1 /* $NetBSD: nfs_subs.c,v 1.134 2004/06/14 12:28:35 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 1.134 2004/06/14 12:28:35 yamt Exp $"); 74 75 #include "fs_nfs.h" 76 #include "opt_nfs.h" 77 #include "opt_nfsserver.h" 78 #include "opt_iso.h" 79 #include "opt_inet.h" 80 81 /* 82 * These functions support the macros and help fiddle mbuf chains for 83 * the nfs op functions. They do things like create the rpc header and 84 * copy data between mbuf chains and uio lists. 85 */ 86 #include <sys/param.h> 87 #include <sys/proc.h> 88 #include <sys/systm.h> 89 #include <sys/kernel.h> 90 #include <sys/mount.h> 91 #include <sys/vnode.h> 92 #include <sys/namei.h> 93 #include <sys/mbuf.h> 94 #include <sys/socket.h> 95 #include <sys/stat.h> 96 #include <sys/malloc.h> 97 #include <sys/filedesc.h> 98 #include <sys/time.h> 99 #include <sys/dirent.h> 100 101 #include <uvm/uvm_extern.h> 102 103 #include <nfs/rpcv2.h> 104 #include <nfs/nfsproto.h> 105 #include <nfs/nfsnode.h> 106 #include <nfs/nfs.h> 107 #include <nfs/xdr_subs.h> 108 #include <nfs/nfsm_subs.h> 109 #include <nfs/nfsmount.h> 110 #include <nfs/nqnfs.h> 111 #include <nfs/nfsrtt.h> 112 #include <nfs/nfs_var.h> 113 114 #include <miscfs/specfs/specdev.h> 115 116 #include <netinet/in.h> 117 #ifdef ISO 118 #include <netiso/iso.h> 119 #endif 120 121 /* 122 * Data items converted to xdr at startup, since they are constant 123 * This is kinda hokey, but may save a little time doing byte swaps 124 */ 125 u_int32_t nfs_xdrneg1; 126 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 127 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 128 rpc_auth_kerb; 129 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; 130 131 /* And other global data */ 132 const nfstype nfsv2_type[9] = 133 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, NFCHR, NFNON }; 134 const nfstype nfsv3_type[9] = 135 { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, NFFIFO, NFNON }; 136 const enum vtype nv2tov_type[8] = 137 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 138 const enum vtype nv3tov_type[8] = 139 { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 140 int nfs_ticks; 141 int nfs_commitsize; 142 143 MALLOC_DEFINE(M_NFSDIROFF, "NFS diroff", "NFS directory cookies"); 144 145 /* NFS client/server stats. */ 146 struct nfsstats nfsstats; 147 148 /* 149 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 150 */ 151 const int nfsv3_procid[NFS_NPROCS] = { 152 NFSPROC_NULL, 153 NFSPROC_GETATTR, 154 NFSPROC_SETATTR, 155 NFSPROC_NOOP, 156 NFSPROC_LOOKUP, 157 NFSPROC_READLINK, 158 NFSPROC_READ, 159 NFSPROC_NOOP, 160 NFSPROC_WRITE, 161 NFSPROC_CREATE, 162 NFSPROC_REMOVE, 163 NFSPROC_RENAME, 164 NFSPROC_LINK, 165 NFSPROC_SYMLINK, 166 NFSPROC_MKDIR, 167 NFSPROC_RMDIR, 168 NFSPROC_READDIR, 169 NFSPROC_FSSTAT, 170 NFSPROC_NOOP, 171 NFSPROC_NOOP, 172 NFSPROC_NOOP, 173 NFSPROC_NOOP, 174 NFSPROC_NOOP, 175 NFSPROC_NOOP, 176 NFSPROC_NOOP, 177 NFSPROC_NOOP 178 }; 179 180 /* 181 * and the reverse mapping from generic to Version 2 procedure numbers 182 */ 183 const int nfsv2_procid[NFS_NPROCS] = { 184 NFSV2PROC_NULL, 185 NFSV2PROC_GETATTR, 186 NFSV2PROC_SETATTR, 187 NFSV2PROC_LOOKUP, 188 NFSV2PROC_NOOP, 189 NFSV2PROC_READLINK, 190 NFSV2PROC_READ, 191 NFSV2PROC_WRITE, 192 NFSV2PROC_CREATE, 193 NFSV2PROC_MKDIR, 194 NFSV2PROC_SYMLINK, 195 NFSV2PROC_CREATE, 196 NFSV2PROC_REMOVE, 197 NFSV2PROC_RMDIR, 198 NFSV2PROC_RENAME, 199 NFSV2PROC_LINK, 200 NFSV2PROC_READDIR, 201 NFSV2PROC_NOOP, 202 NFSV2PROC_STATFS, 203 NFSV2PROC_NOOP, 204 NFSV2PROC_NOOP, 205 NFSV2PROC_NOOP, 206 NFSV2PROC_NOOP, 207 NFSV2PROC_NOOP, 208 NFSV2PROC_NOOP, 209 NFSV2PROC_NOOP, 210 }; 211 212 /* 213 * Maps errno values to nfs error numbers. 214 * Use NFSERR_IO as the catch all for ones not specifically defined in 215 * RFC 1094. 216 */ 217 static const u_char nfsrv_v2errmap[ELAST] = { 218 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 219 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 220 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 221 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 222 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 223 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 224 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 225 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 226 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 227 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 228 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 229 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 230 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, 231 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 232 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 233 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 234 NFSERR_IO, NFSERR_IO, 235 }; 236 237 /* 238 * Maps errno values to nfs error numbers. 239 * Although it is not obvious whether or not NFS clients really care if 240 * a returned error value is in the specified list for the procedure, the 241 * safest thing to do is filter them appropriately. For Version 2, the 242 * X/Open XNFS document is the only specification that defines error values 243 * for each RPC (The RFC simply lists all possible error values for all RPCs), 244 * so I have decided to not do this for Version 2. 245 * The first entry is the default error return and the rest are the valid 246 * errors for that RPC in increasing numeric order. 247 */ 248 static const short nfsv3err_null[] = { 249 0, 250 0, 251 }; 252 253 static const short nfsv3err_getattr[] = { 254 NFSERR_IO, 255 NFSERR_IO, 256 NFSERR_STALE, 257 NFSERR_BADHANDLE, 258 NFSERR_SERVERFAULT, 259 0, 260 }; 261 262 static const short nfsv3err_setattr[] = { 263 NFSERR_IO, 264 NFSERR_PERM, 265 NFSERR_IO, 266 NFSERR_ACCES, 267 NFSERR_INVAL, 268 NFSERR_NOSPC, 269 NFSERR_ROFS, 270 NFSERR_DQUOT, 271 NFSERR_STALE, 272 NFSERR_BADHANDLE, 273 NFSERR_NOT_SYNC, 274 NFSERR_SERVERFAULT, 275 0, 276 }; 277 278 static const short nfsv3err_lookup[] = { 279 NFSERR_IO, 280 NFSERR_NOENT, 281 NFSERR_IO, 282 NFSERR_ACCES, 283 NFSERR_NOTDIR, 284 NFSERR_NAMETOL, 285 NFSERR_STALE, 286 NFSERR_BADHANDLE, 287 NFSERR_SERVERFAULT, 288 0, 289 }; 290 291 static const short nfsv3err_access[] = { 292 NFSERR_IO, 293 NFSERR_IO, 294 NFSERR_STALE, 295 NFSERR_BADHANDLE, 296 NFSERR_SERVERFAULT, 297 0, 298 }; 299 300 static const short nfsv3err_readlink[] = { 301 NFSERR_IO, 302 NFSERR_IO, 303 NFSERR_ACCES, 304 NFSERR_INVAL, 305 NFSERR_STALE, 306 NFSERR_BADHANDLE, 307 NFSERR_NOTSUPP, 308 NFSERR_SERVERFAULT, 309 0, 310 }; 311 312 static const short nfsv3err_read[] = { 313 NFSERR_IO, 314 NFSERR_IO, 315 NFSERR_NXIO, 316 NFSERR_ACCES, 317 NFSERR_INVAL, 318 NFSERR_STALE, 319 NFSERR_BADHANDLE, 320 NFSERR_SERVERFAULT, 321 NFSERR_JUKEBOX, 322 0, 323 }; 324 325 static const short nfsv3err_write[] = { 326 NFSERR_IO, 327 NFSERR_IO, 328 NFSERR_ACCES, 329 NFSERR_INVAL, 330 NFSERR_FBIG, 331 NFSERR_NOSPC, 332 NFSERR_ROFS, 333 NFSERR_DQUOT, 334 NFSERR_STALE, 335 NFSERR_BADHANDLE, 336 NFSERR_SERVERFAULT, 337 NFSERR_JUKEBOX, 338 0, 339 }; 340 341 static const short nfsv3err_create[] = { 342 NFSERR_IO, 343 NFSERR_IO, 344 NFSERR_ACCES, 345 NFSERR_EXIST, 346 NFSERR_NOTDIR, 347 NFSERR_NOSPC, 348 NFSERR_ROFS, 349 NFSERR_NAMETOL, 350 NFSERR_DQUOT, 351 NFSERR_STALE, 352 NFSERR_BADHANDLE, 353 NFSERR_NOTSUPP, 354 NFSERR_SERVERFAULT, 355 0, 356 }; 357 358 static const short nfsv3err_mkdir[] = { 359 NFSERR_IO, 360 NFSERR_IO, 361 NFSERR_ACCES, 362 NFSERR_EXIST, 363 NFSERR_NOTDIR, 364 NFSERR_NOSPC, 365 NFSERR_ROFS, 366 NFSERR_NAMETOL, 367 NFSERR_DQUOT, 368 NFSERR_STALE, 369 NFSERR_BADHANDLE, 370 NFSERR_NOTSUPP, 371 NFSERR_SERVERFAULT, 372 0, 373 }; 374 375 static const short nfsv3err_symlink[] = { 376 NFSERR_IO, 377 NFSERR_IO, 378 NFSERR_ACCES, 379 NFSERR_EXIST, 380 NFSERR_NOTDIR, 381 NFSERR_NOSPC, 382 NFSERR_ROFS, 383 NFSERR_NAMETOL, 384 NFSERR_DQUOT, 385 NFSERR_STALE, 386 NFSERR_BADHANDLE, 387 NFSERR_NOTSUPP, 388 NFSERR_SERVERFAULT, 389 0, 390 }; 391 392 static const short nfsv3err_mknod[] = { 393 NFSERR_IO, 394 NFSERR_IO, 395 NFSERR_ACCES, 396 NFSERR_EXIST, 397 NFSERR_NOTDIR, 398 NFSERR_NOSPC, 399 NFSERR_ROFS, 400 NFSERR_NAMETOL, 401 NFSERR_DQUOT, 402 NFSERR_STALE, 403 NFSERR_BADHANDLE, 404 NFSERR_NOTSUPP, 405 NFSERR_SERVERFAULT, 406 NFSERR_BADTYPE, 407 0, 408 }; 409 410 static const short nfsv3err_remove[] = { 411 NFSERR_IO, 412 NFSERR_NOENT, 413 NFSERR_IO, 414 NFSERR_ACCES, 415 NFSERR_NOTDIR, 416 NFSERR_ROFS, 417 NFSERR_NAMETOL, 418 NFSERR_STALE, 419 NFSERR_BADHANDLE, 420 NFSERR_SERVERFAULT, 421 0, 422 }; 423 424 static const short nfsv3err_rmdir[] = { 425 NFSERR_IO, 426 NFSERR_NOENT, 427 NFSERR_IO, 428 NFSERR_ACCES, 429 NFSERR_EXIST, 430 NFSERR_NOTDIR, 431 NFSERR_INVAL, 432 NFSERR_ROFS, 433 NFSERR_NAMETOL, 434 NFSERR_NOTEMPTY, 435 NFSERR_STALE, 436 NFSERR_BADHANDLE, 437 NFSERR_NOTSUPP, 438 NFSERR_SERVERFAULT, 439 0, 440 }; 441 442 static const short nfsv3err_rename[] = { 443 NFSERR_IO, 444 NFSERR_NOENT, 445 NFSERR_IO, 446 NFSERR_ACCES, 447 NFSERR_EXIST, 448 NFSERR_XDEV, 449 NFSERR_NOTDIR, 450 NFSERR_ISDIR, 451 NFSERR_INVAL, 452 NFSERR_NOSPC, 453 NFSERR_ROFS, 454 NFSERR_MLINK, 455 NFSERR_NAMETOL, 456 NFSERR_NOTEMPTY, 457 NFSERR_DQUOT, 458 NFSERR_STALE, 459 NFSERR_BADHANDLE, 460 NFSERR_NOTSUPP, 461 NFSERR_SERVERFAULT, 462 0, 463 }; 464 465 static const short nfsv3err_link[] = { 466 NFSERR_IO, 467 NFSERR_IO, 468 NFSERR_ACCES, 469 NFSERR_EXIST, 470 NFSERR_XDEV, 471 NFSERR_NOTDIR, 472 NFSERR_INVAL, 473 NFSERR_NOSPC, 474 NFSERR_ROFS, 475 NFSERR_MLINK, 476 NFSERR_NAMETOL, 477 NFSERR_DQUOT, 478 NFSERR_STALE, 479 NFSERR_BADHANDLE, 480 NFSERR_NOTSUPP, 481 NFSERR_SERVERFAULT, 482 0, 483 }; 484 485 static const short nfsv3err_readdir[] = { 486 NFSERR_IO, 487 NFSERR_IO, 488 NFSERR_ACCES, 489 NFSERR_NOTDIR, 490 NFSERR_STALE, 491 NFSERR_BADHANDLE, 492 NFSERR_BAD_COOKIE, 493 NFSERR_TOOSMALL, 494 NFSERR_SERVERFAULT, 495 0, 496 }; 497 498 static const short nfsv3err_readdirplus[] = { 499 NFSERR_IO, 500 NFSERR_IO, 501 NFSERR_ACCES, 502 NFSERR_NOTDIR, 503 NFSERR_STALE, 504 NFSERR_BADHANDLE, 505 NFSERR_BAD_COOKIE, 506 NFSERR_NOTSUPP, 507 NFSERR_TOOSMALL, 508 NFSERR_SERVERFAULT, 509 0, 510 }; 511 512 static const short nfsv3err_fsstat[] = { 513 NFSERR_IO, 514 NFSERR_IO, 515 NFSERR_STALE, 516 NFSERR_BADHANDLE, 517 NFSERR_SERVERFAULT, 518 0, 519 }; 520 521 static const short nfsv3err_fsinfo[] = { 522 NFSERR_STALE, 523 NFSERR_STALE, 524 NFSERR_BADHANDLE, 525 NFSERR_SERVERFAULT, 526 0, 527 }; 528 529 static const short nfsv3err_pathconf[] = { 530 NFSERR_STALE, 531 NFSERR_STALE, 532 NFSERR_BADHANDLE, 533 NFSERR_SERVERFAULT, 534 0, 535 }; 536 537 static const short nfsv3err_commit[] = { 538 NFSERR_IO, 539 NFSERR_IO, 540 NFSERR_STALE, 541 NFSERR_BADHANDLE, 542 NFSERR_SERVERFAULT, 543 0, 544 }; 545 546 static const short * const nfsrv_v3errmap[] = { 547 nfsv3err_null, 548 nfsv3err_getattr, 549 nfsv3err_setattr, 550 nfsv3err_lookup, 551 nfsv3err_access, 552 nfsv3err_readlink, 553 nfsv3err_read, 554 nfsv3err_write, 555 nfsv3err_create, 556 nfsv3err_mkdir, 557 nfsv3err_symlink, 558 nfsv3err_mknod, 559 nfsv3err_remove, 560 nfsv3err_rmdir, 561 nfsv3err_rename, 562 nfsv3err_link, 563 nfsv3err_readdir, 564 nfsv3err_readdirplus, 565 nfsv3err_fsstat, 566 nfsv3err_fsinfo, 567 nfsv3err_pathconf, 568 nfsv3err_commit, 569 }; 570 571 extern struct nfsrtt nfsrtt; 572 extern time_t nqnfsstarttime; 573 extern int nqsrv_clockskew; 574 extern int nqsrv_writeslack; 575 extern int nqsrv_maxlease; 576 extern const int nqnfs_piggy[NFS_NPROCS]; 577 extern struct nfsnodehashhead *nfsnodehashtbl; 578 extern u_long nfsnodehash; 579 580 u_long nfsdirhashmask; 581 582 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 583 584 /* 585 * Create the header for an rpc request packet 586 * The hsiz is the size of the rest of the nfs request header. 587 * (just used to decide if a cluster is a good idea) 588 */ 589 struct mbuf * 590 nfsm_reqh(np, procid, hsiz, bposp) 591 struct nfsnode *np; 592 u_long procid; 593 int hsiz; 594 caddr_t *bposp; 595 { 596 struct mbuf *mb; 597 caddr_t bpos; 598 #ifndef NFS_V2_ONLY 599 struct nfsmount *nmp; 600 u_int32_t *tl; 601 int nqflag; 602 #endif 603 604 mb = m_get(M_WAIT, MT_DATA); 605 MCLAIM(mb, &nfs_mowner); 606 if (hsiz >= MINCLSIZE) 607 m_clget(mb, M_WAIT); 608 mb->m_len = 0; 609 bpos = mtod(mb, caddr_t); 610 611 #ifndef NFS_V2_ONLY 612 /* 613 * For NQNFS, add lease request. 614 */ 615 if (np) { 616 nmp = VFSTONFS(np->n_vnode->v_mount); 617 if (nmp->nm_flag & NFSMNT_NQNFS) { 618 nqflag = NQNFS_NEEDLEASE(np, procid); 619 if (nqflag) { 620 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); 621 *tl++ = txdr_unsigned(nqflag); 622 *tl = txdr_unsigned(nmp->nm_leaseterm); 623 } else { 624 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 625 *tl = 0; 626 } 627 } 628 } 629 #endif 630 /* Finally, return values */ 631 *bposp = bpos; 632 return (mb); 633 } 634 635 /* 636 * Build the RPC header and fill in the authorization info. 637 * The authorization string argument is only used when the credentials 638 * come from outside of the kernel. 639 * Returns the head of the mbuf list. 640 */ 641 struct mbuf * 642 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 643 verf_str, mrest, mrest_len, mbp, xidp) 644 struct ucred *cr; 645 int nmflag; 646 int procid; 647 int auth_type; 648 int auth_len; 649 char *auth_str; 650 int verf_len; 651 char *verf_str; 652 struct mbuf *mrest; 653 int mrest_len; 654 struct mbuf **mbp; 655 u_int32_t *xidp; 656 { 657 struct mbuf *mb; 658 u_int32_t *tl; 659 caddr_t bpos; 660 int i; 661 struct mbuf *mreq; 662 int siz, grpsiz, authsiz; 663 664 authsiz = nfsm_rndup(auth_len); 665 mb = m_gethdr(M_WAIT, MT_DATA); 666 MCLAIM(mb, &nfs_mowner); 667 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 668 m_clget(mb, M_WAIT); 669 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 670 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 671 } else { 672 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 673 } 674 mb->m_len = 0; 675 mreq = mb; 676 bpos = mtod(mb, caddr_t); 677 678 /* 679 * First the RPC header. 680 */ 681 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 682 683 *tl++ = *xidp = nfs_getxid(); 684 *tl++ = rpc_call; 685 *tl++ = rpc_vers; 686 if (nmflag & NFSMNT_NQNFS) { 687 *tl++ = txdr_unsigned(NQNFS_PROG); 688 *tl++ = txdr_unsigned(NQNFS_VER3); 689 } else { 690 *tl++ = txdr_unsigned(NFS_PROG); 691 if (nmflag & NFSMNT_NFSV3) 692 *tl++ = txdr_unsigned(NFS_VER3); 693 else 694 *tl++ = txdr_unsigned(NFS_VER2); 695 } 696 if (nmflag & NFSMNT_NFSV3) 697 *tl++ = txdr_unsigned(procid); 698 else 699 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 700 701 /* 702 * And then the authorization cred. 703 */ 704 *tl++ = txdr_unsigned(auth_type); 705 *tl = txdr_unsigned(authsiz); 706 switch (auth_type) { 707 case RPCAUTH_UNIX: 708 nfsm_build(tl, u_int32_t *, auth_len); 709 *tl++ = 0; /* stamp ?? */ 710 *tl++ = 0; /* NULL hostname */ 711 *tl++ = txdr_unsigned(cr->cr_uid); 712 *tl++ = txdr_unsigned(cr->cr_gid); 713 grpsiz = (auth_len >> 2) - 5; 714 *tl++ = txdr_unsigned(grpsiz); 715 for (i = 0; i < grpsiz; i++) 716 *tl++ = txdr_unsigned(cr->cr_groups[i]); 717 break; 718 case RPCAUTH_KERB4: 719 siz = auth_len; 720 while (siz > 0) { 721 if (M_TRAILINGSPACE(mb) == 0) { 722 struct mbuf *mb2; 723 mb2 = m_get(M_WAIT, MT_DATA); 724 MCLAIM(mb2, &nfs_mowner); 725 if (siz >= MINCLSIZE) 726 m_clget(mb2, M_WAIT); 727 mb->m_next = mb2; 728 mb = mb2; 729 mb->m_len = 0; 730 bpos = mtod(mb, caddr_t); 731 } 732 i = min(siz, M_TRAILINGSPACE(mb)); 733 memcpy(bpos, auth_str, i); 734 mb->m_len += i; 735 auth_str += i; 736 bpos += i; 737 siz -= i; 738 } 739 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 740 for (i = 0; i < siz; i++) 741 *bpos++ = '\0'; 742 mb->m_len += siz; 743 } 744 break; 745 }; 746 747 /* 748 * And the verifier... 749 */ 750 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 751 if (verf_str) { 752 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 753 *tl = txdr_unsigned(verf_len); 754 siz = verf_len; 755 while (siz > 0) { 756 if (M_TRAILINGSPACE(mb) == 0) { 757 struct mbuf *mb2; 758 mb2 = m_get(M_WAIT, MT_DATA); 759 MCLAIM(mb2, &nfs_mowner); 760 if (siz >= MINCLSIZE) 761 m_clget(mb2, M_WAIT); 762 mb->m_next = mb2; 763 mb = mb2; 764 mb->m_len = 0; 765 bpos = mtod(mb, caddr_t); 766 } 767 i = min(siz, M_TRAILINGSPACE(mb)); 768 memcpy(bpos, verf_str, i); 769 mb->m_len += i; 770 verf_str += i; 771 bpos += i; 772 siz -= i; 773 } 774 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 775 for (i = 0; i < siz; i++) 776 *bpos++ = '\0'; 777 mb->m_len += siz; 778 } 779 } else { 780 *tl++ = txdr_unsigned(RPCAUTH_NULL); 781 *tl = 0; 782 } 783 mb->m_next = mrest; 784 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 785 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 786 *mbp = mb; 787 return (mreq); 788 } 789 790 /* 791 * copies mbuf chain to the uio scatter/gather list 792 */ 793 int 794 nfsm_mbuftouio(mrep, uiop, siz, dpos) 795 struct mbuf **mrep; 796 struct uio *uiop; 797 int siz; 798 caddr_t *dpos; 799 { 800 char *mbufcp, *uiocp; 801 int xfer, left, len; 802 struct mbuf *mp; 803 long uiosiz, rem; 804 int error = 0; 805 806 mp = *mrep; 807 mbufcp = *dpos; 808 len = mtod(mp, caddr_t)+mp->m_len-mbufcp; 809 rem = nfsm_rndup(siz)-siz; 810 while (siz > 0) { 811 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 812 return (EFBIG); 813 left = uiop->uio_iov->iov_len; 814 uiocp = uiop->uio_iov->iov_base; 815 if (left > siz) 816 left = siz; 817 uiosiz = left; 818 while (left > 0) { 819 while (len == 0) { 820 mp = mp->m_next; 821 if (mp == NULL) 822 return (EBADRPC); 823 mbufcp = mtod(mp, caddr_t); 824 len = mp->m_len; 825 } 826 xfer = (left > len) ? len : left; 827 #ifdef notdef 828 /* Not Yet.. */ 829 if (uiop->uio_iov->iov_op != NULL) 830 (*(uiop->uio_iov->iov_op)) 831 (mbufcp, uiocp, xfer); 832 else 833 #endif 834 if (uiop->uio_segflg == UIO_SYSSPACE) 835 memcpy(uiocp, mbufcp, xfer); 836 else 837 copyout(mbufcp, uiocp, xfer); 838 left -= xfer; 839 len -= xfer; 840 mbufcp += xfer; 841 uiocp += xfer; 842 uiop->uio_offset += xfer; 843 uiop->uio_resid -= xfer; 844 } 845 if (uiop->uio_iov->iov_len <= siz) { 846 uiop->uio_iovcnt--; 847 uiop->uio_iov++; 848 } else { 849 uiop->uio_iov->iov_base = 850 (caddr_t)uiop->uio_iov->iov_base + uiosiz; 851 uiop->uio_iov->iov_len -= uiosiz; 852 } 853 siz -= uiosiz; 854 } 855 *dpos = mbufcp; 856 *mrep = mp; 857 if (rem > 0) { 858 if (len < rem) 859 error = nfs_adv(mrep, dpos, rem, len); 860 else 861 *dpos += rem; 862 } 863 return (error); 864 } 865 866 /* 867 * copies a uio scatter/gather list to an mbuf chain. 868 * NOTE: can ony handle iovcnt == 1 869 */ 870 int 871 nfsm_uiotombuf(uiop, mq, siz, bpos) 872 struct uio *uiop; 873 struct mbuf **mq; 874 int siz; 875 caddr_t *bpos; 876 { 877 char *uiocp; 878 struct mbuf *mp, *mp2; 879 int xfer, left, mlen; 880 int uiosiz, clflg, rem; 881 char *cp; 882 883 #ifdef DIAGNOSTIC 884 if (uiop->uio_iovcnt != 1) 885 panic("nfsm_uiotombuf: iovcnt != 1"); 886 #endif 887 888 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 889 clflg = 1; 890 else 891 clflg = 0; 892 rem = nfsm_rndup(siz)-siz; 893 mp = mp2 = *mq; 894 while (siz > 0) { 895 left = uiop->uio_iov->iov_len; 896 uiocp = uiop->uio_iov->iov_base; 897 if (left > siz) 898 left = siz; 899 uiosiz = left; 900 while (left > 0) { 901 mlen = M_TRAILINGSPACE(mp); 902 if (mlen == 0) { 903 mp = m_get(M_WAIT, MT_DATA); 904 MCLAIM(mp, &nfs_mowner); 905 if (clflg) 906 m_clget(mp, M_WAIT); 907 mp->m_len = 0; 908 mp2->m_next = mp; 909 mp2 = mp; 910 mlen = M_TRAILINGSPACE(mp); 911 } 912 xfer = (left > mlen) ? mlen : left; 913 #ifdef notdef 914 /* Not Yet.. */ 915 if (uiop->uio_iov->iov_op != NULL) 916 (*(uiop->uio_iov->iov_op)) 917 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 918 else 919 #endif 920 if (uiop->uio_segflg == UIO_SYSSPACE) 921 memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer); 922 else 923 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 924 mp->m_len += xfer; 925 left -= xfer; 926 uiocp += xfer; 927 uiop->uio_offset += xfer; 928 uiop->uio_resid -= xfer; 929 } 930 uiop->uio_iov->iov_base = (caddr_t)uiop->uio_iov->iov_base + 931 uiosiz; 932 uiop->uio_iov->iov_len -= uiosiz; 933 siz -= uiosiz; 934 } 935 if (rem > 0) { 936 if (rem > M_TRAILINGSPACE(mp)) { 937 mp = m_get(M_WAIT, MT_DATA); 938 MCLAIM(mp, &nfs_mowner); 939 mp->m_len = 0; 940 mp2->m_next = mp; 941 } 942 cp = mtod(mp, caddr_t)+mp->m_len; 943 for (left = 0; left < rem; left++) 944 *cp++ = '\0'; 945 mp->m_len += rem; 946 *bpos = cp; 947 } else 948 *bpos = mtod(mp, caddr_t)+mp->m_len; 949 *mq = mp; 950 return (0); 951 } 952 953 /* 954 * Get at least "siz" bytes of correctly aligned data. 955 * When called the mbuf pointers are not necessarily correct, 956 * dsosp points to what ought to be in m_data and left contains 957 * what ought to be in m_len. 958 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 959 * cases. (The macros use the vars. dpos and dpos2) 960 */ 961 int 962 nfsm_disct(mdp, dposp, siz, left, cp2) 963 struct mbuf **mdp; 964 caddr_t *dposp; 965 int siz; 966 int left; 967 caddr_t *cp2; 968 { 969 struct mbuf *m1, *m2; 970 struct mbuf *havebuf = NULL; 971 caddr_t src = *dposp; 972 caddr_t dst; 973 int len; 974 975 #ifdef DEBUG 976 if (left < 0) 977 panic("nfsm_disct: left < 0"); 978 #endif 979 m1 = *mdp; 980 /* 981 * Skip through the mbuf chain looking for an mbuf with 982 * some data. If the first mbuf found has enough data 983 * and it is correctly aligned return it. 984 */ 985 while (left == 0) { 986 havebuf = m1; 987 *mdp = m1 = m1->m_next; 988 if (m1 == NULL) 989 return (EBADRPC); 990 src = mtod(m1, caddr_t); 991 left = m1->m_len; 992 /* 993 * If we start a new mbuf and it is big enough 994 * and correctly aligned just return it, don't 995 * do any pull up. 996 */ 997 if (left >= siz && nfsm_aligned(src)) { 998 *cp2 = src; 999 *dposp = src + siz; 1000 return (0); 1001 } 1002 } 1003 if (m1->m_flags & M_EXT) { 1004 if (havebuf) { 1005 /* If the first mbuf with data has external data 1006 * and there is a previous empty mbuf use it 1007 * to move the data into. 1008 */ 1009 m2 = m1; 1010 *mdp = m1 = havebuf; 1011 if (m1->m_flags & M_EXT) { 1012 MEXTREMOVE(m1); 1013 } 1014 } else { 1015 /* 1016 * If the first mbuf has a external data 1017 * and there is no previous empty mbuf 1018 * allocate a new mbuf and move the external 1019 * data to the new mbuf. Also make the first 1020 * mbuf look empty. 1021 */ 1022 m2 = m_get(M_WAIT, MT_DATA); 1023 m2->m_ext = m1->m_ext; 1024 m2->m_data = src; 1025 m2->m_len = left; 1026 MCLADDREFERENCE(m1, m2); 1027 MEXTREMOVE(m1); 1028 m2->m_next = m1->m_next; 1029 m1->m_next = m2; 1030 } 1031 m1->m_len = 0; 1032 if (m1->m_flags & M_PKTHDR) 1033 dst = m1->m_pktdat; 1034 else 1035 dst = m1->m_dat; 1036 m1->m_data = dst; 1037 } else { 1038 /* 1039 * If the first mbuf has no external data 1040 * move the data to the front of the mbuf. 1041 */ 1042 if (m1->m_flags & M_PKTHDR) 1043 dst = m1->m_pktdat; 1044 else 1045 dst = m1->m_dat; 1046 m1->m_data = dst; 1047 if (dst != src) 1048 memmove(dst, src, left); 1049 dst += left; 1050 m1->m_len = left; 1051 m2 = m1->m_next; 1052 } 1053 *cp2 = m1->m_data; 1054 *dposp = mtod(m1, caddr_t) + siz; 1055 /* 1056 * Loop through mbufs pulling data up into first mbuf until 1057 * the first mbuf is full or there is no more data to 1058 * pullup. 1059 */ 1060 while ((len = M_TRAILINGSPACE(m1)) != 0 && m2) { 1061 if ((len = min(len, m2->m_len)) != 0) 1062 memcpy(dst, m2->m_data, len); 1063 m1->m_len += len; 1064 dst += len; 1065 m2->m_data += len; 1066 m2->m_len -= len; 1067 m2 = m2->m_next; 1068 } 1069 if (m1->m_len < siz) 1070 return (EBADRPC); 1071 return (0); 1072 } 1073 1074 /* 1075 * Advance the position in the mbuf chain. 1076 */ 1077 int 1078 nfs_adv(mdp, dposp, offs, left) 1079 struct mbuf **mdp; 1080 caddr_t *dposp; 1081 int offs; 1082 int left; 1083 { 1084 struct mbuf *m; 1085 int s; 1086 1087 m = *mdp; 1088 s = left; 1089 while (s < offs) { 1090 offs -= s; 1091 m = m->m_next; 1092 if (m == NULL) 1093 return (EBADRPC); 1094 s = m->m_len; 1095 } 1096 *mdp = m; 1097 *dposp = mtod(m, caddr_t)+offs; 1098 return (0); 1099 } 1100 1101 /* 1102 * Copy a string into mbufs for the hard cases... 1103 */ 1104 int 1105 nfsm_strtmbuf(mb, bpos, cp, siz) 1106 struct mbuf **mb; 1107 char **bpos; 1108 const char *cp; 1109 long siz; 1110 { 1111 struct mbuf *m1 = NULL, *m2; 1112 long left, xfer, len, tlen; 1113 u_int32_t *tl; 1114 int putsize; 1115 1116 putsize = 1; 1117 m2 = *mb; 1118 left = M_TRAILINGSPACE(m2); 1119 if (left > 0) { 1120 tl = ((u_int32_t *)(*bpos)); 1121 *tl++ = txdr_unsigned(siz); 1122 putsize = 0; 1123 left -= NFSX_UNSIGNED; 1124 m2->m_len += NFSX_UNSIGNED; 1125 if (left > 0) { 1126 memcpy((caddr_t) tl, cp, left); 1127 siz -= left; 1128 cp += left; 1129 m2->m_len += left; 1130 left = 0; 1131 } 1132 } 1133 /* Loop around adding mbufs */ 1134 while (siz > 0) { 1135 m1 = m_get(M_WAIT, MT_DATA); 1136 MCLAIM(m1, &nfs_mowner); 1137 if (siz > MLEN) 1138 m_clget(m1, M_WAIT); 1139 m1->m_len = NFSMSIZ(m1); 1140 m2->m_next = m1; 1141 m2 = m1; 1142 tl = mtod(m1, u_int32_t *); 1143 tlen = 0; 1144 if (putsize) { 1145 *tl++ = txdr_unsigned(siz); 1146 m1->m_len -= NFSX_UNSIGNED; 1147 tlen = NFSX_UNSIGNED; 1148 putsize = 0; 1149 } 1150 if (siz < m1->m_len) { 1151 len = nfsm_rndup(siz); 1152 xfer = siz; 1153 if (xfer < len) 1154 *(tl+(xfer>>2)) = 0; 1155 } else { 1156 xfer = len = m1->m_len; 1157 } 1158 memcpy((caddr_t) tl, cp, xfer); 1159 m1->m_len = len+tlen; 1160 siz -= xfer; 1161 cp += xfer; 1162 } 1163 *mb = m1; 1164 *bpos = mtod(m1, caddr_t)+m1->m_len; 1165 return (0); 1166 } 1167 1168 /* 1169 * Directory caching routines. They work as follows: 1170 * - a cache is maintained per VDIR nfsnode. 1171 * - for each offset cookie that is exported to userspace, and can 1172 * thus be thrown back at us as an offset to VOP_READDIR, store 1173 * information in the cache. 1174 * - cached are: 1175 * - cookie itself 1176 * - blocknumber (essentially just a search key in the buffer cache) 1177 * - entry number in block. 1178 * - offset cookie of block in which this entry is stored 1179 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1180 * - entries are looked up in a hash table 1181 * - also maintained is an LRU list of entries, used to determine 1182 * which ones to delete if the cache grows too large. 1183 * - if 32 <-> 64 translation mode is requested for a filesystem, 1184 * the cache also functions as a translation table 1185 * - in the translation case, invalidating the cache does not mean 1186 * flushing it, but just marking entries as invalid, except for 1187 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1188 * still be able to use the cache as a translation table. 1189 * - 32 bit cookies are uniquely created by combining the hash table 1190 * entry value, and one generation count per hash table entry, 1191 * incremented each time an entry is appended to the chain. 1192 * - the cache is invalidated each time a direcory is modified 1193 * - sanity checks are also done; if an entry in a block turns 1194 * out not to have a matching cookie, the cache is invalidated 1195 * and a new block starting from the wanted offset is fetched from 1196 * the server. 1197 * - directory entries as read from the server are extended to contain 1198 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1199 * the cache and exporting them to userspace through the cookie 1200 * argument to VOP_READDIR. 1201 */ 1202 1203 u_long 1204 nfs_dirhash(off) 1205 off_t off; 1206 { 1207 int i; 1208 char *cp = (char *)&off; 1209 u_long sum = 0L; 1210 1211 for (i = 0 ; i < sizeof (off); i++) 1212 sum += *cp++; 1213 1214 return sum; 1215 } 1216 1217 void 1218 nfs_initdircache(vp) 1219 struct vnode *vp; 1220 { 1221 struct nfsnode *np = VTONFS(vp); 1222 1223 KASSERT(np->n_dircache == NULL); 1224 1225 np->n_dircachesize = 0; 1226 np->n_dblkno = 1; 1227 np->n_dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, M_NFSDIROFF, 1228 M_WAITOK, &nfsdirhashmask); 1229 TAILQ_INIT(&np->n_dirchain); 1230 } 1231 1232 void 1233 nfs_initdirxlatecookie(vp) 1234 struct vnode *vp; 1235 { 1236 struct nfsnode *np = VTONFS(vp); 1237 1238 KASSERT(VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_XLATECOOKIE); 1239 KASSERT(np->n_dirgens == NULL); 1240 1241 MALLOC(np->n_dirgens, unsigned *, 1242 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, M_WAITOK); 1243 memset((caddr_t)np->n_dirgens, 0, NFS_DIRHASHSIZ * sizeof (unsigned)); 1244 } 1245 1246 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0}; 1247 1248 struct nfsdircache * 1249 nfs_searchdircache(vp, off, do32, hashent) 1250 struct vnode *vp; 1251 off_t off; 1252 int do32; 1253 int *hashent; 1254 { 1255 struct nfsdirhashhead *ndhp; 1256 struct nfsdircache *ndp = NULL; 1257 struct nfsnode *np = VTONFS(vp); 1258 unsigned ent; 1259 1260 /* 1261 * Zero is always a valid cookie. 1262 */ 1263 if (off == 0) 1264 return &dzero; 1265 1266 if (!np->n_dircache) 1267 return NULL; 1268 1269 /* 1270 * We use a 32bit cookie as search key, directly reconstruct 1271 * the hashentry. Else use the hashfunction. 1272 */ 1273 if (do32) { 1274 ent = (u_int32_t)off >> 24; 1275 if (ent >= NFS_DIRHASHSIZ) 1276 return NULL; 1277 ndhp = &np->n_dircache[ent]; 1278 } else { 1279 ndhp = NFSDIRHASH(np, off); 1280 } 1281 1282 if (hashent) 1283 *hashent = (int)(ndhp - np->n_dircache); 1284 if (do32) { 1285 LIST_FOREACH(ndp, ndhp, dc_hash) { 1286 if (ndp->dc_cookie32 == (u_int32_t)off) { 1287 /* 1288 * An invalidated entry will become the 1289 * start of a new block fetched from 1290 * the server. 1291 */ 1292 if (ndp->dc_blkno == -1) { 1293 ndp->dc_blkcookie = ndp->dc_cookie; 1294 ndp->dc_blkno = np->n_dblkno++; 1295 ndp->dc_entry = 0; 1296 } 1297 break; 1298 } 1299 } 1300 } else { 1301 LIST_FOREACH(ndp, ndhp, dc_hash) { 1302 if (ndp->dc_cookie == off) 1303 break; 1304 } 1305 } 1306 return ndp; 1307 } 1308 1309 1310 struct nfsdircache * 1311 nfs_enterdircache(vp, off, blkoff, en, blkno) 1312 struct vnode *vp; 1313 off_t off, blkoff; 1314 int en; 1315 daddr_t blkno; 1316 { 1317 struct nfsnode *np = VTONFS(vp); 1318 struct nfsdirhashhead *ndhp; 1319 struct nfsdircache *ndp = NULL, *first; 1320 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1321 int hashent, gen, overwrite; 1322 1323 if (!np->n_dircache) 1324 /* 1325 * XXX would like to do this in nfs_nget but vtype 1326 * isn't known at that time. 1327 */ 1328 nfs_initdircache(vp); 1329 1330 if ((nmp->nm_flag & NFSMNT_XLATECOOKIE) && !np->n_dirgens) 1331 nfs_initdirxlatecookie(vp); 1332 1333 /* 1334 * XXX refuse entries for offset 0. amd(8) erroneously sets 1335 * cookie 0 for the '.' entry, making this necessary. This 1336 * isn't so bad, as 0 is a special case anyway. 1337 */ 1338 if (off == 0) 1339 return &dzero; 1340 1341 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1342 1343 if (ndp && ndp->dc_blkno != -1) { 1344 /* 1345 * Overwriting an old entry. Check if it's the same. 1346 * If so, just return. If not, remove the old entry. 1347 */ 1348 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1349 return ndp; 1350 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1351 LIST_REMOVE(ndp, dc_hash); 1352 FREE(ndp, M_NFSDIROFF); 1353 ndp = 0; 1354 } 1355 1356 ndhp = &np->n_dircache[hashent]; 1357 1358 if (!ndp) { 1359 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF, 1360 M_WAITOK); 1361 overwrite = 0; 1362 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1363 /* 1364 * We're allocating a new entry, so bump the 1365 * generation number. 1366 */ 1367 gen = ++np->n_dirgens[hashent]; 1368 if (gen == 0) { 1369 np->n_dirgens[hashent]++; 1370 gen++; 1371 } 1372 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1373 } 1374 } else 1375 overwrite = 1; 1376 1377 /* 1378 * If the entry number is 0, we are at the start of a new block, so 1379 * allocate a new blocknumber. 1380 */ 1381 if (en == 0) 1382 ndp->dc_blkno = np->n_dblkno++; 1383 else 1384 ndp->dc_blkno = blkno; 1385 1386 ndp->dc_cookie = off; 1387 ndp->dc_blkcookie = blkoff; 1388 ndp->dc_entry = en; 1389 1390 if (overwrite) 1391 return ndp; 1392 1393 /* 1394 * If the maximum directory cookie cache size has been reached 1395 * for this node, take one off the front. The idea is that 1396 * directories are typically read front-to-back once, so that 1397 * the oldest entries can be thrown away without much performance 1398 * loss. 1399 */ 1400 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1401 first = TAILQ_FIRST(&np->n_dirchain); 1402 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain); 1403 LIST_REMOVE(first, dc_hash); 1404 FREE(first, M_NFSDIROFF); 1405 } else 1406 np->n_dircachesize++; 1407 1408 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1409 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1410 return ndp; 1411 } 1412 1413 void 1414 nfs_invaldircache(vp, forcefree) 1415 struct vnode *vp; 1416 int forcefree; 1417 { 1418 struct nfsnode *np = VTONFS(vp); 1419 struct nfsdircache *ndp = NULL; 1420 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1421 1422 #ifdef DIAGNOSTIC 1423 if (vp->v_type != VDIR) 1424 panic("nfs: invaldircache: not dir"); 1425 #endif 1426 1427 if (!np->n_dircache) 1428 return; 1429 1430 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1431 while ((ndp = TAILQ_FIRST(&np->n_dirchain)) != 0) { 1432 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1433 LIST_REMOVE(ndp, dc_hash); 1434 FREE(ndp, M_NFSDIROFF); 1435 } 1436 np->n_dircachesize = 0; 1437 if (forcefree && np->n_dirgens) { 1438 FREE(np->n_dirgens, M_NFSDIROFF); 1439 np->n_dirgens = NULL; 1440 } 1441 } else { 1442 TAILQ_FOREACH(ndp, &np->n_dirchain, dc_chain) { 1443 ndp->dc_blkno = -1; 1444 } 1445 } 1446 1447 np->n_dblkno = 1; 1448 } 1449 1450 /* 1451 * Called once before VFS init to initialize shared and 1452 * server-specific data structures. 1453 */ 1454 void 1455 nfs_init() 1456 { 1457 nfsrtt.pos = 0; 1458 rpc_vers = txdr_unsigned(RPC_VER2); 1459 rpc_call = txdr_unsigned(RPC_CALL); 1460 rpc_reply = txdr_unsigned(RPC_REPLY); 1461 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1462 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1463 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1464 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1465 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1466 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1467 nfs_prog = txdr_unsigned(NFS_PROG); 1468 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1469 nfs_true = txdr_unsigned(TRUE); 1470 nfs_false = txdr_unsigned(FALSE); 1471 nfs_xdrneg1 = txdr_unsigned(-1); 1472 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1473 if (nfs_ticks < 1) 1474 nfs_ticks = 1; 1475 #ifdef NFSSERVER 1476 nfsrv_init(0); /* Init server data structures */ 1477 nfsrv_initcache(); /* Init the server request cache */ 1478 pool_init(&nfs_srvdesc_pool, sizeof(struct nfsrv_descript), 1479 0, 0, 0, "nfsrvdescpl", &pool_allocator_nointr); 1480 #endif /* NFSSERVER */ 1481 1482 #if defined(NFSSERVER) || !defined(NFS_V2_ONLY) 1483 /* 1484 * Initialize the nqnfs data structures. 1485 */ 1486 if (nqnfsstarttime == 0) { 1487 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1488 + nqsrv_clockskew + nqsrv_writeslack; 1489 NQLOADNOVRAM(nqnfsstarttime); 1490 CIRCLEQ_INIT(&nqtimerhead); 1491 nqfhhashtbl = hashinit(NQLCHSZ, HASH_LIST, M_NQLEASE, 1492 M_WAITOK, &nqfhhash); 1493 } 1494 #endif 1495 1496 exithook_establish(nfs_exit, NULL); 1497 1498 /* 1499 * Initialize reply list and start timer 1500 */ 1501 TAILQ_INIT(&nfs_reqq); 1502 nfs_timer(NULL); 1503 MOWNER_ATTACH(&nfs_mowner); 1504 1505 #ifdef NFS 1506 /* Initialize the kqueue structures */ 1507 nfs_kqinit(); 1508 /* Initialize the iod structures */ 1509 nfs_iodinit(); 1510 #endif 1511 } 1512 1513 #ifdef NFS 1514 /* 1515 * Called once at VFS init to initialize client-specific data structures. 1516 */ 1517 void 1518 nfs_vfs_init() 1519 { 1520 nfs_nhinit(); /* Init the nfsnode table */ 1521 nfs_commitsize = uvmexp.npages << (PAGE_SHIFT - 4); 1522 } 1523 1524 void 1525 nfs_vfs_reinit() 1526 { 1527 nfs_nhreinit(); 1528 } 1529 1530 void 1531 nfs_vfs_done() 1532 { 1533 nfs_nhdone(); 1534 } 1535 1536 /* 1537 * Attribute cache routines. 1538 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1539 * that are on the mbuf list 1540 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1541 * error otherwise 1542 */ 1543 1544 /* 1545 * Load the attribute cache (that lives in the nfsnode entry) with 1546 * the values on the mbuf list and 1547 * Iff vap not NULL 1548 * copy the attributes to *vaper 1549 */ 1550 int 1551 nfsm_loadattrcache(vpp, mdp, dposp, vaper, flags) 1552 struct vnode **vpp; 1553 struct mbuf **mdp; 1554 caddr_t *dposp; 1555 struct vattr *vaper; 1556 int flags; 1557 { 1558 int32_t t1; 1559 caddr_t cp2; 1560 int error = 0; 1561 struct mbuf *md; 1562 int v3 = NFS_ISV3(*vpp); 1563 1564 md = *mdp; 1565 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1566 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1567 if (error) 1568 return (error); 1569 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper, flags); 1570 } 1571 1572 int 1573 nfs_loadattrcache(vpp, fp, vaper, flags) 1574 struct vnode **vpp; 1575 struct nfs_fattr *fp; 1576 struct vattr *vaper; 1577 int flags; 1578 { 1579 struct vnode *vp = *vpp; 1580 struct vattr *vap; 1581 int v3 = NFS_ISV3(vp); 1582 enum vtype vtyp; 1583 u_short vmode; 1584 struct timespec mtime; 1585 struct vnode *nvp; 1586 int32_t rdev; 1587 struct nfsnode *np; 1588 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1589 uid_t uid; 1590 gid_t gid; 1591 1592 if (v3) { 1593 vtyp = nfsv3tov_type(fp->fa_type); 1594 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1595 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1596 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1597 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1598 } else { 1599 vtyp = nfsv2tov_type(fp->fa_type); 1600 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1601 if (vtyp == VNON || vtyp == VREG) 1602 vtyp = IFTOVT(vmode); 1603 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1604 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1605 1606 /* 1607 * Really ugly NFSv2 kludge. 1608 */ 1609 if (vtyp == VCHR && rdev == 0xffffffff) 1610 vtyp = VFIFO; 1611 } 1612 1613 vmode &= ALLPERMS; 1614 1615 /* 1616 * If v_type == VNON it is a new node, so fill in the v_type, 1617 * n_mtime fields. Check to see if it represents a special 1618 * device, and if so, check for a possible alias. Once the 1619 * correct vnode has been obtained, fill in the rest of the 1620 * information. 1621 */ 1622 np = VTONFS(vp); 1623 if (vp->v_type == VNON) { 1624 vp->v_type = vtyp; 1625 if (vp->v_type == VFIFO) { 1626 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1627 vp->v_op = fifo_nfsv2nodeop_p; 1628 } else if (vp->v_type == VREG) { 1629 lockinit(&np->n_commitlock, PINOD, "nfsclock", 0, 0); 1630 } else if (vp->v_type == VCHR || vp->v_type == VBLK) { 1631 vp->v_op = spec_nfsv2nodeop_p; 1632 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1633 if (nvp) { 1634 /* 1635 * Discard unneeded vnode, but save its nfsnode. 1636 * Since the nfsnode does not have a lock, its 1637 * vnode lock has to be carried over. 1638 */ 1639 /* 1640 * XXX is the old node sure to be locked here? 1641 */ 1642 KASSERT(lockstatus(&vp->v_lock) == 1643 LK_EXCLUSIVE); 1644 nvp->v_data = vp->v_data; 1645 vp->v_data = NULL; 1646 VOP_UNLOCK(vp, 0); 1647 vp->v_op = spec_vnodeop_p; 1648 vrele(vp); 1649 vgone(vp); 1650 lockmgr(&nvp->v_lock, LK_EXCLUSIVE, 1651 &nvp->v_interlock); 1652 /* 1653 * Reinitialize aliased node. 1654 */ 1655 np->n_vnode = nvp; 1656 *vpp = vp = nvp; 1657 } 1658 } 1659 np->n_mtime = mtime; 1660 } 1661 uid = fxdr_unsigned(uid_t, fp->fa_uid); 1662 gid = fxdr_unsigned(gid_t, fp->fa_gid); 1663 vap = np->n_vattr; 1664 1665 /* 1666 * Invalidate access cache if uid, gid or mode changed. 1667 */ 1668 if (np->n_accstamp != -1 && 1669 (gid != vap->va_gid || uid != vap->va_uid || vmode != vap->va_mode)) 1670 np->n_accstamp = -1; 1671 1672 vap->va_type = vtyp; 1673 vap->va_mode = vmode; 1674 vap->va_rdev = (dev_t)rdev; 1675 vap->va_mtime = mtime; 1676 vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0]; 1677 switch (vtyp) { 1678 case VDIR: 1679 vap->va_blocksize = NFS_DIRFRAGSIZ; 1680 break; 1681 case VBLK: 1682 vap->va_blocksize = BLKDEV_IOSIZE; 1683 break; 1684 case VCHR: 1685 vap->va_blocksize = MAXBSIZE; 1686 break; 1687 default: 1688 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1689 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1690 break; 1691 } 1692 if (v3) { 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_hyper(&fp->fa3_size); 1697 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1698 vap->va_fileid = fxdr_unsigned(int32_t, 1699 fp->fa3_fileid.nfsuquad[1]); 1700 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1701 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); 1702 vap->va_flags = 0; 1703 vap->va_filerev = 0; 1704 } else { 1705 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1706 vap->va_uid = uid; 1707 vap->va_gid = gid; 1708 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1709 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1710 * NFS_FABLKSIZE; 1711 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1712 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1713 vap->va_flags = 0; 1714 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, 1715 fp->fa2_ctime.nfsv2_sec); 1716 vap->va_ctime.tv_nsec = 0; 1717 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1718 vap->va_filerev = 0; 1719 } 1720 if (vap->va_size != np->n_size) { 1721 if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) { 1722 vap->va_size = np->n_size; 1723 } else { 1724 np->n_size = vap->va_size; 1725 if (vap->va_type == VREG) { 1726 if ((flags & NAC_NOTRUNC) 1727 && np->n_size < vp->v_size) { 1728 /* 1729 * we can't free pages now because 1730 * the pages can be owned by ourselves. 1731 */ 1732 np->n_flag |= NTRUNCDELAYED; 1733 } 1734 else { 1735 uvm_vnp_setsize(vp, np->n_size); 1736 } 1737 } 1738 } 1739 } 1740 np->n_attrstamp = time.tv_sec; 1741 if (vaper != NULL) { 1742 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1743 if (np->n_flag & NCHG) { 1744 if (np->n_flag & NACC) 1745 vaper->va_atime = np->n_atim; 1746 if (np->n_flag & NUPD) 1747 vaper->va_mtime = np->n_mtim; 1748 } 1749 } 1750 return (0); 1751 } 1752 1753 /* 1754 * Check the time stamp 1755 * If the cache is valid, copy contents to *vap and return 0 1756 * otherwise return an error 1757 */ 1758 int 1759 nfs_getattrcache(vp, vaper) 1760 struct vnode *vp; 1761 struct vattr *vaper; 1762 { 1763 struct nfsnode *np = VTONFS(vp); 1764 struct vattr *vap; 1765 1766 if (np->n_attrstamp == 0 || 1767 (time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { 1768 nfsstats.attrcache_misses++; 1769 return (ENOENT); 1770 } 1771 nfsstats.attrcache_hits++; 1772 vap = np->n_vattr; 1773 if (vap->va_size != np->n_size) { 1774 if (vap->va_type == VREG) { 1775 if (np->n_flag & NMODIFIED) { 1776 if (vap->va_size < np->n_size) 1777 vap->va_size = np->n_size; 1778 else 1779 np->n_size = vap->va_size; 1780 } else 1781 np->n_size = vap->va_size; 1782 uvm_vnp_setsize(vp, np->n_size); 1783 } else 1784 np->n_size = vap->va_size; 1785 } 1786 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1787 if (np->n_flag & NCHG) { 1788 if (np->n_flag & NACC) 1789 vaper->va_atime = np->n_atim; 1790 if (np->n_flag & NUPD) 1791 vaper->va_mtime = np->n_mtim; 1792 } 1793 return (0); 1794 } 1795 1796 void 1797 nfs_delayedtruncate(vp) 1798 struct vnode *vp; 1799 { 1800 struct nfsnode *np = VTONFS(vp); 1801 1802 if (np->n_flag & NTRUNCDELAYED) { 1803 np->n_flag &= ~NTRUNCDELAYED; 1804 uvm_vnp_setsize(vp, np->n_size); 1805 } 1806 } 1807 1808 /* 1809 * Heuristic to see if the server XDR encodes directory cookies or not. 1810 * it is not supposed to, but a lot of servers may do this. Also, since 1811 * most/all servers will implement V2 as well, it is expected that they 1812 * may return just 32 bits worth of cookie information, so we need to 1813 * find out in which 32 bits this information is available. We do this 1814 * to avoid trouble with emulated binaries that can't handle 64 bit 1815 * directory offsets. 1816 */ 1817 1818 void 1819 nfs_cookieheuristic(vp, flagp, p, cred) 1820 struct vnode *vp; 1821 int *flagp; 1822 struct proc *p; 1823 struct ucred *cred; 1824 { 1825 struct uio auio; 1826 struct iovec aiov; 1827 caddr_t buf, cp; 1828 struct dirent *dp; 1829 off_t *cookies = NULL, *cop; 1830 int error, eof, nc, len; 1831 1832 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 1833 1834 aiov.iov_base = buf; 1835 aiov.iov_len = NFS_DIRFRAGSIZ; 1836 auio.uio_iov = &aiov; 1837 auio.uio_iovcnt = 1; 1838 auio.uio_rw = UIO_READ; 1839 auio.uio_segflg = UIO_SYSSPACE; 1840 auio.uio_procp = p; 1841 auio.uio_resid = NFS_DIRFRAGSIZ; 1842 auio.uio_offset = 0; 1843 1844 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 1845 1846 len = NFS_DIRFRAGSIZ - auio.uio_resid; 1847 if (error || len == 0) { 1848 FREE(buf, M_TEMP); 1849 if (cookies) 1850 free(cookies, M_TEMP); 1851 return; 1852 } 1853 1854 /* 1855 * Find the first valid entry and look at its offset cookie. 1856 */ 1857 1858 cp = buf; 1859 for (cop = cookies; len > 0; len -= dp->d_reclen) { 1860 dp = (struct dirent *)cp; 1861 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 1862 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 1863 *flagp |= NFSMNT_SWAPCOOKIE; 1864 nfs_invaldircache(vp, 0); 1865 nfs_vinvalbuf(vp, 0, cred, p, 1); 1866 } 1867 break; 1868 } 1869 cop++; 1870 cp += dp->d_reclen; 1871 } 1872 1873 FREE(buf, M_TEMP); 1874 free(cookies, M_TEMP); 1875 } 1876 #endif /* NFS */ 1877 1878 /* 1879 * Set up nameidata for a lookup() call and do it. 1880 * 1881 * If pubflag is set, this call is done for a lookup operation on the 1882 * public filehandle. In that case we allow crossing mountpoints and 1883 * absolute pathnames. However, the caller is expected to check that 1884 * the lookup result is within the public fs, and deny access if 1885 * it is not. 1886 */ 1887 int 1888 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) 1889 struct nameidata *ndp; 1890 fhandle_t *fhp; 1891 uint32_t len; 1892 struct nfssvc_sock *slp; 1893 struct mbuf *nam; 1894 struct mbuf **mdp; 1895 caddr_t *dposp; 1896 struct vnode **retdirp; 1897 struct proc *p; 1898 int kerbflag, pubflag; 1899 { 1900 int i, rem; 1901 struct mbuf *md; 1902 char *fromcp, *tocp, *cp; 1903 struct iovec aiov; 1904 struct uio auio; 1905 struct vnode *dp; 1906 int error, rdonly, linklen; 1907 struct componentname *cnp = &ndp->ni_cnd; 1908 1909 *retdirp = (struct vnode *)0; 1910 1911 if ((len + 1) > MAXPATHLEN) 1912 return (ENAMETOOLONG); 1913 cnp->cn_pnbuf = PNBUF_GET(); 1914 1915 /* 1916 * Copy the name from the mbuf list to ndp->ni_pnbuf 1917 * and set the various ndp fields appropriately. 1918 */ 1919 fromcp = *dposp; 1920 tocp = cnp->cn_pnbuf; 1921 md = *mdp; 1922 rem = mtod(md, caddr_t) + md->m_len - fromcp; 1923 for (i = 0; i < len; i++) { 1924 while (rem == 0) { 1925 md = md->m_next; 1926 if (md == NULL) { 1927 error = EBADRPC; 1928 goto out; 1929 } 1930 fromcp = mtod(md, caddr_t); 1931 rem = md->m_len; 1932 } 1933 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 1934 error = EACCES; 1935 goto out; 1936 } 1937 *tocp++ = *fromcp++; 1938 rem--; 1939 } 1940 *tocp = '\0'; 1941 *mdp = md; 1942 *dposp = fromcp; 1943 len = nfsm_rndup(len)-len; 1944 if (len > 0) { 1945 if (rem >= len) 1946 *dposp += len; 1947 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 1948 goto out; 1949 } 1950 1951 /* 1952 * Extract and set starting directory. 1953 */ 1954 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 1955 nam, &rdonly, kerbflag, pubflag); 1956 if (error) 1957 goto out; 1958 if (dp->v_type != VDIR) { 1959 vrele(dp); 1960 error = ENOTDIR; 1961 goto out; 1962 } 1963 1964 if (rdonly) 1965 cnp->cn_flags |= RDONLY; 1966 1967 *retdirp = dp; 1968 1969 if (pubflag) { 1970 /* 1971 * Oh joy. For WebNFS, handle those pesky '%' escapes, 1972 * and the 'native path' indicator. 1973 */ 1974 cp = PNBUF_GET(); 1975 fromcp = cnp->cn_pnbuf; 1976 tocp = cp; 1977 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 1978 switch ((unsigned char)*fromcp) { 1979 case WEBNFS_NATIVE_CHAR: 1980 /* 1981 * 'Native' path for us is the same 1982 * as a path according to the NFS spec, 1983 * just skip the escape char. 1984 */ 1985 fromcp++; 1986 break; 1987 /* 1988 * More may be added in the future, range 0x80-0xff 1989 */ 1990 default: 1991 error = EIO; 1992 PNBUF_PUT(cp); 1993 goto out; 1994 } 1995 } 1996 /* 1997 * Translate the '%' escapes, URL-style. 1998 */ 1999 while (*fromcp != '\0') { 2000 if (*fromcp == WEBNFS_ESC_CHAR) { 2001 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 2002 fromcp++; 2003 *tocp++ = HEXSTRTOI(fromcp); 2004 fromcp += 2; 2005 continue; 2006 } else { 2007 error = ENOENT; 2008 PNBUF_PUT(cp); 2009 goto out; 2010 } 2011 } else 2012 *tocp++ = *fromcp++; 2013 } 2014 *tocp = '\0'; 2015 PNBUF_PUT(cnp->cn_pnbuf); 2016 cnp->cn_pnbuf = cp; 2017 } 2018 2019 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 2020 ndp->ni_segflg = UIO_SYSSPACE; 2021 ndp->ni_rootdir = rootvnode; 2022 2023 if (pubflag) { 2024 ndp->ni_loopcnt = 0; 2025 if (cnp->cn_pnbuf[0] == '/') 2026 dp = rootvnode; 2027 } else { 2028 cnp->cn_flags |= NOCROSSMOUNT; 2029 } 2030 2031 cnp->cn_proc = p; 2032 VREF(dp); 2033 2034 for (;;) { 2035 cnp->cn_nameptr = cnp->cn_pnbuf; 2036 ndp->ni_startdir = dp; 2037 /* 2038 * And call lookup() to do the real work 2039 */ 2040 error = lookup(ndp); 2041 if (error) { 2042 PNBUF_PUT(cnp->cn_pnbuf); 2043 return (error); 2044 } 2045 /* 2046 * Check for encountering a symbolic link 2047 */ 2048 if ((cnp->cn_flags & ISSYMLINK) == 0) { 2049 if (cnp->cn_flags & (SAVENAME | SAVESTART)) 2050 cnp->cn_flags |= HASBUF; 2051 else 2052 PNBUF_PUT(cnp->cn_pnbuf); 2053 return (0); 2054 } else { 2055 if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN)) 2056 VOP_UNLOCK(ndp->ni_dvp, 0); 2057 if (!pubflag) { 2058 error = EINVAL; 2059 break; 2060 } 2061 2062 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 2063 error = ELOOP; 2064 break; 2065 } 2066 if (ndp->ni_vp->v_mount->mnt_flag & MNT_SYMPERM) { 2067 error = VOP_ACCESS(ndp->ni_vp, VEXEC, cnp->cn_cred, 2068 cnp->cn_proc); 2069 if (error != 0) 2070 break; 2071 } 2072 if (ndp->ni_pathlen > 1) 2073 cp = PNBUF_GET(); 2074 else 2075 cp = cnp->cn_pnbuf; 2076 aiov.iov_base = cp; 2077 aiov.iov_len = MAXPATHLEN; 2078 auio.uio_iov = &aiov; 2079 auio.uio_iovcnt = 1; 2080 auio.uio_offset = 0; 2081 auio.uio_rw = UIO_READ; 2082 auio.uio_segflg = UIO_SYSSPACE; 2083 auio.uio_procp = (struct proc *)0; 2084 auio.uio_resid = MAXPATHLEN; 2085 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 2086 if (error) { 2087 badlink: 2088 if (ndp->ni_pathlen > 1) 2089 PNBUF_PUT(cp); 2090 break; 2091 } 2092 linklen = MAXPATHLEN - auio.uio_resid; 2093 if (linklen == 0) { 2094 error = ENOENT; 2095 goto badlink; 2096 } 2097 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 2098 error = ENAMETOOLONG; 2099 goto badlink; 2100 } 2101 if (ndp->ni_pathlen > 1) { 2102 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 2103 PNBUF_PUT(cnp->cn_pnbuf); 2104 cnp->cn_pnbuf = cp; 2105 } else 2106 cnp->cn_pnbuf[linklen] = '\0'; 2107 ndp->ni_pathlen += linklen; 2108 vput(ndp->ni_vp); 2109 dp = ndp->ni_dvp; 2110 /* 2111 * Check if root directory should replace current directory. 2112 */ 2113 if (cnp->cn_pnbuf[0] == '/') { 2114 vrele(dp); 2115 dp = ndp->ni_rootdir; 2116 VREF(dp); 2117 } 2118 } 2119 } 2120 vrele(ndp->ni_dvp); 2121 vput(ndp->ni_vp); 2122 ndp->ni_vp = NULL; 2123 out: 2124 PNBUF_PUT(cnp->cn_pnbuf); 2125 return (error); 2126 } 2127 2128 /* 2129 * A fiddled version of m_adj() that ensures null fill to a 32-bit 2130 * boundary and only trims off the back end 2131 * 2132 * 1. trim off 'len' bytes as m_adj(mp, -len). 2133 * 2. add zero-padding 'nul' bytes at the end of the mbuf chain. 2134 */ 2135 void 2136 nfs_zeropad(mp, len, nul) 2137 struct mbuf *mp; 2138 int len; 2139 int nul; 2140 { 2141 struct mbuf *m; 2142 int count; 2143 2144 /* 2145 * Trim from tail. Scan the mbuf chain, 2146 * calculating its length and finding the last mbuf. 2147 * If the adjustment only affects this mbuf, then just 2148 * adjust and return. Otherwise, rescan and truncate 2149 * after the remaining size. 2150 */ 2151 count = 0; 2152 m = mp; 2153 for (;;) { 2154 count += m->m_len; 2155 if (m->m_next == NULL) 2156 break; 2157 m = m->m_next; 2158 } 2159 2160 KDASSERT(count >= len); 2161 2162 if (m->m_len >= len) { 2163 m->m_len -= len; 2164 } else { 2165 count -= len; 2166 /* 2167 * Correct length for chain is "count". 2168 * Find the mbuf with last data, adjust its length, 2169 * and toss data from remaining mbufs on chain. 2170 */ 2171 for (m = mp; m; m = m->m_next) { 2172 if (m->m_len >= count) { 2173 m->m_len = count; 2174 break; 2175 } 2176 count -= m->m_len; 2177 } 2178 m_freem(m->m_next); 2179 m->m_next = NULL; 2180 } 2181 2182 KDASSERT(m->m_next == NULL); 2183 2184 /* 2185 * zero-padding. 2186 */ 2187 if (nul > 0) { 2188 char *cp; 2189 int i; 2190 2191 if (M_ROMAP(m) || M_TRAILINGSPACE(m) < nul) { 2192 struct mbuf *n; 2193 2194 KDASSERT(MLEN >= nul); 2195 n = m_get(M_WAIT, MT_DATA); 2196 MCLAIM(n, &nfs_mowner); 2197 n->m_len = nul; 2198 n->m_next = NULL; 2199 m->m_next = n; 2200 cp = mtod(n, caddr_t); 2201 } else { 2202 cp = mtod(m, caddr_t) + m->m_len; 2203 m->m_len += nul; 2204 } 2205 for (i = 0; i < nul; i++) 2206 *cp++ = '\0'; 2207 } 2208 return; 2209 } 2210 2211 /* 2212 * Make these functions instead of macros, so that the kernel text size 2213 * doesn't get too big... 2214 */ 2215 void 2216 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2217 struct nfsrv_descript *nfsd; 2218 int before_ret; 2219 struct vattr *before_vap; 2220 int after_ret; 2221 struct vattr *after_vap; 2222 struct mbuf **mbp; 2223 char **bposp; 2224 { 2225 struct mbuf *mb = *mbp; 2226 char *bpos = *bposp; 2227 u_int32_t *tl; 2228 2229 if (before_ret) { 2230 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2231 *tl = nfs_false; 2232 } else { 2233 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2234 *tl++ = nfs_true; 2235 txdr_hyper(before_vap->va_size, tl); 2236 tl += 2; 2237 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2238 tl += 2; 2239 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2240 } 2241 *bposp = bpos; 2242 *mbp = mb; 2243 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2244 } 2245 2246 void 2247 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2248 struct nfsrv_descript *nfsd; 2249 int after_ret; 2250 struct vattr *after_vap; 2251 struct mbuf **mbp; 2252 char **bposp; 2253 { 2254 struct mbuf *mb = *mbp; 2255 char *bpos = *bposp; 2256 u_int32_t *tl; 2257 struct nfs_fattr *fp; 2258 2259 if (after_ret) { 2260 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2261 *tl = nfs_false; 2262 } else { 2263 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2264 *tl++ = nfs_true; 2265 fp = (struct nfs_fattr *)tl; 2266 nfsm_srvfattr(nfsd, after_vap, fp); 2267 } 2268 *mbp = mb; 2269 *bposp = bpos; 2270 } 2271 2272 void 2273 nfsm_srvfattr(nfsd, vap, fp) 2274 struct nfsrv_descript *nfsd; 2275 struct vattr *vap; 2276 struct nfs_fattr *fp; 2277 { 2278 2279 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2280 fp->fa_uid = txdr_unsigned(vap->va_uid); 2281 fp->fa_gid = txdr_unsigned(vap->va_gid); 2282 if (nfsd->nd_flag & ND_NFSV3) { 2283 fp->fa_type = vtonfsv3_type(vap->va_type); 2284 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2285 txdr_hyper(vap->va_size, &fp->fa3_size); 2286 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2287 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2288 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2289 fp->fa3_fsid.nfsuquad[0] = 0; 2290 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2291 fp->fa3_fileid.nfsuquad[0] = 0; 2292 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); 2293 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2294 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2295 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2296 } else { 2297 fp->fa_type = vtonfsv2_type(vap->va_type); 2298 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2299 fp->fa2_size = txdr_unsigned(vap->va_size); 2300 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2301 if (vap->va_type == VFIFO) 2302 fp->fa2_rdev = 0xffffffff; 2303 else 2304 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2305 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2306 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2307 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2308 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2309 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2310 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2311 } 2312 } 2313 2314 /* 2315 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2316 * - look up fsid in mount list (if not found ret error) 2317 * - get vp and export rights by calling VFS_FHTOVP() 2318 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2319 * - if not lockflag unlock it with VOP_UNLOCK() 2320 */ 2321 int 2322 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) 2323 fhandle_t *fhp; 2324 int lockflag; 2325 struct vnode **vpp; 2326 struct ucred *cred; 2327 struct nfssvc_sock *slp; 2328 struct mbuf *nam; 2329 int *rdonlyp; 2330 int kerbflag; 2331 { 2332 struct mount *mp; 2333 int i; 2334 struct ucred *credanon; 2335 int error, exflags; 2336 struct sockaddr_in *saddr; 2337 2338 *vpp = (struct vnode *)0; 2339 2340 if (nfs_ispublicfh(fhp)) { 2341 if (!pubflag || !nfs_pub.np_valid) 2342 return (ESTALE); 2343 fhp = &nfs_pub.np_handle; 2344 } 2345 2346 mp = vfs_getvfs(&fhp->fh_fsid); 2347 if (!mp) 2348 return (ESTALE); 2349 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); 2350 if (error) 2351 return (error); 2352 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2353 if (error) 2354 return (error); 2355 2356 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2357 saddr = mtod(nam, struct sockaddr_in *); 2358 if ((saddr->sin_family == AF_INET) && 2359 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2360 vput(*vpp); 2361 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2362 } 2363 #ifdef INET6 2364 if ((saddr->sin_family == AF_INET6) && 2365 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2366 vput(*vpp); 2367 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2368 } 2369 #endif 2370 } 2371 /* 2372 * Check/setup credentials. 2373 */ 2374 if (exflags & MNT_EXKERB) { 2375 if (!kerbflag) { 2376 vput(*vpp); 2377 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2378 } 2379 } else if (kerbflag) { 2380 vput(*vpp); 2381 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2382 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { 2383 cred->cr_uid = credanon->cr_uid; 2384 cred->cr_gid = credanon->cr_gid; 2385 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) 2386 cred->cr_groups[i] = credanon->cr_groups[i]; 2387 cred->cr_ngroups = i; 2388 } 2389 if (exflags & MNT_EXRDONLY) 2390 *rdonlyp = 1; 2391 else 2392 *rdonlyp = 0; 2393 if (!lockflag) 2394 VOP_UNLOCK(*vpp, 0); 2395 return (0); 2396 } 2397 2398 /* 2399 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2400 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has 2401 * transformed this to all zeroes in both cases, so check for it. 2402 */ 2403 int 2404 nfs_ispublicfh(fhp) 2405 fhandle_t *fhp; 2406 { 2407 char *cp = (char *)fhp; 2408 int i; 2409 2410 for (i = 0; i < NFSX_V3FH; i++) 2411 if (*cp++ != 0) 2412 return (FALSE); 2413 return (TRUE); 2414 } 2415 2416 /* 2417 * This function compares two net addresses by family and returns TRUE 2418 * if they are the same host. 2419 * If there is any doubt, return FALSE. 2420 * The AF_INET family is handled as a special case so that address mbufs 2421 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2422 */ 2423 int 2424 netaddr_match(family, haddr, nam) 2425 int family; 2426 union nethostaddr *haddr; 2427 struct mbuf *nam; 2428 { 2429 struct sockaddr_in *inetaddr; 2430 2431 switch (family) { 2432 case AF_INET: 2433 inetaddr = mtod(nam, struct sockaddr_in *); 2434 if (inetaddr->sin_family == AF_INET && 2435 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2436 return (1); 2437 break; 2438 #ifdef INET6 2439 case AF_INET6: 2440 { 2441 struct sockaddr_in6 *sin6_1, *sin6_2; 2442 2443 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2444 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2445 if (sin6_1->sin6_family == AF_INET6 && 2446 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2447 return 1; 2448 } 2449 #endif 2450 #ifdef ISO 2451 case AF_ISO: 2452 { 2453 struct sockaddr_iso *isoaddr1, *isoaddr2; 2454 2455 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2456 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2457 if (isoaddr1->siso_family == AF_ISO && 2458 isoaddr1->siso_nlen > 0 && 2459 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2460 SAME_ISOADDR(isoaddr1, isoaddr2)) 2461 return (1); 2462 break; 2463 } 2464 #endif /* ISO */ 2465 default: 2466 break; 2467 }; 2468 return (0); 2469 } 2470 2471 /* 2472 * The write verifier has changed (probably due to a server reboot), so all 2473 * PG_NEEDCOMMIT pages will have to be written again. Since they are marked 2474 * as dirty or are being written out just now, all this takes is clearing 2475 * the PG_NEEDCOMMIT flag. Once done the new write verifier can be set for 2476 * the mount point. 2477 */ 2478 void 2479 nfs_clearcommit(mp) 2480 struct mount *mp; 2481 { 2482 struct vnode *vp; 2483 struct nfsnode *np; 2484 struct vm_page *pg; 2485 struct nfsmount *nmp = VFSTONFS(mp); 2486 2487 lockmgr(&nmp->nm_writeverflock, LK_EXCLUSIVE, NULL); 2488 2489 LIST_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2490 KASSERT(vp->v_mount == mp); 2491 if (vp->v_type == VNON) 2492 continue; 2493 np = VTONFS(vp); 2494 np->n_pushlo = np->n_pushhi = np->n_pushedlo = 2495 np->n_pushedhi = 0; 2496 np->n_commitflags &= 2497 ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID); 2498 simple_lock(&vp->v_uobj.vmobjlock); 2499 TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq) { 2500 pg->flags &= ~PG_NEEDCOMMIT; 2501 } 2502 simple_unlock(&vp->v_uobj.vmobjlock); 2503 } 2504 simple_lock(&nmp->nm_slock); 2505 nmp->nm_iflag &= ~NFSMNT_STALEWRITEVERF; 2506 simple_unlock(&nmp->nm_slock); 2507 lockmgr(&nmp->nm_writeverflock, LK_RELEASE, NULL); 2508 } 2509 2510 void 2511 nfs_merge_commit_ranges(vp) 2512 struct vnode *vp; 2513 { 2514 struct nfsnode *np = VTONFS(vp); 2515 2516 KASSERT(np->n_commitflags & NFS_COMMIT_PUSH_VALID); 2517 2518 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2519 np->n_pushedlo = np->n_pushlo; 2520 np->n_pushedhi = np->n_pushhi; 2521 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2522 } else { 2523 if (np->n_pushlo < np->n_pushedlo) 2524 np->n_pushedlo = np->n_pushlo; 2525 if (np->n_pushhi > np->n_pushedhi) 2526 np->n_pushedhi = np->n_pushhi; 2527 } 2528 2529 np->n_pushlo = np->n_pushhi = 0; 2530 np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID; 2531 2532 #ifdef NFS_DEBUG_COMMIT 2533 printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2534 (unsigned)np->n_pushedhi); 2535 #endif 2536 } 2537 2538 int 2539 nfs_in_committed_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_PUSHED_VALID)) 2547 return 0; 2548 lo = off; 2549 hi = lo + len; 2550 2551 return (lo >= np->n_pushedlo && hi <= np->n_pushedhi); 2552 } 2553 2554 int 2555 nfs_in_tobecommitted_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 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2563 return 0; 2564 lo = off; 2565 hi = lo + len; 2566 2567 return (lo >= np->n_pushlo && hi <= np->n_pushhi); 2568 } 2569 2570 void 2571 nfs_add_committed_range(vp, off, len) 2572 struct vnode *vp; 2573 off_t off, len; 2574 { 2575 struct nfsnode *np = VTONFS(vp); 2576 off_t lo, hi; 2577 2578 lo = off; 2579 hi = lo + len; 2580 2581 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2582 np->n_pushedlo = lo; 2583 np->n_pushedhi = hi; 2584 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2585 } else { 2586 if (hi > np->n_pushedhi) 2587 np->n_pushedhi = hi; 2588 if (lo < np->n_pushedlo) 2589 np->n_pushedlo = lo; 2590 } 2591 #ifdef NFS_DEBUG_COMMIT 2592 printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2593 (unsigned)np->n_pushedhi); 2594 #endif 2595 } 2596 2597 void 2598 nfs_del_committed_range(vp, off, len) 2599 struct vnode *vp; 2600 off_t off, len; 2601 { 2602 struct nfsnode *np = VTONFS(vp); 2603 off_t lo, hi; 2604 2605 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2606 return; 2607 2608 lo = off; 2609 hi = lo + len; 2610 2611 if (lo > np->n_pushedhi || hi < np->n_pushedlo) 2612 return; 2613 if (lo <= np->n_pushedlo) 2614 np->n_pushedlo = hi; 2615 else if (hi >= np->n_pushedhi) 2616 np->n_pushedhi = lo; 2617 else { 2618 /* 2619 * XXX There's only one range. If the deleted range 2620 * is in the middle, pick the largest of the 2621 * contiguous ranges that it leaves. 2622 */ 2623 if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi)) 2624 np->n_pushedhi = lo; 2625 else 2626 np->n_pushedlo = hi; 2627 } 2628 #ifdef NFS_DEBUG_COMMIT 2629 printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2630 (unsigned)np->n_pushedhi); 2631 #endif 2632 } 2633 2634 void 2635 nfs_add_tobecommitted_range(vp, off, len) 2636 struct vnode *vp; 2637 off_t off, len; 2638 { 2639 struct nfsnode *np = VTONFS(vp); 2640 off_t lo, hi; 2641 2642 lo = off; 2643 hi = lo + len; 2644 2645 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) { 2646 np->n_pushlo = lo; 2647 np->n_pushhi = hi; 2648 np->n_commitflags |= NFS_COMMIT_PUSH_VALID; 2649 } else { 2650 if (lo < np->n_pushlo) 2651 np->n_pushlo = lo; 2652 if (hi > np->n_pushhi) 2653 np->n_pushhi = hi; 2654 } 2655 #ifdef NFS_DEBUG_COMMIT 2656 printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2657 (unsigned)np->n_pushhi); 2658 #endif 2659 } 2660 2661 void 2662 nfs_del_tobecommitted_range(vp, off, len) 2663 struct vnode *vp; 2664 off_t off, len; 2665 { 2666 struct nfsnode *np = VTONFS(vp); 2667 off_t lo, hi; 2668 2669 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2670 return; 2671 2672 lo = off; 2673 hi = lo + len; 2674 2675 if (lo > np->n_pushhi || hi < np->n_pushlo) 2676 return; 2677 2678 if (lo <= np->n_pushlo) 2679 np->n_pushlo = hi; 2680 else if (hi >= np->n_pushhi) 2681 np->n_pushhi = lo; 2682 else { 2683 /* 2684 * XXX There's only one range. If the deleted range 2685 * is in the middle, pick the largest of the 2686 * contiguous ranges that it leaves. 2687 */ 2688 if ((np->n_pushlo - lo) > (hi - np->n_pushhi)) 2689 np->n_pushhi = lo; 2690 else 2691 np->n_pushlo = hi; 2692 } 2693 #ifdef NFS_DEBUG_COMMIT 2694 printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2695 (unsigned)np->n_pushhi); 2696 #endif 2697 } 2698 2699 /* 2700 * Map errnos to NFS error numbers. For Version 3 also filter out error 2701 * numbers not specified for the associated procedure. 2702 */ 2703 int 2704 nfsrv_errmap(nd, err) 2705 struct nfsrv_descript *nd; 2706 int err; 2707 { 2708 const short *defaulterrp, *errp; 2709 2710 if (nd->nd_flag & ND_NFSV3) { 2711 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2712 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2713 while (*++errp) { 2714 if (*errp == err) 2715 return (err); 2716 else if (*errp > err) 2717 break; 2718 } 2719 return ((int)*defaulterrp); 2720 } else 2721 return (err & 0xffff); 2722 } 2723 if (err <= ELAST) 2724 return ((int)nfsrv_v2errmap[err - 1]); 2725 return (NFSERR_IO); 2726 } 2727 2728 /* 2729 * Sort the group list in increasing numerical order. 2730 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort 2731 * that used to be here.) 2732 */ 2733 void 2734 nfsrvw_sort(list, num) 2735 gid_t *list; 2736 int num; 2737 { 2738 int i, j; 2739 gid_t v; 2740 2741 /* Insertion sort. */ 2742 for (i = 1; i < num; i++) { 2743 v = list[i]; 2744 /* find correct slot for value v, moving others up */ 2745 for (j = i; --j >= 0 && v < list[j];) 2746 list[j + 1] = list[j]; 2747 list[j + 1] = v; 2748 } 2749 } 2750 2751 /* 2752 * copy credentials making sure that the result can be compared with memcmp(). 2753 */ 2754 void 2755 nfsrv_setcred(incred, outcred) 2756 struct ucred *incred, *outcred; 2757 { 2758 int i; 2759 2760 memset((caddr_t)outcred, 0, sizeof (struct ucred)); 2761 outcred->cr_ref = 1; 2762 outcred->cr_uid = incred->cr_uid; 2763 outcred->cr_gid = incred->cr_gid; 2764 outcred->cr_ngroups = incred->cr_ngroups; 2765 for (i = 0; i < incred->cr_ngroups; i++) 2766 outcred->cr_groups[i] = incred->cr_groups[i]; 2767 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); 2768 } 2769 2770 u_int32_t 2771 nfs_getxid() 2772 { 2773 static u_int32_t base; 2774 static u_int32_t nfs_xid = 0; 2775 static struct simplelock nfs_xidlock = SIMPLELOCK_INITIALIZER; 2776 u_int32_t newxid; 2777 2778 simple_lock(&nfs_xidlock); 2779 /* 2780 * derive initial xid from system time 2781 * XXX time is invalid if root not yet mounted 2782 */ 2783 if (__predict_false(!base && (rootvp))) { 2784 struct timeval tv; 2785 2786 microtime(&tv); 2787 base = tv.tv_sec << 12; 2788 nfs_xid = base; 2789 } 2790 2791 /* 2792 * Skip zero xid if it should ever happen. 2793 */ 2794 if (__predict_false(++nfs_xid == 0)) 2795 nfs_xid++; 2796 newxid = nfs_xid; 2797 simple_unlock(&nfs_xidlock); 2798 2799 return txdr_unsigned(newxid); 2800 } 2801 2802 /* 2803 * assign a new xid for existing request. 2804 * used for NFSERR_JUKEBOX handling. 2805 */ 2806 void 2807 nfs_renewxid(struct nfsreq *req) 2808 { 2809 u_int32_t xid; 2810 int off; 2811 2812 xid = nfs_getxid(); 2813 if (req->r_nmp->nm_sotype == SOCK_STREAM) 2814 off = sizeof(u_int32_t); /* RPC record mark */ 2815 else 2816 off = 0; 2817 2818 m_copyback(req->r_mreq, off, sizeof(xid), (void *)&xid); 2819 req->r_xid = xid; 2820 } 2821