nfs_subs.c revision 1.176
1 /* $NetBSD: nfs_subs.c,v 1.176 2006/10/20 18:58:12 reinoud 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 1.176 2006/10/20 18:58:12 reinoud 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 #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/nqnfs.h> 113 #include <nfs/nfsrtt.h> 114 #include <nfs/nfs_var.h> 115 116 #include <miscfs/specfs/specdev.h> 117 118 #include <netinet/in.h> 119 #ifdef ISO 120 #include <netiso/iso.h> 121 #endif 122 123 /* 124 * Data items converted to xdr at startup, since they are constant 125 * This is kinda hokey, but may save a little time doing byte swaps 126 */ 127 u_int32_t nfs_xdrneg1; 128 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 129 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 130 rpc_auth_kerb; 131 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; 132 133 /* And other global data */ 134 const nfstype nfsv2_type[9] = 135 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON }; 136 const nfstype nfsv3_type[9] = 137 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON }; 138 const enum vtype nv2tov_type[8] = 139 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 140 const enum vtype nv3tov_type[8] = 141 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 142 int nfs_ticks; 143 int nfs_commitsize; 144 145 MALLOC_DEFINE(M_NFSDIROFF, "NFS diroff", "NFS directory cookies"); 146 147 /* NFS client/server stats. */ 148 struct nfsstats nfsstats; 149 150 /* 151 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 152 */ 153 const int nfsv3_procid[NFS_NPROCS] = { 154 NFSPROC_NULL, 155 NFSPROC_GETATTR, 156 NFSPROC_SETATTR, 157 NFSPROC_NOOP, 158 NFSPROC_LOOKUP, 159 NFSPROC_READLINK, 160 NFSPROC_READ, 161 NFSPROC_NOOP, 162 NFSPROC_WRITE, 163 NFSPROC_CREATE, 164 NFSPROC_REMOVE, 165 NFSPROC_RENAME, 166 NFSPROC_LINK, 167 NFSPROC_SYMLINK, 168 NFSPROC_MKDIR, 169 NFSPROC_RMDIR, 170 NFSPROC_READDIR, 171 NFSPROC_FSSTAT, 172 NFSPROC_NOOP, 173 NFSPROC_NOOP, 174 NFSPROC_NOOP, 175 NFSPROC_NOOP, 176 NFSPROC_NOOP, 177 NFSPROC_NOOP, 178 NFSPROC_NOOP, 179 NFSPROC_NOOP 180 }; 181 182 /* 183 * and the reverse mapping from generic to Version 2 procedure numbers 184 */ 185 const int nfsv2_procid[NFS_NPROCS] = { 186 NFSV2PROC_NULL, 187 NFSV2PROC_GETATTR, 188 NFSV2PROC_SETATTR, 189 NFSV2PROC_LOOKUP, 190 NFSV2PROC_NOOP, 191 NFSV2PROC_READLINK, 192 NFSV2PROC_READ, 193 NFSV2PROC_WRITE, 194 NFSV2PROC_CREATE, 195 NFSV2PROC_MKDIR, 196 NFSV2PROC_SYMLINK, 197 NFSV2PROC_CREATE, 198 NFSV2PROC_REMOVE, 199 NFSV2PROC_RMDIR, 200 NFSV2PROC_RENAME, 201 NFSV2PROC_LINK, 202 NFSV2PROC_READDIR, 203 NFSV2PROC_NOOP, 204 NFSV2PROC_STATFS, 205 NFSV2PROC_NOOP, 206 NFSV2PROC_NOOP, 207 NFSV2PROC_NOOP, 208 NFSV2PROC_NOOP, 209 NFSV2PROC_NOOP, 210 NFSV2PROC_NOOP, 211 NFSV2PROC_NOOP, 212 }; 213 214 /* 215 * Maps errno values to nfs error numbers. 216 * Use NFSERR_IO as the catch all for ones not specifically defined in 217 * RFC 1094. 218 */ 219 static const u_char nfsrv_v2errmap[ELAST] = { 220 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 221 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 222 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 223 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 224 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 225 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 226 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 227 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 228 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 229 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 230 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 231 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 232 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, 233 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 234 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 235 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 236 NFSERR_IO, NFSERR_IO, 237 }; 238 239 /* 240 * Maps errno values to nfs error numbers. 241 * Although it is not obvious whether or not NFS clients really care if 242 * a returned error value is in the specified list for the procedure, the 243 * safest thing to do is filter them appropriately. For Version 2, the 244 * X/Open XNFS document is the only specification that defines error values 245 * for each RPC (The RFC simply lists all possible error values for all RPCs), 246 * so I have decided to not do this for Version 2. 247 * The first entry is the default error return and the rest are the valid 248 * errors for that RPC in increasing numeric order. 249 */ 250 static const short nfsv3err_null[] = { 251 0, 252 0, 253 }; 254 255 static const short nfsv3err_getattr[] = { 256 NFSERR_IO, 257 NFSERR_IO, 258 NFSERR_STALE, 259 NFSERR_BADHANDLE, 260 NFSERR_SERVERFAULT, 261 0, 262 }; 263 264 static const short nfsv3err_setattr[] = { 265 NFSERR_IO, 266 NFSERR_PERM, 267 NFSERR_IO, 268 NFSERR_ACCES, 269 NFSERR_INVAL, 270 NFSERR_NOSPC, 271 NFSERR_ROFS, 272 NFSERR_DQUOT, 273 NFSERR_STALE, 274 NFSERR_BADHANDLE, 275 NFSERR_NOT_SYNC, 276 NFSERR_SERVERFAULT, 277 0, 278 }; 279 280 static const short nfsv3err_lookup[] = { 281 NFSERR_IO, 282 NFSERR_NOENT, 283 NFSERR_IO, 284 NFSERR_ACCES, 285 NFSERR_NOTDIR, 286 NFSERR_NAMETOL, 287 NFSERR_STALE, 288 NFSERR_BADHANDLE, 289 NFSERR_SERVERFAULT, 290 0, 291 }; 292 293 static const short nfsv3err_access[] = { 294 NFSERR_IO, 295 NFSERR_IO, 296 NFSERR_STALE, 297 NFSERR_BADHANDLE, 298 NFSERR_SERVERFAULT, 299 0, 300 }; 301 302 static const short nfsv3err_readlink[] = { 303 NFSERR_IO, 304 NFSERR_IO, 305 NFSERR_ACCES, 306 NFSERR_INVAL, 307 NFSERR_STALE, 308 NFSERR_BADHANDLE, 309 NFSERR_NOTSUPP, 310 NFSERR_SERVERFAULT, 311 0, 312 }; 313 314 static const short nfsv3err_read[] = { 315 NFSERR_IO, 316 NFSERR_IO, 317 NFSERR_NXIO, 318 NFSERR_ACCES, 319 NFSERR_INVAL, 320 NFSERR_STALE, 321 NFSERR_BADHANDLE, 322 NFSERR_SERVERFAULT, 323 NFSERR_JUKEBOX, 324 0, 325 }; 326 327 static const short nfsv3err_write[] = { 328 NFSERR_IO, 329 NFSERR_IO, 330 NFSERR_ACCES, 331 NFSERR_INVAL, 332 NFSERR_FBIG, 333 NFSERR_NOSPC, 334 NFSERR_ROFS, 335 NFSERR_DQUOT, 336 NFSERR_STALE, 337 NFSERR_BADHANDLE, 338 NFSERR_SERVERFAULT, 339 NFSERR_JUKEBOX, 340 0, 341 }; 342 343 static const short nfsv3err_create[] = { 344 NFSERR_IO, 345 NFSERR_IO, 346 NFSERR_ACCES, 347 NFSERR_EXIST, 348 NFSERR_NOTDIR, 349 NFSERR_NOSPC, 350 NFSERR_ROFS, 351 NFSERR_NAMETOL, 352 NFSERR_DQUOT, 353 NFSERR_STALE, 354 NFSERR_BADHANDLE, 355 NFSERR_NOTSUPP, 356 NFSERR_SERVERFAULT, 357 0, 358 }; 359 360 static const short nfsv3err_mkdir[] = { 361 NFSERR_IO, 362 NFSERR_IO, 363 NFSERR_ACCES, 364 NFSERR_EXIST, 365 NFSERR_NOTDIR, 366 NFSERR_NOSPC, 367 NFSERR_ROFS, 368 NFSERR_NAMETOL, 369 NFSERR_DQUOT, 370 NFSERR_STALE, 371 NFSERR_BADHANDLE, 372 NFSERR_NOTSUPP, 373 NFSERR_SERVERFAULT, 374 0, 375 }; 376 377 static const short nfsv3err_symlink[] = { 378 NFSERR_IO, 379 NFSERR_IO, 380 NFSERR_ACCES, 381 NFSERR_EXIST, 382 NFSERR_NOTDIR, 383 NFSERR_NOSPC, 384 NFSERR_ROFS, 385 NFSERR_NAMETOL, 386 NFSERR_DQUOT, 387 NFSERR_STALE, 388 NFSERR_BADHANDLE, 389 NFSERR_NOTSUPP, 390 NFSERR_SERVERFAULT, 391 0, 392 }; 393 394 static const short nfsv3err_mknod[] = { 395 NFSERR_IO, 396 NFSERR_IO, 397 NFSERR_ACCES, 398 NFSERR_EXIST, 399 NFSERR_NOTDIR, 400 NFSERR_NOSPC, 401 NFSERR_ROFS, 402 NFSERR_NAMETOL, 403 NFSERR_DQUOT, 404 NFSERR_STALE, 405 NFSERR_BADHANDLE, 406 NFSERR_NOTSUPP, 407 NFSERR_SERVERFAULT, 408 NFSERR_BADTYPE, 409 0, 410 }; 411 412 static const short nfsv3err_remove[] = { 413 NFSERR_IO, 414 NFSERR_NOENT, 415 NFSERR_IO, 416 NFSERR_ACCES, 417 NFSERR_NOTDIR, 418 NFSERR_ROFS, 419 NFSERR_NAMETOL, 420 NFSERR_STALE, 421 NFSERR_BADHANDLE, 422 NFSERR_SERVERFAULT, 423 0, 424 }; 425 426 static const short nfsv3err_rmdir[] = { 427 NFSERR_IO, 428 NFSERR_NOENT, 429 NFSERR_IO, 430 NFSERR_ACCES, 431 NFSERR_EXIST, 432 NFSERR_NOTDIR, 433 NFSERR_INVAL, 434 NFSERR_ROFS, 435 NFSERR_NAMETOL, 436 NFSERR_NOTEMPTY, 437 NFSERR_STALE, 438 NFSERR_BADHANDLE, 439 NFSERR_NOTSUPP, 440 NFSERR_SERVERFAULT, 441 0, 442 }; 443 444 static const short nfsv3err_rename[] = { 445 NFSERR_IO, 446 NFSERR_NOENT, 447 NFSERR_IO, 448 NFSERR_ACCES, 449 NFSERR_EXIST, 450 NFSERR_XDEV, 451 NFSERR_NOTDIR, 452 NFSERR_ISDIR, 453 NFSERR_INVAL, 454 NFSERR_NOSPC, 455 NFSERR_ROFS, 456 NFSERR_MLINK, 457 NFSERR_NAMETOL, 458 NFSERR_NOTEMPTY, 459 NFSERR_DQUOT, 460 NFSERR_STALE, 461 NFSERR_BADHANDLE, 462 NFSERR_NOTSUPP, 463 NFSERR_SERVERFAULT, 464 0, 465 }; 466 467 static const short nfsv3err_link[] = { 468 NFSERR_IO, 469 NFSERR_IO, 470 NFSERR_ACCES, 471 NFSERR_EXIST, 472 NFSERR_XDEV, 473 NFSERR_NOTDIR, 474 NFSERR_INVAL, 475 NFSERR_NOSPC, 476 NFSERR_ROFS, 477 NFSERR_MLINK, 478 NFSERR_NAMETOL, 479 NFSERR_DQUOT, 480 NFSERR_STALE, 481 NFSERR_BADHANDLE, 482 NFSERR_NOTSUPP, 483 NFSERR_SERVERFAULT, 484 0, 485 }; 486 487 static const short nfsv3err_readdir[] = { 488 NFSERR_IO, 489 NFSERR_IO, 490 NFSERR_ACCES, 491 NFSERR_NOTDIR, 492 NFSERR_STALE, 493 NFSERR_BADHANDLE, 494 NFSERR_BAD_COOKIE, 495 NFSERR_TOOSMALL, 496 NFSERR_SERVERFAULT, 497 0, 498 }; 499 500 static const short nfsv3err_readdirplus[] = { 501 NFSERR_IO, 502 NFSERR_IO, 503 NFSERR_ACCES, 504 NFSERR_NOTDIR, 505 NFSERR_STALE, 506 NFSERR_BADHANDLE, 507 NFSERR_BAD_COOKIE, 508 NFSERR_NOTSUPP, 509 NFSERR_TOOSMALL, 510 NFSERR_SERVERFAULT, 511 0, 512 }; 513 514 static const short nfsv3err_fsstat[] = { 515 NFSERR_IO, 516 NFSERR_IO, 517 NFSERR_STALE, 518 NFSERR_BADHANDLE, 519 NFSERR_SERVERFAULT, 520 0, 521 }; 522 523 static const short nfsv3err_fsinfo[] = { 524 NFSERR_STALE, 525 NFSERR_STALE, 526 NFSERR_BADHANDLE, 527 NFSERR_SERVERFAULT, 528 0, 529 }; 530 531 static const short nfsv3err_pathconf[] = { 532 NFSERR_STALE, 533 NFSERR_STALE, 534 NFSERR_BADHANDLE, 535 NFSERR_SERVERFAULT, 536 0, 537 }; 538 539 static const short nfsv3err_commit[] = { 540 NFSERR_IO, 541 NFSERR_IO, 542 NFSERR_STALE, 543 NFSERR_BADHANDLE, 544 NFSERR_SERVERFAULT, 545 0, 546 }; 547 548 static const short * const nfsrv_v3errmap[] = { 549 nfsv3err_null, 550 nfsv3err_getattr, 551 nfsv3err_setattr, 552 nfsv3err_lookup, 553 nfsv3err_access, 554 nfsv3err_readlink, 555 nfsv3err_read, 556 nfsv3err_write, 557 nfsv3err_create, 558 nfsv3err_mkdir, 559 nfsv3err_symlink, 560 nfsv3err_mknod, 561 nfsv3err_remove, 562 nfsv3err_rmdir, 563 nfsv3err_rename, 564 nfsv3err_link, 565 nfsv3err_readdir, 566 nfsv3err_readdirplus, 567 nfsv3err_fsstat, 568 nfsv3err_fsinfo, 569 nfsv3err_pathconf, 570 nfsv3err_commit, 571 }; 572 573 extern struct nfsrtt nfsrtt; 574 extern time_t nqnfsstarttime; 575 extern int nqsrv_clockskew; 576 extern int nqsrv_writeslack; 577 extern int nqsrv_maxlease; 578 extern const int nqnfs_piggy[NFS_NPROCS]; 579 extern struct nfsnodehashhead *nfsnodehashtbl; 580 extern u_long nfsnodehash; 581 582 u_long nfsdirhashmask; 583 584 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 585 586 /* 587 * Create the header for an rpc request packet 588 * The hsiz is the size of the rest of the nfs request header. 589 * (just used to decide if a cluster is a good idea) 590 */ 591 struct mbuf * 592 nfsm_reqh( 593 struct nfsnode *np __unused, 594 u_long procid __unused, 595 int hsiz, 596 caddr_t *bposp 597 ) 598 { 599 struct mbuf *mb; 600 caddr_t bpos; 601 #ifndef NFS_V2_ONLY 602 struct nfsmount *nmp; 603 u_int32_t *tl; 604 int nqflag; 605 #endif 606 607 mb = m_get(M_WAIT, MT_DATA); 608 MCLAIM(mb, &nfs_mowner); 609 if (hsiz >= MINCLSIZE) 610 m_clget(mb, M_WAIT); 611 mb->m_len = 0; 612 bpos = mtod(mb, caddr_t); 613 614 #ifndef NFS_V2_ONLY 615 /* 616 * For NQNFS, add lease request. 617 */ 618 if (np) { 619 nmp = VFSTONFS(np->n_vnode->v_mount); 620 if (nmp->nm_flag & NFSMNT_NQNFS) { 621 nqflag = NQNFS_NEEDLEASE(np, procid); 622 if (nqflag) { 623 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); 624 *tl++ = txdr_unsigned(nqflag); 625 *tl = txdr_unsigned(nmp->nm_leaseterm); 626 } else { 627 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 628 *tl = 0; 629 } 630 } 631 } 632 #endif 633 /* Finally, return values */ 634 *bposp = bpos; 635 return (mb); 636 } 637 638 /* 639 * Build the RPC header and fill in the authorization info. 640 * The authorization string argument is only used when the credentials 641 * come from outside of the kernel. 642 * Returns the head of the mbuf list. 643 */ 644 struct mbuf * 645 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 646 verf_str, mrest, mrest_len, mbp, xidp) 647 kauth_cred_t cr; 648 int nmflag; 649 int procid; 650 int auth_type; 651 int auth_len; 652 char *auth_str; 653 int verf_len; 654 char *verf_str; 655 struct mbuf *mrest; 656 int mrest_len; 657 struct mbuf **mbp; 658 u_int32_t *xidp; 659 { 660 struct mbuf *mb; 661 u_int32_t *tl; 662 caddr_t bpos; 663 int i; 664 struct mbuf *mreq; 665 int siz, grpsiz, authsiz; 666 667 authsiz = nfsm_rndup(auth_len); 668 mb = m_gethdr(M_WAIT, MT_DATA); 669 MCLAIM(mb, &nfs_mowner); 670 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 671 m_clget(mb, M_WAIT); 672 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 673 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 674 } else { 675 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 676 } 677 mb->m_len = 0; 678 mreq = mb; 679 bpos = mtod(mb, caddr_t); 680 681 /* 682 * First the RPC header. 683 */ 684 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 685 686 *tl++ = *xidp = nfs_getxid(); 687 *tl++ = rpc_call; 688 *tl++ = rpc_vers; 689 if (nmflag & NFSMNT_NQNFS) { 690 *tl++ = txdr_unsigned(NQNFS_PROG); 691 *tl++ = txdr_unsigned(NQNFS_VER3); 692 } else { 693 *tl++ = txdr_unsigned(NFS_PROG); 694 if (nmflag & NFSMNT_NFSV3) 695 *tl++ = txdr_unsigned(NFS_VER3); 696 else 697 *tl++ = txdr_unsigned(NFS_VER2); 698 } 699 if (nmflag & NFSMNT_NFSV3) 700 *tl++ = txdr_unsigned(procid); 701 else 702 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 703 704 /* 705 * And then the authorization cred. 706 */ 707 *tl++ = txdr_unsigned(auth_type); 708 *tl = txdr_unsigned(authsiz); 709 switch (auth_type) { 710 case RPCAUTH_UNIX: 711 nfsm_build(tl, u_int32_t *, auth_len); 712 *tl++ = 0; /* stamp ?? */ 713 *tl++ = 0; /* NULL hostname */ 714 *tl++ = txdr_unsigned(kauth_cred_geteuid(cr)); 715 *tl++ = txdr_unsigned(kauth_cred_getegid(cr)); 716 grpsiz = (auth_len >> 2) - 5; 717 *tl++ = txdr_unsigned(grpsiz); 718 for (i = 0; i < grpsiz; i++) 719 *tl++ = txdr_unsigned(kauth_cred_group(cr, i)); /* XXX elad review */ 720 break; 721 case RPCAUTH_KERB4: 722 siz = auth_len; 723 while (siz > 0) { 724 if (M_TRAILINGSPACE(mb) == 0) { 725 struct mbuf *mb2; 726 mb2 = m_get(M_WAIT, MT_DATA); 727 MCLAIM(mb2, &nfs_mowner); 728 if (siz >= MINCLSIZE) 729 m_clget(mb2, M_WAIT); 730 mb->m_next = mb2; 731 mb = mb2; 732 mb->m_len = 0; 733 bpos = mtod(mb, caddr_t); 734 } 735 i = min(siz, M_TRAILINGSPACE(mb)); 736 memcpy(bpos, auth_str, i); 737 mb->m_len += i; 738 auth_str += i; 739 bpos += i; 740 siz -= i; 741 } 742 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 743 for (i = 0; i < siz; i++) 744 *bpos++ = '\0'; 745 mb->m_len += siz; 746 } 747 break; 748 }; 749 750 /* 751 * And the verifier... 752 */ 753 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 754 if (verf_str) { 755 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 756 *tl = txdr_unsigned(verf_len); 757 siz = verf_len; 758 while (siz > 0) { 759 if (M_TRAILINGSPACE(mb) == 0) { 760 struct mbuf *mb2; 761 mb2 = m_get(M_WAIT, MT_DATA); 762 MCLAIM(mb2, &nfs_mowner); 763 if (siz >= MINCLSIZE) 764 m_clget(mb2, M_WAIT); 765 mb->m_next = mb2; 766 mb = mb2; 767 mb->m_len = 0; 768 bpos = mtod(mb, caddr_t); 769 } 770 i = min(siz, M_TRAILINGSPACE(mb)); 771 memcpy(bpos, verf_str, i); 772 mb->m_len += i; 773 verf_str += i; 774 bpos += i; 775 siz -= i; 776 } 777 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 778 for (i = 0; i < siz; i++) 779 *bpos++ = '\0'; 780 mb->m_len += siz; 781 } 782 } else { 783 *tl++ = txdr_unsigned(RPCAUTH_NULL); 784 *tl = 0; 785 } 786 mb->m_next = mrest; 787 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 788 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 789 *mbp = mb; 790 return (mreq); 791 } 792 793 /* 794 * copies mbuf chain to the uio scatter/gather list 795 */ 796 int 797 nfsm_mbuftouio(mrep, uiop, siz, dpos) 798 struct mbuf **mrep; 799 struct uio *uiop; 800 int siz; 801 caddr_t *dpos; 802 { 803 char *mbufcp, *uiocp; 804 int xfer, left, len; 805 struct mbuf *mp; 806 long uiosiz, rem; 807 int error = 0; 808 809 mp = *mrep; 810 mbufcp = *dpos; 811 len = mtod(mp, caddr_t)+mp->m_len-mbufcp; 812 rem = nfsm_rndup(siz)-siz; 813 while (siz > 0) { 814 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 815 return (EFBIG); 816 left = uiop->uio_iov->iov_len; 817 uiocp = uiop->uio_iov->iov_base; 818 if (left > siz) 819 left = siz; 820 uiosiz = left; 821 while (left > 0) { 822 while (len == 0) { 823 mp = mp->m_next; 824 if (mp == NULL) 825 return (EBADRPC); 826 mbufcp = mtod(mp, caddr_t); 827 len = mp->m_len; 828 } 829 xfer = (left > len) ? len : left; 830 error = copyout_vmspace(uiop->uio_vmspace, mbufcp, 831 uiocp, xfer); 832 if (error) { 833 return error; 834 } 835 left -= xfer; 836 len -= xfer; 837 mbufcp += xfer; 838 uiocp += xfer; 839 uiop->uio_offset += xfer; 840 uiop->uio_resid -= xfer; 841 } 842 if (uiop->uio_iov->iov_len <= siz) { 843 uiop->uio_iovcnt--; 844 uiop->uio_iov++; 845 } else { 846 uiop->uio_iov->iov_base = 847 (caddr_t)uiop->uio_iov->iov_base + uiosiz; 848 uiop->uio_iov->iov_len -= uiosiz; 849 } 850 siz -= uiosiz; 851 } 852 *dpos = mbufcp; 853 *mrep = mp; 854 if (rem > 0) { 855 if (len < rem) 856 error = nfs_adv(mrep, dpos, rem, len); 857 else 858 *dpos += rem; 859 } 860 return (error); 861 } 862 863 /* 864 * copies a uio scatter/gather list to an mbuf chain. 865 * NOTE: can ony handle iovcnt == 1 866 */ 867 int 868 nfsm_uiotombuf(uiop, mq, siz, bpos) 869 struct uio *uiop; 870 struct mbuf **mq; 871 int siz; 872 caddr_t *bpos; 873 { 874 char *uiocp; 875 struct mbuf *mp, *mp2; 876 int xfer, left, mlen; 877 int uiosiz, clflg, rem; 878 char *cp; 879 int error; 880 881 #ifdef DIAGNOSTIC 882 if (uiop->uio_iovcnt != 1) 883 panic("nfsm_uiotombuf: iovcnt != 1"); 884 #endif 885 886 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 887 clflg = 1; 888 else 889 clflg = 0; 890 rem = nfsm_rndup(siz)-siz; 891 mp = mp2 = *mq; 892 while (siz > 0) { 893 left = uiop->uio_iov->iov_len; 894 uiocp = uiop->uio_iov->iov_base; 895 if (left > siz) 896 left = siz; 897 uiosiz = left; 898 while (left > 0) { 899 mlen = M_TRAILINGSPACE(mp); 900 if (mlen == 0) { 901 mp = m_get(M_WAIT, MT_DATA); 902 MCLAIM(mp, &nfs_mowner); 903 if (clflg) 904 m_clget(mp, M_WAIT); 905 mp->m_len = 0; 906 mp2->m_next = mp; 907 mp2 = mp; 908 mlen = M_TRAILINGSPACE(mp); 909 } 910 xfer = (left > mlen) ? mlen : left; 911 cp = mtod(mp, caddr_t) + mp->m_len; 912 error = copyin_vmspace(uiop->uio_vmspace, uiocp, cp, 913 xfer); 914 if (error) { 915 /* XXX */ 916 } 917 mp->m_len += xfer; 918 left -= xfer; 919 uiocp += xfer; 920 uiop->uio_offset += xfer; 921 uiop->uio_resid -= xfer; 922 } 923 uiop->uio_iov->iov_base = (caddr_t)uiop->uio_iov->iov_base + 924 uiosiz; 925 uiop->uio_iov->iov_len -= uiosiz; 926 siz -= uiosiz; 927 } 928 if (rem > 0) { 929 if (rem > M_TRAILINGSPACE(mp)) { 930 mp = m_get(M_WAIT, MT_DATA); 931 MCLAIM(mp, &nfs_mowner); 932 mp->m_len = 0; 933 mp2->m_next = mp; 934 } 935 cp = mtod(mp, caddr_t) + mp->m_len; 936 for (left = 0; left < rem; left++) 937 *cp++ = '\0'; 938 mp->m_len += rem; 939 *bpos = cp; 940 } else 941 *bpos = mtod(mp, caddr_t)+mp->m_len; 942 *mq = mp; 943 return (0); 944 } 945 946 /* 947 * Get at least "siz" bytes of correctly aligned data. 948 * When called the mbuf pointers are not necessarily correct, 949 * dsosp points to what ought to be in m_data and left contains 950 * what ought to be in m_len. 951 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 952 * cases. (The macros use the vars. dpos and dpos2) 953 */ 954 int 955 nfsm_disct(mdp, dposp, siz, left, cp2) 956 struct mbuf **mdp; 957 caddr_t *dposp; 958 int siz; 959 int left; 960 caddr_t *cp2; 961 { 962 struct mbuf *m1, *m2; 963 struct mbuf *havebuf = NULL; 964 caddr_t src = *dposp; 965 caddr_t dst; 966 int len; 967 968 #ifdef DEBUG 969 if (left < 0) 970 panic("nfsm_disct: left < 0"); 971 #endif 972 m1 = *mdp; 973 /* 974 * Skip through the mbuf chain looking for an mbuf with 975 * some data. If the first mbuf found has enough data 976 * and it is correctly aligned return it. 977 */ 978 while (left == 0) { 979 havebuf = m1; 980 *mdp = m1 = m1->m_next; 981 if (m1 == NULL) 982 return (EBADRPC); 983 src = mtod(m1, caddr_t); 984 left = m1->m_len; 985 /* 986 * If we start a new mbuf and it is big enough 987 * and correctly aligned just return it, don't 988 * do any pull up. 989 */ 990 if (left >= siz && nfsm_aligned(src)) { 991 *cp2 = src; 992 *dposp = src + siz; 993 return (0); 994 } 995 } 996 if (m1->m_flags & M_EXT) { 997 if (havebuf) { 998 /* If the first mbuf with data has external data 999 * and there is a previous empty mbuf use it 1000 * to move the data into. 1001 */ 1002 m2 = m1; 1003 *mdp = m1 = havebuf; 1004 if (m1->m_flags & M_EXT) { 1005 MEXTREMOVE(m1); 1006 } 1007 } else { 1008 /* 1009 * If the first mbuf has a external data 1010 * and there is no previous empty mbuf 1011 * allocate a new mbuf and move the external 1012 * data to the new mbuf. Also make the first 1013 * mbuf look empty. 1014 */ 1015 m2 = m_get(M_WAIT, MT_DATA); 1016 m2->m_ext = m1->m_ext; 1017 m2->m_data = src; 1018 m2->m_len = left; 1019 MCLADDREFERENCE(m1, m2); 1020 MEXTREMOVE(m1); 1021 m2->m_next = m1->m_next; 1022 m1->m_next = m2; 1023 } 1024 m1->m_len = 0; 1025 if (m1->m_flags & M_PKTHDR) 1026 dst = m1->m_pktdat; 1027 else 1028 dst = m1->m_dat; 1029 m1->m_data = dst; 1030 } else { 1031 /* 1032 * If the first mbuf has no external data 1033 * move the data to the front of the mbuf. 1034 */ 1035 if (m1->m_flags & M_PKTHDR) 1036 dst = m1->m_pktdat; 1037 else 1038 dst = m1->m_dat; 1039 m1->m_data = dst; 1040 if (dst != src) 1041 memmove(dst, src, left); 1042 dst += left; 1043 m1->m_len = left; 1044 m2 = m1->m_next; 1045 } 1046 *cp2 = m1->m_data; 1047 *dposp = mtod(m1, caddr_t) + siz; 1048 /* 1049 * Loop through mbufs pulling data up into first mbuf until 1050 * the first mbuf is full or there is no more data to 1051 * pullup. 1052 */ 1053 while ((len = M_TRAILINGSPACE(m1)) != 0 && m2) { 1054 if ((len = min(len, m2->m_len)) != 0) 1055 memcpy(dst, m2->m_data, len); 1056 m1->m_len += len; 1057 dst += len; 1058 m2->m_data += len; 1059 m2->m_len -= len; 1060 m2 = m2->m_next; 1061 } 1062 if (m1->m_len < siz) 1063 return (EBADRPC); 1064 return (0); 1065 } 1066 1067 /* 1068 * Advance the position in the mbuf chain. 1069 */ 1070 int 1071 nfs_adv(mdp, dposp, offs, left) 1072 struct mbuf **mdp; 1073 caddr_t *dposp; 1074 int offs; 1075 int left; 1076 { 1077 struct mbuf *m; 1078 int s; 1079 1080 m = *mdp; 1081 s = left; 1082 while (s < offs) { 1083 offs -= s; 1084 m = m->m_next; 1085 if (m == NULL) 1086 return (EBADRPC); 1087 s = m->m_len; 1088 } 1089 *mdp = m; 1090 *dposp = mtod(m, caddr_t)+offs; 1091 return (0); 1092 } 1093 1094 /* 1095 * Copy a string into mbufs for the hard cases... 1096 */ 1097 int 1098 nfsm_strtmbuf(mb, bpos, cp, siz) 1099 struct mbuf **mb; 1100 char **bpos; 1101 const char *cp; 1102 long siz; 1103 { 1104 struct mbuf *m1 = NULL, *m2; 1105 long left, xfer, len, tlen; 1106 u_int32_t *tl; 1107 int putsize; 1108 1109 putsize = 1; 1110 m2 = *mb; 1111 left = M_TRAILINGSPACE(m2); 1112 if (left > 0) { 1113 tl = ((u_int32_t *)(*bpos)); 1114 *tl++ = txdr_unsigned(siz); 1115 putsize = 0; 1116 left -= NFSX_UNSIGNED; 1117 m2->m_len += NFSX_UNSIGNED; 1118 if (left > 0) { 1119 memcpy((caddr_t) tl, cp, left); 1120 siz -= left; 1121 cp += left; 1122 m2->m_len += left; 1123 left = 0; 1124 } 1125 } 1126 /* Loop around adding mbufs */ 1127 while (siz > 0) { 1128 m1 = m_get(M_WAIT, MT_DATA); 1129 MCLAIM(m1, &nfs_mowner); 1130 if (siz > MLEN) 1131 m_clget(m1, M_WAIT); 1132 m1->m_len = NFSMSIZ(m1); 1133 m2->m_next = m1; 1134 m2 = m1; 1135 tl = mtod(m1, u_int32_t *); 1136 tlen = 0; 1137 if (putsize) { 1138 *tl++ = txdr_unsigned(siz); 1139 m1->m_len -= NFSX_UNSIGNED; 1140 tlen = NFSX_UNSIGNED; 1141 putsize = 0; 1142 } 1143 if (siz < m1->m_len) { 1144 len = nfsm_rndup(siz); 1145 xfer = siz; 1146 if (xfer < len) 1147 *(tl+(xfer>>2)) = 0; 1148 } else { 1149 xfer = len = m1->m_len; 1150 } 1151 memcpy((caddr_t) tl, cp, xfer); 1152 m1->m_len = len+tlen; 1153 siz -= xfer; 1154 cp += xfer; 1155 } 1156 *mb = m1; 1157 *bpos = mtod(m1, caddr_t)+m1->m_len; 1158 return (0); 1159 } 1160 1161 /* 1162 * Directory caching routines. They work as follows: 1163 * - a cache is maintained per VDIR nfsnode. 1164 * - for each offset cookie that is exported to userspace, and can 1165 * thus be thrown back at us as an offset to VOP_READDIR, store 1166 * information in the cache. 1167 * - cached are: 1168 * - cookie itself 1169 * - blocknumber (essentially just a search key in the buffer cache) 1170 * - entry number in block. 1171 * - offset cookie of block in which this entry is stored 1172 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1173 * - entries are looked up in a hash table 1174 * - also maintained is an LRU list of entries, used to determine 1175 * which ones to delete if the cache grows too large. 1176 * - if 32 <-> 64 translation mode is requested for a filesystem, 1177 * the cache also functions as a translation table 1178 * - in the translation case, invalidating the cache does not mean 1179 * flushing it, but just marking entries as invalid, except for 1180 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1181 * still be able to use the cache as a translation table. 1182 * - 32 bit cookies are uniquely created by combining the hash table 1183 * entry value, and one generation count per hash table entry, 1184 * incremented each time an entry is appended to the chain. 1185 * - the cache is invalidated each time a direcory is modified 1186 * - sanity checks are also done; if an entry in a block turns 1187 * out not to have a matching cookie, the cache is invalidated 1188 * and a new block starting from the wanted offset is fetched from 1189 * the server. 1190 * - directory entries as read from the server are extended to contain 1191 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1192 * the cache and exporting them to userspace through the cookie 1193 * argument to VOP_READDIR. 1194 */ 1195 1196 u_long 1197 nfs_dirhash(off) 1198 off_t off; 1199 { 1200 int i; 1201 char *cp = (char *)&off; 1202 u_long sum = 0L; 1203 1204 for (i = 0 ; i < sizeof (off); i++) 1205 sum += *cp++; 1206 1207 return sum; 1208 } 1209 1210 #define _NFSDC_MTX(np) (&NFSTOV(np)->v_interlock) 1211 #define NFSDC_LOCK(np) simple_lock(_NFSDC_MTX(np)) 1212 #define NFSDC_UNLOCK(np) simple_unlock(_NFSDC_MTX(np)) 1213 #define NFSDC_ASSERT_LOCKED(np) LOCK_ASSERT(simple_lock_held(_NFSDC_MTX(np))) 1214 1215 void 1216 nfs_initdircache(vp) 1217 struct vnode *vp; 1218 { 1219 struct nfsnode *np = VTONFS(vp); 1220 struct nfsdirhashhead *dircache; 1221 1222 dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, M_NFSDIROFF, 1223 M_WAITOK, &nfsdirhashmask); 1224 1225 NFSDC_LOCK(np); 1226 if (np->n_dircache == NULL) { 1227 np->n_dircachesize = 0; 1228 np->n_dircache = dircache; 1229 dircache = NULL; 1230 TAILQ_INIT(&np->n_dirchain); 1231 } 1232 NFSDC_UNLOCK(np); 1233 if (dircache) 1234 hashdone(dircache, M_NFSDIROFF); 1235 } 1236 1237 void 1238 nfs_initdirxlatecookie(vp) 1239 struct vnode *vp; 1240 { 1241 struct nfsnode *np = VTONFS(vp); 1242 unsigned *dirgens; 1243 1244 KASSERT(VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_XLATECOOKIE); 1245 1246 dirgens = malloc(NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, 1247 M_WAITOK|M_ZERO); 1248 NFSDC_LOCK(np); 1249 if (np->n_dirgens == NULL) { 1250 np->n_dirgens = dirgens; 1251 dirgens = NULL; 1252 } 1253 NFSDC_UNLOCK(np); 1254 if (dirgens) 1255 free(dirgens, M_NFSDIROFF); 1256 } 1257 1258 static const struct nfsdircache dzero; 1259 1260 static void nfs_unlinkdircache __P((struct nfsnode *np, struct nfsdircache *)); 1261 static void nfs_putdircache_unlocked __P((struct nfsnode *, 1262 struct nfsdircache *)); 1263 1264 static void 1265 nfs_unlinkdircache(np, ndp) 1266 struct nfsnode *np; 1267 struct nfsdircache *ndp; 1268 { 1269 1270 NFSDC_ASSERT_LOCKED(np); 1271 KASSERT(ndp != &dzero); 1272 1273 if (LIST_NEXT(ndp, dc_hash) == (void *)-1) 1274 return; 1275 1276 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1277 LIST_REMOVE(ndp, dc_hash); 1278 LIST_NEXT(ndp, dc_hash) = (void *)-1; /* mark as unlinked */ 1279 1280 nfs_putdircache_unlocked(np, ndp); 1281 } 1282 1283 void 1284 nfs_putdircache(np, ndp) 1285 struct nfsnode *np; 1286 struct nfsdircache *ndp; 1287 { 1288 int ref; 1289 1290 if (ndp == &dzero) 1291 return; 1292 1293 KASSERT(ndp->dc_refcnt > 0); 1294 NFSDC_LOCK(np); 1295 ref = --ndp->dc_refcnt; 1296 NFSDC_UNLOCK(np); 1297 1298 if (ref == 0) 1299 free(ndp, M_NFSDIROFF); 1300 } 1301 1302 static void 1303 nfs_putdircache_unlocked(struct nfsnode *np __unused, struct nfsdircache *ndp) 1304 { 1305 int ref; 1306 1307 NFSDC_ASSERT_LOCKED(np); 1308 1309 if (ndp == &dzero) 1310 return; 1311 1312 KASSERT(ndp->dc_refcnt > 0); 1313 ref = --ndp->dc_refcnt; 1314 if (ref == 0) 1315 free(ndp, M_NFSDIROFF); 1316 } 1317 1318 struct nfsdircache * 1319 nfs_searchdircache(vp, off, do32, hashent) 1320 struct vnode *vp; 1321 off_t off; 1322 int do32; 1323 int *hashent; 1324 { 1325 struct nfsdirhashhead *ndhp; 1326 struct nfsdircache *ndp = NULL; 1327 struct nfsnode *np = VTONFS(vp); 1328 unsigned ent; 1329 1330 /* 1331 * Zero is always a valid cookie. 1332 */ 1333 if (off == 0) 1334 /* XXXUNCONST */ 1335 return (struct nfsdircache *)__UNCONST(&dzero); 1336 1337 if (!np->n_dircache) 1338 return NULL; 1339 1340 /* 1341 * We use a 32bit cookie as search key, directly reconstruct 1342 * the hashentry. Else use the hashfunction. 1343 */ 1344 if (do32) { 1345 ent = (u_int32_t)off >> 24; 1346 if (ent >= NFS_DIRHASHSIZ) 1347 return NULL; 1348 ndhp = &np->n_dircache[ent]; 1349 } else { 1350 ndhp = NFSDIRHASH(np, off); 1351 } 1352 1353 if (hashent) 1354 *hashent = (int)(ndhp - np->n_dircache); 1355 1356 NFSDC_LOCK(np); 1357 if (do32) { 1358 LIST_FOREACH(ndp, ndhp, dc_hash) { 1359 if (ndp->dc_cookie32 == (u_int32_t)off) { 1360 /* 1361 * An invalidated entry will become the 1362 * start of a new block fetched from 1363 * the server. 1364 */ 1365 if (ndp->dc_flags & NFSDC_INVALID) { 1366 ndp->dc_blkcookie = ndp->dc_cookie; 1367 ndp->dc_entry = 0; 1368 ndp->dc_flags &= ~NFSDC_INVALID; 1369 } 1370 break; 1371 } 1372 } 1373 } else { 1374 LIST_FOREACH(ndp, ndhp, dc_hash) { 1375 if (ndp->dc_cookie == off) 1376 break; 1377 } 1378 } 1379 if (ndp != NULL) 1380 ndp->dc_refcnt++; 1381 NFSDC_UNLOCK(np); 1382 return ndp; 1383 } 1384 1385 1386 struct nfsdircache * 1387 nfs_enterdircache(struct vnode *vp, off_t off, off_t blkoff, int en, 1388 daddr_t blkno __unused) 1389 { 1390 struct nfsnode *np = VTONFS(vp); 1391 struct nfsdirhashhead *ndhp; 1392 struct nfsdircache *ndp = NULL; 1393 struct nfsdircache *newndp = NULL; 1394 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1395 int hashent = 0, gen, overwrite; /* XXX: GCC */ 1396 1397 /* 1398 * XXX refuse entries for offset 0. amd(8) erroneously sets 1399 * cookie 0 for the '.' entry, making this necessary. This 1400 * isn't so bad, as 0 is a special case anyway. 1401 */ 1402 if (off == 0) 1403 /* XXXUNCONST */ 1404 return (struct nfsdircache *)__UNCONST(&dzero); 1405 1406 if (!np->n_dircache) 1407 /* 1408 * XXX would like to do this in nfs_nget but vtype 1409 * isn't known at that time. 1410 */ 1411 nfs_initdircache(vp); 1412 1413 if ((nmp->nm_flag & NFSMNT_XLATECOOKIE) && !np->n_dirgens) 1414 nfs_initdirxlatecookie(vp); 1415 1416 retry: 1417 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1418 1419 NFSDC_LOCK(np); 1420 if (ndp && (ndp->dc_flags & NFSDC_INVALID) == 0) { 1421 /* 1422 * Overwriting an old entry. Check if it's the same. 1423 * If so, just return. If not, remove the old entry. 1424 */ 1425 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1426 goto done; 1427 nfs_unlinkdircache(np, ndp); 1428 nfs_putdircache_unlocked(np, ndp); 1429 ndp = NULL; 1430 } 1431 1432 ndhp = &np->n_dircache[hashent]; 1433 1434 if (!ndp) { 1435 if (newndp == NULL) { 1436 NFSDC_UNLOCK(np); 1437 newndp = malloc(sizeof(*ndp), M_NFSDIROFF, M_WAITOK); 1438 newndp->dc_refcnt = 1; 1439 LIST_NEXT(newndp, dc_hash) = (void *)-1; 1440 goto retry; 1441 } 1442 ndp = newndp; 1443 newndp = NULL; 1444 overwrite = 0; 1445 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1446 /* 1447 * We're allocating a new entry, so bump the 1448 * generation number. 1449 */ 1450 KASSERT(np->n_dirgens); 1451 gen = ++np->n_dirgens[hashent]; 1452 if (gen == 0) { 1453 np->n_dirgens[hashent]++; 1454 gen++; 1455 } 1456 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1457 } 1458 } else 1459 overwrite = 1; 1460 1461 ndp->dc_cookie = off; 1462 ndp->dc_blkcookie = blkoff; 1463 ndp->dc_entry = en; 1464 ndp->dc_flags = 0; 1465 1466 if (overwrite) 1467 goto done; 1468 1469 /* 1470 * If the maximum directory cookie cache size has been reached 1471 * for this node, take one off the front. The idea is that 1472 * directories are typically read front-to-back once, so that 1473 * the oldest entries can be thrown away without much performance 1474 * loss. 1475 */ 1476 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1477 nfs_unlinkdircache(np, TAILQ_FIRST(&np->n_dirchain)); 1478 } else 1479 np->n_dircachesize++; 1480 1481 KASSERT(ndp->dc_refcnt == 1); 1482 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1483 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1484 ndp->dc_refcnt++; 1485 done: 1486 KASSERT(ndp->dc_refcnt > 0); 1487 NFSDC_UNLOCK(np); 1488 if (newndp) 1489 nfs_putdircache(np, newndp); 1490 return ndp; 1491 } 1492 1493 void 1494 nfs_invaldircache(vp, flags) 1495 struct vnode *vp; 1496 int flags; 1497 { 1498 struct nfsnode *np = VTONFS(vp); 1499 struct nfsdircache *ndp = NULL; 1500 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1501 const boolean_t forcefree = flags & NFS_INVALDIRCACHE_FORCE; 1502 1503 #ifdef DIAGNOSTIC 1504 if (vp->v_type != VDIR) 1505 panic("nfs: invaldircache: not dir"); 1506 #endif 1507 1508 if ((flags & NFS_INVALDIRCACHE_KEEPEOF) == 0) 1509 np->n_flag &= ~NEOFVALID; 1510 1511 if (!np->n_dircache) 1512 return; 1513 1514 NFSDC_LOCK(np); 1515 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1516 while ((ndp = TAILQ_FIRST(&np->n_dirchain)) != NULL) { 1517 KASSERT(!forcefree || ndp->dc_refcnt == 1); 1518 nfs_unlinkdircache(np, ndp); 1519 } 1520 np->n_dircachesize = 0; 1521 if (forcefree && np->n_dirgens) { 1522 FREE(np->n_dirgens, M_NFSDIROFF); 1523 np->n_dirgens = NULL; 1524 } 1525 } else { 1526 TAILQ_FOREACH(ndp, &np->n_dirchain, dc_chain) 1527 ndp->dc_flags |= NFSDC_INVALID; 1528 } 1529 1530 NFSDC_UNLOCK(np); 1531 } 1532 1533 /* 1534 * Called once before VFS init to initialize shared and 1535 * server-specific data structures. 1536 */ 1537 static int 1538 nfs_init0(void) 1539 { 1540 nfsrtt.pos = 0; 1541 rpc_vers = txdr_unsigned(RPC_VER2); 1542 rpc_call = txdr_unsigned(RPC_CALL); 1543 rpc_reply = txdr_unsigned(RPC_REPLY); 1544 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1545 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1546 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1547 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1548 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1549 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1550 nfs_prog = txdr_unsigned(NFS_PROG); 1551 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1552 nfs_true = txdr_unsigned(TRUE); 1553 nfs_false = txdr_unsigned(FALSE); 1554 nfs_xdrneg1 = txdr_unsigned(-1); 1555 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1556 if (nfs_ticks < 1) 1557 nfs_ticks = 1; 1558 #ifdef NFSSERVER 1559 nfsrv_init(0); /* Init server data structures */ 1560 nfsrv_initcache(); /* Init the server request cache */ 1561 #endif /* NFSSERVER */ 1562 1563 #if defined(NFSSERVER) || (defined(NFS) && !defined(NFS_V2_ONLY)) 1564 nfsdreq_init(); 1565 #endif /* defined(NFSSERVER) || (defined(NFS) && !defined(NFS_V2_ONLY)) */ 1566 1567 #if defined(NFSSERVER) || !defined(NFS_V2_ONLY) 1568 /* 1569 * Initialize the nqnfs data structures. 1570 */ 1571 if (nqnfsstarttime == 0) { 1572 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1573 + nqsrv_clockskew + nqsrv_writeslack; 1574 NQLOADNOVRAM(nqnfsstarttime); 1575 CIRCLEQ_INIT(&nqtimerhead); 1576 nqfhhashtbl = hashinit(NQLCHSZ, HASH_LIST, M_NQLEASE, 1577 M_WAITOK, &nqfhhash); 1578 } 1579 #endif 1580 1581 exithook_establish(nfs_exit, NULL); 1582 1583 /* 1584 * Initialize reply list and start timer 1585 */ 1586 TAILQ_INIT(&nfs_reqq); 1587 nfs_timer(NULL); 1588 MOWNER_ATTACH(&nfs_mowner); 1589 1590 #ifdef NFS 1591 /* Initialize the kqueue structures */ 1592 nfs_kqinit(); 1593 /* Initialize the iod structures */ 1594 nfs_iodinit(); 1595 #endif 1596 return 0; 1597 } 1598 1599 void 1600 nfs_init(void) 1601 { 1602 static ONCE_DECL(nfs_init_once); 1603 1604 RUN_ONCE(&nfs_init_once, nfs_init0); 1605 } 1606 1607 #ifdef NFS 1608 /* 1609 * Called once at VFS init to initialize client-specific data structures. 1610 */ 1611 void 1612 nfs_vfs_init() 1613 { 1614 /* Initialize NFS server / client shared data. */ 1615 nfs_init(); 1616 1617 nfs_nhinit(); /* Init the nfsnode table */ 1618 nfs_commitsize = uvmexp.npages << (PAGE_SHIFT - 4); 1619 } 1620 1621 void 1622 nfs_vfs_reinit() 1623 { 1624 nfs_nhreinit(); 1625 } 1626 1627 void 1628 nfs_vfs_done() 1629 { 1630 nfs_nhdone(); 1631 } 1632 1633 /* 1634 * Attribute cache routines. 1635 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1636 * that are on the mbuf list 1637 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1638 * error otherwise 1639 */ 1640 1641 /* 1642 * Load the attribute cache (that lives in the nfsnode entry) with 1643 * the values on the mbuf list and 1644 * Iff vap not NULL 1645 * copy the attributes to *vaper 1646 */ 1647 int 1648 nfsm_loadattrcache(vpp, mdp, dposp, vaper, flags) 1649 struct vnode **vpp; 1650 struct mbuf **mdp; 1651 caddr_t *dposp; 1652 struct vattr *vaper; 1653 int flags; 1654 { 1655 int32_t t1; 1656 caddr_t cp2; 1657 int error = 0; 1658 struct mbuf *md; 1659 int v3 = NFS_ISV3(*vpp); 1660 1661 md = *mdp; 1662 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1663 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1664 if (error) 1665 return (error); 1666 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper, flags); 1667 } 1668 1669 int 1670 nfs_loadattrcache(vpp, fp, vaper, flags) 1671 struct vnode **vpp; 1672 struct nfs_fattr *fp; 1673 struct vattr *vaper; 1674 int flags; 1675 { 1676 struct vnode *vp = *vpp; 1677 struct vattr *vap; 1678 int v3 = NFS_ISV3(vp); 1679 enum vtype vtyp; 1680 u_short vmode; 1681 struct timespec mtime; 1682 struct timespec ctime; 1683 struct vnode *nvp; 1684 int32_t rdev; 1685 struct nfsnode *np; 1686 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1687 uid_t uid; 1688 gid_t gid; 1689 1690 if (v3) { 1691 vtyp = nfsv3tov_type(fp->fa_type); 1692 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1693 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1694 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1695 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1696 fxdr_nfsv3time(&fp->fa3_ctime, &ctime); 1697 } else { 1698 vtyp = nfsv2tov_type(fp->fa_type); 1699 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1700 if (vtyp == VNON || vtyp == VREG) 1701 vtyp = IFTOVT(vmode); 1702 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1703 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1704 ctime.tv_sec = fxdr_unsigned(u_int32_t, 1705 fp->fa2_ctime.nfsv2_sec); 1706 ctime.tv_nsec = 0; 1707 1708 /* 1709 * Really ugly NFSv2 kludge. 1710 */ 1711 if (vtyp == VCHR && rdev == 0xffffffff) 1712 vtyp = VFIFO; 1713 } 1714 1715 vmode &= ALLPERMS; 1716 1717 /* 1718 * If v_type == VNON it is a new node, so fill in the v_type, 1719 * n_mtime fields. Check to see if it represents a special 1720 * device, and if so, check for a possible alias. Once the 1721 * correct vnode has been obtained, fill in the rest of the 1722 * information. 1723 */ 1724 np = VTONFS(vp); 1725 if (vp->v_type == VNON) { 1726 vp->v_type = vtyp; 1727 if (vp->v_type == VFIFO) { 1728 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1729 vp->v_op = fifo_nfsv2nodeop_p; 1730 } else if (vp->v_type == VREG) { 1731 lockinit(&np->n_commitlock, PINOD, "nfsclock", 0, 0); 1732 } else if (vp->v_type == VCHR || vp->v_type == VBLK) { 1733 vp->v_op = spec_nfsv2nodeop_p; 1734 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1735 if (nvp) { 1736 /* 1737 * Discard unneeded vnode, but save its nfsnode. 1738 * Since the nfsnode does not have a lock, its 1739 * vnode lock has to be carried over. 1740 */ 1741 /* 1742 * XXX is the old node sure to be locked here? 1743 */ 1744 KASSERT(lockstatus(&vp->v_lock) == 1745 LK_EXCLUSIVE); 1746 nvp->v_data = vp->v_data; 1747 vp->v_data = NULL; 1748 VOP_UNLOCK(vp, 0); 1749 vp->v_op = spec_vnodeop_p; 1750 vrele(vp); 1751 vgone(vp); 1752 lockmgr(&nvp->v_lock, LK_EXCLUSIVE, 1753 &nvp->v_interlock); 1754 /* 1755 * Reinitialize aliased node. 1756 */ 1757 np->n_vnode = nvp; 1758 *vpp = vp = nvp; 1759 } 1760 } 1761 np->n_mtime = mtime; 1762 } 1763 uid = fxdr_unsigned(uid_t, fp->fa_uid); 1764 gid = fxdr_unsigned(gid_t, fp->fa_gid); 1765 vap = np->n_vattr; 1766 1767 /* 1768 * Invalidate access cache if uid, gid, mode or ctime changed. 1769 */ 1770 if (np->n_accstamp != -1 && 1771 (gid != vap->va_gid || uid != vap->va_uid || vmode != vap->va_mode 1772 || timespeccmp(&ctime, &vap->va_ctime, !=))) 1773 np->n_accstamp = -1; 1774 1775 vap->va_type = vtyp; 1776 vap->va_mode = vmode; 1777 vap->va_rdev = (dev_t)rdev; 1778 vap->va_mtime = mtime; 1779 vap->va_ctime = ctime; 1780 vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0]; 1781 switch (vtyp) { 1782 case VDIR: 1783 vap->va_blocksize = NFS_DIRFRAGSIZ; 1784 break; 1785 case VBLK: 1786 vap->va_blocksize = BLKDEV_IOSIZE; 1787 break; 1788 case VCHR: 1789 vap->va_blocksize = MAXBSIZE; 1790 break; 1791 default: 1792 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1793 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1794 break; 1795 } 1796 if (v3) { 1797 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1798 vap->va_uid = uid; 1799 vap->va_gid = gid; 1800 vap->va_size = fxdr_hyper(&fp->fa3_size); 1801 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1802 vap->va_fileid = fxdr_hyper(&fp->fa3_fileid); 1803 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1804 vap->va_flags = 0; 1805 vap->va_filerev = 0; 1806 } else { 1807 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1808 vap->va_uid = uid; 1809 vap->va_gid = gid; 1810 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1811 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1812 * NFS_FABLKSIZE; 1813 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1814 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1815 vap->va_flags = 0; 1816 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1817 vap->va_filerev = 0; 1818 } 1819 if (vap->va_size != np->n_size) { 1820 if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) { 1821 vap->va_size = np->n_size; 1822 } else { 1823 np->n_size = vap->va_size; 1824 if (vap->va_type == VREG) { 1825 /* 1826 * we can't free pages if NAC_NOTRUNC because 1827 * the pages can be owned by ourselves. 1828 */ 1829 if (flags & NAC_NOTRUNC) { 1830 np->n_flag |= NTRUNCDELAYED; 1831 } else { 1832 genfs_node_wrlock(vp); 1833 simple_lock(&vp->v_interlock); 1834 (void)VOP_PUTPAGES(vp, 0, 1835 0, PGO_SYNCIO | PGO_CLEANIT | 1836 PGO_FREE | PGO_ALLPAGES); 1837 uvm_vnp_setsize(vp, np->n_size); 1838 genfs_node_unlock(vp); 1839 } 1840 } 1841 } 1842 } 1843 np->n_attrstamp = time_second; 1844 if (vaper != NULL) { 1845 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1846 if (np->n_flag & NCHG) { 1847 if (np->n_flag & NACC) 1848 vaper->va_atime = np->n_atim; 1849 if (np->n_flag & NUPD) 1850 vaper->va_mtime = np->n_mtim; 1851 } 1852 } 1853 return (0); 1854 } 1855 1856 /* 1857 * Check the time stamp 1858 * If the cache is valid, copy contents to *vap and return 0 1859 * otherwise return an error 1860 */ 1861 int 1862 nfs_getattrcache(vp, vaper) 1863 struct vnode *vp; 1864 struct vattr *vaper; 1865 { 1866 struct nfsnode *np = VTONFS(vp); 1867 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1868 struct vattr *vap; 1869 1870 if (np->n_attrstamp == 0 || 1871 (time_second - np->n_attrstamp) >= NFS_ATTRTIMEO(nmp, np)) { 1872 nfsstats.attrcache_misses++; 1873 return (ENOENT); 1874 } 1875 nfsstats.attrcache_hits++; 1876 vap = np->n_vattr; 1877 if (vap->va_size != np->n_size) { 1878 if (vap->va_type == VREG) { 1879 if (np->n_flag & NMODIFIED) { 1880 if (vap->va_size < np->n_size) 1881 vap->va_size = np->n_size; 1882 else 1883 np->n_size = vap->va_size; 1884 } else 1885 np->n_size = vap->va_size; 1886 genfs_node_wrlock(vp); 1887 uvm_vnp_setsize(vp, np->n_size); 1888 genfs_node_unlock(vp); 1889 } else 1890 np->n_size = vap->va_size; 1891 } 1892 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1893 if (np->n_flag & NCHG) { 1894 if (np->n_flag & NACC) 1895 vaper->va_atime = np->n_atim; 1896 if (np->n_flag & NUPD) 1897 vaper->va_mtime = np->n_mtim; 1898 } 1899 return (0); 1900 } 1901 1902 void 1903 nfs_delayedtruncate(vp) 1904 struct vnode *vp; 1905 { 1906 struct nfsnode *np = VTONFS(vp); 1907 1908 if (np->n_flag & NTRUNCDELAYED) { 1909 np->n_flag &= ~NTRUNCDELAYED; 1910 genfs_node_wrlock(vp); 1911 simple_lock(&vp->v_interlock); 1912 (void)VOP_PUTPAGES(vp, 0, 1913 0, PGO_SYNCIO | PGO_CLEANIT | PGO_FREE | PGO_ALLPAGES); 1914 uvm_vnp_setsize(vp, np->n_size); 1915 genfs_node_unlock(vp); 1916 } 1917 } 1918 1919 #define NFS_WCCKLUDGE_TIMEOUT (24 * 60 * 60) /* 1 day */ 1920 #define NFS_WCCKLUDGE(nmp, now) \ 1921 (((nmp)->nm_iflag & NFSMNT_WCCKLUDGE) && \ 1922 ((now) - (nmp)->nm_wcckludgetime - NFS_WCCKLUDGE_TIMEOUT) < 0) 1923 1924 /* 1925 * nfs_check_wccdata: check inaccurate wcc_data 1926 * 1927 * => return non-zero if we shouldn't trust the wcc_data. 1928 * => NFS_WCCKLUDGE_TIMEOUT is for the case that the server is "fixed". 1929 */ 1930 1931 int 1932 nfs_check_wccdata( 1933 struct nfsnode *np __unused, 1934 const struct timespec *ctime __unused, 1935 struct timespec *mtime __unused, 1936 boolean_t docheck __unused 1937 ) 1938 { 1939 int error = 0; 1940 1941 #if !defined(NFS_V2_ONLY) 1942 1943 if (docheck) { 1944 struct vnode *vp = NFSTOV(np); 1945 struct nfsmount *nmp; 1946 long now = time_second; 1947 #if defined(DEBUG) 1948 const char *reason = NULL; /* XXX: gcc */ 1949 #endif 1950 1951 if (timespeccmp(&np->n_vattr->va_mtime, mtime, <=)) { 1952 #if defined(DEBUG) 1953 reason = "mtime"; 1954 #endif 1955 error = EINVAL; 1956 } 1957 1958 if (vp->v_type == VDIR && 1959 timespeccmp(&np->n_vattr->va_ctime, ctime, <=)) { 1960 #if defined(DEBUG) 1961 reason = "ctime"; 1962 #endif 1963 error = EINVAL; 1964 } 1965 1966 nmp = VFSTONFS(vp->v_mount); 1967 if (error) { 1968 1969 /* 1970 * despite of the fact that we've updated the file, 1971 * timestamps of the file were not updated as we 1972 * expected. 1973 * it means that the server has incompatible 1974 * semantics of timestamps or (more likely) 1975 * the server time is not precise enough to 1976 * track each modifications. 1977 * in that case, we disable wcc processing. 1978 * 1979 * yes, strictly speaking, we should disable all 1980 * caching. it's a compromise. 1981 */ 1982 1983 simple_lock(&nmp->nm_slock); 1984 #if defined(DEBUG) 1985 if (!NFS_WCCKLUDGE(nmp, now)) { 1986 printf("%s: inaccurate wcc data (%s) detected," 1987 " disabling wcc\n", 1988 vp->v_mount->mnt_stat.f_mntfromname, 1989 reason); 1990 } 1991 #endif 1992 nmp->nm_iflag |= NFSMNT_WCCKLUDGE; 1993 nmp->nm_wcckludgetime = now; 1994 simple_unlock(&nmp->nm_slock); 1995 } else if (NFS_WCCKLUDGE(nmp, now)) { 1996 error = EPERM; /* XXX */ 1997 } else if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 1998 simple_lock(&nmp->nm_slock); 1999 if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 2000 #if defined(DEBUG) 2001 printf("%s: re-enabling wcc\n", 2002 vp->v_mount->mnt_stat.f_mntfromname); 2003 #endif 2004 nmp->nm_iflag &= ~NFSMNT_WCCKLUDGE; 2005 } 2006 simple_unlock(&nmp->nm_slock); 2007 } 2008 } 2009 2010 #endif /* !defined(NFS_V2_ONLY) */ 2011 2012 return error; 2013 } 2014 2015 /* 2016 * Heuristic to see if the server XDR encodes directory cookies or not. 2017 * it is not supposed to, but a lot of servers may do this. Also, since 2018 * most/all servers will implement V2 as well, it is expected that they 2019 * may return just 32 bits worth of cookie information, so we need to 2020 * find out in which 32 bits this information is available. We do this 2021 * to avoid trouble with emulated binaries that can't handle 64 bit 2022 * directory offsets. 2023 */ 2024 2025 void 2026 nfs_cookieheuristic(vp, flagp, l, cred) 2027 struct vnode *vp; 2028 int *flagp; 2029 struct lwp *l; 2030 kauth_cred_t cred; 2031 { 2032 struct uio auio; 2033 struct iovec aiov; 2034 caddr_t tbuf, cp; 2035 struct dirent *dp; 2036 off_t *cookies = NULL, *cop; 2037 int error, eof, nc, len; 2038 2039 MALLOC(tbuf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 2040 2041 aiov.iov_base = tbuf; 2042 aiov.iov_len = NFS_DIRFRAGSIZ; 2043 auio.uio_iov = &aiov; 2044 auio.uio_iovcnt = 1; 2045 auio.uio_rw = UIO_READ; 2046 auio.uio_resid = NFS_DIRFRAGSIZ; 2047 auio.uio_offset = 0; 2048 UIO_SETUP_SYSSPACE(&auio); 2049 2050 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 2051 2052 len = NFS_DIRFRAGSIZ - auio.uio_resid; 2053 if (error || len == 0) { 2054 FREE(tbuf, M_TEMP); 2055 if (cookies) 2056 free(cookies, M_TEMP); 2057 return; 2058 } 2059 2060 /* 2061 * Find the first valid entry and look at its offset cookie. 2062 */ 2063 2064 cp = tbuf; 2065 for (cop = cookies; len > 0; len -= dp->d_reclen) { 2066 dp = (struct dirent *)cp; 2067 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 2068 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 2069 *flagp |= NFSMNT_SWAPCOOKIE; 2070 nfs_invaldircache(vp, 0); 2071 nfs_vinvalbuf(vp, 0, cred, l, 1); 2072 } 2073 break; 2074 } 2075 cop++; 2076 cp += dp->d_reclen; 2077 } 2078 2079 FREE(tbuf, M_TEMP); 2080 free(cookies, M_TEMP); 2081 } 2082 #endif /* NFS */ 2083 2084 #ifdef NFSSERVER 2085 /* 2086 * Set up nameidata for a lookup() call and do it. 2087 * 2088 * If pubflag is set, this call is done for a lookup operation on the 2089 * public filehandle. In that case we allow crossing mountpoints and 2090 * absolute pathnames. However, the caller is expected to check that 2091 * the lookup result is within the public fs, and deny access if 2092 * it is not. 2093 */ 2094 int 2095 nfs_namei(ndp, nsfh, len, slp, nam, mdp, dposp, retdirp, l, kerbflag, pubflag) 2096 struct nameidata *ndp; 2097 nfsrvfh_t *nsfh; 2098 uint32_t len; 2099 struct nfssvc_sock *slp; 2100 struct mbuf *nam; 2101 struct mbuf **mdp; 2102 caddr_t *dposp; 2103 struct vnode **retdirp; 2104 struct lwp *l; 2105 int kerbflag, pubflag; 2106 { 2107 int i, rem; 2108 struct mbuf *md; 2109 char *fromcp, *tocp, *cp; 2110 struct iovec aiov; 2111 struct uio auio; 2112 struct vnode *dp; 2113 int error, rdonly, linklen; 2114 struct componentname *cnp = &ndp->ni_cnd; 2115 2116 *retdirp = (struct vnode *)0; 2117 2118 if ((len + 1) > MAXPATHLEN) 2119 return (ENAMETOOLONG); 2120 if (len == 0) 2121 return (EACCES); 2122 cnp->cn_pnbuf = PNBUF_GET(); 2123 2124 /* 2125 * Copy the name from the mbuf list to ndp->ni_pnbuf 2126 * and set the various ndp fields appropriately. 2127 */ 2128 fromcp = *dposp; 2129 tocp = cnp->cn_pnbuf; 2130 md = *mdp; 2131 rem = mtod(md, caddr_t) + md->m_len - fromcp; 2132 for (i = 0; i < len; i++) { 2133 while (rem == 0) { 2134 md = md->m_next; 2135 if (md == NULL) { 2136 error = EBADRPC; 2137 goto out; 2138 } 2139 fromcp = mtod(md, caddr_t); 2140 rem = md->m_len; 2141 } 2142 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 2143 error = EACCES; 2144 goto out; 2145 } 2146 *tocp++ = *fromcp++; 2147 rem--; 2148 } 2149 *tocp = '\0'; 2150 *mdp = md; 2151 *dposp = fromcp; 2152 len = nfsm_rndup(len)-len; 2153 if (len > 0) { 2154 if (rem >= len) 2155 *dposp += len; 2156 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 2157 goto out; 2158 } 2159 2160 /* 2161 * Extract and set starting directory. 2162 */ 2163 error = nfsrv_fhtovp(nsfh, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 2164 nam, &rdonly, kerbflag, pubflag); 2165 if (error) 2166 goto out; 2167 if (dp->v_type != VDIR) { 2168 vrele(dp); 2169 error = ENOTDIR; 2170 goto out; 2171 } 2172 2173 if (rdonly) 2174 cnp->cn_flags |= RDONLY; 2175 2176 *retdirp = dp; 2177 2178 if (pubflag) { 2179 /* 2180 * Oh joy. For WebNFS, handle those pesky '%' escapes, 2181 * and the 'native path' indicator. 2182 */ 2183 cp = PNBUF_GET(); 2184 fromcp = cnp->cn_pnbuf; 2185 tocp = cp; 2186 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 2187 switch ((unsigned char)*fromcp) { 2188 case WEBNFS_NATIVE_CHAR: 2189 /* 2190 * 'Native' path for us is the same 2191 * as a path according to the NFS spec, 2192 * just skip the escape char. 2193 */ 2194 fromcp++; 2195 break; 2196 /* 2197 * More may be added in the future, range 0x80-0xff 2198 */ 2199 default: 2200 error = EIO; 2201 PNBUF_PUT(cp); 2202 goto out; 2203 } 2204 } 2205 /* 2206 * Translate the '%' escapes, URL-style. 2207 */ 2208 while (*fromcp != '\0') { 2209 if (*fromcp == WEBNFS_ESC_CHAR) { 2210 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 2211 fromcp++; 2212 *tocp++ = HEXSTRTOI(fromcp); 2213 fromcp += 2; 2214 continue; 2215 } else { 2216 error = ENOENT; 2217 PNBUF_PUT(cp); 2218 goto out; 2219 } 2220 } else 2221 *tocp++ = *fromcp++; 2222 } 2223 *tocp = '\0'; 2224 PNBUF_PUT(cnp->cn_pnbuf); 2225 cnp->cn_pnbuf = cp; 2226 } 2227 2228 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 2229 ndp->ni_segflg = UIO_SYSSPACE; 2230 ndp->ni_rootdir = rootvnode; 2231 2232 if (pubflag) { 2233 ndp->ni_loopcnt = 0; 2234 if (cnp->cn_pnbuf[0] == '/') 2235 dp = rootvnode; 2236 } else { 2237 cnp->cn_flags |= NOCROSSMOUNT; 2238 } 2239 2240 cnp->cn_lwp = l; 2241 VREF(dp); 2242 2243 for (;;) { 2244 cnp->cn_nameptr = cnp->cn_pnbuf; 2245 ndp->ni_startdir = dp; 2246 /* 2247 * And call lookup() to do the real work 2248 */ 2249 error = lookup(ndp); 2250 if (error) { 2251 PNBUF_PUT(cnp->cn_pnbuf); 2252 return (error); 2253 } 2254 /* 2255 * Check for encountering a symbolic link 2256 */ 2257 if ((cnp->cn_flags & ISSYMLINK) == 0) { 2258 if (cnp->cn_flags & (SAVENAME | SAVESTART)) 2259 cnp->cn_flags |= HASBUF; 2260 else 2261 PNBUF_PUT(cnp->cn_pnbuf); 2262 return (0); 2263 } else { 2264 if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN)) 2265 VOP_UNLOCK(ndp->ni_dvp, 0); 2266 if (!pubflag) { 2267 error = EINVAL; 2268 break; 2269 } 2270 2271 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 2272 error = ELOOP; 2273 break; 2274 } 2275 if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) { 2276 error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred, 2277 cnp->cn_lwp); 2278 if (error != 0) 2279 break; 2280 } 2281 if (ndp->ni_pathlen > 1) 2282 cp = PNBUF_GET(); 2283 else 2284 cp = cnp->cn_pnbuf; 2285 aiov.iov_base = cp; 2286 aiov.iov_len = MAXPATHLEN; 2287 auio.uio_iov = &aiov; 2288 auio.uio_iovcnt = 1; 2289 auio.uio_offset = 0; 2290 auio.uio_rw = UIO_READ; 2291 auio.uio_resid = MAXPATHLEN; 2292 UIO_SETUP_SYSSPACE(&auio); 2293 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 2294 if (error) { 2295 badlink: 2296 if (ndp->ni_pathlen > 1) 2297 PNBUF_PUT(cp); 2298 break; 2299 } 2300 linklen = MAXPATHLEN - auio.uio_resid; 2301 if (linklen == 0) { 2302 error = ENOENT; 2303 goto badlink; 2304 } 2305 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 2306 error = ENAMETOOLONG; 2307 goto badlink; 2308 } 2309 if (ndp->ni_pathlen > 1) { 2310 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 2311 PNBUF_PUT(cnp->cn_pnbuf); 2312 cnp->cn_pnbuf = cp; 2313 } else 2314 cnp->cn_pnbuf[linklen] = '\0'; 2315 ndp->ni_pathlen += linklen; 2316 vput(ndp->ni_vp); 2317 dp = ndp->ni_dvp; 2318 /* 2319 * Check if root directory should replace current directory. 2320 */ 2321 if (cnp->cn_pnbuf[0] == '/') { 2322 vrele(dp); 2323 dp = ndp->ni_rootdir; 2324 VREF(dp); 2325 } 2326 } 2327 } 2328 vrele(ndp->ni_dvp); 2329 vput(ndp->ni_vp); 2330 ndp->ni_vp = NULL; 2331 out: 2332 PNBUF_PUT(cnp->cn_pnbuf); 2333 return (error); 2334 } 2335 #endif /* NFSSERVER */ 2336 2337 /* 2338 * A fiddled version of m_adj() that ensures null fill to a 32-bit 2339 * boundary and only trims off the back end 2340 * 2341 * 1. trim off 'len' bytes as m_adj(mp, -len). 2342 * 2. add zero-padding 'nul' bytes at the end of the mbuf chain. 2343 */ 2344 void 2345 nfs_zeropad(mp, len, nul) 2346 struct mbuf *mp; 2347 int len; 2348 int nul; 2349 { 2350 struct mbuf *m; 2351 int count; 2352 2353 /* 2354 * Trim from tail. Scan the mbuf chain, 2355 * calculating its length and finding the last mbuf. 2356 * If the adjustment only affects this mbuf, then just 2357 * adjust and return. Otherwise, rescan and truncate 2358 * after the remaining size. 2359 */ 2360 count = 0; 2361 m = mp; 2362 for (;;) { 2363 count += m->m_len; 2364 if (m->m_next == NULL) 2365 break; 2366 m = m->m_next; 2367 } 2368 2369 KDASSERT(count >= len); 2370 2371 if (m->m_len >= len) { 2372 m->m_len -= len; 2373 } else { 2374 count -= len; 2375 /* 2376 * Correct length for chain is "count". 2377 * Find the mbuf with last data, adjust its length, 2378 * and toss data from remaining mbufs on chain. 2379 */ 2380 for (m = mp; m; m = m->m_next) { 2381 if (m->m_len >= count) { 2382 m->m_len = count; 2383 break; 2384 } 2385 count -= m->m_len; 2386 } 2387 KASSERT(m && m->m_next); 2388 m_freem(m->m_next); 2389 m->m_next = NULL; 2390 } 2391 2392 KDASSERT(m->m_next == NULL); 2393 2394 /* 2395 * zero-padding. 2396 */ 2397 if (nul > 0) { 2398 char *cp; 2399 int i; 2400 2401 if (M_ROMAP(m) || M_TRAILINGSPACE(m) < nul) { 2402 struct mbuf *n; 2403 2404 KDASSERT(MLEN >= nul); 2405 n = m_get(M_WAIT, MT_DATA); 2406 MCLAIM(n, &nfs_mowner); 2407 n->m_len = nul; 2408 n->m_next = NULL; 2409 m->m_next = n; 2410 cp = mtod(n, caddr_t); 2411 } else { 2412 cp = mtod(m, caddr_t) + m->m_len; 2413 m->m_len += nul; 2414 } 2415 for (i = 0; i < nul; i++) 2416 *cp++ = '\0'; 2417 } 2418 return; 2419 } 2420 2421 /* 2422 * Make these functions instead of macros, so that the kernel text size 2423 * doesn't get too big... 2424 */ 2425 void 2426 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2427 struct nfsrv_descript *nfsd; 2428 int before_ret; 2429 struct vattr *before_vap; 2430 int after_ret; 2431 struct vattr *after_vap; 2432 struct mbuf **mbp; 2433 char **bposp; 2434 { 2435 struct mbuf *mb = *mbp; 2436 char *bpos = *bposp; 2437 u_int32_t *tl; 2438 2439 if (before_ret) { 2440 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2441 *tl = nfs_false; 2442 } else { 2443 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2444 *tl++ = nfs_true; 2445 txdr_hyper(before_vap->va_size, tl); 2446 tl += 2; 2447 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2448 tl += 2; 2449 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2450 } 2451 *bposp = bpos; 2452 *mbp = mb; 2453 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2454 } 2455 2456 void 2457 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2458 struct nfsrv_descript *nfsd; 2459 int after_ret; 2460 struct vattr *after_vap; 2461 struct mbuf **mbp; 2462 char **bposp; 2463 { 2464 struct mbuf *mb = *mbp; 2465 char *bpos = *bposp; 2466 u_int32_t *tl; 2467 struct nfs_fattr *fp; 2468 2469 if (after_ret) { 2470 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2471 *tl = nfs_false; 2472 } else { 2473 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2474 *tl++ = nfs_true; 2475 fp = (struct nfs_fattr *)tl; 2476 nfsm_srvfattr(nfsd, after_vap, fp); 2477 } 2478 *mbp = mb; 2479 *bposp = bpos; 2480 } 2481 2482 void 2483 nfsm_srvfattr(nfsd, vap, fp) 2484 struct nfsrv_descript *nfsd; 2485 struct vattr *vap; 2486 struct nfs_fattr *fp; 2487 { 2488 2489 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2490 fp->fa_uid = txdr_unsigned(vap->va_uid); 2491 fp->fa_gid = txdr_unsigned(vap->va_gid); 2492 if (nfsd->nd_flag & ND_NFSV3) { 2493 fp->fa_type = vtonfsv3_type(vap->va_type); 2494 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2495 txdr_hyper(vap->va_size, &fp->fa3_size); 2496 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2497 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2498 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2499 fp->fa3_fsid.nfsuquad[0] = 0; 2500 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2501 txdr_hyper(vap->va_fileid, &fp->fa3_fileid); 2502 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2503 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2504 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2505 } else { 2506 fp->fa_type = vtonfsv2_type(vap->va_type); 2507 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2508 fp->fa2_size = txdr_unsigned(vap->va_size); 2509 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2510 if (vap->va_type == VFIFO) 2511 fp->fa2_rdev = 0xffffffff; 2512 else 2513 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2514 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2515 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2516 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2517 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2518 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2519 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2520 } 2521 } 2522 2523 #ifdef NFSSERVER 2524 /* 2525 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2526 * - look up fsid in mount list (if not found ret error) 2527 * - get vp and export rights by calling VFS_FHTOVP() 2528 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2529 * - if not lockflag unlock it with VOP_UNLOCK() 2530 */ 2531 int 2532 nfsrv_fhtovp( 2533 nfsrvfh_t *nsfh, 2534 int lockflag, 2535 struct vnode **vpp, 2536 kauth_cred_t cred, 2537 struct nfssvc_sock *slp __unused, 2538 struct mbuf *nam, 2539 int *rdonlyp, 2540 int kerbflag, 2541 int pubflag 2542 ) 2543 { 2544 struct mount *mp; 2545 kauth_cred_t credanon; 2546 int error, exflags; 2547 struct sockaddr_in *saddr; 2548 fhandle_t *fhp; 2549 2550 fhp = NFSRVFH_FHANDLE(nsfh); 2551 *vpp = (struct vnode *)0; 2552 2553 if (nfs_ispublicfh(nsfh)) { 2554 if (!pubflag || !nfs_pub.np_valid) 2555 return (ESTALE); 2556 fhp = nfs_pub.np_handle; 2557 } 2558 2559 error = netexport_check(&fhp->fh_fsid, nam, &mp, &exflags, &credanon); 2560 if (error) { 2561 return error; 2562 } 2563 2564 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2565 if (error) 2566 return (error); 2567 2568 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2569 saddr = mtod(nam, struct sockaddr_in *); 2570 if ((saddr->sin_family == AF_INET) && 2571 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2572 vput(*vpp); 2573 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2574 } 2575 #ifdef INET6 2576 if ((saddr->sin_family == AF_INET6) && 2577 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2578 vput(*vpp); 2579 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2580 } 2581 #endif 2582 } 2583 /* 2584 * Check/setup credentials. 2585 */ 2586 if (exflags & MNT_EXKERB) { 2587 if (!kerbflag) { 2588 vput(*vpp); 2589 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2590 } 2591 } else if (kerbflag) { 2592 vput(*vpp); 2593 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2594 } else if (kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER, 2595 NULL) == 0 || (exflags & MNT_EXPORTANON)) { 2596 kauth_cred_clone(credanon, cred); 2597 } 2598 if (exflags & MNT_EXRDONLY) 2599 *rdonlyp = 1; 2600 else 2601 *rdonlyp = 0; 2602 if (!lockflag) 2603 VOP_UNLOCK(*vpp, 0); 2604 return (0); 2605 } 2606 2607 /* 2608 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2609 * means a length of 0, for v2 it means all zeroes. 2610 */ 2611 int 2612 nfs_ispublicfh(const nfsrvfh_t *nsfh) 2613 { 2614 const char *cp = (const void *)(NFSRVFH_DATA(nsfh)); 2615 int i; 2616 2617 if (NFSRVFH_SIZE(nsfh) == 0) { 2618 return TRUE; 2619 } 2620 if (NFSRVFH_SIZE(nsfh) != NFSX_V2FH) { 2621 return FALSE; 2622 } 2623 for (i = 0; i < NFSX_V2FH; i++) 2624 if (*cp++ != 0) 2625 return FALSE; 2626 return TRUE; 2627 } 2628 #endif /* NFSSERVER */ 2629 2630 /* 2631 * This function compares two net addresses by family and returns TRUE 2632 * if they are the same host. 2633 * If there is any doubt, return FALSE. 2634 * The AF_INET family is handled as a special case so that address mbufs 2635 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2636 */ 2637 int 2638 netaddr_match(family, haddr, nam) 2639 int family; 2640 union nethostaddr *haddr; 2641 struct mbuf *nam; 2642 { 2643 struct sockaddr_in *inetaddr; 2644 2645 switch (family) { 2646 case AF_INET: 2647 inetaddr = mtod(nam, struct sockaddr_in *); 2648 if (inetaddr->sin_family == AF_INET && 2649 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2650 return (1); 2651 break; 2652 #ifdef INET6 2653 case AF_INET6: 2654 { 2655 struct sockaddr_in6 *sin6_1, *sin6_2; 2656 2657 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2658 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2659 if (sin6_1->sin6_family == AF_INET6 && 2660 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2661 return 1; 2662 } 2663 #endif 2664 #ifdef ISO 2665 case AF_ISO: 2666 { 2667 struct sockaddr_iso *isoaddr1, *isoaddr2; 2668 2669 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2670 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2671 if (isoaddr1->siso_family == AF_ISO && 2672 isoaddr1->siso_nlen > 0 && 2673 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2674 SAME_ISOADDR(isoaddr1, isoaddr2)) 2675 return (1); 2676 break; 2677 } 2678 #endif /* ISO */ 2679 default: 2680 break; 2681 }; 2682 return (0); 2683 } 2684 2685 /* 2686 * The write verifier has changed (probably due to a server reboot), so all 2687 * PG_NEEDCOMMIT pages will have to be written again. Since they are marked 2688 * as dirty or are being written out just now, all this takes is clearing 2689 * the PG_NEEDCOMMIT flag. Once done the new write verifier can be set for 2690 * the mount point. 2691 */ 2692 void 2693 nfs_clearcommit(mp) 2694 struct mount *mp; 2695 { 2696 struct vnode *vp; 2697 struct nfsnode *np; 2698 struct vm_page *pg; 2699 struct nfsmount *nmp = VFSTONFS(mp); 2700 2701 lockmgr(&nmp->nm_writeverflock, LK_EXCLUSIVE, NULL); 2702 2703 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2704 KASSERT(vp->v_mount == mp); 2705 if (vp->v_type != VREG) 2706 continue; 2707 np = VTONFS(vp); 2708 np->n_pushlo = np->n_pushhi = np->n_pushedlo = 2709 np->n_pushedhi = 0; 2710 np->n_commitflags &= 2711 ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID); 2712 simple_lock(&vp->v_uobj.vmobjlock); 2713 TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) { 2714 pg->flags &= ~PG_NEEDCOMMIT; 2715 } 2716 simple_unlock(&vp->v_uobj.vmobjlock); 2717 } 2718 simple_lock(&nmp->nm_slock); 2719 nmp->nm_iflag &= ~NFSMNT_STALEWRITEVERF; 2720 simple_unlock(&nmp->nm_slock); 2721 lockmgr(&nmp->nm_writeverflock, LK_RELEASE, NULL); 2722 } 2723 2724 void 2725 nfs_merge_commit_ranges(vp) 2726 struct vnode *vp; 2727 { 2728 struct nfsnode *np = VTONFS(vp); 2729 2730 KASSERT(np->n_commitflags & NFS_COMMIT_PUSH_VALID); 2731 2732 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2733 np->n_pushedlo = np->n_pushlo; 2734 np->n_pushedhi = np->n_pushhi; 2735 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2736 } else { 2737 if (np->n_pushlo < np->n_pushedlo) 2738 np->n_pushedlo = np->n_pushlo; 2739 if (np->n_pushhi > np->n_pushedhi) 2740 np->n_pushedhi = np->n_pushhi; 2741 } 2742 2743 np->n_pushlo = np->n_pushhi = 0; 2744 np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID; 2745 2746 #ifdef NFS_DEBUG_COMMIT 2747 printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2748 (unsigned)np->n_pushedhi); 2749 #endif 2750 } 2751 2752 int 2753 nfs_in_committed_range(vp, off, len) 2754 struct vnode *vp; 2755 off_t off, len; 2756 { 2757 struct nfsnode *np = VTONFS(vp); 2758 off_t lo, hi; 2759 2760 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2761 return 0; 2762 lo = off; 2763 hi = lo + len; 2764 2765 return (lo >= np->n_pushedlo && hi <= np->n_pushedhi); 2766 } 2767 2768 int 2769 nfs_in_tobecommitted_range(vp, off, len) 2770 struct vnode *vp; 2771 off_t off, len; 2772 { 2773 struct nfsnode *np = VTONFS(vp); 2774 off_t lo, hi; 2775 2776 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2777 return 0; 2778 lo = off; 2779 hi = lo + len; 2780 2781 return (lo >= np->n_pushlo && hi <= np->n_pushhi); 2782 } 2783 2784 void 2785 nfs_add_committed_range(vp, off, len) 2786 struct vnode *vp; 2787 off_t off, len; 2788 { 2789 struct nfsnode *np = VTONFS(vp); 2790 off_t lo, hi; 2791 2792 lo = off; 2793 hi = lo + len; 2794 2795 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2796 np->n_pushedlo = lo; 2797 np->n_pushedhi = hi; 2798 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2799 } else { 2800 if (hi > np->n_pushedhi) 2801 np->n_pushedhi = hi; 2802 if (lo < np->n_pushedlo) 2803 np->n_pushedlo = lo; 2804 } 2805 #ifdef NFS_DEBUG_COMMIT 2806 printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2807 (unsigned)np->n_pushedhi); 2808 #endif 2809 } 2810 2811 void 2812 nfs_del_committed_range(vp, off, len) 2813 struct vnode *vp; 2814 off_t off, len; 2815 { 2816 struct nfsnode *np = VTONFS(vp); 2817 off_t lo, hi; 2818 2819 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2820 return; 2821 2822 lo = off; 2823 hi = lo + len; 2824 2825 if (lo > np->n_pushedhi || hi < np->n_pushedlo) 2826 return; 2827 if (lo <= np->n_pushedlo) 2828 np->n_pushedlo = hi; 2829 else if (hi >= np->n_pushedhi) 2830 np->n_pushedhi = lo; 2831 else { 2832 /* 2833 * XXX There's only one range. If the deleted range 2834 * is in the middle, pick the largest of the 2835 * contiguous ranges that it leaves. 2836 */ 2837 if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi)) 2838 np->n_pushedhi = lo; 2839 else 2840 np->n_pushedlo = hi; 2841 } 2842 #ifdef NFS_DEBUG_COMMIT 2843 printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2844 (unsigned)np->n_pushedhi); 2845 #endif 2846 } 2847 2848 void 2849 nfs_add_tobecommitted_range(vp, off, len) 2850 struct vnode *vp; 2851 off_t off, len; 2852 { 2853 struct nfsnode *np = VTONFS(vp); 2854 off_t lo, hi; 2855 2856 lo = off; 2857 hi = lo + len; 2858 2859 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) { 2860 np->n_pushlo = lo; 2861 np->n_pushhi = hi; 2862 np->n_commitflags |= NFS_COMMIT_PUSH_VALID; 2863 } else { 2864 if (lo < np->n_pushlo) 2865 np->n_pushlo = lo; 2866 if (hi > np->n_pushhi) 2867 np->n_pushhi = hi; 2868 } 2869 #ifdef NFS_DEBUG_COMMIT 2870 printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2871 (unsigned)np->n_pushhi); 2872 #endif 2873 } 2874 2875 void 2876 nfs_del_tobecommitted_range(vp, off, len) 2877 struct vnode *vp; 2878 off_t off, len; 2879 { 2880 struct nfsnode *np = VTONFS(vp); 2881 off_t lo, hi; 2882 2883 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2884 return; 2885 2886 lo = off; 2887 hi = lo + len; 2888 2889 if (lo > np->n_pushhi || hi < np->n_pushlo) 2890 return; 2891 2892 if (lo <= np->n_pushlo) 2893 np->n_pushlo = hi; 2894 else if (hi >= np->n_pushhi) 2895 np->n_pushhi = lo; 2896 else { 2897 /* 2898 * XXX There's only one range. If the deleted range 2899 * is in the middle, pick the largest of the 2900 * contiguous ranges that it leaves. 2901 */ 2902 if ((np->n_pushlo - lo) > (hi - np->n_pushhi)) 2903 np->n_pushhi = lo; 2904 else 2905 np->n_pushlo = hi; 2906 } 2907 #ifdef NFS_DEBUG_COMMIT 2908 printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2909 (unsigned)np->n_pushhi); 2910 #endif 2911 } 2912 2913 /* 2914 * Map errnos to NFS error numbers. For Version 3 also filter out error 2915 * numbers not specified for the associated procedure. 2916 */ 2917 int 2918 nfsrv_errmap(nd, err) 2919 struct nfsrv_descript *nd; 2920 int err; 2921 { 2922 const short *defaulterrp, *errp; 2923 2924 if (nd->nd_flag & ND_NFSV3) { 2925 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2926 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2927 while (*++errp) { 2928 if (*errp == err) 2929 return (err); 2930 else if (*errp > err) 2931 break; 2932 } 2933 return ((int)*defaulterrp); 2934 } else 2935 return (err & 0xffff); 2936 } 2937 if (err <= ELAST) 2938 return ((int)nfsrv_v2errmap[err - 1]); 2939 return (NFSERR_IO); 2940 } 2941 2942 u_int32_t 2943 nfs_getxid() 2944 { 2945 static u_int32_t base; 2946 static u_int32_t nfs_xid = 0; 2947 static struct simplelock nfs_xidlock = SIMPLELOCK_INITIALIZER; 2948 u_int32_t newxid; 2949 2950 simple_lock(&nfs_xidlock); 2951 /* 2952 * derive initial xid from system time 2953 * XXX time is invalid if root not yet mounted 2954 */ 2955 if (__predict_false(!base && (rootvp))) { 2956 struct timeval tv; 2957 2958 microtime(&tv); 2959 base = tv.tv_sec << 12; 2960 nfs_xid = base; 2961 } 2962 2963 /* 2964 * Skip zero xid if it should ever happen. 2965 */ 2966 if (__predict_false(++nfs_xid == 0)) 2967 nfs_xid++; 2968 newxid = nfs_xid; 2969 simple_unlock(&nfs_xidlock); 2970 2971 return txdr_unsigned(newxid); 2972 } 2973 2974 /* 2975 * assign a new xid for existing request. 2976 * used for NFSERR_JUKEBOX handling. 2977 */ 2978 void 2979 nfs_renewxid(struct nfsreq *req) 2980 { 2981 u_int32_t xid; 2982 int off; 2983 2984 xid = nfs_getxid(); 2985 if (req->r_nmp->nm_sotype == SOCK_STREAM) 2986 off = sizeof(u_int32_t); /* RPC record mark */ 2987 else 2988 off = 0; 2989 2990 m_copyback(req->r_mreq, off, sizeof(xid), (void *)&xid); 2991 req->r_xid = xid; 2992 } 2993 2994 #if defined(NFSSERVER) 2995 int 2996 nfsrv_composefh(struct vnode *vp, nfsrvfh_t *nsfh, boolean_t v3) 2997 { 2998 int error; 2999 size_t fhsize; 3000 3001 fhsize = NFSD_MAXFHSIZE; 3002 error = vfs_composefh(vp, (void *)NFSRVFH_DATA(nsfh), &fhsize); 3003 if (NFSX_FHTOOBIG_P(fhsize, v3)) { 3004 error = EOPNOTSUPP; 3005 } 3006 if (error != 0) { 3007 return error; 3008 } 3009 if (!v3 && fhsize < NFSX_V2FH) { 3010 memset((char *)NFSRVFH_DATA(nsfh) + fhsize, 0, 3011 NFSX_V2FH - fhsize); 3012 fhsize = NFSX_V2FH; 3013 } 3014 if ((fhsize % NFSX_UNSIGNED) != 0) { 3015 return EOPNOTSUPP; 3016 } 3017 nsfh->nsfh_size = fhsize; 3018 return 0; 3019 } 3020 3021 int 3022 nfsrv_comparefh(const nfsrvfh_t *fh1, const nfsrvfh_t *fh2) 3023 { 3024 3025 if (NFSRVFH_SIZE(fh1) != NFSRVFH_SIZE(fh2)) { 3026 return NFSRVFH_SIZE(fh2) - NFSRVFH_SIZE(fh1); 3027 } 3028 return memcmp(NFSRVFH_DATA(fh1), NFSRVFH_DATA(fh2), NFSRVFH_SIZE(fh1)); 3029 } 3030 3031 void 3032 nfsrv_copyfh(nfsrvfh_t *fh1, const nfsrvfh_t *fh2) 3033 { 3034 size_t size; 3035 3036 fh1->nsfh_size = size = NFSRVFH_SIZE(fh2); 3037 memcpy(NFSRVFH_DATA(fh1), NFSRVFH_DATA(fh2), size); 3038 } 3039 #endif /* defined(NFSSERVER) */ 3040
Indexes created Thu Sep 25 16:09:42 GMT 2025