sys/vfs/nfs/nfs_subs.c

   1 /*
   2  * Copyright (c) 1989, 1993
   3  *      The Regents of the University of California.  All rights reserved.
   4  *
   5  * This code is derived from software contributed to Berkeley by
   6  * Rick Macklem at The University of Guelph.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions
  10  * are met:
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. All advertising materials mentioning features or use of this software
  17  *    must display the following acknowledgement:
  18  *      This product includes software developed by the University of
  19  *      California, Berkeley and its contributors.
  20  * 4. Neither the name of the University nor the names of its contributors
  21  *    may be used to endorse or promote products derived from this software
  22  *    without specific prior written permission.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  34  * SUCH DAMAGE.
  35  *
  36  *      @(#)nfs_subs.c  8.8 (Berkeley) 5/22/95
  37  * $FreeBSD: /repoman/r/ncvs/src/sys/nfsclient/nfs_subs.c,v 1.128 2004/04/14 23:23:55 peadar Exp $
  38  * $DragonFly: src/sys/vfs/nfs/nfs_subs.c,v 1.47 2007/11/02 19:52:28 dillon Exp $
  39  */
  40
  41 /*
  42  * These functions support the macros and help fiddle mbuf chains for
  43  * the nfs op functions. They do things like create the rpc header and
  44  * copy data between mbuf chains and uio lists.
  45  */
  46 #include <sys/param.h>
  47 #include <sys/systm.h>
  48 #include <sys/kernel.h>
  49 #include <sys/buf.h>
  50 #include <sys/proc.h>
  51 #include <sys/mount.h>
  52 #include <sys/vnode.h>
  53 #include <sys/nlookup.h>
  54 #include <sys/namei.h>
  55 #include <sys/mbuf.h>
  56 #include <sys/socket.h>
  57 #include <sys/stat.h>
  58 #include <sys/malloc.h>
  59 #include <sys/sysent.h>
  60 #include <sys/syscall.h>
  61 #include <sys/conf.h>
  62 #include <sys/objcache.h>
  63
  64 #include <vm/vm.h>
  65 #include <vm/vm_object.h>
  66 #include <vm/vm_extern.h>
  67 #include <vm/vm_zone.h>
  68
  69 #include <sys/buf2.h>
  70
  71 #include "rpcv2.h"
  72 #include "nfsproto.h"
  73 #include "nfs.h"
  74 #include "nfsmount.h"
  75 #include "nfsnode.h"
  76 #include "xdr_subs.h"
  77 #include "nfsm_subs.h"
  78 #include "nfsrtt.h"
  79
  80 #include <netinet/in.h>
  81
  82 /*
  83  * Data items converted to xdr at startup, since they are constant
  84  * This is kinda hokey, but may save a little time doing byte swaps
  85  */
  86 u_int32_t nfs_xdrneg1;
  87 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr,
  88         rpc_mismatch, rpc_auth_unix, rpc_msgaccepted,
  89         rpc_auth_kerb;
  90 u_int32_t nfs_prog, nfs_true, nfs_false;
  91
  92 /* And other global data */
  93 static u_int32_t nfs_xid = 0;
  94 static enum vtype nv2tov_type[8]= {
  95         VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON,  VNON
  96 };
  97 enum vtype nv3tov_type[8]= {
  98         VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO
  99 };
 100
 101 int nfs_ticks;
 102 int nfs_pbuf_freecnt = -1;      /* start out unlimited */
 103
 104 struct nfs_reqq nfs_reqq;
 105 struct nfssvc_sockhead nfssvc_sockhead;
 106 int nfssvc_sockhead_flag;
 107 struct nfsd_head nfsd_head;
 108 int nfsd_head_flag;
 109 struct nfs_bufq nfs_bufq;
 110 struct nqfhhashhead *nqfhhashtbl;
 111 u_long nqfhhash;
 112
 113 static int nfs_prev_nfssvc_sy_narg;
 114 static sy_call_t *nfs_prev_nfssvc_sy_call;
 115
 116 #ifndef NFS_NOSERVER
 117
 118 /*
 119  * Mapping of old NFS Version 2 RPC numbers to generic numbers.
 120  */
 121 int nfsv3_procid[NFS_NPROCS] = {
 122         NFSPROC_NULL,
 123         NFSPROC_GETATTR,
 124         NFSPROC_SETATTR,
 125         NFSPROC_NOOP,
 126         NFSPROC_LOOKUP,
 127         NFSPROC_READLINK,
 128         NFSPROC_READ,
 129         NFSPROC_NOOP,
 130         NFSPROC_WRITE,
 131         NFSPROC_CREATE,
 132         NFSPROC_REMOVE,
 133         NFSPROC_RENAME,
 134         NFSPROC_LINK,
 135         NFSPROC_SYMLINK,
 136         NFSPROC_MKDIR,
 137         NFSPROC_RMDIR,
 138         NFSPROC_READDIR,
 139         NFSPROC_FSSTAT,
 140         NFSPROC_NOOP,
 141         NFSPROC_NOOP,
 142         NFSPROC_NOOP,
 143         NFSPROC_NOOP,
 144         NFSPROC_NOOP,
 145         NFSPROC_NOOP,
 146         NFSPROC_NOOP,
 147         NFSPROC_NOOP
 148 };
 149
 150 #endif /* NFS_NOSERVER */
 151 /*
 152  * and the reverse mapping from generic to Version 2 procedure numbers
 153  */
 154 int nfsv2_procid[NFS_NPROCS] = {
 155         NFSV2PROC_NULL,
 156         NFSV2PROC_GETATTR,
 157         NFSV2PROC_SETATTR,
 158         NFSV2PROC_LOOKUP,
 159         NFSV2PROC_NOOP,
 160         NFSV2PROC_READLINK,
 161         NFSV2PROC_READ,
 162         NFSV2PROC_WRITE,
 163         NFSV2PROC_CREATE,
 164         NFSV2PROC_MKDIR,
 165         NFSV2PROC_SYMLINK,
 166         NFSV2PROC_CREATE,
 167         NFSV2PROC_REMOVE,
 168         NFSV2PROC_RMDIR,
 169         NFSV2PROC_RENAME,
 170         NFSV2PROC_LINK,
 171         NFSV2PROC_READDIR,
 172         NFSV2PROC_NOOP,
 173         NFSV2PROC_STATFS,
 174         NFSV2PROC_NOOP,
 175         NFSV2PROC_NOOP,
 176         NFSV2PROC_NOOP,
 177         NFSV2PROC_NOOP,
 178         NFSV2PROC_NOOP,
 179         NFSV2PROC_NOOP,
 180         NFSV2PROC_NOOP,
 181 };
 182
 183 #ifndef NFS_NOSERVER
 184 /*
 185  * Maps errno values to nfs error numbers.
 186  * Use NFSERR_IO as the catch all for ones not specifically defined in
 187  * RFC 1094.
 188  */
 189 static u_char nfsrv_v2errmap[ELAST] = {
 190   NFSERR_PERM,  NFSERR_NOENT,   NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 191   NFSERR_NXIO,  NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 192   NFSERR_IO,    NFSERR_IO,      NFSERR_ACCES,   NFSERR_IO,      NFSERR_IO,
 193   NFSERR_IO,    NFSERR_EXIST,   NFSERR_IO,      NFSERR_NODEV,   NFSERR_NOTDIR,
 194   NFSERR_ISDIR, NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 195   NFSERR_IO,    NFSERR_FBIG,    NFSERR_NOSPC,   NFSERR_IO,      NFSERR_ROFS,
 196   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 197   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 198   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 199   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 200   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 201   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 202   NFSERR_IO,    NFSERR_IO,      NFSERR_NAMETOL, NFSERR_IO,      NFSERR_IO,
 203   NFSERR_NOTEMPTY, NFSERR_IO,   NFSERR_IO,      NFSERR_DQUOT,   NFSERR_STALE,
 204   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 205   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 206   NFSERR_IO,    NFSERR_IO,      NFSERR_IO,      NFSERR_IO,      NFSERR_IO,
 207   NFSERR_IO /* << Last is 86 */
 208 };
 209
 210 /*
 211  * Maps errno values to nfs error numbers.
 212  * Although it is not obvious whether or not NFS clients really care if
 213  * a returned error value is in the specified list for the procedure, the
 214  * safest thing to do is filter them appropriately. For Version 2, the
 215  * X/Open XNFS document is the only specification that defines error values
 216  * for each RPC (The RFC simply lists all possible error values for all RPCs),
 217  * so I have decided to not do this for Version 2.
 218  * The first entry is the default error return and the rest are the valid
 219  * errors for that RPC in increasing numeric order.
 220  */
 221 static short nfsv3err_null[] = {
 222         0,
 223         0,
 224 };
 225
 226 static short nfsv3err_getattr[] = {
 227         NFSERR_IO,
 228         NFSERR_IO,
 229         NFSERR_STALE,
 230         NFSERR_BADHANDLE,
 231         NFSERR_SERVERFAULT,
 232         0,
 233 };
 234
 235 static short nfsv3err_setattr[] = {
 236         NFSERR_IO,
 237         NFSERR_PERM,
 238         NFSERR_IO,
 239         NFSERR_ACCES,
 240         NFSERR_INVAL,
 241         NFSERR_NOSPC,
 242         NFSERR_ROFS,
 243         NFSERR_DQUOT,
 244         NFSERR_STALE,
 245         NFSERR_BADHANDLE,
 246         NFSERR_NOT_SYNC,
 247         NFSERR_SERVERFAULT,
 248         0,
 249 };
 250
 251 static short nfsv3err_lookup[] = {
 252         NFSERR_IO,
 253         NFSERR_NOENT,
 254         NFSERR_IO,
 255         NFSERR_ACCES,
 256         NFSERR_NOTDIR,
 257         NFSERR_NAMETOL,
 258         NFSERR_STALE,
 259         NFSERR_BADHANDLE,
 260         NFSERR_SERVERFAULT,
 261         0,
 262 };
 263
 264 static short nfsv3err_access[] = {
 265         NFSERR_IO,
 266         NFSERR_IO,
 267         NFSERR_STALE,
 268         NFSERR_BADHANDLE,
 269         NFSERR_SERVERFAULT,
 270         0,
 271 };
 272
 273 static short nfsv3err_readlink[] = {
 274         NFSERR_IO,
 275         NFSERR_IO,
 276         NFSERR_ACCES,
 277         NFSERR_INVAL,
 278         NFSERR_STALE,
 279         NFSERR_BADHANDLE,
 280         NFSERR_NOTSUPP,
 281         NFSERR_SERVERFAULT,
 282         0,
 283 };
 284
 285 static short nfsv3err_read[] = {
 286         NFSERR_IO,
 287         NFSERR_IO,
 288         NFSERR_NXIO,
 289         NFSERR_ACCES,
 290         NFSERR_INVAL,
 291         NFSERR_STALE,
 292         NFSERR_BADHANDLE,
 293         NFSERR_SERVERFAULT,
 294         0,
 295 };
 296
 297 static short nfsv3err_write[] = {
 298         NFSERR_IO,
 299         NFSERR_IO,
 300         NFSERR_ACCES,
 301         NFSERR_INVAL,
 302         NFSERR_FBIG,
 303         NFSERR_NOSPC,
 304         NFSERR_ROFS,
 305         NFSERR_DQUOT,
 306         NFSERR_STALE,
 307         NFSERR_BADHANDLE,
 308         NFSERR_SERVERFAULT,
 309         0,
 310 };
 311
 312 static short nfsv3err_create[] = {
 313         NFSERR_IO,
 314         NFSERR_IO,
 315         NFSERR_ACCES,
 316         NFSERR_EXIST,
 317         NFSERR_NOTDIR,
 318         NFSERR_NOSPC,
 319         NFSERR_ROFS,
 320         NFSERR_NAMETOL,
 321         NFSERR_DQUOT,
 322         NFSERR_STALE,
 323         NFSERR_BADHANDLE,
 324         NFSERR_NOTSUPP,
 325         NFSERR_SERVERFAULT,
 326         0,
 327 };
 328
 329 static short nfsv3err_mkdir[] = {
 330         NFSERR_IO,
 331         NFSERR_IO,
 332         NFSERR_ACCES,
 333         NFSERR_EXIST,
 334         NFSERR_NOTDIR,
 335         NFSERR_NOSPC,
 336         NFSERR_ROFS,
 337         NFSERR_NAMETOL,
 338         NFSERR_DQUOT,
 339         NFSERR_STALE,
 340         NFSERR_BADHANDLE,
 341         NFSERR_NOTSUPP,
 342         NFSERR_SERVERFAULT,
 343         0,
 344 };
 345
 346 static short nfsv3err_symlink[] = {
 347         NFSERR_IO,
 348         NFSERR_IO,
 349         NFSERR_ACCES,
 350         NFSERR_EXIST,
 351         NFSERR_NOTDIR,
 352         NFSERR_NOSPC,
 353         NFSERR_ROFS,
 354         NFSERR_NAMETOL,
 355         NFSERR_DQUOT,
 356         NFSERR_STALE,
 357         NFSERR_BADHANDLE,
 358         NFSERR_NOTSUPP,
 359         NFSERR_SERVERFAULT,
 360         0,
 361 };
 362
 363 static short nfsv3err_mknod[] = {
 364         NFSERR_IO,
 365         NFSERR_IO,
 366         NFSERR_ACCES,
 367         NFSERR_EXIST,
 368         NFSERR_NOTDIR,
 369         NFSERR_NOSPC,
 370         NFSERR_ROFS,
 371         NFSERR_NAMETOL,
 372         NFSERR_DQUOT,
 373         NFSERR_STALE,
 374         NFSERR_BADHANDLE,
 375         NFSERR_NOTSUPP,
 376         NFSERR_SERVERFAULT,
 377         NFSERR_BADTYPE,
 378         0,
 379 };
 380
 381 static short nfsv3err_remove[] = {
 382         NFSERR_IO,
 383         NFSERR_NOENT,
 384         NFSERR_IO,
 385         NFSERR_ACCES,
 386         NFSERR_NOTDIR,
 387         NFSERR_ROFS,
 388         NFSERR_NAMETOL,
 389         NFSERR_STALE,
 390         NFSERR_BADHANDLE,
 391         NFSERR_SERVERFAULT,
 392         0,
 393 };
 394
 395 static short nfsv3err_rmdir[] = {
 396         NFSERR_IO,
 397         NFSERR_NOENT,
 398         NFSERR_IO,
 399         NFSERR_ACCES,
 400         NFSERR_EXIST,
 401         NFSERR_NOTDIR,
 402         NFSERR_INVAL,
 403         NFSERR_ROFS,
 404         NFSERR_NAMETOL,
 405         NFSERR_NOTEMPTY,
 406         NFSERR_STALE,
 407         NFSERR_BADHANDLE,
 408         NFSERR_NOTSUPP,
 409         NFSERR_SERVERFAULT,
 410         0,
 411 };
 412
 413 static short nfsv3err_rename[] = {
 414         NFSERR_IO,
 415         NFSERR_NOENT,
 416         NFSERR_IO,
 417         NFSERR_ACCES,
 418         NFSERR_EXIST,
 419         NFSERR_XDEV,
 420         NFSERR_NOTDIR,
 421         NFSERR_ISDIR,
 422         NFSERR_INVAL,
 423         NFSERR_NOSPC,
 424         NFSERR_ROFS,
 425         NFSERR_MLINK,
 426         NFSERR_NAMETOL,
 427         NFSERR_NOTEMPTY,
 428         NFSERR_DQUOT,
 429         NFSERR_STALE,
 430         NFSERR_BADHANDLE,
 431         NFSERR_NOTSUPP,
 432         NFSERR_SERVERFAULT,
 433         0,
 434 };
 435
 436 static short nfsv3err_link[] = {
 437         NFSERR_IO,
 438         NFSERR_IO,
 439         NFSERR_ACCES,
 440         NFSERR_EXIST,
 441         NFSERR_XDEV,
 442         NFSERR_NOTDIR,
 443         NFSERR_INVAL,
 444         NFSERR_NOSPC,
 445         NFSERR_ROFS,
 446         NFSERR_MLINK,
 447         NFSERR_NAMETOL,
 448         NFSERR_DQUOT,
 449         NFSERR_STALE,
 450         NFSERR_BADHANDLE,
 451         NFSERR_NOTSUPP,
 452         NFSERR_SERVERFAULT,
 453         0,
 454 };
 455
 456 static short nfsv3err_readdir[] = {
 457         NFSERR_IO,
 458         NFSERR_IO,
 459         NFSERR_ACCES,
 460         NFSERR_NOTDIR,
 461         NFSERR_STALE,
 462         NFSERR_BADHANDLE,
 463         NFSERR_BAD_COOKIE,
 464         NFSERR_TOOSMALL,
 465         NFSERR_SERVERFAULT,
 466         0,
 467 };
 468
 469 static short nfsv3err_readdirplus[] = {
 470         NFSERR_IO,
 471         NFSERR_IO,
 472         NFSERR_ACCES,
 473         NFSERR_NOTDIR,
 474         NFSERR_STALE,
 475         NFSERR_BADHANDLE,
 476         NFSERR_BAD_COOKIE,
 477         NFSERR_NOTSUPP,
 478         NFSERR_TOOSMALL,
 479         NFSERR_SERVERFAULT,
 480         0,
 481 };
 482
 483 static short nfsv3err_fsstat[] = {
 484         NFSERR_IO,
 485         NFSERR_IO,
 486         NFSERR_STALE,
 487         NFSERR_BADHANDLE,
 488         NFSERR_SERVERFAULT,
 489         0,
 490 };
 491
 492 static short nfsv3err_fsinfo[] = {
 493         NFSERR_STALE,
 494         NFSERR_STALE,
 495         NFSERR_BADHANDLE,
 496         NFSERR_SERVERFAULT,
 497         0,
 498 };
 499
 500 static short nfsv3err_pathconf[] = {
 501         NFSERR_STALE,
 502         NFSERR_STALE,
 503         NFSERR_BADHANDLE,
 504         NFSERR_SERVERFAULT,
 505         0,
 506 };
 507
 508 static short nfsv3err_commit[] = {
 509         NFSERR_IO,
 510         NFSERR_IO,
 511         NFSERR_STALE,
 512         NFSERR_BADHANDLE,
 513         NFSERR_SERVERFAULT,
 514         0,
 515 };
 516
 517 static short *nfsrv_v3errmap[] = {
 518         nfsv3err_null,
 519         nfsv3err_getattr,
 520         nfsv3err_setattr,
 521         nfsv3err_lookup,
 522         nfsv3err_access,
 523         nfsv3err_readlink,
 524         nfsv3err_read,
 525         nfsv3err_write,
 526         nfsv3err_create,
 527         nfsv3err_mkdir,
 528         nfsv3err_symlink,
 529         nfsv3err_mknod,
 530         nfsv3err_remove,
 531         nfsv3err_rmdir,
 532         nfsv3err_rename,
 533         nfsv3err_link,
 534         nfsv3err_readdir,
 535         nfsv3err_readdirplus,
 536         nfsv3err_fsstat,
 537         nfsv3err_fsinfo,
 538         nfsv3err_pathconf,
 539         nfsv3err_commit,
 540 };
 541
 542 #endif /* NFS_NOSERVER */
 543
 544 extern struct nfsrtt nfsrtt;
 545 extern struct nfsstats nfsstats;
 546 extern nfstype nfsv2_type[9];
 547 extern nfstype nfsv3_type[9];
 548 extern struct nfsnodehashhead *nfsnodehashtbl;
 549 extern u_long nfsnodehash;
 550
 551 struct nfssvc_args;
 552 extern int sys_nfssvc(struct proc *, struct nfssvc_args *, int *);
 553
 554 LIST_HEAD(nfsnodehashhead, nfsnode);
 555
 556 /*
 557  * This needs to return a monotonically increasing or close to monotonically
 558  * increasing result, otherwise the write gathering queues won't work
 559  * properly.
 560  */
 561 u_quad_t
 562 nfs_curusec(void)
 563 {
 564         struct timeval tv;
 565
 566         getmicrouptime(&tv);
 567         return ((u_quad_t)tv.tv_sec * 1000000 + (u_quad_t)tv.tv_usec);
 568 }
 569
 570 /*
 571  * Create the header for an rpc request packet
 572  * The hsiz is the size of the rest of the nfs request header.
 573  * (just used to decide if a cluster is a good idea)
 574  */
 575 struct mbuf *
 576 nfsm_reqh(struct vnode *vp, u_long procid, int hsiz, caddr_t *bposp)
 577 {
 578         struct mbuf *mb;
 579         caddr_t bpos;
 580
 581         mb = m_getl(hsiz, MB_WAIT, MT_DATA, 0, NULL);
 582         mb->m_len = 0;
 583         bpos = mtod(mb, caddr_t);
 584
 585         /* Finally, return values */
 586         *bposp = bpos;
 587         return (mb);
 588 }
 589
 590 /*
 591  * Build the RPC header and fill in the authorization info.
 592  * The authorization string argument is only used when the credentials
 593  * come from outside of the kernel.
 594  * Returns the head of the mbuf list.
 595  */
 596 struct mbuf *
 597 nfsm_rpchead(struct ucred *cr, int nmflag, int procid, int auth_type,
 598              int auth_len, char *auth_str, int verf_len, char *verf_str,
 599              struct mbuf *mrest, int mrest_len, struct mbuf **mbp,
 600              u_int32_t *xidp)
 601 {
 602         struct mbuf *mb;
 603         u_int32_t *tl;
 604         caddr_t bpos;
 605         int i;
 606         struct mbuf *mreq, *mb2;
 607         int siz, grpsiz, authsiz, dsiz;
 608
 609         authsiz = nfsm_rndup(auth_len);
 610         dsiz = authsiz + 10 * NFSX_UNSIGNED;
 611         mb = m_getl(dsiz, MB_WAIT, MT_DATA, M_PKTHDR, NULL);
 612         if (dsiz < MINCLSIZE) {
 613                 if (dsiz < MHLEN)
 614                         MH_ALIGN(mb, dsiz);
 615                 else
 616                         MH_ALIGN(mb, 8 * NFSX_UNSIGNED);
 617         }
 618         mb->m_len = mb->m_pkthdr.len = 0;
 619         mreq = mb;
 620         bpos = mtod(mb, caddr_t);
 621
 622         /*
 623          * First the RPC header.
 624          */
 625         nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED);
 626
 627         /* Get a pretty random xid to start with */
 628         if (!nfs_xid)
 629                 nfs_xid = krandom();
 630         /*
 631          * Skip zero xid if it should ever happen.
 632          */
 633         if (++nfs_xid == 0)
 634                 nfs_xid++;
 635
 636         *tl++ = *xidp = txdr_unsigned(nfs_xid);
 637         *tl++ = rpc_call;
 638         *tl++ = rpc_vers;
 639         *tl++ = txdr_unsigned(NFS_PROG);
 640         if (nmflag & NFSMNT_NFSV3)
 641                 *tl++ = txdr_unsigned(NFS_VER3);
 642         else
 643                 *tl++ = txdr_unsigned(NFS_VER2);
 644         if (nmflag & NFSMNT_NFSV3)
 645                 *tl++ = txdr_unsigned(procid);
 646         else
 647                 *tl++ = txdr_unsigned(nfsv2_procid[procid]);
 648
 649         /*
 650          * And then the authorization cred.
 651          */
 652         *tl++ = txdr_unsigned(auth_type);
 653         *tl = txdr_unsigned(authsiz);
 654         switch (auth_type) {
 655         case RPCAUTH_UNIX:
 656                 nfsm_build(tl, u_int32_t *, auth_len);
 657                 *tl++ = 0;              /* stamp ?? */
 658                 *tl++ = 0;              /* NULL hostname */
 659                 *tl++ = txdr_unsigned(cr->cr_uid);
 660                 *tl++ = txdr_unsigned(cr->cr_groups[0]);
 661                 grpsiz = (auth_len >> 2) - 5;
 662                 *tl++ = txdr_unsigned(grpsiz);
 663                 for (i = 1; i <= grpsiz; i++)
 664                         *tl++ = txdr_unsigned(cr->cr_groups[i]);
 665                 break;
 666         case RPCAUTH_KERB4:
 667                 siz = auth_len;
 668                 while (siz > 0) {
 669                         if (M_TRAILINGSPACE(mb) == 0) {
 670                                 mb2 = m_getl(siz, MB_WAIT, MT_DATA, 0, NULL);
 671                                 mb2->m_len = 0;
 672                                 mb->m_next = mb2;
 673                                 mb = mb2;
 674                                 bpos = mtod(mb, caddr_t);
 675                         }
 676                         i = min(siz, M_TRAILINGSPACE(mb));
 677                         bcopy(auth_str, bpos, i);
 678                         mb->m_len += i;
 679                         auth_str += i;
 680                         bpos += i;
 681                         siz -= i;
 682                 }
 683                 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) {
 684                         for (i = 0; i < siz; i++)
 685                                 *bpos++ = '\0';
 686                         mb->m_len += siz;
 687                 }
 688                 break;
 689         };
 690
 691         /*
 692          * And the verifier...
 693          */
 694         nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED);
 695         if (verf_str) {
 696                 *tl++ = txdr_unsigned(RPCAUTH_KERB4);
 697                 *tl = txdr_unsigned(verf_len);
 698                 siz = verf_len;
 699                 while (siz > 0) {
 700                         if (M_TRAILINGSPACE(mb) == 0) {
 701                                 mb2 = m_getl(siz, MB_WAIT, MT_DATA, 0, NULL);
 702                                 mb2->m_len = 0;
 703                                 mb->m_next = mb2;
 704                                 mb = mb2;
 705                                 bpos = mtod(mb, caddr_t);
 706                         }
 707                         i = min(siz, M_TRAILINGSPACE(mb));
 708                         bcopy(verf_str, bpos, i);
 709                         mb->m_len += i;
 710                         verf_str += i;
 711                         bpos += i;
 712                         siz -= i;
 713                 }
 714                 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) {
 715                         for (i = 0; i < siz; i++)
 716                                 *bpos++ = '\0';
 717                         mb->m_len += siz;
 718                 }
 719         } else {
 720                 *tl++ = txdr_unsigned(RPCAUTH_NULL);
 721                 *tl = 0;
 722         }
 723         mb->m_next = mrest;
 724         mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len;
 725         mreq->m_pkthdr.rcvif = (struct ifnet *)0;
 726         *mbp = mb;
 727         return (mreq);
 728 }
 729
 730 /*
 731  * copies mbuf chain to the uio scatter/gather list
 732  */
 733 int
 734 nfsm_mbuftouio(struct mbuf **mrep, struct uio *uiop, int siz, caddr_t *dpos)
 735 {
 736         char *mbufcp, *uiocp;
 737         int xfer, left, len;
 738         struct mbuf *mp;
 739         long uiosiz, rem;
 740         int error = 0;
 741
 742         mp = *mrep;
 743         mbufcp = *dpos;
 744         len = mtod(mp, caddr_t)+mp->m_len-mbufcp;
 745         rem = nfsm_rndup(siz)-siz;
 746         while (siz > 0) {
 747                 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL)
 748                         return (EFBIG);
 749                 left = uiop->uio_iov->iov_len;
 750                 uiocp = uiop->uio_iov->iov_base;
 751                 if (left > siz)
 752                         left = siz;
 753                 uiosiz = left;
 754                 while (left > 0) {
 755                         while (len == 0) {
 756                                 mp = mp->m_next;
 757                                 if (mp == NULL)
 758                                         return (EBADRPC);
 759                                 mbufcp = mtod(mp, caddr_t);
 760                                 len = mp->m_len;
 761                         }
 762                         xfer = (left > len) ? len : left;
 763 #ifdef notdef
 764                         /* Not Yet.. */
 765                         if (uiop->uio_iov->iov_op != NULL)
 766                                 (*(uiop->uio_iov->iov_op))
 767                                 (mbufcp, uiocp, xfer);
 768                         else
 769 #endif
 770                         if (uiop->uio_segflg == UIO_SYSSPACE)
 771                                 bcopy(mbufcp, uiocp, xfer);
 772                         else
 773                                 copyout(mbufcp, uiocp, xfer);
 774                         left -= xfer;
 775                         len -= xfer;
 776                         mbufcp += xfer;
 777                         uiocp += xfer;
 778                         uiop->uio_offset += xfer;
 779                         uiop->uio_resid -= xfer;
 780                 }
 781                 if (uiop->uio_iov->iov_len <= siz) {
 782                         uiop->uio_iovcnt--;
 783                         uiop->uio_iov++;
 784                 } else {
 785                         uiop->uio_iov->iov_base += uiosiz;
 786                         uiop->uio_iov->iov_len -= uiosiz;
 787                 }
 788                 siz -= uiosiz;
 789         }
 790         *dpos = mbufcp;
 791         *mrep = mp;
 792         if (rem > 0) {
 793                 if (len < rem)
 794                         error = nfs_adv(mrep, dpos, rem, len);
 795                 else
 796                         *dpos += rem;
 797         }
 798         return (error);
 799 }
 800
 801 /*
 802  * copies a uio scatter/gather list to an mbuf chain.
 803  * NOTE: can ony handle iovcnt == 1
 804  */
 805 int
 806 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, caddr_t *bpos)
 807 {
 808         char *uiocp;
 809         struct mbuf *mp, *mp2;
 810         int xfer, left, mlen;
 811         int uiosiz, rem;
 812         boolean_t getcluster;
 813         char *cp;
 814
 815 #ifdef DIAGNOSTIC
 816         if (uiop->uio_iovcnt != 1)
 817                 panic("nfsm_uiotombuf: iovcnt != 1");
 818 #endif
 819
 820         if (siz >= MINCLSIZE)
 821                 getcluster = TRUE;
 822         else
 823                 getcluster = FALSE;
 824         rem = nfsm_rndup(siz) - siz;
 825         mp = mp2 = *mq;
 826         while (siz > 0) {
 827                 left = uiop->uio_iov->iov_len;
 828                 uiocp = uiop->uio_iov->iov_base;
 829                 if (left > siz)
 830                         left = siz;
 831                 uiosiz = left;
 832                 while (left > 0) {
 833                         mlen = M_TRAILINGSPACE(mp);
 834                         if (mlen == 0) {
 835                                 if (getcluster)
 836                                         mp = m_getcl(MB_WAIT, MT_DATA, 0);
 837                                 else
 838                                         mp = m_get(MB_WAIT, MT_DATA);
 839                                 mp->m_len = 0;
 840                                 mp2->m_next = mp;
 841                                 mp2 = mp;
 842                                 mlen = M_TRAILINGSPACE(mp);
 843                         }
 844                         xfer = (left > mlen) ? mlen : left;
 845 #ifdef notdef
 846                         /* Not Yet.. */
 847                         if (uiop->uio_iov->iov_op != NULL)
 848                                 (*(uiop->uio_iov->iov_op))
 849                                 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
 850                         else
 851 #endif
 852                         if (uiop->uio_segflg == UIO_SYSSPACE)
 853                                 bcopy(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
 854                         else
 855                                 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer);
 856                         mp->m_len += xfer;
 857                         left -= xfer;
 858                         uiocp += xfer;
 859                         uiop->uio_offset += xfer;
 860                         uiop->uio_resid -= xfer;
 861                 }
 862                 uiop->uio_iov->iov_base += uiosiz;
 863                 uiop->uio_iov->iov_len -= uiosiz;
 864                 siz -= uiosiz;
 865         }
 866         if (rem > 0) {
 867                 if (rem > M_TRAILINGSPACE(mp)) {
 868                         MGET(mp, MB_WAIT, MT_DATA);
 869                         mp->m_len = 0;
 870                         mp2->m_next = mp;
 871                 }
 872                 cp = mtod(mp, caddr_t)+mp->m_len;
 873                 for (left = 0; left < rem; left++)
 874                         *cp++ = '\0';
 875                 mp->m_len += rem;
 876                 *bpos = cp;
 877         } else
 878                 *bpos = mtod(mp, caddr_t)+mp->m_len;
 879         *mq = mp;
 880         return (0);
 881 }
 882
 883 /*
 884  * Help break down an mbuf chain by setting the first siz bytes contiguous
 885  * pointed to by returned val.
 886  * This is used by the macros nfsm_dissect and nfsm_dissecton for tough
 887  * cases. (The macros use the vars. dpos and dpos2)
 888  */
 889 int
 890 nfsm_disct(struct mbuf **mdp, caddr_t *dposp, int siz, int left, caddr_t *cp2)
 891 {
 892         struct mbuf *mp, *mp2;
 893         int siz2, xfer;
 894         caddr_t p;
 895
 896         mp = *mdp;
 897         while (left == 0) {
 898                 *mdp = mp = mp->m_next;
 899                 if (mp == NULL)
 900                         return (EBADRPC);
 901                 left = mp->m_len;
 902                 *dposp = mtod(mp, caddr_t);
 903         }
 904         if (left >= siz) {
 905                 *cp2 = *dposp;
 906                 *dposp += siz;
 907         } else if (mp->m_next == NULL) {
 908                 return (EBADRPC);
 909         } else if (siz > MHLEN) {
 910                 panic("nfs S too big");
 911         } else {
 912                 MGET(mp2, MB_WAIT, MT_DATA);
 913                 mp2->m_next = mp->m_next;
 914                 mp->m_next = mp2;
 915                 mp->m_len -= left;
 916                 mp = mp2;
 917                 *cp2 = p = mtod(mp, caddr_t);
 918                 bcopy(*dposp, p, left);         /* Copy what was left */
 919                 siz2 = siz-left;
 920                 p += left;
 921                 mp2 = mp->m_next;
 922                 /* Loop around copying up the siz2 bytes */
 923                 while (siz2 > 0) {
 924                         if (mp2 == NULL)
 925                                 return (EBADRPC);
 926                         xfer = (siz2 > mp2->m_len) ? mp2->m_len : siz2;
 927                         if (xfer > 0) {
 928                                 bcopy(mtod(mp2, caddr_t), p, xfer);
 929                                 NFSMADV(mp2, xfer);
 930                                 mp2->m_len -= xfer;
 931                                 p += xfer;
 932                                 siz2 -= xfer;
 933                         }
 934                         if (siz2 > 0)
 935                                 mp2 = mp2->m_next;
 936                 }
 937                 mp->m_len = siz;
 938                 *mdp = mp2;
 939                 *dposp = mtod(mp2, caddr_t);
 940         }
 941         return (0);
 942 }
 943
 944 /*
 945  * Advance the position in the mbuf chain.
 946  */
 947 int
 948 nfs_adv(struct mbuf **mdp, caddr_t *dposp, int offs, int left)
 949 {
 950         struct mbuf *m;
 951         int s;
 952
 953         m = *mdp;
 954         s = left;
 955         while (s < offs) {
 956                 offs -= s;
 957                 m = m->m_next;
 958                 if (m == NULL)
 959                         return (EBADRPC);
 960                 s = m->m_len;
 961         }
 962         *mdp = m;
 963         *dposp = mtod(m, caddr_t)+offs;
 964         return (0);
 965 }
 966
 967 /*
 968  * Copy a string into mbufs for the hard cases...
 969  */
 970 int
 971 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz)
 972 {
 973         struct mbuf *m1 = NULL, *m2;
 974         long left, xfer, len, tlen;
 975         u_int32_t *tl;
 976         int putsize;
 977
 978         putsize = 1;
 979         m2 = *mb;
 980         left = M_TRAILINGSPACE(m2);
 981         if (left > 0) {
 982                 tl = ((u_int32_t *)(*bpos));
 983                 *tl++ = txdr_unsigned(siz);
 984                 putsize = 0;
 985                 left -= NFSX_UNSIGNED;
 986                 m2->m_len += NFSX_UNSIGNED;
 987                 if (left > 0) {
 988                         bcopy(cp, (caddr_t) tl, left);
 989                         siz -= left;
 990                         cp += left;
 991                         m2->m_len += left;
 992                         left = 0;
 993                 }
 994         }
 995         /* Loop around adding mbufs */
 996         while (siz > 0) {
 997                 int msize;
 998
 999                 m1 = m_getl(siz, MB_WAIT, MT_DATA, 0, &msize);
1000                 m1->m_len = msize;
1001                 m2->m_next = m1;
1002                 m2 = m1;
1003                 tl = mtod(m1, u_int32_t *);
1004                 tlen = 0;
1005                 if (putsize) {
1006                         *tl++ = txdr_unsigned(siz);
1007                         m1->m_len -= NFSX_UNSIGNED;
1008                         tlen = NFSX_UNSIGNED;
1009                         putsize = 0;
1010                 }
1011                 if (siz < m1->m_len) {
1012                         len = nfsm_rndup(siz);
1013                         xfer = siz;
1014                         if (xfer < len)
1015                                 *(tl+(xfer>>2)) = 0;
1016                 } else {
1017                         xfer = len = m1->m_len;
1018                 }
1019                 bcopy(cp, (caddr_t) tl, xfer);
1020                 m1->m_len = len+tlen;
1021                 siz -= xfer;
1022                 cp += xfer;
1023         }
1024         *mb = m1;
1025         *bpos = mtod(m1, caddr_t)+m1->m_len;
1026         return (0);
1027 }
1028
1029 /*
1030  * Called once to initialize data structures...
1031  */
1032 int
1033 nfs_init(struct vfsconf *vfsp)
1034 {
1035         int i;
1036
1037         callout_init(&nfs_timer_handle);
1038         nfsmount_zone = zinit("NFSMOUNT", sizeof(struct nfsmount), 0, 0, 1);
1039
1040         nfs_mount_type = vfsp->vfc_typenum;
1041         nfsrtt.pos = 0;
1042         rpc_vers = txdr_unsigned(RPC_VER2);
1043         rpc_call = txdr_unsigned(RPC_CALL);
1044         rpc_reply = txdr_unsigned(RPC_REPLY);
1045         rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED);
1046         rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED);
1047         rpc_mismatch = txdr_unsigned(RPC_MISMATCH);
1048         rpc_autherr = txdr_unsigned(RPC_AUTHERR);
1049         rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX);
1050         rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4);
1051         nfs_prog = txdr_unsigned(NFS_PROG);
1052         nfs_true = txdr_unsigned(TRUE);
1053         nfs_false = txdr_unsigned(FALSE);
1054         nfs_xdrneg1 = txdr_unsigned(-1);
1055         nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000;
1056         if (nfs_ticks < 1)
1057                 nfs_ticks = 1;
1058         /* Ensure async daemons disabled */
1059         for (i = 0; i < NFS_MAXASYNCDAEMON; i++) {
1060                 nfs_iodwant[i] = NULL;
1061                 nfs_iodmount[i] = (struct nfsmount *)0;
1062         }
1063         nfs_nhinit();                   /* Init the nfsnode table */
1064 #ifndef NFS_NOSERVER
1065         nfsrv_init(0);                  /* Init server data structures */
1066         nfsrv_initcache();              /* Init the server request cache */
1067 #endif
1068
1069         /*
1070          * Initialize reply list and start timer
1071          */
1072         TAILQ_INIT(&nfs_reqq);
1073
1074         nfs_timer(0);
1075
1076         nfs_prev_nfssvc_sy_narg = sysent[SYS_nfssvc].sy_narg;
1077         sysent[SYS_nfssvc].sy_narg = 2;
1078         nfs_prev_nfssvc_sy_call = sysent[SYS_nfssvc].sy_call;
1079         sysent[SYS_nfssvc].sy_call = (sy_call_t *)sys_nfssvc;
1080
1081         nfs_pbuf_freecnt = nswbuf / 2 + 1;
1082
1083         return (0);
1084 }
1085
1086 int
1087 nfs_uninit(struct vfsconf *vfsp)
1088 {
1089         callout_stop(&nfs_timer_handle);
1090         nfs_mount_type = -1;
1091         sysent[SYS_nfssvc].sy_narg = nfs_prev_nfssvc_sy_narg;
1092         sysent[SYS_nfssvc].sy_call = nfs_prev_nfssvc_sy_call;
1093         return (0);
1094 }
1095
1096 /*
1097  * Attribute cache routines.
1098  * nfs_loadattrcache() - loads or updates the cache contents from attributes
1099  *      that are on the mbuf list
1100  * nfs_getattrcache() - returns valid attributes if found in cache, returns
1101  *      error otherwise
1102  */
1103
1104 /*
1105  * Load the attribute cache (that lives in the nfsnode entry) with
1106  * the values on the mbuf list.  Load *vaper with the attributes.  vaper
1107  * may be NULL.
1108  *
1109  * As a side effect n_mtime, which we use to determine if the file was
1110  * modified by some other host, is set to the attribute timestamp and
1111  * NRMODIFIED is set if the two values differ.
1112  *
1113  * WARNING: the mtime loaded into vaper does not necessarily represent
1114  * n_mtime or n_attr.mtime due to NACC and NUPD.
1115  */
1116 int
1117 nfs_loadattrcache(struct vnode **vpp, struct mbuf **mdp, caddr_t *dposp,
1118                   struct vattr *vaper, int lattr_flags)
1119 {
1120         struct vnode *vp = *vpp;
1121         struct vattr *vap;
1122         struct nfs_fattr *fp;
1123         struct nfsnode *np;
1124         int32_t t1;
1125         caddr_t cp2;
1126         int error = 0;
1127         int rmajor, rminor;
1128         udev_t rdev;
1129         struct mbuf *md;
1130         enum vtype vtyp;
1131         u_short vmode;
1132         struct timespec mtime;
1133         int v3 = NFS_ISV3(vp);
1134
1135         md = *mdp;
1136         t1 = (mtod(md, caddr_t) + md->m_len) - *dposp;
1137         if ((error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2)) != 0)
1138                 return (error);
1139         fp = (struct nfs_fattr *)cp2;
1140         if (v3) {
1141                 vtyp = nfsv3tov_type(fp->fa_type);
1142                 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1143                 rmajor = (int)fxdr_unsigned(int, fp->fa3_rdev.specdata1);
1144                 rminor = (int)fxdr_unsigned(int, fp->fa3_rdev.specdata2);
1145                 fxdr_nfsv3time(&fp->fa3_mtime, &mtime);
1146         } else {
1147                 vtyp = nfsv2tov_type(fp->fa_type);
1148                 vmode = fxdr_unsigned(u_short, fp->fa_mode);
1149                 /*
1150                  * XXX
1151                  *
1152                  * The duplicate information returned in fa_type and fa_mode
1153                  * is an ambiguity in the NFS version 2 protocol.
1154                  *
1155                  * VREG should be taken literally as a regular file.  If a
1156                  * server intents to return some type information differently
1157                  * in the upper bits of the mode field (e.g. for sockets, or
1158                  * FIFOs), NFSv2 mandates fa_type to be VNON.  Anyway, we
1159                  * leave the examination of the mode bits even in the VREG
1160                  * case to avoid breakage for bogus servers, but we make sure
1161                  * that there are actually type bits set in the upper part of
1162                  * fa_mode (and failing that, trust the va_type field).
1163                  *
1164                  * NFSv3 cleared the issue, and requires fa_mode to not
1165                  * contain any type information (while also introduing sockets
1166                  * and FIFOs for fa_type).
1167                  */
1168                 if (vtyp == VNON || (vtyp == VREG && (vmode & S_IFMT) != 0))
1169                         vtyp = IFTOVT(vmode);
1170                 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev);
1171                 rmajor = umajor(rdev);
1172                 rminor = uminor(rdev);
1173                 fxdr_nfsv2time(&fp->fa2_mtime, &mtime);
1174
1175                 /*
1176                  * Really ugly NFSv2 kludge.
1177                  */
1178                 if (vtyp == VCHR && rdev == (udev_t)0xffffffff)
1179                         vtyp = VFIFO;
1180         }
1181
1182         /*
1183          * If v_type == VNON it is a new node, so fill in the v_type,
1184          * n_mtime fields. Check to see if it represents a special
1185          * device, and if so, check for a possible alias. Once the
1186          * correct vnode has been obtained, fill in the rest of the
1187          * information.
1188          */
1189         np = VTONFS(vp);
1190         if (vp->v_type != vtyp) {
1191                 nfs_setvtype(vp, vtyp);
1192                 if (vp->v_type == VFIFO) {
1193                         vp->v_ops = &vp->v_mount->mnt_vn_fifo_ops;
1194                 } else if (vp->v_type == VCHR || vp->v_type == VBLK) {
1195                         vp->v_ops = &vp->v_mount->mnt_vn_spec_ops;
1196                         addaliasu(vp, rmajor, rminor);
1197                 } else {
1198                         vp->v_ops = &vp->v_mount->mnt_vn_use_ops;
1199                 }
1200                 np->n_mtime = mtime.tv_sec;
1201         } else if (np->n_mtime != mtime.tv_sec) {
1202                 /*
1203                  * If we haven't modified the file locally and the server
1204                  * timestamp does not match, then the server probably
1205                  * modified the file.  We must flag this condition so
1206                  * the proper syncnronization can be done.  We do not
1207                  * try to synchronize the state here because that
1208                  * could lead to an endless recursion.
1209                  *
1210                  * XXX loadattrcache can be set during the reply to a write,
1211                  * before the write timestamp is properly processed.  To
1212                  * avoid unconditionally setting the rmodified bit (which
1213                  * has the effect of flushing the cache), we only do this
1214                  * check if the lmodified bit is not set.
1215                  */
1216                 np->n_mtime = mtime.tv_sec;
1217                 if ((lattr_flags & NFS_LATTR_NOMTIMECHECK) == 0)
1218                         np->n_flag |= NRMODIFIED;
1219         }
1220         vap = &np->n_vattr;
1221         vap->va_type = vtyp;
1222         vap->va_mode = (vmode & 07777);
1223         vap->va_rmajor = rmajor;
1224         vap->va_rminor = rminor;
1225         vap->va_mtime = mtime;
1226         vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0];
1227         if (v3) {
1228                 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1229                 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1230                 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1231                 vap->va_size = fxdr_hyper(&fp->fa3_size);
1232                 vap->va_blocksize = NFS_FABLKSIZE;
1233                 vap->va_bytes = fxdr_hyper(&fp->fa3_used);
1234                 vap->va_fileid = fxdr_hyper(&fp->fa3_fileid);
1235                 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime);
1236                 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime);
1237                 vap->va_flags = 0;
1238                 vap->va_filerev = 0;
1239         } else {
1240                 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink);
1241                 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid);
1242                 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid);
1243                 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size);
1244                 vap->va_blocksize = fxdr_unsigned(int32_t, fp->fa2_blocksize);
1245                 vap->va_bytes = (u_quad_t)fxdr_unsigned(int32_t, fp->fa2_blocks)
1246                     * NFS_FABLKSIZE;
1247                 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid);
1248                 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime);
1249                 vap->va_flags = 0;
1250                 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t,
1251                     fp->fa2_ctime.nfsv2_sec);
1252                 vap->va_ctime.tv_nsec = 0;
1253                 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec);
1254                 vap->va_filerev = 0;
1255         }
1256         np->n_attrstamp = time_second;
1257         if (vap->va_size != np->n_size) {
1258                 if (vap->va_type == VREG) {
1259                         if ((lattr_flags & NFS_LATTR_NOSHRINK) &&
1260                             vap->va_size < np->n_size) {
1261                                 /*
1262                                  * We've been told not to shrink the file;
1263                                  * zero np->n_attrstamp to indicate that
1264                                  * the attributes are stale.
1265                                  *
1266                                  * This occurs primarily due to recursive
1267                                  * NFS ops that are executed during periods
1268                                  * where we cannot safely reduce the size of
1269                                  * the file.
1270                                  *
1271                                  * Additionally, write rpcs are broken down
1272                                  * into buffers and np->n_size is
1273                                  * pre-extended.  Setting NRMODIFIED here
1274                                  * can result in n_size getting reset to a
1275                                  * lower value, which is NOT what we want.
1276                                  * XXX this needs to be cleaned up a lot
1277                                  * more.
1278                                  */
1279                                 vap->va_size = np->n_size;
1280                                 np->n_attrstamp = 0;
1281                                 if ((np->n_flag & NLMODIFIED) == 0)
1282                                         np->n_flag |= NRMODIFIED;
1283                         } else if (np->n_flag & NLMODIFIED) {
1284                                 /*
1285                                  * We've modified the file: Use the larger
1286                                  * of our size, and the server's size.  At
1287                                  * this point the cache coherency is all
1288                                  * shot to hell.  To try to handle multiple
1289                                  * clients appending to the file at the same
1290                                  * time mark that the server has changed
1291                                  * the file if the server's notion of the
1292                                  * file size is larger then our notion.
1293                                  *
1294                                  * XXX this needs work.
1295                                  */
1296                                 if (vap->va_size < np->n_size) {
1297                                         vap->va_size = np->n_size;
1298                                 } else {
1299                                         np->n_size = vap->va_size;
1300                                         np->n_flag |= NRMODIFIED;
1301                                 }
1302                         } else {
1303                                 /*
1304                                  * Someone changed the file's size on the
1305                                  * server and there are no local changes
1306                                  * to get in the way, set the size and mark
1307                                  * it.
1308                                  */
1309                                 np->n_size = vap->va_size;
1310                                 np->n_flag |= NRMODIFIED;
1311                         }
1312                         vnode_pager_setsize(vp, np->n_size);
1313                 } else {
1314                         np->n_size = vap->va_size;
1315                 }
1316         }
1317         if (vaper != NULL) {
1318                 bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(*vap));
1319                 if (np->n_flag & NCHG) {
1320                         if (np->n_flag & NACC)
1321                                 vaper->va_atime = np->n_atim;
1322                         if (np->n_flag & NUPD)
1323                                 vaper->va_mtime = np->n_mtim;
1324                 }
1325         }
1326         return (0);
1327 }
1328
1329 #ifdef NFS_ACDEBUG
1330 #include <sys/sysctl.h>
1331 SYSCTL_DECL(_vfs_nfs);
1332 static int nfs_acdebug;
1333 SYSCTL_INT(_vfs_nfs, OID_AUTO, acdebug, CTLFLAG_RW, &nfs_acdebug, 0, "");
1334 #endif
1335
1336 /*
1337  * Check the time stamp
1338  * If the cache is valid, copy contents to *vap and return 0
1339  * otherwise return an error
1340  */
1341 int
1342 nfs_getattrcache(struct vnode *vp, struct vattr *vaper)
1343 {
1344         struct nfsnode *np;
1345         struct vattr *vap;
1346         struct nfsmount *nmp;
1347         int timeo;
1348
1349         np = VTONFS(vp);
1350         vap = &np->n_vattr;
1351         nmp = VFSTONFS(vp->v_mount);
1352
1353         /*
1354          * Dynamic timeout based on how recently the file was modified.
1355          * n_mtime is always valid.
1356          */
1357         timeo = (get_approximate_time_t() - np->n_mtime) / 60;
1358
1359 #ifdef NFS_ACDEBUG
1360         if (nfs_acdebug>1)
1361                 kprintf("nfs_getattrcache: initial timeo = %d\n", timeo);
1362 #endif
1363
1364         if (vap->va_type == VDIR) {
1365                 if ((np->n_flag & NLMODIFIED) || timeo < nmp->nm_acdirmin)
1366                         timeo = nmp->nm_acdirmin;
1367                 else if (timeo > nmp->nm_acdirmax)
1368                         timeo = nmp->nm_acdirmax;
1369         } else {
1370                 if ((np->n_flag & NLMODIFIED) || timeo < nmp->nm_acregmin)
1371                         timeo = nmp->nm_acregmin;
1372                 else if (timeo > nmp->nm_acregmax)
1373                         timeo = nmp->nm_acregmax;
1374         }
1375
1376 #ifdef NFS_ACDEBUG
1377         if (nfs_acdebug > 2)
1378                 kprintf("acregmin %d; acregmax %d; acdirmin %d; acdirmax %d\n",
1379                         nmp->nm_acregmin, nmp->nm_acregmax,
1380                         nmp->nm_acdirmin, nmp->nm_acdirmax);
1381
1382         if (nfs_acdebug)
1383                 kprintf("nfs_getattrcache: age = %d; final timeo = %d\n",
1384                         (int)(time_second - np->n_attrstamp), timeo);
1385 #endif
1386
1387         if (np->n_attrstamp == 0 || (time_second - np->n_attrstamp) >= timeo) {
1388                 nfsstats.attrcache_misses++;
1389                 return (ENOENT);
1390         }
1391         nfsstats.attrcache_hits++;
1392
1393         /*
1394          * Our attribute cache can be stale due to modifications made on
1395          * this host.  XXX this is a bad hack.  We need a more deterministic
1396          * means of finding out which np fields are valid verses attr cache
1397          * fields.  We really should update the vattr info on the fly when
1398          * making local changes.
1399          */
1400         if (vap->va_size != np->n_size) {
1401                 if (vap->va_type == VREG) {
1402                         if (np->n_flag & NLMODIFIED) {
1403                                 if (vap->va_size < np->n_size)
1404                                         vap->va_size = np->n_size;
1405                                 else
1406                                         np->n_size = vap->va_size;
1407                         } else {
1408                                 np->n_size = vap->va_size;
1409                         }
1410                         vnode_pager_setsize(vp, np->n_size);
1411                 } else {
1412                         np->n_size = vap->va_size;
1413                 }
1414         }
1415         bcopy((caddr_t)vap, (caddr_t)vaper, sizeof(struct vattr));
1416         if (np->n_flag & NCHG) {
1417                 if (np->n_flag & NACC)
1418                         vaper->va_atime = np->n_atim;
1419                 if (np->n_flag & NUPD)
1420                         vaper->va_mtime = np->n_mtim;
1421         }
1422         return (0);
1423 }
1424
1425 #ifndef NFS_NOSERVER
1426
1427 /*
1428  * Set up nameidata for a lookup() call and do it.
1429  *
1430  * If pubflag is set, this call is done for a lookup operation on the
1431  * public filehandle. In that case we allow crossing mountpoints and
1432  * absolute pathnames. However, the caller is expected to check that
1433  * the lookup result is within the public fs, and deny access if
1434  * it is not.
1435  *
1436  * dirp may be set whether an error is returned or not, and must be
1437  * released by the caller.
1438  *
1439  * On return nd->nl_nch usually points to the target ncp, which may represent
1440  * a negative hit.
1441  *
1442  * NOTE: the caller must call nlookup_done(nd) unconditionally on return
1443  * to cleanup.
1444  */
1445 int
1446 nfs_namei(struct nlookupdata *nd, struct ucred *cred, int nameiop,
1447         struct vnode **dvpp, struct vnode **vpp,
1448         fhandle_t *fhp, int len,
1449         struct nfssvc_sock *slp, struct sockaddr *nam, struct mbuf **mdp,
1450         caddr_t *dposp, struct vnode **dirpp, struct thread *td,
1451         int kerbflag, int pubflag)
1452 {
1453         int i, rem;
1454         int flags;
1455         struct mbuf *md;
1456         char *fromcp, *tocp, *cp;
1457         char *namebuf;
1458         struct nchandle nch;
1459         struct vnode *dp;
1460         int error, rdonly;
1461
1462         namebuf = objcache_get(namei_oc, M_WAITOK);
1463         flags = 0;
1464         *dirpp = NULL;
1465
1466         /*
1467          * Copy the name from the mbuf list to namebuf.
1468          */
1469         fromcp = *dposp;
1470         tocp = namebuf;
1471         md = *mdp;
1472         rem = mtod(md, caddr_t) + md->m_len - fromcp;
1473         for (i = 0; i < len; i++) {
1474                 while (rem == 0) {
1475                         md = md->m_next;
1476                         if (md == NULL) {
1477                                 error = EBADRPC;
1478                                 goto out;
1479                         }
1480                         fromcp = mtod(md, caddr_t);
1481                         rem = md->m_len;
1482                 }
1483                 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) {
1484                         error = EACCES;
1485                         goto out;
1486                 }
1487                 *tocp++ = *fromcp++;
1488                 rem--;
1489         }
1490         *tocp = '\0';
1491         *mdp = md;
1492         *dposp = fromcp;
1493         len = nfsm_rndup(len)-len;
1494         if (len > 0) {
1495                 if (rem >= len)
1496                         *dposp += len;
1497                 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0)
1498                         goto out;
1499         }
1500
1501         /*
1502          * Extract and set starting directory.  The returned dp is refd
1503          * but not locked.
1504          */
1505         error = nfsrv_fhtovp(fhp, FALSE, &dp, cred, slp,
1506                                 nam, &rdonly, kerbflag, pubflag);
1507         if (error)
1508                 goto out;
1509         if (dp->v_type != VDIR) {
1510                 vrele(dp);
1511                 error = ENOTDIR;
1512                 goto out;
1513         }
1514
1515         /*
1516          * Set return directory.  Reference to dp is implicitly transfered
1517          * to the returned pointer.  This must be set before we potentially
1518          * goto out below.
1519          */
1520         *dirpp = dp;
1521
1522         if (pubflag) {
1523                 /*
1524                  * Oh joy. For WebNFS, handle those pesky '%' escapes,
1525                  * and the 'native path' indicator.
1526                  */
1527                 cp = objcache_get(namei_oc, M_WAITOK);
1528                 fromcp = namebuf;
1529                 tocp = cp;
1530                 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) {
1531                         switch ((unsigned char)*fromcp) {
1532                         case WEBNFS_NATIVE_CHAR:
1533                                 /*
1534                                  * 'Native' path for us is the same
1535                                  * as a path according to the NFS spec,
1536                                  * just skip the escape char.
1537                                  */
1538                                 fromcp++;
1539                                 break;
1540                         /*
1541                          * More may be added in the future, range 0x80-0xff
1542                          */
1543                         default:
1544                                 error = EIO;
1545                                 objcache_put(namei_oc, cp);
1546                                 goto out;
1547                         }
1548                 }
1549                 /*
1550                  * Translate the '%' escapes, URL-style.
1551                  */
1552                 while (*fromcp != '\0') {
1553                         if (*fromcp == WEBNFS_ESC_CHAR) {
1554                                 if (fromcp[1] != '\0' && fromcp[2] != '\0') {
1555                                         fromcp++;
1556                                         *tocp++ = HEXSTRTOI(fromcp);
1557                                         fromcp += 2;
1558                                         continue;
1559                                 } else {
1560                                         error = ENOENT;
1561                                         objcache_put(namei_oc, cp);
1562                                         goto out;
1563                                 }
1564                         } else
1565                                 *tocp++ = *fromcp++;
1566                 }
1567                 *tocp = '\0';
1568                 objcache_put(namei_oc, namebuf);
1569                 namebuf = cp;
1570         }
1571
1572         /*
1573          * Setup for search.  We need to get a start directory from dp.  Note
1574          * that dp is ref'd, but we no longer 'own' the ref (*dirpp owns it).
1575          */
1576         if (pubflag == 0) {
1577                 flags |= NLC_NFS_NOSOFTLINKTRAV;
1578                 flags |= NLC_NOCROSSMOUNT;
1579         }
1580         if (rdonly)
1581                 flags |= NLC_NFS_RDONLY;
1582         if (nameiop == NAMEI_CREATE || nameiop == NAMEI_RENAME)
1583                 flags |= NLC_CREATE;
1584
1585         /*
1586          * We need a starting ncp from the directory vnode dp.  dp must not
1587          * be locked.  The returned ncp will be refd but not locked.
1588          *
1589          * If no suitable ncp is found we instruct cache_fromdvp() to create
1590          * one.  If this fails the directory has probably been removed while
1591          * the target was chdir'd into it and any further lookup will fail.
1592          */
1593         if ((error = cache_fromdvp(dp, cred, 1, &nch)) != 0)
1594                 goto out;
1595         nlookup_init_raw(nd, namebuf, UIO_SYSSPACE, flags, cred, &nch);
1596         cache_drop(&nch);
1597
1598         /*
1599          * Ok, do the lookup.
1600          */
1601         error = nlookup(nd);
1602
1603         /*
1604          * If no error occured return the requested dvpp and vpp.  If
1605          * NLC_CREATE was specified nd->nl_nch may represent a negative
1606          * cache hit in which case we do not attempt to obtain the vp.
1607          */
1608         if (error == 0) {
1609                 if (dvpp) {
1610                         if (nd->nl_nch.ncp->nc_parent) {
1611                                 nch = nd->nl_nch;
1612                                 nch.ncp = nch.ncp->nc_parent;
1613                                 error = cache_vget(&nch, nd->nl_cred,
1614                                                    LK_EXCLUSIVE, dvpp);
1615                         } else {
1616                                 error = ENXIO;
1617                         }
1618                 }
1619                 if (vpp && nd->nl_nch.ncp->nc_vp) {
1620                         error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, vpp);
1621                 }
1622                 if (error) {
1623                         if (dvpp && *dvpp) {
1624                                 vput(*dvpp);
1625                                 *dvpp = NULL;
1626                         }
1627                         if (vpp && *vpp) {
1628                                 vput(*vpp);
1629                                 *vpp = NULL;
1630                         }
1631                 }
1632         }
1633
1634         /*
1635          * Finish up.
1636          */
1637 out:
1638         objcache_put(namei_oc, namebuf);
1639         return (error);
1640 }
1641
1642 /*
1643  * A fiddled version of m_adj() that ensures null fill to a long
1644  * boundary and only trims off the back end
1645  */
1646 void
1647 nfsm_adj(struct mbuf *mp, int len, int nul)
1648 {
1649         struct mbuf *m;
1650         int count, i;
1651         char *cp;
1652
1653         /*
1654          * Trim from tail.  Scan the mbuf chain,
1655          * calculating its length and finding the last mbuf.
1656          * If the adjustment only affects this mbuf, then just
1657          * adjust and return.  Otherwise, rescan and truncate
1658          * after the remaining size.
1659          */
1660         count = 0;
1661         m = mp;
1662         for (;;) {
1663                 count += m->m_len;
1664                 if (m->m_next == (struct mbuf *)0)
1665                         break;
1666                 m = m->m_next;
1667         }
1668         if (m->m_len > len) {
1669                 m->m_len -= len;
1670                 if (nul > 0) {
1671                         cp = mtod(m, caddr_t)+m->m_len-nul;
1672                         for (i = 0; i < nul; i++)
1673                                 *cp++ = '\0';
1674                 }
1675                 return;
1676         }
1677         count -= len;
1678         if (count < 0)
1679                 count = 0;
1680         /*
1681          * Correct length for chain is "count".
1682          * Find the mbuf with last data, adjust its length,
1683          * and toss data from remaining mbufs on chain.
1684          */
1685         for (m = mp; m; m = m->m_next) {
1686                 if (m->m_len >= count) {
1687                         m->m_len = count;
1688                         if (nul > 0) {
1689                                 cp = mtod(m, caddr_t)+m->m_len-nul;
1690                                 for (i = 0; i < nul; i++)
1691                                         *cp++ = '\0';
1692                         }
1693                         break;
1694                 }
1695                 count -= m->m_len;
1696         }
1697         for (m = m->m_next;m;m = m->m_next)
1698                 m->m_len = 0;
1699 }
1700
1701 /*
1702  * Make these functions instead of macros, so that the kernel text size
1703  * doesn't get too big...
1704  */
1705 void
1706 nfsm_srvwcc(struct nfsrv_descript *nfsd, int before_ret,
1707             struct vattr *before_vap, int after_ret, struct vattr *after_vap,
1708             struct mbuf **mbp, char **bposp)
1709 {
1710         struct mbuf *mb = *mbp, *mb2;
1711         char *bpos = *bposp;
1712         u_int32_t *tl;
1713
1714         /*
1715          * before_ret is 0 if before_vap is valid, non-zero if it isn't.
1716          */
1717         if (before_ret) {
1718                 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1719                 *tl = nfs_false;
1720         } else {
1721                 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED);
1722                 *tl++ = nfs_true;
1723                 txdr_hyper(before_vap->va_size, tl);
1724                 tl += 2;
1725                 txdr_nfsv3time(&(before_vap->va_mtime), tl);
1726                 tl += 2;
1727                 txdr_nfsv3time(&(before_vap->va_ctime), tl);
1728         }
1729         *bposp = bpos;
1730         *mbp = mb;
1731         nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp);
1732 }
1733
1734 void
1735 nfsm_srvpostopattr(struct nfsrv_descript *nfsd, int after_ret,
1736                    struct vattr *after_vap, struct mbuf **mbp, char **bposp)
1737 {
1738         struct mbuf *mb = *mbp, *mb2;
1739         char *bpos = *bposp;
1740         u_int32_t *tl;
1741         struct nfs_fattr *fp;
1742
1743         if (after_ret) {
1744                 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED);
1745                 *tl = nfs_false;
1746         } else {
1747                 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR);
1748                 *tl++ = nfs_true;
1749                 fp = (struct nfs_fattr *)tl;
1750                 nfsm_srvfattr(nfsd, after_vap, fp);
1751         }
1752         *mbp = mb;
1753         *bposp = bpos;
1754 }
1755
1756 void
1757 nfsm_srvfattr(struct nfsrv_descript *nfsd, struct vattr *vap,
1758               struct nfs_fattr *fp)
1759 {
1760         /*
1761          * NFS seems to truncate nlink to 16 bits, don't let it overflow.
1762          */
1763         if (vap->va_nlink > 65535)
1764                 fp->fa_nlink = 65535;
1765         else
1766                 fp->fa_nlink = txdr_unsigned(vap->va_nlink);
1767         fp->fa_uid = txdr_unsigned(vap->va_uid);
1768         fp->fa_gid = txdr_unsigned(vap->va_gid);
1769         if (nfsd->nd_flag & ND_NFSV3) {
1770                 fp->fa_type = vtonfsv3_type(vap->va_type);
1771                 fp->fa_mode = vtonfsv3_mode(vap->va_mode);
1772                 txdr_hyper(vap->va_size, &fp->fa3_size);
1773                 txdr_hyper(vap->va_bytes, &fp->fa3_used);
1774                 fp->fa3_rdev.specdata1 = txdr_unsigned(vap->va_rmajor);
1775                 fp->fa3_rdev.specdata2 = txdr_unsigned(vap->va_rminor);
1776                 fp->fa3_fsid.nfsuquad[0] = 0;
1777                 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid);
1778                 txdr_hyper(vap->va_fileid, &fp->fa3_fileid);
1779                 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime);
1780                 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime);
1781                 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime);
1782         } else {
1783                 fp->fa_type = vtonfsv2_type(vap->va_type);
1784                 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode);
1785                 fp->fa2_size = txdr_unsigned(vap->va_size);
1786                 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize);
1787                 if (vap->va_type == VFIFO)
1788                         fp->fa2_rdev = 0xffffffff;
1789                 else
1790                         fp->fa2_rdev = txdr_unsigned(makeudev(vap->va_rmajor, vap->va_rminor));
1791                 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE);
1792                 fp->fa2_fsid = txdr_unsigned(vap->va_fsid);
1793                 fp->fa2_fileid = txdr_unsigned(vap->va_fileid);
1794                 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime);
1795                 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime);
1796                 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime);
1797         }
1798 }
1799
1800 /*
1801  * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked)
1802  *      - look up fsid in mount list (if not found ret error)
1803  *      - get vp and export rights by calling VFS_FHTOVP()
1804  *      - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon
1805  *      - if not lockflag unlock it with vn_unlock()
1806  */
1807 int
1808 nfsrv_fhtovp(fhandle_t *fhp, int lockflag, struct vnode **vpp,
1809              struct ucred *cred, struct nfssvc_sock *slp, struct sockaddr *nam,
1810              int *rdonlyp, int kerbflag, int pubflag)
1811 {
1812         struct mount *mp;
1813         int i;
1814         struct ucred *credanon;
1815         int error, exflags;
1816 #ifdef MNT_EXNORESPORT          /* XXX needs mountd and /etc/exports help yet */
1817         struct sockaddr_int *saddr;
1818 #endif
1819
1820         *vpp = (struct vnode *)0;
1821
1822         if (nfs_ispublicfh(fhp)) {
1823                 if (!pubflag || !nfs_pub.np_valid)
1824                         return (ESTALE);
1825                 fhp = &nfs_pub.np_handle;
1826         }
1827
1828         mp = vfs_getvfs(&fhp->fh_fsid);
1829         if (!mp)
1830                 return (ESTALE);
1831         error = VFS_CHECKEXP(mp, nam, &exflags, &credanon);
1832         if (error)
1833                 return (error);
1834         error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp);
1835         if (error)
1836                 return (error);
1837 #ifdef MNT_EXNORESPORT
1838         if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) {
1839                 saddr = (struct sockaddr_in *)nam;
1840                 if (saddr->sin_family == AF_INET &&
1841                     ntohs(saddr->sin_port) >= IPPORT_RESERVED) {
1842                         vput(*vpp);
1843                         *vpp = NULL;
1844                         return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1845                 }
1846         }
1847 #endif
1848         /*
1849          * Check/setup credentials.
1850          */
1851         if (exflags & MNT_EXKERB) {
1852                 if (!kerbflag) {
1853                         vput(*vpp);
1854                         *vpp = NULL;
1855                         return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1856                 }
1857         } else if (kerbflag) {
1858                 vput(*vpp);
1859                 *vpp = NULL;
1860                 return (NFSERR_AUTHERR | AUTH_TOOWEAK);
1861         } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) {
1862                 cred->cr_uid = credanon->cr_uid;
1863                 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++)
1864                         cred->cr_groups[i] = credanon->cr_groups[i];
1865                 cred->cr_ngroups = i;
1866         }
1867         if (exflags & MNT_EXRDONLY)
1868                 *rdonlyp = 1;
1869         else
1870                 *rdonlyp = 0;
1871
1872         if (!lockflag)
1873                 vn_unlock(*vpp);
1874         return (0);
1875 }
1876
1877
1878 /*
1879  * WebNFS: check if a filehandle is a public filehandle. For v3, this
1880  * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has
1881  * transformed this to all zeroes in both cases, so check for it.
1882  */
1883 int
1884 nfs_ispublicfh(fhandle_t *fhp)
1885 {
1886         char *cp = (char *)fhp;
1887         int i;
1888
1889         for (i = 0; i < NFSX_V3FH; i++)
1890                 if (*cp++ != 0)
1891                         return (FALSE);
1892         return (TRUE);
1893 }
1894
1895 #endif /* NFS_NOSERVER */
1896 /*
1897  * This function compares two net addresses by family and returns TRUE
1898  * if they are the same host.
1899  * If there is any doubt, return FALSE.
1900  * The AF_INET family is handled as a special case so that address mbufs
1901  * don't need to be saved to store "struct in_addr", which is only 4 bytes.
1902  */
1903 int
1904 netaddr_match(int family, union nethostaddr *haddr, struct sockaddr *nam)
1905 {
1906         struct sockaddr_in *inetaddr;
1907
1908         switch (family) {
1909         case AF_INET:
1910                 inetaddr = (struct sockaddr_in *)nam;
1911                 if (inetaddr->sin_family == AF_INET &&
1912                     inetaddr->sin_addr.s_addr == haddr->had_inetaddr)
1913                         return (1);
1914                 break;
1915         default:
1916                 break;
1917         };
1918         return (0);
1919 }
1920
1921 static nfsuint64 nfs_nullcookie = { { 0, 0 } };
1922 /*
1923  * This function finds the directory cookie that corresponds to the
1924  * logical byte offset given.
1925  */
1926 nfsuint64 *
1927 nfs_getcookie(struct nfsnode *np, off_t off, int add)
1928 {
1929         struct nfsdmap *dp, *dp2;
1930         int pos;
1931
1932         pos = (uoff_t)off / NFS_DIRBLKSIZ;
1933         if (pos == 0 || off < 0) {
1934 #ifdef DIAGNOSTIC
1935                 if (add)
1936                         panic("nfs getcookie add at <= 0");
1937 #endif
1938                 return (&nfs_nullcookie);
1939         }
1940         pos--;
1941         dp = np->n_cookies.lh_first;
1942         if (!dp) {
1943                 if (add) {
1944                         MALLOC(dp, struct nfsdmap *, sizeof (struct nfsdmap),
1945                                 M_NFSDIROFF, M_WAITOK);
1946                         dp->ndm_eocookie = 0;
1947                         LIST_INSERT_HEAD(&np->n_cookies, dp, ndm_list);
1948                 } else
1949                         return ((nfsuint64 *)0);
1950         }
1951         while (pos >= NFSNUMCOOKIES) {
1952                 pos -= NFSNUMCOOKIES;
1953                 if (dp->ndm_list.le_next) {
1954                         if (!add && dp->ndm_eocookie < NFSNUMCOOKIES &&
1955                                 pos >= dp->ndm_eocookie)
1956                                 return ((nfsuint64 *)0);
1957                         dp = dp->ndm_list.le_next;
1958                 } else if (add) {
1959                         MALLOC(dp2, struct nfsdmap *, sizeof (struct nfsdmap),
1960                                 M_NFSDIROFF, M_WAITOK);
1961                         dp2->ndm_eocookie = 0;
1962                         LIST_INSERT_AFTER(dp, dp2, ndm_list);
1963                         dp = dp2;
1964                 } else
1965                         return ((nfsuint64 *)0);
1966         }
1967         if (pos >= dp->ndm_eocookie) {
1968                 if (add)
1969                         dp->ndm_eocookie = pos + 1;
1970                 else
1971                         return ((nfsuint64 *)0);
1972         }
1973         return (&dp->ndm_cookies[pos]);
1974 }
1975
1976 /*
1977  * Invalidate cached directory information, except for the actual directory
1978  * blocks (which are invalidated separately).
1979  * Done mainly to avoid the use of stale offset cookies.
1980  */
1981 void
1982 nfs_invaldir(struct vnode *vp)
1983 {
1984         struct nfsnode *np = VTONFS(vp);
1985
1986 #ifdef DIAGNOSTIC
1987         if (vp->v_type != VDIR)
1988                 panic("nfs: invaldir not dir");
1989 #endif
1990         np->n_direofoffset = 0;
1991         np->n_cookieverf.nfsuquad[0] = 0;
1992         np->n_cookieverf.nfsuquad[1] = 0;
1993         if (np->n_cookies.lh_first)
1994                 np->n_cookies.lh_first->ndm_eocookie = 0;
1995 }
1996
1997 /*
1998  * Set the v_type field for an NFS client's vnode and initialize for
1999  * buffer cache operations if necessary.
2000  */
2001 void
2002 nfs_setvtype(struct vnode *vp, enum vtype vtyp)
2003 {
2004         vp->v_type = vtyp;
2005
2006         switch(vtyp) {
2007         case VREG:
2008         case VDIR:
2009         case VLNK:
2010                 vinitvmio(vp, 0);       /* needs VMIO, size not yet known */
2011                 break;
2012         default:
2013                 break;
2014         }
2015 }
2016
2017 /*
2018  * The write verifier has changed (probably due to a server reboot), so all
2019  * B_NEEDCOMMIT blocks will have to be written again. Since they are on the
2020  * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT
2021  * and B_CLUSTEROK flags.  Once done the new write verifier can be set for the
2022  * mount point.
2023  *
2024  * B_CLUSTEROK must be cleared along with B_NEEDCOMMIT because stage 1 data
2025  * writes are not clusterable.
2026  */
2027
2028 static int nfs_clearcommit_bp(struct buf *bp, void *data __unused);
2029
2030 void
2031 nfs_clearcommit(struct mount *mp)
2032 {
2033         struct vnode *vp, *nvp;
2034         lwkt_tokref ilock;
2035
2036         lwkt_gettoken(&ilock, &mntvnode_token);
2037         crit_enter();
2038         for (vp = TAILQ_FIRST(&mp->mnt_nvnodelist); vp; vp = nvp) {
2039                 nvp = TAILQ_NEXT(vp, v_nmntvnodes);     /* ZZZ */
2040                 RB_SCAN(buf_rb_tree, &vp->v_rbdirty_tree, NULL,
2041                         nfs_clearcommit_bp, NULL);
2042         }
2043         crit_exit();
2044         lwkt_reltoken(&ilock);
2045 }
2046
2047 static int
2048 nfs_clearcommit_bp(struct buf *bp, void *data __unused)
2049 {
2050         if (BUF_REFCNT(bp) == 0 &&
2051             (bp->b_flags & (B_DELWRI | B_NEEDCOMMIT))
2052              == (B_DELWRI | B_NEEDCOMMIT)) {
2053                 bp->b_flags &= ~(B_NEEDCOMMIT | B_CLUSTEROK);
2054         }
2055         return(0);
2056 }
2057
2058 #ifndef NFS_NOSERVER
2059 /*
2060  * Map errnos to NFS error numbers. For Version 3 also filter out error
2061  * numbers not specified for the associated procedure.
2062  */
2063 int
2064 nfsrv_errmap(struct nfsrv_descript *nd, int err)
2065 {
2066         short *defaulterrp, *errp;
2067
2068         if (nd->nd_flag & ND_NFSV3) {
2069             if (nd->nd_procnum <= NFSPROC_COMMIT) {
2070                 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum];
2071                 while (*++errp) {
2072                         if (*errp == err)
2073                                 return (err);
2074                         else if (*errp > err)
2075                                 break;
2076                 }
2077                 return ((int)*defaulterrp);
2078             } else
2079                 return (err & 0xffff);
2080         }
2081         if (err <= ELAST)
2082                 return ((int)nfsrv_v2errmap[err - 1]);
2083         return (NFSERR_IO);
2084 }
2085
2086 /*
2087  * Sort the group list in increasing numerical order.
2088  * (Insertion sort by Chris Torek, who was grossed out by the bubble sort
2089  *  that used to be here.)
2090  */
2091 void
2092 nfsrvw_sort(gid_t *list, int num)
2093 {
2094         int i, j;
2095         gid_t v;
2096
2097         /* Insertion sort. */
2098         for (i = 1; i < num; i++) {
2099                 v = list[i];
2100                 /* find correct slot for value v, moving others up */
2101                 for (j = i; --j >= 0 && v < list[j];)
2102                         list[j + 1] = list[j];
2103                 list[j + 1] = v;
2104         }
2105 }
2106
2107 /*
2108  * copy credentials making sure that the result can be compared with bcmp().
2109  */
2110 void
2111 nfsrv_setcred(struct ucred *incred, struct ucred *outcred)
2112 {
2113         int i;
2114
2115         bzero((caddr_t)outcred, sizeof (struct ucred));
2116         outcred->cr_ref = 1;
2117         outcred->cr_uid = incred->cr_uid;
2118         outcred->cr_ngroups = incred->cr_ngroups;
2119         for (i = 0; i < incred->cr_ngroups; i++)
2120                 outcred->cr_groups[i] = incred->cr_groups[i];
2121         nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups);
2122 }
2123 #endif /* NFS_NOSERVER */