nfs_subs.c revision 1.128
1 /* $NetBSD: nfs_subs.c,v 1.128 2003/09/26 11:51:53 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 1.128 2003/09/26 11:51:53 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 dst = m1->m_dat; 1033 } else { 1034 /* 1035 * If the first mbuf has no external data 1036 * move the data to the front of the mbuf. 1037 */ 1038 if ((dst = m1->m_dat) != src) 1039 memmove(dst, src, left); 1040 dst += left; 1041 m1->m_len = left; 1042 m2 = m1->m_next; 1043 } 1044 m1->m_flags &= ~M_PKTHDR; 1045 *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */ 1046 *dposp = mtod(m1, caddr_t) + siz; 1047 /* 1048 * Loop through mbufs pulling data up into first mbuf until 1049 * the first mbuf is full or there is no more data to 1050 * pullup. 1051 */ 1052 while ((len = (MLEN - m1->m_len)) != 0 && m2) { 1053 if ((len = min(len, m2->m_len)) != 0) 1054 memcpy(dst, m2->m_data, len); 1055 m1->m_len += len; 1056 dst += len; 1057 m2->m_data += len; 1058 m2->m_len -= len; 1059 m2 = m2->m_next; 1060 } 1061 if (m1->m_len < siz) 1062 return (EBADRPC); 1063 return (0); 1064 } 1065 1066 /* 1067 * Advance the position in the mbuf chain. 1068 */ 1069 int 1070 nfs_adv(mdp, dposp, offs, left) 1071 struct mbuf **mdp; 1072 caddr_t *dposp; 1073 int offs; 1074 int left; 1075 { 1076 struct mbuf *m; 1077 int s; 1078 1079 m = *mdp; 1080 s = left; 1081 while (s < offs) { 1082 offs -= s; 1083 m = m->m_next; 1084 if (m == NULL) 1085 return (EBADRPC); 1086 s = m->m_len; 1087 } 1088 *mdp = m; 1089 *dposp = mtod(m, caddr_t)+offs; 1090 return (0); 1091 } 1092 1093 /* 1094 * Copy a string into mbufs for the hard cases... 1095 */ 1096 int 1097 nfsm_strtmbuf(mb, bpos, cp, siz) 1098 struct mbuf **mb; 1099 char **bpos; 1100 const char *cp; 1101 long siz; 1102 { 1103 struct mbuf *m1 = NULL, *m2; 1104 long left, xfer, len, tlen; 1105 u_int32_t *tl; 1106 int putsize; 1107 1108 putsize = 1; 1109 m2 = *mb; 1110 left = M_TRAILINGSPACE(m2); 1111 if (left > 0) { 1112 tl = ((u_int32_t *)(*bpos)); 1113 *tl++ = txdr_unsigned(siz); 1114 putsize = 0; 1115 left -= NFSX_UNSIGNED; 1116 m2->m_len += NFSX_UNSIGNED; 1117 if (left > 0) { 1118 memcpy((caddr_t) tl, cp, left); 1119 siz -= left; 1120 cp += left; 1121 m2->m_len += left; 1122 left = 0; 1123 } 1124 } 1125 /* Loop around adding mbufs */ 1126 while (siz > 0) { 1127 m1 = m_get(M_WAIT, MT_DATA); 1128 MCLAIM(m1, &nfs_mowner); 1129 if (siz > MLEN) 1130 m_clget(m1, M_WAIT); 1131 m1->m_len = NFSMSIZ(m1); 1132 m2->m_next = m1; 1133 m2 = m1; 1134 tl = mtod(m1, u_int32_t *); 1135 tlen = 0; 1136 if (putsize) { 1137 *tl++ = txdr_unsigned(siz); 1138 m1->m_len -= NFSX_UNSIGNED; 1139 tlen = NFSX_UNSIGNED; 1140 putsize = 0; 1141 } 1142 if (siz < m1->m_len) { 1143 len = nfsm_rndup(siz); 1144 xfer = siz; 1145 if (xfer < len) 1146 *(tl+(xfer>>2)) = 0; 1147 } else { 1148 xfer = len = m1->m_len; 1149 } 1150 memcpy((caddr_t) tl, cp, xfer); 1151 m1->m_len = len+tlen; 1152 siz -= xfer; 1153 cp += xfer; 1154 } 1155 *mb = m1; 1156 *bpos = mtod(m1, caddr_t)+m1->m_len; 1157 return (0); 1158 } 1159 1160 /* 1161 * Directory caching routines. They work as follows: 1162 * - a cache is maintained per VDIR nfsnode. 1163 * - for each offset cookie that is exported to userspace, and can 1164 * thus be thrown back at us as an offset to VOP_READDIR, store 1165 * information in the cache. 1166 * - cached are: 1167 * - cookie itself 1168 * - blocknumber (essentially just a search key in the buffer cache) 1169 * - entry number in block. 1170 * - offset cookie of block in which this entry is stored 1171 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1172 * - entries are looked up in a hash table 1173 * - also maintained is an LRU list of entries, used to determine 1174 * which ones to delete if the cache grows too large. 1175 * - if 32 <-> 64 translation mode is requested for a filesystem, 1176 * the cache also functions as a translation table 1177 * - in the translation case, invalidating the cache does not mean 1178 * flushing it, but just marking entries as invalid, except for 1179 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1180 * still be able to use the cache as a translation table. 1181 * - 32 bit cookies are uniquely created by combining the hash table 1182 * entry value, and one generation count per hash table entry, 1183 * incremented each time an entry is appended to the chain. 1184 * - the cache is invalidated each time a direcory is modified 1185 * - sanity checks are also done; if an entry in a block turns 1186 * out not to have a matching cookie, the cache is invalidated 1187 * and a new block starting from the wanted offset is fetched from 1188 * the server. 1189 * - directory entries as read from the server are extended to contain 1190 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1191 * the cache and exporting them to userspace through the cookie 1192 * argument to VOP_READDIR. 1193 */ 1194 1195 u_long 1196 nfs_dirhash(off) 1197 off_t off; 1198 { 1199 int i; 1200 char *cp = (char *)&off; 1201 u_long sum = 0L; 1202 1203 for (i = 0 ; i < sizeof (off); i++) 1204 sum += *cp++; 1205 1206 return sum; 1207 } 1208 1209 void 1210 nfs_initdircache(vp) 1211 struct vnode *vp; 1212 { 1213 struct nfsnode *np = VTONFS(vp); 1214 1215 KASSERT(np->n_dircache == NULL); 1216 1217 np->n_dircachesize = 0; 1218 np->n_dblkno = 1; 1219 np->n_dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, M_NFSDIROFF, 1220 M_WAITOK, &nfsdirhashmask); 1221 TAILQ_INIT(&np->n_dirchain); 1222 } 1223 1224 void 1225 nfs_initdirxlatecookie(vp) 1226 struct vnode *vp; 1227 { 1228 struct nfsnode *np = VTONFS(vp); 1229 1230 KASSERT(VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_XLATECOOKIE); 1231 KASSERT(np->n_dirgens == NULL); 1232 1233 MALLOC(np->n_dirgens, unsigned *, 1234 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, M_WAITOK); 1235 memset((caddr_t)np->n_dirgens, 0, NFS_DIRHASHSIZ * sizeof (unsigned)); 1236 } 1237 1238 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0}; 1239 1240 struct nfsdircache * 1241 nfs_searchdircache(vp, off, do32, hashent) 1242 struct vnode *vp; 1243 off_t off; 1244 int do32; 1245 int *hashent; 1246 { 1247 struct nfsdirhashhead *ndhp; 1248 struct nfsdircache *ndp = NULL; 1249 struct nfsnode *np = VTONFS(vp); 1250 unsigned ent; 1251 1252 /* 1253 * Zero is always a valid cookie. 1254 */ 1255 if (off == 0) 1256 return &dzero; 1257 1258 /* 1259 * We use a 32bit cookie as search key, directly reconstruct 1260 * the hashentry. Else use the hashfunction. 1261 */ 1262 if (do32) { 1263 ent = (u_int32_t)off >> 24; 1264 if (ent >= NFS_DIRHASHSIZ) 1265 return NULL; 1266 ndhp = &np->n_dircache[ent]; 1267 } else { 1268 ndhp = NFSDIRHASH(np, off); 1269 } 1270 1271 if (hashent) 1272 *hashent = (int)(ndhp - np->n_dircache); 1273 if (do32) { 1274 LIST_FOREACH(ndp, ndhp, dc_hash) { 1275 if (ndp->dc_cookie32 == (u_int32_t)off) { 1276 /* 1277 * An invalidated entry will become the 1278 * start of a new block fetched from 1279 * the server. 1280 */ 1281 if (ndp->dc_blkno == -1) { 1282 ndp->dc_blkcookie = ndp->dc_cookie; 1283 ndp->dc_blkno = np->n_dblkno++; 1284 ndp->dc_entry = 0; 1285 } 1286 break; 1287 } 1288 } 1289 } else { 1290 LIST_FOREACH(ndp, ndhp, dc_hash) { 1291 if (ndp->dc_cookie == off) 1292 break; 1293 } 1294 } 1295 return ndp; 1296 } 1297 1298 1299 struct nfsdircache * 1300 nfs_enterdircache(vp, off, blkoff, en, blkno) 1301 struct vnode *vp; 1302 off_t off, blkoff; 1303 int en; 1304 daddr_t blkno; 1305 { 1306 struct nfsnode *np = VTONFS(vp); 1307 struct nfsdirhashhead *ndhp; 1308 struct nfsdircache *ndp = NULL, *first; 1309 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1310 int hashent, gen, overwrite; 1311 1312 if (!np->n_dircache) 1313 /* 1314 * XXX would like to do this in nfs_nget but vtype 1315 * isn't known at that time. 1316 */ 1317 nfs_initdircache(vp); 1318 1319 if ((nmp->nm_flag & NFSMNT_XLATECOOKIE) && !np->n_dirgens) 1320 nfs_initdirxlatecookie(vp); 1321 1322 /* 1323 * XXX refuse entries for offset 0. amd(8) erroneously sets 1324 * cookie 0 for the '.' entry, making this necessary. This 1325 * isn't so bad, as 0 is a special case anyway. 1326 */ 1327 if (off == 0) 1328 return &dzero; 1329 1330 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1331 1332 if (ndp && ndp->dc_blkno != -1) { 1333 /* 1334 * Overwriting an old entry. Check if it's the same. 1335 * If so, just return. If not, remove the old entry. 1336 */ 1337 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1338 return ndp; 1339 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1340 LIST_REMOVE(ndp, dc_hash); 1341 FREE(ndp, M_NFSDIROFF); 1342 ndp = 0; 1343 } 1344 1345 ndhp = &np->n_dircache[hashent]; 1346 1347 if (!ndp) { 1348 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF, 1349 M_WAITOK); 1350 overwrite = 0; 1351 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1352 /* 1353 * We're allocating a new entry, so bump the 1354 * generation number. 1355 */ 1356 gen = ++np->n_dirgens[hashent]; 1357 if (gen == 0) { 1358 np->n_dirgens[hashent]++; 1359 gen++; 1360 } 1361 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1362 } 1363 } else 1364 overwrite = 1; 1365 1366 /* 1367 * If the entry number is 0, we are at the start of a new block, so 1368 * allocate a new blocknumber. 1369 */ 1370 if (en == 0) 1371 ndp->dc_blkno = np->n_dblkno++; 1372 else 1373 ndp->dc_blkno = blkno; 1374 1375 ndp->dc_cookie = off; 1376 ndp->dc_blkcookie = blkoff; 1377 ndp->dc_entry = en; 1378 1379 if (overwrite) 1380 return ndp; 1381 1382 /* 1383 * If the maximum directory cookie cache size has been reached 1384 * for this node, take one off the front. The idea is that 1385 * directories are typically read front-to-back once, so that 1386 * the oldest entries can be thrown away without much performance 1387 * loss. 1388 */ 1389 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1390 first = TAILQ_FIRST(&np->n_dirchain); 1391 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain); 1392 LIST_REMOVE(first, dc_hash); 1393 FREE(first, M_NFSDIROFF); 1394 } else 1395 np->n_dircachesize++; 1396 1397 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1398 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1399 return ndp; 1400 } 1401 1402 void 1403 nfs_invaldircache(vp, forcefree) 1404 struct vnode *vp; 1405 int forcefree; 1406 { 1407 struct nfsnode *np = VTONFS(vp); 1408 struct nfsdircache *ndp = NULL; 1409 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1410 1411 #ifdef DIAGNOSTIC 1412 if (vp->v_type != VDIR) 1413 panic("nfs: invaldircache: not dir"); 1414 #endif 1415 1416 if (!np->n_dircache) 1417 return; 1418 1419 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1420 while ((ndp = TAILQ_FIRST(&np->n_dirchain)) != 0) { 1421 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1422 LIST_REMOVE(ndp, dc_hash); 1423 FREE(ndp, M_NFSDIROFF); 1424 } 1425 np->n_dircachesize = 0; 1426 if (forcefree && np->n_dirgens) { 1427 FREE(np->n_dirgens, M_NFSDIROFF); 1428 np->n_dirgens = NULL; 1429 } 1430 } else { 1431 TAILQ_FOREACH(ndp, &np->n_dirchain, dc_chain) { 1432 ndp->dc_blkno = -1; 1433 } 1434 } 1435 1436 np->n_dblkno = 1; 1437 } 1438 1439 /* 1440 * Called once before VFS init to initialize shared and 1441 * server-specific data structures. 1442 */ 1443 void 1444 nfs_init() 1445 { 1446 nfsrtt.pos = 0; 1447 rpc_vers = txdr_unsigned(RPC_VER2); 1448 rpc_call = txdr_unsigned(RPC_CALL); 1449 rpc_reply = txdr_unsigned(RPC_REPLY); 1450 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1451 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1452 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1453 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1454 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1455 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1456 nfs_prog = txdr_unsigned(NFS_PROG); 1457 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1458 nfs_true = txdr_unsigned(TRUE); 1459 nfs_false = txdr_unsigned(FALSE); 1460 nfs_xdrneg1 = txdr_unsigned(-1); 1461 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1462 if (nfs_ticks < 1) 1463 nfs_ticks = 1; 1464 #ifdef NFSSERVER 1465 nfsrv_init(0); /* Init server data structures */ 1466 nfsrv_initcache(); /* Init the server request cache */ 1467 pool_init(&nfs_srvdesc_pool, sizeof(struct nfsrv_descript), 1468 0, 0, 0, "nfsrvdescpl", &pool_allocator_nointr); 1469 #endif /* NFSSERVER */ 1470 1471 #if defined(NFSSERVER) || !defined(NFS_V2_ONLY) 1472 /* 1473 * Initialize the nqnfs data structures. 1474 */ 1475 if (nqnfsstarttime == 0) { 1476 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1477 + nqsrv_clockskew + nqsrv_writeslack; 1478 NQLOADNOVRAM(nqnfsstarttime); 1479 CIRCLEQ_INIT(&nqtimerhead); 1480 nqfhhashtbl = hashinit(NQLCHSZ, HASH_LIST, M_NQLEASE, 1481 M_WAITOK, &nqfhhash); 1482 } 1483 #endif 1484 1485 exithook_establish(nfs_exit, NULL); 1486 1487 /* 1488 * Initialize reply list and start timer 1489 */ 1490 TAILQ_INIT(&nfs_reqq); 1491 nfs_timer(NULL); 1492 MOWNER_ATTACH(&nfs_mowner); 1493 1494 #ifdef NFS 1495 /* Initialize the kqueue structures */ 1496 nfs_kqinit(); 1497 /* Initialize the iod structures */ 1498 nfs_iodinit(); 1499 #endif 1500 } 1501 1502 #ifdef NFS 1503 /* 1504 * Called once at VFS init to initialize client-specific data structures. 1505 */ 1506 void 1507 nfs_vfs_init() 1508 { 1509 nfs_nhinit(); /* Init the nfsnode table */ 1510 nfs_commitsize = uvmexp.npages << (PAGE_SHIFT - 4); 1511 } 1512 1513 void 1514 nfs_vfs_reinit() 1515 { 1516 nfs_nhreinit(); 1517 } 1518 1519 void 1520 nfs_vfs_done() 1521 { 1522 nfs_nhdone(); 1523 } 1524 1525 /* 1526 * Attribute cache routines. 1527 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1528 * that are on the mbuf list 1529 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1530 * error otherwise 1531 */ 1532 1533 /* 1534 * Load the attribute cache (that lives in the nfsnode entry) with 1535 * the values on the mbuf list and 1536 * Iff vap not NULL 1537 * copy the attributes to *vaper 1538 */ 1539 int 1540 nfsm_loadattrcache(vpp, mdp, dposp, vaper, flags) 1541 struct vnode **vpp; 1542 struct mbuf **mdp; 1543 caddr_t *dposp; 1544 struct vattr *vaper; 1545 int flags; 1546 { 1547 int32_t t1; 1548 caddr_t cp2; 1549 int error = 0; 1550 struct mbuf *md; 1551 int v3 = NFS_ISV3(*vpp); 1552 1553 md = *mdp; 1554 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1555 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1556 if (error) 1557 return (error); 1558 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper, flags); 1559 } 1560 1561 int 1562 nfs_loadattrcache(vpp, fp, vaper, flags) 1563 struct vnode **vpp; 1564 struct nfs_fattr *fp; 1565 struct vattr *vaper; 1566 int flags; 1567 { 1568 struct vnode *vp = *vpp; 1569 struct vattr *vap; 1570 int v3 = NFS_ISV3(vp); 1571 enum vtype vtyp; 1572 u_short vmode; 1573 struct timespec mtime; 1574 struct vnode *nvp; 1575 int32_t rdev; 1576 struct nfsnode *np; 1577 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1578 uid_t uid; 1579 gid_t gid; 1580 1581 if (v3) { 1582 vtyp = nfsv3tov_type(fp->fa_type); 1583 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1584 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1585 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1586 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1587 } else { 1588 vtyp = nfsv2tov_type(fp->fa_type); 1589 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1590 if (vtyp == VNON || vtyp == VREG) 1591 vtyp = IFTOVT(vmode); 1592 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1593 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1594 1595 /* 1596 * Really ugly NFSv2 kludge. 1597 */ 1598 if (vtyp == VCHR && rdev == 0xffffffff) 1599 vtyp = VFIFO; 1600 } 1601 1602 vmode &= ALLPERMS; 1603 1604 /* 1605 * If v_type == VNON it is a new node, so fill in the v_type, 1606 * n_mtime fields. Check to see if it represents a special 1607 * device, and if so, check for a possible alias. Once the 1608 * correct vnode has been obtained, fill in the rest of the 1609 * information. 1610 */ 1611 np = VTONFS(vp); 1612 if (vp->v_type == VNON) { 1613 vp->v_type = vtyp; 1614 if (vp->v_type == VFIFO) { 1615 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1616 vp->v_op = fifo_nfsv2nodeop_p; 1617 } 1618 if (vp->v_type == VCHR || vp->v_type == VBLK) { 1619 vp->v_op = spec_nfsv2nodeop_p; 1620 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1621 if (nvp) { 1622 /* 1623 * Discard unneeded vnode, but save its nfsnode. 1624 * Since the nfsnode does not have a lock, its 1625 * vnode lock has to be carried over. 1626 */ 1627 /* 1628 * XXX is the old node sure to be locked here? 1629 */ 1630 KASSERT(lockstatus(&vp->v_lock) == 1631 LK_EXCLUSIVE); 1632 nvp->v_data = vp->v_data; 1633 vp->v_data = NULL; 1634 VOP_UNLOCK(vp, 0); 1635 vp->v_op = spec_vnodeop_p; 1636 vrele(vp); 1637 vgone(vp); 1638 lockmgr(&nvp->v_lock, LK_EXCLUSIVE, 1639 &nvp->v_interlock); 1640 /* 1641 * Reinitialize aliased node. 1642 */ 1643 np->n_vnode = nvp; 1644 *vpp = vp = nvp; 1645 } 1646 } 1647 np->n_mtime = mtime; 1648 } 1649 uid = fxdr_unsigned(uid_t, fp->fa_uid); 1650 gid = fxdr_unsigned(gid_t, fp->fa_gid); 1651 vap = np->n_vattr; 1652 1653 /* 1654 * Invalidate access cache if uid, gid or mode changed. 1655 */ 1656 if (np->n_accstamp != -1 && 1657 (gid != vap->va_gid || uid != vap->va_uid || vmode != vap->va_mode)) 1658 np->n_accstamp = -1; 1659 1660 vap->va_type = vtyp; 1661 vap->va_mode = vmode; 1662 vap->va_rdev = (dev_t)rdev; 1663 vap->va_mtime = mtime; 1664 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 1665 switch (vtyp) { 1666 case VDIR: 1667 vap->va_blocksize = NFS_DIRFRAGSIZ; 1668 break; 1669 case VBLK: 1670 vap->va_blocksize = BLKDEV_IOSIZE; 1671 break; 1672 case VCHR: 1673 vap->va_blocksize = MAXBSIZE; 1674 break; 1675 default: 1676 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1677 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1678 break; 1679 } 1680 if (v3) { 1681 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1682 vap->va_uid = uid; 1683 vap->va_gid = gid; 1684 vap->va_size = fxdr_hyper(&fp->fa3_size); 1685 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1686 vap->va_fileid = fxdr_unsigned(int32_t, 1687 fp->fa3_fileid.nfsuquad[1]); 1688 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1689 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); 1690 vap->va_flags = 0; 1691 vap->va_filerev = 0; 1692 } else { 1693 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1694 vap->va_uid = uid; 1695 vap->va_gid = gid; 1696 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1697 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1698 * NFS_FABLKSIZE; 1699 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1700 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1701 vap->va_flags = 0; 1702 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, 1703 fp->fa2_ctime.nfsv2_sec); 1704 vap->va_ctime.tv_nsec = 0; 1705 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1706 vap->va_filerev = 0; 1707 } 1708 if (vap->va_size != np->n_size) { 1709 if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) { 1710 vap->va_size = np->n_size; 1711 } else { 1712 np->n_size = vap->va_size; 1713 if (vap->va_type == VREG) { 1714 if ((flags & NAC_NOTRUNC) 1715 && np->n_size < vp->v_size) { 1716 /* 1717 * we can't free pages now because 1718 * the pages can be owned by ourselves. 1719 */ 1720 np->n_flag |= NTRUNCDELAYED; 1721 } 1722 else { 1723 uvm_vnp_setsize(vp, np->n_size); 1724 } 1725 } 1726 } 1727 } 1728 np->n_attrstamp = time.tv_sec; 1729 if (vaper != NULL) { 1730 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1731 if (np->n_flag & NCHG) { 1732 if (np->n_flag & NACC) 1733 vaper->va_atime = np->n_atim; 1734 if (np->n_flag & NUPD) 1735 vaper->va_mtime = np->n_mtim; 1736 } 1737 } 1738 return (0); 1739 } 1740 1741 /* 1742 * Check the time stamp 1743 * If the cache is valid, copy contents to *vap and return 0 1744 * otherwise return an error 1745 */ 1746 int 1747 nfs_getattrcache(vp, vaper) 1748 struct vnode *vp; 1749 struct vattr *vaper; 1750 { 1751 struct nfsnode *np = VTONFS(vp); 1752 struct vattr *vap; 1753 1754 if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { 1755 nfsstats.attrcache_misses++; 1756 return (ENOENT); 1757 } 1758 nfsstats.attrcache_hits++; 1759 vap = np->n_vattr; 1760 if (vap->va_size != np->n_size) { 1761 if (vap->va_type == VREG) { 1762 if (np->n_flag & NMODIFIED) { 1763 if (vap->va_size < np->n_size) 1764 vap->va_size = np->n_size; 1765 else 1766 np->n_size = vap->va_size; 1767 } else 1768 np->n_size = vap->va_size; 1769 uvm_vnp_setsize(vp, np->n_size); 1770 } else 1771 np->n_size = vap->va_size; 1772 } 1773 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1774 if (np->n_flag & NCHG) { 1775 if (np->n_flag & NACC) 1776 vaper->va_atime = np->n_atim; 1777 if (np->n_flag & NUPD) 1778 vaper->va_mtime = np->n_mtim; 1779 } 1780 return (0); 1781 } 1782 1783 void 1784 nfs_delayedtruncate(vp) 1785 struct vnode *vp; 1786 { 1787 struct nfsnode *np = VTONFS(vp); 1788 1789 if (np->n_flag & NTRUNCDELAYED) { 1790 np->n_flag &= ~NTRUNCDELAYED; 1791 uvm_vnp_setsize(vp, np->n_size); 1792 } 1793 } 1794 1795 /* 1796 * Heuristic to see if the server XDR encodes directory cookies or not. 1797 * it is not supposed to, but a lot of servers may do this. Also, since 1798 * most/all servers will implement V2 as well, it is expected that they 1799 * may return just 32 bits worth of cookie information, so we need to 1800 * find out in which 32 bits this information is available. We do this 1801 * to avoid trouble with emulated binaries that can't handle 64 bit 1802 * directory offsets. 1803 */ 1804 1805 void 1806 nfs_cookieheuristic(vp, flagp, p, cred) 1807 struct vnode *vp; 1808 int *flagp; 1809 struct proc *p; 1810 struct ucred *cred; 1811 { 1812 struct uio auio; 1813 struct iovec aiov; 1814 caddr_t buf, cp; 1815 struct dirent *dp; 1816 off_t *cookies = NULL, *cop; 1817 int error, eof, nc, len; 1818 1819 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 1820 1821 aiov.iov_base = buf; 1822 aiov.iov_len = NFS_DIRFRAGSIZ; 1823 auio.uio_iov = &aiov; 1824 auio.uio_iovcnt = 1; 1825 auio.uio_rw = UIO_READ; 1826 auio.uio_segflg = UIO_SYSSPACE; 1827 auio.uio_procp = p; 1828 auio.uio_resid = NFS_DIRFRAGSIZ; 1829 auio.uio_offset = 0; 1830 1831 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 1832 1833 len = NFS_DIRFRAGSIZ - auio.uio_resid; 1834 if (error || len == 0) { 1835 FREE(buf, M_TEMP); 1836 if (cookies) 1837 free(cookies, M_TEMP); 1838 return; 1839 } 1840 1841 /* 1842 * Find the first valid entry and look at its offset cookie. 1843 */ 1844 1845 cp = buf; 1846 for (cop = cookies; len > 0; len -= dp->d_reclen) { 1847 dp = (struct dirent *)cp; 1848 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 1849 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 1850 *flagp |= NFSMNT_SWAPCOOKIE; 1851 nfs_invaldircache(vp, 0); 1852 nfs_vinvalbuf(vp, 0, cred, p, 1); 1853 } 1854 break; 1855 } 1856 cop++; 1857 cp += dp->d_reclen; 1858 } 1859 1860 FREE(buf, M_TEMP); 1861 free(cookies, M_TEMP); 1862 } 1863 #endif /* NFS */ 1864 1865 /* 1866 * Set up nameidata for a lookup() call and do it. 1867 * 1868 * If pubflag is set, this call is done for a lookup operation on the 1869 * public filehandle. In that case we allow crossing mountpoints and 1870 * absolute pathnames. However, the caller is expected to check that 1871 * the lookup result is within the public fs, and deny access if 1872 * it is not. 1873 */ 1874 int 1875 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) 1876 struct nameidata *ndp; 1877 fhandle_t *fhp; 1878 uint32_t len; 1879 struct nfssvc_sock *slp; 1880 struct mbuf *nam; 1881 struct mbuf **mdp; 1882 caddr_t *dposp; 1883 struct vnode **retdirp; 1884 struct proc *p; 1885 int kerbflag, pubflag; 1886 { 1887 int i, rem; 1888 struct mbuf *md; 1889 char *fromcp, *tocp, *cp; 1890 struct iovec aiov; 1891 struct uio auio; 1892 struct vnode *dp; 1893 int error, rdonly, linklen; 1894 struct componentname *cnp = &ndp->ni_cnd; 1895 1896 *retdirp = (struct vnode *)0; 1897 1898 if ((len + 1) > MAXPATHLEN) 1899 return (ENAMETOOLONG); 1900 cnp->cn_pnbuf = PNBUF_GET(); 1901 1902 /* 1903 * Copy the name from the mbuf list to ndp->ni_pnbuf 1904 * and set the various ndp fields appropriately. 1905 */ 1906 fromcp = *dposp; 1907 tocp = cnp->cn_pnbuf; 1908 md = *mdp; 1909 rem = mtod(md, caddr_t) + md->m_len - fromcp; 1910 for (i = 0; i < len; i++) { 1911 while (rem == 0) { 1912 md = md->m_next; 1913 if (md == NULL) { 1914 error = EBADRPC; 1915 goto out; 1916 } 1917 fromcp = mtod(md, caddr_t); 1918 rem = md->m_len; 1919 } 1920 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 1921 error = EACCES; 1922 goto out; 1923 } 1924 *tocp++ = *fromcp++; 1925 rem--; 1926 } 1927 *tocp = '\0'; 1928 *mdp = md; 1929 *dposp = fromcp; 1930 len = nfsm_rndup(len)-len; 1931 if (len > 0) { 1932 if (rem >= len) 1933 *dposp += len; 1934 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 1935 goto out; 1936 } 1937 1938 /* 1939 * Extract and set starting directory. 1940 */ 1941 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 1942 nam, &rdonly, kerbflag, pubflag); 1943 if (error) 1944 goto out; 1945 if (dp->v_type != VDIR) { 1946 vrele(dp); 1947 error = ENOTDIR; 1948 goto out; 1949 } 1950 1951 if (rdonly) 1952 cnp->cn_flags |= RDONLY; 1953 1954 *retdirp = dp; 1955 1956 if (pubflag) { 1957 /* 1958 * Oh joy. For WebNFS, handle those pesky '%' escapes, 1959 * and the 'native path' indicator. 1960 */ 1961 cp = PNBUF_GET(); 1962 fromcp = cnp->cn_pnbuf; 1963 tocp = cp; 1964 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 1965 switch ((unsigned char)*fromcp) { 1966 case WEBNFS_NATIVE_CHAR: 1967 /* 1968 * 'Native' path for us is the same 1969 * as a path according to the NFS spec, 1970 * just skip the escape char. 1971 */ 1972 fromcp++; 1973 break; 1974 /* 1975 * More may be added in the future, range 0x80-0xff 1976 */ 1977 default: 1978 error = EIO; 1979 PNBUF_PUT(cp); 1980 goto out; 1981 } 1982 } 1983 /* 1984 * Translate the '%' escapes, URL-style. 1985 */ 1986 while (*fromcp != '\0') { 1987 if (*fromcp == WEBNFS_ESC_CHAR) { 1988 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 1989 fromcp++; 1990 *tocp++ = HEXSTRTOI(fromcp); 1991 fromcp += 2; 1992 continue; 1993 } else { 1994 error = ENOENT; 1995 PNBUF_PUT(cp); 1996 goto out; 1997 } 1998 } else 1999 *tocp++ = *fromcp++; 2000 } 2001 *tocp = '\0'; 2002 PNBUF_PUT(cnp->cn_pnbuf); 2003 cnp->cn_pnbuf = cp; 2004 } 2005 2006 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 2007 ndp->ni_segflg = UIO_SYSSPACE; 2008 ndp->ni_rootdir = rootvnode; 2009 2010 if (pubflag) { 2011 ndp->ni_loopcnt = 0; 2012 if (cnp->cn_pnbuf[0] == '/') 2013 dp = rootvnode; 2014 } else { 2015 cnp->cn_flags |= NOCROSSMOUNT; 2016 } 2017 2018 cnp->cn_proc = p; 2019 VREF(dp); 2020 2021 for (;;) { 2022 cnp->cn_nameptr = cnp->cn_pnbuf; 2023 ndp->ni_startdir = dp; 2024 /* 2025 * And call lookup() to do the real work 2026 */ 2027 error = lookup(ndp); 2028 if (error) { 2029 PNBUF_PUT(cnp->cn_pnbuf); 2030 return (error); 2031 } 2032 /* 2033 * Check for encountering a symbolic link 2034 */ 2035 if ((cnp->cn_flags & ISSYMLINK) == 0) { 2036 if (cnp->cn_flags & (SAVENAME | SAVESTART)) 2037 cnp->cn_flags |= HASBUF; 2038 else 2039 PNBUF_PUT(cnp->cn_pnbuf); 2040 return (0); 2041 } else { 2042 if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN)) 2043 VOP_UNLOCK(ndp->ni_dvp, 0); 2044 if (!pubflag) { 2045 error = EINVAL; 2046 break; 2047 } 2048 2049 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 2050 error = ELOOP; 2051 break; 2052 } 2053 if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) { 2054 error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred, 2055 cnp->cn_proc); 2056 if (error != 0) 2057 break; 2058 } 2059 if (ndp->ni_pathlen > 1) 2060 cp = PNBUF_GET(); 2061 else 2062 cp = cnp->cn_pnbuf; 2063 aiov.iov_base = cp; 2064 aiov.iov_len = MAXPATHLEN; 2065 auio.uio_iov = &aiov; 2066 auio.uio_iovcnt = 1; 2067 auio.uio_offset = 0; 2068 auio.uio_rw = UIO_READ; 2069 auio.uio_segflg = UIO_SYSSPACE; 2070 auio.uio_procp = (struct proc *)0; 2071 auio.uio_resid = MAXPATHLEN; 2072 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 2073 if (error) { 2074 badlink: 2075 if (ndp->ni_pathlen > 1) 2076 PNBUF_PUT(cp); 2077 break; 2078 } 2079 linklen = MAXPATHLEN - auio.uio_resid; 2080 if (linklen == 0) { 2081 error = ENOENT; 2082 goto badlink; 2083 } 2084 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 2085 error = ENAMETOOLONG; 2086 goto badlink; 2087 } 2088 if (ndp->ni_pathlen > 1) { 2089 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 2090 PNBUF_PUT(cnp->cn_pnbuf); 2091 cnp->cn_pnbuf = cp; 2092 } else 2093 cnp->cn_pnbuf[linklen] = '\0'; 2094 ndp->ni_pathlen += linklen; 2095 vput(ndp->ni_vp); 2096 dp = ndp->ni_dvp; 2097 /* 2098 * Check if root directory should replace current directory. 2099 */ 2100 if (cnp->cn_pnbuf[0] == '/') { 2101 vrele(dp); 2102 dp = ndp->ni_rootdir; 2103 VREF(dp); 2104 } 2105 } 2106 } 2107 vrele(ndp->ni_dvp); 2108 vput(ndp->ni_vp); 2109 ndp->ni_vp = NULL; 2110 out: 2111 PNBUF_PUT(cnp->cn_pnbuf); 2112 return (error); 2113 } 2114 2115 /* 2116 * A fiddled version of m_adj() that ensures null fill to a 32-bit 2117 * boundary and only trims off the back end 2118 * 2119 * 1. trim off 'len' bytes as m_adj(mp, -len). 2120 * 2. add zero-padding 'nul' bytes at the end of the mbuf chain. 2121 */ 2122 void 2123 nfs_zeropad(mp, len, nul) 2124 struct mbuf *mp; 2125 int len; 2126 int nul; 2127 { 2128 struct mbuf *m; 2129 int count, i; 2130 char *cp; 2131 2132 /* 2133 * Trim from tail. Scan the mbuf chain, 2134 * calculating its length and finding the last mbuf. 2135 * If the adjustment only affects this mbuf, then just 2136 * adjust and return. Otherwise, rescan and truncate 2137 * after the remaining size. 2138 */ 2139 count = 0; 2140 m = mp; 2141 for (;;) { 2142 count += m->m_len; 2143 if (m->m_next == NULL) 2144 break; 2145 m = m->m_next; 2146 } 2147 2148 KDASSERT(count >= len); 2149 2150 if (m->m_len >= len) { 2151 m->m_len -= len; 2152 } else { 2153 count -= len; 2154 /* 2155 * Correct length for chain is "count". 2156 * Find the mbuf with last data, adjust its length, 2157 * and toss data from remaining mbufs on chain. 2158 */ 2159 for (m = mp; m; m = m->m_next) { 2160 if (m->m_len >= count) { 2161 m->m_len = count; 2162 break; 2163 } 2164 count -= m->m_len; 2165 } 2166 m_freem(m->m_next); 2167 m->m_next = NULL; 2168 } 2169 2170 /* 2171 * zero-padding. 2172 */ 2173 if (nul > 0) { 2174 if (M_ROMAP(m) || M_TRAILINGSPACE(m) < nul) { 2175 struct mbuf *n; 2176 2177 KDASSERT(MLEN >= nul); 2178 n = m_get(M_WAIT, MT_DATA); 2179 MCLAIM(n, &nfs_mowner); 2180 n->m_len = nul; 2181 n->m_next = m->m_next; 2182 m->m_next = n; 2183 m = n; 2184 cp = mtod(n, caddr_t); 2185 } else { 2186 cp = mtod(m, caddr_t) + m->m_len; 2187 m->m_len += nul; 2188 } 2189 for (i = 0; i < nul; i++) 2190 *cp++ = '\0'; 2191 } 2192 return; 2193 } 2194 2195 /* 2196 * Make these functions instead of macros, so that the kernel text size 2197 * doesn't get too big... 2198 */ 2199 void 2200 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2201 struct nfsrv_descript *nfsd; 2202 int before_ret; 2203 struct vattr *before_vap; 2204 int after_ret; 2205 struct vattr *after_vap; 2206 struct mbuf **mbp; 2207 char **bposp; 2208 { 2209 struct mbuf *mb = *mbp; 2210 char *bpos = *bposp; 2211 u_int32_t *tl; 2212 2213 if (before_ret) { 2214 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2215 *tl = nfs_false; 2216 } else { 2217 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2218 *tl++ = nfs_true; 2219 txdr_hyper(before_vap->va_size, tl); 2220 tl += 2; 2221 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2222 tl += 2; 2223 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2224 } 2225 *bposp = bpos; 2226 *mbp = mb; 2227 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2228 } 2229 2230 void 2231 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2232 struct nfsrv_descript *nfsd; 2233 int after_ret; 2234 struct vattr *after_vap; 2235 struct mbuf **mbp; 2236 char **bposp; 2237 { 2238 struct mbuf *mb = *mbp; 2239 char *bpos = *bposp; 2240 u_int32_t *tl; 2241 struct nfs_fattr *fp; 2242 2243 if (after_ret) { 2244 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2245 *tl = nfs_false; 2246 } else { 2247 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2248 *tl++ = nfs_true; 2249 fp = (struct nfs_fattr *)tl; 2250 nfsm_srvfattr(nfsd, after_vap, fp); 2251 } 2252 *mbp = mb; 2253 *bposp = bpos; 2254 } 2255 2256 void 2257 nfsm_srvfattr(nfsd, vap, fp) 2258 struct nfsrv_descript *nfsd; 2259 struct vattr *vap; 2260 struct nfs_fattr *fp; 2261 { 2262 2263 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2264 fp->fa_uid = txdr_unsigned(vap->va_uid); 2265 fp->fa_gid = txdr_unsigned(vap->va_gid); 2266 if (nfsd->nd_flag & ND_NFSV3) { 2267 fp->fa_type = vtonfsv3_type(vap->va_type); 2268 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2269 txdr_hyper(vap->va_size, &fp->fa3_size); 2270 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2271 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2272 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2273 fp->fa3_fsid.nfsuquad[0] = 0; 2274 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2275 fp->fa3_fileid.nfsuquad[0] = 0; 2276 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); 2277 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2278 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2279 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2280 } else { 2281 fp->fa_type = vtonfsv2_type(vap->va_type); 2282 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2283 fp->fa2_size = txdr_unsigned(vap->va_size); 2284 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2285 if (vap->va_type == VFIFO) 2286 fp->fa2_rdev = 0xffffffff; 2287 else 2288 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2289 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2290 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2291 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2292 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2293 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2294 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2295 } 2296 } 2297 2298 /* 2299 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2300 * - look up fsid in mount list (if not found ret error) 2301 * - get vp and export rights by calling VFS_FHTOVP() 2302 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2303 * - if not lockflag unlock it with VOP_UNLOCK() 2304 */ 2305 int 2306 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) 2307 fhandle_t *fhp; 2308 int lockflag; 2309 struct vnode **vpp; 2310 struct ucred *cred; 2311 struct nfssvc_sock *slp; 2312 struct mbuf *nam; 2313 int *rdonlyp; 2314 int kerbflag; 2315 { 2316 struct mount *mp; 2317 int i; 2318 struct ucred *credanon; 2319 int error, exflags; 2320 struct sockaddr_in *saddr; 2321 2322 *vpp = (struct vnode *)0; 2323 2324 if (nfs_ispublicfh(fhp)) { 2325 if (!pubflag || !nfs_pub.np_valid) 2326 return (ESTALE); 2327 fhp = &nfs_pub.np_handle; 2328 } 2329 2330 mp = vfs_getvfs(&fhp->fh_fsid); 2331 if (!mp) 2332 return (ESTALE); 2333 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); 2334 if (error) 2335 return (error); 2336 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2337 if (error) 2338 return (error); 2339 2340 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2341 saddr = mtod(nam, struct sockaddr_in *); 2342 if ((saddr->sin_family == AF_INET) && 2343 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2344 vput(*vpp); 2345 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2346 } 2347 #ifdef INET6 2348 if ((saddr->sin_family == AF_INET6) && 2349 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2350 vput(*vpp); 2351 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2352 } 2353 #endif 2354 } 2355 /* 2356 * Check/setup credentials. 2357 */ 2358 if (exflags & MNT_EXKERB) { 2359 if (!kerbflag) { 2360 vput(*vpp); 2361 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2362 } 2363 } else if (kerbflag) { 2364 vput(*vpp); 2365 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2366 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { 2367 cred->cr_uid = credanon->cr_uid; 2368 cred->cr_gid = credanon->cr_gid; 2369 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) 2370 cred->cr_groups[i] = credanon->cr_groups[i]; 2371 cred->cr_ngroups = i; 2372 } 2373 if (exflags & MNT_EXRDONLY) 2374 *rdonlyp = 1; 2375 else 2376 *rdonlyp = 0; 2377 if (!lockflag) 2378 VOP_UNLOCK(*vpp, 0); 2379 return (0); 2380 } 2381 2382 /* 2383 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2384 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has 2385 * transformed this to all zeroes in both cases, so check for it. 2386 */ 2387 int 2388 nfs_ispublicfh(fhp) 2389 fhandle_t *fhp; 2390 { 2391 char *cp = (char *)fhp; 2392 int i; 2393 2394 for (i = 0; i < NFSX_V3FH; i++) 2395 if (*cp++ != 0) 2396 return (FALSE); 2397 return (TRUE); 2398 } 2399 2400 /* 2401 * This function compares two net addresses by family and returns TRUE 2402 * if they are the same host. 2403 * If there is any doubt, return FALSE. 2404 * The AF_INET family is handled as a special case so that address mbufs 2405 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2406 */ 2407 int 2408 netaddr_match(family, haddr, nam) 2409 int family; 2410 union nethostaddr *haddr; 2411 struct mbuf *nam; 2412 { 2413 struct sockaddr_in *inetaddr; 2414 2415 switch (family) { 2416 case AF_INET: 2417 inetaddr = mtod(nam, struct sockaddr_in *); 2418 if (inetaddr->sin_family == AF_INET && 2419 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2420 return (1); 2421 break; 2422 #ifdef INET6 2423 case AF_INET6: 2424 { 2425 struct sockaddr_in6 *sin6_1, *sin6_2; 2426 2427 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2428 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2429 if (sin6_1->sin6_family == AF_INET6 && 2430 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2431 return 1; 2432 } 2433 #endif 2434 #ifdef ISO 2435 case AF_ISO: 2436 { 2437 struct sockaddr_iso *isoaddr1, *isoaddr2; 2438 2439 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2440 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2441 if (isoaddr1->siso_family == AF_ISO && 2442 isoaddr1->siso_nlen > 0 && 2443 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2444 SAME_ISOADDR(isoaddr1, isoaddr2)) 2445 return (1); 2446 break; 2447 } 2448 #endif /* ISO */ 2449 default: 2450 break; 2451 }; 2452 return (0); 2453 } 2454 2455 /* 2456 * The write verifier has changed (probably due to a server reboot), so all 2457 * PG_NEEDCOMMIT pages will have to be written again. Since they are marked 2458 * as dirty or are being written out just now, all this takes is clearing 2459 * the PG_NEEDCOMMIT flag. Once done the new write verifier can be set for 2460 * the mount point. 2461 */ 2462 void 2463 nfs_clearcommit(mp) 2464 struct mount *mp; 2465 { 2466 struct vnode *vp; 2467 struct nfsnode *np; 2468 struct vm_page *pg; 2469 struct nfsmount *nmp = VFSTONFS(mp); 2470 2471 lockmgr(&nmp->nm_writeverflock, LK_EXCLUSIVE, NULL); 2472 2473 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2474 KASSERT(vp->v_mount == mp); 2475 if (vp->v_type == VNON) 2476 continue; 2477 np = VTONFS(vp); 2478 np->n_pushlo = np->n_pushhi = np->n_pushedlo = 2479 np->n_pushedhi = 0; 2480 np->n_commitflags &= 2481 ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID); 2482 simple_lock(&vp->v_uobj.vmobjlock); 2483 TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) { 2484 pg->flags &= ~PG_NEEDCOMMIT; 2485 } 2486 simple_unlock(&vp->v_uobj.vmobjlock); 2487 } 2488 simple_lock(&nmp->nm_slock); 2489 nmp->nm_iflag &= ~NFSMNT_STALEWRITEVERF; 2490 simple_unlock(&nmp->nm_slock); 2491 lockmgr(&nmp->nm_writeverflock, LK_RELEASE, NULL); 2492 } 2493 2494 void 2495 nfs_merge_commit_ranges(vp) 2496 struct vnode *vp; 2497 { 2498 struct nfsnode *np = VTONFS(vp); 2499 2500 KASSERT(np->n_commitflags & NFS_COMMIT_PUSH_VALID); 2501 2502 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2503 np->n_pushedlo = np->n_pushlo; 2504 np->n_pushedhi = np->n_pushhi; 2505 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2506 } else { 2507 if (np->n_pushlo < np->n_pushedlo) 2508 np->n_pushedlo = np->n_pushlo; 2509 if (np->n_pushhi > np->n_pushedhi) 2510 np->n_pushedhi = np->n_pushhi; 2511 } 2512 2513 np->n_pushlo = np->n_pushhi = 0; 2514 np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID; 2515 2516 #ifdef NFS_DEBUG_COMMIT 2517 printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2518 (unsigned)np->n_pushedhi); 2519 #endif 2520 } 2521 2522 int 2523 nfs_in_committed_range(vp, off, len) 2524 struct vnode *vp; 2525 off_t off, len; 2526 { 2527 struct nfsnode *np = VTONFS(vp); 2528 off_t lo, hi; 2529 2530 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2531 return 0; 2532 lo = off; 2533 hi = lo + len; 2534 2535 return (lo >= np->n_pushedlo && hi <= np->n_pushedhi); 2536 } 2537 2538 int 2539 nfs_in_tobecommitted_range(vp, off, len) 2540 struct vnode *vp; 2541 off_t off, len; 2542 { 2543 struct nfsnode *np = VTONFS(vp); 2544 off_t lo, hi; 2545 2546 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2547 return 0; 2548 lo = off; 2549 hi = lo + len; 2550 2551 return (lo >= np->n_pushlo && hi <= np->n_pushhi); 2552 } 2553 2554 void 2555 nfs_add_committed_range(vp, off, len) 2556 struct vnode *vp; 2557 off_t off, len; 2558 { 2559 struct nfsnode *np = VTONFS(vp); 2560 off_t lo, hi; 2561 2562 lo = off; 2563 hi = lo + len; 2564 2565 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2566 np->n_pushedlo = lo; 2567 np->n_pushedhi = hi; 2568 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2569 } else { 2570 if (hi > np->n_pushedhi) 2571 np->n_pushedhi = hi; 2572 if (lo < np->n_pushedlo) 2573 np->n_pushedlo = lo; 2574 } 2575 #ifdef NFS_DEBUG_COMMIT 2576 printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2577 (unsigned)np->n_pushedhi); 2578 #endif 2579 } 2580 2581 void 2582 nfs_del_committed_range(vp, off, len) 2583 struct vnode *vp; 2584 off_t off, len; 2585 { 2586 struct nfsnode *np = VTONFS(vp); 2587 off_t lo, hi; 2588 2589 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2590 return; 2591 2592 lo = off; 2593 hi = lo + len; 2594 2595 if (lo > np->n_pushedhi || hi < np->n_pushedlo) 2596 return; 2597 if (lo <= np->n_pushedlo) 2598 np->n_pushedlo = hi; 2599 else if (hi >= np->n_pushedhi) 2600 np->n_pushedhi = lo; 2601 else { 2602 /* 2603 * XXX There's only one range. If the deleted range 2604 * is in the middle, pick the largest of the 2605 * contiguous ranges that it leaves. 2606 */ 2607 if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi)) 2608 np->n_pushedhi = lo; 2609 else 2610 np->n_pushedlo = hi; 2611 } 2612 #ifdef NFS_DEBUG_COMMIT 2613 printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2614 (unsigned)np->n_pushedhi); 2615 #endif 2616 } 2617 2618 void 2619 nfs_add_tobecommitted_range(vp, off, len) 2620 struct vnode *vp; 2621 off_t off, len; 2622 { 2623 struct nfsnode *np = VTONFS(vp); 2624 off_t lo, hi; 2625 2626 lo = off; 2627 hi = lo + len; 2628 2629 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) { 2630 np->n_pushlo = lo; 2631 np->n_pushhi = hi; 2632 np->n_commitflags |= NFS_COMMIT_PUSH_VALID; 2633 } else { 2634 if (lo < np->n_pushlo) 2635 np->n_pushlo = lo; 2636 if (hi > np->n_pushhi) 2637 np->n_pushhi = hi; 2638 } 2639 #ifdef NFS_DEBUG_COMMIT 2640 printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2641 (unsigned)np->n_pushhi); 2642 #endif 2643 } 2644 2645 void 2646 nfs_del_tobecommitted_range(vp, off, len) 2647 struct vnode *vp; 2648 off_t off, len; 2649 { 2650 struct nfsnode *np = VTONFS(vp); 2651 off_t lo, hi; 2652 2653 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2654 return; 2655 2656 lo = off; 2657 hi = lo + len; 2658 2659 if (lo > np->n_pushhi || hi < np->n_pushlo) 2660 return; 2661 2662 if (lo <= np->n_pushlo) 2663 np->n_pushlo = hi; 2664 else if (hi >= np->n_pushhi) 2665 np->n_pushhi = lo; 2666 else { 2667 /* 2668 * XXX There's only one range. If the deleted range 2669 * is in the middle, pick the largest of the 2670 * contiguous ranges that it leaves. 2671 */ 2672 if ((np->n_pushlo - lo) > (hi - np->n_pushhi)) 2673 np->n_pushhi = lo; 2674 else 2675 np->n_pushlo = hi; 2676 } 2677 #ifdef NFS_DEBUG_COMMIT 2678 printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2679 (unsigned)np->n_pushhi); 2680 #endif 2681 } 2682 2683 /* 2684 * Map errnos to NFS error numbers. For Version 3 also filter out error 2685 * numbers not specified for the associated procedure. 2686 */ 2687 int 2688 nfsrv_errmap(nd, err) 2689 struct nfsrv_descript *nd; 2690 int err; 2691 { 2692 const short *defaulterrp, *errp; 2693 2694 if (nd->nd_flag & ND_NFSV3) { 2695 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2696 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2697 while (*++errp) { 2698 if (*errp == err) 2699 return (err); 2700 else if (*errp > err) 2701 break; 2702 } 2703 return ((int)*defaulterrp); 2704 } else 2705 return (err & 0xffff); 2706 } 2707 if (err <= ELAST) 2708 return ((int)nfsrv_v2errmap[err - 1]); 2709 return (NFSERR_IO); 2710 } 2711 2712 /* 2713 * Sort the group list in increasing numerical order. 2714 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort 2715 * that used to be here.) 2716 */ 2717 void 2718 nfsrvw_sort(list, num) 2719 gid_t *list; 2720 int num; 2721 { 2722 int i, j; 2723 gid_t v; 2724 2725 /* Insertion sort. */ 2726 for (i = 1; i < num; i++) { 2727 v = list[i]; 2728 /* find correct slot for value v, moving others up */ 2729 for (j = i; --j >= 0 && v < list[j];) 2730 list[j + 1] = list[j]; 2731 list[j + 1] = v; 2732 } 2733 } 2734 2735 /* 2736 * copy credentials making sure that the result can be compared with memcmp(). 2737 */ 2738 void 2739 nfsrv_setcred(incred, outcred) 2740 struct ucred *incred, *outcred; 2741 { 2742 int i; 2743 2744 memset((caddr_t)outcred, 0, sizeof (struct ucred)); 2745 outcred->cr_ref = 1; 2746 outcred->cr_uid = incred->cr_uid; 2747 outcred->cr_gid = incred->cr_gid; 2748 outcred->cr_ngroups = incred->cr_ngroups; 2749 for (i = 0; i < incred->cr_ngroups; i++) 2750 outcred->cr_groups[i] = incred->cr_groups[i]; 2751 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); 2752 } 2753 2754 u_int32_t 2755 nfs_getxid() 2756 { 2757 static u_int32_t base; 2758 static u_int32_t nfs_xid = 0; 2759 static struct simplelock nfs_xidlock = SIMPLELOCK_INITIALIZER; 2760 u_int32_t newxid; 2761 2762 simple_lock(&nfs_xidlock); 2763 /* 2764 * derive initial xid from system time 2765 * XXX time is invalid if root not yet mounted 2766 */ 2767 if (__predict_false(!base && (rootvp))) { 2768 struct timeval tv; 2769 2770 microtime(&tv); 2771 base = tv.tv_sec << 12; 2772 nfs_xid = base; 2773 } 2774 2775 /* 2776 * Skip zero xid if it should ever happen. 2777 */ 2778 if (__predict_false(++nfs_xid == 0)) 2779 nfs_xid++; 2780 newxid = nfs_xid; 2781 simple_unlock(&nfs_xidlock); 2782 2783 return txdr_unsigned(newxid); 2784 } 2785 2786 /* 2787 * assign a new xid for existing request. 2788 * used for NFSERR_JUKEBOX handling. 2789 */ 2790 void 2791 nfs_renewxid(struct nfsreq *req) 2792 { 2793 u_int32_t xid; 2794 int off; 2795 2796 xid = nfs_getxid(); 2797 if (req->r_nmp->nm_sotype == SOCK_STREAM) 2798 off = sizeof(u_int32_t); /* RPC record mark */ 2799 else 2800 off = 0; 2801 2802 m_copyback(req->r_mreq, off, sizeof(xid), (void *)&xid); 2803 req->r_xid = xid; 2804 } 2805
Indexes created Thu Sep 25 16:09:42 GMT 2025