nfs_subs.c revision 1.216
1 /* $NetBSD: nfs_subs.c,v 1.216 2009/03/15 17:20:10 cegger 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 1.216 2009/03/15 17:20:10 cegger Exp $"); 74 75 #ifdef _KERNEL_OPT 76 #include "fs_nfs.h" 77 #include "opt_nfs.h" 78 #endif 79 80 /* 81 * These functions support the macros and help fiddle mbuf chains for 82 * the nfs op functions. They do things like create the rpc header and 83 * copy data between mbuf chains and uio lists. 84 */ 85 #include <sys/param.h> 86 #include <sys/proc.h> 87 #include <sys/systm.h> 88 #include <sys/kernel.h> 89 #include <sys/kmem.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/filedesc.h> 97 #include <sys/time.h> 98 #include <sys/dirent.h> 99 #include <sys/once.h> 100 #include <sys/kauth.h> 101 #include <sys/atomic.h> 102 103 #include <uvm/uvm_extern.h> 104 105 #include <nfs/rpcv2.h> 106 #include <nfs/nfsproto.h> 107 #include <nfs/nfsnode.h> 108 #include <nfs/nfs.h> 109 #include <nfs/xdr_subs.h> 110 #include <nfs/nfsm_subs.h> 111 #include <nfs/nfsmount.h> 112 #include <nfs/nfsrtt.h> 113 #include <nfs/nfs_var.h> 114 115 #include <miscfs/specfs/specdev.h> 116 117 #include <netinet/in.h> 118 119 static u_int32_t nfs_xid; 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, 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 /* NFS client/server stats. */ 144 struct nfsstats nfsstats; 145 146 /* 147 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 148 */ 149 const int nfsv3_procid[NFS_NPROCS] = { 150 NFSPROC_NULL, 151 NFSPROC_GETATTR, 152 NFSPROC_SETATTR, 153 NFSPROC_NOOP, 154 NFSPROC_LOOKUP, 155 NFSPROC_READLINK, 156 NFSPROC_READ, 157 NFSPROC_NOOP, 158 NFSPROC_WRITE, 159 NFSPROC_CREATE, 160 NFSPROC_REMOVE, 161 NFSPROC_RENAME, 162 NFSPROC_LINK, 163 NFSPROC_SYMLINK, 164 NFSPROC_MKDIR, 165 NFSPROC_RMDIR, 166 NFSPROC_READDIR, 167 NFSPROC_FSSTAT, 168 NFSPROC_NOOP, 169 NFSPROC_NOOP, 170 NFSPROC_NOOP, 171 NFSPROC_NOOP, 172 NFSPROC_NOOP 173 }; 174 175 /* 176 * and the reverse mapping from generic to Version 2 procedure numbers 177 */ 178 const int nfsv2_procid[NFS_NPROCS] = { 179 NFSV2PROC_NULL, 180 NFSV2PROC_GETATTR, 181 NFSV2PROC_SETATTR, 182 NFSV2PROC_LOOKUP, 183 NFSV2PROC_NOOP, 184 NFSV2PROC_READLINK, 185 NFSV2PROC_READ, 186 NFSV2PROC_WRITE, 187 NFSV2PROC_CREATE, 188 NFSV2PROC_MKDIR, 189 NFSV2PROC_SYMLINK, 190 NFSV2PROC_CREATE, 191 NFSV2PROC_REMOVE, 192 NFSV2PROC_RMDIR, 193 NFSV2PROC_RENAME, 194 NFSV2PROC_LINK, 195 NFSV2PROC_READDIR, 196 NFSV2PROC_NOOP, 197 NFSV2PROC_STATFS, 198 NFSV2PROC_NOOP, 199 NFSV2PROC_NOOP, 200 NFSV2PROC_NOOP, 201 NFSV2PROC_NOOP, 202 }; 203 204 /* 205 * Maps errno values to nfs error numbers. 206 * Use NFSERR_IO as the catch all for ones not specifically defined in 207 * RFC 1094. 208 */ 209 static const u_char nfsrv_v2errmap[ELAST] = { 210 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 211 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 212 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 213 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 214 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 215 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 216 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 217 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 218 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 219 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 220 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 221 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 222 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, 223 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 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, 227 }; 228 229 /* 230 * Maps errno values to nfs error numbers. 231 * Although it is not obvious whether or not NFS clients really care if 232 * a returned error value is in the specified list for the procedure, the 233 * safest thing to do is filter them appropriately. For Version 2, the 234 * X/Open XNFS document is the only specification that defines error values 235 * for each RPC (The RFC simply lists all possible error values for all RPCs), 236 * so I have decided to not do this for Version 2. 237 * The first entry is the default error return and the rest are the valid 238 * errors for that RPC in increasing numeric order. 239 */ 240 static const short nfsv3err_null[] = { 241 0, 242 0, 243 }; 244 245 static const short nfsv3err_getattr[] = { 246 NFSERR_IO, 247 NFSERR_IO, 248 NFSERR_STALE, 249 NFSERR_BADHANDLE, 250 NFSERR_SERVERFAULT, 251 0, 252 }; 253 254 static const short nfsv3err_setattr[] = { 255 NFSERR_IO, 256 NFSERR_PERM, 257 NFSERR_IO, 258 NFSERR_ACCES, 259 NFSERR_INVAL, 260 NFSERR_NOSPC, 261 NFSERR_ROFS, 262 NFSERR_DQUOT, 263 NFSERR_STALE, 264 NFSERR_BADHANDLE, 265 NFSERR_NOT_SYNC, 266 NFSERR_SERVERFAULT, 267 0, 268 }; 269 270 static const short nfsv3err_lookup[] = { 271 NFSERR_IO, 272 NFSERR_NOENT, 273 NFSERR_IO, 274 NFSERR_ACCES, 275 NFSERR_NOTDIR, 276 NFSERR_NAMETOL, 277 NFSERR_STALE, 278 NFSERR_BADHANDLE, 279 NFSERR_SERVERFAULT, 280 0, 281 }; 282 283 static const short nfsv3err_access[] = { 284 NFSERR_IO, 285 NFSERR_IO, 286 NFSERR_STALE, 287 NFSERR_BADHANDLE, 288 NFSERR_SERVERFAULT, 289 0, 290 }; 291 292 static const short nfsv3err_readlink[] = { 293 NFSERR_IO, 294 NFSERR_IO, 295 NFSERR_ACCES, 296 NFSERR_INVAL, 297 NFSERR_STALE, 298 NFSERR_BADHANDLE, 299 NFSERR_NOTSUPP, 300 NFSERR_SERVERFAULT, 301 0, 302 }; 303 304 static const short nfsv3err_read[] = { 305 NFSERR_IO, 306 NFSERR_IO, 307 NFSERR_NXIO, 308 NFSERR_ACCES, 309 NFSERR_INVAL, 310 NFSERR_STALE, 311 NFSERR_BADHANDLE, 312 NFSERR_SERVERFAULT, 313 NFSERR_JUKEBOX, 314 0, 315 }; 316 317 static const short nfsv3err_write[] = { 318 NFSERR_IO, 319 NFSERR_IO, 320 NFSERR_ACCES, 321 NFSERR_INVAL, 322 NFSERR_FBIG, 323 NFSERR_NOSPC, 324 NFSERR_ROFS, 325 NFSERR_DQUOT, 326 NFSERR_STALE, 327 NFSERR_BADHANDLE, 328 NFSERR_SERVERFAULT, 329 NFSERR_JUKEBOX, 330 0, 331 }; 332 333 static const short nfsv3err_create[] = { 334 NFSERR_IO, 335 NFSERR_IO, 336 NFSERR_ACCES, 337 NFSERR_EXIST, 338 NFSERR_NOTDIR, 339 NFSERR_NOSPC, 340 NFSERR_ROFS, 341 NFSERR_NAMETOL, 342 NFSERR_DQUOT, 343 NFSERR_STALE, 344 NFSERR_BADHANDLE, 345 NFSERR_NOTSUPP, 346 NFSERR_SERVERFAULT, 347 0, 348 }; 349 350 static const short nfsv3err_mkdir[] = { 351 NFSERR_IO, 352 NFSERR_IO, 353 NFSERR_ACCES, 354 NFSERR_EXIST, 355 NFSERR_NOTDIR, 356 NFSERR_NOSPC, 357 NFSERR_ROFS, 358 NFSERR_NAMETOL, 359 NFSERR_DQUOT, 360 NFSERR_STALE, 361 NFSERR_BADHANDLE, 362 NFSERR_NOTSUPP, 363 NFSERR_SERVERFAULT, 364 0, 365 }; 366 367 static const short nfsv3err_symlink[] = { 368 NFSERR_IO, 369 NFSERR_IO, 370 NFSERR_ACCES, 371 NFSERR_EXIST, 372 NFSERR_NOTDIR, 373 NFSERR_NOSPC, 374 NFSERR_ROFS, 375 NFSERR_NAMETOL, 376 NFSERR_DQUOT, 377 NFSERR_STALE, 378 NFSERR_BADHANDLE, 379 NFSERR_NOTSUPP, 380 NFSERR_SERVERFAULT, 381 0, 382 }; 383 384 static const short nfsv3err_mknod[] = { 385 NFSERR_IO, 386 NFSERR_IO, 387 NFSERR_ACCES, 388 NFSERR_EXIST, 389 NFSERR_NOTDIR, 390 NFSERR_NOSPC, 391 NFSERR_ROFS, 392 NFSERR_NAMETOL, 393 NFSERR_DQUOT, 394 NFSERR_STALE, 395 NFSERR_BADHANDLE, 396 NFSERR_NOTSUPP, 397 NFSERR_SERVERFAULT, 398 NFSERR_BADTYPE, 399 0, 400 }; 401 402 static const short nfsv3err_remove[] = { 403 NFSERR_IO, 404 NFSERR_NOENT, 405 NFSERR_IO, 406 NFSERR_ACCES, 407 NFSERR_NOTDIR, 408 NFSERR_ROFS, 409 NFSERR_NAMETOL, 410 NFSERR_STALE, 411 NFSERR_BADHANDLE, 412 NFSERR_SERVERFAULT, 413 0, 414 }; 415 416 static const short nfsv3err_rmdir[] = { 417 NFSERR_IO, 418 NFSERR_NOENT, 419 NFSERR_IO, 420 NFSERR_ACCES, 421 NFSERR_EXIST, 422 NFSERR_NOTDIR, 423 NFSERR_INVAL, 424 NFSERR_ROFS, 425 NFSERR_NAMETOL, 426 NFSERR_NOTEMPTY, 427 NFSERR_STALE, 428 NFSERR_BADHANDLE, 429 NFSERR_NOTSUPP, 430 NFSERR_SERVERFAULT, 431 0, 432 }; 433 434 static const short nfsv3err_rename[] = { 435 NFSERR_IO, 436 NFSERR_NOENT, 437 NFSERR_IO, 438 NFSERR_ACCES, 439 NFSERR_EXIST, 440 NFSERR_XDEV, 441 NFSERR_NOTDIR, 442 NFSERR_ISDIR, 443 NFSERR_INVAL, 444 NFSERR_NOSPC, 445 NFSERR_ROFS, 446 NFSERR_MLINK, 447 NFSERR_NAMETOL, 448 NFSERR_NOTEMPTY, 449 NFSERR_DQUOT, 450 NFSERR_STALE, 451 NFSERR_BADHANDLE, 452 NFSERR_NOTSUPP, 453 NFSERR_SERVERFAULT, 454 0, 455 }; 456 457 static const short nfsv3err_link[] = { 458 NFSERR_IO, 459 NFSERR_IO, 460 NFSERR_ACCES, 461 NFSERR_EXIST, 462 NFSERR_XDEV, 463 NFSERR_NOTDIR, 464 NFSERR_INVAL, 465 NFSERR_NOSPC, 466 NFSERR_ROFS, 467 NFSERR_MLINK, 468 NFSERR_NAMETOL, 469 NFSERR_DQUOT, 470 NFSERR_STALE, 471 NFSERR_BADHANDLE, 472 NFSERR_NOTSUPP, 473 NFSERR_SERVERFAULT, 474 0, 475 }; 476 477 static const short nfsv3err_readdir[] = { 478 NFSERR_IO, 479 NFSERR_IO, 480 NFSERR_ACCES, 481 NFSERR_NOTDIR, 482 NFSERR_STALE, 483 NFSERR_BADHANDLE, 484 NFSERR_BAD_COOKIE, 485 NFSERR_TOOSMALL, 486 NFSERR_SERVERFAULT, 487 0, 488 }; 489 490 static const short nfsv3err_readdirplus[] = { 491 NFSERR_IO, 492 NFSERR_IO, 493 NFSERR_ACCES, 494 NFSERR_NOTDIR, 495 NFSERR_STALE, 496 NFSERR_BADHANDLE, 497 NFSERR_BAD_COOKIE, 498 NFSERR_NOTSUPP, 499 NFSERR_TOOSMALL, 500 NFSERR_SERVERFAULT, 501 0, 502 }; 503 504 static const short nfsv3err_fsstat[] = { 505 NFSERR_IO, 506 NFSERR_IO, 507 NFSERR_STALE, 508 NFSERR_BADHANDLE, 509 NFSERR_SERVERFAULT, 510 0, 511 }; 512 513 static const short nfsv3err_fsinfo[] = { 514 NFSERR_STALE, 515 NFSERR_STALE, 516 NFSERR_BADHANDLE, 517 NFSERR_SERVERFAULT, 518 0, 519 }; 520 521 static const short nfsv3err_pathconf[] = { 522 NFSERR_STALE, 523 NFSERR_STALE, 524 NFSERR_BADHANDLE, 525 NFSERR_SERVERFAULT, 526 0, 527 }; 528 529 static const short nfsv3err_commit[] = { 530 NFSERR_IO, 531 NFSERR_IO, 532 NFSERR_STALE, 533 NFSERR_BADHANDLE, 534 NFSERR_SERVERFAULT, 535 0, 536 }; 537 538 static const short * const nfsrv_v3errmap[] = { 539 nfsv3err_null, 540 nfsv3err_getattr, 541 nfsv3err_setattr, 542 nfsv3err_lookup, 543 nfsv3err_access, 544 nfsv3err_readlink, 545 nfsv3err_read, 546 nfsv3err_write, 547 nfsv3err_create, 548 nfsv3err_mkdir, 549 nfsv3err_symlink, 550 nfsv3err_mknod, 551 nfsv3err_remove, 552 nfsv3err_rmdir, 553 nfsv3err_rename, 554 nfsv3err_link, 555 nfsv3err_readdir, 556 nfsv3err_readdirplus, 557 nfsv3err_fsstat, 558 nfsv3err_fsinfo, 559 nfsv3err_pathconf, 560 nfsv3err_commit, 561 }; 562 563 extern struct nfsrtt nfsrtt; 564 565 u_long nfsdirhashmask; 566 567 int nfs_webnamei(struct nameidata *, struct vnode *, struct proc *); 568 569 /* 570 * Create the header for an rpc request packet 571 * The hsiz is the size of the rest of the nfs request header. 572 * (just used to decide if a cluster is a good idea) 573 */ 574 struct mbuf * 575 nfsm_reqh(struct nfsnode *np, u_long procid, int hsiz, char **bposp) 576 { 577 struct mbuf *mb; 578 char *bpos; 579 580 mb = m_get(M_WAIT, MT_DATA); 581 MCLAIM(mb, &nfs_mowner); 582 if (hsiz >= MINCLSIZE) 583 m_clget(mb, M_WAIT); 584 mb->m_len = 0; 585 bpos = mtod(mb, void *); 586 587 /* Finally, return values */ 588 *bposp = bpos; 589 return (mb); 590 } 591 592 /* 593 * Build the RPC header and fill in the authorization info. 594 * The authorization string argument is only used when the credentials 595 * come from outside of the kernel. 596 * Returns the head of the mbuf list. 597 */ 598 struct mbuf * 599 nfsm_rpchead(kauth_cred_t cr, int nmflag, int procid, 600 int auth_type, int auth_len, char *auth_str, int verf_len, 601 char *verf_str, struct mbuf *mrest, int mrest_len, 602 struct mbuf **mbp, uint32_t *xidp) 603 { 604 struct mbuf *mb; 605 u_int32_t *tl; 606 char *bpos; 607 int i; 608 struct mbuf *mreq; 609 int siz, grpsiz, authsiz; 610 611 authsiz = nfsm_rndup(auth_len); 612 mb = m_gethdr(M_WAIT, MT_DATA); 613 MCLAIM(mb, &nfs_mowner); 614 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 615 m_clget(mb, M_WAIT); 616 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 617 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 618 } else { 619 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 620 } 621 mb->m_len = 0; 622 mreq = mb; 623 bpos = mtod(mb, void *); 624 625 /* 626 * First the RPC header. 627 */ 628 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 629 630 *tl++ = *xidp = nfs_getxid(); 631 *tl++ = rpc_call; 632 *tl++ = rpc_vers; 633 *tl++ = txdr_unsigned(NFS_PROG); 634 if (nmflag & NFSMNT_NFSV3) 635 *tl++ = txdr_unsigned(NFS_VER3); 636 else 637 *tl++ = txdr_unsigned(NFS_VER2); 638 if (nmflag & NFSMNT_NFSV3) 639 *tl++ = txdr_unsigned(procid); 640 else 641 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 642 643 /* 644 * And then the authorization cred. 645 */ 646 *tl++ = txdr_unsigned(auth_type); 647 *tl = txdr_unsigned(authsiz); 648 switch (auth_type) { 649 case RPCAUTH_UNIX: 650 nfsm_build(tl, u_int32_t *, auth_len); 651 *tl++ = 0; /* stamp ?? */ 652 *tl++ = 0; /* NULL hostname */ 653 *tl++ = txdr_unsigned(kauth_cred_geteuid(cr)); 654 *tl++ = txdr_unsigned(kauth_cred_getegid(cr)); 655 grpsiz = (auth_len >> 2) - 5; 656 *tl++ = txdr_unsigned(grpsiz); 657 for (i = 0; i < grpsiz; i++) 658 *tl++ = txdr_unsigned(kauth_cred_group(cr, i)); /* XXX elad review */ 659 break; 660 case RPCAUTH_KERB4: 661 siz = auth_len; 662 while (siz > 0) { 663 if (M_TRAILINGSPACE(mb) == 0) { 664 struct mbuf *mb2; 665 mb2 = m_get(M_WAIT, MT_DATA); 666 MCLAIM(mb2, &nfs_mowner); 667 if (siz >= MINCLSIZE) 668 m_clget(mb2, M_WAIT); 669 mb->m_next = mb2; 670 mb = mb2; 671 mb->m_len = 0; 672 bpos = mtod(mb, void *); 673 } 674 i = min(siz, M_TRAILINGSPACE(mb)); 675 memcpy(bpos, auth_str, i); 676 mb->m_len += i; 677 auth_str += i; 678 bpos += i; 679 siz -= i; 680 } 681 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 682 for (i = 0; i < siz; i++) 683 *bpos++ = '\0'; 684 mb->m_len += siz; 685 } 686 break; 687 }; 688 689 /* 690 * And the verifier... 691 */ 692 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 693 if (verf_str) { 694 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 695 *tl = txdr_unsigned(verf_len); 696 siz = verf_len; 697 while (siz > 0) { 698 if (M_TRAILINGSPACE(mb) == 0) { 699 struct mbuf *mb2; 700 mb2 = m_get(M_WAIT, MT_DATA); 701 MCLAIM(mb2, &nfs_mowner); 702 if (siz >= MINCLSIZE) 703 m_clget(mb2, M_WAIT); 704 mb->m_next = mb2; 705 mb = mb2; 706 mb->m_len = 0; 707 bpos = mtod(mb, void *); 708 } 709 i = min(siz, M_TRAILINGSPACE(mb)); 710 memcpy(bpos, verf_str, i); 711 mb->m_len += i; 712 verf_str += i; 713 bpos += i; 714 siz -= i; 715 } 716 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 717 for (i = 0; i < siz; i++) 718 *bpos++ = '\0'; 719 mb->m_len += siz; 720 } 721 } else { 722 *tl++ = txdr_unsigned(RPCAUTH_NULL); 723 *tl = 0; 724 } 725 mb->m_next = mrest; 726 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 727 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 728 *mbp = mb; 729 return (mreq); 730 } 731 732 /* 733 * copies mbuf chain to the uio scatter/gather list 734 */ 735 int 736 nfsm_mbuftouio(struct mbuf **mrep, struct uio *uiop, int siz, char **dpos) 737 { 738 char *mbufcp, *uiocp; 739 int xfer, left, len; 740 struct mbuf *mp; 741 long uiosiz, rem; 742 int error = 0; 743 744 mp = *mrep; 745 mbufcp = *dpos; 746 len = mtod(mp, char *) + mp->m_len - mbufcp; 747 rem = nfsm_rndup(siz)-siz; 748 while (siz > 0) { 749 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 750 return (EFBIG); 751 left = uiop->uio_iov->iov_len; 752 uiocp = uiop->uio_iov->iov_base; 753 if (left > siz) 754 left = siz; 755 uiosiz = left; 756 while (left > 0) { 757 while (len == 0) { 758 mp = mp->m_next; 759 if (mp == NULL) 760 return (EBADRPC); 761 mbufcp = mtod(mp, void *); 762 len = mp->m_len; 763 } 764 xfer = (left > len) ? len : left; 765 error = copyout_vmspace(uiop->uio_vmspace, mbufcp, 766 uiocp, xfer); 767 if (error) { 768 return error; 769 } 770 left -= xfer; 771 len -= xfer; 772 mbufcp += xfer; 773 uiocp += xfer; 774 uiop->uio_offset += xfer; 775 uiop->uio_resid -= xfer; 776 } 777 if (uiop->uio_iov->iov_len <= siz) { 778 uiop->uio_iovcnt--; 779 uiop->uio_iov++; 780 } else { 781 uiop->uio_iov->iov_base = 782 (char *)uiop->uio_iov->iov_base + uiosiz; 783 uiop->uio_iov->iov_len -= uiosiz; 784 } 785 siz -= uiosiz; 786 } 787 *dpos = mbufcp; 788 *mrep = mp; 789 if (rem > 0) { 790 if (len < rem) 791 error = nfs_adv(mrep, dpos, rem, len); 792 else 793 *dpos += rem; 794 } 795 return (error); 796 } 797 798 /* 799 * copies a uio scatter/gather list to an mbuf chain. 800 * NOTE: can ony handle iovcnt == 1 801 */ 802 int 803 nfsm_uiotombuf(struct uio *uiop, struct mbuf **mq, int siz, char **bpos) 804 { 805 char *uiocp; 806 struct mbuf *mp, *mp2; 807 int xfer, left, mlen; 808 int uiosiz, clflg, rem; 809 char *cp; 810 int error; 811 812 #ifdef DIAGNOSTIC 813 if (uiop->uio_iovcnt != 1) 814 panic("nfsm_uiotombuf: iovcnt != 1"); 815 #endif 816 817 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 818 clflg = 1; 819 else 820 clflg = 0; 821 rem = nfsm_rndup(siz)-siz; 822 mp = mp2 = *mq; 823 while (siz > 0) { 824 left = uiop->uio_iov->iov_len; 825 uiocp = uiop->uio_iov->iov_base; 826 if (left > siz) 827 left = siz; 828 uiosiz = left; 829 while (left > 0) { 830 mlen = M_TRAILINGSPACE(mp); 831 if (mlen == 0) { 832 mp = m_get(M_WAIT, MT_DATA); 833 MCLAIM(mp, &nfs_mowner); 834 if (clflg) 835 m_clget(mp, M_WAIT); 836 mp->m_len = 0; 837 mp2->m_next = mp; 838 mp2 = mp; 839 mlen = M_TRAILINGSPACE(mp); 840 } 841 xfer = (left > mlen) ? mlen : left; 842 cp = mtod(mp, char *) + mp->m_len; 843 error = copyin_vmspace(uiop->uio_vmspace, uiocp, cp, 844 xfer); 845 if (error) { 846 /* XXX */ 847 } 848 mp->m_len += xfer; 849 left -= xfer; 850 uiocp += xfer; 851 uiop->uio_offset += xfer; 852 uiop->uio_resid -= xfer; 853 } 854 uiop->uio_iov->iov_base = (char *)uiop->uio_iov->iov_base + 855 uiosiz; 856 uiop->uio_iov->iov_len -= uiosiz; 857 siz -= uiosiz; 858 } 859 if (rem > 0) { 860 if (rem > M_TRAILINGSPACE(mp)) { 861 mp = m_get(M_WAIT, MT_DATA); 862 MCLAIM(mp, &nfs_mowner); 863 mp->m_len = 0; 864 mp2->m_next = mp; 865 } 866 cp = mtod(mp, char *) + mp->m_len; 867 for (left = 0; left < rem; left++) 868 *cp++ = '\0'; 869 mp->m_len += rem; 870 *bpos = cp; 871 } else 872 *bpos = mtod(mp, char *) + mp->m_len; 873 *mq = mp; 874 return (0); 875 } 876 877 /* 878 * Get at least "siz" bytes of correctly aligned data. 879 * When called the mbuf pointers are not necessarily correct, 880 * dsosp points to what ought to be in m_data and left contains 881 * what ought to be in m_len. 882 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 883 * cases. (The macros use the vars. dpos and dpos2) 884 */ 885 int 886 nfsm_disct(struct mbuf **mdp, char **dposp, int siz, int left, char **cp2) 887 { 888 struct mbuf *m1, *m2; 889 struct mbuf *havebuf = NULL; 890 char *src = *dposp; 891 char *dst; 892 int len; 893 894 #ifdef DEBUG 895 if (left < 0) 896 panic("nfsm_disct: left < 0"); 897 #endif 898 m1 = *mdp; 899 /* 900 * Skip through the mbuf chain looking for an mbuf with 901 * some data. If the first mbuf found has enough data 902 * and it is correctly aligned return it. 903 */ 904 while (left == 0) { 905 havebuf = m1; 906 *mdp = m1 = m1->m_next; 907 if (m1 == NULL) 908 return (EBADRPC); 909 src = mtod(m1, void *); 910 left = m1->m_len; 911 /* 912 * If we start a new mbuf and it is big enough 913 * and correctly aligned just return it, don't 914 * do any pull up. 915 */ 916 if (left >= siz && nfsm_aligned(src)) { 917 *cp2 = src; 918 *dposp = src + siz; 919 return (0); 920 } 921 } 922 if ((m1->m_flags & M_EXT) != 0) { 923 if (havebuf && M_TRAILINGSPACE(havebuf) >= siz && 924 nfsm_aligned(mtod(havebuf, char *) + havebuf->m_len)) { 925 /* 926 * If the first mbuf with data has external data 927 * and there is a previous mbuf with some trailing 928 * space, use it to move the data into. 929 */ 930 m2 = m1; 931 *mdp = m1 = havebuf; 932 *cp2 = mtod(m1, char *) + m1->m_len; 933 } else if (havebuf) { 934 /* 935 * If the first mbuf has a external data 936 * and there is no previous empty mbuf 937 * allocate a new mbuf and move the external 938 * data to the new mbuf. Also make the first 939 * mbuf look empty. 940 */ 941 m2 = m1; 942 *mdp = m1 = m_get(M_WAIT, MT_DATA); 943 MCLAIM(m1, m2->m_owner); 944 if ((m2->m_flags & M_PKTHDR) != 0) { 945 /* XXX MOVE */ 946 M_COPY_PKTHDR(m1, m2); 947 m_tag_delete_chain(m2, NULL); 948 m2->m_flags &= ~M_PKTHDR; 949 } 950 if (havebuf) { 951 havebuf->m_next = m1; 952 } 953 m1->m_next = m2; 954 MRESETDATA(m1); 955 m1->m_len = 0; 956 m2->m_data = src; 957 m2->m_len = left; 958 *cp2 = mtod(m1, char *); 959 } else { 960 struct mbuf **nextp = &m1->m_next; 961 962 m1->m_len -= left; 963 do { 964 m2 = m_get(M_WAIT, MT_DATA); 965 MCLAIM(m2, m1->m_owner); 966 if (left >= MINCLSIZE) { 967 MCLGET(m2, M_WAIT); 968 } 969 m2->m_next = *nextp; 970 *nextp = m2; 971 nextp = &m2->m_next; 972 len = (m2->m_flags & M_EXT) != 0 ? 973 MCLBYTES : MLEN; 974 if (len > left) { 975 len = left; 976 } 977 memcpy(mtod(m2, char *), src, len); 978 m2->m_len = len; 979 src += len; 980 left -= len; 981 } while (left > 0); 982 *mdp = m1 = m1->m_next; 983 m2 = m1->m_next; 984 *cp2 = mtod(m1, char *); 985 } 986 } else { 987 /* 988 * If the first mbuf has no external data 989 * move the data to the front of the mbuf. 990 */ 991 MRESETDATA(m1); 992 dst = mtod(m1, char *); 993 if (dst != src) { 994 memmove(dst, src, left); 995 } 996 m1->m_len = left; 997 m2 = m1->m_next; 998 *cp2 = m1->m_data; 999 } 1000 *dposp = *cp2 + siz; 1001 /* 1002 * Loop through mbufs pulling data up into first mbuf until 1003 * the first mbuf is full or there is no more data to 1004 * pullup. 1005 */ 1006 dst = mtod(m1, char *) + m1->m_len; 1007 while ((len = M_TRAILINGSPACE(m1)) != 0 && m2) { 1008 if ((len = min(len, m2->m_len)) != 0) { 1009 memcpy(dst, mtod(m2, char *), len); 1010 } 1011 m1->m_len += len; 1012 dst += len; 1013 m2->m_data += len; 1014 m2->m_len -= len; 1015 m2 = m2->m_next; 1016 } 1017 if (m1->m_len < siz) 1018 return (EBADRPC); 1019 return (0); 1020 } 1021 1022 /* 1023 * Advance the position in the mbuf chain. 1024 */ 1025 int 1026 nfs_adv(struct mbuf **mdp, char **dposp, int offs, int left) 1027 { 1028 struct mbuf *m; 1029 int s; 1030 1031 m = *mdp; 1032 s = left; 1033 while (s < offs) { 1034 offs -= s; 1035 m = m->m_next; 1036 if (m == NULL) 1037 return (EBADRPC); 1038 s = m->m_len; 1039 } 1040 *mdp = m; 1041 *dposp = mtod(m, char *) + offs; 1042 return (0); 1043 } 1044 1045 /* 1046 * Copy a string into mbufs for the hard cases... 1047 */ 1048 int 1049 nfsm_strtmbuf(struct mbuf **mb, char **bpos, const char *cp, long siz) 1050 { 1051 struct mbuf *m1 = NULL, *m2; 1052 long left, xfer, len, tlen; 1053 u_int32_t *tl; 1054 int putsize; 1055 1056 putsize = 1; 1057 m2 = *mb; 1058 left = M_TRAILINGSPACE(m2); 1059 if (left > 0) { 1060 tl = ((u_int32_t *)(*bpos)); 1061 *tl++ = txdr_unsigned(siz); 1062 putsize = 0; 1063 left -= NFSX_UNSIGNED; 1064 m2->m_len += NFSX_UNSIGNED; 1065 if (left > 0) { 1066 memcpy((void *) tl, cp, left); 1067 siz -= left; 1068 cp += left; 1069 m2->m_len += left; 1070 left = 0; 1071 } 1072 } 1073 /* Loop around adding mbufs */ 1074 while (siz > 0) { 1075 m1 = m_get(M_WAIT, MT_DATA); 1076 MCLAIM(m1, &nfs_mowner); 1077 if (siz > MLEN) 1078 m_clget(m1, M_WAIT); 1079 m1->m_len = NFSMSIZ(m1); 1080 m2->m_next = m1; 1081 m2 = m1; 1082 tl = mtod(m1, u_int32_t *); 1083 tlen = 0; 1084 if (putsize) { 1085 *tl++ = txdr_unsigned(siz); 1086 m1->m_len -= NFSX_UNSIGNED; 1087 tlen = NFSX_UNSIGNED; 1088 putsize = 0; 1089 } 1090 if (siz < m1->m_len) { 1091 len = nfsm_rndup(siz); 1092 xfer = siz; 1093 if (xfer < len) 1094 *(tl+(xfer>>2)) = 0; 1095 } else { 1096 xfer = len = m1->m_len; 1097 } 1098 memcpy((void *) tl, cp, xfer); 1099 m1->m_len = len+tlen; 1100 siz -= xfer; 1101 cp += xfer; 1102 } 1103 *mb = m1; 1104 *bpos = mtod(m1, char *) + m1->m_len; 1105 return (0); 1106 } 1107 1108 /* 1109 * Directory caching routines. They work as follows: 1110 * - a cache is maintained per VDIR nfsnode. 1111 * - for each offset cookie that is exported to userspace, and can 1112 * thus be thrown back at us as an offset to VOP_READDIR, store 1113 * information in the cache. 1114 * - cached are: 1115 * - cookie itself 1116 * - blocknumber (essentially just a search key in the buffer cache) 1117 * - entry number in block. 1118 * - offset cookie of block in which this entry is stored 1119 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1120 * - entries are looked up in a hash table 1121 * - also maintained is an LRU list of entries, used to determine 1122 * which ones to delete if the cache grows too large. 1123 * - if 32 <-> 64 translation mode is requested for a filesystem, 1124 * the cache also functions as a translation table 1125 * - in the translation case, invalidating the cache does not mean 1126 * flushing it, but just marking entries as invalid, except for 1127 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1128 * still be able to use the cache as a translation table. 1129 * - 32 bit cookies are uniquely created by combining the hash table 1130 * entry value, and one generation count per hash table entry, 1131 * incremented each time an entry is appended to the chain. 1132 * - the cache is invalidated each time a direcory is modified 1133 * - sanity checks are also done; if an entry in a block turns 1134 * out not to have a matching cookie, the cache is invalidated 1135 * and a new block starting from the wanted offset is fetched from 1136 * the server. 1137 * - directory entries as read from the server are extended to contain 1138 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1139 * the cache and exporting them to userspace through the cookie 1140 * argument to VOP_READDIR. 1141 */ 1142 1143 u_long 1144 nfs_dirhash(off_t off) 1145 { 1146 int i; 1147 char *cp = (char *)&off; 1148 u_long sum = 0L; 1149 1150 for (i = 0 ; i < sizeof (off); i++) 1151 sum += *cp++; 1152 1153 return sum; 1154 } 1155 1156 #define _NFSDC_MTX(np) (&NFSTOV(np)->v_interlock) 1157 #define NFSDC_LOCK(np) mutex_enter(_NFSDC_MTX(np)) 1158 #define NFSDC_UNLOCK(np) mutex_exit(_NFSDC_MTX(np)) 1159 #define NFSDC_ASSERT_LOCKED(np) KASSERT(mutex_owned(_NFSDC_MTX(np))) 1160 1161 void 1162 nfs_initdircache(struct vnode *vp) 1163 { 1164 struct nfsnode *np = VTONFS(vp); 1165 struct nfsdirhashhead *dircache; 1166 1167 dircache = hashinit(NFS_DIRHASHSIZ, HASH_LIST, true, 1168 &nfsdirhashmask); 1169 1170 NFSDC_LOCK(np); 1171 if (np->n_dircache == NULL) { 1172 np->n_dircachesize = 0; 1173 np->n_dircache = dircache; 1174 dircache = NULL; 1175 TAILQ_INIT(&np->n_dirchain); 1176 } 1177 NFSDC_UNLOCK(np); 1178 if (dircache) 1179 hashdone(dircache, HASH_LIST, nfsdirhashmask); 1180 } 1181 1182 void 1183 nfs_initdirxlatecookie(struct vnode *vp) 1184 { 1185 struct nfsnode *np = VTONFS(vp); 1186 unsigned *dirgens; 1187 1188 KASSERT(VFSTONFS(vp->v_mount)->nm_flag & NFSMNT_XLATECOOKIE); 1189 1190 dirgens = kmem_zalloc(NFS_DIRHASHSIZ * sizeof(unsigned), KM_SLEEP); 1191 NFSDC_LOCK(np); 1192 if (np->n_dirgens == NULL) { 1193 np->n_dirgens = dirgens; 1194 dirgens = NULL; 1195 } 1196 NFSDC_UNLOCK(np); 1197 if (dirgens) 1198 kmem_free(dirgens, NFS_DIRHASHSIZ * sizeof(unsigned)); 1199 } 1200 1201 static const struct nfsdircache dzero; 1202 1203 static void nfs_unlinkdircache(struct nfsnode *np, struct nfsdircache *); 1204 static void nfs_putdircache_unlocked(struct nfsnode *, 1205 struct nfsdircache *); 1206 1207 static void 1208 nfs_unlinkdircache(struct nfsnode *np, struct nfsdircache *ndp) 1209 { 1210 1211 NFSDC_ASSERT_LOCKED(np); 1212 KASSERT(ndp != &dzero); 1213 1214 if (LIST_NEXT(ndp, dc_hash) == (void *)-1) 1215 return; 1216 1217 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1218 LIST_REMOVE(ndp, dc_hash); 1219 LIST_NEXT(ndp, dc_hash) = (void *)-1; /* mark as unlinked */ 1220 1221 nfs_putdircache_unlocked(np, ndp); 1222 } 1223 1224 void 1225 nfs_putdircache(struct nfsnode *np, struct nfsdircache *ndp) 1226 { 1227 int ref; 1228 1229 if (ndp == &dzero) 1230 return; 1231 1232 KASSERT(ndp->dc_refcnt > 0); 1233 NFSDC_LOCK(np); 1234 ref = --ndp->dc_refcnt; 1235 NFSDC_UNLOCK(np); 1236 1237 if (ref == 0) 1238 kmem_free(ndp, sizeof(*ndp)); 1239 } 1240 1241 static void 1242 nfs_putdircache_unlocked(struct nfsnode *np, struct nfsdircache *ndp) 1243 { 1244 int ref; 1245 1246 NFSDC_ASSERT_LOCKED(np); 1247 1248 if (ndp == &dzero) 1249 return; 1250 1251 KASSERT(ndp->dc_refcnt > 0); 1252 ref = --ndp->dc_refcnt; 1253 if (ref == 0) 1254 kmem_free(ndp, sizeof(*ndp)); 1255 } 1256 1257 struct nfsdircache * 1258 nfs_searchdircache(struct vnode *vp, off_t off, int do32, int *hashent) 1259 { 1260 struct nfsdirhashhead *ndhp; 1261 struct nfsdircache *ndp = NULL; 1262 struct nfsnode *np = VTONFS(vp); 1263 unsigned ent; 1264 1265 /* 1266 * Zero is always a valid cookie. 1267 */ 1268 if (off == 0) 1269 /* XXXUNCONST */ 1270 return (struct nfsdircache *)__UNCONST(&dzero); 1271 1272 if (!np->n_dircache) 1273 return NULL; 1274 1275 /* 1276 * We use a 32bit cookie as search key, directly reconstruct 1277 * the hashentry. Else use the hashfunction. 1278 */ 1279 if (do32) { 1280 ent = (u_int32_t)off >> 24; 1281 if (ent >= NFS_DIRHASHSIZ) 1282 return NULL; 1283 ndhp = &np->n_dircache[ent]; 1284 } else { 1285 ndhp = NFSDIRHASH(np, off); 1286 } 1287 1288 if (hashent) 1289 *hashent = (int)(ndhp - np->n_dircache); 1290 1291 NFSDC_LOCK(np); 1292 if (do32) { 1293 LIST_FOREACH(ndp, ndhp, dc_hash) { 1294 if (ndp->dc_cookie32 == (u_int32_t)off) { 1295 /* 1296 * An invalidated entry will become the 1297 * start of a new block fetched from 1298 * the server. 1299 */ 1300 if (ndp->dc_flags & NFSDC_INVALID) { 1301 ndp->dc_blkcookie = ndp->dc_cookie; 1302 ndp->dc_entry = 0; 1303 ndp->dc_flags &= ~NFSDC_INVALID; 1304 } 1305 break; 1306 } 1307 } 1308 } else { 1309 LIST_FOREACH(ndp, ndhp, dc_hash) { 1310 if (ndp->dc_cookie == off) 1311 break; 1312 } 1313 } 1314 if (ndp != NULL) 1315 ndp->dc_refcnt++; 1316 NFSDC_UNLOCK(np); 1317 return ndp; 1318 } 1319 1320 1321 struct nfsdircache * 1322 nfs_enterdircache(struct vnode *vp, off_t off, off_t blkoff, int en, 1323 daddr_t blkno) 1324 { 1325 struct nfsnode *np = VTONFS(vp); 1326 struct nfsdirhashhead *ndhp; 1327 struct nfsdircache *ndp = NULL; 1328 struct nfsdircache *newndp = NULL; 1329 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1330 int hashent = 0, gen, overwrite; /* XXX: GCC */ 1331 1332 /* 1333 * XXX refuse entries for offset 0. amd(8) erroneously sets 1334 * cookie 0 for the '.' entry, making this necessary. This 1335 * isn't so bad, as 0 is a special case anyway. 1336 */ 1337 if (off == 0) 1338 /* XXXUNCONST */ 1339 return (struct nfsdircache *)__UNCONST(&dzero); 1340 1341 if (!np->n_dircache) 1342 /* 1343 * XXX would like to do this in nfs_nget but vtype 1344 * isn't known at that time. 1345 */ 1346 nfs_initdircache(vp); 1347 1348 if ((nmp->nm_flag & NFSMNT_XLATECOOKIE) && !np->n_dirgens) 1349 nfs_initdirxlatecookie(vp); 1350 1351 retry: 1352 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1353 1354 NFSDC_LOCK(np); 1355 if (ndp && (ndp->dc_flags & NFSDC_INVALID) == 0) { 1356 /* 1357 * Overwriting an old entry. Check if it's the same. 1358 * If so, just return. If not, remove the old entry. 1359 */ 1360 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1361 goto done; 1362 nfs_unlinkdircache(np, ndp); 1363 nfs_putdircache_unlocked(np, ndp); 1364 ndp = NULL; 1365 } 1366 1367 ndhp = &np->n_dircache[hashent]; 1368 1369 if (!ndp) { 1370 if (newndp == NULL) { 1371 NFSDC_UNLOCK(np); 1372 newndp = kmem_alloc(sizeof(*newndp), KM_SLEEP); 1373 newndp->dc_refcnt = 1; 1374 LIST_NEXT(newndp, dc_hash) = (void *)-1; 1375 goto retry; 1376 } 1377 ndp = newndp; 1378 newndp = NULL; 1379 overwrite = 0; 1380 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1381 /* 1382 * We're allocating a new entry, so bump the 1383 * generation number. 1384 */ 1385 KASSERT(np->n_dirgens); 1386 gen = ++np->n_dirgens[hashent]; 1387 if (gen == 0) { 1388 np->n_dirgens[hashent]++; 1389 gen++; 1390 } 1391 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1392 } 1393 } else 1394 overwrite = 1; 1395 1396 ndp->dc_cookie = off; 1397 ndp->dc_blkcookie = blkoff; 1398 ndp->dc_entry = en; 1399 ndp->dc_flags = 0; 1400 1401 if (overwrite) 1402 goto done; 1403 1404 /* 1405 * If the maximum directory cookie cache size has been reached 1406 * for this node, take one off the front. The idea is that 1407 * directories are typically read front-to-back once, so that 1408 * the oldest entries can be thrown away without much performance 1409 * loss. 1410 */ 1411 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1412 nfs_unlinkdircache(np, TAILQ_FIRST(&np->n_dirchain)); 1413 } else 1414 np->n_dircachesize++; 1415 1416 KASSERT(ndp->dc_refcnt == 1); 1417 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1418 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1419 ndp->dc_refcnt++; 1420 done: 1421 KASSERT(ndp->dc_refcnt > 0); 1422 NFSDC_UNLOCK(np); 1423 if (newndp) 1424 nfs_putdircache(np, newndp); 1425 return ndp; 1426 } 1427 1428 void 1429 nfs_invaldircache(struct vnode *vp, int flags) 1430 { 1431 struct nfsnode *np = VTONFS(vp); 1432 struct nfsdircache *ndp = NULL; 1433 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1434 const bool forcefree = flags & NFS_INVALDIRCACHE_FORCE; 1435 1436 #ifdef DIAGNOSTIC 1437 if (vp->v_type != VDIR) 1438 panic("nfs: invaldircache: not dir"); 1439 #endif 1440 1441 if ((flags & NFS_INVALDIRCACHE_KEEPEOF) == 0) 1442 np->n_flag &= ~NEOFVALID; 1443 1444 if (!np->n_dircache) 1445 return; 1446 1447 NFSDC_LOCK(np); 1448 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1449 while ((ndp = TAILQ_FIRST(&np->n_dirchain)) != NULL) { 1450 KASSERT(!forcefree || ndp->dc_refcnt == 1); 1451 nfs_unlinkdircache(np, ndp); 1452 } 1453 np->n_dircachesize = 0; 1454 if (forcefree && np->n_dirgens) { 1455 kmem_free(np->n_dirgens, 1456 NFS_DIRHASHSIZ * sizeof(unsigned)); 1457 np->n_dirgens = NULL; 1458 } 1459 } else { 1460 TAILQ_FOREACH(ndp, &np->n_dirchain, dc_chain) 1461 ndp->dc_flags |= NFSDC_INVALID; 1462 } 1463 1464 NFSDC_UNLOCK(np); 1465 } 1466 1467 /* 1468 * Called once before VFS init to initialize shared and 1469 * server-specific data structures. 1470 */ 1471 static int 1472 nfs_init0(void) 1473 { 1474 1475 nfsrtt.pos = 0; 1476 rpc_vers = txdr_unsigned(RPC_VER2); 1477 rpc_call = txdr_unsigned(RPC_CALL); 1478 rpc_reply = txdr_unsigned(RPC_REPLY); 1479 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1480 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1481 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1482 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1483 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1484 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1485 nfs_prog = txdr_unsigned(NFS_PROG); 1486 nfs_true = txdr_unsigned(true); 1487 nfs_false = txdr_unsigned(false); 1488 nfs_xdrneg1 = txdr_unsigned(-1); 1489 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1490 if (nfs_ticks < 1) 1491 nfs_ticks = 1; 1492 nfs_xid = arc4random(); 1493 nfsdreq_init(); 1494 1495 /* 1496 * Initialize reply list and start timer 1497 */ 1498 TAILQ_INIT(&nfs_reqq); 1499 nfs_timer_init(); 1500 MOWNER_ATTACH(&nfs_mowner); 1501 1502 #ifdef NFS 1503 /* Initialize the kqueue structures */ 1504 nfs_kqinit(); 1505 /* Initialize the iod structures */ 1506 nfs_iodinit(); 1507 #endif 1508 1509 return 0; 1510 } 1511 1512 /* 1513 * This is disgusting, but it must support both modular and monolothic 1514 * configurations. For monolithic builds NFSSERVER may not imply NFS. 1515 * 1516 * Yuck. 1517 */ 1518 void 1519 nfs_init(void) 1520 { 1521 static ONCE_DECL(nfs_init_once); 1522 1523 RUN_ONCE(&nfs_init_once, nfs_init0); 1524 } 1525 1526 void 1527 nfs_fini(void) 1528 { 1529 1530 #ifdef NFS 1531 nfs_kqfini(); 1532 nfs_iodfini(); 1533 #endif 1534 nfsdreq_fini(); 1535 nfs_timer_fini(); 1536 MOWNER_DETACH(&nfs_mowner); 1537 } 1538 1539 #ifdef NFS 1540 /* 1541 * Called once at VFS init to initialize client-specific data structures. 1542 */ 1543 void 1544 nfs_vfs_init(void) 1545 { 1546 1547 /* Initialize NFS server / client shared data. */ 1548 nfs_init(); 1549 nfs_node_init(); 1550 1551 nfs_commitsize = uvmexp.npages << (PAGE_SHIFT - 4); 1552 } 1553 1554 void 1555 nfs_vfs_done(void) 1556 { 1557 1558 nfs_node_done(); 1559 } 1560 1561 /* 1562 * Attribute cache routines. 1563 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1564 * that are on the mbuf list 1565 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1566 * error otherwise 1567 */ 1568 1569 /* 1570 * Load the attribute cache (that lives in the nfsnode entry) with 1571 * the values on the mbuf list and 1572 * Iff vap not NULL 1573 * copy the attributes to *vaper 1574 */ 1575 int 1576 nfsm_loadattrcache(struct vnode **vpp, struct mbuf **mdp, char **dposp, struct vattr *vaper, int flags) 1577 { 1578 int32_t t1; 1579 char *cp2; 1580 int error = 0; 1581 struct mbuf *md; 1582 int v3 = NFS_ISV3(*vpp); 1583 1584 md = *mdp; 1585 t1 = (mtod(md, char *) + md->m_len) - *dposp; 1586 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1587 if (error) 1588 return (error); 1589 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper, flags); 1590 } 1591 1592 int 1593 nfs_loadattrcache(struct vnode **vpp, struct nfs_fattr *fp, struct vattr *vaper, int flags) 1594 { 1595 struct vnode *vp = *vpp; 1596 struct vattr *vap; 1597 int v3 = NFS_ISV3(vp); 1598 enum vtype vtyp; 1599 u_short vmode; 1600 struct timespec mtime; 1601 struct timespec ctime; 1602 int32_t rdev; 1603 struct nfsnode *np; 1604 extern int (**spec_nfsv2nodeop_p)(void *); 1605 uid_t uid; 1606 gid_t gid; 1607 1608 if (v3) { 1609 vtyp = nfsv3tov_type(fp->fa_type); 1610 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1611 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1612 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1613 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1614 fxdr_nfsv3time(&fp->fa3_ctime, &ctime); 1615 } else { 1616 vtyp = nfsv2tov_type(fp->fa_type); 1617 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1618 if (vtyp == VNON || vtyp == VREG) 1619 vtyp = IFTOVT(vmode); 1620 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1621 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1622 ctime.tv_sec = fxdr_unsigned(u_int32_t, 1623 fp->fa2_ctime.nfsv2_sec); 1624 ctime.tv_nsec = 0; 1625 1626 /* 1627 * Really ugly NFSv2 kludge. 1628 */ 1629 if (vtyp == VCHR && rdev == 0xffffffff) 1630 vtyp = VFIFO; 1631 } 1632 1633 vmode &= ALLPERMS; 1634 1635 /* 1636 * If v_type == VNON it is a new node, so fill in the v_type, 1637 * n_mtime fields. Check to see if it represents a special 1638 * device, and if so, check for a possible alias. Once the 1639 * correct vnode has been obtained, fill in the rest of the 1640 * information. 1641 */ 1642 np = VTONFS(vp); 1643 if (vp->v_type == VNON) { 1644 vp->v_type = vtyp; 1645 if (vp->v_type == VFIFO) { 1646 extern int (**fifo_nfsv2nodeop_p)(void *); 1647 vp->v_op = fifo_nfsv2nodeop_p; 1648 } else if (vp->v_type == VREG) { 1649 mutex_init(&np->n_commitlock, MUTEX_DEFAULT, IPL_NONE); 1650 } else if (vp->v_type == VCHR || vp->v_type == VBLK) { 1651 vp->v_op = spec_nfsv2nodeop_p; 1652 spec_node_init(vp, (dev_t)rdev); 1653 } 1654 np->n_mtime = mtime; 1655 } 1656 uid = fxdr_unsigned(uid_t, fp->fa_uid); 1657 gid = fxdr_unsigned(gid_t, fp->fa_gid); 1658 vap = np->n_vattr; 1659 1660 /* 1661 * Invalidate access cache if uid, gid, mode or ctime changed. 1662 */ 1663 if (np->n_accstamp != -1 && 1664 (gid != vap->va_gid || uid != vap->va_uid || vmode != vap->va_mode 1665 || timespeccmp(&ctime, &vap->va_ctime, !=))) 1666 np->n_accstamp = -1; 1667 1668 vap->va_type = vtyp; 1669 vap->va_mode = vmode; 1670 vap->va_rdev = (dev_t)rdev; 1671 vap->va_mtime = mtime; 1672 vap->va_ctime = ctime; 1673 vap->va_birthtime.tv_sec = VNOVAL; 1674 vap->va_birthtime.tv_nsec = VNOVAL; 1675 vap->va_fsid = vp->v_mount->mnt_stat.f_fsidx.__fsid_val[0]; 1676 switch (vtyp) { 1677 case VDIR: 1678 vap->va_blocksize = NFS_DIRFRAGSIZ; 1679 break; 1680 case VBLK: 1681 vap->va_blocksize = BLKDEV_IOSIZE; 1682 break; 1683 case VCHR: 1684 vap->va_blocksize = MAXBSIZE; 1685 break; 1686 default: 1687 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1688 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1689 break; 1690 } 1691 if (v3) { 1692 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1693 vap->va_uid = uid; 1694 vap->va_gid = gid; 1695 vap->va_size = fxdr_hyper(&fp->fa3_size); 1696 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1697 vap->va_fileid = fxdr_hyper(&fp->fa3_fileid); 1698 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1699 vap->va_flags = 0; 1700 vap->va_filerev = 0; 1701 } else { 1702 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1703 vap->va_uid = uid; 1704 vap->va_gid = gid; 1705 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1706 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1707 * NFS_FABLKSIZE; 1708 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1709 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1710 vap->va_flags = 0; 1711 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1712 vap->va_filerev = 0; 1713 } 1714 if (vap->va_size > VFSTONFS(vp->v_mount)->nm_maxfilesize) { 1715 return EFBIG; 1716 } 1717 if (vap->va_size != np->n_size) { 1718 if ((np->n_flag & NMODIFIED) && vap->va_size < np->n_size) { 1719 vap->va_size = np->n_size; 1720 } else { 1721 np->n_size = vap->va_size; 1722 if (vap->va_type == VREG) { 1723 /* 1724 * we can't free pages if NAC_NOTRUNC because 1725 * the pages can be owned by ourselves. 1726 */ 1727 if (flags & NAC_NOTRUNC) { 1728 np->n_flag |= NTRUNCDELAYED; 1729 } else { 1730 genfs_node_wrlock(vp); 1731 mutex_enter(&vp->v_interlock); 1732 (void)VOP_PUTPAGES(vp, 0, 1733 0, PGO_SYNCIO | PGO_CLEANIT | 1734 PGO_FREE | PGO_ALLPAGES); 1735 uvm_vnp_setsize(vp, np->n_size); 1736 genfs_node_unlock(vp); 1737 } 1738 } 1739 } 1740 } 1741 np->n_attrstamp = time_second; 1742 if (vaper != NULL) { 1743 memcpy((void *)vaper, (void *)vap, sizeof(*vap)); 1744 if (np->n_flag & NCHG) { 1745 if (np->n_flag & NACC) 1746 vaper->va_atime = np->n_atim; 1747 if (np->n_flag & NUPD) 1748 vaper->va_mtime = np->n_mtim; 1749 } 1750 } 1751 return (0); 1752 } 1753 1754 /* 1755 * Check the time stamp 1756 * If the cache is valid, copy contents to *vap and return 0 1757 * otherwise return an error 1758 */ 1759 int 1760 nfs_getattrcache(struct vnode *vp, struct vattr *vaper) 1761 { 1762 struct nfsnode *np = VTONFS(vp); 1763 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1764 struct vattr *vap; 1765 1766 if (np->n_attrstamp == 0 || 1767 (time_second - np->n_attrstamp) >= nfs_attrtimeo(nmp, 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) != 0 && 1776 vap->va_size < np->n_size) { 1777 vap->va_size = np->n_size; 1778 } else { 1779 np->n_size = vap->va_size; 1780 } 1781 genfs_node_wrlock(vp); 1782 uvm_vnp_setsize(vp, np->n_size); 1783 genfs_node_unlock(vp); 1784 } else 1785 np->n_size = vap->va_size; 1786 } 1787 memcpy((void *)vaper, (void *)vap, sizeof(struct vattr)); 1788 if (np->n_flag & NCHG) { 1789 if (np->n_flag & NACC) 1790 vaper->va_atime = np->n_atim; 1791 if (np->n_flag & NUPD) 1792 vaper->va_mtime = np->n_mtim; 1793 } 1794 return (0); 1795 } 1796 1797 void 1798 nfs_delayedtruncate(struct vnode *vp) 1799 { 1800 struct nfsnode *np = VTONFS(vp); 1801 1802 if (np->n_flag & NTRUNCDELAYED) { 1803 np->n_flag &= ~NTRUNCDELAYED; 1804 genfs_node_wrlock(vp); 1805 mutex_enter(&vp->v_interlock); 1806 (void)VOP_PUTPAGES(vp, 0, 1807 0, PGO_SYNCIO | PGO_CLEANIT | PGO_FREE | PGO_ALLPAGES); 1808 uvm_vnp_setsize(vp, np->n_size); 1809 genfs_node_unlock(vp); 1810 } 1811 } 1812 1813 #define NFS_WCCKLUDGE_TIMEOUT (24 * 60 * 60) /* 1 day */ 1814 #define NFS_WCCKLUDGE(nmp, now) \ 1815 (((nmp)->nm_iflag & NFSMNT_WCCKLUDGE) && \ 1816 ((now) - (nmp)->nm_wcckludgetime - NFS_WCCKLUDGE_TIMEOUT) < 0) 1817 1818 /* 1819 * nfs_check_wccdata: check inaccurate wcc_data 1820 * 1821 * => return non-zero if we shouldn't trust the wcc_data. 1822 * => NFS_WCCKLUDGE_TIMEOUT is for the case that the server is "fixed". 1823 */ 1824 1825 int 1826 nfs_check_wccdata(struct nfsnode *np, const struct timespec *ctime, 1827 struct timespec *mtime, bool docheck) 1828 { 1829 int error = 0; 1830 1831 #if !defined(NFS_V2_ONLY) 1832 1833 if (docheck) { 1834 struct vnode *vp = NFSTOV(np); 1835 struct nfsmount *nmp; 1836 long now = time_second; 1837 const struct timespec *omtime = &np->n_vattr->va_mtime; 1838 const struct timespec *octime = &np->n_vattr->va_ctime; 1839 const char *reason = NULL; /* XXX: gcc */ 1840 1841 if (timespeccmp(omtime, mtime, <=)) { 1842 reason = "mtime"; 1843 error = EINVAL; 1844 } 1845 1846 if (vp->v_type == VDIR && timespeccmp(octime, ctime, <=)) { 1847 reason = "ctime"; 1848 error = EINVAL; 1849 } 1850 1851 nmp = VFSTONFS(vp->v_mount); 1852 if (error) { 1853 1854 /* 1855 * despite of the fact that we've updated the file, 1856 * timestamps of the file were not updated as we 1857 * expected. 1858 * it means that the server has incompatible 1859 * semantics of timestamps or (more likely) 1860 * the server time is not precise enough to 1861 * track each modifications. 1862 * in that case, we disable wcc processing. 1863 * 1864 * yes, strictly speaking, we should disable all 1865 * caching. it's a compromise. 1866 */ 1867 1868 mutex_enter(&nmp->nm_lock); 1869 if (!NFS_WCCKLUDGE(nmp, now)) { 1870 printf("%s: inaccurate wcc data (%s) detected," 1871 " disabling wcc" 1872 " (ctime %u.%09u %u.%09u," 1873 " mtime %u.%09u %u.%09u)\n", 1874 vp->v_mount->mnt_stat.f_mntfromname, 1875 reason, 1876 (unsigned int)octime->tv_sec, 1877 (unsigned int)octime->tv_nsec, 1878 (unsigned int)ctime->tv_sec, 1879 (unsigned int)ctime->tv_nsec, 1880 (unsigned int)omtime->tv_sec, 1881 (unsigned int)omtime->tv_nsec, 1882 (unsigned int)mtime->tv_sec, 1883 (unsigned int)mtime->tv_nsec); 1884 } 1885 nmp->nm_iflag |= NFSMNT_WCCKLUDGE; 1886 nmp->nm_wcckludgetime = now; 1887 mutex_exit(&nmp->nm_lock); 1888 } else if (NFS_WCCKLUDGE(nmp, now)) { 1889 error = EPERM; /* XXX */ 1890 } else if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 1891 mutex_enter(&nmp->nm_lock); 1892 if (nmp->nm_iflag & NFSMNT_WCCKLUDGE) { 1893 printf("%s: re-enabling wcc\n", 1894 vp->v_mount->mnt_stat.f_mntfromname); 1895 nmp->nm_iflag &= ~NFSMNT_WCCKLUDGE; 1896 } 1897 mutex_exit(&nmp->nm_lock); 1898 } 1899 } 1900 1901 #endif /* !defined(NFS_V2_ONLY) */ 1902 1903 return error; 1904 } 1905 1906 /* 1907 * Heuristic to see if the server XDR encodes directory cookies or not. 1908 * it is not supposed to, but a lot of servers may do this. Also, since 1909 * most/all servers will implement V2 as well, it is expected that they 1910 * may return just 32 bits worth of cookie information, so we need to 1911 * find out in which 32 bits this information is available. We do this 1912 * to avoid trouble with emulated binaries that can't handle 64 bit 1913 * directory offsets. 1914 */ 1915 1916 void 1917 nfs_cookieheuristic(struct vnode *vp, int *flagp, struct lwp *l, kauth_cred_t cred) 1918 { 1919 struct uio auio; 1920 struct iovec aiov; 1921 char *tbuf, *cp; 1922 struct dirent *dp; 1923 off_t *cookies = NULL, *cop; 1924 int error, eof, nc, len; 1925 1926 tbuf = malloc(NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 1927 1928 aiov.iov_base = tbuf; 1929 aiov.iov_len = NFS_DIRFRAGSIZ; 1930 auio.uio_iov = &aiov; 1931 auio.uio_iovcnt = 1; 1932 auio.uio_rw = UIO_READ; 1933 auio.uio_resid = NFS_DIRFRAGSIZ; 1934 auio.uio_offset = 0; 1935 UIO_SETUP_SYSSPACE(&auio); 1936 1937 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 1938 1939 len = NFS_DIRFRAGSIZ - auio.uio_resid; 1940 if (error || len == 0) { 1941 free(tbuf, M_TEMP); 1942 if (cookies) 1943 free(cookies, M_TEMP); 1944 return; 1945 } 1946 1947 /* 1948 * Find the first valid entry and look at its offset cookie. 1949 */ 1950 1951 cp = tbuf; 1952 for (cop = cookies; len > 0; len -= dp->d_reclen) { 1953 dp = (struct dirent *)cp; 1954 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 1955 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 1956 *flagp |= NFSMNT_SWAPCOOKIE; 1957 nfs_invaldircache(vp, 0); 1958 nfs_vinvalbuf(vp, 0, cred, l, 1); 1959 } 1960 break; 1961 } 1962 cop++; 1963 cp += dp->d_reclen; 1964 } 1965 1966 free(tbuf, M_TEMP); 1967 free(cookies, M_TEMP); 1968 } 1969 #endif /* NFS */ 1970 1971 /* 1972 * A fiddled version of m_adj() that ensures null fill to a 32-bit 1973 * boundary and only trims off the back end 1974 * 1975 * 1. trim off 'len' bytes as m_adj(mp, -len). 1976 * 2. add zero-padding 'nul' bytes at the end of the mbuf chain. 1977 */ 1978 void 1979 nfs_zeropad(struct mbuf *mp, int len, int nul) 1980 { 1981 struct mbuf *m; 1982 int count; 1983 1984 /* 1985 * Trim from tail. Scan the mbuf chain, 1986 * calculating its length and finding the last mbuf. 1987 * If the adjustment only affects this mbuf, then just 1988 * adjust and return. Otherwise, rescan and truncate 1989 * after the remaining size. 1990 */ 1991 count = 0; 1992 m = mp; 1993 for (;;) { 1994 count += m->m_len; 1995 if (m->m_next == NULL) 1996 break; 1997 m = m->m_next; 1998 } 1999 2000 KDASSERT(count >= len); 2001 2002 if (m->m_len >= len) { 2003 m->m_len -= len; 2004 } else { 2005 count -= len; 2006 /* 2007 * Correct length for chain is "count". 2008 * Find the mbuf with last data, adjust its length, 2009 * and toss data from remaining mbufs on chain. 2010 */ 2011 for (m = mp; m; m = m->m_next) { 2012 if (m->m_len >= count) { 2013 m->m_len = count; 2014 break; 2015 } 2016 count -= m->m_len; 2017 } 2018 KASSERT(m && m->m_next); 2019 m_freem(m->m_next); 2020 m->m_next = NULL; 2021 } 2022 2023 KDASSERT(m->m_next == NULL); 2024 2025 /* 2026 * zero-padding. 2027 */ 2028 if (nul > 0) { 2029 char *cp; 2030 int i; 2031 2032 if (M_ROMAP(m) || M_TRAILINGSPACE(m) < nul) { 2033 struct mbuf *n; 2034 2035 KDASSERT(MLEN >= nul); 2036 n = m_get(M_WAIT, MT_DATA); 2037 MCLAIM(n, &nfs_mowner); 2038 n->m_len = nul; 2039 n->m_next = NULL; 2040 m->m_next = n; 2041 cp = mtod(n, void *); 2042 } else { 2043 cp = mtod(m, char *) + m->m_len; 2044 m->m_len += nul; 2045 } 2046 for (i = 0; i < nul; i++) 2047 *cp++ = '\0'; 2048 } 2049 return; 2050 } 2051 2052 /* 2053 * Make these functions instead of macros, so that the kernel text size 2054 * doesn't get too big... 2055 */ 2056 void 2057 nfsm_srvwcc(struct nfsrv_descript *nfsd, int before_ret, struct vattr *before_vap, int after_ret, struct vattr *after_vap, struct mbuf **mbp, char **bposp) 2058 { 2059 struct mbuf *mb = *mbp; 2060 char *bpos = *bposp; 2061 u_int32_t *tl; 2062 2063 if (before_ret) { 2064 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2065 *tl = nfs_false; 2066 } else { 2067 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2068 *tl++ = nfs_true; 2069 txdr_hyper(before_vap->va_size, tl); 2070 tl += 2; 2071 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2072 tl += 2; 2073 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2074 } 2075 *bposp = bpos; 2076 *mbp = mb; 2077 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2078 } 2079 2080 void 2081 nfsm_srvpostopattr(struct nfsrv_descript *nfsd, int after_ret, struct vattr *after_vap, struct mbuf **mbp, char **bposp) 2082 { 2083 struct mbuf *mb = *mbp; 2084 char *bpos = *bposp; 2085 u_int32_t *tl; 2086 struct nfs_fattr *fp; 2087 2088 if (after_ret) { 2089 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2090 *tl = nfs_false; 2091 } else { 2092 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2093 *tl++ = nfs_true; 2094 fp = (struct nfs_fattr *)tl; 2095 nfsm_srvfattr(nfsd, after_vap, fp); 2096 } 2097 *mbp = mb; 2098 *bposp = bpos; 2099 } 2100 2101 void 2102 nfsm_srvfattr(struct nfsrv_descript *nfsd, struct vattr *vap, struct nfs_fattr *fp) 2103 { 2104 2105 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2106 fp->fa_uid = txdr_unsigned(vap->va_uid); 2107 fp->fa_gid = txdr_unsigned(vap->va_gid); 2108 if (nfsd->nd_flag & ND_NFSV3) { 2109 fp->fa_type = vtonfsv3_type(vap->va_type); 2110 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2111 txdr_hyper(vap->va_size, &fp->fa3_size); 2112 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2113 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2114 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2115 fp->fa3_fsid.nfsuquad[0] = 0; 2116 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2117 txdr_hyper(vap->va_fileid, &fp->fa3_fileid); 2118 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2119 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2120 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2121 } else { 2122 fp->fa_type = vtonfsv2_type(vap->va_type); 2123 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2124 fp->fa2_size = txdr_unsigned(vap->va_size); 2125 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2126 if (vap->va_type == VFIFO) 2127 fp->fa2_rdev = 0xffffffff; 2128 else 2129 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2130 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2131 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2132 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2133 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2134 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2135 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2136 } 2137 } 2138 2139 /* 2140 * This function compares two net addresses by family and returns true 2141 * if they are the same host. 2142 * If there is any doubt, return false. 2143 * The AF_INET family is handled as a special case so that address mbufs 2144 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2145 */ 2146 int 2147 netaddr_match(int family, union nethostaddr *haddr, struct mbuf *nam) 2148 { 2149 struct sockaddr_in *inetaddr; 2150 2151 switch (family) { 2152 case AF_INET: 2153 inetaddr = mtod(nam, struct sockaddr_in *); 2154 if (inetaddr->sin_family == AF_INET && 2155 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2156 return (1); 2157 break; 2158 case AF_INET6: 2159 { 2160 struct sockaddr_in6 *sin6_1, *sin6_2; 2161 2162 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2163 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2164 if (sin6_1->sin6_family == AF_INET6 && 2165 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2166 return 1; 2167 } 2168 default: 2169 break; 2170 }; 2171 return (0); 2172 } 2173 2174 /* 2175 * The write verifier has changed (probably due to a server reboot), so all 2176 * PG_NEEDCOMMIT pages will have to be written again. Since they are marked 2177 * as dirty or are being written out just now, all this takes is clearing 2178 * the PG_NEEDCOMMIT flag. Once done the new write verifier can be set for 2179 * the mount point. 2180 */ 2181 void 2182 nfs_clearcommit(struct mount *mp) 2183 { 2184 struct vnode *vp; 2185 struct nfsnode *np; 2186 struct vm_page *pg; 2187 struct nfsmount *nmp = VFSTONFS(mp); 2188 2189 rw_enter(&nmp->nm_writeverflock, RW_WRITER); 2190 mutex_enter(&mntvnode_lock); 2191 TAILQ_FOREACH(vp, &mp->mnt_vnodelist, v_mntvnodes) { 2192 KASSERT(vp->v_mount == mp); 2193 if (vp->v_type != VREG) 2194 continue; 2195 np = VTONFS(vp); 2196 np->n_pushlo = np->n_pushhi = np->n_pushedlo = 2197 np->n_pushedhi = 0; 2198 np->n_commitflags &= 2199 ~(NFS_COMMIT_PUSH_VALID | NFS_COMMIT_PUSHED_VALID); 2200 mutex_enter(&vp->v_uobj.vmobjlock); 2201 TAILQ_FOREACH(pg, &vp->v_uobj.memq, listq.queue) { 2202 pg->flags &= ~PG_NEEDCOMMIT; 2203 } 2204 mutex_exit(&vp->v_uobj.vmobjlock); 2205 } 2206 mutex_exit(&mntvnode_lock); 2207 mutex_enter(&nmp->nm_lock); 2208 nmp->nm_iflag &= ~NFSMNT_STALEWRITEVERF; 2209 mutex_exit(&nmp->nm_lock); 2210 rw_exit(&nmp->nm_writeverflock); 2211 } 2212 2213 void 2214 nfs_merge_commit_ranges(struct vnode *vp) 2215 { 2216 struct nfsnode *np = VTONFS(vp); 2217 2218 KASSERT(np->n_commitflags & NFS_COMMIT_PUSH_VALID); 2219 2220 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2221 np->n_pushedlo = np->n_pushlo; 2222 np->n_pushedhi = np->n_pushhi; 2223 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2224 } else { 2225 if (np->n_pushlo < np->n_pushedlo) 2226 np->n_pushedlo = np->n_pushlo; 2227 if (np->n_pushhi > np->n_pushedhi) 2228 np->n_pushedhi = np->n_pushhi; 2229 } 2230 2231 np->n_pushlo = np->n_pushhi = 0; 2232 np->n_commitflags &= ~NFS_COMMIT_PUSH_VALID; 2233 2234 #ifdef NFS_DEBUG_COMMIT 2235 printf("merge: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2236 (unsigned)np->n_pushedhi); 2237 #endif 2238 } 2239 2240 int 2241 nfs_in_committed_range(struct vnode *vp, off_t off, off_t len) 2242 { 2243 struct nfsnode *np = VTONFS(vp); 2244 off_t lo, hi; 2245 2246 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2247 return 0; 2248 lo = off; 2249 hi = lo + len; 2250 2251 return (lo >= np->n_pushedlo && hi <= np->n_pushedhi); 2252 } 2253 2254 int 2255 nfs_in_tobecommitted_range(struct vnode *vp, off_t off, off_t len) 2256 { 2257 struct nfsnode *np = VTONFS(vp); 2258 off_t lo, hi; 2259 2260 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2261 return 0; 2262 lo = off; 2263 hi = lo + len; 2264 2265 return (lo >= np->n_pushlo && hi <= np->n_pushhi); 2266 } 2267 2268 void 2269 nfs_add_committed_range(struct vnode *vp, off_t off, off_t len) 2270 { 2271 struct nfsnode *np = VTONFS(vp); 2272 off_t lo, hi; 2273 2274 lo = off; 2275 hi = lo + len; 2276 2277 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) { 2278 np->n_pushedlo = lo; 2279 np->n_pushedhi = hi; 2280 np->n_commitflags |= NFS_COMMIT_PUSHED_VALID; 2281 } else { 2282 if (hi > np->n_pushedhi) 2283 np->n_pushedhi = hi; 2284 if (lo < np->n_pushedlo) 2285 np->n_pushedlo = lo; 2286 } 2287 #ifdef NFS_DEBUG_COMMIT 2288 printf("add: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2289 (unsigned)np->n_pushedhi); 2290 #endif 2291 } 2292 2293 void 2294 nfs_del_committed_range(struct vnode *vp, off_t off, off_t len) 2295 { 2296 struct nfsnode *np = VTONFS(vp); 2297 off_t lo, hi; 2298 2299 if (!(np->n_commitflags & NFS_COMMIT_PUSHED_VALID)) 2300 return; 2301 2302 lo = off; 2303 hi = lo + len; 2304 2305 if (lo > np->n_pushedhi || hi < np->n_pushedlo) 2306 return; 2307 if (lo <= np->n_pushedlo) 2308 np->n_pushedlo = hi; 2309 else if (hi >= np->n_pushedhi) 2310 np->n_pushedhi = lo; 2311 else { 2312 /* 2313 * XXX There's only one range. If the deleted range 2314 * is in the middle, pick the largest of the 2315 * contiguous ranges that it leaves. 2316 */ 2317 if ((np->n_pushedlo - lo) > (hi - np->n_pushedhi)) 2318 np->n_pushedhi = lo; 2319 else 2320 np->n_pushedlo = hi; 2321 } 2322 #ifdef NFS_DEBUG_COMMIT 2323 printf("del: committed: %u - %u\n", (unsigned)np->n_pushedlo, 2324 (unsigned)np->n_pushedhi); 2325 #endif 2326 } 2327 2328 void 2329 nfs_add_tobecommitted_range(struct vnode *vp, off_t off, off_t len) 2330 { 2331 struct nfsnode *np = VTONFS(vp); 2332 off_t lo, hi; 2333 2334 lo = off; 2335 hi = lo + len; 2336 2337 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) { 2338 np->n_pushlo = lo; 2339 np->n_pushhi = hi; 2340 np->n_commitflags |= NFS_COMMIT_PUSH_VALID; 2341 } else { 2342 if (lo < np->n_pushlo) 2343 np->n_pushlo = lo; 2344 if (hi > np->n_pushhi) 2345 np->n_pushhi = hi; 2346 } 2347 #ifdef NFS_DEBUG_COMMIT 2348 printf("add: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2349 (unsigned)np->n_pushhi); 2350 #endif 2351 } 2352 2353 void 2354 nfs_del_tobecommitted_range(struct vnode *vp, off_t off, off_t len) 2355 { 2356 struct nfsnode *np = VTONFS(vp); 2357 off_t lo, hi; 2358 2359 if (!(np->n_commitflags & NFS_COMMIT_PUSH_VALID)) 2360 return; 2361 2362 lo = off; 2363 hi = lo + len; 2364 2365 if (lo > np->n_pushhi || hi < np->n_pushlo) 2366 return; 2367 2368 if (lo <= np->n_pushlo) 2369 np->n_pushlo = hi; 2370 else if (hi >= np->n_pushhi) 2371 np->n_pushhi = lo; 2372 else { 2373 /* 2374 * XXX There's only one range. If the deleted range 2375 * is in the middle, pick the largest of the 2376 * contiguous ranges that it leaves. 2377 */ 2378 if ((np->n_pushlo - lo) > (hi - np->n_pushhi)) 2379 np->n_pushhi = lo; 2380 else 2381 np->n_pushlo = hi; 2382 } 2383 #ifdef NFS_DEBUG_COMMIT 2384 printf("del: tobecommitted: %u - %u\n", (unsigned)np->n_pushlo, 2385 (unsigned)np->n_pushhi); 2386 #endif 2387 } 2388 2389 /* 2390 * Map errnos to NFS error numbers. For Version 3 also filter out error 2391 * numbers not specified for the associated procedure. 2392 */ 2393 int 2394 nfsrv_errmap(struct nfsrv_descript *nd, int err) 2395 { 2396 const short *defaulterrp, *errp; 2397 2398 if (nd->nd_flag & ND_NFSV3) { 2399 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2400 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2401 while (*++errp) { 2402 if (*errp == err) 2403 return (err); 2404 else if (*errp > err) 2405 break; 2406 } 2407 return ((int)*defaulterrp); 2408 } else 2409 return (err & 0xffff); 2410 } 2411 if (err <= ELAST) 2412 return ((int)nfsrv_v2errmap[err - 1]); 2413 return (NFSERR_IO); 2414 } 2415 2416 u_int32_t 2417 nfs_getxid(void) 2418 { 2419 u_int32_t newxid; 2420 2421 /* get next xid. skip 0 */ 2422 do { 2423 newxid = atomic_inc_32_nv(&nfs_xid); 2424 } while (__predict_false(newxid == 0)); 2425 2426 return txdr_unsigned(newxid); 2427 } 2428 2429 /* 2430 * assign a new xid for existing request. 2431 * used for NFSERR_JUKEBOX handling. 2432 */ 2433 void 2434 nfs_renewxid(struct nfsreq *req) 2435 { 2436 u_int32_t xid; 2437 int off; 2438 2439 xid = nfs_getxid(); 2440 if (req->r_nmp->nm_sotype == SOCK_STREAM) 2441 off = sizeof(u_int32_t); /* RPC record mark */ 2442 else 2443 off = 0; 2444 2445 m_copyback(req->r_mreq, off, sizeof(xid), (void *)&xid); 2446 req->r_xid = xid; 2447 } 2448 2449 #if defined(NFS) 2450 /* 2451 * Set the attribute timeout based on how recently the file has been modified. 2452 */ 2453 2454 time_t 2455 nfs_attrtimeo(struct nfsmount *nmp, struct nfsnode *np) 2456 { 2457 time_t timeo; 2458 2459 if ((nmp->nm_flag & NFSMNT_NOAC) != 0) 2460 return 0; 2461 2462 if (((np)->n_flag & NMODIFIED) != 0) 2463 return NFS_MINATTRTIMO; 2464 2465 timeo = (time_second - np->n_mtime.tv_sec) / 10; 2466 timeo = max(timeo, NFS_MINATTRTIMO); 2467 timeo = min(timeo, NFS_MAXATTRTIMO); 2468 return timeo; 2469 } 2470 #endif /* defined(NFS) */ 2471
Indexes created Wed Sep 24 02:09:48 GMT 2025