cred)
   1755 	struct vnode *vp;
   1756 	int *flagp;
   1757 	struct proc *p;
   1758 	struct ucred *cred;
   1759 {
   1760 	struct uio auio;
   1761 	struct iovec aiov;
   1762 	caddr_t buf, cp;
   1763 	struct dirent *dp;
   1764 	off_t *cookies = NULL, *cop;
   1765 	int error, eof, nc, len;
   1766 
   1767 	MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK);
   1768 
   1769 	aiov.iov_base = buf;
   1770 	aiov.iov_len = NFS_DIRFRAGSIZ;
   1771 	auio.uio_iov = &aiov;
   1772 	auio.uio_iovcnt = 1;
   1773 	auio.uio_rw = UIO_READ;
   1774 	auio.uio_segflg = UIO_SYSSPACE;
   1775 	auio.uio_procp = p;
   1776 	auio.uio_resid = NFS_DIRFRAGSIZ;
   1777 	auio.uio_offset = 0;
   1778 
   1779 	error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc);
   1780 
   1781 	len = NFS_DIRFRAGSIZ - auio.uio_resid;
   1782 	if (error || len == 0) {
   1783 		FREE(buf, M_TEMP);
   1784 		if (cookies)
   1785 			free(cookies, M_TEMP);
   1786 		return;
   1787 	}
   1788 
   1789 	/*
   1790 	 * Find the first valid entry and look at its offset cookie.
   1791 	 */
   1792 
   1793 	cp = buf;
   1794 	for (cop = cookies; len > 0; len -= dp->d_reclen) {
   1795 		dp = (struct dirent *)cp;
   1796 		if (dp->d_fileno != 0 && len >= dp->d_reclen) {
   1797 			if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) {
   1798 				*flagp |= NFSMNT_SWAPCOOKIE;
   1799 				nfs_invaldircache(vp, 0);
   1800 				nfs_vinvalbuf(vp, 0, cred, p, 1);
   1801 			}
   1802 			break;
   1803 		}
   1804 		cop++;
   1805 		cp += dp->d_reclen;
   1806 	}
   1807 
   1808 	FREE(buf, M_TEMP);
   1809 	free(cookies, M_TEMP);
   1810 }
   1811 #endif /* NFS */
   1812 
   1813 /*
   1814  * Set up nameidata for a lookup() call and do it.
   1815  *
   1816  * If pubflag is set, this call is done for a lookup operation on the
   1817  * public filehandle. In that case we allow crossing mountpoints and
   1818  * absolute pathnames. However, the caller is expected to check that
   1819  * the lookup result is within the public fs, and deny access if
   1820  * it is not.
   1821  */
   1822 int
   1823 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag)
   1824 	struct nameidata *ndp;
   1825 	fhandle_t *fhp;
   1826 	int len;
   1827 	struct nfssvc_sock *slp;
   1828 	struct mbuf *nam;
   1829 	struct mbuf **mdp;
   1830 	caddr_t *dposp;
   1831 	struct vnode **retdirp;
   1832 	struct proc *p;
   1833 	int kerbflag, pubflag;
   1834 {
   1835 	int i, rem;
   1836 	struct mbuf *md;
   1837 	char *fromcp, *tocp, *cp;
   1838 	struct iovec aiov;
   1839 	struct uio auio;
   1840 	struct vnode *dp;
   1841 	int error, rdonly, linklen;
   1842 	struct componentname *cnp = &ndp->ni_cnd;
   1843 
   1844 	*retdirp = (struct vnode *)0;
   1845 
   1846 	if ((len + 1) > MAXPATHLEN)
   1847 		return (ENAMETOOLONG);
   1848 	cnp->cn_pnbuf = PNBUF_GET();
   1849 
   1850 	/*
   1851 	 * Copy the name from the mbuf list to ndp->ni_pnbuf
   1852 	 * and set the various ndp fields appropriately.
   1853 	 */
   1854 	fromcp = *dposp;
   1855 	tocp = cnp->cn_pnbuf;
   1856 	md = *mdp;
   1857 	rem = mtod(md, caddr_t) + md->m_len - fromcp;
   1858 	for (i = 0; i < len; i++) {
   1859 		while (rem == 0) {
   1860 			md = md->m_next;
   1861 			if (md == NULL) {
   1862 				error = EBADRPC;
   1863 				goto out;
   1864 			}
   1865 			fromcp = mtod(md, caddr_t);
   1866 			rem = md->m_len;
   1867 		}
   1868 		if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
   1869 			error = EACCES;
   1870 			goto out;
   1871 		}
   1872 		*tocp++ = *fromcp++;
   1873 		rem--;
   1874 	}
   1875 	*tocp = '\0';
   1876 	*mdp = md;
   1877 	*dposp = fromcp;
   1878 	len = nfsm_rndup(len)-len;
   1879 	if (len > 0) {
   1880 		if (rem >= len)
   1881 			*dposp += len;
   1882 		else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
   1883 			goto out;
   1884 	}
   1885 
   1886 	/*
   1887 	 * Extract and set starting directory.
   1888 	 */
   1889 	error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp,
   1890 	    nam, &rdonly, kerbflag, pubflag);
   1891 	if (error)
   1892 		goto out;
   1893 	if (dp->v_type != VDIR) {
   1894 		vrele(dp);
   1895 		error = ENOTDIR;
   1896 		goto out;
   1897 	}
   1898 
   1899 	if (rdonly)
   1900 		cnp->cn_flags |= RDONLY;
   1901 
   1902 	*retdirp = dp;
   1903 
   1904 	if (pubflag) {
   1905 		/*
   1906 		 * Oh joy. For WebNFS, handle those pesky '%' escapes,
   1907 		 * and the 'native path' indicator.
   1908 		 */
   1909 		cp = PNBUF_GET();
   1910 		fromcp = cnp->cn_pnbuf;
   1911 		tocp = cp;
   1912 		if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
   1913 			switch ((unsigned char)*fromcp) {
   1914 			case WEBNFS_NATIVE_CHAR:
   1915 				/*
   1916 				 * 'Native' path for us is the same
   1917 				 * as a path according to the NFS spec,
   1918 				 * just skip the escape char.
   1919 				 */
   1920 				fromcp++;
   1921 				break;
   1922 			/*
   1923 			 * More may be added in the future, range 0x80-0xff
   1924 			 */
   1925 			default:
   1926 				error = EIO;
   1927 				FREE(cp, M_NAMEI);
   1928 				goto out;
   1929 			}
   1930 		}
   1931 		/*
   1932 		 * Translate the '%' escapes, URL-style.
   1933 		 */
   1934 		while (*fromcp != '\0') {
   1935 			if (*fromcp == WEBNFS_ESC_CHAR) {
   1936 				if (fromcp[1] != '\0' && fromcp[2] != '\0') {
   1937 					fromcp++;
   1938 					*tocp++ = HEXSTRTOI(fromcp);
   1939 					fromcp += 2;
   1940 					continue;
   1941 				} else {
   1942 					error = ENOENT;
   1943 					FREE(cp, M_NAMEI);
   1944 					goto out;
   1945 				}
   1946 			} else
   1947 				*tocp++ = *fromcp++;
   1948 		}
   1949 		*tocp = '\0';
   1950 		PNBUF_PUT(cnp->cn_pnbuf);
   1951 		cnp->cn_pnbuf = cp;
   1952 	}
   1953 
   1954 	ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1;
   1955 	ndp->ni_segflg = UIO_SYSSPACE;
   1956 	ndp->ni_rootdir = rootvnode;
   1957 
   1958 	if (pubflag) {
   1959 		ndp->ni_loopcnt = 0;
   1960 		if (cnp->cn_pnbuf[0] == '/')
   1961 			dp = rootvnode;
   1962 	} else {
   1963 		cnp->cn_flags |= NOCROSSMOUNT;
   1964 	}
   1965 
   1966 	cnp->cn_proc = p;
   1967 	VREF(dp);
   1968 
   1969     for (;;) {
   1970 	cnp->cn_nameptr = cnp->cn_pnbuf;
   1971 	ndp->ni_startdir = dp;
   1972 	/*
   1973 	 * And call lookup() to do the real work
   1974 	 */
   1975 	error = lookup(ndp);
   1976 	if (error) {
   1977 		PNBUF_PUT(cnp->cn_pnbuf);
   1978 		return (error);
   1979 	}
   1980 	/*
   1981 	 * Check for encountering a symbolic link
   1982 	 */
   1983 	if ((cnp->cn_flags & ISSYMLINK) == 0) {
   1984 		if (cnp->cn_flags & (SAVENAME | SAVESTART))
   1985 			cnp->cn_flags |= HASBUF;
   1986 		else
   1987 			PNBUF_PUT(cnp->cn_pnbuf);
   1988 		return (0);
   1989 	} else {
   1990 		if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN))
   1991 			VOP_UNLOCK(ndp->ni_dvp, 0);
   1992 		if (!pubflag) {
   1993 			error = EINVAL;
   1994 			break;
   1995 		}
   1996 
   1997 		if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
   1998 			error = ELOOP;
   1999 			break;
   2000 		}
   2001 		if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) {
   2002 			error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred,
   2003 			    cnp->cn_proc);
   2004 			if (error != 0)
   2005 				break;
   2006 		}
   2007 		if (ndp->ni_pathlen > 1)
   2008 			cp = PNBUF_GET();
   2009 		else
   2010 			cp = cnp->cn_pnbuf;
   2011 		aiov.iov_base = cp;
   2012 		aiov.iov_len = MAXPATHLEN;
   2013 		auio.uio_iov = &aiov;
   2014 		auio.uio_iovcnt = 1;
   2015 		auio.uio_offset = 0;
   2016 		auio.uio_rw = UIO_READ;
   2017 		auio.uio_segflg = UIO_SYSSPACE;
   2018 		auio.uio_procp = (struct proc *)0;
   2019 		auio.uio_resid = MAXPATHLEN;
   2020 		error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred);
   2021 		if (error) {
   2022 		badlink:
   2023 			if (ndp->ni_pathlen > 1)
   2024 				PNBUF_PUT(cp);
   2025 			break;
   2026 		}
   2027 		linklen = MAXPATHLEN - auio.uio_resid;
   2028 		if (linklen == 0) {
   2029 			error = ENOENT;
   2030 			goto badlink;
   2031 		}
   2032 		if (linklen + ndp->ni_pathlen >= MAXPATHLEN) {
   2033 			error = ENAMETOOLONG;
   2034 			goto badlink;
   2035 		}
   2036 		if (ndp->ni_pathlen > 1) {
   2037 			memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen);
   2038 			PNBUF_PUT(cnp->cn_pnbuf);
   2039 			cnp->cn_pnbuf = cp;
   2040 		} else
   2041 			cnp->cn_pnbuf[linklen] = '\0';
   2042 		ndp->ni_pathlen += linklen;
   2043 		vput(ndp->ni_vp);
   2044 		dp = ndp->ni_dvp;
   2045 		/*
   2046 		 * Check if root directory should replace current directory.
   2047 		 */
   2048 		if (cnp->cn_pnbuf[0] == '/') {
   2049 			vrele(dp);
   2050 			dp = ndp->ni_rootdir;
   2051 			VREF(dp);
   2052 		}
   2053 	}
   2054    }
   2055 	vrele(ndp->ni_dvp);
   2056 	vput(ndp->ni_vp);
   2057 	ndp->ni_vp = NULL;
   2058 out:
   2059 	PNBUF_PUT(cnp->cn_pnbuf);
   2060 	return (error);
   2061 }
   2062 
   2063 /*
   2064  * A fiddled version of m_adj() that ensures null fill to a long
   2065  * boundary and only trims off the back end
   2066  */
   2067 void
   2068 nfsm_adj(mp, len, nul)
   2069 	struct mbuf *mp;
   2070 	int len;
   2071 	int nul;
   2072 {
   2073 	struct mbuf *m;
   2074 	int count, i;
   2075 	char *cp;
   2076 
   2077 	/*
   2078 	 * Trim from tail.  Scan the mbuf chain,
   2079 	 * calculating its length and finding the last mbuf.
   2080 	 * If the adjustment only affects this mbuf, then just
   2081 	 * adjust and return.  Otherwise, rescan and truncate
   2082 	 * after the remaining size.
   2083 	 */
   2084 	count = 0;
   2085 	m = mp;
   2086 	for (;;) {
   2087 		count += m->m_len;
   2088 		if (m->m_next == (struct mbuf *)0)
   2089 			break;
   2090 		m = m->m_next;
   2091 	}
   2092 	if (m->m_len > len) {
   2093 		m->m_len -= len;
   2094 		if (nul > 0) {
   2095 			cp = mtod(m, caddr_t)+m->m_len-nul;
   2096 			for (i = 0; i < nul; i++)
   2097 				*cp++ = '\0';
   2098 		}
   2099 		return;
   2100 	}
   2101 	count -= len;
   2102 	if (count < 0)
   2103 		count = 0;
   2104 	/*
   2105 	 * Correct length for chain is "count".
   2106 	 * Find the mbuf with last data, adjust its length,
   2107 	 * and toss data from remaining mbufs on chain.
   2108 	 */
   2109 	for (m = mp; m; m = m->m_next) {
   2110 		if (m->m_len >= count) {
   2111 			m->m_len = count;
   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 			break;
   2118 		}
   2119 		count -= m->m_len;
   2120 	}
   2121 	for (m = m->m_next;m;m = m->m_next)
   2122 		m->m_len = 0;
   2123 }
   2124 
   2125 /*
   2126  * Make these functions instead of macros, so that the kernel text size
   2127  * doesn't get too big...
   2128  */
   2129 void
   2130 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp)
   2131 	struct nfsrv_descript *nfsd;
   2132 	int before_ret;
   2133 	struct vattr *before_vap;
   2134 	int after_ret;
   2135 	struct vattr *after_vap;
   2136 	struct mbuf **mbp;
   2137 	char **bposp;
   2138 {
   2139 	struct mbuf *mb = *mbp, *mb2;
   2140 	char *bpos = *bposp;
   2141 	u_int32_t *tl;
   2142 
   2143 	if (before_ret) {
   2144 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
   2145 		*tl = nfs_false;
   2146 	} else {
   2147 		nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
   2148 		*tl++ = nfs_true;
   2149 		txdr_hyper(before_vap->va_size, tl);
   2150 		tl += 2;
   2151 		txdr_nfsv3time(&(before_vap->va_mtime), tl);
   2152 		tl += 2;
   2153 		txdr_nfsv3time(&(before_vap->va_ctime), tl);
   2154 	}
   2155 	*bposp = bpos;
   2156 	*mbp = mb;
   2157 	nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
   2158 }
   2159 
   2160 void
   2161 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp)
   2162 	struct nfsrv_descript *nfsd;
   2163 	int after_ret;
   2164 	struct vattr *after_vap;
   2165 	struct mbuf **mbp;
   2166 	char **bposp;
   2167 {
   2168 	struct mbuf *mb = *mbp, *mb2;
   2169 	char *bpos = *bposp;
   2170 	u_int32_t *tl;
   2171 	struct nfs_fattr *fp;
   2172 
   2173 	if (after_ret) {
   2174 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
   2175 		*tl = nfs_false;
   2176 	} else {
   2177 		nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
   2178 		*tl++ = nfs_true;
   2179 		fp = (struct nfs_fattr *)tl;
   2180 		nfsm_srvfattr(nfsd, after_vap, fp);
   2181 	}
   2182 	*mbp = mb;
   2183 	*bposp = bpos;
   2184 }
   2185 
   2186 void
   2187 nfsm_srvfattr(nfsd, vap, fp)
   2188 	struct nfsrv_descript *nfsd;
   2189 	struct vattr *vap;
   2190 	struct nfs_fattr *fp;
   2191 {
   2192 
   2193 	fp->fa_nlink = txdr_unsigned(vap->va_nlink);
   2194 	fp->fa_uid = txdr_unsigned(vap->va_uid);
   2195 	fp->fa_gid = txdr_unsigned(vap->va_gid);
   2196 	if (nfsd->nd_flag & ND_NFSV3) {
   2197 		fp->fa_type = vtonfsv3_type(vap->va_type);
   2198 		fp->fa_mode = vtonfsv3_mode(vap->va_mode);
   2199 		txdr_hyper(vap->va_size, &fp->fa3_size);
   2200 		txdr_hyper(vap->va_bytes, &fp->fa3_used);
   2201 		fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev));
   2202 		fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev));
   2203 		fp->fa3_fsid.nfsuquad[0] = 0;
   2204 		fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
   2205 		fp->fa3_fileid.nfsuquad[0] = 0;
   2206 		fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid);
   2207 		txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
   2208 		txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
   2209 		txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
   2210 	} else {
   2211 		fp->fa_type = vtonfsv2_type(vap->va_type);
   2212 		fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
   2213 		fp->fa2_size = txdr_unsigned(vap->va_size);
   2214 		fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
   2215 		if (vap->va_type == VFIFO)
   2216 			fp->fa2_rdev = 0xffffffff;
   2217 		else
   2218 			fp->fa2_rdev = txdr_unsigned(vap->va_rdev);
   2219 		fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
   2220 		fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
   2221 		fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
   2222 		txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
   2223 		txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
   2224 		txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
   2225 	}
   2226 }
   2227 
   2228 /*
   2229  * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
   2230  * 	- look up fsid in mount list (if not found ret error)
   2231  *	- get vp and export rights by calling VFS_FHTOVP()
   2232  *	- if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
   2233  *	- if not lockflag unlock it with VOP_UNLOCK()
   2234  */
   2235 int
   2236 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag)
   2237 	fhandle_t *fhp;
   2238 	int lockflag;
   2239 	struct vnode **vpp;
   2240 	struct ucred *cred;
   2241 	struct nfssvc_sock *slp;
   2242 	struct mbuf *nam;
   2243 	int *rdonlyp;
   2244 	int kerbflag;
   2245 {
   2246 	struct mount *mp;
   2247 	int i;
   2248 	struct ucred *credanon;
   2249 	int error, exflags;
   2250 	struct sockaddr_in *saddr;
   2251 
   2252 	*vpp = (struct vnode *)0;
   2253 
   2254 	if (nfs_ispublicfh(fhp)) {
   2255 		if (!pubflag || !nfs_pub.np_valid)
   2256 			return (ESTALE);
   2257 		fhp = &nfs_pub.np_handle;
   2258 	}
   2259 
   2260 	mp = vfs_getvfs(&fhp->fh_fsid);
   2261 	if (!mp)
   2262 		return (ESTALE);
   2263 	error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
   2264 	if (error)
   2265 		return (error);
   2266 	error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
   2267 	if (error)
   2268 		return (error);
   2269 
   2270 	if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
   2271 		saddr = mtod(nam, struct sockaddr_in *);
   2272 		if ((saddr->sin_family == AF_INET) &&
   2273 		    ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
   2274 			vput(*vpp);
   2275 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2276 		}
   2277 #ifdef INET6
   2278 		if ((saddr->sin_family == AF_INET6) &&
   2279 		    ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) {
   2280 			vput(*vpp);
   2281 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2282 		}
   2283 #endif
   2284 	}
   2285 	/*
   2286 	 * Check/setup credentials.
   2287 	 */
   2288 	if (exflags & MNT_EXKERB) {
   2289 		if (!kerbflag) {
   2290 			vput(*vpp);
   2291 			return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2292 		}
   2293 	} else if (kerbflag) {
   2294 		vput(*vpp);
   2295 		return (NFSERR_AUTHERR | AUTH_TOOWEAK);
   2296 	} else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
   2297 		cred->cr_uid = credanon->cr_uid;
   2298 		cred->cr_gid = credanon->cr_gid;
   2299 		for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
   2300 			cred->cr_groups[i] = credanon->cr_groups[i];
   2301 		cred->cr_ngroups = i;
   2302 	}
   2303 	if (exflags & MNT_EXRDONLY)
   2304 		*rdonlyp = 1;
   2305 	else
   2306 		*rdonlyp = 0;
   2307 	if (!lockflag)
   2308 		VOP_UNLOCK(*vpp, 0);
   2309 	return (0);
   2310 }
   2311 
   2312 /*
   2313  * WebNFS: check if a filehandle is a public filehandle. For v3, this
   2314  * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
   2315  * transformed this to all zeroes in both cases, so check for it.
   2316  */
   2317 int
   2318 nfs_ispublicfh(fhp)
   2319 	fhandle_t *fhp;
   2320 {
   2321 	char *cp = (char *)fhp;
   2322 	int i;
   2323 
   2324 	for (i = 0; i < NFSX_V3FH; i++)
   2325 		if (*cp++ != 0)
   2326 			return (FALSE);
   2327 	return (TRUE);
   2328 }
   2329 
   2330 /*
   2331  * This function compares two net addresses by family and returns TRUE
   2332  * if they are the same host.
   2333  * If there is any doubt, return FALSE.
   2334  * The AF_INET family is handled as a special case so that address mbufs
   2335  * don't need to be saved to store "struct in_addr", which is only 4 bytes.
   2336  */
   2337 int
   2338 netaddr_match(family, haddr, nam)
   2339 	int family;
   2340 	union nethostaddr *haddr;
   2341 	struct mbuf *nam;
   2342 {
   2343 	struct sockaddr_in *inetaddr;
   2344 
   2345 	switch (family) {
   2346 	case AF_INET:
   2347 		inetaddr = mtod(nam, struct sockaddr_in *);
   2348 		if (inetaddr->sin_family == AF_INET &&
   2349 		    inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
   2350 			return (1);
   2351 		break;
   2352 #ifdef INET6
   2353 	case AF_INET6:
   2354 	    {
   2355 		struct sockaddr_in6 *sin6_1, *sin6_2;
   2356 
   2357 		sin6_1 = mtod(nam, struct sockaddr_in6 *);
   2358 		sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *);
   2359 		if (sin6_1->sin6_family == AF_INET6 &&
   2360 		    IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr))
   2361 			return 1;
   2362 	    }
   2363 #endif
   2364 #ifdef ISO
   2365 	case AF_ISO:
   2366 	    {
   2367 		struct sockaddr_iso *isoaddr1, *isoaddr2;
   2368 
   2369 		isoaddr1 = mtod(nam, struct sockaddr_iso *);
   2370 		isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *);
   2371 		if (isoaddr1->siso_family == AF_ISO &&
   2372 		    isoaddr1->siso_nlen > 0 &&
   2373 		    isoaddr1->siso_nlen == isoaddr2->siso_nlen &&
   2374 		    SAME_ISOADDR(isoaddr1, isoaddr2))
   2375 			return (1);
   2376 		break;
   2377 	    }
   2378 #endif	/* ISO */
   2379 	default:
   2380 		break;
   2381 	};
   2382 	return (0);
   2383 }
   2384 
   2385 /*
   2386  * The write verifier has changed (probably due to a server reboot), so all
   2387  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
   2388  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
   2389  * flag. Once done the new write verifier can be set for the mount point.
   2390  */
   2391 void
   2392 nfs_clearcommit(mp)
   2393 	struct mount *mp;
   2394 {
   2395 	struct vnode *vp;
   2396 	struct nfsnode *np;
   2397 	struct vm_page *pg;
   2398 	int s;
   2399 
   2400 	s = splbio();
   2401 	LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) {
   2402 		KASSERT(vp->v_mount == mp);
   2403 		if (vp->v_type == VNON)
   2404 			continue;
   2405 		np = VTONFS(vp);
   2406 		np->n_pushlo = np->n_pushhi = np->n_pushedlo =
   2407 		    np->n_pushedhi = 0;
   2408 		np->n_commitflags &=
   2409 		    ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID);
   2410 		simple_lock(&vp->v_uobj.vmobjlock);
   2411 		TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) {
   2412 			pg->flags &= ~PG_NEEDCOMMIT;
   2413 		}
   2414 		simple_unlock(&vp->v_uobj.vmobjlock);
   2415 	}
   2416 	splx(s);
   2417 }
   2418 
   2419 void
   2420 nfs_merge_commit_ranges(vp)
   2421 	struct vnode *vp;
   2422 {
   2423 	struct nfsnode *np = VTONFS(vp);
   2424 
   2425 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) {
   2426 		np->n_pushedlo = np->n_pushlo;
   2427 		np->n_pushedhi = np->n_pushhi;
   2428 		np->n_commitflags |= NFS_COMMIT_PUSHED_VALID;
   2429 	} else {
   2430 		if (np->n_pushlo < np->n_pushedlo)
   2431 			np->n_pushedlo = np->n_pushlo;
   2432 		if (np->n_pushhi > np->n_pushedhi)
   2433 			np->n_pushedhi = np->n_pushhi;
   2434 	}
   2435 
   2436 	np->n_pushlo = np->n_pushhi = 0;
   2437 	np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID;
   2438 
   2439 #ifdef fvdl_debug
   2440 	printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2441 	    (unsigned)np->n_pushedhi);
   2442 #endif
   2443 }
   2444 
   2445 int
   2446 nfs_in_committed_range(vp, off, len)
   2447 	struct vnode *vp;
   2448 	off_t off, len;
   2449 {
   2450 	struct nfsnode *np = VTONFS(vp);
   2451 	off_t lo, hi;
   2452 
   2453 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID))
   2454 		return 0;
   2455 	lo = off;
   2456 	hi = lo + len;
   2457 
   2458 	return (lo >= np->n_pushedlo && hi <= np->n_pushedhi);
   2459 }
   2460 
   2461 int
   2462 nfs_in_tobecommitted_range(vp, off, len)
   2463 	struct vnode *vp;
   2464 	off_t off, len;
   2465 {
   2466 	struct nfsnode *np = VTONFS(vp);
   2467 	off_t lo, hi;
   2468 
   2469 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID))
   2470 		return 0;
   2471 	lo = off;
   2472 	hi = lo + len;
   2473 
   2474 	return (lo >= np->n_pushlo && hi <= np->n_pushhi);
   2475 }
   2476 
   2477 void
   2478 nfs_add_committed_range(vp, off, len)
   2479 	struct vnode *vp;
   2480 	off_t off, len;
   2481 {
   2482 	struct nfsnode *np = VTONFS(vp);
   2483 	off_t lo, hi;
   2484 
   2485 	lo = off;
   2486 	hi = lo + len;
   2487 
   2488 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) {
   2489 		np->n_pushedlo = lo;
   2490 		np->n_pushedhi = hi;
   2491 		np->n_commitflags |= NFS_COMMIT_PUSHED_VALID;
   2492 	} else {
   2493 		if (hi > np->n_pushedhi)
   2494 			np->n_pushedhi = hi;
   2495 		if (lo < np->n_pushedlo)
   2496 			np->n_pushedlo = lo;
   2497 	}
   2498 #ifdef fvdl_debug
   2499 	printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2500 	    (unsigned)np->n_pushedhi);
   2501 #endif
   2502 }
   2503 
   2504 void
   2505 nfs_del_committed_range(vp, off, len)
   2506 	struct vnode *vp;
   2507 	off_t off, len;
   2508 {
   2509 	struct nfsnode *np = VTONFS(vp);
   2510 	off_t lo, hi;
   2511 
   2512 	if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID))
   2513 		return;
   2514 
   2515 	lo = off;
   2516 	hi = lo + len;
   2517 
   2518 	if (lo > np->n_pushedhi || hi < np->n_pushedlo)
   2519 		return;
   2520 	if (lo <= np->n_pushedlo)
   2521 		np->n_pushedlo = hi;
   2522 	else if (hi >= np->n_pushedhi)
   2523 		np->n_pushedhi = lo;
   2524 	else {
   2525 		/*
   2526 		 * XXX There's only one range. If the deleted range
   2527 		 * is in the middle, pick the largest of the
   2528 		 * contiguous ranges that it leaves.
   2529 		 */
   2530 		if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi))
   2531 			np->n_pushedhi = lo;
   2532 		else
   2533 			np->n_pushedlo = hi;
   2534 	}
   2535 #ifdef fvdl_debug
   2536 	printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo,
   2537 	    (unsigned)np->n_pushedhi);
   2538 #endif
   2539 }
   2540 
   2541 void
   2542 nfs_add_tobecommitted_range(vp, off, len)
   2543 	struct vnode *vp;
   2544 	off_t off, len;
   2545 {
   2546 	struct nfsnode *np = VTONFS(vp);
   2547 	off_t lo, hi;
   2548 
   2549 	lo = off;
   2550 	hi = lo + len;
   2551 
   2552 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) {
   2553 		np->n_pushlo = lo;
   2554 		np->n_pushhi = hi;
   2555 		np->n_commitflags |= NFS_COMMIT_PUSH_VALID;
   2556 	} else {
   2557 		if (lo < np->n_pushlo)
   2558 			np->n_pushlo = lo;
   2559 		if (hi > np->n_pushhi)
   2560 			np->n_pushhi = hi;
   2561 	}
   2562 #ifdef fvdl_debug
   2563 	printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo,
   2564 	    (unsigned)np->n_pushhi);
   2565 #endif
   2566 }
   2567 
   2568 void
   2569 nfs_del_tobecommitted_range(vp, off, len)
   2570 	struct vnode *vp;
   2571 	off_t off, len;
   2572 {
   2573 	struct nfsnode *np = VTONFS(vp);
   2574 	off_t lo, hi;
   2575 
   2576 	if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID))
   2577 		return;
   2578 
   2579 	lo = off;
   2580 	hi = lo + len;
   2581 
   2582 	if (lo > np->n_pushhi || hi < np->n_pushlo)
   2583 		return;
   2584 
   2585 	if (lo <= np->n_pushlo)
   2586 		np->n_pushlo = hi;
   2587 	else if (hi >= np->n_pushhi)
   2588 		np->n_pushhi = lo;
   2589 	else {
   2590 		/*
   2591 		 * XXX There's only one range. If the deleted range
   2592 		 * is in the middle, pick the largest of the
   2593 		 * contiguous ranges that it leaves.
   2594 		 */
   2595 		if ((np->n_pushlo - lo) > (hi - np->n_pushhi))
   2596 			np->n_pushhi = lo;
   2597 		else
   2598 			np->n_pushlo = hi;
   2599 	}
   2600 #ifdef fvdl_debug
   2601 	printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo,
   2602 	    (unsigned)np->n_pushhi);
   2603 #endif
   2604 }
   2605 
   2606 /*
   2607  * Map errnos to NFS error numbers. For Version 3 also filter out error
   2608  * numbers not specified for the associated procedure.
   2609  */
   2610 int
   2611 nfsrv_errmap(nd, err)
   2612 	struct nfsrv_descript *nd;
   2613 	int err;
   2614 {
   2615 	const short *defaulterrp, *errp;
   2616 
   2617 	if (nd->nd_flag & ND_NFSV3) {
   2618 	    if (nd->nd_procnum <= NFSPROC_COMMIT) {
   2619 		errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
   2620 		while (*++errp) {
   2621 			if (*errp == err)
   2622 				return (err);
   2623 			else if (*errp > err)
   2624 				break;
   2625 		}
   2626 		return ((int)*defaulterrp);
   2627 	    } else
   2628 		return (err & 0xffff);
   2629 	}
   2630 	if (err <= ELAST)
   2631 		return ((int)nfsrv_v2errmap[err - 1]);
   2632 	return (NFSERR_IO);
   2633 }
   2634 
   2635 /*
   2636  * Sort the group list in increasing numerical order.
   2637  * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
   2638  *  that used to be here.)
   2639  */
   2640 void
   2641 nfsrvw_sort(list, num)
   2642         gid_t *list;
   2643         int num;
   2644 {
   2645 	int i, j;
   2646 	gid_t v;
   2647 
   2648 	/* Insertion sort. */
   2649 	for (i = 1; i < num; i++) {
   2650 		v = list[i];
   2651 		/* find correct slot for value v, moving others up */
   2652 		for (j = i; --j >= 0 && v < list[j];)
   2653 			list[j + 1] = list[j];
   2654 		list[j + 1] = v;
   2655 	}
   2656 }
   2657 
   2658 /*
   2659  * copy credentials making sure that the result can be compared with memcmp().
   2660  */
   2661 void
   2662 nfsrv_setcred(incred, outcred)
   2663 	struct ucred *incred, *outcred;
   2664 {
   2665 	int i;
   2666 
   2667 	memset((caddr_t)outcred, 0, sizeof (struct ucred));
   2668 	outcred->cr_ref = 1;
   2669 	outcred->cr_uid = incred->cr_uid;
   2670 	outcred->cr_gid = incred->cr_gid;
   2671 	outcred->cr_ngroups = incred->cr_ngroups;
   2672 	for (i = 0; i < incred->cr_ngroups; i++)
   2673 		outcred->cr_groups[i] = incred->cr_groups[i];
   2674 	nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
   2675 }
   2676