nfs_subs.c revision 1.201.4.2
1 /* $NetBSD: nfs_subs.c,v 1.201.4.2 2008/05/16 02:25:49 yamt 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. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95 35 */ 36 37 /* 38 * Copyright 2000 Wasabi Systems, Inc. 39 * All rights reserved. 40 * 41 * Written by Frank van der Linden for Wasabi Systems, Inc. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 3. All advertising materials mentioning features or use of this software 52 * must display the following acknowledgement: 53 * This product includes software developed for the NetBSD Project by 54 * Wasabi Systems, Inc. 55 * 4. The name of Wasabi Systems, Inc. may not be used to endorse 56 * or promote products derived from this software without specific prior 57 * written permission. 58 * 59 * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND 60 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 61 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 62 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC 63 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 64 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 65 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 66 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 67 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 68 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 69 * POSSIBILITY OF SUCH DAMAGE. 70 */ 71 72 #include <sys/cdefs.h> 73 __KERNEL_RCSID(0, "$NetBSD: nfs_subs.c,v 1.201.4.2 2008/05/16 02:25:49 yamt Exp $"); 74 75 #include "fs_nfs.h" 76 #include "opt_nfs.h" 77 #include "opt_nfsserver.h" 78 #include "opt_iso.h" 79 #include "opt_inet.h" 80 81 /* 82 * These functions support the macros and help fiddle mbuf chains for 83 * the nfs op functions. They do things like create the rpc header and 84 * copy data between mbuf chains and uio lists. 85 */ 86 #include <sys/param.h> 87 #include <sys/proc.h> 88 #include <sys/systm.h> 89 #include <sys/kernel.h> 90 #include <sys/kmem.h> 91 #include <sys/mount.h> 92 #include <sys/vnode.h> 93 #include <sys/namei.h> 94 #include <sys/mbuf.h> 95 #include <sys/socket.h> 96 #include <sys/stat.h> 97 #include <sys/filedesc.h> 98 #include <sys/time.h> 99 #include <sys/dirent.h> 100 #include <sys/once.h> 101 #include <sys/kauth.h> 102 103 #include <uvm/uvm_extern.h> 104 105 #include <nfs/rpcv2.h> 106 #include <nfs/nfsproto.h> 107 #include <nfs/nfsnode.h> 108 #include <nfs/nfs.h> 109 #include <nfs/xdr_subs.h> 110 #include <nfs/nfsm_subs.h> 111 #include <nfs/nfsmount.h> 112 #include <nfs/nfsrtt.h> 113 #include <nfs/nfs_var.h> 114 115 #include <miscfs/specfs/specdev.h> 116 117 #include <netinet/in.h> 118 #ifdef ISO 119 #include <netiso/iso.h> 120 #endif 121 122 /* 123 * Data items converted to xdr at startup, since they are constant 124 * This is kinda hokey, but may save a little time doing byte swaps 125 */ 126 u_int32_t nfs_xdrneg1; 127 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 128 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 129 rpc_auth_kerb; 130 u_int32_t nfs_prog, nfs_true, nfs_false; 131 132 /* And other global data */ 133 const nfstype nfsv2_type[9] = 134 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON }; 135 const nfstype nfsv3_type[9] = 136 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON }; 137 const enum vtype nv2tov_type[8] = 138 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 139 const enum vtype nv3tov_type[8] = 140 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 141 int nfs_ticks; 142 int nfs_commitsize; 143 144 MALLOC_DEFINE(M_NFSDIROFF, "NFS diroff", "NFS directory cookies"); 145 146 /* NFS client/server stats. */ 147 struct nfsstats nfsstats; 148 149 /* 150 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 151 */ 152 const int nfsv3_procid[NFS_NPROCS] = { 153 NFSPROC_NULL, 154 NFSPROC_GETATTR, 155 NFSPROC_SETATTR, 156 NFSPROC_NOOP, 157 NFSPROC_LOOKUP, 158 NFSPROC_READLINK, 159 NFSPROC_READ, 160 NFSPROC_NOOP, 161 NFSPROC_WRITE, 162 NFSPROC_CREATE, 163 NFSPROC_REMOVE, 164 NFSPROC_RENAME, 165 NFSPROC_LINK, 166 NFSPROC_SYMLINK, 167 NFSPROC_MKDIR, 168 NFSPROC_RMDIR, 169 NFSPROC_READDIR, 170 NFSPROC_FSSTAT, 171 NFSPROC_NOOP, 172 NFSPROC_NOOP, 173 NFSPROC_NOOP, 174 NFSPROC_NOOP, 175 NFSPROC_NOOP 176 }; 177 178 /* 179 * and the reverse mapping from generic to Version 2 procedure numbers 180 */ 181 const int nfsv2_procid[NFS_NPROCS] = { 182 NFSV2PROC_NULL, 183 NFSV2PROC_GETATTR, 184 NFSV2PROC_SETATTR, 185 NFSV2PROC_LOOKUP, 186 NFSV2PROC_NOOP, 187 NFSV2PROC_READLINK, 188 NFSV2PROC_READ, 189 NFSV2PROC_WRITE, 190 NFSV2PROC_CREATE, 191 NFSV2PROC_MKDIR, 192 NFSV2PROC_SYMLINK, 193 NFSV2PROC_CREATE, 194 NFSV2PROC_REMOVE, 195 NFSV2PROC_RMDIR, 196 NFSV2PROC_RENAME, 197 NFSV2PROC_LINK, 198 NFSV2PROC_READDIR, 199 NFSV2PROC_NOOP, 200 NFSV2PROC_STATFS, 201 NFSV2PROC_NOOP, 202 NFSV2PROC_NOOP, 203 NFSV2PROC_NOOP, 204 NFSV2PROC_NOOP, 205 }; 206 207 /* 208 * Maps errno values to nfs error numbers. 209 * Use NFSERR_IO as the catch all for ones not specifically defined in 210 * RFC 1094. 211 */ 212 static const u_char nfsrv_v2errmap[ELAST] = { 213 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 214 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 215 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 216 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 217 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 218 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 219 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 220 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 221 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 222 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 223 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 224 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 225 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, 226 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 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, 230 }; 231 232 /* 233 * Maps errno values to nfs error numbers. 234 * Although it is not obvious whether or not NFS clients really care if 235 * a returned error value is in the specified list for the procedure, the 236 * safest thing to do is filter them appropriately. For Version 2, the 237 * X/Open XNFS document is the only specification that defines error values 238 * for each RPC (The RFC simply lists all possible error values for all RPCs), 239 * so I have decided to not do this for Version 2. 240 * The first entry is the default error return and the rest are the valid 241 * errors for that RPC in increasing numeric order. 242 */ 243 static const short nfsv3err_null[] = { 244 0, 245 0, 246 }; 247 248 static const short nfsv3err_getattr[] = { 249 NFSERR_IO, 250 NFSERR_IO, 251 NFSERR_STALE, 252 NFSERR_BADHANDLE, 253 NFSERR_SERVERFAULT, 254 0, 255 }; 256 257 static const short nfsv3err_setattr[] = { 258 NFSERR_IO, 259 NFSERR_PERM, 260 NFSERR_IO, 261 NFSERR_ACCES, 262 NFSERR_INVAL, 263 NFSERR_NOSPC, 264 NFSERR_ROFS, 265 NFSERR_DQUOT, 266 NFSERR_STALE, 267 NFSERR_BADHANDLE, 268 NFSERR_NOT_SYNC, 269 NFSERR_SERVERFAULT, 270 0, 271 }; 272 273 static const short nfsv3err_lookup[] = { 274 NFSERR_IO, 275 NFSERR_NOENT, 276 NFSERR_IO, 277 NFSERR_ACCES, 278 NFSERR_NOTDIR, 279 NFSERR_NAMETOL, 280 NFSERR_STALE, 281 NFSERR_BADHANDLE, 282 NFSERR_SERVERFAULT, 283 0, 284 }; 285 286 static const short nfsv3err_access[] = { 287 NFSERR_IO, 288 NFSERR_IO, 289 NFSERR_STALE, 290 NFSERR_BADHANDLE, 291 NFSERR_SERVERFAULT, 292 0, 293 }; 294 295 static const short nfsv3err_readlink[] = { 296 NFSERR_IO, 297 NFSERR_IO, 298 NFSERR_ACCES, 299 NFSERR_INVAL, 300 NFSERR_STALE, 301 NFSERR_BADHANDLE, 302 NFSERR_NOTSUPP, 303 NFSERR_SERVERFAULT, 304 0, 305 }; 306 307 static const short nfsv3err_read[] = { 308 NFSERR_IO, 309 NFSERR_IO, 310 NFSERR_NXIO, 311 NFSERR_ACCES, 312 NFSERR_INVAL, 313 NFSERR_STALE, 314 NFSERR_BADHANDLE, 315 NFSERR_SERVERFAULT, 316 NFSERR_JUKEBOX, 317 0, 318 }; 319 320 static const short nfsv3err_write[] = { 321 NFSERR_IO, 322 NFSERR_IO, 323 NFSERR_ACCES, 324 NFSERR_INVAL, 325 NFSERR_FBIG, 326 NFSERR_NOSPC, 327 NFSERR_ROFS, 328 NFSERR_DQUOT, 329 NFSERR_STALE, 330 NFSERR_BADHANDLE, 331 NFSERR_SERVERFAULT, 332 NFSERR_JUKEBOX, 333 0, 334 }; 335 336 static const short nfsv3err_create[] = { 337 NFSERR_IO, 338 NFSERR_IO, 339 NFSERR_ACCES, 340 NFSERR_EXIST, 341 NFSERR_NOTDIR, 342 NFSERR_NOSPC, 343 NFSERR_ROFS, 344 NFSERR_NAMETOL, 345 NFSERR_DQUOT, 346 NFSERR_STALE, 347 NFSERR_BADHANDLE, 348 NFSERR_NOTSUPP, 349 NFSERR_SERVERFAULT, 350 0, 351 }; 352 353 static const short nfsv3err_mkdir[] = { 354 NFSERR_IO, 355 NFSERR_IO, 356 NFSERR_ACCES, 357 NFSERR_EXIST, 358 NFSERR_NOTDIR, 359 NFSERR_NOSPC, 360 NFSERR_ROFS, 361 NFSERR_NAMETOL, 362 NFSERR_DQUOT, 363 NFSERR_STALE, 364 NFSERR_BADHANDLE, 365 NFSERR_NOTSUPP, 366 NFSERR_SERVERFAULT, 367 0, 368 }; 369 370 static const short nfsv3err_symlink[] = { 371 NFSERR_IO, 372 NFSERR_IO, 373 NFSERR_ACCES, 374 NFSERR_EXIST, 375 NFSERR_NOTDIR, 376 NFSERR_NOSPC, 377 NFSERR_ROFS, 378 NFSERR_NAMETOL, 379 NFSERR_DQUOT, 380 NFSERR_STALE, 381 NFSERR_BADHANDLE, 382 NFSERR_NOTSUPP, 383 NFSERR_SERVERFAULT, 384 0, 385 }; 386 387 static const short nfsv3err_mknod[] = { 388 NFSERR_IO, 389 NFSERR_IO, 390 NFSERR_ACCES, 391 NFSERR_EXIST, 392 NFSERR_NOTDIR, 393 NFSERR_NOSPC, 394 NFSERR_ROFS, 395 NFSERR_NAMETOL, 396 NFSERR_DQUOT, 397 NFSERR_STALE, 398 NFSERR_BADHANDLE, 399 NFSERR_NOTSUPP, 400 NFSERR_SERVERFAULT, 401 NFSERR_BADTYPE, 402 0, 403 }; 404 405 static const short nfsv3err_remove[] = { 406 NFSERR_IO, 407 NFSERR_NOENT, 408 NFSERR_IO, 409 NFSERR_ACCES, 410 NFSERR_NOTDIR, 411 NFSERR_ROFS, 412 NFSERR_NAMETOL, 413 NFSERR_STALE, 414 NFSERR_BADHANDLE, 415 NFSERR_SERVERFAULT, 416 0, 417 }; 418 419 static const short nfsv3err_rmdir[] = { 420 NFSERR_IO, 421 NFSERR_NOENT, 422 NFSERR_IO, 423 NFSERR_ACCES, 424 NFSERR_EXIST, 425 NFSERR_NOTDIR, 426 NFSERR_INVAL, 427 NFSERR_ROFS, 428 NFSERR_NAMETOL, 429 NFSERR_NOTEMPTY, 430 NFSERR_STALE, 431 NFSERR_BADHANDLE, 432 NFSERR_NOTSUPP, 433 NFSERR_SERVERFAULT, 434 0, 435 }; 436 437 static const short nfsv3err_rename[] = { 438 NFSERR_IO, 439 NFSERR_NOENT, 440 NFSERR_IO, 441 NFSERR_ACCES, 442 NFSERR_EXIST, 443 NFSERR_XDEV, 444 NFSERR_NOTDIR, 445 NFSERR_ISDIR, 446 NFSERR_INVAL, 447 NFSERR_NOSPC, 448 NFSERR_ROFS, 449 NFSERR_MLINK, 450 NFSERR_NAMETOL, 451 NFSERR_NOTEMPTY, 452 NFSERR_DQUOT, 453 NFSERR_STALE, 454 NFSERR_BADHANDLE, 455 NFSERR_NOTSUPP, 456 NFSERR_SERVERFAULT, 457 0, 458 }; 459 460 static const short nfsv3err_link[] = { 461 NFSERR_IO, 462 NFSERR_IO, 463 NFSERR_ACCES, 464 NFSERR_EXIST, 465 NFSERR_XDEV, 466 NFSERR_NOTDIR, 467 NFSERR_INVAL, 468 NFSERR_NOSPC, 469 NFSERR_ROFS, 470 NFSERR_MLINK, 471 NFSERR_NAMETOL, 472 NFSERR_DQUOT, 473 NFSERR_STALE, 474 NFSERR_BADHANDLE, 475 NFSERR_NOTSUPP, 476 NFSERR_SERVERFAULT, 477 0, 478 }; 479 480 static const short nfsv3err_readdir[] = { 481 NFSERR_IO, 482 NFSERR_IO, 483 NFSERR_ACCES, 484 NFSERR_NOTDIR, 485 NFSERR_STALE, 486 NFSERR_BADHANDLE, 487 NFSERR_BAD_COOKIE, 488 NFSERR_TOOSMALL, 489 NFSERR_SERVERFAULT, 490 0, 491 }; 492 493 static const short nfsv3err_readdirplus[] = { 494 NFSERR_IO, 495 NFSERR_IO, 496 NFSERR_ACCES, 497 NFSERR_NOTDIR, 498 NFSERR_STALE, 499 NFSERR_BADHANDLE, 500 NFSERR_BAD_COOKIE, 501 NFSERR_NOTSUPP, 502 NFSERR_TOOSMALL, 503 NFSERR_SERVERFAULT, 504 0, 505 }; 506 507 static const short nfsv3err_fsstat[] = { 508 NFSERR_IO, 509 NFSERR_IO, 510 NFSERR_STALE, 511 NFSERR_BADHANDLE, 512 NFSERR_SERVERFAULT, 513 0, 514 }; 515 516 static const short nfsv3err_fsinfo[] = { 517 NFSERR_STALE, 518 NFSERR_STALE, 519 NFSERR_BADHANDLE, 520 NFSERR_SERVERFAULT, 521 0, 522 }; 523 524 static const short nfsv3err_pathconf[] = { 525 NFSERR_STALE, 526 NFSERR_STALE, 527 NFSERR_BADHANDLE, 528 NFSERR_SERVERFAULT, 529 0, 530 }; 531 532 static const short nfsv3err_commit[] = { 533 NFSERR_IO, 534 NFSERR_IO, 535 NFSERR_STALE, 536 NFSERR_BADHANDLE, 537 NFSERR_SERVERFAULT, 538 0, 539 }; 540 541 static const short * const nfsrv_v3errmap[] = { 542 nfsv3err_null, 543 nfsv3err_getattr, 544 nfsv3err_setattr, 545 nfsv3err_lookup, 546 nfsv3err_access, 547 nfsv3err_readlink, 548 nfsv3err_read, 549 nfsv3err_write, 550 nfsv3err_create, 551 nfsv3err_mkdir, 552 nfsv3err_symlink, 553 nfsv3err_mknod, 554 nfsv3err_remove, 555 nfsv3err_rmdir, 556 nfsv3err_rename, 557 nfsv3err_link, 558 nfsv3err_readdir, 559 nfsv3err_readdirplus, 560 nfsv3err_fsstat, 561 nfsv3err_fsinfo, 562 nfsv3err_pathconf, 563 nfsv3err_commit, 564 }; 565 566 extern struct vfs_hooks nfs_export_hooks; 567 extern struct nfsrtt nfsrtt; 568 extern struct nfsnodehashhead *nfsnodehashtbl; 569 extern u_long nfsnodehash; 570 571 u_long nfsdirhashmask; 572 573 static kmutex_t nfs_xidlock; 574 575 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 576 577 /* 578 * Create the header for an rpc request packet 579 * The hsiz is the size of the rest of the nfs request header. 580 * (just used to decide if a cluster is a good idea) 581 */ 582 struct mbuf * 583 nfsm_reqh(struct nfsnode *np, u_long procid, int hsiz, char **bposp) 584 { 585 struct mbuf *mb; 586 char *bpos; 587 588 mb = m_get(M_WAIT, MT_DATA); 589 MCLAIM(mb, &nfs_mowner); 590 if (hsiz >= MINCLSIZE) 591 m_clget(mb, M_WAIT); 592 mb->m_len = 0; 593 bpos = mtod(mb, void *); 594 595 /* Finally, return values */ 596 *bposp = bpos; 597 return (mb); 598 } 599 600 /* 601 * Build the RPC header and fill in the authorization info. 602 * The authorization string argument is only used when the credentials 603 * come from outside of the kernel. 604 * Returns the head of the mbuf list. 605 */ 606 struct mbuf * 607 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 608 verf_str, mrest, mrest_len, mbp, xidp) 609 kauth_cred_t cr; 610 int nmflag; 611 int procid; 612 int auth_type; 613 int auth_len; 614 char *auth_str; 615 int verf_len; 616 char *verf_str; 617 struct mbuf *mrest; 618 int mrest_len; 619 struct mbuf **mbp; 620 u_int32_t *xidp; 621 { 622 struct mbuf *mb; 623 u_int32_t *tl; 624 char *bpos; 625 int i; 626 struct mbuf *mreq; 627 int siz, grpsiz, authsiz; 628 629 authsiz = nfsm_rndup(auth_len); 630 mb = m_gethdr(M_WAIT, MT_DATA); 631 MCLAIM(mb, &nfs_mowner); 632 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 633 m_clget(mb, M_WAIT); 634 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 635 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 636 } else { 637 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 638 } 639 mb->m_len = 0; 640 mreq = mb; 641 bpos = mtod(mb, void *); 642 643 /* 644 * First the RPC header. 645 */ 646 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 647 648 *tl++ = *xidp = nfs_getxid(); 649 *tl++ = rpc_call; 650 *tl++ = rpc_vers; 651 *tl++ = txdr_unsigned(NFS_PROG); 652 if (nmflag & NFSMNT_NFSV3) 653 *tl++ = txdr_unsigned(NFS_VER3); 654 else 655 *tl++ = txdr_unsigned(NFS_VER2); 656 if (nmflag & NFSMNT_NFSV3) 657 *tl++ = txdr_unsigned(procid); 658 else 659 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 660 661 /* 662 * And then the authorization cred. 663 */ 664 *tl++ = txdr_unsigned(auth_type); 665 *tl = txdr_unsigned(authsiz); 666 switch (auth_type) { 667 case RPCAUTH_UNIX: 668 nfsm_build(tl, u_int32_t *, auth_len); 669 *tl++ = 0; /* stamp ?? */ 670 *tl++ = 0; /* NULL hostname */ 671 *tl++ = txdr_unsigned(kauth_cred_geteuid(cr)); 672 *tl++ = txdr_unsigned(kauth_cred_getegid(cr)); 673 grpsiz = (auth_len >> 2) - 5; 674 *tl++ = txdr_unsigned(grpsiz); 675 for (i = 0; i < grpsiz; i++) 676 *tl++ = txdr_unsigned(kauth_cred_group(cr, i)); /* XXX elad review */ 677 break; 678 case RPCAUTH_KERB4: 679 siz = auth_len; 680 while (siz > 0) { 681 if (M_TRAILINGSPACE(mb) == 0) { 682 struct mbuf *mb2; 683 mb2 = m_get(M_WAIT, MT_DATA); 684 MCLAIM(mb2, &nfs_mowner); 685 if (siz >= MINCLSIZE) 686 m_clget(mb2, M_WAIT); 687 mb->m_next = mb2; 688 mb = mb2; 689 mb->m_len = 0; 690 bpos = mtod(mb, void *); 691 } 692 i = min(siz, M_TRAILINGSPACE(mb)); 693 memcpy(bpos, auth_str, i); 694 mb->m_len += i; 695 auth_str += i; 696 bpos += i; 697 siz -= i; 698 } 699 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 700 for (i = 0; i < siz; i++) 701 *bpos++ = '\0'; 702 mb->m_len += siz; 703 } 704 break; 705 }; 706 707 /* 708 * And the verifier... 709 */ 710 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 711 if (verf_str) { 712 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 713 *tl = txdr_unsigned(verf_len); 714 siz = verf_len; 715 while (siz > 0) { 716 if (M_TRAILINGSPACE(mb) == 0) { 717 struct mbuf *mb2; 718 mb2 = m_get(M_WAIT, MT_DATA); 719 MCLAIM(mb2, &nfs_mowner); 720 if (siz >= MINCLSIZE) 721 m_clget(mb2, M_WAIT); 722 mb->m_next = mb2; 723 mb = mb2; 724 mb->m_len = 0; 725 bpos = mtod(mb, void *); 726 } 727 i = min(siz, M_TRAILINGSPACE(mb)); 728 memcpy(bpos, verf_str, i); 729 mb->m_len += i; 730 verf_str += i; 731 bpos += i; 732 siz -= i; 733 } 734 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 735 for (i = 0; i < siz; i++) 736 *bpos++ = '\0'; 737 mb->m_len += siz; 738 } 739 } else { 740 *tl++ = txdr_unsigned(RPCAUTH_NULL); 741 *tl = 0; 742 } 743 mb->m_next = mrest; 744 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 745 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 746 *mbp = mb; 747 return (mreq); 748 } 749 750 /* 751 * copies mbuf chain to the uio scatter/gather list 752 */ 753 int 754 nfsm_mbuftouio(mrep, uiop, siz, dpos) 755 struct mbuf **mrep; 756 struct uio *uiop; 757 int siz; 758 char **dpos; 759 { 760 char *mbufcp, *uiocp; 761 int xfer, left, len; 762 struct mbuf *mp; 763 long uiosiz, rem; 764 int error = 0; 765 766 mp = *mrep; 767 mbufcp = *dpos; 768 len = mtod(mp, char *) + mp->m_len - mbufcp; 769 rem = nfsm_rndup(siz)-siz; 770 while (siz > 0) { 771 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 772 return (EFBIG); 773 left = uiop->uio_iov->iov_len; 774 uiocp = uiop->uio_iov->iov_base; 775 if (left > siz) 776 left = siz; 777 uiosiz = left; 778 while (left > 0) { 779 while (len == 0) { 780 mp = mp->m_next; 781 if (mp == NULL) 782 return (EBADRPC); 783 mbufcp = mtod(mp, void *); 784 len = mp->m_len; 785 } 786 xfer = (left > len) ? len : left; 787 error = copyout_vmspace(uiop->uio_vmspace, mbufcp, 788 uiocp, xfer); 789 if (error) { 790 return error; 791 } 792 left -= xfer; 793 len -= xfer; 794 mbufcp += xfer; 795 uiocp += xfer; 796 uiop->uio_offset += xfer; 797 uiop->uio_resid -= xfer; 798 } 799 if (uiop->uio_iov->iov_len <= siz) { 800 uiop->uio_iovcnt--; 801 uiop->uio_iov++; 802 } else { 803 uiop->uio_iov->iov_base = 804 (char *)uiop->uio_iov->iov_base + uiosiz; 805 uiop->uio_iov->iov_len -= uiosiz; 806 } 807 siz -= uiosiz; 808 } 809 *dpos = mbufcp; 810 *mrep = mp; 811 if (rem > 0) { 812 if (len < rem) 813 error = nfs_adv(mrep, dpos, rem, len); 814 else 815 *dpos += rem; 816 } 817 return (error); 818 } 819 820 /* 821 * copies a uio scatter/gather list to an mbuf chain. 822 * NOTE: can ony handle iovcnt == 1 823 */ 824 int 825 nfsm_uiotombuf(uiop, mq, siz, bpos) 826 struct uio *uiop; 827 struct mbuf **mq; 828 int siz; 829 char **bpos; 830 { 831 char *uiocp; 832 struct mbuf *mp, *mp2; 833 int xfer, left, mlen; 834 int uiosiz, clflg, rem; 835 char *cp; 836 int error; 837 838 #ifdef DIAGNOSTIC 839 if (uiop->uio_iovcnt != 1) 840 panic("nfsm_uiotombuf: iovcnt != 1"); 841 #endif 842 843 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 844 clflg = 1; 845 else 846 clflg = 0; 847 rem = nfsm_rndup(siz)-siz; 848 mp = mp2 = *mq; 849 while (siz > 0) { 850 left = uiop->uio_iov->iov_len; 851 uiocp = uiop->uio_iov->iov_base; 852 if (left > siz) 853 left = siz; 854 uiosiz = left; 855 while (left > 0) { 856 mlen = M_TRAILINGSPACE(mp); 857 if (mlen == 0) { 858 mp = m_get(M_WAIT, MT_DATA); 859 MCLAIM(mp, &nfs_mowner); 860 if (clflg) 861 m_clget(mp, M_WAIT); 862 mp->m_len = 0; 863 mp2->m_next = mp; 864 mp2 = mp; 865 mlen = M_TRAILINGSPACE(mp); 866 } 867 xfer = (left > mlen) ? mlen : left; 868 cp = mtod(mp, char *) + mp->m_len; 869 error = copyin_vmspace(uiop->uio_vmspace, uiocp, cp, 870 xfer); 871 if (error) { 872 /* XXX */ 873 } 874 mp->m_len += xfer; 875 left -= xfer; 876 uiocp += xfer; 877 uiop->uio_offset += xfer; 878 uiop->uio_resid -= xfer; 879 } 880 uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base + 881 uiosiz; 882 uiop->uio_iov->iov_len -= uiosiz; 883 siz -= uiosiz; 884 } 885 if (rem > 0) { 886 if (rem > M_TRAILINGSPACE(mp)) { 887 mp = m_get(M_WAIT, MT_DATA); 888 MCLAIM(mp, &nfs_mowner); 889 mp->m_len = 0; 890 mp2->m_next = mp; 891 } 892 cp = mtod(mp, char *) + mp->m_len; 893 for (left = 0; left < rem; left++) 894 *cp++ = '\0'; 895 mp->m_len += rem; 896 *bpos = cp; 897 } else 898 *bpos = mtod(mp, char *) + mp->m_len; 899 *mq = mp; 900 return (0); 901 } 902 903 /* 904 * Get at least "siz" bytes of correctly aligned data. 905 * When called the mbuf pointers are not necessarily correct, 906 * dsosp points to what ought to be in m_data and left contains 907 * what ought to be in m_len. 908 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 909 * cases. (The macros use the vars. dpos and dpos2) 910 */ 911 int 912 nfsm_disct(mdp, dposp, siz, left, cp2) 913 struct mbuf **mdp; 914 char **dposp; 915 int siz; 916 int left; 917 char **cp2; 918 { 919 struct mbuf *m1, *m2; 920 struct mbuf *havebuf = NULL; 921 char *src = *dposp; 922 char *dst; 923 int len; 924 925 #ifdef DEBUG 926 if (left < 0) 927 panic("nfsm_disct: left < 0"); 928 #endif 929 m1 = *mdp; 930 /* 931 * Skip through the mbuf chain looking for an mbuf with 932 * some data. If the first mbuf found has enough data 933 * and it is correctly aligned return it. 934 */ 935 while (left == 0) { 936 havebuf = m1; 937 *mdp = m1 = m1->m_next; 938 if (m1 == NULL) 939 return (EBADRPC); 940 src = mtod(m1, void *); 941 left = m1->m_len; 942 /* 943 * If we start a new mbuf and it is big enough 944 * and correctly aligned just return it, don't 945 * do any pull up. 946 */ 947 if (left >= siz && nfsm_aligned(src)) { 948 *cp2 = src; 949 *dposp = src + siz; 950 return (0); 951 } 952 } 953 if ((m1->m_flags & M_EXT) != 0) { 954 if (havebuf && M_TRAILINGSPACE(havebuf) >= siz && 955 nfsm_aligned(mtod(havebuf, char *) + havebuf->m_len)) { 956 /* 957 * If the first mbuf with data has external data 958 * and there is a previous mbuf with some trailing 959 * space, use it to move the data into. 960 */ 961 m2 = m1; 962 *mdp = m1 = havebuf; 963 *cp2 = mtod(m1, char *) + m1->m_len; 964 } else if (havebuf) { 965 /* 966 * If the first mbuf has a external data 967 * and there is no previous empty mbuf 968 * allocate a new mbuf and move the external 969 * data to the new mbuf. Also make the first 970 * mbuf look empty. 971 */ 972 m2 = m1; 973 *mdp = m1 = m_get(M_WAIT, MT_DATA); 974 MCLAIM(m1, m2->m_owner); 975 if ((m2->m_flags & M_PKTHDR) != 0) { 976 /* XXX MOVE */ 977 M_COPY_PKTHDR(m1, m2); 978 m_tag_delete_chain(m2, NULL); 979 m2->m_flags &= ~M_PKTHDR; 980 } 981 if (havebuf) { 982 havebuf->m_next = m1; 983 } 984 m1->m_next = m2; 985 MRESETDATA(m1); 986 m1->m_len = 0; 987 m2->m_data = src; 988 m2->m_len = left; 989 *cp2 = mtod(m1, char *); 990 } else { 991 struct mbuf **nextp = &m1->m_next; 992 993 m1->m_len -= left; 994 do { 995 m2 = m_get(M_WAIT, MT_DATA); 996 MCLAIM(m2, m1->m_owner); 997 if (left >= MINCLSIZE) { 998 MCLGET(m2, M_WAIT); 999 } 1000 m2->m_next = *nextp; 1001 *nextp = m2; 1002 nextp = &m2->m_next; 1003 len = (m2->m_flags & M_EXT) != 0 ? 1004 MCLBYTES : MLEN; 1005 if (len > left) { 1006 len = left; 1007 } 1008 memcpy(mtod(m2, char *), src, len); 1009 m2->m_len = len; 1010 src += len; 1011 left -= len; 1012 } while (left > 0); 1013 *mdp = m1 = m1->m_next; 1014 m2 = m1->m_next; 1015 *cp2 = mtod(m1, char *); 1016 } 1017 } else { 1018 /* 1019 * If the first mbuf has no external data 1020 * move the data to the front of the mbuf. 1021 */ 1022 MRESETDATA(m1); 1023 dst = mtod(m1, char *); 1024 if (dst != src) { 1025 memmove(dst, src, left); 1026 } 1027 m1->m_len = left; 1028 m2 = m1->m_next; 1029 *cp2 = m1->m_data; 1030 } 1031 *dposp = *cp2 + siz; 1032 /* 1033 * Loop through mbufs pulling data up into first mbuf until 1034 * the first mbuf is full or there is no more data to 1035 * pullup. 1036 */ 1037 dst = mtod(m1, char *) + m1->m_len; 1038 while ((len = M_TRAILINGSPACE(m1)) != 0 && m2) { 1039 if ((len = min(len, m2->m_len)) != 0) { 1040 memcpy(dst, mtod(m2, char *), len); 1041 } 1042 m1->m_len += len; 1043 dst += len; 1044 m2->m_data += len; 1045 m2->m_len -= len; 1046 m2 = m2->m_next; 1047 } 1048 if (m1->m_len < siz) 1049 return (EBADRPC); 1050 return (0); 1051 } 1052 1053 /* 1054 * Advance the position in the mbuf chain. 1055 */ 1056 int 1057 nfs_adv(mdp, dposp, offs, left) 1058 struct mbuf **mdp; 1059 char **dposp; 1060 int offs; 1061 int left; 1062 { 1063 struct mbuf *m; 1064 int s; 1065 1066 m = *mdp; 1067 s = left; 1068 while (s < offs) { 1069 offs -= s; 1070 m = m->m_next; 1071 if (m == NULL) 1072 return (EBADRPC); 1073 s = m->m_len; 1074 } 1075 *mdp = m; 1076 *dposp = mtod(m, char *) + offs; 1077 return (0); 1078 } 1079 1080 /* 1081 * Copy a string into mbufs for the hard cases... 1082 */ 1083 int 1084 nfsm_strtmbuf(mb, bpos, cp, siz) 1085 struct mbuf **mb; 1086 char **bpos; 1087 const char *cp; 1088 long siz; 1089 { 1090 struct mbuf *m1 = NULL, *m2; 1091 long left, xfer, len, tlen; 1092 u_int32_t *tl; 1093 int putsize; 1094 1095 putsize = 1; 1096 m2 = *mb; 1097 left = M_TRAILINGSPACE(m2); 1098 if (left > 0) { 1099 tl = ((u_int32_t *)(*bpos)); 1100 *tl++ = txdr_unsigned(siz); 1101 putsize = 0; 1102 left -= NFSX_UNSIGNED; 1103 m2->m_len += NFSX_UNSIGNED; 1104 if (left > 0) { 1105 memcpy((void *) tl, cp, left); 1106 siz -= left; 1107 cp += left; 1108 m2->m_len += left; 1109 left = 0; 1110 } 1111 } 1112 /* Loop around adding mbufs */ 1113 while (siz > 0) { 1114 m1 = m_get(M_WAIT, MT_DATA); 1115 MCLAIM(m1, &nfs_mowner); 1116 if (siz > MLEN) 1117 m_clget(m1, M_WAIT); 1118 m1->m_len = NFSMSIZ(m1); 1119 m2->m_next = m1; 1120 m2 = m1; 1121 tl = mtod(m1, u_int32_t *); 1122 tlen = 0; 1123 if (putsize) { 1124 *tl++ = txdr_unsigned(siz); 1125 m1->m_len -= NFSX_UNSIGNED; 1126 tlen = NFSX_UNSIGNED; 1127 putsize = 0; 1128 } 1129 if (siz < m1->m_len) { 1130 len = nfsm_rndup(siz); 1131 xfer = siz; 1132 if (xfer < len) 1133 *(tl+(xfer>>2)) = 0; 1134 } else { 1135 xfer = len = m1->m_len; 1136 } 1137 memcpy((void *) tl, cp, xfer); 1138 m1->m_len = len+tlen; 1139 siz -= xfer; 1140 cp += xfer; 1141 } 1142 *mb = m1; 1143 *bpos = mtod(m1, char *) + m1->m_len; 1144 return (0); 1145 } 1146 1147 /* 1148 * Directory caching routines. They work as follows: 1149 * - a cache is maintained per VDIR nfsnode. 1150 * - for each offset cookie that is exported to userspace, and can 1151 * thus be thrown back at us as an offset to VOP_READDIR, store 1152 * information in the cache. 1153 * - cached are: 1154 * - cookie itself 1155 * - blocknumber (essentially just a search key in the buffer cache) 1156 * - entry number in block. 1157 * - offset cookie of block in which this entry is stored 1158 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1159 * - entries are looked up in a hash table 1160 * - also maintained is an LRU list of entries, used to determine 1161 * which ones to delete if the cache grows too large. 1162 * - if 32 <-> 64 translation mode is requested for a filesystem, 1163 * the cache also functions as a translation table 1164 * - in the translation case, invalidating the cache does not mean 1165 * flushing it, but just marking entries as invalid, except for 1166 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1167 * still be able to use the cache as a translation table. 1168 * - 32 bit cookies are uniquely created by combining the hash table 1169 * entry value, and one generation count per hash table entry, 1170 * incremented each time an entry is appended to the chain. 1171 * - the cache is invalidated each time a direcory is modified 1172 * - sanity checks are also done; if an entry in a block turns 1173 * out not to have a matching cookie, the cache is invalidated 1174 * and a new block starting from the wanted offset is fetched from 1175 * the server. 1176 * - directory entries as read from the server are extended to contain 1177 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1178 * the cache and exporting them to userspace through the cookie 1179 * argument to VOP_READDIR. 1180 */ 1181 1182 u_long 1183 nfs_dirhash(off) 1184 off_t off; 1185 { 1186 int i; 1187 char *cp = (char *)&off; 1188 u_long sum = 0L; 1189 1190 for (i = 0 ; i < sizeof (off); i++) 1191 sum += *cp++; 1192 1193 return sum; 1194 } 1195 1196 #define _NFSDC_MTX(np) (&NFSTOV(np)->v_interlock) 1197 #define NFSDC_LOCK(np) mutex_enter(_NFSDC_MTX(np)) 1198 #define NFSDC_UNLOCK(np) mutex_exit(_NFSDC_MTX(np)) 1199 #define NFSDC_ASSERT_LOCKED(np) KASSERT(mutex_owned(_NFSDC_MTX(np))) 1200 1201 void 1202 nfs_initdircache(vp) 1203 struct vnode *vp; 1204 { 1205 struct nfsnode *np = VTONFS(vp); 1206 struct nfsdirhashhead *dircache; 1207 1208 dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, true, 1209 &nfsdirhashmask); 1210 1211 NFSDC_LOCK(np); 1212 if (np->n_dircache == NULL) { 1213 np->n_dircachesize = 0; 1214 np->n_dircache = dircache; 1215 dircache = NULL; 1216 TAILQ_INIT(&np->n_dirchain); 1217 } 1218 NFSDC_UNLOCK(np); 1219 if (dircache) 1220 hashdone(dircache, HASH_LIST, nfsdirhashmask); 1221 } 1222 1223 void 1224 nfs_initdirxlatecookie(vp) 1225 struct vnode *vp; 1226 { 1227 struct nfsnode *np = VTONFS(vp); 1228 unsigned *dirgens; 1229 1230 KASSERT(VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_XLATECOOKIE); 1231 1232 dirgens = kmem_zalloc(NFS_DIRHASHSIZ * sizeof(unsigned), KM_SLEEP); 1233 NFSDC_LOCK(np); 1234 if (np->n_dirgens == NULL) { 1235 np->n_dirgens = dirgens; 1236 dirgens = NULL; 1237 } 1238 NFSDC_UNLOCK(np); 1239 if (dirgens) 1240 kmem_free(dirgens, NFS_DIRHASHSIZ * sizeof(unsigned)); 1241 } 1242 1243 static const struct nfsdircache dzero; 1244 1245 static void nfs_unlinkdircache __P((struct nfsnode *np, struct nfsdircache *)); 1246 static void nfs_putdircache_unlocked __P((struct nfsnode *, 1247 struct nfsdircache *)); 1248 1249 static void 1250 nfs_unlinkdircache(np, ndp) 1251 struct nfsnode *np; 1252 struct nfsdircache *ndp; 1253 { 1254 1255 NFSDC_ASSERT_LOCKED(np); 1256 KASSERT(ndp != &dzero); 1257 1258 if (LIST_NEXT(ndp, dc_hash) == (void *)-1) 1259 return; 1260 1261 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1262 LIST_REMOVE(ndp, dc_hash); 1263 LIST_NEXT(ndp, dc_hash) = (void *)-1; /* mark as unlinked */ 1264 1265 nfs_putdircache_unlocked(np, ndp); 1266 } 1267 1268 void 1269 nfs_putdircache(np, ndp) 1270 struct nfsnode *np; 1271 struct nfsdircache *ndp; 1272 { 1273 int ref; 1274 1275 if (ndp == &dzero) 1276 return; 1277 1278 KASSERT(ndp->dc_refcnt > 0); 1279 NFSDC_LOCK(np); 1280 ref = --ndp->dc_refcnt; 1281 NFSDC_UNLOCK(np); 1282 1283 if (ref == 0) 1284 kmem_free(ndp, sizeof(*ndp)); 1285 } 1286 1287 static void 1288 nfs_putdircache_unlocked(struct nfsnode *np, struct nfsdircache *ndp) 1289 { 1290 int ref; 1291 1292 NFSDC_ASSERT_LOCKED(np); 1293 1294 if (ndp == &dzero) 1295 return; 1296 1297 KASSERT(ndp->dc_refcnt > 0); 1298 ref = --ndp->dc_refcnt; 1299 if (ref == 0) 1300 kmem_free(ndp, sizeof(*ndp)); 1301 } 1302 1303 struct nfsdircache * 1304 nfs_searchdircache(vp, off, do32, hashent) 1305 struct vnode *vp; 1306 off_t off; 1307 int do32; 1308 int *hashent; 1309 { 1310 struct nfsdirhashhead *ndhp; 1311 struct nfsdircache *ndp = NULL; 1312 struct nfsnode *np = VTONFS(vp); 1313 unsigned ent; 1314 1315 /* 1316 * Zero is always a valid cookie. 1317 */ 1318 if (off == 0) 1319 /* XXXUNCONST */ 1320 return (struct nfsdircache *)__UNCONST(&dzero); 1321 1322 if (!np->n_dircache) 1323 return NULL; 1324 1325 /* 1326 * We use a 32bit cookie as search key, directly reconstruct 1327 * the hashentry. Else use the hashfunction. 1328 */ 1329 if (do32) { 1330 ent = (u_int32_t)off >> 24; 1331 if (ent >= NFS_DIRHASHSIZ) 1332 return NULL; 1333 ndhp = &np->n_dircache[ent]; 1334 } else { 1335 ndhp = NFSDIRHASH(np, off); 1336 } 1337 1338 if (hashent) 1339 *hashent = (int)(ndhp - np->n_dircache); 1340 1341 NFSDC_LOCK(np); 1342 if (do32) { 1343 LIST_FOREACH(ndp, ndhp, dc_hash) { 1344 if (ndp->dc_cookie32 == (u_int32_t)off) { 1345 /* 1346 * An invalidated entry will become the 1347 * start of a new block fetched from 1348 * the server. 1349 */ 1350 if (ndp->dc_flags & NFSDC_INVALID) { 1351 ndp->dc_blkcookie = ndp->dc_cookie; 1352 ndp->dc_entry = 0; 1353 ndp->dc_flags &= ~NFSDC_INVALID; 1354 } 1355 break; 1356 } 1357 } 1358 } else { 1359 LIST_FOREACH(ndp, ndhp, dc_hash) { 1360 if (ndp->dc_cookie == off) 1361 break; 1362 } 1363 } 1364 if (ndp != NULL) 1365 ndp->dc_refcnt++; 1366 NFSDC_UNLOCK(np); 1367 return ndp; 1368 } 1369 1370 1371 struct nfsdircache * 1372 nfs_enterdircache(struct vnode *vp, off_t off, off_t blkoff, int en, 1373 daddr_t blkno) 1374 { 1375 struct nfsnode *np = VTONFS(vp); 1376 struct nfsdirhashhead *ndhp; 1377 struct nfsdircache *ndp = NULL; 1378 struct nfsdircache *newndp = NULL; 1379 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1380 int hashent = 0, gen, overwrite; /* XXX: GCC */ 1381 1382 /* 1383 * XXX refuse entries for offset 0. amd(8) erroneously sets 1384 * cookie 0 for the '.' entry, making this necessary. This 1385 * isn't so bad, as 0 is a special case anyway. 1386 */ 1387 if (off == 0) 1388 /* XXXUNCONST */ 1389 return (struct nfsdircache *)__UNCONST(&dzero); 1390 1391 if (!np->n_dircache) 1392 /* 1393 * XXX would like to do this in nfs_nget but vtype 1394 * isn't known at that time. 1395 */ 1396 nfs_initdircache(vp); 1397 1398 if ((nmp->nm_flag & NFSMNT_XLATECOOKIE) && !np->n_dirgens) 1399 nfs_initdirxlatecookie(vp); 1400 1401 retry: 1402 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1403 1404 NFSDC_LOCK(np); 1405 if (ndp && (ndp->dc_flags & NFSDC_INVALID) == 0) { 1406 /* 1407 * Overwriting an old entry. Check if it's the same. 1408 * If so, just return. If not, remove the old entry. 1409 */ 1410 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1411 goto done; 1412 nfs_unlinkdircache(np, ndp); 1413 nfs_putdircache_unlocked(np, ndp); 1414 ndp = NULL; 1415 } 1416 1417 ndhp = &np->n_dircache[hashent]; 1418 1419 if (!ndp) { 1420 if (newndp == NULL) { 1421 NFSDC_UNLOCK(np); 1422 newndp = kmem_alloc(sizeof(*newndp), KM_SLEEP); 1423 newndp->dc_refcnt = 1; 1424 LIST_NEXT(newndp, dc_hash) = (void *)-1; 1425 goto retry; 1426 } 1427 ndp = newndp; 1428 newndp = NULL; 1429 overwrite = 0; 1430 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1431 /* 1432 * We're allocating a new entry, so bump the 1433 * generation number. 1434 */ 1435 KASSERT(np->n_dirgens); 1436 gen = ++np->n_dirgens[hashent]; 1437 if (gen == 0) { 1438 np->n_dirgens[hashent]++; 1439 gen++; 1440 } 1441 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1442 } 1443 } else 1444 overwrite = 1; 1445 1446 ndp->dc_cookie = off; 1447 ndp->dc_blkcookie = blkoff; 1448 ndp->dc_entry = en; 1449 ndp->dc_flags = 0; 1450 1451 if (overwrite) 1452 goto done; 1453 1454 /* 1455 * If the maximum directory cookie cache size has been reached 1456 * for this node, take one off the front. The idea is that 1457 * directories are typically read front-to-back once, so that 1458 * the oldest entries can be thrown away without much performance 1459 * loss. 1460 */ 1461 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1462 nfs_unlinkdircache(np, TAILQ_FIRST(&np->n_dirchain)); 1463 } else 1464 np->n_dircachesize++; 1465 1466 KASSERT(ndp->dc_refcnt == 1); 1467 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1468 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1469 ndp->dc_refcnt++; 1470 done: 1471 KASSERT(ndp->dc_refcnt > 0); 1472 NFSDC_UNLOCK(np); 1473 if (newndp) 1474 nfs_putdircache(np, newndp); 1475 return ndp; 1476 } 1477 1478 void 1479 nfs_invaldircache(vp, flags) 1480 struct vnode *vp; 1481 int flags; 1482 { 1483 struct nfsnode *np = VTONFS(vp); 1484 struct nfsdircache *ndp = NULL; 1485 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1486 const bool forcefree = flags & NFS_INVALDIRCACHE_FORCE; 1487 1488 #ifdef DIAGNOSTIC 1489 if (vp->v_type != VDIR) 1490 panic("nfs: invaldircache: not dir"); 1491 #endif 1492 1493 if ((flags & NFS_INVALDIRCACHE_KEEPEOF) == 0) 1494 np->n_flag &= ~NEOFVALID; 1495 1496 if (!np->n_dircache) 1497 return; 1498 1499 NFSDC_LOCK(np); 1500 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1501 while ((ndp = TAILQ_FIRST(&np->n_dirchain)) != NULL) { 1502 KASSERT(!forcefree || ndp->dc_refcnt == 1); 1503 nfs_unlinkdircache(np, ndp); 1504 } 1505 np->n_dircachesize = 0; 1506 if (forcefree && np->n_dirgens) { 1507 kmem_free(np->n_dirgens, 1508 NFS_DIRHASHSIZ * sizeof(unsigned)); 1509 np->n_dirgens = NULL; 1510 } 1511 } else { 1512 TAILQ_FOREACH(ndp, &np->n_dirchain, dc_chain) 1513 ndp->dc_flags |= NFSDC_INVALID; 1514 } 1515 1516 NFSDC_UNLOCK(np); 1517 } 1518 1519 /* 1520 * Called once before VFS init to initialize shared and 1521 * server-specific data structures. 1522 */ 1523 static int 1524 nfs_init0(void) 1525 { 1526 1527 nfsrtt.pos = 0; 1528 rpc_vers = txdr_unsigned(RPC_VER2); 1529 rpc_call = txdr_unsigned(RPC_CALL); 1530 rpc_reply = txdr_unsigned(RPC_REPLY); 1531 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1532 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1533 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1534 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1535 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1536 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1537 nfs_prog = txdr_unsigned(NFS_PROG); 1538 nfs_true = txdr_unsigned(true); 1539 nfs_false = txdr_unsigned(false); 1540 nfs_xdrneg1 = txdr_unsigned(-1); 1541 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1542 if (nfs_ticks < 1) 1543 nfs_ticks = 1; 1544 #ifdef NFSSERVER 1545 vfs_hooks_attach(&nfs_export_hooks); 1546 nfsrv_init(0); /* Init server data structures */ 1547 nfsrv_initcache(); /* Init the server request cache */ 1548 { 1549 extern krwlock_t netexport_lock; /* XXX */ 1550 rw_init(&netexport_lock); 1551 } 1552 #endif /* NFSSERVER */ 1553 1554 #if defined(NFSSERVER) || (defined(NFS) && !defined(NFS_V2_ONLY)) 1555 nfsdreq_init(); 1556 #endif /* defined(NFSSERVER) || (defined(NFS) && !defined(NFS_V2_ONLY)) */ 1557 1558 /* 1559 * Initialize reply list and start timer 1560 */ 1561 mutex_init(&nfs_reqq_lock, MUTEX_DEFAULT, IPL_SOFTCLOCK); 1562 mutex_init(&nfs_xidlock, MUTEX_DEFAULT, IPL_NONE); 1563 TAILQ_INIT(&nfs_reqq); 1564 nfs_timer_init(); 1565 MOWNER_ATTACH(&nfs_mowner); 1566 1567 #ifdef NFS 1568 /* Initialize the kqueue structures */ 1569 nfs_kqinit(); 1570 /* Initialize the iod structures */ 1571 nfs_iodinit(); 1572 #endif 1573 return 0; 1574 } 1575 1576 void 1577 nfs_init(void) 1578 { 1579 static ONCE_DECL(nfs_init_once); 1580 1581 RUN_ONCE(&nfs_init_once, nfs_init0); 1582 } 1583 1584 #ifdef NFS 1585 /* 1586 * Called once at VFS init to initialize client-specific data structures. 1587 */ 1588 void 1589 nfs_vfs_init() 1590 { 1591 /* Initialize NFS server / client shared data. */ 1592 nfs_init(); 1593 1594 nfs_nhinit(); /* Init the nfsnode table */ 1595 nfs_commitsize = uvmexp.npages << (PAGE_SHIFT - 4); 1596 } 1597 1598 void 1599 nfs_vfs_reinit() 1600 { 1601 nfs_nhreinit(); 1602 } 1603 1604 void 1605 nfs_vfs_done() 1606 { 1607 nfs_nhdone(); 1608 } 1609 1610 /* 1611 * Attribute cache routines. 1612 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1613 * that are on the mbuf list 1614 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1615 * error otherwise 1616 */ 1617 1618 /* 1619 * Load the attribute cache (that lives in the nfsnode entry) with 1620 * the values on the mbuf list and 1621 * Iff vap not NULL 1622 * copy the attributes to *vaper 1623 */ 1624 int 1625 nfsm_loadattrcache(vpp, mdp, dposp, vaper, flags) 1626 struct vnode **vpp; 1627 struct mbuf **mdp; 1628 char **dposp; 1629 struct vattr *vaper; 1630 int flags; 1631 { 1632 int32_t t1; 1633 char *cp2; 1634 int error = 0; 1635 struct mbuf *md; 1636 int v3 = NFS_ISV3(*vpp); 1637 1638 md = *mdp; 1639 t1 = (mtod(md, char *) + md->m_len) - *dposp; 1640 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1641 if (error) 1642 return (error); 1643 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper, flags); 1644 } 1645 1646 int 1647 nfs_loadattrcache(vpp, fp, vaper, flags) 1648 struct vnode **vpp; 1649 struct nfs_fattr *fp; 1650 struct vattr *vaper; 1651 int flags; 1652 { 1653 struct vnode *vp = *vpp; 1654 struct vattr *vap; 1655 int v3 = NFS_ISV3(vp); 1656 enum vtype vtyp; 1657 u_short vmode; 1658 struct timespec mtime; 1659 struct timespec ctime; 1660 int32_t rdev; 1661 struct nfsnode *np; 1662 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1663 uid_t uid; 1664 gid_t gid; 1665 1666 if (v3) { 1667 vtyp = nfsv3tov_type(fp->fa_type); 1668 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1669 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1670 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1671 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1672 fxdr_nfsv3time(&fp->fa3_ctime, &ctime); 1673 } else { 1674 vtyp = nfsv2tov_type(fp->fa_type); 1675 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1676 if (vtyp == VNON || vtyp == VREG) 1677 vtyp = IFTOVT(vmode); 1678 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1679 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1680 ctime.tv_sec = fxdr_unsigned(u_int32_t, 1681 fp->fa2_ctime.nfsv2_sec); 1682 ctime.tv_nsec = 0; 1683 1684 /* 1685 * Really ugly NFSv2 kludge. 1686 */ 1687 if (vtyp == VCHR && rdev == 0xffffffff) 1688 vtyp = VFIFO; 1689 } 1690 1691 vmode &= ALLPERMS; 1692 1693 /* 1694 * If v_type == VNON it is a new node, so fill in the v_type, 1695 * n_mtime fields. Check to see if it represents a special 1696 * device, and if so, check for a possible alias. Once the 1697 * correct vnode has been obtained, fill in the rest of the 1698 * information. 1699 */ 1700 np = VTONFS(vp); 1701 if (vp->v_type == VNON) { 1702 vp->v_type = vtyp; 1703 if (vp->v_type == VFIFO) { 1704 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1705 vp->v_op = fifo_nfsv2nodeop_p; 1706 } else if (vp->v_type == VREG) { 1707 mutex_init(&np->n_commitlock, MUTEX_DEFAULT, IPL_NONE); 1708 } else if (vp->v_type == VCHR || vp->v_type == VBLK) { 1709 vp->v_op = spec_nfsv2nodeop_p; 1710 spec_node_init(vp, (dev_t)rdev); 1711 } 1712 np->n_mtime = mtime; 1713 } 1714 uid = fxdr_unsigned(uid_t, fp->fa_uid); 1715 gid = fxdr_unsigned(gid_t, fp->fa_gid); 1716 vap = np->n_vattr; 1717 1718 mutex_enter(&np->n_attrlock); 1719 /* 1720 * Invalidate access cache if uid, gid, mode or ctime changed. 1721 */ 1722 if (np->n_accstamp != -1 && 1723 (gid != vap->va_gid || uid != vap->va_uid || vmode != vap->va_mode 1724 || timespeccmp(&ctime, &vap->va_ctime, !=))) 1725 np->n_accstamp = -1; 1726 1727 vap->va_type = vtyp; 1728 vap->va_mode = vmode; 1729 vap->va_rdev = (dev_t)rdev; 1730 vap->va_mtime = mtime; 1731 vap->va_ctime = ctime; 1732 vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0]; 1733 switch (vtyp) { 1734 case VDIR: 1735 vap->va_blocksize = NFS_DIRFRAGSIZ; 1736 break; 1737 case VBLK: 1738 vap->va_blocksize = BLKDEV_IOSIZE; 1739 break; 1740 case VCHR: 1741 vap->va_blocksize = MAXBSIZE; 1742 break; 1743 default: 1744 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1745 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1746 break; 1747 } 1748 if (v3) { 1749 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1750 vap->va_uid = uid; 1751 vap->va_gid = gid; 1752 vap->va_size = fxdr_hyper(&fp->fa3_size); 1753 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1754 vap->va_fileid = fxdr_hyper(&fp->fa3_fileid); 1755 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1756 vap->va_flags = 0; 1757 vap->va_filerev = 0; 1758 } else { 1759 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1760 vap->va_uid = uid; 1761 vap->va_gid = gid; 1762 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1763 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1764 * NFS_FABLKSIZE; 1765 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1766 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1767 vap->va_flags = 0; 1768 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1769 vap->va_filerev = 0; 1770 } 1771 if (vap->va_size > VFSTONFS(vp->v_mount)->nm_maxfilesize) { 1772 mutex_exit(&np->n_attrlock); 1773 return EFBIG; 1774 } 1775 if (vap->va_size != np->n_size) { 1776 if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) { 1777 vap->va_size = np->n_size; 1778 } else { 1779 np->n_size = vap->va_size; 1780 if (vap->va_type == VREG) { 1781 /* 1782 * we can't free pages if NAC_NOTRUNC because 1783 * the pages can be owned by ourselves. 1784 */ 1785 if (flags & NAC_NOTRUNC) { 1786 np->n_flag |= NTRUNCDELAYED; 1787 } else { 1788 mutex_exit(&np->n_attrlock); /* XXX */ 1789 genfs_node_wrlock(vp); 1790 mutex_enter(&vp->v_interlock); 1791 (void)VOP_PUTPAGES(vp, 0, 1792 0, PGO_SYNCIO | PGO_CLEANIT | 1793 PGO_FREE | PGO_ALLPAGES); 1794 uvm_vnp_setsize(vp, np->n_size); 1795 genfs_node_unlock(vp); 1796 mutex_enter(&np->n_attrlock); 1797 } 1798 } 1799 } 1800 } 1801 np->n_attrstamp = time_second; 1802 if (vaper != NULL) { 1803 memcpy((void *)vaper, (void *)vap, sizeof(*vap)); 1804 if (np->n_flag & NCHG) { 1805 if (np->n_flag & NACC) 1806 vaper->va_atime = np->n_atim; 1807 if (np->n_flag & NUPD) 1808 vaper->va_mtime = np->n_mtim; 1809 } 1810 } 1811 mutex_exit(&np->n_attrlock); 1812 return (0); 1813 } 1814 1815 /* 1816 * Check the time stamp 1817 * If the cache is valid, copy contents to *vap and return 0 1818 * otherwise return an error 1819 */ 1820 int 1821 nfs_getattrcache(vp, vaper) 1822 struct vnode *vp; 1823 struct vattr *vaper; 1824 { 1825 struct nfsnode *np = VTONFS(vp); 1826 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1827 struct vattr *vap; 1828 1829 mutex_enter(&np->n_attrlock); 1830 if (np->n_attrstamp == 0 || 1831 (time_second - np->n_attrstamp) >= nfs_attrtimeo(nmp, np)) { 1832 mutex_exit(&np->n_attrlock); 1833 nfsstats.attrcache_misses++; 1834 return (ENOENT); 1835 } 1836 nfsstats.attrcache_hits++; 1837 1838 vap = np->n_vattr; 1839 if (vap->va_size != np->n_size) { 1840 if (vap->va_type == VREG) { 1841 voff_t size; 1842 1843 if ((np->n_flag & NMODIFIED) != 0 && 1844 vap->va_size < np->n_size) { 1845 vap->va_size = np->n_size; 1846 } else { 1847 np->n_size = vap->va_size; 1848 } 1849 size = np->n_size; 1850 mutex_exit(&np->n_attrlock); /* XXX */ 1851 genfs_node_wrlock(vp); 1852 uvm_vnp_setsize(vp, size); 1853 genfs_node_unlock(vp); 1854 mutex_enter(&np->n_attrlock); 1855 } else 1856 np->n_size = vap->va_size; 1857 } 1858 memcpy(vaper, vap, sizeof(struct vattr)); 1859 if (np->n_flag & NCHG) { 1860 if (np->n_flag & NACC) 1861 vaper->va_atime = np->n_atim; 1862 if (np->n_flag & NUPD) 1863 vaper->va_mtime = np->n_mtim; 1864 } 1865 mutex_exit(&np->n_attrlock); 1866 return (0); 1867 } 1868 1869 void 1870 nfs_delayedtruncate(vp) 1871 struct vnode *vp; 1872 { 1873 struct nfsnode *np = VTONFS(vp); 1874 1875 if (np->n_flag & NTRUNCDELAYED) { 1876 np->n_flag &= ~NTRUNCDELAYED; 1877 genfs_node_wrlock(vp); 1878 mutex_enter(&vp->v_interlock); 1879 (void)VOP_PUTPAGES(vp, 0, 1880 0, PGO_SYNCIO | PGO_CLEANIT | PGO_FREE | PGO_ALLPAGES); 1881 uvm_vnp_setsize(vp, np->n_size); 1882 genfs_node_unlock(vp); 1883 } 1884 } 1885 1886 #define NFS_WCCKLUDGE_TIMEOUT (24 * 60 * 60) /* 1 day */ 1887 #define NFS_WCCKLUDGE(nmp, now) \ 1888 (((nmp)->nm_iflag & NFSMNT_WCCKLUDGE) && \ 1889 ((now) - (nmp)->nm_wcckludgetime - NFS_WCCKLUDGE_TIMEOUT) < 0) 1890 1891 /* 1892 * nfs_check_wccdata: check inaccurate wcc_data 1893 * 1894 * => return non-zero if we shouldn't trust the wcc_data. 1895 * => NFS_WCCKLUDGE_TIMEOUT is for the case that the server is "fixed". 1896 */ 1897 1898 int 1899 nfs_check_wccdata(struct nfsnode *np, const struct timespec *ctime, 1900 struct timespec *mtime, bool docheck) 1901 { 1902 int error = 0; 1903 1904 #if !defined(NFS_V2_ONLY) 1905 1906 if (docheck) { 1907 struct vnode *vp = NFSTOV(np); 1908 struct nfsmount *nmp; 1909 long now = time_second; 1910 const struct timespec *omtime = &np->n_vattr->va_mtime; 1911 const struct timespec *octime = &np->n_vattr->va_ctime; 1912 const char *reason = NULL; /* XXX: gcc */ 1913 1914 if (timespeccmp(omtime, mtime, <=)) { 1915 reason = "mtime"; 1916 error = EINVAL; 1917 } 1918 1919 if (vp->v_type == VDIR && timespeccmp(octime, ctime, <=)) { 1920 reason = "ctime"; 1921 error = EINVAL; 1922 } 1923 1924 nmp = VFSTONFS(vp->v_mount); 1925 if (error) { 1926 1927 /* 1928 * despite of the fact that we've updated the file, 1929 * timestamps of the file were not updated as we 1930 * expected. 1931 * it means that the server has incompatible 1932 * semantics of timestamps or (more likely) 1933 * the server time is not precise enough to 1934 * track each modifications. 1935 * in that case, we disable wcc processing. 1936 * 1937 * yes, strictly speaking, we should disable all 1938 * caching. it's a compromise. 1939 */ 1940 1941 mutex_enter(&nmp->nm_lock); 1942 if (!NFS_WCCKLUDGE(nmp, now)) { 1943 printf("%s: inaccurate wcc data (%s) detected," 1944 " disabling wcc" 1945 " (ctime %u.%09u %u.%09u," 1946 " mtime %u.%09u %u.%09u)\n", 1947 vp->v_mount->mnt_stat.f_mntfromname, 1948 reason, 1949 (unsigned int)octime->tv_sec, 1950 (unsigned int)octime->tv_nsec, 1951 (unsigned int)ctime->tv_sec, 1952 (unsigned int)ctime->tv_nsec, 1953 (unsigned int)omtime->tv_sec, 1954 (unsigned int)omtime->tv_nsec, 1955 (unsigned int)mtime->tv_sec, 1956 (unsigned int)mtime->tv_nsec); 1957 } 1958 nmp->nm_iflag |= NFSMNT_WCCKLUDGE; 1959 nmp->nm_wcckludgetime = now; 1960 mutex_exit(&nmp->nm_lock); 1961 } else if (NFS_WCCKLUDGE(nmp, now)) { 1962 error = EPERM; /* XXX */ 1963 } else if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 1964 mutex_enter(&nmp->nm_lock); 1965 if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 1966 printf("%s: re-enabling wcc\n", 1967 vp->v_mount->mnt_stat.f_mntfromname); 1968 nmp->nm_iflag &= ~NFSMNT_WCCKLUDGE; 1969 } 1970 mutex_exit(&nmp->nm_lock); 1971 } 1972 } 1973 1974 #endif /* !defined(NFS_V2_ONLY) */ 1975 1976 return error; 1977 } 1978 1979 /* 1980 * Heuristic to see if the server XDR encodes directory cookies or not. 1981 * it is not supposed to, but a lot of servers may do this. Also, since 1982 * most/all servers will implement V2 as well, it is expected that they 1983 * may return just 32 bits worth of cookie information, so we need to 1984 * find out in which 32 bits this information is available. We do this 1985 * to avoid trouble with emulated binaries that can't handle 64 bit 1986 * directory offsets. 1987 */ 1988 1989 void 1990 nfs_cookieheuristic(vp, flagp, l, cred) 1991 struct vnode *vp; 1992 int *flagp; 1993 struct lwp *l; 1994 kauth_cred_t cred; 1995 { 1996 struct uio auio; 1997 struct iovec aiov; 1998 char *tbuf, *cp; 1999 struct dirent *dp; 2000 off_t *cookies = NULL, *cop; 2001 int error, eof, nc, len; 2002 2003 MALLOC(tbuf, void *, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 2004 2005 aiov.iov_base = tbuf; 2006 aiov.iov_len = NFS_DIRFRAGSIZ; 2007 auio.uio_iov = &aiov; 2008 auio.uio_iovcnt = 1; 2009 auio.uio_rw = UIO_READ; 2010 auio.uio_resid = NFS_DIRFRAGSIZ; 2011 auio.uio_offset = 0; 2012 UIO_SETUP_SYSSPACE(&auio); 2013 2014 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 2015 2016 len = NFS_DIRFRAGSIZ - auio.uio_resid; 2017 if (error || len == 0) { 2018 FREE(tbuf, M_TEMP); 2019 if (cookies) 2020 free(cookies, M_TEMP); 2021 return; 2022 } 2023 2024 /* 2025 * Find the first valid entry and look at its offset cookie. 2026 */ 2027 2028 cp = tbuf; 2029 for (cop = cookies; len > 0; len -= dp->d_reclen) { 2030 dp = (struct dirent *)cp; 2031 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 2032 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 2033 *flagp |= NFSMNT_SWAPCOOKIE; 2034 nfs_invaldircache(vp, 0); 2035 nfs_vinvalbuf(vp, 0, cred, l, 1); 2036 } 2037 break; 2038 } 2039 cop++; 2040 cp += dp->d_reclen; 2041 } 2042 2043 FREE(tbuf, M_TEMP); 2044 free(cookies, M_TEMP); 2045 } 2046 #endif /* NFS */ 2047 2048 #ifdef NFSSERVER 2049 /* 2050 * Set up nameidata for a lookup() call and do it. 2051 * 2052 * If pubflag is set, this call is done for a lookup operation on the 2053 * public filehandle. In that case we allow crossing mountpoints and 2054 * absolute pathnames. However, the caller is expected to check that 2055 * the lookup result is within the public fs, and deny access if 2056 * it is not. 2057 */ 2058 int 2059 nfs_namei(ndp, nsfh, len, slp, nam, mdp, dposp, retdirp, l, kerbflag, pubflag) 2060 struct nameidata *ndp; 2061 nfsrvfh_t *nsfh; 2062 uint32_t len; 2063 struct nfssvc_sock *slp; 2064 struct mbuf *nam; 2065 struct mbuf **mdp; 2066 char **dposp; 2067 struct vnode **retdirp; 2068 struct lwp *l; 2069 int kerbflag, pubflag; 2070 { 2071 int i, rem; 2072 struct mbuf *md; 2073 char *fromcp, *tocp, *cp; 2074 struct iovec aiov; 2075 struct uio auio; 2076 struct vnode *dp; 2077 int error, rdonly, linklen; 2078 struct componentname *cnp = &ndp->ni_cnd; 2079 2080 *retdirp = NULL; 2081 2082 if ((len + 1) > MAXPATHLEN) 2083 return (ENAMETOOLONG); 2084 if (len == 0) 2085 return (EACCES); 2086 cnp->cn_pnbuf = PNBUF_GET(); 2087 2088 /* 2089 * Copy the name from the mbuf list to ndp->ni_pnbuf 2090 * and set the various ndp fields appropriately. 2091 */ 2092 fromcp = *dposp; 2093 tocp = cnp->cn_pnbuf; 2094 md = *mdp; 2095 rem = mtod(md, char *) + md->m_len - fromcp; 2096 for (i = 0; i < len; i++) { 2097 while (rem == 0) { 2098 md = md->m_next; 2099 if (md == NULL) { 2100 error = EBADRPC; 2101 goto out; 2102 } 2103 fromcp = mtod(md, void *); 2104 rem = md->m_len; 2105 } 2106 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 2107 error = EACCES; 2108 goto out; 2109 } 2110 *tocp++ = *fromcp++; 2111 rem--; 2112 } 2113 *tocp = '\0'; 2114 *mdp = md; 2115 *dposp = fromcp; 2116 len = nfsm_rndup(len)-len; 2117 if (len > 0) { 2118 if (rem >= len) 2119 *dposp += len; 2120 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 2121 goto out; 2122 } 2123 2124 /* 2125 * Extract and set starting directory. 2126 */ 2127 error = nfsrv_fhtovp(nsfh, false, &dp, ndp->ni_cnd.cn_cred, slp, 2128 nam, &rdonly, kerbflag, pubflag); 2129 if (error) 2130 goto out; 2131 if (dp->v_type != VDIR) { 2132 vrele(dp); 2133 error = ENOTDIR; 2134 goto out; 2135 } 2136 2137 if (rdonly) 2138 cnp->cn_flags |= RDONLY; 2139 2140 *retdirp = dp; 2141 2142 if (pubflag) { 2143 /* 2144 * Oh joy. For WebNFS, handle those pesky '%' escapes, 2145 * and the 'native path' indicator. 2146 */ 2147 cp = PNBUF_GET(); 2148 fromcp = cnp->cn_pnbuf; 2149 tocp = cp; 2150 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 2151 switch ((unsigned char)*fromcp) { 2152 case WEBNFS_NATIVE_CHAR: 2153 /* 2154 * 'Native' path for us is the same 2155 * as a path according to the NFS spec, 2156 * just skip the escape char. 2157 */ 2158 fromcp++; 2159 break; 2160 /* 2161 * More may be added in the future, range 0x80-0xff 2162 */ 2163 default: 2164 error = EIO; 2165 vrele(dp); 2166 PNBUF_PUT(cp); 2167 goto out; 2168 } 2169 } 2170 /* 2171 * Translate the '%' escapes, URL-style. 2172 */ 2173 while (*fromcp != '\0') { 2174 if (*fromcp == WEBNFS_ESC_CHAR) { 2175 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 2176 fromcp++; 2177 *tocp++ = HEXSTRTOI(fromcp); 2178 fromcp += 2; 2179 continue; 2180 } else { 2181 error = ENOENT; 2182 vrele(dp); 2183 PNBUF_PUT(cp); 2184 goto out; 2185 } 2186 } else 2187 *tocp++ = *fromcp++; 2188 } 2189 *tocp = '\0'; 2190 PNBUF_PUT(cnp->cn_pnbuf); 2191 cnp->cn_pnbuf = cp; 2192 } 2193 2194 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 2195 ndp->ni_segflg = UIO_SYSSPACE; 2196 ndp->ni_rootdir = rootvnode; 2197 ndp->ni_erootdir = NULL; 2198 2199 if (pubflag) { 2200 ndp->ni_loopcnt = 0; 2201 if (cnp->cn_pnbuf[0] == '/') 2202 dp = rootvnode; 2203 } else { 2204 cnp->cn_flags |= NOCROSSMOUNT; 2205 } 2206 2207 VREF(dp); 2208 vn_lock(dp, LK_EXCLUSIVE | LK_RETRY); 2209 2210 for (;;) { 2211 cnp->cn_nameptr = cnp->cn_pnbuf; 2212 ndp->ni_startdir = dp; 2213 2214 /* 2215 * And call lookup() to do the real work 2216 */ 2217 error = lookup(ndp); 2218 if (error) { 2219 if (ndp->ni_dvp) { 2220 vput(ndp->ni_dvp); 2221 } 2222 PNBUF_PUT(cnp->cn_pnbuf); 2223 return (error); 2224 } 2225 2226 /* 2227 * Check for encountering a symbolic link 2228 */ 2229 if ((cnp->cn_flags & ISSYMLINK) == 0) { 2230 if ((cnp->cn_flags & LOCKPARENT) == 0 && ndp->ni_dvp) { 2231 if (ndp->ni_dvp == ndp->ni_vp) { 2232 vrele(ndp->ni_dvp); 2233 } else { 2234 vput(ndp->ni_dvp); 2235 } 2236 } 2237 if (cnp->cn_flags & (SAVENAME | SAVESTART)) 2238 cnp->cn_flags |= HASBUF; 2239 else 2240 PNBUF_PUT(cnp->cn_pnbuf); 2241 return (0); 2242 } else { 2243 if (!pubflag) { 2244 error = EINVAL; 2245 break; 2246 } 2247 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 2248 error = ELOOP; 2249 break; 2250 } 2251 if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) { 2252 error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred); 2253 if (error != 0) 2254 break; 2255 } 2256 if (ndp->ni_pathlen > 1) 2257 cp = PNBUF_GET(); 2258 else 2259 cp = cnp->cn_pnbuf; 2260 aiov.iov_base = cp; 2261 aiov.iov_len = MAXPATHLEN; 2262 auio.uio_iov = &aiov; 2263 auio.uio_iovcnt = 1; 2264 auio.uio_offset = 0; 2265 auio.uio_rw = UIO_READ; 2266 auio.uio_resid = MAXPATHLEN; 2267 UIO_SETUP_SYSSPACE(&auio); 2268 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 2269 if (error) { 2270 badlink: 2271 if (ndp->ni_pathlen > 1) 2272 PNBUF_PUT(cp); 2273 break; 2274 } 2275 linklen = MAXPATHLEN - auio.uio_resid; 2276 if (linklen == 0) { 2277 error = ENOENT; 2278 goto badlink; 2279 } 2280 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 2281 error = ENAMETOOLONG; 2282 goto badlink; 2283 } 2284 if (ndp->ni_pathlen > 1) { 2285 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 2286 PNBUF_PUT(cnp->cn_pnbuf); 2287 cnp->cn_pnbuf = cp; 2288 } else 2289 cnp->cn_pnbuf[linklen] = '\0'; 2290 ndp->ni_pathlen += linklen; 2291 vput(ndp->ni_vp); 2292 dp = ndp->ni_dvp; 2293 2294 /* 2295 * Check if root directory should replace current directory. 2296 */ 2297 if (cnp->cn_pnbuf[0] == '/') { 2298 vput(dp); 2299 dp = ndp->ni_rootdir; 2300 VREF(dp); 2301 vn_lock(dp, LK_EXCLUSIVE | LK_RETRY); 2302 } 2303 } 2304 } 2305 vput(ndp->ni_dvp); 2306 vput(ndp->ni_vp); 2307 ndp->ni_vp = NULL; 2308 out: 2309 PNBUF_PUT(cnp->cn_pnbuf); 2310 return (error); 2311 } 2312 #endif /* NFSSERVER */ 2313 2314 /* 2315 * A fiddled version of m_adj() that ensures null fill to a 32-bit 2316 * boundary and only trims off the back end 2317 * 2318 * 1. trim off 'len' bytes as m_adj(mp, -len). 2319 * 2. add zero-padding 'nul' bytes at the end of the mbuf chain. 2320 */ 2321 void 2322 nfs_zeropad(mp, len, nul) 2323 struct mbuf *mp; 2324 int len; 2325 int nul; 2326 { 2327 struct mbuf *m; 2328 int count; 2329 2330 /* 2331 * Trim from tail. Scan the mbuf chain, 2332 * calculating its length and finding the last mbuf. 2333 * If the adjustment only affects this mbuf, then just 2334 * adjust and return. Otherwise, rescan and truncate 2335 * after the remaining size. 2336 */ 2337 count = 0; 2338 m = mp; 2339 for (;;) { 2340 count += m->m_len; 2341 if (m->m_next == NULL) 2342 break; 2343 m = m->m_next; 2344 } 2345 2346 KDASSERT(count >= len); 2347 2348 if (m->m_len >= len) { 2349 m->m_len -= len; 2350 } else { 2351 count -= len; 2352 /* 2353 * Correct length for chain is "count". 2354 * Find the mbuf with last data, adjust its length, 2355 * and toss data from remaining mbufs on chain. 2356 */ 2357 for (m = mp; m; m = m->m_next) { 2358 if (m->m_len >= count) { 2359 m->m_len = count; 2360 break; 2361 } 2362 count -= m->m_len; 2363 } 2364 KASSERT(m && m->m_next); 2365 m_freem(m->m_next); 2366 m->m_next = NULL; 2367 } 2368 2369 KDASSERT(m->m_next == NULL); 2370 2371 /* 2372 * zero-padding. 2373 */ 2374 if (nul > 0) { 2375 char *cp; 2376 int i; 2377 2378 if (M_ROMAP(m) || M_TRAILINGSPACE(m) < nul) { 2379 struct mbuf *n; 2380 2381 KDASSERT(MLEN >= nul); 2382 n = m_get(M_WAIT, MT_DATA); 2383 MCLAIM(n, &nfs_mowner); 2384 n->m_len = nul; 2385 n->m_next = NULL; 2386 m->m_next = n; 2387 cp = mtod(n, void *); 2388 } else { 2389 cp = mtod(m, char *) + m->m_len; 2390 m->m_len += nul; 2391 } 2392 for (i = 0; i < nul; i++) 2393 *cp++ = '\0'; 2394 } 2395 return; 2396 } 2397 2398 /* 2399 * Make these functions instead of macros, so that the kernel text size 2400 * doesn't get too big... 2401 */ 2402 void 2403 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2404 struct nfsrv_descript *nfsd; 2405 int before_ret; 2406 struct vattr *before_vap; 2407 int after_ret; 2408 struct vattr *after_vap; 2409 struct mbuf **mbp; 2410 char **bposp; 2411 { 2412 struct mbuf *mb = *mbp; 2413 char *bpos = *bposp; 2414 u_int32_t *tl; 2415 2416 if (before_ret) { 2417 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2418 *tl = nfs_false; 2419 } else { 2420 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2421 *tl++ = nfs_true; 2422 txdr_hyper(before_vap->va_size, tl); 2423 tl += 2; 2424 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2425 tl += 2; 2426 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2427 } 2428 *bposp = bpos; 2429 *mbp = mb; 2430 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2431 } 2432 2433 void 2434 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2435 struct nfsrv_descript *nfsd; 2436 int after_ret; 2437 struct vattr *after_vap; 2438 struct mbuf **mbp; 2439 char **bposp; 2440 { 2441 struct mbuf *mb = *mbp; 2442 char *bpos = *bposp; 2443 u_int32_t *tl; 2444 struct nfs_fattr *fp; 2445 2446 if (after_ret) { 2447 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2448 *tl = nfs_false; 2449 } else { 2450 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2451 *tl++ = nfs_true; 2452 fp = (struct nfs_fattr *)tl; 2453 nfsm_srvfattr(nfsd, after_vap, fp); 2454 } 2455 *mbp = mb; 2456 *bposp = bpos; 2457 } 2458 2459 void 2460 nfsm_srvfattr(nfsd, vap, fp) 2461 struct nfsrv_descript *nfsd; 2462 struct vattr *vap; 2463 struct nfs_fattr *fp; 2464 { 2465 2466 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2467 fp->fa_uid = txdr_unsigned(vap->va_uid); 2468 fp->fa_gid = txdr_unsigned(vap->va_gid); 2469 if (nfsd->nd_flag & ND_NFSV3) { 2470 fp->fa_type = vtonfsv3_type(vap->va_type); 2471 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2472 txdr_hyper(vap->va_size, &fp->fa3_size); 2473 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2474 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2475 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2476 fp->fa3_fsid.nfsuquad[0] = 0; 2477 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2478 txdr_hyper(vap->va_fileid, &fp->fa3_fileid); 2479 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2480 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2481 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2482 } else { 2483 fp->fa_type = vtonfsv2_type(vap->va_type); 2484 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2485 fp->fa2_size = txdr_unsigned(vap->va_size); 2486 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2487 if (vap->va_type == VFIFO) 2488 fp->fa2_rdev = 0xffffffff; 2489 else 2490 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2491 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2492 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2493 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2494 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2495 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2496 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2497 } 2498 } 2499 2500 #ifdef NFSSERVER 2501 /* 2502 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2503 * - look up fsid in mount list (if not found ret error) 2504 * - get vp and export rights by calling VFS_FHTOVP() 2505 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2506 * - if not lockflag unlock it with VOP_UNLOCK() 2507 */ 2508 int 2509 nfsrv_fhtovp(nfsrvfh_t *nsfh, int lockflag, struct vnode **vpp, 2510 kauth_cred_t cred, struct nfssvc_sock *slp, struct mbuf *nam, int *rdonlyp, 2511 int kerbflag, int pubflag) 2512 { 2513 struct mount *mp; 2514 kauth_cred_t credanon; 2515 int error, exflags; 2516 struct sockaddr_in *saddr; 2517 fhandle_t *fhp; 2518 2519 fhp = NFSRVFH_FHANDLE(nsfh); 2520 *vpp = (struct vnode *)0; 2521 2522 if (nfs_ispublicfh(nsfh)) { 2523 if (!pubflag || !nfs_pub.np_valid) 2524 return (ESTALE); 2525 fhp = nfs_pub.np_handle; 2526 } 2527 2528 error = netexport_check(&fhp->fh_fsid, nam, &mp, &exflags, &credanon); 2529 if (error) { 2530 return error; 2531 } 2532 2533 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2534 if (error) 2535 return (error); 2536 2537 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2538 saddr = mtod(nam, struct sockaddr_in *); 2539 if ((saddr->sin_family == AF_INET) && 2540 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2541 vput(*vpp); 2542 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2543 } 2544 #ifdef INET6 2545 if ((saddr->sin_family == AF_INET6) && 2546 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2547 vput(*vpp); 2548 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2549 } 2550 #endif 2551 } 2552 /* 2553 * Check/setup credentials. 2554 */ 2555 if (exflags & MNT_EXKERB) { 2556 if (!kerbflag) { 2557 vput(*vpp); 2558 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2559 } 2560 } else if (kerbflag) { 2561 vput(*vpp); 2562 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2563 } else if (kauth_cred_geteuid(cred) == 0 || /* NFS maproot, see below */ 2564 (exflags & MNT_EXPORTANON)) { 2565 /* 2566 * This is used by the NFS maproot option. While we can change 2567 * the secmodel on our own host, we can't change it on the 2568 * clients. As means of least surprise, we're doing the 2569 * traditional thing here. 2570 * Should look into adding a "mapprivileged" or similar where 2571 * the users can be explicitly specified... 2572 * [elad, yamt 2008-03-05] 2573 */ 2574 kauth_cred_clone(credanon, cred); 2575 } 2576 if (exflags & MNT_EXRDONLY) 2577 *rdonlyp = 1; 2578 else 2579 *rdonlyp = 0; 2580 if (!lockflag) 2581 VOP_UNLOCK(*vpp, 0); 2582 return (0); 2583 } 2584 2585 /* 2586 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2587 * means a length of 0, for v2 it means all zeroes. 2588 */ 2589 int 2590 nfs_ispublicfh(const nfsrvfh_t *nsfh) 2591 { 2592 const char *cp = (const void *)(NFSRVFH_DATA(nsfh)); 2593 int i; 2594 2595 if (NFSRVFH_SIZE(nsfh) == 0) { 2596 return true; 2597 } 2598 if (NFSRVFH_SIZE(nsfh) != NFSX_V2FH) { 2599 return false; 2600 } 2601 for (i = 0; i < NFSX_V2FH; i++) 2602 if (*cp++ != 0) 2603 return false; 2604 return true; 2605 } 2606 #endif /* NFSSERVER */ 2607 2608 /* 2609 * This function compares two net addresses by family and returns true 2610 * if they are the same host. 2611 * If there is any doubt, return false. 2612 * The AF_INET family is handled as a special case so that address mbufs 2613 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2614 */ 2615 int 2616 netaddr_match(family, haddr, nam) 2617 int family; 2618 union nethostaddr *haddr; 2619 struct mbuf *nam; 2620 { 2621 struct sockaddr_in *inetaddr; 2622 2623 switch (family) { 2624 case AF_INET: 2625 inetaddr = mtod(nam, struct sockaddr_in *); 2626 if (inetaddr->sin_family == AF_INET && 2627 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2628 return (1); 2629 break; 2630 #ifdef INET6 2631 case AF_INET6: 2632 { 2633 struct sockaddr_in6 *sin6_1, *sin6_2; 2634 2635 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2636 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2637 if (sin6_1->sin6_family == AF_INET6 && 2638 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2639 return 1; 2640 } 2641 #endif 2642 #ifdef ISO 2643 case AF_ISO: 2644 { 2645 struct sockaddr_iso *isoaddr1, *isoaddr2; 2646 2647 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2648 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2649 if (isoaddr1->siso_family == AF_ISO && 2650 isoaddr1->siso_nlen > 0 && 2651 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2652 SAME_ISOADDR(isoaddr1, isoaddr2)) 2653 return (1); 2654 break; 2655 } 2656 #endif /* ISO */ 2657 default: 2658 break; 2659 }; 2660 return (0); 2661 } 2662 2663 /* 2664 * The write verifier has changed (probably due to a server reboot), so all 2665 * PG_NEEDCOMMIT pages will have to be written again. Since they are marked 2666 * as dirty or are being written out just now, all this takes is clearing 2667 * the PG_NEEDCOMMIT flag. Once done the new write verifier can be set for 2668 * the mount point. 2669 */ 2670 void 2671 nfs_clearcommit(mp) 2672 struct mount *mp; 2673 { 2674 struct vnode *vp; 2675 struct nfsnode *np; 2676 struct vm_page *pg; 2677 struct nfsmount *nmp = VFSTONFS(mp); 2678 2679 rw_enter(&nmp->nm_writeverflock, RW_WRITER); 2680 mutex_enter(&mntvnode_lock); 2681 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2682 KASSERT(vp->v_mount == mp); 2683 if (vp->v_type != VREG) 2684 continue; 2685 np = VTONFS(vp); 2686 np->n_pushlo = np->n_pushhi = np->n_pushedlo = 2687 np->n_pushedhi = 0; 2688 np->n_commitflags &= 2689 ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID); 2690 mutex_enter(&vp->v_uobj.vmobjlock); 2691 TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) { 2692 pg->flags &= ~PG_NEEDCOMMIT; 2693 } 2694 mutex_exit(&vp->v_uobj.vmobjlock); 2695 } 2696 mutex_exit(&mntvnode_lock); 2697 mutex_enter(&nmp->nm_lock); 2698 nmp->nm_iflag &= ~NFSMNT_STALEWRITEVERF; 2699 mutex_exit(&nmp->nm_lock); 2700 rw_exit(&nmp->nm_writeverflock); 2701 } 2702 2703 void 2704 nfs_merge_commit_ranges(vp) 2705 struct vnode *vp; 2706 { 2707 struct nfsnode *np = VTONFS(vp); 2708 2709 KASSERT(np->n_commitflags & NFS_COMMIT_PUSH_VALID); 2710 2711 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2712 np->n_pushedlo = np->n_pushlo; 2713 np->n_pushedhi = np->n_pushhi; 2714 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2715 } else { 2716 if (np->n_pushlo < np->n_pushedlo) 2717 np->n_pushedlo = np->n_pushlo; 2718 if (np->n_pushhi > np->n_pushedhi) 2719 np->n_pushedhi = np->n_pushhi; 2720 } 2721 2722 np->n_pushlo = np->n_pushhi = 0; 2723 np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID; 2724 2725 #ifdef NFS_DEBUG_COMMIT 2726 printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2727 (unsigned)np->n_pushedhi); 2728 #endif 2729 } 2730 2731 int 2732 nfs_in_committed_range(vp, off, len) 2733 struct vnode *vp; 2734 off_t off, len; 2735 { 2736 struct nfsnode *np = VTONFS(vp); 2737 off_t lo, hi; 2738 2739 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2740 return 0; 2741 lo = off; 2742 hi = lo + len; 2743 2744 return (lo >= np->n_pushedlo && hi <= np->n_pushedhi); 2745 } 2746 2747 int 2748 nfs_in_tobecommitted_range(vp, off, len) 2749 struct vnode *vp; 2750 off_t off, len; 2751 { 2752 struct nfsnode *np = VTONFS(vp); 2753 off_t lo, hi; 2754 2755 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2756 return 0; 2757 lo = off; 2758 hi = lo + len; 2759 2760 return (lo >= np->n_pushlo && hi <= np->n_pushhi); 2761 } 2762 2763 void 2764 nfs_add_committed_range(vp, off, len) 2765 struct vnode *vp; 2766 off_t off, len; 2767 { 2768 struct nfsnode *np = VTONFS(vp); 2769 off_t lo, hi; 2770 2771 lo = off; 2772 hi = lo + len; 2773 2774 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2775 np->n_pushedlo = lo; 2776 np->n_pushedhi = hi; 2777 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2778 } else { 2779 if (hi > np->n_pushedhi) 2780 np->n_pushedhi = hi; 2781 if (lo < np->n_pushedlo) 2782 np->n_pushedlo = lo; 2783 } 2784 #ifdef NFS_DEBUG_COMMIT 2785 printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2786 (unsigned)np->n_pushedhi); 2787 #endif 2788 } 2789 2790 void 2791 nfs_del_committed_range(vp, off, len) 2792 struct vnode *vp; 2793 off_t off, len; 2794 { 2795 struct nfsnode *np = VTONFS(vp); 2796 off_t lo, hi; 2797 2798 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2799 return; 2800 2801 lo = off; 2802 hi = lo + len; 2803 2804 if (lo > np->n_pushedhi || hi < np->n_pushedlo) 2805 return; 2806 if (lo <= np->n_pushedlo) 2807 np->n_pushedlo = hi; 2808 else if (hi >= np->n_pushedhi) 2809 np->n_pushedhi = lo; 2810 else { 2811 /* 2812 * XXX There's only one range. If the deleted range 2813 * is in the middle, pick the largest of the 2814 * contiguous ranges that it leaves. 2815 */ 2816 if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi)) 2817 np->n_pushedhi = lo; 2818 else 2819 np->n_pushedlo = hi; 2820 } 2821 #ifdef NFS_DEBUG_COMMIT 2822 printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2823 (unsigned)np->n_pushedhi); 2824 #endif 2825 } 2826 2827 void 2828 nfs_add_tobecommitted_range(vp, off, len) 2829 struct vnode *vp; 2830 off_t off, len; 2831 { 2832 struct nfsnode *np = VTONFS(vp); 2833 off_t lo, hi; 2834 2835 lo = off; 2836 hi = lo + len; 2837 2838 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) { 2839 np->n_pushlo = lo; 2840 np->n_pushhi = hi; 2841 np->n_commitflags |= NFS_COMMIT_PUSH_VALID; 2842 } else { 2843 if (lo < np->n_pushlo) 2844 np->n_pushlo = lo; 2845 if (hi > np->n_pushhi) 2846 np->n_pushhi = hi; 2847 } 2848 #ifdef NFS_DEBUG_COMMIT 2849 printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2850 (unsigned)np->n_pushhi); 2851 #endif 2852 } 2853 2854 void 2855 nfs_del_tobecommitted_range(vp, off, len) 2856 struct vnode *vp; 2857 off_t off, len; 2858 { 2859 struct nfsnode *np = VTONFS(vp); 2860 off_t lo, hi; 2861 2862 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2863 return; 2864 2865 lo = off; 2866 hi = lo + len; 2867 2868 if (lo > np->n_pushhi || hi < np->n_pushlo) 2869 return; 2870 2871 if (lo <= np->n_pushlo) 2872 np->n_pushlo = hi; 2873 else if (hi >= np->n_pushhi) 2874 np->n_pushhi = lo; 2875 else { 2876 /* 2877 * XXX There's only one range. If the deleted range 2878 * is in the middle, pick the largest of the 2879 * contiguous ranges that it leaves. 2880 */ 2881 if ((np->n_pushlo - lo) > (hi - np->n_pushhi)) 2882 np->n_pushhi = lo; 2883 else 2884 np->n_pushlo = hi; 2885 } 2886 #ifdef NFS_DEBUG_COMMIT 2887 printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2888 (unsigned)np->n_pushhi); 2889 #endif 2890 } 2891 2892 /* 2893 * Map errnos to NFS error numbers. For Version 3 also filter out error 2894 * numbers not specified for the associated procedure. 2895 */ 2896 int 2897 nfsrv_errmap(nd, err) 2898 struct nfsrv_descript *nd; 2899 int err; 2900 { 2901 const short *defaulterrp, *errp; 2902 2903 if (nd->nd_flag & ND_NFSV3) { 2904 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2905 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2906 while (*++errp) { 2907 if (*errp == err) 2908 return (err); 2909 else if (*errp > err) 2910 break; 2911 } 2912 return ((int)*defaulterrp); 2913 } else 2914 return (err & 0xffff); 2915 } 2916 if (err <= ELAST) 2917 return ((int)nfsrv_v2errmap[err - 1]); 2918 return (NFSERR_IO); 2919 } 2920 2921 u_int32_t 2922 nfs_getxid() 2923 { 2924 static u_int32_t base; 2925 static u_int32_t nfs_xid = 0; 2926 u_int32_t newxid; 2927 2928 mutex_enter(&nfs_xidlock); 2929 /* 2930 * derive initial xid from system time 2931 * XXX time is invalid if root not yet mounted 2932 */ 2933 if (__predict_false(!base && (rootvp))) { 2934 struct timeval tv; 2935 2936 microtime(&tv); 2937 base = tv.tv_sec << 12; 2938 nfs_xid = base; 2939 } 2940 2941 /* 2942 * Skip zero xid if it should ever happen. 2943 */ 2944 if (__predict_false(++nfs_xid == 0)) 2945 nfs_xid++; 2946 newxid = nfs_xid; 2947 mutex_exit(&nfs_xidlock); 2948 2949 return txdr_unsigned(newxid); 2950 } 2951 2952 /* 2953 * assign a new xid for existing request. 2954 * used for NFSERR_JUKEBOX handling. 2955 */ 2956 void 2957 nfs_renewxid(struct nfsreq *req) 2958 { 2959 u_int32_t xid; 2960 int off; 2961 2962 xid = nfs_getxid(); 2963 if (req->r_nmp->nm_sotype == SOCK_STREAM) 2964 off = sizeof(u_int32_t); /* RPC record mark */ 2965 else 2966 off = 0; 2967 2968 m_copyback(req->r_mreq, off, sizeof(xid), (void *)&xid); 2969 req->r_xid = xid; 2970 } 2971 2972 #if defined(NFSSERVER) 2973 int 2974 nfsrv_composefh(struct vnode *vp, nfsrvfh_t *nsfh, bool v3) 2975 { 2976 int error; 2977 size_t fhsize; 2978 2979 fhsize = NFSD_MAXFHSIZE; 2980 error = vfs_composefh(vp, (void *)NFSRVFH_DATA(nsfh), &fhsize); 2981 if (NFSX_FHTOOBIG_P(fhsize, v3)) { 2982 error = EOPNOTSUPP; 2983 } 2984 if (error != 0) { 2985 return error; 2986 } 2987 if (!v3 && fhsize < NFSX_V2FH) { 2988 memset((char *)NFSRVFH_DATA(nsfh) + fhsize, 0, 2989 NFSX_V2FH - fhsize); 2990 fhsize = NFSX_V2FH; 2991 } 2992 if ((fhsize % NFSX_UNSIGNED) != 0) { 2993 return EOPNOTSUPP; 2994 } 2995 nsfh->nsfh_size = fhsize; 2996 return 0; 2997 } 2998 2999 int 3000 nfsrv_comparefh(const nfsrvfh_t *fh1, const nfsrvfh_t *fh2) 3001 { 3002 3003 if (NFSRVFH_SIZE(fh1) != NFSRVFH_SIZE(fh2)) { 3004 return NFSRVFH_SIZE(fh2) - NFSRVFH_SIZE(fh1); 3005 } 3006 return memcmp(NFSRVFH_DATA(fh1), NFSRVFH_DATA(fh2), NFSRVFH_SIZE(fh1)); 3007 } 3008 3009 void 3010 nfsrv_copyfh(nfsrvfh_t *fh1, const nfsrvfh_t *fh2) 3011 { 3012 size_t size; 3013 3014 fh1->nsfh_size = size = NFSRVFH_SIZE(fh2); 3015 memcpy(NFSRVFH_DATA(fh1), NFSRVFH_DATA(fh2), size); 3016 } 3017 #endif /* defined(NFSSERVER) */ 3018 3019 #if defined(NFS) 3020 /* 3021 * Set the attribute timeout based on how recently the file has been modified. 3022 */ 3023 3024 time_t 3025 nfs_attrtimeo(struct nfsmount *nmp, struct nfsnode *np) 3026 { 3027 time_t timeo; 3028 3029 if ((nmp->nm_flag & NFSMNT_NOAC) != 0) 3030 return 0; 3031 3032 if (((np)->n_flag & NMODIFIED) != 0) 3033 return NFS_MINATTRTIMO; 3034 3035 timeo = (time_second - np->n_mtime.tv_sec) / 10; 3036 timeo = max(timeo, NFS_MINATTRTIMO); 3037 timeo = min(timeo, NFS_MAXATTRTIMO); 3038 return timeo; 3039 } 3040 #endif /* defined(NFS) */ 3041
Indexes created Thu Oct 02 14:10:14 GMT 2025