nfs_subs.c revision 1.77
1 /* $NetBSD: nfs_subs.c,v 1.77 2000/06/20 02:29:51 mrg 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. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)nfs_subs.c 8.8 (Berkeley) 5/22/95 39 */ 40 41 #include "fs_nfs.h" 42 #include "opt_nfsserver.h" 43 #include "opt_iso.h" 44 #include "opt_inet.h" 45 46 /* 47 * These functions support the macros and help fiddle mbuf chains for 48 * the nfs op functions. They do things like create the rpc header and 49 * copy data between mbuf chains and uio lists. 50 */ 51 #include <sys/param.h> 52 #include <sys/proc.h> 53 #include <sys/systm.h> 54 #include <sys/kernel.h> 55 #include <sys/mount.h> 56 #include <sys/vnode.h> 57 #include <sys/namei.h> 58 #include <sys/mbuf.h> 59 #include <sys/socket.h> 60 #include <sys/stat.h> 61 #include <sys/malloc.h> 62 #include <sys/time.h> 63 #include <sys/dirent.h> 64 65 #include <vm/vm.h> 66 67 #include <uvm/uvm_extern.h> 68 69 #include <nfs/rpcv2.h> 70 #include <nfs/nfsproto.h> 71 #include <nfs/nfsnode.h> 72 #include <nfs/nfs.h> 73 #include <nfs/xdr_subs.h> 74 #include <nfs/nfsm_subs.h> 75 #include <nfs/nfsmount.h> 76 #include <nfs/nqnfs.h> 77 #include <nfs/nfsrtt.h> 78 #include <nfs/nfs_var.h> 79 80 #include <miscfs/specfs/specdev.h> 81 82 #include <vm/vm.h> 83 84 #include <netinet/in.h> 85 #ifdef ISO 86 #include <netiso/iso.h> 87 #endif 88 89 /* 90 * Data items converted to xdr at startup, since they are constant 91 * This is kinda hokey, but may save a little time doing byte swaps 92 */ 93 u_int32_t nfs_xdrneg1; 94 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 95 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 96 rpc_auth_kerb; 97 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; 98 99 /* And other global data */ 100 static u_int32_t nfs_xid = 0; 101 nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, 102 NFCHR, NFNON }; 103 nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, 104 NFFIFO, NFNON }; 105 enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 106 enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 107 int nfs_ticks; 108 109 /* NFS client/server stats. */ 110 struct nfsstats nfsstats; 111 112 /* 113 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 114 */ 115 int nfsv3_procid[NFS_NPROCS] = { 116 NFSPROC_NULL, 117 NFSPROC_GETATTR, 118 NFSPROC_SETATTR, 119 NFSPROC_NOOP, 120 NFSPROC_LOOKUP, 121 NFSPROC_READLINK, 122 NFSPROC_READ, 123 NFSPROC_NOOP, 124 NFSPROC_WRITE, 125 NFSPROC_CREATE, 126 NFSPROC_REMOVE, 127 NFSPROC_RENAME, 128 NFSPROC_LINK, 129 NFSPROC_SYMLINK, 130 NFSPROC_MKDIR, 131 NFSPROC_RMDIR, 132 NFSPROC_READDIR, 133 NFSPROC_FSSTAT, 134 NFSPROC_NOOP, 135 NFSPROC_NOOP, 136 NFSPROC_NOOP, 137 NFSPROC_NOOP, 138 NFSPROC_NOOP, 139 NFSPROC_NOOP, 140 NFSPROC_NOOP, 141 NFSPROC_NOOP 142 }; 143 144 /* 145 * and the reverse mapping from generic to Version 2 procedure numbers 146 */ 147 int nfsv2_procid[NFS_NPROCS] = { 148 NFSV2PROC_NULL, 149 NFSV2PROC_GETATTR, 150 NFSV2PROC_SETATTR, 151 NFSV2PROC_LOOKUP, 152 NFSV2PROC_NOOP, 153 NFSV2PROC_READLINK, 154 NFSV2PROC_READ, 155 NFSV2PROC_WRITE, 156 NFSV2PROC_CREATE, 157 NFSV2PROC_MKDIR, 158 NFSV2PROC_SYMLINK, 159 NFSV2PROC_CREATE, 160 NFSV2PROC_REMOVE, 161 NFSV2PROC_RMDIR, 162 NFSV2PROC_RENAME, 163 NFSV2PROC_LINK, 164 NFSV2PROC_READDIR, 165 NFSV2PROC_NOOP, 166 NFSV2PROC_STATFS, 167 NFSV2PROC_NOOP, 168 NFSV2PROC_NOOP, 169 NFSV2PROC_NOOP, 170 NFSV2PROC_NOOP, 171 NFSV2PROC_NOOP, 172 NFSV2PROC_NOOP, 173 NFSV2PROC_NOOP, 174 }; 175 176 /* 177 * Maps errno values to nfs error numbers. 178 * Use NFSERR_IO as the catch all for ones not specifically defined in 179 * RFC 1094. 180 */ 181 static u_char nfsrv_v2errmap[ELAST] = { 182 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 183 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 184 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 185 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 186 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 187 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 188 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 189 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 190 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 191 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 192 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 193 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 194 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, 195 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 196 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 197 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 198 NFSERR_IO, NFSERR_IO, 199 }; 200 201 /* 202 * Maps errno values to nfs error numbers. 203 * Although it is not obvious whether or not NFS clients really care if 204 * a returned error value is in the specified list for the procedure, the 205 * safest thing to do is filter them appropriately. For Version 2, the 206 * X/Open XNFS document is the only specification that defines error values 207 * for each RPC (The RFC simply lists all possible error values for all RPCs), 208 * so I have decided to not do this for Version 2. 209 * The first entry is the default error return and the rest are the valid 210 * errors for that RPC in increasing numeric order. 211 */ 212 static short nfsv3err_null[] = { 213 0, 214 0, 215 }; 216 217 static short nfsv3err_getattr[] = { 218 NFSERR_IO, 219 NFSERR_IO, 220 NFSERR_STALE, 221 NFSERR_BADHANDLE, 222 NFSERR_SERVERFAULT, 223 0, 224 }; 225 226 static short nfsv3err_setattr[] = { 227 NFSERR_IO, 228 NFSERR_PERM, 229 NFSERR_IO, 230 NFSERR_ACCES, 231 NFSERR_INVAL, 232 NFSERR_NOSPC, 233 NFSERR_ROFS, 234 NFSERR_DQUOT, 235 NFSERR_STALE, 236 NFSERR_BADHANDLE, 237 NFSERR_NOT_SYNC, 238 NFSERR_SERVERFAULT, 239 0, 240 }; 241 242 static short nfsv3err_lookup[] = { 243 NFSERR_IO, 244 NFSERR_NOENT, 245 NFSERR_IO, 246 NFSERR_ACCES, 247 NFSERR_NOTDIR, 248 NFSERR_NAMETOL, 249 NFSERR_STALE, 250 NFSERR_BADHANDLE, 251 NFSERR_SERVERFAULT, 252 0, 253 }; 254 255 static short nfsv3err_access[] = { 256 NFSERR_IO, 257 NFSERR_IO, 258 NFSERR_STALE, 259 NFSERR_BADHANDLE, 260 NFSERR_SERVERFAULT, 261 0, 262 }; 263 264 static short nfsv3err_readlink[] = { 265 NFSERR_IO, 266 NFSERR_IO, 267 NFSERR_ACCES, 268 NFSERR_INVAL, 269 NFSERR_STALE, 270 NFSERR_BADHANDLE, 271 NFSERR_NOTSUPP, 272 NFSERR_SERVERFAULT, 273 0, 274 }; 275 276 static short nfsv3err_read[] = { 277 NFSERR_IO, 278 NFSERR_IO, 279 NFSERR_NXIO, 280 NFSERR_ACCES, 281 NFSERR_INVAL, 282 NFSERR_STALE, 283 NFSERR_BADHANDLE, 284 NFSERR_SERVERFAULT, 285 NFSERR_JUKEBOX, 286 0, 287 }; 288 289 static short nfsv3err_write[] = { 290 NFSERR_IO, 291 NFSERR_IO, 292 NFSERR_ACCES, 293 NFSERR_INVAL, 294 NFSERR_FBIG, 295 NFSERR_NOSPC, 296 NFSERR_ROFS, 297 NFSERR_DQUOT, 298 NFSERR_STALE, 299 NFSERR_BADHANDLE, 300 NFSERR_SERVERFAULT, 301 NFSERR_JUKEBOX, 302 0, 303 }; 304 305 static short nfsv3err_create[] = { 306 NFSERR_IO, 307 NFSERR_IO, 308 NFSERR_ACCES, 309 NFSERR_EXIST, 310 NFSERR_NOTDIR, 311 NFSERR_NOSPC, 312 NFSERR_ROFS, 313 NFSERR_NAMETOL, 314 NFSERR_DQUOT, 315 NFSERR_STALE, 316 NFSERR_BADHANDLE, 317 NFSERR_NOTSUPP, 318 NFSERR_SERVERFAULT, 319 0, 320 }; 321 322 static short nfsv3err_mkdir[] = { 323 NFSERR_IO, 324 NFSERR_IO, 325 NFSERR_ACCES, 326 NFSERR_EXIST, 327 NFSERR_NOTDIR, 328 NFSERR_NOSPC, 329 NFSERR_ROFS, 330 NFSERR_NAMETOL, 331 NFSERR_DQUOT, 332 NFSERR_STALE, 333 NFSERR_BADHANDLE, 334 NFSERR_NOTSUPP, 335 NFSERR_SERVERFAULT, 336 0, 337 }; 338 339 static short nfsv3err_symlink[] = { 340 NFSERR_IO, 341 NFSERR_IO, 342 NFSERR_ACCES, 343 NFSERR_EXIST, 344 NFSERR_NOTDIR, 345 NFSERR_NOSPC, 346 NFSERR_ROFS, 347 NFSERR_NAMETOL, 348 NFSERR_DQUOT, 349 NFSERR_STALE, 350 NFSERR_BADHANDLE, 351 NFSERR_NOTSUPP, 352 NFSERR_SERVERFAULT, 353 0, 354 }; 355 356 static short nfsv3err_mknod[] = { 357 NFSERR_IO, 358 NFSERR_IO, 359 NFSERR_ACCES, 360 NFSERR_EXIST, 361 NFSERR_NOTDIR, 362 NFSERR_NOSPC, 363 NFSERR_ROFS, 364 NFSERR_NAMETOL, 365 NFSERR_DQUOT, 366 NFSERR_STALE, 367 NFSERR_BADHANDLE, 368 NFSERR_NOTSUPP, 369 NFSERR_SERVERFAULT, 370 NFSERR_BADTYPE, 371 0, 372 }; 373 374 static short nfsv3err_remove[] = { 375 NFSERR_IO, 376 NFSERR_NOENT, 377 NFSERR_IO, 378 NFSERR_ACCES, 379 NFSERR_NOTDIR, 380 NFSERR_ROFS, 381 NFSERR_NAMETOL, 382 NFSERR_STALE, 383 NFSERR_BADHANDLE, 384 NFSERR_SERVERFAULT, 385 0, 386 }; 387 388 static short nfsv3err_rmdir[] = { 389 NFSERR_IO, 390 NFSERR_NOENT, 391 NFSERR_IO, 392 NFSERR_ACCES, 393 NFSERR_EXIST, 394 NFSERR_NOTDIR, 395 NFSERR_INVAL, 396 NFSERR_ROFS, 397 NFSERR_NAMETOL, 398 NFSERR_NOTEMPTY, 399 NFSERR_STALE, 400 NFSERR_BADHANDLE, 401 NFSERR_NOTSUPP, 402 NFSERR_SERVERFAULT, 403 0, 404 }; 405 406 static short nfsv3err_rename[] = { 407 NFSERR_IO, 408 NFSERR_NOENT, 409 NFSERR_IO, 410 NFSERR_ACCES, 411 NFSERR_EXIST, 412 NFSERR_XDEV, 413 NFSERR_NOTDIR, 414 NFSERR_ISDIR, 415 NFSERR_INVAL, 416 NFSERR_NOSPC, 417 NFSERR_ROFS, 418 NFSERR_MLINK, 419 NFSERR_NAMETOL, 420 NFSERR_NOTEMPTY, 421 NFSERR_DQUOT, 422 NFSERR_STALE, 423 NFSERR_BADHANDLE, 424 NFSERR_NOTSUPP, 425 NFSERR_SERVERFAULT, 426 0, 427 }; 428 429 static short nfsv3err_link[] = { 430 NFSERR_IO, 431 NFSERR_IO, 432 NFSERR_ACCES, 433 NFSERR_EXIST, 434 NFSERR_XDEV, 435 NFSERR_NOTDIR, 436 NFSERR_INVAL, 437 NFSERR_NOSPC, 438 NFSERR_ROFS, 439 NFSERR_MLINK, 440 NFSERR_NAMETOL, 441 NFSERR_DQUOT, 442 NFSERR_STALE, 443 NFSERR_BADHANDLE, 444 NFSERR_NOTSUPP, 445 NFSERR_SERVERFAULT, 446 0, 447 }; 448 449 static short nfsv3err_readdir[] = { 450 NFSERR_IO, 451 NFSERR_IO, 452 NFSERR_ACCES, 453 NFSERR_NOTDIR, 454 NFSERR_STALE, 455 NFSERR_BADHANDLE, 456 NFSERR_BAD_COOKIE, 457 NFSERR_TOOSMALL, 458 NFSERR_SERVERFAULT, 459 0, 460 }; 461 462 static short nfsv3err_readdirplus[] = { 463 NFSERR_IO, 464 NFSERR_IO, 465 NFSERR_ACCES, 466 NFSERR_NOTDIR, 467 NFSERR_STALE, 468 NFSERR_BADHANDLE, 469 NFSERR_BAD_COOKIE, 470 NFSERR_NOTSUPP, 471 NFSERR_TOOSMALL, 472 NFSERR_SERVERFAULT, 473 0, 474 }; 475 476 static short nfsv3err_fsstat[] = { 477 NFSERR_IO, 478 NFSERR_IO, 479 NFSERR_STALE, 480 NFSERR_BADHANDLE, 481 NFSERR_SERVERFAULT, 482 0, 483 }; 484 485 static short nfsv3err_fsinfo[] = { 486 NFSERR_STALE, 487 NFSERR_STALE, 488 NFSERR_BADHANDLE, 489 NFSERR_SERVERFAULT, 490 0, 491 }; 492 493 static short nfsv3err_pathconf[] = { 494 NFSERR_STALE, 495 NFSERR_STALE, 496 NFSERR_BADHANDLE, 497 NFSERR_SERVERFAULT, 498 0, 499 }; 500 501 static short nfsv3err_commit[] = { 502 NFSERR_IO, 503 NFSERR_IO, 504 NFSERR_STALE, 505 NFSERR_BADHANDLE, 506 NFSERR_SERVERFAULT, 507 0, 508 }; 509 510 static short *nfsrv_v3errmap[] = { 511 nfsv3err_null, 512 nfsv3err_getattr, 513 nfsv3err_setattr, 514 nfsv3err_lookup, 515 nfsv3err_access, 516 nfsv3err_readlink, 517 nfsv3err_read, 518 nfsv3err_write, 519 nfsv3err_create, 520 nfsv3err_mkdir, 521 nfsv3err_symlink, 522 nfsv3err_mknod, 523 nfsv3err_remove, 524 nfsv3err_rmdir, 525 nfsv3err_rename, 526 nfsv3err_link, 527 nfsv3err_readdir, 528 nfsv3err_readdirplus, 529 nfsv3err_fsstat, 530 nfsv3err_fsinfo, 531 nfsv3err_pathconf, 532 nfsv3err_commit, 533 }; 534 535 extern struct nfsrtt nfsrtt; 536 extern time_t nqnfsstarttime; 537 extern int nqsrv_clockskew; 538 extern int nqsrv_writeslack; 539 extern int nqsrv_maxlease; 540 extern int nqnfs_piggy[NFS_NPROCS]; 541 extern struct nfsnodehashhead *nfsnodehashtbl; 542 extern u_long nfsnodehash; 543 544 LIST_HEAD(nfsnodehashhead, nfsnode); 545 u_long nfsdirhashmask; 546 547 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 548 549 /* 550 * Create the header for an rpc request packet 551 * The hsiz is the size of the rest of the nfs request header. 552 * (just used to decide if a cluster is a good idea) 553 */ 554 struct mbuf * 555 nfsm_reqh(vp, procid, hsiz, bposp) 556 struct vnode *vp; 557 u_long procid; 558 int hsiz; 559 caddr_t *bposp; 560 { 561 struct mbuf *mb; 562 u_int32_t *tl; 563 caddr_t bpos; 564 struct mbuf *mb2; 565 struct nfsmount *nmp; 566 int nqflag; 567 568 MGET(mb, M_WAIT, MT_DATA); 569 if (hsiz >= MINCLSIZE) 570 MCLGET(mb, M_WAIT); 571 mb->m_len = 0; 572 bpos = mtod(mb, caddr_t); 573 574 /* 575 * For NQNFS, add lease request. 576 */ 577 if (vp) { 578 nmp = VFSTONFS(vp->v_mount); 579 if (nmp->nm_flag & NFSMNT_NQNFS) { 580 nqflag = NQNFS_NEEDLEASE(vp, procid); 581 if (nqflag) { 582 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); 583 *tl++ = txdr_unsigned(nqflag); 584 *tl = txdr_unsigned(nmp->nm_leaseterm); 585 } else { 586 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 587 *tl = 0; 588 } 589 } 590 } 591 /* Finally, return values */ 592 *bposp = bpos; 593 return (mb); 594 } 595 596 /* 597 * Build the RPC header and fill in the authorization info. 598 * The authorization string argument is only used when the credentials 599 * come from outside of the kernel. 600 * Returns the head of the mbuf list. 601 */ 602 struct mbuf * 603 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 604 verf_str, mrest, mrest_len, mbp, xidp) 605 struct ucred *cr; 606 int nmflag; 607 int procid; 608 int auth_type; 609 int auth_len; 610 char *auth_str; 611 int verf_len; 612 char *verf_str; 613 struct mbuf *mrest; 614 int mrest_len; 615 struct mbuf **mbp; 616 u_int32_t *xidp; 617 { 618 struct mbuf *mb; 619 u_int32_t *tl; 620 caddr_t bpos; 621 int i; 622 struct mbuf *mreq, *mb2; 623 int siz, grpsiz, authsiz; 624 struct timeval tv; 625 static u_int32_t base; 626 627 authsiz = nfsm_rndup(auth_len); 628 MGETHDR(mb, M_WAIT, MT_DATA); 629 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 630 MCLGET(mb, M_WAIT); 631 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 632 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 633 } else { 634 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 635 } 636 mb->m_len = 0; 637 mreq = mb; 638 bpos = mtod(mb, caddr_t); 639 640 /* 641 * First the RPC header. 642 */ 643 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 644 645 /* 646 * derive initial xid from system time 647 * XXX time is invalid if root not yet mounted 648 */ 649 if (!base && (rootvp)) { 650 microtime(&tv); 651 base = tv.tv_sec << 12; 652 nfs_xid = base; 653 } 654 /* 655 * Skip zero xid if it should ever happen. 656 */ 657 if (++nfs_xid == 0) 658 nfs_xid++; 659 660 *tl++ = *xidp = txdr_unsigned(nfs_xid); 661 *tl++ = rpc_call; 662 *tl++ = rpc_vers; 663 if (nmflag & NFSMNT_NQNFS) { 664 *tl++ = txdr_unsigned(NQNFS_PROG); 665 *tl++ = txdr_unsigned(NQNFS_VER3); 666 } else { 667 *tl++ = txdr_unsigned(NFS_PROG); 668 if (nmflag & NFSMNT_NFSV3) 669 *tl++ = txdr_unsigned(NFS_VER3); 670 else 671 *tl++ = txdr_unsigned(NFS_VER2); 672 } 673 if (nmflag & NFSMNT_NFSV3) 674 *tl++ = txdr_unsigned(procid); 675 else 676 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 677 678 /* 679 * And then the authorization cred. 680 */ 681 *tl++ = txdr_unsigned(auth_type); 682 *tl = txdr_unsigned(authsiz); 683 switch (auth_type) { 684 case RPCAUTH_UNIX: 685 nfsm_build(tl, u_int32_t *, auth_len); 686 *tl++ = 0; /* stamp ?? */ 687 *tl++ = 0; /* NULL hostname */ 688 *tl++ = txdr_unsigned(cr->cr_uid); 689 *tl++ = txdr_unsigned(cr->cr_gid); 690 grpsiz = (auth_len >> 2) - 5; 691 *tl++ = txdr_unsigned(grpsiz); 692 for (i = 0; i < grpsiz; i++) 693 *tl++ = txdr_unsigned(cr->cr_groups[i]); 694 break; 695 case RPCAUTH_KERB4: 696 siz = auth_len; 697 while (siz > 0) { 698 if (M_TRAILINGSPACE(mb) == 0) { 699 MGET(mb2, M_WAIT, MT_DATA); 700 if (siz >= MINCLSIZE) 701 MCLGET(mb2, M_WAIT); 702 mb->m_next = mb2; 703 mb = mb2; 704 mb->m_len = 0; 705 bpos = mtod(mb, caddr_t); 706 } 707 i = min(siz, M_TRAILINGSPACE(mb)); 708 memcpy(bpos, auth_str, i); 709 mb->m_len += i; 710 auth_str += i; 711 bpos += i; 712 siz -= i; 713 } 714 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 715 for (i = 0; i < siz; i++) 716 *bpos++ = '\0'; 717 mb->m_len += siz; 718 } 719 break; 720 }; 721 722 /* 723 * And the verifier... 724 */ 725 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 726 if (verf_str) { 727 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 728 *tl = txdr_unsigned(verf_len); 729 siz = verf_len; 730 while (siz > 0) { 731 if (M_TRAILINGSPACE(mb) == 0) { 732 MGET(mb2, M_WAIT, MT_DATA); 733 if (siz >= MINCLSIZE) 734 MCLGET(mb2, M_WAIT); 735 mb->m_next = mb2; 736 mb = mb2; 737 mb->m_len = 0; 738 bpos = mtod(mb, caddr_t); 739 } 740 i = min(siz, M_TRAILINGSPACE(mb)); 741 memcpy(bpos, verf_str, i); 742 mb->m_len += i; 743 verf_str += i; 744 bpos += i; 745 siz -= i; 746 } 747 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 748 for (i = 0; i < siz; i++) 749 *bpos++ = '\0'; 750 mb->m_len += siz; 751 } 752 } else { 753 *tl++ = txdr_unsigned(RPCAUTH_NULL); 754 *tl = 0; 755 } 756 mb->m_next = mrest; 757 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 758 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 759 *mbp = mb; 760 return (mreq); 761 } 762 763 /* 764 * copies mbuf chain to the uio scatter/gather list 765 */ 766 int 767 nfsm_mbuftouio(mrep, uiop, siz, dpos) 768 struct mbuf **mrep; 769 struct uio *uiop; 770 int siz; 771 caddr_t *dpos; 772 { 773 char *mbufcp, *uiocp; 774 int xfer, left, len; 775 struct mbuf *mp; 776 long uiosiz, rem; 777 int error = 0; 778 779 mp = *mrep; 780 mbufcp = *dpos; 781 len = mtod(mp, caddr_t)+mp->m_len-mbufcp; 782 rem = nfsm_rndup(siz)-siz; 783 while (siz > 0) { 784 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 785 return (EFBIG); 786 left = uiop->uio_iov->iov_len; 787 uiocp = uiop->uio_iov->iov_base; 788 if (left > siz) 789 left = siz; 790 uiosiz = left; 791 while (left > 0) { 792 while (len == 0) { 793 mp = mp->m_next; 794 if (mp == NULL) 795 return (EBADRPC); 796 mbufcp = mtod(mp, caddr_t); 797 len = mp->m_len; 798 } 799 xfer = (left > len) ? len : left; 800 #ifdef notdef 801 /* Not Yet.. */ 802 if (uiop->uio_iov->iov_op != NULL) 803 (*(uiop->uio_iov->iov_op)) 804 (mbufcp, uiocp, xfer); 805 else 806 #endif 807 if (uiop->uio_segflg == UIO_SYSSPACE) 808 memcpy(uiocp, mbufcp, xfer); 809 else 810 copyout(mbufcp, uiocp, xfer); 811 left -= xfer; 812 len -= xfer; 813 mbufcp += xfer; 814 uiocp += xfer; 815 uiop->uio_offset += xfer; 816 uiop->uio_resid -= xfer; 817 } 818 if (uiop->uio_iov->iov_len <= siz) { 819 uiop->uio_iovcnt--; 820 uiop->uio_iov++; 821 } else { 822 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 823 uiop->uio_iov->iov_len -= uiosiz; 824 } 825 siz -= uiosiz; 826 } 827 *dpos = mbufcp; 828 *mrep = mp; 829 if (rem > 0) { 830 if (len < rem) 831 error = nfs_adv(mrep, dpos, rem, len); 832 else 833 *dpos += rem; 834 } 835 return (error); 836 } 837 838 /* 839 * copies a uio scatter/gather list to an mbuf chain. 840 * NOTE: can ony handle iovcnt == 1 841 */ 842 int 843 nfsm_uiotombuf(uiop, mq, siz, bpos) 844 struct uio *uiop; 845 struct mbuf **mq; 846 int siz; 847 caddr_t *bpos; 848 { 849 char *uiocp; 850 struct mbuf *mp, *mp2; 851 int xfer, left, mlen; 852 int uiosiz, clflg, rem; 853 char *cp; 854 855 #ifdef DIAGNOSTIC 856 if (uiop->uio_iovcnt != 1) 857 panic("nfsm_uiotombuf: iovcnt != 1"); 858 #endif 859 860 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 861 clflg = 1; 862 else 863 clflg = 0; 864 rem = nfsm_rndup(siz)-siz; 865 mp = mp2 = *mq; 866 while (siz > 0) { 867 left = uiop->uio_iov->iov_len; 868 uiocp = uiop->uio_iov->iov_base; 869 if (left > siz) 870 left = siz; 871 uiosiz = left; 872 while (left > 0) { 873 mlen = M_TRAILINGSPACE(mp); 874 if (mlen == 0) { 875 MGET(mp, M_WAIT, MT_DATA); 876 if (clflg) 877 MCLGET(mp, M_WAIT); 878 mp->m_len = 0; 879 mp2->m_next = mp; 880 mp2 = mp; 881 mlen = M_TRAILINGSPACE(mp); 882 } 883 xfer = (left > mlen) ? mlen : left; 884 #ifdef notdef 885 /* Not Yet.. */ 886 if (uiop->uio_iov->iov_op != NULL) 887 (*(uiop->uio_iov->iov_op)) 888 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 889 else 890 #endif 891 if (uiop->uio_segflg == UIO_SYSSPACE) 892 memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer); 893 else 894 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 895 mp->m_len += xfer; 896 left -= xfer; 897 uiocp += xfer; 898 uiop->uio_offset += xfer; 899 uiop->uio_resid -= xfer; 900 } 901 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 902 uiop->uio_iov->iov_len -= uiosiz; 903 siz -= uiosiz; 904 } 905 if (rem > 0) { 906 if (rem > M_TRAILINGSPACE(mp)) { 907 MGET(mp, M_WAIT, MT_DATA); 908 mp->m_len = 0; 909 mp2->m_next = mp; 910 } 911 cp = mtod(mp, caddr_t)+mp->m_len; 912 for (left = 0; left < rem; left++) 913 *cp++ = '\0'; 914 mp->m_len += rem; 915 *bpos = cp; 916 } else 917 *bpos = mtod(mp, caddr_t)+mp->m_len; 918 *mq = mp; 919 return (0); 920 } 921 922 /* 923 * Get at least "siz" bytes of correctly aligned data. 924 * When called the mbuf pointers are not necessarily correct, 925 * dsosp points to what ought to be in m_data and left contains 926 * what ought to be in m_len. 927 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 928 * cases. (The macros use the vars. dpos and dpos2) 929 */ 930 int 931 nfsm_disct(mdp, dposp, siz, left, cp2) 932 struct mbuf **mdp; 933 caddr_t *dposp; 934 int siz; 935 int left; 936 caddr_t *cp2; 937 { 938 struct mbuf *m1, *m2; 939 struct mbuf *havebuf = NULL; 940 caddr_t src = *dposp; 941 caddr_t dst; 942 int len; 943 944 #ifdef DEBUG 945 if (left < 0) 946 panic("nfsm_disct: left < 0"); 947 #endif 948 m1 = *mdp; 949 /* 950 * Skip through the mbuf chain looking for an mbuf with 951 * some data. If the first mbuf found has enough data 952 * and it is correctly aligned return it. 953 */ 954 while (left == 0) { 955 havebuf = m1; 956 *mdp = m1 = m1->m_next; 957 if (m1 == NULL) 958 return (EBADRPC); 959 src = mtod(m1, caddr_t); 960 left = m1->m_len; 961 /* 962 * If we start a new mbuf and it is big enough 963 * and correctly aligned just return it, don't 964 * do any pull up. 965 */ 966 if (left >= siz && nfsm_aligned(src)) { 967 *cp2 = src; 968 *dposp = src + siz; 969 return (0); 970 } 971 } 972 if (m1->m_flags & M_EXT) { 973 if (havebuf) { 974 /* If the first mbuf with data has external data 975 * and there is a previous empty mbuf use it 976 * to move the data into. 977 */ 978 m2 = m1; 979 *mdp = m1 = havebuf; 980 if (m1->m_flags & M_EXT) { 981 MEXTREMOVE(m1); 982 } 983 } else { 984 /* 985 * If the first mbuf has a external data 986 * and there is no previous empty mbuf 987 * allocate a new mbuf and move the external 988 * data to the new mbuf. Also make the first 989 * mbuf look empty. 990 */ 991 m2 = m_get(M_WAIT, MT_DATA); 992 m2->m_ext = m1->m_ext; 993 m2->m_data = src; 994 m2->m_len = left; 995 MCLADDREFERENCE(m1, m2); 996 MEXTREMOVE(m1); 997 m2->m_next = m1->m_next; 998 m1->m_next = m2; 999 } 1000 m1->m_len = 0; 1001 dst = m1->m_dat; 1002 } else { 1003 /* 1004 * If the first mbuf has no external data 1005 * move the data to the front of the mbuf. 1006 */ 1007 if ((dst = m1->m_dat) != src) 1008 memmove(dst, src, left); 1009 dst += left; 1010 m1->m_len = left; 1011 m2 = m1->m_next; 1012 } 1013 m1->m_flags &= ~M_PKTHDR; 1014 *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */ 1015 *dposp = mtod(m1, caddr_t) + siz; 1016 /* 1017 * Loop through mbufs pulling data up into first mbuf until 1018 * the first mbuf is full or there is no more data to 1019 * pullup. 1020 */ 1021 while ((len = (MLEN - m1->m_len)) != 0 && m2) { 1022 if ((len = min(len, m2->m_len)) != 0) 1023 memcpy(dst, m2->m_data, len); 1024 m1->m_len += len; 1025 dst += len; 1026 m2->m_data += len; 1027 m2->m_len -= len; 1028 m2 = m2->m_next; 1029 } 1030 if (m1->m_len < siz) 1031 return (EBADRPC); 1032 return (0); 1033 } 1034 1035 /* 1036 * Advance the position in the mbuf chain. 1037 */ 1038 int 1039 nfs_adv(mdp, dposp, offs, left) 1040 struct mbuf **mdp; 1041 caddr_t *dposp; 1042 int offs; 1043 int left; 1044 { 1045 struct mbuf *m; 1046 int s; 1047 1048 m = *mdp; 1049 s = left; 1050 while (s < offs) { 1051 offs -= s; 1052 m = m->m_next; 1053 if (m == NULL) 1054 return (EBADRPC); 1055 s = m->m_len; 1056 } 1057 *mdp = m; 1058 *dposp = mtod(m, caddr_t)+offs; 1059 return (0); 1060 } 1061 1062 /* 1063 * Copy a string into mbufs for the hard cases... 1064 */ 1065 int 1066 nfsm_strtmbuf(mb, bpos, cp, siz) 1067 struct mbuf **mb; 1068 char **bpos; 1069 const char *cp; 1070 long siz; 1071 { 1072 struct mbuf *m1 = NULL, *m2; 1073 long left, xfer, len, tlen; 1074 u_int32_t *tl; 1075 int putsize; 1076 1077 putsize = 1; 1078 m2 = *mb; 1079 left = M_TRAILINGSPACE(m2); 1080 if (left > 0) { 1081 tl = ((u_int32_t *)(*bpos)); 1082 *tl++ = txdr_unsigned(siz); 1083 putsize = 0; 1084 left -= NFSX_UNSIGNED; 1085 m2->m_len += NFSX_UNSIGNED; 1086 if (left > 0) { 1087 memcpy((caddr_t) tl, cp, left); 1088 siz -= left; 1089 cp += left; 1090 m2->m_len += left; 1091 left = 0; 1092 } 1093 } 1094 /* Loop around adding mbufs */ 1095 while (siz > 0) { 1096 MGET(m1, M_WAIT, MT_DATA); 1097 if (siz > MLEN) 1098 MCLGET(m1, M_WAIT); 1099 m1->m_len = NFSMSIZ(m1); 1100 m2->m_next = m1; 1101 m2 = m1; 1102 tl = mtod(m1, u_int32_t *); 1103 tlen = 0; 1104 if (putsize) { 1105 *tl++ = txdr_unsigned(siz); 1106 m1->m_len -= NFSX_UNSIGNED; 1107 tlen = NFSX_UNSIGNED; 1108 putsize = 0; 1109 } 1110 if (siz < m1->m_len) { 1111 len = nfsm_rndup(siz); 1112 xfer = siz; 1113 if (xfer < len) 1114 *(tl+(xfer>>2)) = 0; 1115 } else { 1116 xfer = len = m1->m_len; 1117 } 1118 memcpy((caddr_t) tl, cp, xfer); 1119 m1->m_len = len+tlen; 1120 siz -= xfer; 1121 cp += xfer; 1122 } 1123 *mb = m1; 1124 *bpos = mtod(m1, caddr_t)+m1->m_len; 1125 return (0); 1126 } 1127 1128 /* 1129 * Directory caching routines. They work as follows: 1130 * - a cache is maintained per VDIR nfsnode. 1131 * - for each offset cookie that is exported to userspace, and can 1132 * thus be thrown back at us as an offset to VOP_READDIR, store 1133 * information in the cache. 1134 * - cached are: 1135 * - cookie itself 1136 * - blocknumber (essentially just a search key in the buffer cache) 1137 * - entry number in block. 1138 * - offset cookie of block in which this entry is stored 1139 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1140 * - entries are looked up in a hash table 1141 * - also maintained is an LRU list of entries, used to determine 1142 * which ones to delete if the cache grows too large. 1143 * - if 32 <-> 64 translation mode is requested for a filesystem, 1144 * the cache also functions as a translation table 1145 * - in the translation case, invalidating the cache does not mean 1146 * flushing it, but just marking entries as invalid, except for 1147 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1148 * still be able to use the cache as a translation table. 1149 * - 32 bit cookies are uniquely created by combining the hash table 1150 * entry value, and one generation count per hash table entry, 1151 * incremented each time an entry is appended to the chain. 1152 * - the cache is invalidated each time a direcory is modified 1153 * - sanity checks are also done; if an entry in a block turns 1154 * out not to have a matching cookie, the cache is invalidated 1155 * and a new block starting from the wanted offset is fetched from 1156 * the server. 1157 * - directory entries as read from the server are extended to contain 1158 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1159 * the cache and exporting them to userspace through the cookie 1160 * argument to VOP_READDIR. 1161 */ 1162 1163 u_long 1164 nfs_dirhash(off) 1165 off_t off; 1166 { 1167 int i; 1168 char *cp = (char *)&off; 1169 u_long sum = 0L; 1170 1171 for (i = 0 ; i < sizeof (off); i++) 1172 sum += *cp++; 1173 1174 return sum; 1175 } 1176 1177 void 1178 nfs_initdircache(vp) 1179 struct vnode *vp; 1180 { 1181 struct nfsnode *np = VTONFS(vp); 1182 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1183 1184 np->n_dircachesize = 0; 1185 np->n_dblkno = 1; 1186 np->n_dircache = 1187 hashinit(NFS_DIRHASHSIZ, M_NFSDIROFF, M_WAITOK, &nfsdirhashmask); 1188 TAILQ_INIT(&np->n_dirchain); 1189 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1190 MALLOC(np->n_dirgens, unsigned *, 1191 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, 1192 M_WAITOK); 1193 memset((caddr_t)np->n_dirgens, 0, 1194 NFS_DIRHASHSIZ * sizeof (unsigned)); 1195 } 1196 } 1197 1198 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0}; 1199 1200 struct nfsdircache * 1201 nfs_searchdircache(vp, off, do32, hashent) 1202 struct vnode *vp; 1203 off_t off; 1204 int do32; 1205 int *hashent; 1206 { 1207 struct nfsdirhashhead *ndhp; 1208 struct nfsdircache *ndp = NULL; 1209 struct nfsnode *np = VTONFS(vp); 1210 unsigned ent; 1211 1212 /* 1213 * Zero is always a valid cookie. 1214 */ 1215 if (off == 0) 1216 return &dzero; 1217 1218 /* 1219 * We use a 32bit cookie as search key, directly reconstruct 1220 * the hashentry. Else use the hashfunction. 1221 */ 1222 if (do32) { 1223 ent = (u_int32_t)off >> 24; 1224 if (ent >= NFS_DIRHASHSIZ) 1225 return NULL; 1226 ndhp = &np->n_dircache[ent]; 1227 } else { 1228 ndhp = NFSDIRHASH(np, off); 1229 } 1230 1231 if (hashent) 1232 *hashent = (int)(ndhp - np->n_dircache); 1233 if (do32) { 1234 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) { 1235 if (ndp->dc_cookie32 == (u_int32_t)off) { 1236 /* 1237 * An invalidated entry will become the 1238 * start of a new block fetched from 1239 * the server. 1240 */ 1241 if (ndp->dc_blkno == -1) { 1242 ndp->dc_blkcookie = ndp->dc_cookie; 1243 ndp->dc_blkno = np->n_dblkno++; 1244 ndp->dc_entry = 0; 1245 } 1246 break; 1247 } 1248 } 1249 } else { 1250 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) 1251 if (ndp->dc_cookie == off) 1252 break; 1253 } 1254 return ndp; 1255 } 1256 1257 1258 struct nfsdircache * 1259 nfs_enterdircache(vp, off, blkoff, en, blkno) 1260 struct vnode *vp; 1261 off_t off, blkoff; 1262 daddr_t blkno; 1263 int en; 1264 { 1265 struct nfsnode *np = VTONFS(vp); 1266 struct nfsdirhashhead *ndhp; 1267 struct nfsdircache *ndp = NULL, *first; 1268 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1269 int hashent, gen, overwrite; 1270 1271 if (!np->n_dircache) 1272 /* 1273 * XXX would like to do this in nfs_nget but vtype 1274 * isn't known at that time. 1275 */ 1276 nfs_initdircache(vp); 1277 1278 /* 1279 * XXX refuse entries for offset 0. amd(8) erroneously sets 1280 * cookie 0 for the '.' entry, making this necessary. This 1281 * isn't so bad, as 0 is a special case anyway. 1282 */ 1283 if (off == 0) 1284 return &dzero; 1285 1286 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1287 1288 if (ndp && ndp->dc_blkno != -1) { 1289 /* 1290 * Overwriting an old entry. Check if it's the same. 1291 * If so, just return. If not, remove the old entry. 1292 */ 1293 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1294 return ndp; 1295 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1296 LIST_REMOVE(ndp, dc_hash); 1297 FREE(ndp, M_NFSDIROFF); 1298 ndp = 0; 1299 } 1300 1301 ndhp = &np->n_dircache[hashent]; 1302 1303 if (!ndp) { 1304 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF, 1305 M_WAITOK); 1306 overwrite = 0; 1307 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1308 /* 1309 * We're allocating a new entry, so bump the 1310 * generation number. 1311 */ 1312 gen = ++np->n_dirgens[hashent]; 1313 if (gen == 0) { 1314 np->n_dirgens[hashent]++; 1315 gen++; 1316 } 1317 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1318 } 1319 } else 1320 overwrite = 1; 1321 1322 /* 1323 * If the entry number is 0, we are at the start of a new block, so 1324 * allocate a new blocknumber. 1325 */ 1326 if (en == 0) 1327 ndp->dc_blkno = np->n_dblkno++; 1328 else 1329 ndp->dc_blkno = blkno; 1330 1331 ndp->dc_cookie = off; 1332 ndp->dc_blkcookie = blkoff; 1333 ndp->dc_entry = en; 1334 1335 if (overwrite) 1336 return ndp; 1337 1338 /* 1339 * If the maximum directory cookie cache size has been reached 1340 * for this node, take one off the front. The idea is that 1341 * directories are typically read front-to-back once, so that 1342 * the oldest entries can be thrown away without much performance 1343 * loss. 1344 */ 1345 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1346 first = np->n_dirchain.tqh_first; 1347 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain); 1348 LIST_REMOVE(first, dc_hash); 1349 FREE(first, M_NFSDIROFF); 1350 } else 1351 np->n_dircachesize++; 1352 1353 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1354 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1355 return ndp; 1356 } 1357 1358 void 1359 nfs_invaldircache(vp, forcefree) 1360 struct vnode *vp; 1361 int forcefree; 1362 { 1363 struct nfsnode *np = VTONFS(vp); 1364 struct nfsdircache *ndp = NULL; 1365 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1366 1367 #ifdef DIAGNOSTIC 1368 if (vp->v_type != VDIR) 1369 panic("nfs: invaldircache: not dir"); 1370 #endif 1371 1372 if (!np->n_dircache) 1373 return; 1374 1375 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1376 while ((ndp = np->n_dirchain.tqh_first)) { 1377 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1378 LIST_REMOVE(ndp, dc_hash); 1379 FREE(ndp, M_NFSDIROFF); 1380 } 1381 np->n_dircachesize = 0; 1382 if (forcefree && np->n_dirgens) { 1383 FREE(np->n_dirgens, M_NFSDIROFF); 1384 } 1385 } else { 1386 for (ndp = np->n_dirchain.tqh_first; ndp; 1387 ndp = ndp->dc_chain.tqe_next) 1388 ndp->dc_blkno = -1; 1389 } 1390 1391 np->n_dblkno = 1; 1392 } 1393 1394 /* 1395 * Called once before VFS init to initialize shared and 1396 * server-specific data structures. 1397 */ 1398 void 1399 nfs_init() 1400 { 1401 1402 #if !defined(alpha) && !defined(_LP64) && defined(DIAGNOSTIC) 1403 /* 1404 * Check to see if major data structures haven't bloated. 1405 */ 1406 if (sizeof (struct nfsnode) > NFS_NODEALLOC) { 1407 printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); 1408 printf("Try reducing NFS_SMALLFH\n"); 1409 } 1410 if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { 1411 printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); 1412 printf("Try reducing NFS_UIDHASHSIZ\n"); 1413 } 1414 if (sizeof (struct nfsuid) > NFS_UIDALLOC) { 1415 printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); 1416 printf("Try unionizing the nu_nickname and nu_flag fields\n"); 1417 } 1418 #endif 1419 1420 nfsrtt.pos = 0; 1421 rpc_vers = txdr_unsigned(RPC_VER2); 1422 rpc_call = txdr_unsigned(RPC_CALL); 1423 rpc_reply = txdr_unsigned(RPC_REPLY); 1424 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1425 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1426 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1427 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1428 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1429 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1430 nfs_prog = txdr_unsigned(NFS_PROG); 1431 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1432 nfs_true = txdr_unsigned(TRUE); 1433 nfs_false = txdr_unsigned(FALSE); 1434 nfs_xdrneg1 = txdr_unsigned(-1); 1435 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1436 if (nfs_ticks < 1) 1437 nfs_ticks = 1; 1438 #ifdef NFSSERVER 1439 nfsrv_init(0); /* Init server data structures */ 1440 nfsrv_initcache(); /* Init the server request cache */ 1441 #endif /* NFSSERVER */ 1442 1443 /* 1444 * Initialize the nqnfs data structures. 1445 */ 1446 if (nqnfsstarttime == 0) { 1447 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1448 + nqsrv_clockskew + nqsrv_writeslack; 1449 NQLOADNOVRAM(nqnfsstarttime); 1450 CIRCLEQ_INIT(&nqtimerhead); 1451 nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, M_WAITOK, &nqfhhash); 1452 } 1453 1454 /* 1455 * Initialize reply list and start timer 1456 */ 1457 TAILQ_INIT(&nfs_reqq); 1458 nfs_timer(NULL); 1459 } 1460 1461 #ifdef NFS 1462 /* 1463 * Called once at VFS init to initialize client-specific data structures. 1464 */ 1465 void 1466 nfs_vfs_init() 1467 { 1468 int i; 1469 1470 /* Ensure async daemons disabled */ 1471 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { 1472 nfs_iodwant[i] = (struct proc *)0; 1473 nfs_iodmount[i] = (struct nfsmount *)0; 1474 } 1475 nfs_nhinit(); /* Init the nfsnode table */ 1476 } 1477 1478 void 1479 nfs_vfs_done() 1480 { 1481 nfs_nhdone(); 1482 } 1483 1484 /* 1485 * Attribute cache routines. 1486 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1487 * that are on the mbuf list 1488 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1489 * error otherwise 1490 */ 1491 1492 /* 1493 * Load the attribute cache (that lives in the nfsnode entry) with 1494 * the values on the mbuf list and 1495 * Iff vap not NULL 1496 * copy the attributes to *vaper 1497 */ 1498 int 1499 nfsm_loadattrcache(vpp, mdp, dposp, vaper) 1500 struct vnode **vpp; 1501 struct mbuf **mdp; 1502 caddr_t *dposp; 1503 struct vattr *vaper; 1504 { 1505 int32_t t1; 1506 caddr_t cp2; 1507 int error = 0; 1508 struct mbuf *md; 1509 int v3 = NFS_ISV3(*vpp); 1510 1511 md = *mdp; 1512 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1513 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1514 if (error) 1515 return (error); 1516 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper); 1517 } 1518 1519 int 1520 nfs_loadattrcache(vpp, fp, vaper) 1521 struct vnode **vpp; 1522 struct nfs_fattr *fp; 1523 struct vattr *vaper; 1524 { 1525 struct vnode *vp = *vpp; 1526 struct vattr *vap; 1527 int v3 = NFS_ISV3(vp); 1528 enum vtype vtyp; 1529 u_short vmode; 1530 struct timespec mtime; 1531 struct vnode *nvp; 1532 int32_t rdev; 1533 struct nfsnode *np; 1534 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1535 1536 if (v3) { 1537 vtyp = nfsv3tov_type(fp->fa_type); 1538 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1539 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1540 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1541 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1542 } else { 1543 vtyp = nfsv2tov_type(fp->fa_type); 1544 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1545 if (vtyp == VNON || vtyp == VREG) 1546 vtyp = IFTOVT(vmode); 1547 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1548 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1549 1550 /* 1551 * Really ugly NFSv2 kludge. 1552 */ 1553 if (vtyp == VCHR && rdev == 0xffffffff) 1554 vtyp = VFIFO; 1555 } 1556 1557 /* 1558 * If v_type == VNON it is a new node, so fill in the v_type, 1559 * n_mtime fields. Check to see if it represents a special 1560 * device, and if so, check for a possible alias. Once the 1561 * correct vnode has been obtained, fill in the rest of the 1562 * information. 1563 */ 1564 np = VTONFS(vp); 1565 if (vp->v_type != vtyp) { 1566 vp->v_type = vtyp; 1567 if (vp->v_type == VFIFO) { 1568 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1569 vp->v_op = fifo_nfsv2nodeop_p; 1570 } 1571 if (vp->v_type == VCHR || vp->v_type == VBLK) { 1572 vp->v_op = spec_nfsv2nodeop_p; 1573 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1574 if (nvp) { 1575 /* 1576 * Discard unneeded vnode, but save its nfsnode. 1577 * Since the nfsnode does not have a lock, its 1578 * vnode lock has to be carried over. 1579 */ 1580 nvp->v_data = vp->v_data; 1581 vp->v_data = NULL; 1582 vp->v_op = spec_vnodeop_p; 1583 vput(vp); 1584 vgone(vp); 1585 /* 1586 * XXX When nfs starts locking, we need to 1587 * lock the new node here. 1588 */ 1589 /* 1590 * Reinitialize aliased node. 1591 */ 1592 np->n_vnode = nvp; 1593 *vpp = vp = nvp; 1594 } 1595 } 1596 np->n_mtime = mtime.tv_sec; 1597 } 1598 vap = np->n_vattr; 1599 vap->va_type = vtyp; 1600 vap->va_mode = vmode & ALLPERMS; 1601 vap->va_rdev = (dev_t)rdev; 1602 vap->va_mtime = mtime; 1603 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 1604 switch (vtyp) { 1605 case VDIR: 1606 vap->va_blocksize = NFS_DIRFRAGSIZ; 1607 break; 1608 case VBLK: 1609 vap->va_blocksize = BLKDEV_IOSIZE; 1610 break; 1611 case VCHR: 1612 vap->va_blocksize = MAXBSIZE; 1613 break; 1614 default: 1615 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1616 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1617 break; 1618 } 1619 if (v3) { 1620 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1621 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1622 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1623 vap->va_size = fxdr_hyper(&fp->fa3_size); 1624 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1625 vap->va_fileid = fxdr_unsigned(int32_t, 1626 fp->fa3_fileid.nfsuquad[1]); 1627 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1628 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); 1629 vap->va_flags = 0; 1630 vap->va_filerev = 0; 1631 } else { 1632 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1633 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1634 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1635 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1636 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1637 * NFS_FABLKSIZE; 1638 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1639 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1640 vap->va_flags = 0; 1641 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, 1642 fp->fa2_ctime.nfsv2_sec); 1643 vap->va_ctime.tv_nsec = 0; 1644 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1645 vap->va_filerev = 0; 1646 } 1647 if (vap->va_size != np->n_size) { 1648 if (vap->va_type == VREG) { 1649 if (np->n_flag & NMODIFIED) { 1650 if (vap->va_size < np->n_size) 1651 vap->va_size = np->n_size; 1652 else 1653 np->n_size = vap->va_size; 1654 } else 1655 np->n_size = vap->va_size; 1656 uvm_vnp_setsize(vp, np->n_size); 1657 } else 1658 np->n_size = vap->va_size; 1659 } 1660 np->n_attrstamp = time.tv_sec; 1661 if (vaper != NULL) { 1662 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1663 if (np->n_flag & NCHG) { 1664 if (np->n_flag & NACC) 1665 vaper->va_atime = np->n_atim; 1666 if (np->n_flag & NUPD) 1667 vaper->va_mtime = np->n_mtim; 1668 } 1669 } 1670 return (0); 1671 } 1672 1673 /* 1674 * Check the time stamp 1675 * If the cache is valid, copy contents to *vap and return 0 1676 * otherwise return an error 1677 */ 1678 int 1679 nfs_getattrcache(vp, vaper) 1680 struct vnode *vp; 1681 struct vattr *vaper; 1682 { 1683 struct nfsnode *np = VTONFS(vp); 1684 struct vattr *vap; 1685 1686 if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { 1687 nfsstats.attrcache_misses++; 1688 return (ENOENT); 1689 } 1690 nfsstats.attrcache_hits++; 1691 vap = np->n_vattr; 1692 if (vap->va_size != np->n_size) { 1693 if (vap->va_type == VREG) { 1694 if (np->n_flag & NMODIFIED) { 1695 if (vap->va_size < np->n_size) 1696 vap->va_size = np->n_size; 1697 else 1698 np->n_size = vap->va_size; 1699 } else 1700 np->n_size = vap->va_size; 1701 uvm_vnp_setsize(vp, np->n_size); 1702 } else 1703 np->n_size = vap->va_size; 1704 } 1705 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1706 if (np->n_flag & NCHG) { 1707 if (np->n_flag & NACC) 1708 vaper->va_atime = np->n_atim; 1709 if (np->n_flag & NUPD) 1710 vaper->va_mtime = np->n_mtim; 1711 } 1712 return (0); 1713 } 1714 1715 /* 1716 * Heuristic to see if the server XDR encodes directory cookies or not. 1717 * it is not supposed to, but a lot of servers may do this. Also, since 1718 * most/all servers will implement V2 as well, it is expected that they 1719 * may return just 32 bits worth of cookie information, so we need to 1720 * find out in which 32 bits this information is available. We do this 1721 * to avoid trouble with emulated binaries that can't handle 64 bit 1722 * directory offsets. 1723 */ 1724 1725 void 1726 nfs_cookieheuristic(vp, flagp, p, cred) 1727 struct vnode *vp; 1728 int *flagp; 1729 struct proc *p; 1730 struct ucred *cred; 1731 { 1732 struct uio auio; 1733 struct iovec aiov; 1734 caddr_t buf, cp; 1735 struct dirent *dp; 1736 off_t *cookies = NULL, *cop; 1737 int error, eof, nc, len; 1738 1739 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 1740 1741 aiov.iov_base = buf; 1742 aiov.iov_len = NFS_DIRFRAGSIZ; 1743 auio.uio_iov = &aiov; 1744 auio.uio_iovcnt = 1; 1745 auio.uio_rw = UIO_READ; 1746 auio.uio_segflg = UIO_SYSSPACE; 1747 auio.uio_procp = p; 1748 auio.uio_resid = NFS_DIRFRAGSIZ; 1749 auio.uio_offset = 0; 1750 1751 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 1752 1753 len = NFS_DIRFRAGSIZ - auio.uio_resid; 1754 if (error || len == 0) { 1755 FREE(buf, M_TEMP); 1756 if (cookies) 1757 FREE(cookies, M_TEMP); 1758 return; 1759 } 1760 1761 /* 1762 * Find the first valid entry and look at its offset cookie. 1763 */ 1764 1765 cp = buf; 1766 for (cop = cookies; len > 0; len -= dp->d_reclen) { 1767 dp = (struct dirent *)cp; 1768 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 1769 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 1770 *flagp |= NFSMNT_SWAPCOOKIE; 1771 nfs_invaldircache(vp, 0); 1772 nfs_vinvalbuf(vp, 0, cred, p, 1); 1773 } 1774 break; 1775 } 1776 cop++; 1777 cp += dp->d_reclen; 1778 } 1779 1780 FREE(buf, M_TEMP); 1781 FREE(cookies, M_TEMP); 1782 } 1783 #endif /* NFS */ 1784 1785 /* 1786 * Set up nameidata for a lookup() call and do it. 1787 * 1788 * If pubflag is set, this call is done for a lookup operation on the 1789 * public filehandle. In that case we allow crossing mountpoints and 1790 * absolute pathnames. However, the caller is expected to check that 1791 * the lookup result is within the public fs, and deny access if 1792 * it is not. 1793 */ 1794 int 1795 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) 1796 struct nameidata *ndp; 1797 fhandle_t *fhp; 1798 int len; 1799 struct nfssvc_sock *slp; 1800 struct mbuf *nam; 1801 struct mbuf **mdp; 1802 caddr_t *dposp; 1803 struct vnode **retdirp; 1804 struct proc *p; 1805 int kerbflag, pubflag; 1806 { 1807 int i, rem; 1808 struct mbuf *md; 1809 char *fromcp, *tocp, *cp; 1810 struct iovec aiov; 1811 struct uio auio; 1812 struct vnode *dp; 1813 int error, rdonly, linklen; 1814 struct componentname *cnp = &ndp->ni_cnd; 1815 1816 *retdirp = (struct vnode *)0; 1817 MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); 1818 /* 1819 * Copy the name from the mbuf list to ndp->ni_pnbuf 1820 * and set the various ndp fields appropriately. 1821 */ 1822 fromcp = *dposp; 1823 tocp = cnp->cn_pnbuf; 1824 md = *mdp; 1825 rem = mtod(md, caddr_t) + md->m_len - fromcp; 1826 for (i = 0; i < len; i++) { 1827 while (rem == 0) { 1828 md = md->m_next; 1829 if (md == NULL) { 1830 error = EBADRPC; 1831 goto out; 1832 } 1833 fromcp = mtod(md, caddr_t); 1834 rem = md->m_len; 1835 } 1836 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 1837 error = EACCES; 1838 goto out; 1839 } 1840 *tocp++ = *fromcp++; 1841 rem--; 1842 } 1843 *tocp = '\0'; 1844 *mdp = md; 1845 *dposp = fromcp; 1846 len = nfsm_rndup(len)-len; 1847 if (len > 0) { 1848 if (rem >= len) 1849 *dposp += len; 1850 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 1851 goto out; 1852 } 1853 1854 /* 1855 * Extract and set starting directory. 1856 */ 1857 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 1858 nam, &rdonly, kerbflag, pubflag); 1859 if (error) 1860 goto out; 1861 if (dp->v_type != VDIR) { 1862 vrele(dp); 1863 error = ENOTDIR; 1864 goto out; 1865 } 1866 1867 if (rdonly) 1868 cnp->cn_flags |= RDONLY; 1869 1870 *retdirp = dp; 1871 1872 if (pubflag) { 1873 /* 1874 * Oh joy. For WebNFS, handle those pesky '%' escapes, 1875 * and the 'native path' indicator. 1876 */ 1877 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); 1878 fromcp = cnp->cn_pnbuf; 1879 tocp = cp; 1880 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 1881 switch ((unsigned char)*fromcp) { 1882 case WEBNFS_NATIVE_CHAR: 1883 /* 1884 * 'Native' path for us is the same 1885 * as a path according to the NFS spec, 1886 * just skip the escape char. 1887 */ 1888 fromcp++; 1889 break; 1890 /* 1891 * More may be added in the future, range 0x80-0xff 1892 */ 1893 default: 1894 error = EIO; 1895 FREE(cp, M_NAMEI); 1896 goto out; 1897 } 1898 } 1899 /* 1900 * Translate the '%' escapes, URL-style. 1901 */ 1902 while (*fromcp != '\0') { 1903 if (*fromcp == WEBNFS_ESC_CHAR) { 1904 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 1905 fromcp++; 1906 *tocp++ = HEXSTRTOI(fromcp); 1907 fromcp += 2; 1908 continue; 1909 } else { 1910 error = ENOENT; 1911 FREE(cp, M_NAMEI); 1912 goto out; 1913 } 1914 } else 1915 *tocp++ = *fromcp++; 1916 } 1917 *tocp = '\0'; 1918 FREE(cnp->cn_pnbuf, M_NAMEI); 1919 cnp->cn_pnbuf = cp; 1920 } 1921 1922 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 1923 ndp->ni_segflg = UIO_SYSSPACE; 1924 1925 if (pubflag) { 1926 ndp->ni_rootdir = rootvnode; 1927 ndp->ni_loopcnt = 0; 1928 if (cnp->cn_pnbuf[0] == '/') 1929 dp = rootvnode; 1930 } else { 1931 cnp->cn_flags |= NOCROSSMOUNT; 1932 } 1933 1934 cnp->cn_proc = p; 1935 VREF(dp); 1936 1937 for (;;) { 1938 cnp->cn_nameptr = cnp->cn_pnbuf; 1939 ndp->ni_startdir = dp; 1940 /* 1941 * And call lookup() to do the real work 1942 */ 1943 error = lookup(ndp); 1944 if (error) 1945 break; 1946 /* 1947 * Check for encountering a symbolic link 1948 */ 1949 if ((cnp->cn_flags & ISSYMLINK) == 0) { 1950 if (cnp->cn_flags & (SAVENAME | SAVESTART)) { 1951 cnp->cn_flags |= HASBUF; 1952 return (0); 1953 } 1954 break; 1955 } else { 1956 if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) 1957 VOP_UNLOCK(ndp->ni_dvp, 0); 1958 if (!pubflag) { 1959 vrele(ndp->ni_dvp); 1960 vput(ndp->ni_vp); 1961 ndp->ni_vp = NULL; 1962 error = EINVAL; 1963 break; 1964 } 1965 1966 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 1967 error = ELOOP; 1968 break; 1969 } 1970 if (ndp->ni_pathlen > 1) 1971 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); 1972 else 1973 cp = cnp->cn_pnbuf; 1974 aiov.iov_base = cp; 1975 aiov.iov_len = MAXPATHLEN; 1976 auio.uio_iov = &aiov; 1977 auio.uio_iovcnt = 1; 1978 auio.uio_offset = 0; 1979 auio.uio_rw = UIO_READ; 1980 auio.uio_segflg = UIO_SYSSPACE; 1981 auio.uio_procp = (struct proc *)0; 1982 auio.uio_resid = MAXPATHLEN; 1983 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 1984 if (error) { 1985 badlink: 1986 if (ndp->ni_pathlen > 1) 1987 FREE(cp, M_NAMEI); 1988 break; 1989 } 1990 linklen = MAXPATHLEN - auio.uio_resid; 1991 if (linklen == 0) { 1992 error = ENOENT; 1993 goto badlink; 1994 } 1995 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 1996 error = ENAMETOOLONG; 1997 goto badlink; 1998 } 1999 if (ndp->ni_pathlen > 1) { 2000 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 2001 FREE(cnp->cn_pnbuf, M_NAMEI); 2002 cnp->cn_pnbuf = cp; 2003 } else 2004 cnp->cn_pnbuf[linklen] = '\0'; 2005 ndp->ni_pathlen += linklen; 2006 vput(ndp->ni_vp); 2007 dp = ndp->ni_dvp; 2008 /* 2009 * Check if root directory should replace current directory. 2010 */ 2011 if (cnp->cn_pnbuf[0] == '/') { 2012 vrele(dp); 2013 dp = ndp->ni_rootdir; 2014 VREF(dp); 2015 } 2016 } 2017 } 2018 out: 2019 FREE(cnp->cn_pnbuf, M_NAMEI); 2020 return (error); 2021 } 2022 2023 /* 2024 * A fiddled version of m_adj() that ensures null fill to a long 2025 * boundary and only trims off the back end 2026 */ 2027 void 2028 nfsm_adj(mp, len, nul) 2029 struct mbuf *mp; 2030 int len; 2031 int nul; 2032 { 2033 struct mbuf *m; 2034 int count, i; 2035 char *cp; 2036 2037 /* 2038 * Trim from tail. Scan the mbuf chain, 2039 * calculating its length and finding the last mbuf. 2040 * If the adjustment only affects this mbuf, then just 2041 * adjust and return. Otherwise, rescan and truncate 2042 * after the remaining size. 2043 */ 2044 count = 0; 2045 m = mp; 2046 for (;;) { 2047 count += m->m_len; 2048 if (m->m_next == (struct mbuf *)0) 2049 break; 2050 m = m->m_next; 2051 } 2052 if (m->m_len > len) { 2053 m->m_len -= len; 2054 if (nul > 0) { 2055 cp = mtod(m, caddr_t)+m->m_len-nul; 2056 for (i = 0; i < nul; i++) 2057 *cp++ = '\0'; 2058 } 2059 return; 2060 } 2061 count -= len; 2062 if (count < 0) 2063 count = 0; 2064 /* 2065 * Correct length for chain is "count". 2066 * Find the mbuf with last data, adjust its length, 2067 * and toss data from remaining mbufs on chain. 2068 */ 2069 for (m = mp; m; m = m->m_next) { 2070 if (m->m_len >= count) { 2071 m->m_len = count; 2072 if (nul > 0) { 2073 cp = mtod(m, caddr_t)+m->m_len-nul; 2074 for (i = 0; i < nul; i++) 2075 *cp++ = '\0'; 2076 } 2077 break; 2078 } 2079 count -= m->m_len; 2080 } 2081 for (m = m->m_next;m;m = m->m_next) 2082 m->m_len = 0; 2083 } 2084 2085 /* 2086 * Make these functions instead of macros, so that the kernel text size 2087 * doesn't get too big... 2088 */ 2089 void 2090 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2091 struct nfsrv_descript *nfsd; 2092 int before_ret; 2093 struct vattr *before_vap; 2094 int after_ret; 2095 struct vattr *after_vap; 2096 struct mbuf **mbp; 2097 char **bposp; 2098 { 2099 struct mbuf *mb = *mbp, *mb2; 2100 char *bpos = *bposp; 2101 u_int32_t *tl; 2102 2103 if (before_ret) { 2104 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2105 *tl = nfs_false; 2106 } else { 2107 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2108 *tl++ = nfs_true; 2109 txdr_hyper(before_vap->va_size, tl); 2110 tl += 2; 2111 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2112 tl += 2; 2113 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2114 } 2115 *bposp = bpos; 2116 *mbp = mb; 2117 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2118 } 2119 2120 void 2121 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2122 struct nfsrv_descript *nfsd; 2123 int after_ret; 2124 struct vattr *after_vap; 2125 struct mbuf **mbp; 2126 char **bposp; 2127 { 2128 struct mbuf *mb = *mbp, *mb2; 2129 char *bpos = *bposp; 2130 u_int32_t *tl; 2131 struct nfs_fattr *fp; 2132 2133 if (after_ret) { 2134 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2135 *tl = nfs_false; 2136 } else { 2137 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2138 *tl++ = nfs_true; 2139 fp = (struct nfs_fattr *)tl; 2140 nfsm_srvfattr(nfsd, after_vap, fp); 2141 } 2142 *mbp = mb; 2143 *bposp = bpos; 2144 } 2145 2146 void 2147 nfsm_srvfattr(nfsd, vap, fp) 2148 struct nfsrv_descript *nfsd; 2149 struct vattr *vap; 2150 struct nfs_fattr *fp; 2151 { 2152 2153 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2154 fp->fa_uid = txdr_unsigned(vap->va_uid); 2155 fp->fa_gid = txdr_unsigned(vap->va_gid); 2156 if (nfsd->nd_flag & ND_NFSV3) { 2157 fp->fa_type = vtonfsv3_type(vap->va_type); 2158 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2159 txdr_hyper(vap->va_size, &fp->fa3_size); 2160 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2161 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2162 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2163 fp->fa3_fsid.nfsuquad[0] = 0; 2164 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2165 fp->fa3_fileid.nfsuquad[0] = 0; 2166 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); 2167 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2168 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2169 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2170 } else { 2171 fp->fa_type = vtonfsv2_type(vap->va_type); 2172 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2173 fp->fa2_size = txdr_unsigned(vap->va_size); 2174 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2175 if (vap->va_type == VFIFO) 2176 fp->fa2_rdev = 0xffffffff; 2177 else 2178 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2179 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2180 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2181 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2182 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2183 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2184 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2185 } 2186 } 2187 2188 /* 2189 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2190 * - look up fsid in mount list (if not found ret error) 2191 * - get vp and export rights by calling VFS_FHTOVP() 2192 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2193 * - if not lockflag unlock it with VOP_UNLOCK() 2194 */ 2195 int 2196 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) 2197 fhandle_t *fhp; 2198 int lockflag; 2199 struct vnode **vpp; 2200 struct ucred *cred; 2201 struct nfssvc_sock *slp; 2202 struct mbuf *nam; 2203 int *rdonlyp; 2204 int kerbflag; 2205 { 2206 struct mount *mp; 2207 int i; 2208 struct ucred *credanon; 2209 int error, exflags; 2210 struct sockaddr_in *saddr; 2211 2212 *vpp = (struct vnode *)0; 2213 2214 if (nfs_ispublicfh(fhp)) { 2215 if (!pubflag || !nfs_pub.np_valid) 2216 return (ESTALE); 2217 fhp = &nfs_pub.np_handle; 2218 } 2219 2220 mp = vfs_getvfs(&fhp->fh_fsid); 2221 if (!mp) 2222 return (ESTALE); 2223 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); 2224 if (error) 2225 return (error); 2226 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2227 if (error) 2228 return (error); 2229 2230 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2231 saddr = mtod(nam, struct sockaddr_in *); 2232 if ((saddr->sin_family == AF_INET) && 2233 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2234 vput(*vpp); 2235 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2236 } 2237 #ifdef INET6 2238 if ((saddr->sin_family == AF_INET6) && 2239 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2240 vput(*vpp); 2241 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2242 } 2243 #endif 2244 } 2245 /* 2246 * Check/setup credentials. 2247 */ 2248 if (exflags & MNT_EXKERB) { 2249 if (!kerbflag) { 2250 vput(*vpp); 2251 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2252 } 2253 } else if (kerbflag) { 2254 vput(*vpp); 2255 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2256 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { 2257 cred->cr_uid = credanon->cr_uid; 2258 cred->cr_gid = credanon->cr_gid; 2259 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) 2260 cred->cr_groups[i] = credanon->cr_groups[i]; 2261 cred->cr_ngroups = i; 2262 } 2263 if (exflags & MNT_EXRDONLY) 2264 *rdonlyp = 1; 2265 else 2266 *rdonlyp = 0; 2267 if (!lockflag) 2268 VOP_UNLOCK(*vpp, 0); 2269 return (0); 2270 } 2271 2272 /* 2273 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2274 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has 2275 * transformed this to all zeroes in both cases, so check for it. 2276 */ 2277 int 2278 nfs_ispublicfh(fhp) 2279 fhandle_t *fhp; 2280 { 2281 char *cp = (char *)fhp; 2282 int i; 2283 2284 for (i = 0; i < NFSX_V3FH; i++) 2285 if (*cp++ != 0) 2286 return (FALSE); 2287 return (TRUE); 2288 } 2289 2290 /* 2291 * This function compares two net addresses by family and returns TRUE 2292 * if they are the same host. 2293 * If there is any doubt, return FALSE. 2294 * The AF_INET family is handled as a special case so that address mbufs 2295 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2296 */ 2297 int 2298 netaddr_match(family, haddr, nam) 2299 int family; 2300 union nethostaddr *haddr; 2301 struct mbuf *nam; 2302 { 2303 struct sockaddr_in *inetaddr; 2304 2305 switch (family) { 2306 case AF_INET: 2307 inetaddr = mtod(nam, struct sockaddr_in *); 2308 if (inetaddr->sin_family == AF_INET && 2309 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2310 return (1); 2311 break; 2312 #ifdef INET6 2313 case AF_INET6: 2314 { 2315 struct sockaddr_in6 *sin6_1, *sin6_2; 2316 2317 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2318 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2319 if (sin6_1->sin6_family == AF_INET6 && 2320 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2321 return 1; 2322 } 2323 #endif 2324 #ifdef ISO 2325 case AF_ISO: 2326 { 2327 struct sockaddr_iso *isoaddr1, *isoaddr2; 2328 2329 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2330 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2331 if (isoaddr1->siso_family == AF_ISO && 2332 isoaddr1->siso_nlen > 0 && 2333 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2334 SAME_ISOADDR(isoaddr1, isoaddr2)) 2335 return (1); 2336 break; 2337 } 2338 #endif /* ISO */ 2339 default: 2340 break; 2341 }; 2342 return (0); 2343 } 2344 2345 2346 /* 2347 * The write verifier has changed (probably due to a server reboot), so all 2348 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 2349 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 2350 * flag. Once done the new write verifier can be set for the mount point. 2351 */ 2352 void 2353 nfs_clearcommit(mp) 2354 struct mount *mp; 2355 { 2356 struct vnode *vp, *nvp; 2357 struct buf *bp, *nbp; 2358 int s; 2359 2360 s = splbio(); 2361 loop: 2362 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 2363 if (vp->v_mount != mp) /* Paranoia */ 2364 goto loop; 2365 nvp = vp->v_mntvnodes.le_next; 2366 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2367 nbp = bp->b_vnbufs.le_next; 2368 if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) 2369 == (B_DELWRI | B_NEEDCOMMIT)) 2370 bp->b_flags &= ~B_NEEDCOMMIT; 2371 } 2372 } 2373 splx(s); 2374 } 2375 2376 /* 2377 * Map errnos to NFS error numbers. For Version 3 also filter out error 2378 * numbers not specified for the associated procedure. 2379 */ 2380 int 2381 nfsrv_errmap(nd, err) 2382 struct nfsrv_descript *nd; 2383 int err; 2384 { 2385 short *defaulterrp, *errp; 2386 2387 if (nd->nd_flag & ND_NFSV3) { 2388 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2389 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2390 while (*++errp) { 2391 if (*errp == err) 2392 return (err); 2393 else if (*errp > err) 2394 break; 2395 } 2396 return ((int)*defaulterrp); 2397 } else 2398 return (err & 0xffff); 2399 } 2400 if (err <= ELAST) 2401 return ((int)nfsrv_v2errmap[err - 1]); 2402 return (NFSERR_IO); 2403 } 2404 2405 /* 2406 * Sort the group list in increasing numerical order. 2407 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort 2408 * that used to be here.) 2409 */ 2410 void 2411 nfsrvw_sort(list, num) 2412 gid_t *list; 2413 int num; 2414 { 2415 int i, j; 2416 gid_t v; 2417 2418 /* Insertion sort. */ 2419 for (i = 1; i < num; i++) { 2420 v = list[i]; 2421 /* find correct slot for value v, moving others up */ 2422 for (j = i; --j >= 0 && v < list[j];) 2423 list[j + 1] = list[j]; 2424 list[j + 1] = v; 2425 } 2426 } 2427 2428 /* 2429 * copy credentials making sure that the result can be compared with memcmp(). 2430 */ 2431 void 2432 nfsrv_setcred(incred, outcred) 2433 struct ucred *incred, *outcred; 2434 { 2435 int i; 2436 2437 memset((caddr_t)outcred, 0, sizeof (struct ucred)); 2438 outcred->cr_ref = 1; 2439 outcred->cr_uid = incred->cr_uid; 2440 outcred->cr_gid = incred->cr_gid; 2441 outcred->cr_ngroups = incred->cr_ngroups; 2442 for (i = 0; i < incred->cr_ngroups; i++) 2443 outcred->cr_groups[i] = incred->cr_groups[i]; 2444 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); 2445 } 2446
Indexes created Wed Sep 24 05:09:52 GMT 2025