nfs_subs.c revision 1.144
1 /* $NetBSD: nfs_subs.c,v 1.144 2005/01/26 10:30:58 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 1.144 2005/01/26 10:30:58 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/mount.h> 91 #include <sys/vnode.h> 92 #include <sys/namei.h> 93 #include <sys/mbuf.h> 94 #include <sys/socket.h> 95 #include <sys/stat.h> 96 #include <sys/malloc.h> 97 #include <sys/filedesc.h> 98 #include <sys/time.h> 99 #include <sys/dirent.h> 100 101 #include <uvm/uvm_extern.h> 102 103 #include <nfs/rpcv2.h> 104 #include <nfs/nfsproto.h> 105 #include <nfs/nfsnode.h> 106 #include <nfs/nfs.h> 107 #include <nfs/xdr_subs.h> 108 #include <nfs/nfsm_subs.h> 109 #include <nfs/nfsmount.h> 110 #include <nfs/nqnfs.h> 111 #include <nfs/nfsrtt.h> 112 #include <nfs/nfs_var.h> 113 114 #include <miscfs/specfs/specdev.h> 115 116 #include <netinet/in.h> 117 #ifdef ISO 118 #include <netiso/iso.h> 119 #endif 120 121 /* 122 * Data items converted to xdr at startup, since they are constant 123 * This is kinda hokey, but may save a little time doing byte swaps 124 */ 125 u_int32_t nfs_xdrneg1; 126 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 127 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 128 rpc_auth_kerb; 129 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; 130 131 /* And other global data */ 132 const nfstype nfsv2_type[9] = 133 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON }; 134 const nfstype nfsv3_type[9] = 135 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON }; 136 const enum vtype nv2tov_type[8] = 137 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 138 const enum vtype nv3tov_type[8] = 139 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 140 int nfs_ticks; 141 int nfs_commitsize; 142 143 MALLOC_DEFINE(M_NFSDIROFF, "NFS diroff", "NFS directory cookies"); 144 145 /* NFS client/server stats. */ 146 struct nfsstats nfsstats; 147 148 /* 149 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 150 */ 151 const int nfsv3_procid[NFS_NPROCS] = { 152 NFSPROC_NULL, 153 NFSPROC_GETATTR, 154 NFSPROC_SETATTR, 155 NFSPROC_NOOP, 156 NFSPROC_LOOKUP, 157 NFSPROC_READLINK, 158 NFSPROC_READ, 159 NFSPROC_NOOP, 160 NFSPROC_WRITE, 161 NFSPROC_CREATE, 162 NFSPROC_REMOVE, 163 NFSPROC_RENAME, 164 NFSPROC_LINK, 165 NFSPROC_SYMLINK, 166 NFSPROC_MKDIR, 167 NFSPROC_RMDIR, 168 NFSPROC_READDIR, 169 NFSPROC_FSSTAT, 170 NFSPROC_NOOP, 171 NFSPROC_NOOP, 172 NFSPROC_NOOP, 173 NFSPROC_NOOP, 174 NFSPROC_NOOP, 175 NFSPROC_NOOP, 176 NFSPROC_NOOP, 177 NFSPROC_NOOP 178 }; 179 180 /* 181 * and the reverse mapping from generic to Version 2 procedure numbers 182 */ 183 const int nfsv2_procid[NFS_NPROCS] = { 184 NFSV2PROC_NULL, 185 NFSV2PROC_GETATTR, 186 NFSV2PROC_SETATTR, 187 NFSV2PROC_LOOKUP, 188 NFSV2PROC_NOOP, 189 NFSV2PROC_READLINK, 190 NFSV2PROC_READ, 191 NFSV2PROC_WRITE, 192 NFSV2PROC_CREATE, 193 NFSV2PROC_MKDIR, 194 NFSV2PROC_SYMLINK, 195 NFSV2PROC_CREATE, 196 NFSV2PROC_REMOVE, 197 NFSV2PROC_RMDIR, 198 NFSV2PROC_RENAME, 199 NFSV2PROC_LINK, 200 NFSV2PROC_READDIR, 201 NFSV2PROC_NOOP, 202 NFSV2PROC_STATFS, 203 NFSV2PROC_NOOP, 204 NFSV2PROC_NOOP, 205 NFSV2PROC_NOOP, 206 NFSV2PROC_NOOP, 207 NFSV2PROC_NOOP, 208 NFSV2PROC_NOOP, 209 NFSV2PROC_NOOP, 210 }; 211 212 /* 213 * Maps errno values to nfs error numbers. 214 * Use NFSERR_IO as the catch all for ones not specifically defined in 215 * RFC 1094. 216 */ 217 static const u_char nfsrv_v2errmap[ELAST] = { 218 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 219 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 220 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 221 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 222 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 223 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 224 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 225 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 226 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 227 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 228 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 229 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 230 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, 231 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 232 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 233 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 234 NFSERR_IO, NFSERR_IO, 235 }; 236 237 /* 238 * Maps errno values to nfs error numbers. 239 * Although it is not obvious whether or not NFS clients really care if 240 * a returned error value is in the specified list for the procedure, the 241 * safest thing to do is filter them appropriately. For Version 2, the 242 * X/Open XNFS document is the only specification that defines error values 243 * for each RPC (The RFC simply lists all possible error values for all RPCs), 244 * so I have decided to not do this for Version 2. 245 * The first entry is the default error return and the rest are the valid 246 * errors for that RPC in increasing numeric order. 247 */ 248 static const short nfsv3err_null[] = { 249 0, 250 0, 251 }; 252 253 static const short nfsv3err_getattr[] = { 254 NFSERR_IO, 255 NFSERR_IO, 256 NFSERR_STALE, 257 NFSERR_BADHANDLE, 258 NFSERR_SERVERFAULT, 259 0, 260 }; 261 262 static const short nfsv3err_setattr[] = { 263 NFSERR_IO, 264 NFSERR_PERM, 265 NFSERR_IO, 266 NFSERR_ACCES, 267 NFSERR_INVAL, 268 NFSERR_NOSPC, 269 NFSERR_ROFS, 270 NFSERR_DQUOT, 271 NFSERR_STALE, 272 NFSERR_BADHANDLE, 273 NFSERR_NOT_SYNC, 274 NFSERR_SERVERFAULT, 275 0, 276 }; 277 278 static const short nfsv3err_lookup[] = { 279 NFSERR_IO, 280 NFSERR_NOENT, 281 NFSERR_IO, 282 NFSERR_ACCES, 283 NFSERR_NOTDIR, 284 NFSERR_NAMETOL, 285 NFSERR_STALE, 286 NFSERR_BADHANDLE, 287 NFSERR_SERVERFAULT, 288 0, 289 }; 290 291 static const short nfsv3err_access[] = { 292 NFSERR_IO, 293 NFSERR_IO, 294 NFSERR_STALE, 295 NFSERR_BADHANDLE, 296 NFSERR_SERVERFAULT, 297 0, 298 }; 299 300 static const short nfsv3err_readlink[] = { 301 NFSERR_IO, 302 NFSERR_IO, 303 NFSERR_ACCES, 304 NFSERR_INVAL, 305 NFSERR_STALE, 306 NFSERR_BADHANDLE, 307 NFSERR_NOTSUPP, 308 NFSERR_SERVERFAULT, 309 0, 310 }; 311 312 static const short nfsv3err_read[] = { 313 NFSERR_IO, 314 NFSERR_IO, 315 NFSERR_NXIO, 316 NFSERR_ACCES, 317 NFSERR_INVAL, 318 NFSERR_STALE, 319 NFSERR_BADHANDLE, 320 NFSERR_SERVERFAULT, 321 NFSERR_JUKEBOX, 322 0, 323 }; 324 325 static const short nfsv3err_write[] = { 326 NFSERR_IO, 327 NFSERR_IO, 328 NFSERR_ACCES, 329 NFSERR_INVAL, 330 NFSERR_FBIG, 331 NFSERR_NOSPC, 332 NFSERR_ROFS, 333 NFSERR_DQUOT, 334 NFSERR_STALE, 335 NFSERR_BADHANDLE, 336 NFSERR_SERVERFAULT, 337 NFSERR_JUKEBOX, 338 0, 339 }; 340 341 static const short nfsv3err_create[] = { 342 NFSERR_IO, 343 NFSERR_IO, 344 NFSERR_ACCES, 345 NFSERR_EXIST, 346 NFSERR_NOTDIR, 347 NFSERR_NOSPC, 348 NFSERR_ROFS, 349 NFSERR_NAMETOL, 350 NFSERR_DQUOT, 351 NFSERR_STALE, 352 NFSERR_BADHANDLE, 353 NFSERR_NOTSUPP, 354 NFSERR_SERVERFAULT, 355 0, 356 }; 357 358 static const short nfsv3err_mkdir[] = { 359 NFSERR_IO, 360 NFSERR_IO, 361 NFSERR_ACCES, 362 NFSERR_EXIST, 363 NFSERR_NOTDIR, 364 NFSERR_NOSPC, 365 NFSERR_ROFS, 366 NFSERR_NAMETOL, 367 NFSERR_DQUOT, 368 NFSERR_STALE, 369 NFSERR_BADHANDLE, 370 NFSERR_NOTSUPP, 371 NFSERR_SERVERFAULT, 372 0, 373 }; 374 375 static const short nfsv3err_symlink[] = { 376 NFSERR_IO, 377 NFSERR_IO, 378 NFSERR_ACCES, 379 NFSERR_EXIST, 380 NFSERR_NOTDIR, 381 NFSERR_NOSPC, 382 NFSERR_ROFS, 383 NFSERR_NAMETOL, 384 NFSERR_DQUOT, 385 NFSERR_STALE, 386 NFSERR_BADHANDLE, 387 NFSERR_NOTSUPP, 388 NFSERR_SERVERFAULT, 389 0, 390 }; 391 392 static const short nfsv3err_mknod[] = { 393 NFSERR_IO, 394 NFSERR_IO, 395 NFSERR_ACCES, 396 NFSERR_EXIST, 397 NFSERR_NOTDIR, 398 NFSERR_NOSPC, 399 NFSERR_ROFS, 400 NFSERR_NAMETOL, 401 NFSERR_DQUOT, 402 NFSERR_STALE, 403 NFSERR_BADHANDLE, 404 NFSERR_NOTSUPP, 405 NFSERR_SERVERFAULT, 406 NFSERR_BADTYPE, 407 0, 408 }; 409 410 static const short nfsv3err_remove[] = { 411 NFSERR_IO, 412 NFSERR_NOENT, 413 NFSERR_IO, 414 NFSERR_ACCES, 415 NFSERR_NOTDIR, 416 NFSERR_ROFS, 417 NFSERR_NAMETOL, 418 NFSERR_STALE, 419 NFSERR_BADHANDLE, 420 NFSERR_SERVERFAULT, 421 0, 422 }; 423 424 static const short nfsv3err_rmdir[] = { 425 NFSERR_IO, 426 NFSERR_NOENT, 427 NFSERR_IO, 428 NFSERR_ACCES, 429 NFSERR_EXIST, 430 NFSERR_NOTDIR, 431 NFSERR_INVAL, 432 NFSERR_ROFS, 433 NFSERR_NAMETOL, 434 NFSERR_NOTEMPTY, 435 NFSERR_STALE, 436 NFSERR_BADHANDLE, 437 NFSERR_NOTSUPP, 438 NFSERR_SERVERFAULT, 439 0, 440 }; 441 442 static const short nfsv3err_rename[] = { 443 NFSERR_IO, 444 NFSERR_NOENT, 445 NFSERR_IO, 446 NFSERR_ACCES, 447 NFSERR_EXIST, 448 NFSERR_XDEV, 449 NFSERR_NOTDIR, 450 NFSERR_ISDIR, 451 NFSERR_INVAL, 452 NFSERR_NOSPC, 453 NFSERR_ROFS, 454 NFSERR_MLINK, 455 NFSERR_NAMETOL, 456 NFSERR_NOTEMPTY, 457 NFSERR_DQUOT, 458 NFSERR_STALE, 459 NFSERR_BADHANDLE, 460 NFSERR_NOTSUPP, 461 NFSERR_SERVERFAULT, 462 0, 463 }; 464 465 static const short nfsv3err_link[] = { 466 NFSERR_IO, 467 NFSERR_IO, 468 NFSERR_ACCES, 469 NFSERR_EXIST, 470 NFSERR_XDEV, 471 NFSERR_NOTDIR, 472 NFSERR_INVAL, 473 NFSERR_NOSPC, 474 NFSERR_ROFS, 475 NFSERR_MLINK, 476 NFSERR_NAMETOL, 477 NFSERR_DQUOT, 478 NFSERR_STALE, 479 NFSERR_BADHANDLE, 480 NFSERR_NOTSUPP, 481 NFSERR_SERVERFAULT, 482 0, 483 }; 484 485 static const short nfsv3err_readdir[] = { 486 NFSERR_IO, 487 NFSERR_IO, 488 NFSERR_ACCES, 489 NFSERR_NOTDIR, 490 NFSERR_STALE, 491 NFSERR_BADHANDLE, 492 NFSERR_BAD_COOKIE, 493 NFSERR_TOOSMALL, 494 NFSERR_SERVERFAULT, 495 0, 496 }; 497 498 static const short nfsv3err_readdirplus[] = { 499 NFSERR_IO, 500 NFSERR_IO, 501 NFSERR_ACCES, 502 NFSERR_NOTDIR, 503 NFSERR_STALE, 504 NFSERR_BADHANDLE, 505 NFSERR_BAD_COOKIE, 506 NFSERR_NOTSUPP, 507 NFSERR_TOOSMALL, 508 NFSERR_SERVERFAULT, 509 0, 510 }; 511 512 static const short nfsv3err_fsstat[] = { 513 NFSERR_IO, 514 NFSERR_IO, 515 NFSERR_STALE, 516 NFSERR_BADHANDLE, 517 NFSERR_SERVERFAULT, 518 0, 519 }; 520 521 static const short nfsv3err_fsinfo[] = { 522 NFSERR_STALE, 523 NFSERR_STALE, 524 NFSERR_BADHANDLE, 525 NFSERR_SERVERFAULT, 526 0, 527 }; 528 529 static const short nfsv3err_pathconf[] = { 530 NFSERR_STALE, 531 NFSERR_STALE, 532 NFSERR_BADHANDLE, 533 NFSERR_SERVERFAULT, 534 0, 535 }; 536 537 static const short nfsv3err_commit[] = { 538 NFSERR_IO, 539 NFSERR_IO, 540 NFSERR_STALE, 541 NFSERR_BADHANDLE, 542 NFSERR_SERVERFAULT, 543 0, 544 }; 545 546 static const short * const nfsrv_v3errmap[] = { 547 nfsv3err_null, 548 nfsv3err_getattr, 549 nfsv3err_setattr, 550 nfsv3err_lookup, 551 nfsv3err_access, 552 nfsv3err_readlink, 553 nfsv3err_read, 554 nfsv3err_write, 555 nfsv3err_create, 556 nfsv3err_mkdir, 557 nfsv3err_symlink, 558 nfsv3err_mknod, 559 nfsv3err_remove, 560 nfsv3err_rmdir, 561 nfsv3err_rename, 562 nfsv3err_link, 563 nfsv3err_readdir, 564 nfsv3err_readdirplus, 565 nfsv3err_fsstat, 566 nfsv3err_fsinfo, 567 nfsv3err_pathconf, 568 nfsv3err_commit, 569 }; 570 571 extern struct nfsrtt nfsrtt; 572 extern time_t nqnfsstarttime; 573 extern int nqsrv_clockskew; 574 extern int nqsrv_writeslack; 575 extern int nqsrv_maxlease; 576 extern const int nqnfs_piggy[NFS_NPROCS]; 577 extern struct nfsnodehashhead *nfsnodehashtbl; 578 extern u_long nfsnodehash; 579 580 u_long nfsdirhashmask; 581 582 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 583 584 /* 585 * Create the header for an rpc request packet 586 * The hsiz is the size of the rest of the nfs request header. 587 * (just used to decide if a cluster is a good idea) 588 */ 589 struct mbuf * 590 nfsm_reqh(np, procid, hsiz, bposp) 591 struct nfsnode *np; 592 u_long procid; 593 int hsiz; 594 caddr_t *bposp; 595 { 596 struct mbuf *mb; 597 caddr_t bpos; 598 #ifndef NFS_V2_ONLY 599 struct nfsmount *nmp; 600 u_int32_t *tl; 601 int nqflag; 602 #endif 603 604 mb = m_get(M_WAIT, MT_DATA); 605 MCLAIM(mb, &nfs_mowner); 606 if (hsiz >= MINCLSIZE) 607 m_clget(mb, M_WAIT); 608 mb->m_len = 0; 609 bpos = mtod(mb, caddr_t); 610 611 #ifndef NFS_V2_ONLY 612 /* 613 * For NQNFS, add lease request. 614 */ 615 if (np) { 616 nmp = VFSTONFS(np->n_vnode->v_mount); 617 if (nmp->nm_flag & NFSMNT_NQNFS) { 618 nqflag = NQNFS_NEEDLEASE(np, procid); 619 if (nqflag) { 620 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); 621 *tl++ = txdr_unsigned(nqflag); 622 *tl = txdr_unsigned(nmp->nm_leaseterm); 623 } else { 624 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 625 *tl = 0; 626 } 627 } 628 } 629 #endif 630 /* Finally, return values */ 631 *bposp = bpos; 632 return (mb); 633 } 634 635 /* 636 * Build the RPC header and fill in the authorization info. 637 * The authorization string argument is only used when the credentials 638 * come from outside of the kernel. 639 * Returns the head of the mbuf list. 640 */ 641 struct mbuf * 642 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 643 verf_str, mrest, mrest_len, mbp, xidp) 644 struct ucred *cr; 645 int nmflag; 646 int procid; 647 int auth_type; 648 int auth_len; 649 char *auth_str; 650 int verf_len; 651 char *verf_str; 652 struct mbuf *mrest; 653 int mrest_len; 654 struct mbuf **mbp; 655 u_int32_t *xidp; 656 { 657 struct mbuf *mb; 658 u_int32_t *tl; 659 caddr_t bpos; 660 int i; 661 struct mbuf *mreq; 662 int siz, grpsiz, authsiz; 663 664 authsiz = nfsm_rndup(auth_len); 665 mb = m_gethdr(M_WAIT, MT_DATA); 666 MCLAIM(mb, &nfs_mowner); 667 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 668 m_clget(mb, M_WAIT); 669 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 670 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 671 } else { 672 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 673 } 674 mb->m_len = 0; 675 mreq = mb; 676 bpos = mtod(mb, caddr_t); 677 678 /* 679 * First the RPC header. 680 */ 681 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 682 683 *tl++ = *xidp = nfs_getxid(); 684 *tl++ = rpc_call; 685 *tl++ = rpc_vers; 686 if (nmflag & NFSMNT_NQNFS) { 687 *tl++ = txdr_unsigned(NQNFS_PROG); 688 *tl++ = txdr_unsigned(NQNFS_VER3); 689 } else { 690 *tl++ = txdr_unsigned(NFS_PROG); 691 if (nmflag & NFSMNT_NFSV3) 692 *tl++ = txdr_unsigned(NFS_VER3); 693 else 694 *tl++ = txdr_unsigned(NFS_VER2); 695 } 696 if (nmflag & NFSMNT_NFSV3) 697 *tl++ = txdr_unsigned(procid); 698 else 699 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 700 701 /* 702 * And then the authorization cred. 703 */ 704 *tl++ = txdr_unsigned(auth_type); 705 *tl = txdr_unsigned(authsiz); 706 switch (auth_type) { 707 case RPCAUTH_UNIX: 708 nfsm_build(tl, u_int32_t *, auth_len); 709 *tl++ = 0; /* stamp ?? */ 710 *tl++ = 0; /* NULL hostname */ 711 *tl++ = txdr_unsigned(cr->cr_uid); 712 *tl++ = txdr_unsigned(cr->cr_gid); 713 grpsiz = (auth_len >> 2) - 5; 714 *tl++ = txdr_unsigned(grpsiz); 715 for (i = 0; i < grpsiz; i++) 716 *tl++ = txdr_unsigned(cr->cr_groups[i]); 717 break; 718 case RPCAUTH_KERB4: 719 siz = auth_len; 720 while (siz > 0) { 721 if (M_TRAILINGSPACE(mb) == 0) { 722 struct mbuf *mb2; 723 mb2 = m_get(M_WAIT, MT_DATA); 724 MCLAIM(mb2, &nfs_mowner); 725 if (siz >= MINCLSIZE) 726 m_clget(mb2, M_WAIT); 727 mb->m_next = mb2; 728 mb = mb2; 729 mb->m_len = 0; 730 bpos = mtod(mb, caddr_t); 731 } 732 i = min(siz, M_TRAILINGSPACE(mb)); 733 memcpy(bpos, auth_str, i); 734 mb->m_len += i; 735 auth_str += i; 736 bpos += i; 737 siz -= i; 738 } 739 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 740 for (i = 0; i < siz; i++) 741 *bpos++ = '\0'; 742 mb->m_len += siz; 743 } 744 break; 745 }; 746 747 /* 748 * And the verifier... 749 */ 750 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 751 if (verf_str) { 752 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 753 *tl = txdr_unsigned(verf_len); 754 siz = verf_len; 755 while (siz > 0) { 756 if (M_TRAILINGSPACE(mb) == 0) { 757 struct mbuf *mb2; 758 mb2 = m_get(M_WAIT, MT_DATA); 759 MCLAIM(mb2, &nfs_mowner); 760 if (siz >= MINCLSIZE) 761 m_clget(mb2, M_WAIT); 762 mb->m_next = mb2; 763 mb = mb2; 764 mb->m_len = 0; 765 bpos = mtod(mb, caddr_t); 766 } 767 i = min(siz, M_TRAILINGSPACE(mb)); 768 memcpy(bpos, verf_str, i); 769 mb->m_len += i; 770 verf_str += i; 771 bpos += i; 772 siz -= i; 773 } 774 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 775 for (i = 0; i < siz; i++) 776 *bpos++ = '\0'; 777 mb->m_len += siz; 778 } 779 } else { 780 *tl++ = txdr_unsigned(RPCAUTH_NULL); 781 *tl = 0; 782 } 783 mb->m_next = mrest; 784 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 785 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 786 *mbp = mb; 787 return (mreq); 788 } 789 790 /* 791 * copies mbuf chain to the uio scatter/gather list 792 */ 793 int 794 nfsm_mbuftouio(mrep, uiop, siz, dpos) 795 struct mbuf **mrep; 796 struct uio *uiop; 797 int siz; 798 caddr_t *dpos; 799 { 800 char *mbufcp, *uiocp; 801 int xfer, left, len; 802 struct mbuf *mp; 803 long uiosiz, rem; 804 int error = 0; 805 806 mp = *mrep; 807 mbufcp = *dpos; 808 len = mtod(mp, caddr_t)+mp->m_len-mbufcp; 809 rem = nfsm_rndup(siz)-siz; 810 while (siz > 0) { 811 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 812 return (EFBIG); 813 left = uiop->uio_iov->iov_len; 814 uiocp = uiop->uio_iov->iov_base; 815 if (left > siz) 816 left = siz; 817 uiosiz = left; 818 while (left > 0) { 819 while (len == 0) { 820 mp = mp->m_next; 821 if (mp == NULL) 822 return (EBADRPC); 823 mbufcp = mtod(mp, caddr_t); 824 len = mp->m_len; 825 } 826 xfer = (left > len) ? len : left; 827 #ifdef notdef 828 /* Not Yet.. */ 829 if (uiop->uio_iov->iov_op != NULL) 830 (*(uiop->uio_iov->iov_op)) 831 (mbufcp, uiocp, xfer); 832 else 833 #endif 834 if (uiop->uio_segflg == UIO_SYSSPACE) 835 memcpy(uiocp, mbufcp, xfer); 836 else 837 copyout(mbufcp, uiocp, xfer); 838 left -= xfer; 839 len -= xfer; 840 mbufcp += xfer; 841 uiocp += xfer; 842 uiop->uio_offset += xfer; 843 uiop->uio_resid -= xfer; 844 } 845 if (uiop->uio_iov->iov_len <= siz) { 846 uiop->uio_iovcnt--; 847 uiop->uio_iov++; 848 } else { 849 uiop->uio_iov->iov_base = 850 (caddr_t)uiop->uio_iov->iov_base + uiosiz; 851 uiop->uio_iov->iov_len -= uiosiz; 852 } 853 siz -= uiosiz; 854 } 855 *dpos = mbufcp; 856 *mrep = mp; 857 if (rem > 0) { 858 if (len < rem) 859 error = nfs_adv(mrep, dpos, rem, len); 860 else 861 *dpos += rem; 862 } 863 return (error); 864 } 865 866 /* 867 * copies a uio scatter/gather list to an mbuf chain. 868 * NOTE: can ony handle iovcnt == 1 869 */ 870 int 871 nfsm_uiotombuf(uiop, mq, siz, bpos) 872 struct uio *uiop; 873 struct mbuf **mq; 874 int siz; 875 caddr_t *bpos; 876 { 877 char *uiocp; 878 struct mbuf *mp, *mp2; 879 int xfer, left, mlen; 880 int uiosiz, clflg, rem; 881 char *cp; 882 883 #ifdef DIAGNOSTIC 884 if (uiop->uio_iovcnt != 1) 885 panic("nfsm_uiotombuf: iovcnt != 1"); 886 #endif 887 888 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 889 clflg = 1; 890 else 891 clflg = 0; 892 rem = nfsm_rndup(siz)-siz; 893 mp = mp2 = *mq; 894 while (siz > 0) { 895 left = uiop->uio_iov->iov_len; 896 uiocp = uiop->uio_iov->iov_base; 897 if (left > siz) 898 left = siz; 899 uiosiz = left; 900 while (left > 0) { 901 mlen = M_TRAILINGSPACE(mp); 902 if (mlen == 0) { 903 mp = m_get(M_WAIT, MT_DATA); 904 MCLAIM(mp, &nfs_mowner); 905 if (clflg) 906 m_clget(mp, M_WAIT); 907 mp->m_len = 0; 908 mp2->m_next = mp; 909 mp2 = mp; 910 mlen = M_TRAILINGSPACE(mp); 911 } 912 xfer = (left > mlen) ? mlen : left; 913 #ifdef notdef 914 /* Not Yet.. */ 915 if (uiop->uio_iov->iov_op != NULL) 916 (*(uiop->uio_iov->iov_op)) 917 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 918 else 919 #endif 920 if (uiop->uio_segflg == UIO_SYSSPACE) 921 memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer); 922 else 923 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 924 mp->m_len += xfer; 925 left -= xfer; 926 uiocp += xfer; 927 uiop->uio_offset += xfer; 928 uiop->uio_resid -= xfer; 929 } 930 uiop->uio_iov->iov_base = (caddr_t)uiop->uio_iov->iov_base + 931 uiosiz; 932 uiop->uio_iov->iov_len -= uiosiz; 933 siz -= uiosiz; 934 } 935 if (rem > 0) { 936 if (rem > M_TRAILINGSPACE(mp)) { 937 mp = m_get(M_WAIT, MT_DATA); 938 MCLAIM(mp, &nfs_mowner); 939 mp->m_len = 0; 940 mp2->m_next = mp; 941 } 942 cp = mtod(mp, caddr_t)+mp->m_len; 943 for (left = 0; left < rem; left++) 944 *cp++ = '\0'; 945 mp->m_len += rem; 946 *bpos = cp; 947 } else 948 *bpos = mtod(mp, caddr_t)+mp->m_len; 949 *mq = mp; 950 return (0); 951 } 952 953 /* 954 * Get at least "siz" bytes of correctly aligned data. 955 * When called the mbuf pointers are not necessarily correct, 956 * dsosp points to what ought to be in m_data and left contains 957 * what ought to be in m_len. 958 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 959 * cases. (The macros use the vars. dpos and dpos2) 960 */ 961 int 962 nfsm_disct(mdp, dposp, siz, left, cp2) 963 struct mbuf **mdp; 964 caddr_t *dposp; 965 int siz; 966 int left; 967 caddr_t *cp2; 968 { 969 struct mbuf *m1, *m2; 970 struct mbuf *havebuf = NULL; 971 caddr_t src = *dposp; 972 caddr_t dst; 973 int len; 974 975 #ifdef DEBUG 976 if (left < 0) 977 panic("nfsm_disct: left < 0"); 978 #endif 979 m1 = *mdp; 980 /* 981 * Skip through the mbuf chain looking for an mbuf with 982 * some data. If the first mbuf found has enough data 983 * and it is correctly aligned return it. 984 */ 985 while (left == 0) { 986 havebuf = m1; 987 *mdp = m1 = m1->m_next; 988 if (m1 == NULL) 989 return (EBADRPC); 990 src = mtod(m1, caddr_t); 991 left = m1->m_len; 992 /* 993 * If we start a new mbuf and it is big enough 994 * and correctly aligned just return it, don't 995 * do any pull up. 996 */ 997 if (left >= siz && nfsm_aligned(src)) { 998 *cp2 = src; 999 *dposp = src + siz; 1000 return (0); 1001 } 1002 } 1003 if (m1->m_flags & M_EXT) { 1004 if (havebuf) { 1005 /* If the first mbuf with data has external data 1006 * and there is a previous empty mbuf use it 1007 * to move the data into. 1008 */ 1009 m2 = m1; 1010 *mdp = m1 = havebuf; 1011 if (m1->m_flags & M_EXT) { 1012 MEXTREMOVE(m1); 1013 } 1014 } else { 1015 /* 1016 * If the first mbuf has a external data 1017 * and there is no previous empty mbuf 1018 * allocate a new mbuf and move the external 1019 * data to the new mbuf. Also make the first 1020 * mbuf look empty. 1021 */ 1022 m2 = m_get(M_WAIT, MT_DATA); 1023 m2->m_ext = m1->m_ext; 1024 m2->m_data = src; 1025 m2->m_len = left; 1026 MCLADDREFERENCE(m1, m2); 1027 MEXTREMOVE(m1); 1028 m2->m_next = m1->m_next; 1029 m1->m_next = m2; 1030 } 1031 m1->m_len = 0; 1032 if (m1->m_flags & M_PKTHDR) 1033 dst = m1->m_pktdat; 1034 else 1035 dst = m1->m_dat; 1036 m1->m_data = dst; 1037 } else { 1038 /* 1039 * If the first mbuf has no external data 1040 * move the data to the front of the mbuf. 1041 */ 1042 if (m1->m_flags & M_PKTHDR) 1043 dst = m1->m_pktdat; 1044 else 1045 dst = m1->m_dat; 1046 m1->m_data = dst; 1047 if (dst != src) 1048 memmove(dst, src, left); 1049 dst += left; 1050 m1->m_len = left; 1051 m2 = m1->m_next; 1052 } 1053 *cp2 = m1->m_data; 1054 *dposp = mtod(m1, caddr_t) + siz; 1055 /* 1056 * Loop through mbufs pulling data up into first mbuf until 1057 * the first mbuf is full or there is no more data to 1058 * pullup. 1059 */ 1060 while ((len = M_TRAILINGSPACE(m1)) != 0 && m2) { 1061 if ((len = min(len, m2->m_len)) != 0) 1062 memcpy(dst, m2->m_data, len); 1063 m1->m_len += len; 1064 dst += len; 1065 m2->m_data += len; 1066 m2->m_len -= len; 1067 m2 = m2->m_next; 1068 } 1069 if (m1->m_len < siz) 1070 return (EBADRPC); 1071 return (0); 1072 } 1073 1074 /* 1075 * Advance the position in the mbuf chain. 1076 */ 1077 int 1078 nfs_adv(mdp, dposp, offs, left) 1079 struct mbuf **mdp; 1080 caddr_t *dposp; 1081 int offs; 1082 int left; 1083 { 1084 struct mbuf *m; 1085 int s; 1086 1087 m = *mdp; 1088 s = left; 1089 while (s < offs) { 1090 offs -= s; 1091 m = m->m_next; 1092 if (m == NULL) 1093 return (EBADRPC); 1094 s = m->m_len; 1095 } 1096 *mdp = m; 1097 *dposp = mtod(m, caddr_t)+offs; 1098 return (0); 1099 } 1100 1101 /* 1102 * Copy a string into mbufs for the hard cases... 1103 */ 1104 int 1105 nfsm_strtmbuf(mb, bpos, cp, siz) 1106 struct mbuf **mb; 1107 char **bpos; 1108 const char *cp; 1109 long siz; 1110 { 1111 struct mbuf *m1 = NULL, *m2; 1112 long left, xfer, len, tlen; 1113 u_int32_t *tl; 1114 int putsize; 1115 1116 putsize = 1; 1117 m2 = *mb; 1118 left = M_TRAILINGSPACE(m2); 1119 if (left > 0) { 1120 tl = ((u_int32_t *)(*bpos)); 1121 *tl++ = txdr_unsigned(siz); 1122 putsize = 0; 1123 left -= NFSX_UNSIGNED; 1124 m2->m_len += NFSX_UNSIGNED; 1125 if (left > 0) { 1126 memcpy((caddr_t) tl, cp, left); 1127 siz -= left; 1128 cp += left; 1129 m2->m_len += left; 1130 left = 0; 1131 } 1132 } 1133 /* Loop around adding mbufs */ 1134 while (siz > 0) { 1135 m1 = m_get(M_WAIT, MT_DATA); 1136 MCLAIM(m1, &nfs_mowner); 1137 if (siz > MLEN) 1138 m_clget(m1, M_WAIT); 1139 m1->m_len = NFSMSIZ(m1); 1140 m2->m_next = m1; 1141 m2 = m1; 1142 tl = mtod(m1, u_int32_t *); 1143 tlen = 0; 1144 if (putsize) { 1145 *tl++ = txdr_unsigned(siz); 1146 m1->m_len -= NFSX_UNSIGNED; 1147 tlen = NFSX_UNSIGNED; 1148 putsize = 0; 1149 } 1150 if (siz < m1->m_len) { 1151 len = nfsm_rndup(siz); 1152 xfer = siz; 1153 if (xfer < len) 1154 *(tl+(xfer>>2)) = 0; 1155 } else { 1156 xfer = len = m1->m_len; 1157 } 1158 memcpy((caddr_t) tl, cp, xfer); 1159 m1->m_len = len+tlen; 1160 siz -= xfer; 1161 cp += xfer; 1162 } 1163 *mb = m1; 1164 *bpos = mtod(m1, caddr_t)+m1->m_len; 1165 return (0); 1166 } 1167 1168 /* 1169 * Directory caching routines. They work as follows: 1170 * - a cache is maintained per VDIR nfsnode. 1171 * - for each offset cookie that is exported to userspace, and can 1172 * thus be thrown back at us as an offset to VOP_READDIR, store 1173 * information in the cache. 1174 * - cached are: 1175 * - cookie itself 1176 * - blocknumber (essentially just a search key in the buffer cache) 1177 * - entry number in block. 1178 * - offset cookie of block in which this entry is stored 1179 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1180 * - entries are looked up in a hash table 1181 * - also maintained is an LRU list of entries, used to determine 1182 * which ones to delete if the cache grows too large. 1183 * - if 32 <-> 64 translation mode is requested for a filesystem, 1184 * the cache also functions as a translation table 1185 * - in the translation case, invalidating the cache does not mean 1186 * flushing it, but just marking entries as invalid, except for 1187 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1188 * still be able to use the cache as a translation table. 1189 * - 32 bit cookies are uniquely created by combining the hash table 1190 * entry value, and one generation count per hash table entry, 1191 * incremented each time an entry is appended to the chain. 1192 * - the cache is invalidated each time a direcory is modified 1193 * - sanity checks are also done; if an entry in a block turns 1194 * out not to have a matching cookie, the cache is invalidated 1195 * and a new block starting from the wanted offset is fetched from 1196 * the server. 1197 * - directory entries as read from the server are extended to contain 1198 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1199 * the cache and exporting them to userspace through the cookie 1200 * argument to VOP_READDIR. 1201 */ 1202 1203 u_long 1204 nfs_dirhash(off) 1205 off_t off; 1206 { 1207 int i; 1208 char *cp = (char *)&off; 1209 u_long sum = 0L; 1210 1211 for (i = 0 ; i < sizeof (off); i++) 1212 sum += *cp++; 1213 1214 return sum; 1215 } 1216 1217 #define _NFSDC_MTX(np) (&NFSTOV(np)->v_interlock) 1218 #define NFSDC_LOCK(np) simple_lock(_NFSDC_MTX(np)) 1219 #define NFSDC_UNLOCK(np) simple_unlock(_NFSDC_MTX(np)) 1220 #define NFSDC_ASSERT_LOCKED(np) LOCK_ASSERT(simple_lock_held(_NFSDC_MTX(np))) 1221 1222 void 1223 nfs_initdircache(vp) 1224 struct vnode *vp; 1225 { 1226 struct nfsnode *np = VTONFS(vp); 1227 struct nfsdirhashhead *dircache; 1228 1229 dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, M_NFSDIROFF, 1230 M_WAITOK, &nfsdirhashmask); 1231 1232 NFSDC_LOCK(np); 1233 if (np->n_dircache == NULL) { 1234 np->n_dircachesize = 0; 1235 np->n_dircache = dircache; 1236 dircache = NULL; 1237 TAILQ_INIT(&np->n_dirchain); 1238 } 1239 NFSDC_UNLOCK(np); 1240 if (dircache) 1241 hashdone(dircache, M_NFSDIROFF); 1242 } 1243 1244 void 1245 nfs_initdirxlatecookie(vp) 1246 struct vnode *vp; 1247 { 1248 struct nfsnode *np = VTONFS(vp); 1249 unsigned *dirgens; 1250 1251 KASSERT(VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_XLATECOOKIE); 1252 1253 dirgens = malloc(NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, 1254 M_WAITOK|M_ZERO); 1255 NFSDC_LOCK(np); 1256 if (np->n_dirgens == NULL) { 1257 np->n_dirgens = dirgens; 1258 dirgens = NULL; 1259 } 1260 NFSDC_UNLOCK(np); 1261 if (dirgens) 1262 free(dirgens, M_NFSDIROFF); 1263 } 1264 1265 static const struct nfsdircache dzero; 1266 1267 static void nfs_unlinkdircache __P((struct nfsnode *np, struct nfsdircache *)); 1268 static void nfs_putdircache_unlocked __P((struct nfsnode *, 1269 struct nfsdircache *)); 1270 1271 static void 1272 nfs_unlinkdircache(np, ndp) 1273 struct nfsnode *np; 1274 struct nfsdircache *ndp; 1275 { 1276 1277 NFSDC_ASSERT_LOCKED(np); 1278 KASSERT(ndp != &dzero); 1279 1280 if (LIST_NEXT(ndp, dc_hash) == (void *)-1) 1281 return; 1282 1283 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1284 LIST_REMOVE(ndp, dc_hash); 1285 LIST_NEXT(ndp, dc_hash) = (void *)-1; /* mark as unlinked */ 1286 1287 nfs_putdircache_unlocked(np, ndp); 1288 } 1289 1290 void 1291 nfs_putdircache(np, ndp) 1292 struct nfsnode *np; 1293 struct nfsdircache *ndp; 1294 { 1295 int ref; 1296 1297 if (ndp == &dzero) 1298 return; 1299 1300 KASSERT(ndp->dc_refcnt > 0); 1301 NFSDC_LOCK(np); 1302 ref = --ndp->dc_refcnt; 1303 NFSDC_UNLOCK(np); 1304 1305 if (ref == 0) 1306 free(ndp, M_NFSDIROFF); 1307 } 1308 1309 static void 1310 nfs_putdircache_unlocked(np, ndp) 1311 struct nfsnode *np; 1312 struct nfsdircache *ndp; 1313 { 1314 int ref; 1315 1316 NFSDC_ASSERT_LOCKED(np); 1317 1318 if (ndp == &dzero) 1319 return; 1320 1321 KASSERT(ndp->dc_refcnt > 0); 1322 ref = --ndp->dc_refcnt; 1323 if (ref == 0) 1324 free(ndp, M_NFSDIROFF); 1325 } 1326 1327 struct nfsdircache * 1328 nfs_searchdircache(vp, off, do32, hashent) 1329 struct vnode *vp; 1330 off_t off; 1331 int do32; 1332 int *hashent; 1333 { 1334 struct nfsdirhashhead *ndhp; 1335 struct nfsdircache *ndp = NULL; 1336 struct nfsnode *np = VTONFS(vp); 1337 unsigned ent; 1338 1339 /* 1340 * Zero is always a valid cookie. 1341 */ 1342 if (off == 0) 1343 /* LINTED const cast away */ 1344 return (struct nfsdircache *)&dzero; 1345 1346 if (!np->n_dircache) 1347 return NULL; 1348 1349 /* 1350 * We use a 32bit cookie as search key, directly reconstruct 1351 * the hashentry. Else use the hashfunction. 1352 */ 1353 if (do32) { 1354 ent = (u_int32_t)off >> 24; 1355 if (ent >= NFS_DIRHASHSIZ) 1356 return NULL; 1357 ndhp = &np->n_dircache[ent]; 1358 } else { 1359 ndhp = NFSDIRHASH(np, off); 1360 } 1361 1362 if (hashent) 1363 *hashent = (int)(ndhp - np->n_dircache); 1364 1365 NFSDC_LOCK(np); 1366 if (do32) { 1367 LIST_FOREACH(ndp, ndhp, dc_hash) { 1368 if (ndp->dc_cookie32 == (u_int32_t)off) { 1369 /* 1370 * An invalidated entry will become the 1371 * start of a new block fetched from 1372 * the server. 1373 */ 1374 if (ndp->dc_flags & NFSDC_INVALID) { 1375 ndp->dc_blkcookie = ndp->dc_cookie; 1376 ndp->dc_entry = 0; 1377 ndp->dc_flags &= ~NFSDC_INVALID; 1378 } 1379 break; 1380 } 1381 } 1382 } else { 1383 LIST_FOREACH(ndp, ndhp, dc_hash) { 1384 if (ndp->dc_cookie == off) 1385 break; 1386 } 1387 } 1388 if (ndp != NULL) 1389 ndp->dc_refcnt++; 1390 NFSDC_UNLOCK(np); 1391 return ndp; 1392 } 1393 1394 1395 struct nfsdircache * 1396 nfs_enterdircache(vp, off, blkoff, en, blkno) 1397 struct vnode *vp; 1398 off_t off, blkoff; 1399 int en; 1400 daddr_t blkno; 1401 { 1402 struct nfsnode *np = VTONFS(vp); 1403 struct nfsdirhashhead *ndhp; 1404 struct nfsdircache *ndp = NULL; 1405 struct nfsdircache *newndp = NULL; 1406 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1407 int hashent, gen, overwrite; 1408 1409 /* 1410 * XXX refuse entries for offset 0. amd(8) erroneously sets 1411 * cookie 0 for the '.' entry, making this necessary. This 1412 * isn't so bad, as 0 is a special case anyway. 1413 */ 1414 if (off == 0) 1415 /* LINTED const cast away */ 1416 return (struct nfsdircache *)&dzero; 1417 1418 if (!np->n_dircache) 1419 /* 1420 * XXX would like to do this in nfs_nget but vtype 1421 * isn't known at that time. 1422 */ 1423 nfs_initdircache(vp); 1424 1425 if ((nmp->nm_flag & NFSMNT_XLATECOOKIE) && !np->n_dirgens) 1426 nfs_initdirxlatecookie(vp); 1427 1428 retry: 1429 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1430 1431 NFSDC_LOCK(np); 1432 if (ndp && (ndp->dc_flags & NFSDC_INVALID) == 0) { 1433 /* 1434 * Overwriting an old entry. Check if it's the same. 1435 * If so, just return. If not, remove the old entry. 1436 */ 1437 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1438 goto done; 1439 nfs_unlinkdircache(np, ndp); 1440 nfs_putdircache_unlocked(np, ndp); 1441 ndp = NULL; 1442 } 1443 1444 ndhp = &np->n_dircache[hashent]; 1445 1446 if (!ndp) { 1447 if (newndp == NULL) { 1448 NFSDC_UNLOCK(np); 1449 newndp = malloc(sizeof(*ndp), M_NFSDIROFF, M_WAITOK); 1450 newndp->dc_refcnt = 1; 1451 LIST_NEXT(newndp, dc_hash) = (void *)-1; 1452 goto retry; 1453 } 1454 ndp = newndp; 1455 newndp = NULL; 1456 overwrite = 0; 1457 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1458 /* 1459 * We're allocating a new entry, so bump the 1460 * generation number. 1461 */ 1462 gen = ++np->n_dirgens[hashent]; 1463 if (gen == 0) { 1464 np->n_dirgens[hashent]++; 1465 gen++; 1466 } 1467 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1468 } 1469 } else 1470 overwrite = 1; 1471 1472 ndp->dc_cookie = off; 1473 ndp->dc_blkcookie = blkoff; 1474 ndp->dc_entry = en; 1475 ndp->dc_flags = 0; 1476 1477 if (overwrite) 1478 goto done; 1479 1480 /* 1481 * If the maximum directory cookie cache size has been reached 1482 * for this node, take one off the front. The idea is that 1483 * directories are typically read front-to-back once, so that 1484 * the oldest entries can be thrown away without much performance 1485 * loss. 1486 */ 1487 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1488 nfs_unlinkdircache(np, TAILQ_FIRST(&np->n_dirchain)); 1489 } else 1490 np->n_dircachesize++; 1491 1492 KASSERT(ndp->dc_refcnt == 1); 1493 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1494 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1495 ndp->dc_refcnt++; 1496 done: 1497 KASSERT(ndp->dc_refcnt > 0); 1498 NFSDC_UNLOCK(np); 1499 if (newndp) 1500 nfs_putdircache(np, newndp); 1501 return ndp; 1502 } 1503 1504 void 1505 nfs_invaldircache(vp, forcefree) 1506 struct vnode *vp; 1507 int forcefree; 1508 { 1509 struct nfsnode *np = VTONFS(vp); 1510 struct nfsdircache *ndp = NULL; 1511 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1512 1513 #ifdef DIAGNOSTIC 1514 if (vp->v_type != VDIR) 1515 panic("nfs: invaldircache: not dir"); 1516 #endif 1517 1518 np->n_flag &= ~NEOFVALID; 1519 1520 if (!np->n_dircache) 1521 return; 1522 1523 NFSDC_LOCK(np); 1524 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1525 while ((ndp = TAILQ_FIRST(&np->n_dirchain)) != NULL) { 1526 KASSERT(!forcefree || ndp->dc_refcnt == 1); 1527 nfs_unlinkdircache(np, ndp); 1528 } 1529 np->n_dircachesize = 0; 1530 if (forcefree && np->n_dirgens) { 1531 FREE(np->n_dirgens, M_NFSDIROFF); 1532 np->n_dirgens = NULL; 1533 } 1534 } else { 1535 TAILQ_FOREACH(ndp, &np->n_dirchain, dc_chain) 1536 ndp->dc_flags |= NFSDC_INVALID; 1537 } 1538 1539 NFSDC_UNLOCK(np); 1540 } 1541 1542 /* 1543 * Called once before VFS init to initialize shared and 1544 * server-specific data structures. 1545 */ 1546 void 1547 nfs_init() 1548 { 1549 nfsrtt.pos = 0; 1550 rpc_vers = txdr_unsigned(RPC_VER2); 1551 rpc_call = txdr_unsigned(RPC_CALL); 1552 rpc_reply = txdr_unsigned(RPC_REPLY); 1553 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1554 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1555 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1556 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1557 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1558 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1559 nfs_prog = txdr_unsigned(NFS_PROG); 1560 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1561 nfs_true = txdr_unsigned(TRUE); 1562 nfs_false = txdr_unsigned(FALSE); 1563 nfs_xdrneg1 = txdr_unsigned(-1); 1564 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1565 if (nfs_ticks < 1) 1566 nfs_ticks = 1; 1567 #ifdef NFSSERVER 1568 nfsrv_init(0); /* Init server data structures */ 1569 nfsrv_initcache(); /* Init the server request cache */ 1570 pool_init(&nfs_srvdesc_pool, sizeof(struct nfsrv_descript), 1571 0, 0, 0, "nfsrvdescpl", &pool_allocator_nointr); 1572 #endif /* NFSSERVER */ 1573 1574 #if defined(NFSSERVER) || !defined(NFS_V2_ONLY) 1575 /* 1576 * Initialize the nqnfs data structures. 1577 */ 1578 if (nqnfsstarttime == 0) { 1579 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1580 + nqsrv_clockskew + nqsrv_writeslack; 1581 NQLOADNOVRAM(nqnfsstarttime); 1582 CIRCLEQ_INIT(&nqtimerhead); 1583 nqfhhashtbl = hashinit(NQLCHSZ, HASH_LIST, M_NQLEASE, 1584 M_WAITOK, &nqfhhash); 1585 } 1586 #endif 1587 1588 exithook_establish(nfs_exit, NULL); 1589 1590 /* 1591 * Initialize reply list and start timer 1592 */ 1593 TAILQ_INIT(&nfs_reqq); 1594 nfs_timer(NULL); 1595 MOWNER_ATTACH(&nfs_mowner); 1596 1597 #ifdef NFS 1598 /* Initialize the kqueue structures */ 1599 nfs_kqinit(); 1600 /* Initialize the iod structures */ 1601 nfs_iodinit(); 1602 #endif 1603 } 1604 1605 #ifdef NFS 1606 /* 1607 * Called once at VFS init to initialize client-specific data structures. 1608 */ 1609 void 1610 nfs_vfs_init() 1611 { 1612 nfs_nhinit(); /* Init the nfsnode table */ 1613 nfs_commitsize = uvmexp.npages << (PAGE_SHIFT - 4); 1614 } 1615 1616 void 1617 nfs_vfs_reinit() 1618 { 1619 nfs_nhreinit(); 1620 } 1621 1622 void 1623 nfs_vfs_done() 1624 { 1625 nfs_nhdone(); 1626 } 1627 1628 /* 1629 * Attribute cache routines. 1630 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1631 * that are on the mbuf list 1632 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1633 * error otherwise 1634 */ 1635 1636 /* 1637 * Load the attribute cache (that lives in the nfsnode entry) with 1638 * the values on the mbuf list and 1639 * Iff vap not NULL 1640 * copy the attributes to *vaper 1641 */ 1642 int 1643 nfsm_loadattrcache(vpp, mdp, dposp, vaper, flags) 1644 struct vnode **vpp; 1645 struct mbuf **mdp; 1646 caddr_t *dposp; 1647 struct vattr *vaper; 1648 int flags; 1649 { 1650 int32_t t1; 1651 caddr_t cp2; 1652 int error = 0; 1653 struct mbuf *md; 1654 int v3 = NFS_ISV3(*vpp); 1655 1656 md = *mdp; 1657 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1658 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1659 if (error) 1660 return (error); 1661 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper, flags); 1662 } 1663 1664 int 1665 nfs_loadattrcache(vpp, fp, vaper, flags) 1666 struct vnode **vpp; 1667 struct nfs_fattr *fp; 1668 struct vattr *vaper; 1669 int flags; 1670 { 1671 struct vnode *vp = *vpp; 1672 struct vattr *vap; 1673 int v3 = NFS_ISV3(vp); 1674 enum vtype vtyp; 1675 u_short vmode; 1676 struct timespec mtime; 1677 struct timespec ctime; 1678 struct vnode *nvp; 1679 int32_t rdev; 1680 struct nfsnode *np; 1681 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1682 uid_t uid; 1683 gid_t gid; 1684 1685 if (v3) { 1686 vtyp = nfsv3tov_type(fp->fa_type); 1687 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1688 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1689 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1690 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1691 fxdr_nfsv3time(&fp->fa3_ctime, &ctime); 1692 } else { 1693 vtyp = nfsv2tov_type(fp->fa_type); 1694 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1695 if (vtyp == VNON || vtyp == VREG) 1696 vtyp = IFTOVT(vmode); 1697 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1698 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1699 ctime.tv_sec = fxdr_unsigned(u_int32_t, 1700 fp->fa2_ctime.nfsv2_sec); 1701 ctime.tv_nsec = 0; 1702 1703 /* 1704 * Really ugly NFSv2 kludge. 1705 */ 1706 if (vtyp == VCHR && rdev == 0xffffffff) 1707 vtyp = VFIFO; 1708 } 1709 1710 vmode &= ALLPERMS; 1711 1712 /* 1713 * If v_type == VNON it is a new node, so fill in the v_type, 1714 * n_mtime fields. Check to see if it represents a special 1715 * device, and if so, check for a possible alias. Once the 1716 * correct vnode has been obtained, fill in the rest of the 1717 * information. 1718 */ 1719 np = VTONFS(vp); 1720 if (vp->v_type == VNON) { 1721 vp->v_type = vtyp; 1722 if (vp->v_type == VFIFO) { 1723 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1724 vp->v_op = fifo_nfsv2nodeop_p; 1725 } else if (vp->v_type == VREG) { 1726 lockinit(&np->n_commitlock, PINOD, "nfsclock", 0, 0); 1727 } else if (vp->v_type == VCHR || vp->v_type == VBLK) { 1728 vp->v_op = spec_nfsv2nodeop_p; 1729 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1730 if (nvp) { 1731 /* 1732 * Discard unneeded vnode, but save its nfsnode. 1733 * Since the nfsnode does not have a lock, its 1734 * vnode lock has to be carried over. 1735 */ 1736 /* 1737 * XXX is the old node sure to be locked here? 1738 */ 1739 KASSERT(lockstatus(&vp->v_lock) == 1740 LK_EXCLUSIVE); 1741 nvp->v_data = vp->v_data; 1742 vp->v_data = NULL; 1743 VOP_UNLOCK(vp, 0); 1744 vp->v_op = spec_vnodeop_p; 1745 vrele(vp); 1746 vgone(vp); 1747 lockmgr(&nvp->v_lock, LK_EXCLUSIVE, 1748 &nvp->v_interlock); 1749 /* 1750 * Reinitialize aliased node. 1751 */ 1752 np->n_vnode = nvp; 1753 *vpp = vp = nvp; 1754 } 1755 } 1756 np->n_mtime = mtime; 1757 } 1758 uid = fxdr_unsigned(uid_t, fp->fa_uid); 1759 gid = fxdr_unsigned(gid_t, fp->fa_gid); 1760 vap = np->n_vattr; 1761 1762 /* 1763 * Invalidate access cache if uid, gid, mode or ctime changed. 1764 */ 1765 if (np->n_accstamp != -1 && 1766 (gid != vap->va_gid || uid != vap->va_uid || vmode != vap->va_mode 1767 || timespeccmp(&ctime, &vap->va_ctime, !=))) 1768 np->n_accstamp = -1; 1769 1770 vap->va_type = vtyp; 1771 vap->va_mode = vmode; 1772 vap->va_rdev = (dev_t)rdev; 1773 vap->va_mtime = mtime; 1774 vap->va_ctime = ctime; 1775 vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0]; 1776 switch (vtyp) { 1777 case VDIR: 1778 vap->va_blocksize = NFS_DIRFRAGSIZ; 1779 break; 1780 case VBLK: 1781 vap->va_blocksize = BLKDEV_IOSIZE; 1782 break; 1783 case VCHR: 1784 vap->va_blocksize = MAXBSIZE; 1785 break; 1786 default: 1787 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1788 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1789 break; 1790 } 1791 if (v3) { 1792 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1793 vap->va_uid = uid; 1794 vap->va_gid = gid; 1795 vap->va_size = fxdr_hyper(&fp->fa3_size); 1796 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1797 vap->va_fileid = fxdr_unsigned(int32_t, 1798 fp->fa3_fileid.nfsuquad[1]); 1799 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1800 vap->va_flags = 0; 1801 vap->va_filerev = 0; 1802 } else { 1803 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1804 vap->va_uid = uid; 1805 vap->va_gid = gid; 1806 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1807 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1808 * NFS_FABLKSIZE; 1809 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1810 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1811 vap->va_flags = 0; 1812 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1813 vap->va_filerev = 0; 1814 } 1815 if (vap->va_size != np->n_size) { 1816 if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) { 1817 vap->va_size = np->n_size; 1818 } else { 1819 np->n_size = vap->va_size; 1820 if (vap->va_type == VREG) { 1821 /* 1822 * we can't free pages if NAC_NOTRUNC because 1823 * the pages can be owned by ourselves. 1824 */ 1825 if (flags & NAC_NOTRUNC) { 1826 np->n_flag |= NTRUNCDELAYED; 1827 } else { 1828 simple_lock(&vp->v_interlock); 1829 (void)VOP_PUTPAGES(vp, 0, 1830 0, PGO_SYNCIO | PGO_CLEANIT | 1831 PGO_FREE | PGO_ALLPAGES); 1832 uvm_vnp_setsize(vp, np->n_size); 1833 } 1834 } 1835 } 1836 } 1837 np->n_attrstamp = mono_time.tv_sec; 1838 if (vaper != NULL) { 1839 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1840 if (np->n_flag & NCHG) { 1841 if (np->n_flag & NACC) 1842 vaper->va_atime = np->n_atim; 1843 if (np->n_flag & NUPD) 1844 vaper->va_mtime = np->n_mtim; 1845 } 1846 } 1847 return (0); 1848 } 1849 1850 /* 1851 * Check the time stamp 1852 * If the cache is valid, copy contents to *vap and return 0 1853 * otherwise return an error 1854 */ 1855 int 1856 nfs_getattrcache(vp, vaper) 1857 struct vnode *vp; 1858 struct vattr *vaper; 1859 { 1860 struct nfsnode *np = VTONFS(vp); 1861 struct vattr *vap; 1862 1863 if (np->n_attrstamp == 0 || 1864 (mono_time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { 1865 nfsstats.attrcache_misses++; 1866 return (ENOENT); 1867 } 1868 nfsstats.attrcache_hits++; 1869 vap = np->n_vattr; 1870 if (vap->va_size != np->n_size) { 1871 if (vap->va_type == VREG) { 1872 if (np->n_flag & NMODIFIED) { 1873 if (vap->va_size < np->n_size) 1874 vap->va_size = np->n_size; 1875 else 1876 np->n_size = vap->va_size; 1877 } else 1878 np->n_size = vap->va_size; 1879 uvm_vnp_setsize(vp, np->n_size); 1880 } else 1881 np->n_size = vap->va_size; 1882 } 1883 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1884 if (np->n_flag & NCHG) { 1885 if (np->n_flag & NACC) 1886 vaper->va_atime = np->n_atim; 1887 if (np->n_flag & NUPD) 1888 vaper->va_mtime = np->n_mtim; 1889 } 1890 return (0); 1891 } 1892 1893 void 1894 nfs_delayedtruncate(vp) 1895 struct vnode *vp; 1896 { 1897 struct nfsnode *np = VTONFS(vp); 1898 1899 if (np->n_flag & NTRUNCDELAYED) { 1900 np->n_flag &= ~NTRUNCDELAYED; 1901 simple_lock(&vp->v_interlock); 1902 (void)VOP_PUTPAGES(vp, 0, 1903 0, PGO_SYNCIO | PGO_CLEANIT | PGO_FREE | PGO_ALLPAGES); 1904 uvm_vnp_setsize(vp, np->n_size); 1905 } 1906 } 1907 1908 #define NFS_WCCKLUDGE_TIMEOUT (24 * 60 * 60) /* 1 day */ 1909 #define NFS_WCCKLUDGE(nmp, now) \ 1910 (((nmp)->nm_iflag & NFSMNT_WCCKLUDGE) && \ 1911 ((now) - (nmp)->nm_wcckludgetime - NFS_WCCKLUDGE_TIMEOUT) < 0) 1912 1913 /* 1914 * nfs_check_wccdata: check inaccurate wcc_data 1915 * 1916 * => return non-zero if we shouldn't trust the wcc_data. 1917 * => NFS_WCCKLUDGE_TIMEOUT is for the case that the server is "fixed". 1918 */ 1919 1920 int 1921 nfs_check_wccdata(struct nfsnode *np, const struct timespec *ctime, 1922 struct timespec *mtime, boolean_t docheck) 1923 { 1924 int error = 0; 1925 1926 #if !defined(NFS_V2_ONLY) 1927 1928 if (docheck) { 1929 struct vnode *vp = NFSTOV(np); 1930 struct nfsmount *nmp; 1931 long now = mono_time.tv_sec; 1932 #if defined(DEBUG) 1933 const char *reason = NULL; /* XXX: gcc */ 1934 #endif 1935 1936 if (timespeccmp(&np->n_vattr->va_mtime, mtime, <=)) { 1937 #if defined(DEBUG) 1938 reason = "mtime"; 1939 #endif 1940 error = EINVAL; 1941 } 1942 1943 if (vp->v_type == VDIR && 1944 timespeccmp(&np->n_vattr->va_ctime, ctime, <=)) { 1945 #if defined(DEBUG) 1946 reason = "ctime"; 1947 #endif 1948 error = EINVAL; 1949 } 1950 1951 nmp = VFSTONFS(vp->v_mount); 1952 if (error) { 1953 1954 /* 1955 * despite of the fact that we've updated the file, 1956 * timestamps of the file were not updated as we 1957 * expected. 1958 * it means that the server has incompatible 1959 * semantics of timestamps or (more likely) 1960 * the server time is not precise enough to 1961 * track each modifications. 1962 * in that case, we disable wcc processing. 1963 * 1964 * yes, strictly speaking, we should disable all 1965 * caching. it's a compromise. 1966 */ 1967 1968 simple_lock(&nmp->nm_slock); 1969 #if defined(DEBUG) 1970 if (!NFS_WCCKLUDGE(nmp, now)) { 1971 printf("%s: inaccurate wcc data (%s) detected," 1972 " disabling wcc\n", 1973 vp->v_mount->mnt_stat.f_mntfromname, 1974 reason); 1975 } 1976 #endif 1977 nmp->nm_iflag |= NFSMNT_WCCKLUDGE; 1978 nmp->nm_wcckludgetime = now; 1979 simple_unlock(&nmp->nm_slock); 1980 } else if (NFS_WCCKLUDGE(nmp, now)) { 1981 error = EPERM; /* XXX */ 1982 } else if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 1983 simple_lock(&nmp->nm_slock); 1984 if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 1985 #if defined(DEBUG) 1986 printf("%s: re-enabling wcc\n", 1987 vp->v_mount->mnt_stat.f_mntfromname); 1988 #endif 1989 nmp->nm_iflag &= ~NFSMNT_WCCKLUDGE; 1990 } 1991 simple_unlock(&nmp->nm_slock); 1992 } 1993 } 1994 1995 #endif /* !defined(NFS_V2_ONLY) */ 1996 1997 return error; 1998 } 1999 2000 /* 2001 * Heuristic to see if the server XDR encodes directory cookies or not. 2002 * it is not supposed to, but a lot of servers may do this. Also, since 2003 * most/all servers will implement V2 as well, it is expected that they 2004 * may return just 32 bits worth of cookie information, so we need to 2005 * find out in which 32 bits this information is available. We do this 2006 * to avoid trouble with emulated binaries that can't handle 64 bit 2007 * directory offsets. 2008 */ 2009 2010 void 2011 nfs_cookieheuristic(vp, flagp, p, cred) 2012 struct vnode *vp; 2013 int *flagp; 2014 struct proc *p; 2015 struct ucred *cred; 2016 { 2017 struct uio auio; 2018 struct iovec aiov; 2019 caddr_t buf, cp; 2020 struct dirent *dp; 2021 off_t *cookies = NULL, *cop; 2022 int error, eof, nc, len; 2023 2024 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 2025 2026 aiov.iov_base = buf; 2027 aiov.iov_len = NFS_DIRFRAGSIZ; 2028 auio.uio_iov = &aiov; 2029 auio.uio_iovcnt = 1; 2030 auio.uio_rw = UIO_READ; 2031 auio.uio_segflg = UIO_SYSSPACE; 2032 auio.uio_procp = NULL; 2033 auio.uio_resid = NFS_DIRFRAGSIZ; 2034 auio.uio_offset = 0; 2035 2036 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 2037 2038 len = NFS_DIRFRAGSIZ - auio.uio_resid; 2039 if (error || len == 0) { 2040 FREE(buf, M_TEMP); 2041 if (cookies) 2042 free(cookies, M_TEMP); 2043 return; 2044 } 2045 2046 /* 2047 * Find the first valid entry and look at its offset cookie. 2048 */ 2049 2050 cp = buf; 2051 for (cop = cookies; len > 0; len -= dp->d_reclen) { 2052 dp = (struct dirent *)cp; 2053 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 2054 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 2055 *flagp |= NFSMNT_SWAPCOOKIE; 2056 nfs_invaldircache(vp, 0); 2057 nfs_vinvalbuf(vp, 0, cred, p, 1); 2058 } 2059 break; 2060 } 2061 cop++; 2062 cp += dp->d_reclen; 2063 } 2064 2065 FREE(buf, M_TEMP); 2066 free(cookies, M_TEMP); 2067 } 2068 #endif /* NFS */ 2069 2070 /* 2071 * Set up nameidata for a lookup() call and do it. 2072 * 2073 * If pubflag is set, this call is done for a lookup operation on the 2074 * public filehandle. In that case we allow crossing mountpoints and 2075 * absolute pathnames. However, the caller is expected to check that 2076 * the lookup result is within the public fs, and deny access if 2077 * it is not. 2078 */ 2079 int 2080 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) 2081 struct nameidata *ndp; 2082 fhandle_t *fhp; 2083 uint32_t len; 2084 struct nfssvc_sock *slp; 2085 struct mbuf *nam; 2086 struct mbuf **mdp; 2087 caddr_t *dposp; 2088 struct vnode **retdirp; 2089 struct proc *p; 2090 int kerbflag, pubflag; 2091 { 2092 int i, rem; 2093 struct mbuf *md; 2094 char *fromcp, *tocp, *cp; 2095 struct iovec aiov; 2096 struct uio auio; 2097 struct vnode *dp; 2098 int error, rdonly, linklen; 2099 struct componentname *cnp = &ndp->ni_cnd; 2100 2101 *retdirp = (struct vnode *)0; 2102 2103 if ((len + 1) > MAXPATHLEN) 2104 return (ENAMETOOLONG); 2105 cnp->cn_pnbuf = PNBUF_GET(); 2106 2107 /* 2108 * Copy the name from the mbuf list to ndp->ni_pnbuf 2109 * and set the various ndp fields appropriately. 2110 */ 2111 fromcp = *dposp; 2112 tocp = cnp->cn_pnbuf; 2113 md = *mdp; 2114 rem = mtod(md, caddr_t) + md->m_len - fromcp; 2115 for (i = 0; i < len; i++) { 2116 while (rem == 0) { 2117 md = md->m_next; 2118 if (md == NULL) { 2119 error = EBADRPC; 2120 goto out; 2121 } 2122 fromcp = mtod(md, caddr_t); 2123 rem = md->m_len; 2124 } 2125 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 2126 error = EACCES; 2127 goto out; 2128 } 2129 *tocp++ = *fromcp++; 2130 rem--; 2131 } 2132 *tocp = '\0'; 2133 *mdp = md; 2134 *dposp = fromcp; 2135 len = nfsm_rndup(len)-len; 2136 if (len > 0) { 2137 if (rem >= len) 2138 *dposp += len; 2139 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 2140 goto out; 2141 } 2142 2143 /* 2144 * Extract and set starting directory. 2145 */ 2146 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 2147 nam, &rdonly, kerbflag, pubflag); 2148 if (error) 2149 goto out; 2150 if (dp->v_type != VDIR) { 2151 vrele(dp); 2152 error = ENOTDIR; 2153 goto out; 2154 } 2155 2156 if (rdonly) 2157 cnp->cn_flags |= RDONLY; 2158 2159 *retdirp = dp; 2160 2161 if (pubflag) { 2162 /* 2163 * Oh joy. For WebNFS, handle those pesky '%' escapes, 2164 * and the 'native path' indicator. 2165 */ 2166 cp = PNBUF_GET(); 2167 fromcp = cnp->cn_pnbuf; 2168 tocp = cp; 2169 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 2170 switch ((unsigned char)*fromcp) { 2171 case WEBNFS_NATIVE_CHAR: 2172 /* 2173 * 'Native' path for us is the same 2174 * as a path according to the NFS spec, 2175 * just skip the escape char. 2176 */ 2177 fromcp++; 2178 break; 2179 /* 2180 * More may be added in the future, range 0x80-0xff 2181 */ 2182 default: 2183 error = EIO; 2184 PNBUF_PUT(cp); 2185 goto out; 2186 } 2187 } 2188 /* 2189 * Translate the '%' escapes, URL-style. 2190 */ 2191 while (*fromcp != '\0') { 2192 if (*fromcp == WEBNFS_ESC_CHAR) { 2193 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 2194 fromcp++; 2195 *tocp++ = HEXSTRTOI(fromcp); 2196 fromcp += 2; 2197 continue; 2198 } else { 2199 error = ENOENT; 2200 PNBUF_PUT(cp); 2201 goto out; 2202 } 2203 } else 2204 *tocp++ = *fromcp++; 2205 } 2206 *tocp = '\0'; 2207 PNBUF_PUT(cnp->cn_pnbuf); 2208 cnp->cn_pnbuf = cp; 2209 } 2210 2211 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 2212 ndp->ni_segflg = UIO_SYSSPACE; 2213 ndp->ni_rootdir = rootvnode; 2214 2215 if (pubflag) { 2216 ndp->ni_loopcnt = 0; 2217 if (cnp->cn_pnbuf[0] == '/') 2218 dp = rootvnode; 2219 } else { 2220 cnp->cn_flags |= NOCROSSMOUNT; 2221 } 2222 2223 cnp->cn_proc = p; 2224 VREF(dp); 2225 2226 for (;;) { 2227 cnp->cn_nameptr = cnp->cn_pnbuf; 2228 ndp->ni_startdir = dp; 2229 /* 2230 * And call lookup() to do the real work 2231 */ 2232 error = lookup(ndp); 2233 if (error) { 2234 PNBUF_PUT(cnp->cn_pnbuf); 2235 return (error); 2236 } 2237 /* 2238 * Check for encountering a symbolic link 2239 */ 2240 if ((cnp->cn_flags & ISSYMLINK) == 0) { 2241 if (cnp->cn_flags & (SAVENAME | SAVESTART)) 2242 cnp->cn_flags |= HASBUF; 2243 else 2244 PNBUF_PUT(cnp->cn_pnbuf); 2245 return (0); 2246 } else { 2247 if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN)) 2248 VOP_UNLOCK(ndp->ni_dvp, 0); 2249 if (!pubflag) { 2250 error = EINVAL; 2251 break; 2252 } 2253 2254 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 2255 error = ELOOP; 2256 break; 2257 } 2258 if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) { 2259 error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred, 2260 cnp->cn_proc); 2261 if (error != 0) 2262 break; 2263 } 2264 if (ndp->ni_pathlen > 1) 2265 cp = PNBUF_GET(); 2266 else 2267 cp = cnp->cn_pnbuf; 2268 aiov.iov_base = cp; 2269 aiov.iov_len = MAXPATHLEN; 2270 auio.uio_iov = &aiov; 2271 auio.uio_iovcnt = 1; 2272 auio.uio_offset = 0; 2273 auio.uio_rw = UIO_READ; 2274 auio.uio_segflg = UIO_SYSSPACE; 2275 auio.uio_procp = NULL; 2276 auio.uio_resid = MAXPATHLEN; 2277 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 2278 if (error) { 2279 badlink: 2280 if (ndp->ni_pathlen > 1) 2281 PNBUF_PUT(cp); 2282 break; 2283 } 2284 linklen = MAXPATHLEN - auio.uio_resid; 2285 if (linklen == 0) { 2286 error = ENOENT; 2287 goto badlink; 2288 } 2289 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 2290 error = ENAMETOOLONG; 2291 goto badlink; 2292 } 2293 if (ndp->ni_pathlen > 1) { 2294 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 2295 PNBUF_PUT(cnp->cn_pnbuf); 2296 cnp->cn_pnbuf = cp; 2297 } else 2298 cnp->cn_pnbuf[linklen] = '\0'; 2299 ndp->ni_pathlen += linklen; 2300 vput(ndp->ni_vp); 2301 dp = ndp->ni_dvp; 2302 /* 2303 * Check if root directory should replace current directory. 2304 */ 2305 if (cnp->cn_pnbuf[0] == '/') { 2306 vrele(dp); 2307 dp = ndp->ni_rootdir; 2308 VREF(dp); 2309 } 2310 } 2311 } 2312 vrele(ndp->ni_dvp); 2313 vput(ndp->ni_vp); 2314 ndp->ni_vp = NULL; 2315 out: 2316 PNBUF_PUT(cnp->cn_pnbuf); 2317 return (error); 2318 } 2319 2320 /* 2321 * A fiddled version of m_adj() that ensures null fill to a 32-bit 2322 * boundary and only trims off the back end 2323 * 2324 * 1. trim off 'len' bytes as m_adj(mp, -len). 2325 * 2. add zero-padding 'nul' bytes at the end of the mbuf chain. 2326 */ 2327 void 2328 nfs_zeropad(mp, len, nul) 2329 struct mbuf *mp; 2330 int len; 2331 int nul; 2332 { 2333 struct mbuf *m; 2334 int count; 2335 2336 /* 2337 * Trim from tail. Scan the mbuf chain, 2338 * calculating its length and finding the last mbuf. 2339 * If the adjustment only affects this mbuf, then just 2340 * adjust and return. Otherwise, rescan and truncate 2341 * after the remaining size. 2342 */ 2343 count = 0; 2344 m = mp; 2345 for (;;) { 2346 count += m->m_len; 2347 if (m->m_next == NULL) 2348 break; 2349 m = m->m_next; 2350 } 2351 2352 KDASSERT(count >= len); 2353 2354 if (m->m_len >= len) { 2355 m->m_len -= len; 2356 } else { 2357 count -= len; 2358 /* 2359 * Correct length for chain is "count". 2360 * Find the mbuf with last data, adjust its length, 2361 * and toss data from remaining mbufs on chain. 2362 */ 2363 for (m = mp; m; m = m->m_next) { 2364 if (m->m_len >= count) { 2365 m->m_len = count; 2366 break; 2367 } 2368 count -= m->m_len; 2369 } 2370 m_freem(m->m_next); 2371 m->m_next = NULL; 2372 } 2373 2374 KDASSERT(m->m_next == NULL); 2375 2376 /* 2377 * zero-padding. 2378 */ 2379 if (nul > 0) { 2380 char *cp; 2381 int i; 2382 2383 if (M_ROMAP(m) || M_TRAILINGSPACE(m) < nul) { 2384 struct mbuf *n; 2385 2386 KDASSERT(MLEN >= nul); 2387 n = m_get(M_WAIT, MT_DATA); 2388 MCLAIM(n, &nfs_mowner); 2389 n->m_len = nul; 2390 n->m_next = NULL; 2391 m->m_next = n; 2392 cp = mtod(n, caddr_t); 2393 } else { 2394 cp = mtod(m, caddr_t) + m->m_len; 2395 m->m_len += nul; 2396 } 2397 for (i = 0; i < nul; i++) 2398 *cp++ = '\0'; 2399 } 2400 return; 2401 } 2402 2403 /* 2404 * Make these functions instead of macros, so that the kernel text size 2405 * doesn't get too big... 2406 */ 2407 void 2408 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2409 struct nfsrv_descript *nfsd; 2410 int before_ret; 2411 struct vattr *before_vap; 2412 int after_ret; 2413 struct vattr *after_vap; 2414 struct mbuf **mbp; 2415 char **bposp; 2416 { 2417 struct mbuf *mb = *mbp; 2418 char *bpos = *bposp; 2419 u_int32_t *tl; 2420 2421 if (before_ret) { 2422 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2423 *tl = nfs_false; 2424 } else { 2425 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2426 *tl++ = nfs_true; 2427 txdr_hyper(before_vap->va_size, tl); 2428 tl += 2; 2429 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2430 tl += 2; 2431 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2432 } 2433 *bposp = bpos; 2434 *mbp = mb; 2435 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2436 } 2437 2438 void 2439 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2440 struct nfsrv_descript *nfsd; 2441 int after_ret; 2442 struct vattr *after_vap; 2443 struct mbuf **mbp; 2444 char **bposp; 2445 { 2446 struct mbuf *mb = *mbp; 2447 char *bpos = *bposp; 2448 u_int32_t *tl; 2449 struct nfs_fattr *fp; 2450 2451 if (after_ret) { 2452 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2453 *tl = nfs_false; 2454 } else { 2455 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2456 *tl++ = nfs_true; 2457 fp = (struct nfs_fattr *)tl; 2458 nfsm_srvfattr(nfsd, after_vap, fp); 2459 } 2460 *mbp = mb; 2461 *bposp = bpos; 2462 } 2463 2464 void 2465 nfsm_srvfattr(nfsd, vap, fp) 2466 struct nfsrv_descript *nfsd; 2467 struct vattr *vap; 2468 struct nfs_fattr *fp; 2469 { 2470 2471 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2472 fp->fa_uid = txdr_unsigned(vap->va_uid); 2473 fp->fa_gid = txdr_unsigned(vap->va_gid); 2474 if (nfsd->nd_flag & ND_NFSV3) { 2475 fp->fa_type = vtonfsv3_type(vap->va_type); 2476 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2477 txdr_hyper(vap->va_size, &fp->fa3_size); 2478 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2479 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2480 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2481 fp->fa3_fsid.nfsuquad[0] = 0; 2482 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2483 fp->fa3_fileid.nfsuquad[0] = 0; 2484 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); 2485 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2486 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2487 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2488 } else { 2489 fp->fa_type = vtonfsv2_type(vap->va_type); 2490 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2491 fp->fa2_size = txdr_unsigned(vap->va_size); 2492 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2493 if (vap->va_type == VFIFO) 2494 fp->fa2_rdev = 0xffffffff; 2495 else 2496 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2497 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2498 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2499 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2500 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2501 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2502 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2503 } 2504 } 2505 2506 /* 2507 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2508 * - look up fsid in mount list (if not found ret error) 2509 * - get vp and export rights by calling VFS_FHTOVP() 2510 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2511 * - if not lockflag unlock it with VOP_UNLOCK() 2512 */ 2513 int 2514 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) 2515 fhandle_t *fhp; 2516 int lockflag; 2517 struct vnode **vpp; 2518 struct ucred *cred; 2519 struct nfssvc_sock *slp; 2520 struct mbuf *nam; 2521 int *rdonlyp; 2522 int kerbflag; 2523 { 2524 struct mount *mp; 2525 int i; 2526 struct ucred *credanon; 2527 int error, exflags; 2528 struct sockaddr_in *saddr; 2529 2530 *vpp = (struct vnode *)0; 2531 2532 if (nfs_ispublicfh(fhp)) { 2533 if (!pubflag || !nfs_pub.np_valid) 2534 return (ESTALE); 2535 fhp = &nfs_pub.np_handle; 2536 } 2537 2538 mp = vfs_getvfs(&fhp->fh_fsid); 2539 if (!mp) 2540 return (ESTALE); 2541 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); 2542 if (error) 2543 return (error); 2544 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2545 if (error) 2546 return (error); 2547 2548 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2549 saddr = mtod(nam, struct sockaddr_in *); 2550 if ((saddr->sin_family == AF_INET) && 2551 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2552 vput(*vpp); 2553 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2554 } 2555 #ifdef INET6 2556 if ((saddr->sin_family == AF_INET6) && 2557 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2558 vput(*vpp); 2559 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2560 } 2561 #endif 2562 } 2563 /* 2564 * Check/setup credentials. 2565 */ 2566 if (exflags & MNT_EXKERB) { 2567 if (!kerbflag) { 2568 vput(*vpp); 2569 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2570 } 2571 } else if (kerbflag) { 2572 vput(*vpp); 2573 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2574 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { 2575 cred->cr_uid = credanon->cr_uid; 2576 cred->cr_gid = credanon->cr_gid; 2577 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) 2578 cred->cr_groups[i] = credanon->cr_groups[i]; 2579 cred->cr_ngroups = i; 2580 } 2581 if (exflags & MNT_EXRDONLY) 2582 *rdonlyp = 1; 2583 else 2584 *rdonlyp = 0; 2585 if (!lockflag) 2586 VOP_UNLOCK(*vpp, 0); 2587 return (0); 2588 } 2589 2590 /* 2591 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2592 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has 2593 * transformed this to all zeroes in both cases, so check for it. 2594 */ 2595 int 2596 nfs_ispublicfh(fhp) 2597 fhandle_t *fhp; 2598 { 2599 char *cp = (char *)fhp; 2600 int i; 2601 2602 for (i = 0; i < NFSX_V3FH; i++) 2603 if (*cp++ != 0) 2604 return (FALSE); 2605 return (TRUE); 2606 } 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 lockmgr(&nmp->nm_writeverflock, LK_EXCLUSIVE, NULL); 2680 2681 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2682 KASSERT(vp->v_mount == mp); 2683 if (vp->v_type == VNON) 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 simple_lock(&vp->v_uobj.vmobjlock); 2691 TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) { 2692 pg->flags &= ~PG_NEEDCOMMIT; 2693 } 2694 simple_unlock(&vp->v_uobj.vmobjlock); 2695 } 2696 simple_lock(&nmp->nm_slock); 2697 nmp->nm_iflag &= ~NFSMNT_STALEWRITEVERF; 2698 simple_unlock(&nmp->nm_slock); 2699 lockmgr(&nmp->nm_writeverflock, LK_RELEASE, NULL); 2700 } 2701 2702 void 2703 nfs_merge_commit_ranges(vp) 2704 struct vnode *vp; 2705 { 2706 struct nfsnode *np = VTONFS(vp); 2707 2708 KASSERT(np->n_commitflags & NFS_COMMIT_PUSH_VALID); 2709 2710 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2711 np->n_pushedlo = np->n_pushlo; 2712 np->n_pushedhi = np->n_pushhi; 2713 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2714 } else { 2715 if (np->n_pushlo < np->n_pushedlo) 2716 np->n_pushedlo = np->n_pushlo; 2717 if (np->n_pushhi > np->n_pushedhi) 2718 np->n_pushedhi = np->n_pushhi; 2719 } 2720 2721 np->n_pushlo = np->n_pushhi = 0; 2722 np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID; 2723 2724 #ifdef NFS_DEBUG_COMMIT 2725 printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2726 (unsigned)np->n_pushedhi); 2727 #endif 2728 } 2729 2730 int 2731 nfs_in_committed_range(vp, off, len) 2732 struct vnode *vp; 2733 off_t off, len; 2734 { 2735 struct nfsnode *np = VTONFS(vp); 2736 off_t lo, hi; 2737 2738 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2739 return 0; 2740 lo = off; 2741 hi = lo + len; 2742 2743 return (lo >= np->n_pushedlo && hi <= np->n_pushedhi); 2744 } 2745 2746 int 2747 nfs_in_tobecommitted_range(vp, off, len) 2748 struct vnode *vp; 2749 off_t off, len; 2750 { 2751 struct nfsnode *np = VTONFS(vp); 2752 off_t lo, hi; 2753 2754 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2755 return 0; 2756 lo = off; 2757 hi = lo + len; 2758 2759 return (lo >= np->n_pushlo && hi <= np->n_pushhi); 2760 } 2761 2762 void 2763 nfs_add_committed_range(vp, off, len) 2764 struct vnode *vp; 2765 off_t off, len; 2766 { 2767 struct nfsnode *np = VTONFS(vp); 2768 off_t lo, hi; 2769 2770 lo = off; 2771 hi = lo + len; 2772 2773 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2774 np->n_pushedlo = lo; 2775 np->n_pushedhi = hi; 2776 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2777 } else { 2778 if (hi > np->n_pushedhi) 2779 np->n_pushedhi = hi; 2780 if (lo < np->n_pushedlo) 2781 np->n_pushedlo = lo; 2782 } 2783 #ifdef NFS_DEBUG_COMMIT 2784 printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2785 (unsigned)np->n_pushedhi); 2786 #endif 2787 } 2788 2789 void 2790 nfs_del_committed_range(vp, off, len) 2791 struct vnode *vp; 2792 off_t off, len; 2793 { 2794 struct nfsnode *np = VTONFS(vp); 2795 off_t lo, hi; 2796 2797 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2798 return; 2799 2800 lo = off; 2801 hi = lo + len; 2802 2803 if (lo > np->n_pushedhi || hi < np->n_pushedlo) 2804 return; 2805 if (lo <= np->n_pushedlo) 2806 np->n_pushedlo = hi; 2807 else if (hi >= np->n_pushedhi) 2808 np->n_pushedhi = lo; 2809 else { 2810 /* 2811 * XXX There's only one range. If the deleted range 2812 * is in the middle, pick the largest of the 2813 * contiguous ranges that it leaves. 2814 */ 2815 if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi)) 2816 np->n_pushedhi = lo; 2817 else 2818 np->n_pushedlo = hi; 2819 } 2820 #ifdef NFS_DEBUG_COMMIT 2821 printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2822 (unsigned)np->n_pushedhi); 2823 #endif 2824 } 2825 2826 void 2827 nfs_add_tobecommitted_range(vp, off, len) 2828 struct vnode *vp; 2829 off_t off, len; 2830 { 2831 struct nfsnode *np = VTONFS(vp); 2832 off_t lo, hi; 2833 2834 lo = off; 2835 hi = lo + len; 2836 2837 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) { 2838 np->n_pushlo = lo; 2839 np->n_pushhi = hi; 2840 np->n_commitflags |= NFS_COMMIT_PUSH_VALID; 2841 } else { 2842 if (lo < np->n_pushlo) 2843 np->n_pushlo = lo; 2844 if (hi > np->n_pushhi) 2845 np->n_pushhi = hi; 2846 } 2847 #ifdef NFS_DEBUG_COMMIT 2848 printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2849 (unsigned)np->n_pushhi); 2850 #endif 2851 } 2852 2853 void 2854 nfs_del_tobecommitted_range(vp, off, len) 2855 struct vnode *vp; 2856 off_t off, len; 2857 { 2858 struct nfsnode *np = VTONFS(vp); 2859 off_t lo, hi; 2860 2861 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2862 return; 2863 2864 lo = off; 2865 hi = lo + len; 2866 2867 if (lo > np->n_pushhi || hi < np->n_pushlo) 2868 return; 2869 2870 if (lo <= np->n_pushlo) 2871 np->n_pushlo = hi; 2872 else if (hi >= np->n_pushhi) 2873 np->n_pushhi = lo; 2874 else { 2875 /* 2876 * XXX There's only one range. If the deleted range 2877 * is in the middle, pick the largest of the 2878 * contiguous ranges that it leaves. 2879 */ 2880 if ((np->n_pushlo - lo) > (hi - np->n_pushhi)) 2881 np->n_pushhi = lo; 2882 else 2883 np->n_pushlo = hi; 2884 } 2885 #ifdef NFS_DEBUG_COMMIT 2886 printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2887 (unsigned)np->n_pushhi); 2888 #endif 2889 } 2890 2891 /* 2892 * Map errnos to NFS error numbers. For Version 3 also filter out error 2893 * numbers not specified for the associated procedure. 2894 */ 2895 int 2896 nfsrv_errmap(nd, err) 2897 struct nfsrv_descript *nd; 2898 int err; 2899 { 2900 const short *defaulterrp, *errp; 2901 2902 if (nd->nd_flag & ND_NFSV3) { 2903 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2904 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2905 while (*++errp) { 2906 if (*errp == err) 2907 return (err); 2908 else if (*errp > err) 2909 break; 2910 } 2911 return ((int)*defaulterrp); 2912 } else 2913 return (err & 0xffff); 2914 } 2915 if (err <= ELAST) 2916 return ((int)nfsrv_v2errmap[err - 1]); 2917 return (NFSERR_IO); 2918 } 2919 2920 /* 2921 * Sort the group list in increasing numerical order. 2922 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort 2923 * that used to be here.) 2924 */ 2925 void 2926 nfsrvw_sort(list, num) 2927 gid_t *list; 2928 int num; 2929 { 2930 int i, j; 2931 gid_t v; 2932 2933 /* Insertion sort. */ 2934 for (i = 1; i < num; i++) { 2935 v = list[i]; 2936 /* find correct slot for value v, moving others up */ 2937 for (j = i; --j >= 0 && v < list[j];) 2938 list[j + 1] = list[j]; 2939 list[j + 1] = v; 2940 } 2941 } 2942 2943 /* 2944 * copy credentials making sure that the result can be compared with memcmp(). 2945 */ 2946 void 2947 nfsrv_setcred(incred, outcred) 2948 struct ucred *incred, *outcred; 2949 { 2950 int i; 2951 2952 memset((caddr_t)outcred, 0, sizeof (struct ucred)); 2953 outcred->cr_ref = 1; 2954 outcred->cr_uid = incred->cr_uid; 2955 outcred->cr_gid = incred->cr_gid; 2956 outcred->cr_ngroups = incred->cr_ngroups; 2957 for (i = 0; i < incred->cr_ngroups; i++) 2958 outcred->cr_groups[i] = incred->cr_groups[i]; 2959 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); 2960 } 2961 2962 u_int32_t 2963 nfs_getxid() 2964 { 2965 static u_int32_t base; 2966 static u_int32_t nfs_xid = 0; 2967 static struct simplelock nfs_xidlock = SIMPLELOCK_INITIALIZER; 2968 u_int32_t newxid; 2969 2970 simple_lock(&nfs_xidlock); 2971 /* 2972 * derive initial xid from system time 2973 * XXX time is invalid if root not yet mounted 2974 */ 2975 if (__predict_false(!base && (rootvp))) { 2976 struct timeval tv; 2977 2978 microtime(&tv); 2979 base = tv.tv_sec << 12; 2980 nfs_xid = base; 2981 } 2982 2983 /* 2984 * Skip zero xid if it should ever happen. 2985 */ 2986 if (__predict_false(++nfs_xid == 0)) 2987 nfs_xid++; 2988 newxid = nfs_xid; 2989 simple_unlock(&nfs_xidlock); 2990 2991 return txdr_unsigned(newxid); 2992 } 2993 2994 /* 2995 * assign a new xid for existing request. 2996 * used for NFSERR_JUKEBOX handling. 2997 */ 2998 void 2999 nfs_renewxid(struct nfsreq *req) 3000 { 3001 u_int32_t xid; 3002 int off; 3003 3004 xid = nfs_getxid(); 3005 if (req->r_nmp->nm_sotype == SOCK_STREAM) 3006 off = sizeof(u_int32_t); /* RPC record mark */ 3007 else 3008 off = 0; 3009 3010 m_copyback(req->r_mreq, off, sizeof(xid), (void *)&xid); 3011 req->r_xid = xid; 3012 } 3013
Indexes created Mon Sep 22 13:09:51 GMT 2025