nfs_subs.c revision 1.82
1 /* $NetBSD: nfs_subs.c,v 1.82 2000/09/19 17:04:51 bjh21 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_nfs.h" 43 #include "opt_nfsserver.h" 44 #include "opt_iso.h" 45 #include "opt_inet.h" 46 47 /* 48 * These functions support the macros and help fiddle mbuf chains for 49 * the nfs op functions. They do things like create the rpc header and 50 * copy data between mbuf chains and uio lists. 51 */ 52 #include <sys/param.h> 53 #include <sys/proc.h> 54 #include <sys/systm.h> 55 #include <sys/kernel.h> 56 #include <sys/mount.h> 57 #include <sys/vnode.h> 58 #include <sys/namei.h> 59 #include <sys/mbuf.h> 60 #include <sys/socket.h> 61 #include <sys/stat.h> 62 #include <sys/malloc.h> 63 #include <sys/time.h> 64 #include <sys/dirent.h> 65 66 #include <uvm/uvm_extern.h> 67 68 #include <nfs/rpcv2.h> 69 #include <nfs/nfsproto.h> 70 #include <nfs/nfsnode.h> 71 #include <nfs/nfs.h> 72 #include <nfs/xdr_subs.h> 73 #include <nfs/nfsm_subs.h> 74 #include <nfs/nfsmount.h> 75 #include <nfs/nqnfs.h> 76 #include <nfs/nfsrtt.h> 77 #include <nfs/nfs_var.h> 78 79 #include <miscfs/specfs/specdev.h> 80 81 #include <netinet/in.h> 82 #ifdef ISO 83 #include <netiso/iso.h> 84 #endif 85 86 /* 87 * Data items converted to xdr at startup, since they are constant 88 * This is kinda hokey, but may save a little time doing byte swaps 89 */ 90 u_int32_t nfs_xdrneg1; 91 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 92 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 93 rpc_auth_kerb; 94 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; 95 96 /* And other global data */ 97 static u_int32_t nfs_xid = 0; 98 nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, 99 NFCHR, NFNON }; 100 nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, 101 NFFIFO, NFNON }; 102 enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 103 enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 104 int nfs_ticks; 105 106 /* NFS client/server stats. */ 107 struct nfsstats nfsstats; 108 109 /* 110 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 111 */ 112 int nfsv3_procid[NFS_NPROCS] = { 113 NFSPROC_NULL, 114 NFSPROC_GETATTR, 115 NFSPROC_SETATTR, 116 NFSPROC_NOOP, 117 NFSPROC_LOOKUP, 118 NFSPROC_READLINK, 119 NFSPROC_READ, 120 NFSPROC_NOOP, 121 NFSPROC_WRITE, 122 NFSPROC_CREATE, 123 NFSPROC_REMOVE, 124 NFSPROC_RENAME, 125 NFSPROC_LINK, 126 NFSPROC_SYMLINK, 127 NFSPROC_MKDIR, 128 NFSPROC_RMDIR, 129 NFSPROC_READDIR, 130 NFSPROC_FSSTAT, 131 NFSPROC_NOOP, 132 NFSPROC_NOOP, 133 NFSPROC_NOOP, 134 NFSPROC_NOOP, 135 NFSPROC_NOOP, 136 NFSPROC_NOOP, 137 NFSPROC_NOOP, 138 NFSPROC_NOOP 139 }; 140 141 /* 142 * and the reverse mapping from generic to Version 2 procedure numbers 143 */ 144 int nfsv2_procid[NFS_NPROCS] = { 145 NFSV2PROC_NULL, 146 NFSV2PROC_GETATTR, 147 NFSV2PROC_SETATTR, 148 NFSV2PROC_LOOKUP, 149 NFSV2PROC_NOOP, 150 NFSV2PROC_READLINK, 151 NFSV2PROC_READ, 152 NFSV2PROC_WRITE, 153 NFSV2PROC_CREATE, 154 NFSV2PROC_MKDIR, 155 NFSV2PROC_SYMLINK, 156 NFSV2PROC_CREATE, 157 NFSV2PROC_REMOVE, 158 NFSV2PROC_RMDIR, 159 NFSV2PROC_RENAME, 160 NFSV2PROC_LINK, 161 NFSV2PROC_READDIR, 162 NFSV2PROC_NOOP, 163 NFSV2PROC_STATFS, 164 NFSV2PROC_NOOP, 165 NFSV2PROC_NOOP, 166 NFSV2PROC_NOOP, 167 NFSV2PROC_NOOP, 168 NFSV2PROC_NOOP, 169 NFSV2PROC_NOOP, 170 NFSV2PROC_NOOP, 171 }; 172 173 /* 174 * Maps errno values to nfs error numbers. 175 * Use NFSERR_IO as the catch all for ones not specifically defined in 176 * RFC 1094. 177 */ 178 static u_char nfsrv_v2errmap[ELAST] = { 179 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 180 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 181 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 182 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 183 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 184 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 185 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 186 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 187 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 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_NAMETOL, NFSERR_IO, NFSERR_IO, 192 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 193 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 194 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 195 NFSERR_IO, NFSERR_IO, 196 }; 197 198 /* 199 * Maps errno values to nfs error numbers. 200 * Although it is not obvious whether or not NFS clients really care if 201 * a returned error value is in the specified list for the procedure, the 202 * safest thing to do is filter them appropriately. For Version 2, the 203 * X/Open XNFS document is the only specification that defines error values 204 * for each RPC (The RFC simply lists all possible error values for all RPCs), 205 * so I have decided to not do this for Version 2. 206 * The first entry is the default error return and the rest are the valid 207 * errors for that RPC in increasing numeric order. 208 */ 209 static short nfsv3err_null[] = { 210 0, 211 0, 212 }; 213 214 static short nfsv3err_getattr[] = { 215 NFSERR_IO, 216 NFSERR_IO, 217 NFSERR_STALE, 218 NFSERR_BADHANDLE, 219 NFSERR_SERVERFAULT, 220 0, 221 }; 222 223 static short nfsv3err_setattr[] = { 224 NFSERR_IO, 225 NFSERR_PERM, 226 NFSERR_IO, 227 NFSERR_ACCES, 228 NFSERR_INVAL, 229 NFSERR_NOSPC, 230 NFSERR_ROFS, 231 NFSERR_DQUOT, 232 NFSERR_STALE, 233 NFSERR_BADHANDLE, 234 NFSERR_NOT_SYNC, 235 NFSERR_SERVERFAULT, 236 0, 237 }; 238 239 static short nfsv3err_lookup[] = { 240 NFSERR_IO, 241 NFSERR_NOENT, 242 NFSERR_IO, 243 NFSERR_ACCES, 244 NFSERR_NOTDIR, 245 NFSERR_NAMETOL, 246 NFSERR_STALE, 247 NFSERR_BADHANDLE, 248 NFSERR_SERVERFAULT, 249 0, 250 }; 251 252 static short nfsv3err_access[] = { 253 NFSERR_IO, 254 NFSERR_IO, 255 NFSERR_STALE, 256 NFSERR_BADHANDLE, 257 NFSERR_SERVERFAULT, 258 0, 259 }; 260 261 static short nfsv3err_readlink[] = { 262 NFSERR_IO, 263 NFSERR_IO, 264 NFSERR_ACCES, 265 NFSERR_INVAL, 266 NFSERR_STALE, 267 NFSERR_BADHANDLE, 268 NFSERR_NOTSUPP, 269 NFSERR_SERVERFAULT, 270 0, 271 }; 272 273 static short nfsv3err_read[] = { 274 NFSERR_IO, 275 NFSERR_IO, 276 NFSERR_NXIO, 277 NFSERR_ACCES, 278 NFSERR_INVAL, 279 NFSERR_STALE, 280 NFSERR_BADHANDLE, 281 NFSERR_SERVERFAULT, 282 NFSERR_JUKEBOX, 283 0, 284 }; 285 286 static short nfsv3err_write[] = { 287 NFSERR_IO, 288 NFSERR_IO, 289 NFSERR_ACCES, 290 NFSERR_INVAL, 291 NFSERR_FBIG, 292 NFSERR_NOSPC, 293 NFSERR_ROFS, 294 NFSERR_DQUOT, 295 NFSERR_STALE, 296 NFSERR_BADHANDLE, 297 NFSERR_SERVERFAULT, 298 NFSERR_JUKEBOX, 299 0, 300 }; 301 302 static short nfsv3err_create[] = { 303 NFSERR_IO, 304 NFSERR_IO, 305 NFSERR_ACCES, 306 NFSERR_EXIST, 307 NFSERR_NOTDIR, 308 NFSERR_NOSPC, 309 NFSERR_ROFS, 310 NFSERR_NAMETOL, 311 NFSERR_DQUOT, 312 NFSERR_STALE, 313 NFSERR_BADHANDLE, 314 NFSERR_NOTSUPP, 315 NFSERR_SERVERFAULT, 316 0, 317 }; 318 319 static short nfsv3err_mkdir[] = { 320 NFSERR_IO, 321 NFSERR_IO, 322 NFSERR_ACCES, 323 NFSERR_EXIST, 324 NFSERR_NOTDIR, 325 NFSERR_NOSPC, 326 NFSERR_ROFS, 327 NFSERR_NAMETOL, 328 NFSERR_DQUOT, 329 NFSERR_STALE, 330 NFSERR_BADHANDLE, 331 NFSERR_NOTSUPP, 332 NFSERR_SERVERFAULT, 333 0, 334 }; 335 336 static short nfsv3err_symlink[] = { 337 NFSERR_IO, 338 NFSERR_IO, 339 NFSERR_ACCES, 340 NFSERR_EXIST, 341 NFSERR_NOTDIR, 342 NFSERR_NOSPC, 343 NFSERR_ROFS, 344 NFSERR_NAMETOL, 345 NFSERR_DQUOT, 346 NFSERR_STALE, 347 NFSERR_BADHANDLE, 348 NFSERR_NOTSUPP, 349 NFSERR_SERVERFAULT, 350 0, 351 }; 352 353 static short nfsv3err_mknod[] = { 354 NFSERR_IO, 355 NFSERR_IO, 356 NFSERR_ACCES, 357 NFSERR_EXIST, 358 NFSERR_NOTDIR, 359 NFSERR_NOSPC, 360 NFSERR_ROFS, 361 NFSERR_NAMETOL, 362 NFSERR_DQUOT, 363 NFSERR_STALE, 364 NFSERR_BADHANDLE, 365 NFSERR_NOTSUPP, 366 NFSERR_SERVERFAULT, 367 NFSERR_BADTYPE, 368 0, 369 }; 370 371 static short nfsv3err_remove[] = { 372 NFSERR_IO, 373 NFSERR_NOENT, 374 NFSERR_IO, 375 NFSERR_ACCES, 376 NFSERR_NOTDIR, 377 NFSERR_ROFS, 378 NFSERR_NAMETOL, 379 NFSERR_STALE, 380 NFSERR_BADHANDLE, 381 NFSERR_SERVERFAULT, 382 0, 383 }; 384 385 static short nfsv3err_rmdir[] = { 386 NFSERR_IO, 387 NFSERR_NOENT, 388 NFSERR_IO, 389 NFSERR_ACCES, 390 NFSERR_EXIST, 391 NFSERR_NOTDIR, 392 NFSERR_INVAL, 393 NFSERR_ROFS, 394 NFSERR_NAMETOL, 395 NFSERR_NOTEMPTY, 396 NFSERR_STALE, 397 NFSERR_BADHANDLE, 398 NFSERR_NOTSUPP, 399 NFSERR_SERVERFAULT, 400 0, 401 }; 402 403 static short nfsv3err_rename[] = { 404 NFSERR_IO, 405 NFSERR_NOENT, 406 NFSERR_IO, 407 NFSERR_ACCES, 408 NFSERR_EXIST, 409 NFSERR_XDEV, 410 NFSERR_NOTDIR, 411 NFSERR_ISDIR, 412 NFSERR_INVAL, 413 NFSERR_NOSPC, 414 NFSERR_ROFS, 415 NFSERR_MLINK, 416 NFSERR_NAMETOL, 417 NFSERR_NOTEMPTY, 418 NFSERR_DQUOT, 419 NFSERR_STALE, 420 NFSERR_BADHANDLE, 421 NFSERR_NOTSUPP, 422 NFSERR_SERVERFAULT, 423 0, 424 }; 425 426 static short nfsv3err_link[] = { 427 NFSERR_IO, 428 NFSERR_IO, 429 NFSERR_ACCES, 430 NFSERR_EXIST, 431 NFSERR_XDEV, 432 NFSERR_NOTDIR, 433 NFSERR_INVAL, 434 NFSERR_NOSPC, 435 NFSERR_ROFS, 436 NFSERR_MLINK, 437 NFSERR_NAMETOL, 438 NFSERR_DQUOT, 439 NFSERR_STALE, 440 NFSERR_BADHANDLE, 441 NFSERR_NOTSUPP, 442 NFSERR_SERVERFAULT, 443 0, 444 }; 445 446 static short nfsv3err_readdir[] = { 447 NFSERR_IO, 448 NFSERR_IO, 449 NFSERR_ACCES, 450 NFSERR_NOTDIR, 451 NFSERR_STALE, 452 NFSERR_BADHANDLE, 453 NFSERR_BAD_COOKIE, 454 NFSERR_TOOSMALL, 455 NFSERR_SERVERFAULT, 456 0, 457 }; 458 459 static short nfsv3err_readdirplus[] = { 460 NFSERR_IO, 461 NFSERR_IO, 462 NFSERR_ACCES, 463 NFSERR_NOTDIR, 464 NFSERR_STALE, 465 NFSERR_BADHANDLE, 466 NFSERR_BAD_COOKIE, 467 NFSERR_NOTSUPP, 468 NFSERR_TOOSMALL, 469 NFSERR_SERVERFAULT, 470 0, 471 }; 472 473 static short nfsv3err_fsstat[] = { 474 NFSERR_IO, 475 NFSERR_IO, 476 NFSERR_STALE, 477 NFSERR_BADHANDLE, 478 NFSERR_SERVERFAULT, 479 0, 480 }; 481 482 static short nfsv3err_fsinfo[] = { 483 NFSERR_STALE, 484 NFSERR_STALE, 485 NFSERR_BADHANDLE, 486 NFSERR_SERVERFAULT, 487 0, 488 }; 489 490 static short nfsv3err_pathconf[] = { 491 NFSERR_STALE, 492 NFSERR_STALE, 493 NFSERR_BADHANDLE, 494 NFSERR_SERVERFAULT, 495 0, 496 }; 497 498 static short nfsv3err_commit[] = { 499 NFSERR_IO, 500 NFSERR_IO, 501 NFSERR_STALE, 502 NFSERR_BADHANDLE, 503 NFSERR_SERVERFAULT, 504 0, 505 }; 506 507 static short *nfsrv_v3errmap[] = { 508 nfsv3err_null, 509 nfsv3err_getattr, 510 nfsv3err_setattr, 511 nfsv3err_lookup, 512 nfsv3err_access, 513 nfsv3err_readlink, 514 nfsv3err_read, 515 nfsv3err_write, 516 nfsv3err_create, 517 nfsv3err_mkdir, 518 nfsv3err_symlink, 519 nfsv3err_mknod, 520 nfsv3err_remove, 521 nfsv3err_rmdir, 522 nfsv3err_rename, 523 nfsv3err_link, 524 nfsv3err_readdir, 525 nfsv3err_readdirplus, 526 nfsv3err_fsstat, 527 nfsv3err_fsinfo, 528 nfsv3err_pathconf, 529 nfsv3err_commit, 530 }; 531 532 extern struct nfsrtt nfsrtt; 533 extern time_t nqnfsstarttime; 534 extern int nqsrv_clockskew; 535 extern int nqsrv_writeslack; 536 extern int nqsrv_maxlease; 537 extern int nqnfs_piggy[NFS_NPROCS]; 538 extern struct nfsnodehashhead *nfsnodehashtbl; 539 extern u_long nfsnodehash; 540 541 LIST_HEAD(nfsnodehashhead, nfsnode); 542 u_long nfsdirhashmask; 543 544 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 545 546 /* 547 * Create the header for an rpc request packet 548 * The hsiz is the size of the rest of the nfs request header. 549 * (just used to decide if a cluster is a good idea) 550 */ 551 struct mbuf * 552 nfsm_reqh(vp, procid, hsiz, bposp) 553 struct vnode *vp; 554 u_long procid; 555 int hsiz; 556 caddr_t *bposp; 557 { 558 struct mbuf *mb; 559 u_int32_t *tl; 560 caddr_t bpos; 561 struct mbuf *mb2; 562 struct nfsmount *nmp; 563 int nqflag; 564 565 MGET(mb, M_WAIT, MT_DATA); 566 if (hsiz >= MINCLSIZE) 567 MCLGET(mb, M_WAIT); 568 mb->m_len = 0; 569 bpos = mtod(mb, caddr_t); 570 571 /* 572 * For NQNFS, add lease request. 573 */ 574 if (vp) { 575 nmp = VFSTONFS(vp->v_mount); 576 if (nmp->nm_flag & NFSMNT_NQNFS) { 577 nqflag = NQNFS_NEEDLEASE(vp, procid); 578 if (nqflag) { 579 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); 580 *tl++ = txdr_unsigned(nqflag); 581 *tl = txdr_unsigned(nmp->nm_leaseterm); 582 } else { 583 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 584 *tl = 0; 585 } 586 } 587 } 588 /* Finally, return values */ 589 *bposp = bpos; 590 return (mb); 591 } 592 593 /* 594 * Build the RPC header and fill in the authorization info. 595 * The authorization string argument is only used when the credentials 596 * come from outside of the kernel. 597 * Returns the head of the mbuf list. 598 */ 599 struct mbuf * 600 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 601 verf_str, mrest, mrest_len, mbp, xidp) 602 struct ucred *cr; 603 int nmflag; 604 int procid; 605 int auth_type; 606 int auth_len; 607 char *auth_str; 608 int verf_len; 609 char *verf_str; 610 struct mbuf *mrest; 611 int mrest_len; 612 struct mbuf **mbp; 613 u_int32_t *xidp; 614 { 615 struct mbuf *mb; 616 u_int32_t *tl; 617 caddr_t bpos; 618 int i; 619 struct mbuf *mreq, *mb2; 620 int siz, grpsiz, authsiz; 621 struct timeval tv; 622 static u_int32_t base; 623 624 authsiz = nfsm_rndup(auth_len); 625 MGETHDR(mb, M_WAIT, MT_DATA); 626 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 627 MCLGET(mb, M_WAIT); 628 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 629 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 630 } else { 631 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 632 } 633 mb->m_len = 0; 634 mreq = mb; 635 bpos = mtod(mb, caddr_t); 636 637 /* 638 * First the RPC header. 639 */ 640 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 641 642 /* 643 * derive initial xid from system time 644 * XXX time is invalid if root not yet mounted 645 */ 646 if (!base && (rootvp)) { 647 microtime(&tv); 648 base = tv.tv_sec << 12; 649 nfs_xid = base; 650 } 651 /* 652 * Skip zero xid if it should ever happen. 653 */ 654 if (++nfs_xid == 0) 655 nfs_xid++; 656 657 *tl++ = *xidp = txdr_unsigned(nfs_xid); 658 *tl++ = rpc_call; 659 *tl++ = rpc_vers; 660 if (nmflag & NFSMNT_NQNFS) { 661 *tl++ = txdr_unsigned(NQNFS_PROG); 662 *tl++ = txdr_unsigned(NQNFS_VER3); 663 } else { 664 *tl++ = txdr_unsigned(NFS_PROG); 665 if (nmflag & NFSMNT_NFSV3) 666 *tl++ = txdr_unsigned(NFS_VER3); 667 else 668 *tl++ = txdr_unsigned(NFS_VER2); 669 } 670 if (nmflag & NFSMNT_NFSV3) 671 *tl++ = txdr_unsigned(procid); 672 else 673 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 674 675 /* 676 * And then the authorization cred. 677 */ 678 *tl++ = txdr_unsigned(auth_type); 679 *tl = txdr_unsigned(authsiz); 680 switch (auth_type) { 681 case RPCAUTH_UNIX: 682 nfsm_build(tl, u_int32_t *, auth_len); 683 *tl++ = 0; /* stamp ?? */ 684 *tl++ = 0; /* NULL hostname */ 685 *tl++ = txdr_unsigned(cr->cr_uid); 686 *tl++ = txdr_unsigned(cr->cr_gid); 687 grpsiz = (auth_len >> 2) - 5; 688 *tl++ = txdr_unsigned(grpsiz); 689 for (i = 0; i < grpsiz; i++) 690 *tl++ = txdr_unsigned(cr->cr_groups[i]); 691 break; 692 case RPCAUTH_KERB4: 693 siz = auth_len; 694 while (siz > 0) { 695 if (M_TRAILINGSPACE(mb) == 0) { 696 MGET(mb2, M_WAIT, MT_DATA); 697 if (siz >= MINCLSIZE) 698 MCLGET(mb2, M_WAIT); 699 mb->m_next = mb2; 700 mb = mb2; 701 mb->m_len = 0; 702 bpos = mtod(mb, caddr_t); 703 } 704 i = min(siz, M_TRAILINGSPACE(mb)); 705 memcpy(bpos, auth_str, i); 706 mb->m_len += i; 707 auth_str += i; 708 bpos += i; 709 siz -= i; 710 } 711 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 712 for (i = 0; i < siz; i++) 713 *bpos++ = '\0'; 714 mb->m_len += siz; 715 } 716 break; 717 }; 718 719 /* 720 * And the verifier... 721 */ 722 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 723 if (verf_str) { 724 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 725 *tl = txdr_unsigned(verf_len); 726 siz = verf_len; 727 while (siz > 0) { 728 if (M_TRAILINGSPACE(mb) == 0) { 729 MGET(mb2, M_WAIT, MT_DATA); 730 if (siz >= MINCLSIZE) 731 MCLGET(mb2, M_WAIT); 732 mb->m_next = mb2; 733 mb = mb2; 734 mb->m_len = 0; 735 bpos = mtod(mb, caddr_t); 736 } 737 i = min(siz, M_TRAILINGSPACE(mb)); 738 memcpy(bpos, verf_str, i); 739 mb->m_len += i; 740 verf_str += i; 741 bpos += i; 742 siz -= i; 743 } 744 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 745 for (i = 0; i < siz; i++) 746 *bpos++ = '\0'; 747 mb->m_len += siz; 748 } 749 } else { 750 *tl++ = txdr_unsigned(RPCAUTH_NULL); 751 *tl = 0; 752 } 753 mb->m_next = mrest; 754 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 755 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 756 *mbp = mb; 757 return (mreq); 758 } 759 760 /* 761 * copies mbuf chain to the uio scatter/gather list 762 */ 763 int 764 nfsm_mbuftouio(mrep, uiop, siz, dpos) 765 struct mbuf **mrep; 766 struct uio *uiop; 767 int siz; 768 caddr_t *dpos; 769 { 770 char *mbufcp, *uiocp; 771 int xfer, left, len; 772 struct mbuf *mp; 773 long uiosiz, rem; 774 int error = 0; 775 776 mp = *mrep; 777 mbufcp = *dpos; 778 len = mtod(mp, caddr_t)+mp->m_len-mbufcp; 779 rem = nfsm_rndup(siz)-siz; 780 while (siz > 0) { 781 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 782 return (EFBIG); 783 left = uiop->uio_iov->iov_len; 784 uiocp = uiop->uio_iov->iov_base; 785 if (left > siz) 786 left = siz; 787 uiosiz = left; 788 while (left > 0) { 789 while (len == 0) { 790 mp = mp->m_next; 791 if (mp == NULL) 792 return (EBADRPC); 793 mbufcp = mtod(mp, caddr_t); 794 len = mp->m_len; 795 } 796 xfer = (left > len) ? len : left; 797 #ifdef notdef 798 /* Not Yet.. */ 799 if (uiop->uio_iov->iov_op != NULL) 800 (*(uiop->uio_iov->iov_op)) 801 (mbufcp, uiocp, xfer); 802 else 803 #endif 804 if (uiop->uio_segflg == UIO_SYSSPACE) 805 memcpy(uiocp, mbufcp, xfer); 806 else 807 copyout(mbufcp, uiocp, xfer); 808 left -= xfer; 809 len -= xfer; 810 mbufcp += xfer; 811 uiocp += xfer; 812 uiop->uio_offset += xfer; 813 uiop->uio_resid -= xfer; 814 } 815 if (uiop->uio_iov->iov_len <= siz) { 816 uiop->uio_iovcnt--; 817 uiop->uio_iov++; 818 } else { 819 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 820 uiop->uio_iov->iov_len -= uiosiz; 821 } 822 siz -= uiosiz; 823 } 824 *dpos = mbufcp; 825 *mrep = mp; 826 if (rem > 0) { 827 if (len < rem) 828 error = nfs_adv(mrep, dpos, rem, len); 829 else 830 *dpos += rem; 831 } 832 return (error); 833 } 834 835 /* 836 * copies a uio scatter/gather list to an mbuf chain. 837 * NOTE: can ony handle iovcnt == 1 838 */ 839 int 840 nfsm_uiotombuf(uiop, mq, siz, bpos) 841 struct uio *uiop; 842 struct mbuf **mq; 843 int siz; 844 caddr_t *bpos; 845 { 846 char *uiocp; 847 struct mbuf *mp, *mp2; 848 int xfer, left, mlen; 849 int uiosiz, clflg, rem; 850 char *cp; 851 852 #ifdef DIAGNOSTIC 853 if (uiop->uio_iovcnt != 1) 854 panic("nfsm_uiotombuf: iovcnt != 1"); 855 #endif 856 857 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 858 clflg = 1; 859 else 860 clflg = 0; 861 rem = nfsm_rndup(siz)-siz; 862 mp = mp2 = *mq; 863 while (siz > 0) { 864 left = uiop->uio_iov->iov_len; 865 uiocp = uiop->uio_iov->iov_base; 866 if (left > siz) 867 left = siz; 868 uiosiz = left; 869 while (left > 0) { 870 mlen = M_TRAILINGSPACE(mp); 871 if (mlen == 0) { 872 MGET(mp, M_WAIT, MT_DATA); 873 if (clflg) 874 MCLGET(mp, M_WAIT); 875 mp->m_len = 0; 876 mp2->m_next = mp; 877 mp2 = mp; 878 mlen = M_TRAILINGSPACE(mp); 879 } 880 xfer = (left > mlen) ? mlen : left; 881 #ifdef notdef 882 /* Not Yet.. */ 883 if (uiop->uio_iov->iov_op != NULL) 884 (*(uiop->uio_iov->iov_op)) 885 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 886 else 887 #endif 888 if (uiop->uio_segflg == UIO_SYSSPACE) 889 memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer); 890 else 891 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 892 mp->m_len += xfer; 893 left -= xfer; 894 uiocp += xfer; 895 uiop->uio_offset += xfer; 896 uiop->uio_resid -= xfer; 897 } 898 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 899 uiop->uio_iov->iov_len -= uiosiz; 900 siz -= uiosiz; 901 } 902 if (rem > 0) { 903 if (rem > M_TRAILINGSPACE(mp)) { 904 MGET(mp, M_WAIT, MT_DATA); 905 mp->m_len = 0; 906 mp2->m_next = mp; 907 } 908 cp = mtod(mp, caddr_t)+mp->m_len; 909 for (left = 0; left < rem; left++) 910 *cp++ = '\0'; 911 mp->m_len += rem; 912 *bpos = cp; 913 } else 914 *bpos = mtod(mp, caddr_t)+mp->m_len; 915 *mq = mp; 916 return (0); 917 } 918 919 /* 920 * Get at least "siz" bytes of correctly aligned data. 921 * When called the mbuf pointers are not necessarily correct, 922 * dsosp points to what ought to be in m_data and left contains 923 * what ought to be in m_len. 924 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 925 * cases. (The macros use the vars. dpos and dpos2) 926 */ 927 int 928 nfsm_disct(mdp, dposp, siz, left, cp2) 929 struct mbuf **mdp; 930 caddr_t *dposp; 931 int siz; 932 int left; 933 caddr_t *cp2; 934 { 935 struct mbuf *m1, *m2; 936 struct mbuf *havebuf = NULL; 937 caddr_t src = *dposp; 938 caddr_t dst; 939 int len; 940 941 #ifdef DEBUG 942 if (left < 0) 943 panic("nfsm_disct: left < 0"); 944 #endif 945 m1 = *mdp; 946 /* 947 * Skip through the mbuf chain looking for an mbuf with 948 * some data. If the first mbuf found has enough data 949 * and it is correctly aligned return it. 950 */ 951 while (left == 0) { 952 havebuf = m1; 953 *mdp = m1 = m1->m_next; 954 if (m1 == NULL) 955 return (EBADRPC); 956 src = mtod(m1, caddr_t); 957 left = m1->m_len; 958 /* 959 * If we start a new mbuf and it is big enough 960 * and correctly aligned just return it, don't 961 * do any pull up. 962 */ 963 if (left >= siz && nfsm_aligned(src)) { 964 *cp2 = src; 965 *dposp = src + siz; 966 return (0); 967 } 968 } 969 if (m1->m_flags & M_EXT) { 970 if (havebuf) { 971 /* If the first mbuf with data has external data 972 * and there is a previous empty mbuf use it 973 * to move the data into. 974 */ 975 m2 = m1; 976 *mdp = m1 = havebuf; 977 if (m1->m_flags & M_EXT) { 978 MEXTREMOVE(m1); 979 } 980 } else { 981 /* 982 * If the first mbuf has a external data 983 * and there is no previous empty mbuf 984 * allocate a new mbuf and move the external 985 * data to the new mbuf. Also make the first 986 * mbuf look empty. 987 */ 988 m2 = m_get(M_WAIT, MT_DATA); 989 m2->m_ext = m1->m_ext; 990 m2->m_data = src; 991 m2->m_len = left; 992 MCLADDREFERENCE(m1, m2); 993 MEXTREMOVE(m1); 994 m2->m_next = m1->m_next; 995 m1->m_next = m2; 996 } 997 m1->m_len = 0; 998 dst = m1->m_dat; 999 } else { 1000 /* 1001 * If the first mbuf has no external data 1002 * move the data to the front of the mbuf. 1003 */ 1004 if ((dst = m1->m_dat) != src) 1005 memmove(dst, src, left); 1006 dst += left; 1007 m1->m_len = left; 1008 m2 = m1->m_next; 1009 } 1010 m1->m_flags &= ~M_PKTHDR; 1011 *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */ 1012 *dposp = mtod(m1, caddr_t) + siz; 1013 /* 1014 * Loop through mbufs pulling data up into first mbuf until 1015 * the first mbuf is full or there is no more data to 1016 * pullup. 1017 */ 1018 while ((len = (MLEN - m1->m_len)) != 0 && m2) { 1019 if ((len = min(len, m2->m_len)) != 0) 1020 memcpy(dst, m2->m_data, len); 1021 m1->m_len += len; 1022 dst += len; 1023 m2->m_data += len; 1024 m2->m_len -= len; 1025 m2 = m2->m_next; 1026 } 1027 if (m1->m_len < siz) 1028 return (EBADRPC); 1029 return (0); 1030 } 1031 1032 /* 1033 * Advance the position in the mbuf chain. 1034 */ 1035 int 1036 nfs_adv(mdp, dposp, offs, left) 1037 struct mbuf **mdp; 1038 caddr_t *dposp; 1039 int offs; 1040 int left; 1041 { 1042 struct mbuf *m; 1043 int s; 1044 1045 m = *mdp; 1046 s = left; 1047 while (s < offs) { 1048 offs -= s; 1049 m = m->m_next; 1050 if (m == NULL) 1051 return (EBADRPC); 1052 s = m->m_len; 1053 } 1054 *mdp = m; 1055 *dposp = mtod(m, caddr_t)+offs; 1056 return (0); 1057 } 1058 1059 /* 1060 * Copy a string into mbufs for the hard cases... 1061 */ 1062 int 1063 nfsm_strtmbuf(mb, bpos, cp, siz) 1064 struct mbuf **mb; 1065 char **bpos; 1066 const char *cp; 1067 long siz; 1068 { 1069 struct mbuf *m1 = NULL, *m2; 1070 long left, xfer, len, tlen; 1071 u_int32_t *tl; 1072 int putsize; 1073 1074 putsize = 1; 1075 m2 = *mb; 1076 left = M_TRAILINGSPACE(m2); 1077 if (left > 0) { 1078 tl = ((u_int32_t *)(*bpos)); 1079 *tl++ = txdr_unsigned(siz); 1080 putsize = 0; 1081 left -= NFSX_UNSIGNED; 1082 m2->m_len += NFSX_UNSIGNED; 1083 if (left > 0) { 1084 memcpy((caddr_t) tl, cp, left); 1085 siz -= left; 1086 cp += left; 1087 m2->m_len += left; 1088 left = 0; 1089 } 1090 } 1091 /* Loop around adding mbufs */ 1092 while (siz > 0) { 1093 MGET(m1, M_WAIT, MT_DATA); 1094 if (siz > MLEN) 1095 MCLGET(m1, M_WAIT); 1096 m1->m_len = NFSMSIZ(m1); 1097 m2->m_next = m1; 1098 m2 = m1; 1099 tl = mtod(m1, u_int32_t *); 1100 tlen = 0; 1101 if (putsize) { 1102 *tl++ = txdr_unsigned(siz); 1103 m1->m_len -= NFSX_UNSIGNED; 1104 tlen = NFSX_UNSIGNED; 1105 putsize = 0; 1106 } 1107 if (siz < m1->m_len) { 1108 len = nfsm_rndup(siz); 1109 xfer = siz; 1110 if (xfer < len) 1111 *(tl+(xfer>>2)) = 0; 1112 } else { 1113 xfer = len = m1->m_len; 1114 } 1115 memcpy((caddr_t) tl, cp, xfer); 1116 m1->m_len = len+tlen; 1117 siz -= xfer; 1118 cp += xfer; 1119 } 1120 *mb = m1; 1121 *bpos = mtod(m1, caddr_t)+m1->m_len; 1122 return (0); 1123 } 1124 1125 /* 1126 * Directory caching routines. They work as follows: 1127 * - a cache is maintained per VDIR nfsnode. 1128 * - for each offset cookie that is exported to userspace, and can 1129 * thus be thrown back at us as an offset to VOP_READDIR, store 1130 * information in the cache. 1131 * - cached are: 1132 * - cookie itself 1133 * - blocknumber (essentially just a search key in the buffer cache) 1134 * - entry number in block. 1135 * - offset cookie of block in which this entry is stored 1136 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1137 * - entries are looked up in a hash table 1138 * - also maintained is an LRU list of entries, used to determine 1139 * which ones to delete if the cache grows too large. 1140 * - if 32 <-> 64 translation mode is requested for a filesystem, 1141 * the cache also functions as a translation table 1142 * - in the translation case, invalidating the cache does not mean 1143 * flushing it, but just marking entries as invalid, except for 1144 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1145 * still be able to use the cache as a translation table. 1146 * - 32 bit cookies are uniquely created by combining the hash table 1147 * entry value, and one generation count per hash table entry, 1148 * incremented each time an entry is appended to the chain. 1149 * - the cache is invalidated each time a direcory is modified 1150 * - sanity checks are also done; if an entry in a block turns 1151 * out not to have a matching cookie, the cache is invalidated 1152 * and a new block starting from the wanted offset is fetched from 1153 * the server. 1154 * - directory entries as read from the server are extended to contain 1155 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1156 * the cache and exporting them to userspace through the cookie 1157 * argument to VOP_READDIR. 1158 */ 1159 1160 u_long 1161 nfs_dirhash(off) 1162 off_t off; 1163 { 1164 int i; 1165 char *cp = (char *)&off; 1166 u_long sum = 0L; 1167 1168 for (i = 0 ; i < sizeof (off); i++) 1169 sum += *cp++; 1170 1171 return sum; 1172 } 1173 1174 void 1175 nfs_initdircache(vp) 1176 struct vnode *vp; 1177 { 1178 struct nfsnode *np = VTONFS(vp); 1179 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1180 1181 np->n_dircachesize = 0; 1182 np->n_dblkno = 1; 1183 np->n_dircache = 1184 hashinit(NFS_DIRHASHSIZ, M_NFSDIROFF, M_WAITOK, &nfsdirhashmask); 1185 TAILQ_INIT(&np->n_dirchain); 1186 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1187 MALLOC(np->n_dirgens, unsigned *, 1188 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, 1189 M_WAITOK); 1190 memset((caddr_t)np->n_dirgens, 0, 1191 NFS_DIRHASHSIZ * sizeof (unsigned)); 1192 } 1193 } 1194 1195 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0}; 1196 1197 struct nfsdircache * 1198 nfs_searchdircache(vp, off, do32, hashent) 1199 struct vnode *vp; 1200 off_t off; 1201 int do32; 1202 int *hashent; 1203 { 1204 struct nfsdirhashhead *ndhp; 1205 struct nfsdircache *ndp = NULL; 1206 struct nfsnode *np = VTONFS(vp); 1207 unsigned ent; 1208 1209 /* 1210 * Zero is always a valid cookie. 1211 */ 1212 if (off == 0) 1213 return &dzero; 1214 1215 /* 1216 * We use a 32bit cookie as search key, directly reconstruct 1217 * the hashentry. Else use the hashfunction. 1218 */ 1219 if (do32) { 1220 ent = (u_int32_t)off >> 24; 1221 if (ent >= NFS_DIRHASHSIZ) 1222 return NULL; 1223 ndhp = &np->n_dircache[ent]; 1224 } else { 1225 ndhp = NFSDIRHASH(np, off); 1226 } 1227 1228 if (hashent) 1229 *hashent = (int)(ndhp - np->n_dircache); 1230 if (do32) { 1231 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) { 1232 if (ndp->dc_cookie32 == (u_int32_t)off) { 1233 /* 1234 * An invalidated entry will become the 1235 * start of a new block fetched from 1236 * the server. 1237 */ 1238 if (ndp->dc_blkno == -1) { 1239 ndp->dc_blkcookie = ndp->dc_cookie; 1240 ndp->dc_blkno = np->n_dblkno++; 1241 ndp->dc_entry = 0; 1242 } 1243 break; 1244 } 1245 } 1246 } else { 1247 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) 1248 if (ndp->dc_cookie == off) 1249 break; 1250 } 1251 return ndp; 1252 } 1253 1254 1255 struct nfsdircache * 1256 nfs_enterdircache(vp, off, blkoff, en, blkno) 1257 struct vnode *vp; 1258 off_t off, blkoff; 1259 daddr_t blkno; 1260 int en; 1261 { 1262 struct nfsnode *np = VTONFS(vp); 1263 struct nfsdirhashhead *ndhp; 1264 struct nfsdircache *ndp = NULL, *first; 1265 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1266 int hashent, gen, overwrite; 1267 1268 if (!np->n_dircache) 1269 /* 1270 * XXX would like to do this in nfs_nget but vtype 1271 * isn't known at that time. 1272 */ 1273 nfs_initdircache(vp); 1274 1275 /* 1276 * XXX refuse entries for offset 0. amd(8) erroneously sets 1277 * cookie 0 for the '.' entry, making this necessary. This 1278 * isn't so bad, as 0 is a special case anyway. 1279 */ 1280 if (off == 0) 1281 return &dzero; 1282 1283 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1284 1285 if (ndp && ndp->dc_blkno != -1) { 1286 /* 1287 * Overwriting an old entry. Check if it's the same. 1288 * If so, just return. If not, remove the old entry. 1289 */ 1290 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1291 return ndp; 1292 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1293 LIST_REMOVE(ndp, dc_hash); 1294 FREE(ndp, M_NFSDIROFF); 1295 ndp = 0; 1296 } 1297 1298 ndhp = &np->n_dircache[hashent]; 1299 1300 if (!ndp) { 1301 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF, 1302 M_WAITOK); 1303 overwrite = 0; 1304 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1305 /* 1306 * We're allocating a new entry, so bump the 1307 * generation number. 1308 */ 1309 gen = ++np->n_dirgens[hashent]; 1310 if (gen == 0) { 1311 np->n_dirgens[hashent]++; 1312 gen++; 1313 } 1314 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1315 } 1316 } else 1317 overwrite = 1; 1318 1319 /* 1320 * If the entry number is 0, we are at the start of a new block, so 1321 * allocate a new blocknumber. 1322 */ 1323 if (en == 0) 1324 ndp->dc_blkno = np->n_dblkno++; 1325 else 1326 ndp->dc_blkno = blkno; 1327 1328 ndp->dc_cookie = off; 1329 ndp->dc_blkcookie = blkoff; 1330 ndp->dc_entry = en; 1331 1332 if (overwrite) 1333 return ndp; 1334 1335 /* 1336 * If the maximum directory cookie cache size has been reached 1337 * for this node, take one off the front. The idea is that 1338 * directories are typically read front-to-back once, so that 1339 * the oldest entries can be thrown away without much performance 1340 * loss. 1341 */ 1342 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1343 first = np->n_dirchain.tqh_first; 1344 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain); 1345 LIST_REMOVE(first, dc_hash); 1346 FREE(first, M_NFSDIROFF); 1347 } else 1348 np->n_dircachesize++; 1349 1350 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1351 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1352 return ndp; 1353 } 1354 1355 void 1356 nfs_invaldircache(vp, forcefree) 1357 struct vnode *vp; 1358 int forcefree; 1359 { 1360 struct nfsnode *np = VTONFS(vp); 1361 struct nfsdircache *ndp = NULL; 1362 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1363 1364 #ifdef DIAGNOSTIC 1365 if (vp->v_type != VDIR) 1366 panic("nfs: invaldircache: not dir"); 1367 #endif 1368 1369 if (!np->n_dircache) 1370 return; 1371 1372 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1373 while ((ndp = np->n_dirchain.tqh_first)) { 1374 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1375 LIST_REMOVE(ndp, dc_hash); 1376 FREE(ndp, M_NFSDIROFF); 1377 } 1378 np->n_dircachesize = 0; 1379 if (forcefree && np->n_dirgens) { 1380 FREE(np->n_dirgens, M_NFSDIROFF); 1381 } 1382 } else { 1383 for (ndp = np->n_dirchain.tqh_first; ndp; 1384 ndp = ndp->dc_chain.tqe_next) 1385 ndp->dc_blkno = -1; 1386 } 1387 1388 np->n_dblkno = 1; 1389 } 1390 1391 /* 1392 * Called once before VFS init to initialize shared and 1393 * server-specific data structures. 1394 */ 1395 void 1396 nfs_init() 1397 { 1398 1399 #if !defined(alpha) && !defined(_LP64) && defined(DIAGNOSTIC) 1400 /* 1401 * Check to see if major data structures haven't bloated. 1402 */ 1403 if (sizeof (struct nfsnode) > NFS_NODEALLOC) { 1404 printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); 1405 printf("Try reducing NFS_SMALLFH\n"); 1406 } 1407 if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { 1408 printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); 1409 printf("Try reducing NFS_UIDHASHSIZ\n"); 1410 } 1411 if (sizeof (struct nfsuid) > NFS_UIDALLOC) { 1412 printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); 1413 printf("Try unionizing the nu_nickname and nu_flag fields\n"); 1414 } 1415 #endif 1416 1417 nfsrtt.pos = 0; 1418 rpc_vers = txdr_unsigned(RPC_VER2); 1419 rpc_call = txdr_unsigned(RPC_CALL); 1420 rpc_reply = txdr_unsigned(RPC_REPLY); 1421 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1422 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1423 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1424 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1425 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1426 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1427 nfs_prog = txdr_unsigned(NFS_PROG); 1428 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1429 nfs_true = txdr_unsigned(TRUE); 1430 nfs_false = txdr_unsigned(FALSE); 1431 nfs_xdrneg1 = txdr_unsigned(-1); 1432 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1433 if (nfs_ticks < 1) 1434 nfs_ticks = 1; 1435 #ifdef NFSSERVER 1436 nfsrv_init(0); /* Init server data structures */ 1437 nfsrv_initcache(); /* Init the server request cache */ 1438 #endif /* NFSSERVER */ 1439 1440 /* 1441 * Initialize the nqnfs data structures. 1442 */ 1443 if (nqnfsstarttime == 0) { 1444 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1445 + nqsrv_clockskew + nqsrv_writeslack; 1446 NQLOADNOVRAM(nqnfsstarttime); 1447 CIRCLEQ_INIT(&nqtimerhead); 1448 nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, M_WAITOK, &nqfhhash); 1449 } 1450 1451 /* 1452 * Initialize reply list and start timer 1453 */ 1454 TAILQ_INIT(&nfs_reqq); 1455 nfs_timer(NULL); 1456 } 1457 1458 #ifdef NFS 1459 /* 1460 * Called once at VFS init to initialize client-specific data structures. 1461 */ 1462 void 1463 nfs_vfs_init() 1464 { 1465 int i; 1466 1467 /* Ensure async daemons disabled */ 1468 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { 1469 nfs_iodwant[i] = (struct proc *)0; 1470 nfs_iodmount[i] = (struct nfsmount *)0; 1471 } 1472 nfs_nhinit(); /* Init the nfsnode table */ 1473 } 1474 1475 void 1476 nfs_vfs_done() 1477 { 1478 nfs_nhdone(); 1479 } 1480 1481 /* 1482 * Attribute cache routines. 1483 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1484 * that are on the mbuf list 1485 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1486 * error otherwise 1487 */ 1488 1489 /* 1490 * Load the attribute cache (that lives in the nfsnode entry) with 1491 * the values on the mbuf list and 1492 * Iff vap not NULL 1493 * copy the attributes to *vaper 1494 */ 1495 int 1496 nfsm_loadattrcache(vpp, mdp, dposp, vaper) 1497 struct vnode **vpp; 1498 struct mbuf **mdp; 1499 caddr_t *dposp; 1500 struct vattr *vaper; 1501 { 1502 int32_t t1; 1503 caddr_t cp2; 1504 int error = 0; 1505 struct mbuf *md; 1506 int v3 = NFS_ISV3(*vpp); 1507 1508 md = *mdp; 1509 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1510 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1511 if (error) 1512 return (error); 1513 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper); 1514 } 1515 1516 int 1517 nfs_loadattrcache(vpp, fp, vaper) 1518 struct vnode **vpp; 1519 struct nfs_fattr *fp; 1520 struct vattr *vaper; 1521 { 1522 struct vnode *vp = *vpp; 1523 struct vattr *vap; 1524 int v3 = NFS_ISV3(vp); 1525 enum vtype vtyp; 1526 u_short vmode; 1527 struct timespec mtime; 1528 struct vnode *nvp; 1529 int32_t rdev; 1530 struct nfsnode *np; 1531 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1532 1533 if (v3) { 1534 vtyp = nfsv3tov_type(fp->fa_type); 1535 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1536 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1537 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1538 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1539 } else { 1540 vtyp = nfsv2tov_type(fp->fa_type); 1541 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1542 if (vtyp == VNON || vtyp == VREG) 1543 vtyp = IFTOVT(vmode); 1544 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1545 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1546 1547 /* 1548 * Really ugly NFSv2 kludge. 1549 */ 1550 if (vtyp == VCHR && rdev == 0xffffffff) 1551 vtyp = VFIFO; 1552 } 1553 1554 /* 1555 * If v_type == VNON it is a new node, so fill in the v_type, 1556 * n_mtime fields. Check to see if it represents a special 1557 * device, and if so, check for a possible alias. Once the 1558 * correct vnode has been obtained, fill in the rest of the 1559 * information. 1560 */ 1561 np = VTONFS(vp); 1562 if (vp->v_type != vtyp) { 1563 vp->v_type = vtyp; 1564 if (vp->v_type == VFIFO) { 1565 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1566 vp->v_op = fifo_nfsv2nodeop_p; 1567 } 1568 if (vp->v_type == VCHR || vp->v_type == VBLK) { 1569 vp->v_op = spec_nfsv2nodeop_p; 1570 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1571 if (nvp) { 1572 /* 1573 * Discard unneeded vnode, but save its nfsnode. 1574 * Since the nfsnode does not have a lock, its 1575 * vnode lock has to be carried over. 1576 */ 1577 nvp->v_data = vp->v_data; 1578 vp->v_data = NULL; 1579 vp->v_op = spec_vnodeop_p; 1580 vput(vp); 1581 vgone(vp); 1582 /* 1583 * XXX When nfs starts locking, we need to 1584 * lock the new node here. 1585 */ 1586 /* 1587 * Reinitialize aliased node. 1588 */ 1589 np->n_vnode = nvp; 1590 *vpp = vp = nvp; 1591 } 1592 } 1593 np->n_mtime = mtime.tv_sec; 1594 } 1595 vap = np->n_vattr; 1596 vap->va_type = vtyp; 1597 vap->va_mode = vmode & ALLPERMS; 1598 vap->va_rdev = (dev_t)rdev; 1599 vap->va_mtime = mtime; 1600 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 1601 switch (vtyp) { 1602 case VDIR: 1603 vap->va_blocksize = NFS_DIRFRAGSIZ; 1604 break; 1605 case VBLK: 1606 vap->va_blocksize = BLKDEV_IOSIZE; 1607 break; 1608 case VCHR: 1609 vap->va_blocksize = MAXBSIZE; 1610 break; 1611 default: 1612 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1613 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1614 break; 1615 } 1616 if (v3) { 1617 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1618 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1619 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1620 vap->va_size = fxdr_hyper(&fp->fa3_size); 1621 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1622 vap->va_fileid = fxdr_unsigned(int32_t, 1623 fp->fa3_fileid.nfsuquad[1]); 1624 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1625 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); 1626 vap->va_flags = 0; 1627 vap->va_filerev = 0; 1628 } else { 1629 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1630 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1631 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1632 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1633 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1634 * NFS_FABLKSIZE; 1635 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1636 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1637 vap->va_flags = 0; 1638 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, 1639 fp->fa2_ctime.nfsv2_sec); 1640 vap->va_ctime.tv_nsec = 0; 1641 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1642 vap->va_filerev = 0; 1643 } 1644 if (vap->va_size != np->n_size) { 1645 if (vap->va_type == VREG) { 1646 if (np->n_flag & NMODIFIED) { 1647 if (vap->va_size < np->n_size) 1648 vap->va_size = np->n_size; 1649 else 1650 np->n_size = vap->va_size; 1651 } else 1652 np->n_size = vap->va_size; 1653 uvm_vnp_setsize(vp, np->n_size); 1654 } else 1655 np->n_size = vap->va_size; 1656 } 1657 np->n_attrstamp = time.tv_sec; 1658 if (vaper != NULL) { 1659 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1660 if (np->n_flag & NCHG) { 1661 if (np->n_flag & NACC) 1662 vaper->va_atime = np->n_atim; 1663 if (np->n_flag & NUPD) 1664 vaper->va_mtime = np->n_mtim; 1665 } 1666 } 1667 return (0); 1668 } 1669 1670 /* 1671 * Check the time stamp 1672 * If the cache is valid, copy contents to *vap and return 0 1673 * otherwise return an error 1674 */ 1675 int 1676 nfs_getattrcache(vp, vaper) 1677 struct vnode *vp; 1678 struct vattr *vaper; 1679 { 1680 struct nfsnode *np = VTONFS(vp); 1681 struct vattr *vap; 1682 1683 if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { 1684 nfsstats.attrcache_misses++; 1685 return (ENOENT); 1686 } 1687 nfsstats.attrcache_hits++; 1688 vap = np->n_vattr; 1689 if (vap->va_size != np->n_size) { 1690 if (vap->va_type == VREG) { 1691 if (np->n_flag & NMODIFIED) { 1692 if (vap->va_size < np->n_size) 1693 vap->va_size = np->n_size; 1694 else 1695 np->n_size = vap->va_size; 1696 } else 1697 np->n_size = vap->va_size; 1698 uvm_vnp_setsize(vp, np->n_size); 1699 } else 1700 np->n_size = vap->va_size; 1701 } 1702 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1703 if (np->n_flag & NCHG) { 1704 if (np->n_flag & NACC) 1705 vaper->va_atime = np->n_atim; 1706 if (np->n_flag & NUPD) 1707 vaper->va_mtime = np->n_mtim; 1708 } 1709 return (0); 1710 } 1711 1712 /* 1713 * Heuristic to see if the server XDR encodes directory cookies or not. 1714 * it is not supposed to, but a lot of servers may do this. Also, since 1715 * most/all servers will implement V2 as well, it is expected that they 1716 * may return just 32 bits worth of cookie information, so we need to 1717 * find out in which 32 bits this information is available. We do this 1718 * to avoid trouble with emulated binaries that can't handle 64 bit 1719 * directory offsets. 1720 */ 1721 1722 void 1723 nfs_cookieheuristic(vp, flagp, p, cred) 1724 struct vnode *vp; 1725 int *flagp; 1726 struct proc *p; 1727 struct ucred *cred; 1728 { 1729 struct uio auio; 1730 struct iovec aiov; 1731 caddr_t buf, cp; 1732 struct dirent *dp; 1733 off_t *cookies = NULL, *cop; 1734 int error, eof, nc, len; 1735 1736 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 1737 1738 aiov.iov_base = buf; 1739 aiov.iov_len = NFS_DIRFRAGSIZ; 1740 auio.uio_iov = &aiov; 1741 auio.uio_iovcnt = 1; 1742 auio.uio_rw = UIO_READ; 1743 auio.uio_segflg = UIO_SYSSPACE; 1744 auio.uio_procp = p; 1745 auio.uio_resid = NFS_DIRFRAGSIZ; 1746 auio.uio_offset = 0; 1747 1748 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 1749 1750 len = NFS_DIRFRAGSIZ - auio.uio_resid; 1751 if (error || len == 0) { 1752 FREE(buf, M_TEMP); 1753 if (cookies) 1754 free(cookies, M_TEMP); 1755 return; 1756 } 1757 1758 /* 1759 * Find the first valid entry and look at its offset cookie. 1760 */ 1761 1762 cp = buf; 1763 for (cop = cookies; len > 0; len -= dp->d_reclen) { 1764 dp = (struct dirent *)cp; 1765 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 1766 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 1767 *flagp |= NFSMNT_SWAPCOOKIE; 1768 nfs_invaldircache(vp, 0); 1769 nfs_vinvalbuf(vp, 0, cred, p, 1); 1770 } 1771 break; 1772 } 1773 cop++; 1774 cp += dp->d_reclen; 1775 } 1776 1777 FREE(buf, M_TEMP); 1778 free(cookies, M_TEMP); 1779 } 1780 #endif /* NFS */ 1781 1782 /* 1783 * Set up nameidata for a lookup() call and do it. 1784 * 1785 * If pubflag is set, this call is done for a lookup operation on the 1786 * public filehandle. In that case we allow crossing mountpoints and 1787 * absolute pathnames. However, the caller is expected to check that 1788 * the lookup result is within the public fs, and deny access if 1789 * it is not. 1790 */ 1791 int 1792 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) 1793 struct nameidata *ndp; 1794 fhandle_t *fhp; 1795 int len; 1796 struct nfssvc_sock *slp; 1797 struct mbuf *nam; 1798 struct mbuf **mdp; 1799 caddr_t *dposp; 1800 struct vnode **retdirp; 1801 struct proc *p; 1802 int kerbflag, pubflag; 1803 { 1804 int i, rem; 1805 struct mbuf *md; 1806 char *fromcp, *tocp, *cp; 1807 struct iovec aiov; 1808 struct uio auio; 1809 struct vnode *dp; 1810 int error, rdonly, linklen; 1811 struct componentname *cnp = &ndp->ni_cnd; 1812 1813 *retdirp = (struct vnode *)0; 1814 1815 if ((len + 1) > MAXPATHLEN) 1816 return (ENAMETOOLONG); 1817 cnp->cn_pnbuf = PNBUF_GET(); 1818 1819 /* 1820 * Copy the name from the mbuf list to ndp->ni_pnbuf 1821 * and set the various ndp fields appropriately. 1822 */ 1823 fromcp = *dposp; 1824 tocp = cnp->cn_pnbuf; 1825 md = *mdp; 1826 rem = mtod(md, caddr_t) + md->m_len - fromcp; 1827 for (i = 0; i < len; i++) { 1828 while (rem == 0) { 1829 md = md->m_next; 1830 if (md == NULL) { 1831 error = EBADRPC; 1832 goto out; 1833 } 1834 fromcp = mtod(md, caddr_t); 1835 rem = md->m_len; 1836 } 1837 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 1838 error = EACCES; 1839 goto out; 1840 } 1841 *tocp++ = *fromcp++; 1842 rem--; 1843 } 1844 *tocp = '\0'; 1845 *mdp = md; 1846 *dposp = fromcp; 1847 len = nfsm_rndup(len)-len; 1848 if (len > 0) { 1849 if (rem >= len) 1850 *dposp += len; 1851 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 1852 goto out; 1853 } 1854 1855 /* 1856 * Extract and set starting directory. 1857 */ 1858 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 1859 nam, &rdonly, kerbflag, pubflag); 1860 if (error) 1861 goto out; 1862 if (dp->v_type != VDIR) { 1863 vrele(dp); 1864 error = ENOTDIR; 1865 goto out; 1866 } 1867 1868 if (rdonly) 1869 cnp->cn_flags |= RDONLY; 1870 1871 *retdirp = dp; 1872 1873 if (pubflag) { 1874 /* 1875 * Oh joy. For WebNFS, handle those pesky '%' escapes, 1876 * and the 'native path' indicator. 1877 */ 1878 cp = PNBUF_GET(); 1879 fromcp = cnp->cn_pnbuf; 1880 tocp = cp; 1881 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 1882 switch ((unsigned char)*fromcp) { 1883 case WEBNFS_NATIVE_CHAR: 1884 /* 1885 * 'Native' path for us is the same 1886 * as a path according to the NFS spec, 1887 * just skip the escape char. 1888 */ 1889 fromcp++; 1890 break; 1891 /* 1892 * More may be added in the future, range 0x80-0xff 1893 */ 1894 default: 1895 error = EIO; 1896 FREE(cp, M_NAMEI); 1897 goto out; 1898 } 1899 } 1900 /* 1901 * Translate the '%' escapes, URL-style. 1902 */ 1903 while (*fromcp != '\0') { 1904 if (*fromcp == WEBNFS_ESC_CHAR) { 1905 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 1906 fromcp++; 1907 *tocp++ = HEXSTRTOI(fromcp); 1908 fromcp += 2; 1909 continue; 1910 } else { 1911 error = ENOENT; 1912 FREE(cp, M_NAMEI); 1913 goto out; 1914 } 1915 } else 1916 *tocp++ = *fromcp++; 1917 } 1918 *tocp = '\0'; 1919 PNBUF_PUT(cnp->cn_pnbuf); 1920 cnp->cn_pnbuf = cp; 1921 } 1922 1923 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 1924 ndp->ni_segflg = UIO_SYSSPACE; 1925 1926 if (pubflag) { 1927 ndp->ni_rootdir = rootvnode; 1928 ndp->ni_loopcnt = 0; 1929 if (cnp->cn_pnbuf[0] == '/') 1930 dp = rootvnode; 1931 } else { 1932 cnp->cn_flags |= NOCROSSMOUNT; 1933 } 1934 1935 cnp->cn_proc = p; 1936 VREF(dp); 1937 1938 for (;;) { 1939 cnp->cn_nameptr = cnp->cn_pnbuf; 1940 ndp->ni_startdir = dp; 1941 /* 1942 * And call lookup() to do the real work 1943 */ 1944 error = lookup(ndp); 1945 if (error) 1946 break; 1947 /* 1948 * Check for encountering a symbolic link 1949 */ 1950 if ((cnp->cn_flags & ISSYMLINK) == 0) { 1951 if (cnp->cn_flags & (SAVENAME | SAVESTART)) { 1952 cnp->cn_flags |= HASBUF; 1953 return (0); 1954 } 1955 break; 1956 } else { 1957 if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) 1958 VOP_UNLOCK(ndp->ni_dvp, 0); 1959 if (!pubflag) { 1960 vrele(ndp->ni_dvp); 1961 vput(ndp->ni_vp); 1962 ndp->ni_vp = NULL; 1963 error = EINVAL; 1964 break; 1965 } 1966 1967 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 1968 error = ELOOP; 1969 break; 1970 } 1971 if (ndp->ni_pathlen > 1) 1972 cp = PNBUF_GET(); 1973 else 1974 cp = cnp->cn_pnbuf; 1975 aiov.iov_base = cp; 1976 aiov.iov_len = MAXPATHLEN; 1977 auio.uio_iov = &aiov; 1978 auio.uio_iovcnt = 1; 1979 auio.uio_offset = 0; 1980 auio.uio_rw = UIO_READ; 1981 auio.uio_segflg = UIO_SYSSPACE; 1982 auio.uio_procp = (struct proc *)0; 1983 auio.uio_resid = MAXPATHLEN; 1984 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 1985 if (error) { 1986 badlink: 1987 if (ndp->ni_pathlen > 1) 1988 PNBUF_PUT(cp); 1989 break; 1990 } 1991 linklen = MAXPATHLEN - auio.uio_resid; 1992 if (linklen == 0) { 1993 error = ENOENT; 1994 goto badlink; 1995 } 1996 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 1997 error = ENAMETOOLONG; 1998 goto badlink; 1999 } 2000 if (ndp->ni_pathlen > 1) { 2001 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 2002 PNBUF_PUT(cnp->cn_pnbuf); 2003 cnp->cn_pnbuf = cp; 2004 } else 2005 cnp->cn_pnbuf[linklen] = '\0'; 2006 ndp->ni_pathlen += linklen; 2007 vput(ndp->ni_vp); 2008 dp = ndp->ni_dvp; 2009 /* 2010 * Check if root directory should replace current directory. 2011 */ 2012 if (cnp->cn_pnbuf[0] == '/') { 2013 vrele(dp); 2014 dp = ndp->ni_rootdir; 2015 VREF(dp); 2016 } 2017 } 2018 } 2019 out: 2020 PNBUF_PUT(cnp->cn_pnbuf); 2021 return (error); 2022 } 2023 2024 /* 2025 * A fiddled version of m_adj() that ensures null fill to a long 2026 * boundary and only trims off the back end 2027 */ 2028 void 2029 nfsm_adj(mp, len, nul) 2030 struct mbuf *mp; 2031 int len; 2032 int nul; 2033 { 2034 struct mbuf *m; 2035 int count, i; 2036 char *cp; 2037 2038 /* 2039 * Trim from tail. Scan the mbuf chain, 2040 * calculating its length and finding the last mbuf. 2041 * If the adjustment only affects this mbuf, then just 2042 * adjust and return. Otherwise, rescan and truncate 2043 * after the remaining size. 2044 */ 2045 count = 0; 2046 m = mp; 2047 for (;;) { 2048 count += m->m_len; 2049 if (m->m_next == (struct mbuf *)0) 2050 break; 2051 m = m->m_next; 2052 } 2053 if (m->m_len > len) { 2054 m->m_len -= len; 2055 if (nul > 0) { 2056 cp = mtod(m, caddr_t)+m->m_len-nul; 2057 for (i = 0; i < nul; i++) 2058 *cp++ = '\0'; 2059 } 2060 return; 2061 } 2062 count -= len; 2063 if (count < 0) 2064 count = 0; 2065 /* 2066 * Correct length for chain is "count". 2067 * Find the mbuf with last data, adjust its length, 2068 * and toss data from remaining mbufs on chain. 2069 */ 2070 for (m = mp; m; m = m->m_next) { 2071 if (m->m_len >= count) { 2072 m->m_len = count; 2073 if (nul > 0) { 2074 cp = mtod(m, caddr_t)+m->m_len-nul; 2075 for (i = 0; i < nul; i++) 2076 *cp++ = '\0'; 2077 } 2078 break; 2079 } 2080 count -= m->m_len; 2081 } 2082 for (m = m->m_next;m;m = m->m_next) 2083 m->m_len = 0; 2084 } 2085 2086 /* 2087 * Make these functions instead of macros, so that the kernel text size 2088 * doesn't get too big... 2089 */ 2090 void 2091 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2092 struct nfsrv_descript *nfsd; 2093 int before_ret; 2094 struct vattr *before_vap; 2095 int after_ret; 2096 struct vattr *after_vap; 2097 struct mbuf **mbp; 2098 char **bposp; 2099 { 2100 struct mbuf *mb = *mbp, *mb2; 2101 char *bpos = *bposp; 2102 u_int32_t *tl; 2103 2104 if (before_ret) { 2105 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2106 *tl = nfs_false; 2107 } else { 2108 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2109 *tl++ = nfs_true; 2110 txdr_hyper(before_vap->va_size, tl); 2111 tl += 2; 2112 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2113 tl += 2; 2114 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2115 } 2116 *bposp = bpos; 2117 *mbp = mb; 2118 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2119 } 2120 2121 void 2122 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2123 struct nfsrv_descript *nfsd; 2124 int after_ret; 2125 struct vattr *after_vap; 2126 struct mbuf **mbp; 2127 char **bposp; 2128 { 2129 struct mbuf *mb = *mbp, *mb2; 2130 char *bpos = *bposp; 2131 u_int32_t *tl; 2132 struct nfs_fattr *fp; 2133 2134 if (after_ret) { 2135 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2136 *tl = nfs_false; 2137 } else { 2138 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2139 *tl++ = nfs_true; 2140 fp = (struct nfs_fattr *)tl; 2141 nfsm_srvfattr(nfsd, after_vap, fp); 2142 } 2143 *mbp = mb; 2144 *bposp = bpos; 2145 } 2146 2147 void 2148 nfsm_srvfattr(nfsd, vap, fp) 2149 struct nfsrv_descript *nfsd; 2150 struct vattr *vap; 2151 struct nfs_fattr *fp; 2152 { 2153 2154 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2155 fp->fa_uid = txdr_unsigned(vap->va_uid); 2156 fp->fa_gid = txdr_unsigned(vap->va_gid); 2157 if (nfsd->nd_flag & ND_NFSV3) { 2158 fp->fa_type = vtonfsv3_type(vap->va_type); 2159 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2160 txdr_hyper(vap->va_size, &fp->fa3_size); 2161 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2162 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2163 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2164 fp->fa3_fsid.nfsuquad[0] = 0; 2165 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2166 fp->fa3_fileid.nfsuquad[0] = 0; 2167 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); 2168 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2169 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2170 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2171 } else { 2172 fp->fa_type = vtonfsv2_type(vap->va_type); 2173 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2174 fp->fa2_size = txdr_unsigned(vap->va_size); 2175 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2176 if (vap->va_type == VFIFO) 2177 fp->fa2_rdev = 0xffffffff; 2178 else 2179 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2180 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2181 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2182 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2183 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2184 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2185 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2186 } 2187 } 2188 2189 /* 2190 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2191 * - look up fsid in mount list (if not found ret error) 2192 * - get vp and export rights by calling VFS_FHTOVP() 2193 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2194 * - if not lockflag unlock it with VOP_UNLOCK() 2195 */ 2196 int 2197 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) 2198 fhandle_t *fhp; 2199 int lockflag; 2200 struct vnode **vpp; 2201 struct ucred *cred; 2202 struct nfssvc_sock *slp; 2203 struct mbuf *nam; 2204 int *rdonlyp; 2205 int kerbflag; 2206 { 2207 struct mount *mp; 2208 int i; 2209 struct ucred *credanon; 2210 int error, exflags; 2211 struct sockaddr_in *saddr; 2212 2213 *vpp = (struct vnode *)0; 2214 2215 if (nfs_ispublicfh(fhp)) { 2216 if (!pubflag || !nfs_pub.np_valid) 2217 return (ESTALE); 2218 fhp = &nfs_pub.np_handle; 2219 } 2220 2221 mp = vfs_getvfs(&fhp->fh_fsid); 2222 if (!mp) 2223 return (ESTALE); 2224 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); 2225 if (error) 2226 return (error); 2227 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2228 if (error) 2229 return (error); 2230 2231 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2232 saddr = mtod(nam, struct sockaddr_in *); 2233 if ((saddr->sin_family == AF_INET) && 2234 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2235 vput(*vpp); 2236 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2237 } 2238 #ifdef INET6 2239 if ((saddr->sin_family == AF_INET6) && 2240 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2241 vput(*vpp); 2242 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2243 } 2244 #endif 2245 } 2246 /* 2247 * Check/setup credentials. 2248 */ 2249 if (exflags & MNT_EXKERB) { 2250 if (!kerbflag) { 2251 vput(*vpp); 2252 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2253 } 2254 } else if (kerbflag) { 2255 vput(*vpp); 2256 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2257 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { 2258 cred->cr_uid = credanon->cr_uid; 2259 cred->cr_gid = credanon->cr_gid; 2260 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) 2261 cred->cr_groups[i] = credanon->cr_groups[i]; 2262 cred->cr_ngroups = i; 2263 } 2264 if (exflags & MNT_EXRDONLY) 2265 *rdonlyp = 1; 2266 else 2267 *rdonlyp = 0; 2268 if (!lockflag) 2269 VOP_UNLOCK(*vpp, 0); 2270 return (0); 2271 } 2272 2273 /* 2274 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2275 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has 2276 * transformed this to all zeroes in both cases, so check for it. 2277 */ 2278 int 2279 nfs_ispublicfh(fhp) 2280 fhandle_t *fhp; 2281 { 2282 char *cp = (char *)fhp; 2283 int i; 2284 2285 for (i = 0; i < NFSX_V3FH; i++) 2286 if (*cp++ != 0) 2287 return (FALSE); 2288 return (TRUE); 2289 } 2290 2291 /* 2292 * This function compares two net addresses by family and returns TRUE 2293 * if they are the same host. 2294 * If there is any doubt, return FALSE. 2295 * The AF_INET family is handled as a special case so that address mbufs 2296 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2297 */ 2298 int 2299 netaddr_match(family, haddr, nam) 2300 int family; 2301 union nethostaddr *haddr; 2302 struct mbuf *nam; 2303 { 2304 struct sockaddr_in *inetaddr; 2305 2306 switch (family) { 2307 case AF_INET: 2308 inetaddr = mtod(nam, struct sockaddr_in *); 2309 if (inetaddr->sin_family == AF_INET && 2310 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2311 return (1); 2312 break; 2313 #ifdef INET6 2314 case AF_INET6: 2315 { 2316 struct sockaddr_in6 *sin6_1, *sin6_2; 2317 2318 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2319 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2320 if (sin6_1->sin6_family == AF_INET6 && 2321 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2322 return 1; 2323 } 2324 #endif 2325 #ifdef ISO 2326 case AF_ISO: 2327 { 2328 struct sockaddr_iso *isoaddr1, *isoaddr2; 2329 2330 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2331 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2332 if (isoaddr1->siso_family == AF_ISO && 2333 isoaddr1->siso_nlen > 0 && 2334 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2335 SAME_ISOADDR(isoaddr1, isoaddr2)) 2336 return (1); 2337 break; 2338 } 2339 #endif /* ISO */ 2340 default: 2341 break; 2342 }; 2343 return (0); 2344 } 2345 2346 2347 /* 2348 * The write verifier has changed (probably due to a server reboot), so all 2349 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 2350 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 2351 * flag. Once done the new write verifier can be set for the mount point. 2352 */ 2353 void 2354 nfs_clearcommit(mp) 2355 struct mount *mp; 2356 { 2357 struct vnode *vp, *nvp; 2358 struct buf *bp, *nbp; 2359 int s; 2360 2361 s = splbio(); 2362 loop: 2363 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 2364 if (vp->v_mount != mp) /* Paranoia */ 2365 goto loop; 2366 nvp = vp->v_mntvnodes.le_next; 2367 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2368 nbp = bp->b_vnbufs.le_next; 2369 if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) 2370 == (B_DELWRI | B_NEEDCOMMIT)) 2371 bp->b_flags &= ~B_NEEDCOMMIT; 2372 } 2373 } 2374 splx(s); 2375 } 2376 2377 /* 2378 * Map errnos to NFS error numbers. For Version 3 also filter out error 2379 * numbers not specified for the associated procedure. 2380 */ 2381 int 2382 nfsrv_errmap(nd, err) 2383 struct nfsrv_descript *nd; 2384 int err; 2385 { 2386 short *defaulterrp, *errp; 2387 2388 if (nd->nd_flag & ND_NFSV3) { 2389 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2390 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2391 while (*++errp) { 2392 if (*errp == err) 2393 return (err); 2394 else if (*errp > err) 2395 break; 2396 } 2397 return ((int)*defaulterrp); 2398 } else 2399 return (err & 0xffff); 2400 } 2401 if (err <= ELAST) 2402 return ((int)nfsrv_v2errmap[err - 1]); 2403 return (NFSERR_IO); 2404 } 2405 2406 /* 2407 * Sort the group list in increasing numerical order. 2408 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort 2409 * that used to be here.) 2410 */ 2411 void 2412 nfsrvw_sort(list, num) 2413 gid_t *list; 2414 int num; 2415 { 2416 int i, j; 2417 gid_t v; 2418 2419 /* Insertion sort. */ 2420 for (i = 1; i < num; i++) { 2421 v = list[i]; 2422 /* find correct slot for value v, moving others up */ 2423 for (j = i; --j >= 0 && v < list[j];) 2424 list[j + 1] = list[j]; 2425 list[j + 1] = v; 2426 } 2427 } 2428 2429 /* 2430 * copy credentials making sure that the result can be compared with memcmp(). 2431 */ 2432 void 2433 nfsrv_setcred(incred, outcred) 2434 struct ucred *incred, *outcred; 2435 { 2436 int i; 2437 2438 memset((caddr_t)outcred, 0, sizeof (struct ucred)); 2439 outcred->cr_ref = 1; 2440 outcred->cr_uid = incred->cr_uid; 2441 outcred->cr_gid = incred->cr_gid; 2442 outcred->cr_ngroups = incred->cr_ngroups; 2443 for (i = 0; i < incred->cr_ngroups; i++) 2444 outcred->cr_groups[i] = incred->cr_groups[i]; 2445 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); 2446 } 2447
Indexes created Thu Sep 25 15:09:46 GMT 2025