nfs_subs.c revision 1.78
1 /* $NetBSD: nfs_subs.c,v 1.78 2000/06/27 17:52:34 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 <uvm/uvm_extern.h> 66 67 #include <nfs/rpcv2.h> 68 #include <nfs/nfsproto.h> 69 #include <nfs/nfsnode.h> 70 #include <nfs/nfs.h> 71 #include <nfs/xdr_subs.h> 72 #include <nfs/nfsm_subs.h> 73 #include <nfs/nfsmount.h> 74 #include <nfs/nqnfs.h> 75 #include <nfs/nfsrtt.h> 76 #include <nfs/nfs_var.h> 77 78 #include <miscfs/specfs/specdev.h> 79 80 #include <vm/vm.h> 81 82 #include <netinet/in.h> 83 #ifdef ISO 84 #include <netiso/iso.h> 85 #endif 86 87 /* 88 * Data items converted to xdr at startup, since they are constant 89 * This is kinda hokey, but may save a little time doing byte swaps 90 */ 91 u_int32_t nfs_xdrneg1; 92 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 93 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 94 rpc_auth_kerb; 95 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; 96 97 /* And other global data */ 98 static u_int32_t nfs_xid = 0; 99 nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, 100 NFCHR, NFNON }; 101 nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, 102 NFFIFO, NFNON }; 103 enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 104 enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 105 int nfs_ticks; 106 107 /* NFS client/server stats. */ 108 struct nfsstats nfsstats; 109 110 /* 111 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 112 */ 113 int nfsv3_procid[NFS_NPROCS] = { 114 NFSPROC_NULL, 115 NFSPROC_GETATTR, 116 NFSPROC_SETATTR, 117 NFSPROC_NOOP, 118 NFSPROC_LOOKUP, 119 NFSPROC_READLINK, 120 NFSPROC_READ, 121 NFSPROC_NOOP, 122 NFSPROC_WRITE, 123 NFSPROC_CREATE, 124 NFSPROC_REMOVE, 125 NFSPROC_RENAME, 126 NFSPROC_LINK, 127 NFSPROC_SYMLINK, 128 NFSPROC_MKDIR, 129 NFSPROC_RMDIR, 130 NFSPROC_READDIR, 131 NFSPROC_FSSTAT, 132 NFSPROC_NOOP, 133 NFSPROC_NOOP, 134 NFSPROC_NOOP, 135 NFSPROC_NOOP, 136 NFSPROC_NOOP, 137 NFSPROC_NOOP, 138 NFSPROC_NOOP, 139 NFSPROC_NOOP 140 }; 141 142 /* 143 * and the reverse mapping from generic to Version 2 procedure numbers 144 */ 145 int nfsv2_procid[NFS_NPROCS] = { 146 NFSV2PROC_NULL, 147 NFSV2PROC_GETATTR, 148 NFSV2PROC_SETATTR, 149 NFSV2PROC_LOOKUP, 150 NFSV2PROC_NOOP, 151 NFSV2PROC_READLINK, 152 NFSV2PROC_READ, 153 NFSV2PROC_WRITE, 154 NFSV2PROC_CREATE, 155 NFSV2PROC_MKDIR, 156 NFSV2PROC_SYMLINK, 157 NFSV2PROC_CREATE, 158 NFSV2PROC_REMOVE, 159 NFSV2PROC_RMDIR, 160 NFSV2PROC_RENAME, 161 NFSV2PROC_LINK, 162 NFSV2PROC_READDIR, 163 NFSV2PROC_NOOP, 164 NFSV2PROC_STATFS, 165 NFSV2PROC_NOOP, 166 NFSV2PROC_NOOP, 167 NFSV2PROC_NOOP, 168 NFSV2PROC_NOOP, 169 NFSV2PROC_NOOP, 170 NFSV2PROC_NOOP, 171 NFSV2PROC_NOOP, 172 }; 173 174 /* 175 * Maps errno values to nfs error numbers. 176 * Use NFSERR_IO as the catch all for ones not specifically defined in 177 * RFC 1094. 178 */ 179 static u_char nfsrv_v2errmap[ELAST] = { 180 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 181 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 182 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 183 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 184 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 185 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 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_IO, NFSERR_IO, NFSERR_IO, 192 NFSERR_IO, NFSERR_IO, NFSERR_NAMETOL, NFSERR_IO, NFSERR_IO, 193 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 194 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 195 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 196 NFSERR_IO, NFSERR_IO, 197 }; 198 199 /* 200 * Maps errno values to nfs error numbers. 201 * Although it is not obvious whether or not NFS clients really care if 202 * a returned error value is in the specified list for the procedure, the 203 * safest thing to do is filter them appropriately. For Version 2, the 204 * X/Open XNFS document is the only specification that defines error values 205 * for each RPC (The RFC simply lists all possible error values for all RPCs), 206 * so I have decided to not do this for Version 2. 207 * The first entry is the default error return and the rest are the valid 208 * errors for that RPC in increasing numeric order. 209 */ 210 static short nfsv3err_null[] = { 211 0, 212 0, 213 }; 214 215 static short nfsv3err_getattr[] = { 216 NFSERR_IO, 217 NFSERR_IO, 218 NFSERR_STALE, 219 NFSERR_BADHANDLE, 220 NFSERR_SERVERFAULT, 221 0, 222 }; 223 224 static short nfsv3err_setattr[] = { 225 NFSERR_IO, 226 NFSERR_PERM, 227 NFSERR_IO, 228 NFSERR_ACCES, 229 NFSERR_INVAL, 230 NFSERR_NOSPC, 231 NFSERR_ROFS, 232 NFSERR_DQUOT, 233 NFSERR_STALE, 234 NFSERR_BADHANDLE, 235 NFSERR_NOT_SYNC, 236 NFSERR_SERVERFAULT, 237 0, 238 }; 239 240 static short nfsv3err_lookup[] = { 241 NFSERR_IO, 242 NFSERR_NOENT, 243 NFSERR_IO, 244 NFSERR_ACCES, 245 NFSERR_NOTDIR, 246 NFSERR_NAMETOL, 247 NFSERR_STALE, 248 NFSERR_BADHANDLE, 249 NFSERR_SERVERFAULT, 250 0, 251 }; 252 253 static short nfsv3err_access[] = { 254 NFSERR_IO, 255 NFSERR_IO, 256 NFSERR_STALE, 257 NFSERR_BADHANDLE, 258 NFSERR_SERVERFAULT, 259 0, 260 }; 261 262 static short nfsv3err_readlink[] = { 263 NFSERR_IO, 264 NFSERR_IO, 265 NFSERR_ACCES, 266 NFSERR_INVAL, 267 NFSERR_STALE, 268 NFSERR_BADHANDLE, 269 NFSERR_NOTSUPP, 270 NFSERR_SERVERFAULT, 271 0, 272 }; 273 274 static short nfsv3err_read[] = { 275 NFSERR_IO, 276 NFSERR_IO, 277 NFSERR_NXIO, 278 NFSERR_ACCES, 279 NFSERR_INVAL, 280 NFSERR_STALE, 281 NFSERR_BADHANDLE, 282 NFSERR_SERVERFAULT, 283 NFSERR_JUKEBOX, 284 0, 285 }; 286 287 static short nfsv3err_write[] = { 288 NFSERR_IO, 289 NFSERR_IO, 290 NFSERR_ACCES, 291 NFSERR_INVAL, 292 NFSERR_FBIG, 293 NFSERR_NOSPC, 294 NFSERR_ROFS, 295 NFSERR_DQUOT, 296 NFSERR_STALE, 297 NFSERR_BADHANDLE, 298 NFSERR_SERVERFAULT, 299 NFSERR_JUKEBOX, 300 0, 301 }; 302 303 static short nfsv3err_create[] = { 304 NFSERR_IO, 305 NFSERR_IO, 306 NFSERR_ACCES, 307 NFSERR_EXIST, 308 NFSERR_NOTDIR, 309 NFSERR_NOSPC, 310 NFSERR_ROFS, 311 NFSERR_NAMETOL, 312 NFSERR_DQUOT, 313 NFSERR_STALE, 314 NFSERR_BADHANDLE, 315 NFSERR_NOTSUPP, 316 NFSERR_SERVERFAULT, 317 0, 318 }; 319 320 static short nfsv3err_mkdir[] = { 321 NFSERR_IO, 322 NFSERR_IO, 323 NFSERR_ACCES, 324 NFSERR_EXIST, 325 NFSERR_NOTDIR, 326 NFSERR_NOSPC, 327 NFSERR_ROFS, 328 NFSERR_NAMETOL, 329 NFSERR_DQUOT, 330 NFSERR_STALE, 331 NFSERR_BADHANDLE, 332 NFSERR_NOTSUPP, 333 NFSERR_SERVERFAULT, 334 0, 335 }; 336 337 static short nfsv3err_symlink[] = { 338 NFSERR_IO, 339 NFSERR_IO, 340 NFSERR_ACCES, 341 NFSERR_EXIST, 342 NFSERR_NOTDIR, 343 NFSERR_NOSPC, 344 NFSERR_ROFS, 345 NFSERR_NAMETOL, 346 NFSERR_DQUOT, 347 NFSERR_STALE, 348 NFSERR_BADHANDLE, 349 NFSERR_NOTSUPP, 350 NFSERR_SERVERFAULT, 351 0, 352 }; 353 354 static short nfsv3err_mknod[] = { 355 NFSERR_IO, 356 NFSERR_IO, 357 NFSERR_ACCES, 358 NFSERR_EXIST, 359 NFSERR_NOTDIR, 360 NFSERR_NOSPC, 361 NFSERR_ROFS, 362 NFSERR_NAMETOL, 363 NFSERR_DQUOT, 364 NFSERR_STALE, 365 NFSERR_BADHANDLE, 366 NFSERR_NOTSUPP, 367 NFSERR_SERVERFAULT, 368 NFSERR_BADTYPE, 369 0, 370 }; 371 372 static short nfsv3err_remove[] = { 373 NFSERR_IO, 374 NFSERR_NOENT, 375 NFSERR_IO, 376 NFSERR_ACCES, 377 NFSERR_NOTDIR, 378 NFSERR_ROFS, 379 NFSERR_NAMETOL, 380 NFSERR_STALE, 381 NFSERR_BADHANDLE, 382 NFSERR_SERVERFAULT, 383 0, 384 }; 385 386 static short nfsv3err_rmdir[] = { 387 NFSERR_IO, 388 NFSERR_NOENT, 389 NFSERR_IO, 390 NFSERR_ACCES, 391 NFSERR_EXIST, 392 NFSERR_NOTDIR, 393 NFSERR_INVAL, 394 NFSERR_ROFS, 395 NFSERR_NAMETOL, 396 NFSERR_NOTEMPTY, 397 NFSERR_STALE, 398 NFSERR_BADHANDLE, 399 NFSERR_NOTSUPP, 400 NFSERR_SERVERFAULT, 401 0, 402 }; 403 404 static short nfsv3err_rename[] = { 405 NFSERR_IO, 406 NFSERR_NOENT, 407 NFSERR_IO, 408 NFSERR_ACCES, 409 NFSERR_EXIST, 410 NFSERR_XDEV, 411 NFSERR_NOTDIR, 412 NFSERR_ISDIR, 413 NFSERR_INVAL, 414 NFSERR_NOSPC, 415 NFSERR_ROFS, 416 NFSERR_MLINK, 417 NFSERR_NAMETOL, 418 NFSERR_NOTEMPTY, 419 NFSERR_DQUOT, 420 NFSERR_STALE, 421 NFSERR_BADHANDLE, 422 NFSERR_NOTSUPP, 423 NFSERR_SERVERFAULT, 424 0, 425 }; 426 427 static short nfsv3err_link[] = { 428 NFSERR_IO, 429 NFSERR_IO, 430 NFSERR_ACCES, 431 NFSERR_EXIST, 432 NFSERR_XDEV, 433 NFSERR_NOTDIR, 434 NFSERR_INVAL, 435 NFSERR_NOSPC, 436 NFSERR_ROFS, 437 NFSERR_MLINK, 438 NFSERR_NAMETOL, 439 NFSERR_DQUOT, 440 NFSERR_STALE, 441 NFSERR_BADHANDLE, 442 NFSERR_NOTSUPP, 443 NFSERR_SERVERFAULT, 444 0, 445 }; 446 447 static short nfsv3err_readdir[] = { 448 NFSERR_IO, 449 NFSERR_IO, 450 NFSERR_ACCES, 451 NFSERR_NOTDIR, 452 NFSERR_STALE, 453 NFSERR_BADHANDLE, 454 NFSERR_BAD_COOKIE, 455 NFSERR_TOOSMALL, 456 NFSERR_SERVERFAULT, 457 0, 458 }; 459 460 static short nfsv3err_readdirplus[] = { 461 NFSERR_IO, 462 NFSERR_IO, 463 NFSERR_ACCES, 464 NFSERR_NOTDIR, 465 NFSERR_STALE, 466 NFSERR_BADHANDLE, 467 NFSERR_BAD_COOKIE, 468 NFSERR_NOTSUPP, 469 NFSERR_TOOSMALL, 470 NFSERR_SERVERFAULT, 471 0, 472 }; 473 474 static short nfsv3err_fsstat[] = { 475 NFSERR_IO, 476 NFSERR_IO, 477 NFSERR_STALE, 478 NFSERR_BADHANDLE, 479 NFSERR_SERVERFAULT, 480 0, 481 }; 482 483 static short nfsv3err_fsinfo[] = { 484 NFSERR_STALE, 485 NFSERR_STALE, 486 NFSERR_BADHANDLE, 487 NFSERR_SERVERFAULT, 488 0, 489 }; 490 491 static short nfsv3err_pathconf[] = { 492 NFSERR_STALE, 493 NFSERR_STALE, 494 NFSERR_BADHANDLE, 495 NFSERR_SERVERFAULT, 496 0, 497 }; 498 499 static short nfsv3err_commit[] = { 500 NFSERR_IO, 501 NFSERR_IO, 502 NFSERR_STALE, 503 NFSERR_BADHANDLE, 504 NFSERR_SERVERFAULT, 505 0, 506 }; 507 508 static short *nfsrv_v3errmap[] = { 509 nfsv3err_null, 510 nfsv3err_getattr, 511 nfsv3err_setattr, 512 nfsv3err_lookup, 513 nfsv3err_access, 514 nfsv3err_readlink, 515 nfsv3err_read, 516 nfsv3err_write, 517 nfsv3err_create, 518 nfsv3err_mkdir, 519 nfsv3err_symlink, 520 nfsv3err_mknod, 521 nfsv3err_remove, 522 nfsv3err_rmdir, 523 nfsv3err_rename, 524 nfsv3err_link, 525 nfsv3err_readdir, 526 nfsv3err_readdirplus, 527 nfsv3err_fsstat, 528 nfsv3err_fsinfo, 529 nfsv3err_pathconf, 530 nfsv3err_commit, 531 }; 532 533 extern struct nfsrtt nfsrtt; 534 extern time_t nqnfsstarttime; 535 extern int nqsrv_clockskew; 536 extern int nqsrv_writeslack; 537 extern int nqsrv_maxlease; 538 extern int nqnfs_piggy[NFS_NPROCS]; 539 extern struct nfsnodehashhead *nfsnodehashtbl; 540 extern u_long nfsnodehash; 541 542 LIST_HEAD(nfsnodehashhead, nfsnode); 543 u_long nfsdirhashmask; 544 545 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 546 547 /* 548 * Create the header for an rpc request packet 549 * The hsiz is the size of the rest of the nfs request header. 550 * (just used to decide if a cluster is a good idea) 551 */ 552 struct mbuf * 553 nfsm_reqh(vp, procid, hsiz, bposp) 554 struct vnode *vp; 555 u_long procid; 556 int hsiz; 557 caddr_t *bposp; 558 { 559 struct mbuf *mb; 560 u_int32_t *tl; 561 caddr_t bpos; 562 struct mbuf *mb2; 563 struct nfsmount *nmp; 564 int nqflag; 565 566 MGET(mb, M_WAIT, MT_DATA); 567 if (hsiz >= MINCLSIZE) 568 MCLGET(mb, M_WAIT); 569 mb->m_len = 0; 570 bpos = mtod(mb, caddr_t); 571 572 /* 573 * For NQNFS, add lease request. 574 */ 575 if (vp) { 576 nmp = VFSTONFS(vp->v_mount); 577 if (nmp->nm_flag & NFSMNT_NQNFS) { 578 nqflag = NQNFS_NEEDLEASE(vp, procid); 579 if (nqflag) { 580 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); 581 *tl++ = txdr_unsigned(nqflag); 582 *tl = txdr_unsigned(nmp->nm_leaseterm); 583 } else { 584 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 585 *tl = 0; 586 } 587 } 588 } 589 /* Finally, return values */ 590 *bposp = bpos; 591 return (mb); 592 } 593 594 /* 595 * Build the RPC header and fill in the authorization info. 596 * The authorization string argument is only used when the credentials 597 * come from outside of the kernel. 598 * Returns the head of the mbuf list. 599 */ 600 struct mbuf * 601 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 602 verf_str, mrest, mrest_len, mbp, xidp) 603 struct ucred *cr; 604 int nmflag; 605 int procid; 606 int auth_type; 607 int auth_len; 608 char *auth_str; 609 int verf_len; 610 char *verf_str; 611 struct mbuf *mrest; 612 int mrest_len; 613 struct mbuf **mbp; 614 u_int32_t *xidp; 615 { 616 struct mbuf *mb; 617 u_int32_t *tl; 618 caddr_t bpos; 619 int i; 620 struct mbuf *mreq, *mb2; 621 int siz, grpsiz, authsiz; 622 struct timeval tv; 623 static u_int32_t base; 624 625 authsiz = nfsm_rndup(auth_len); 626 MGETHDR(mb, M_WAIT, MT_DATA); 627 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 628 MCLGET(mb, M_WAIT); 629 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 630 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 631 } else { 632 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 633 } 634 mb->m_len = 0; 635 mreq = mb; 636 bpos = mtod(mb, caddr_t); 637 638 /* 639 * First the RPC header. 640 */ 641 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 642 643 /* 644 * derive initial xid from system time 645 * XXX time is invalid if root not yet mounted 646 */ 647 if (!base && (rootvp)) { 648 microtime(&tv); 649 base = tv.tv_sec << 12; 650 nfs_xid = base; 651 } 652 /* 653 * Skip zero xid if it should ever happen. 654 */ 655 if (++nfs_xid == 0) 656 nfs_xid++; 657 658 *tl++ = *xidp = txdr_unsigned(nfs_xid); 659 *tl++ = rpc_call; 660 *tl++ = rpc_vers; 661 if (nmflag & NFSMNT_NQNFS) { 662 *tl++ = txdr_unsigned(NQNFS_PROG); 663 *tl++ = txdr_unsigned(NQNFS_VER3); 664 } else { 665 *tl++ = txdr_unsigned(NFS_PROG); 666 if (nmflag & NFSMNT_NFSV3) 667 *tl++ = txdr_unsigned(NFS_VER3); 668 else 669 *tl++ = txdr_unsigned(NFS_VER2); 670 } 671 if (nmflag & NFSMNT_NFSV3) 672 *tl++ = txdr_unsigned(procid); 673 else 674 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 675 676 /* 677 * And then the authorization cred. 678 */ 679 *tl++ = txdr_unsigned(auth_type); 680 *tl = txdr_unsigned(authsiz); 681 switch (auth_type) { 682 case RPCAUTH_UNIX: 683 nfsm_build(tl, u_int32_t *, auth_len); 684 *tl++ = 0; /* stamp ?? */ 685 *tl++ = 0; /* NULL hostname */ 686 *tl++ = txdr_unsigned(cr->cr_uid); 687 *tl++ = txdr_unsigned(cr->cr_gid); 688 grpsiz = (auth_len >> 2) - 5; 689 *tl++ = txdr_unsigned(grpsiz); 690 for (i = 0; i < grpsiz; i++) 691 *tl++ = txdr_unsigned(cr->cr_groups[i]); 692 break; 693 case RPCAUTH_KERB4: 694 siz = auth_len; 695 while (siz > 0) { 696 if (M_TRAILINGSPACE(mb) == 0) { 697 MGET(mb2, M_WAIT, MT_DATA); 698 if (siz >= MINCLSIZE) 699 MCLGET(mb2, M_WAIT); 700 mb->m_next = mb2; 701 mb = mb2; 702 mb->m_len = 0; 703 bpos = mtod(mb, caddr_t); 704 } 705 i = min(siz, M_TRAILINGSPACE(mb)); 706 memcpy(bpos, auth_str, i); 707 mb->m_len += i; 708 auth_str += i; 709 bpos += i; 710 siz -= i; 711 } 712 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 713 for (i = 0; i < siz; i++) 714 *bpos++ = '\0'; 715 mb->m_len += siz; 716 } 717 break; 718 }; 719 720 /* 721 * And the verifier... 722 */ 723 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 724 if (verf_str) { 725 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 726 *tl = txdr_unsigned(verf_len); 727 siz = verf_len; 728 while (siz > 0) { 729 if (M_TRAILINGSPACE(mb) == 0) { 730 MGET(mb2, M_WAIT, MT_DATA); 731 if (siz >= MINCLSIZE) 732 MCLGET(mb2, M_WAIT); 733 mb->m_next = mb2; 734 mb = mb2; 735 mb->m_len = 0; 736 bpos = mtod(mb, caddr_t); 737 } 738 i = min(siz, M_TRAILINGSPACE(mb)); 739 memcpy(bpos, verf_str, i); 740 mb->m_len += i; 741 verf_str += i; 742 bpos += i; 743 siz -= i; 744 } 745 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 746 for (i = 0; i < siz; i++) 747 *bpos++ = '\0'; 748 mb->m_len += siz; 749 } 750 } else { 751 *tl++ = txdr_unsigned(RPCAUTH_NULL); 752 *tl = 0; 753 } 754 mb->m_next = mrest; 755 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 756 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 757 *mbp = mb; 758 return (mreq); 759 } 760 761 /* 762 * copies mbuf chain to the uio scatter/gather list 763 */ 764 int 765 nfsm_mbuftouio(mrep, uiop, siz, dpos) 766 struct mbuf **mrep; 767 struct uio *uiop; 768 int siz; 769 caddr_t *dpos; 770 { 771 char *mbufcp, *uiocp; 772 int xfer, left, len; 773 struct mbuf *mp; 774 long uiosiz, rem; 775 int error = 0; 776 777 mp = *mrep; 778 mbufcp = *dpos; 779 len = mtod(mp, caddr_t)+mp->m_len-mbufcp; 780 rem = nfsm_rndup(siz)-siz; 781 while (siz > 0) { 782 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 783 return (EFBIG); 784 left = uiop->uio_iov->iov_len; 785 uiocp = uiop->uio_iov->iov_base; 786 if (left > siz) 787 left = siz; 788 uiosiz = left; 789 while (left > 0) { 790 while (len == 0) { 791 mp = mp->m_next; 792 if (mp == NULL) 793 return (EBADRPC); 794 mbufcp = mtod(mp, caddr_t); 795 len = mp->m_len; 796 } 797 xfer = (left > len) ? len : left; 798 #ifdef notdef 799 /* Not Yet.. */ 800 if (uiop->uio_iov->iov_op != NULL) 801 (*(uiop->uio_iov->iov_op)) 802 (mbufcp, uiocp, xfer); 803 else 804 #endif 805 if (uiop->uio_segflg == UIO_SYSSPACE) 806 memcpy(uiocp, mbufcp, xfer); 807 else 808 copyout(mbufcp, uiocp, xfer); 809 left -= xfer; 810 len -= xfer; 811 mbufcp += xfer; 812 uiocp += xfer; 813 uiop->uio_offset += xfer; 814 uiop->uio_resid -= xfer; 815 } 816 if (uiop->uio_iov->iov_len <= siz) { 817 uiop->uio_iovcnt--; 818 uiop->uio_iov++; 819 } else { 820 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 821 uiop->uio_iov->iov_len -= uiosiz; 822 } 823 siz -= uiosiz; 824 } 825 *dpos = mbufcp; 826 *mrep = mp; 827 if (rem > 0) { 828 if (len < rem) 829 error = nfs_adv(mrep, dpos, rem, len); 830 else 831 *dpos += rem; 832 } 833 return (error); 834 } 835 836 /* 837 * copies a uio scatter/gather list to an mbuf chain. 838 * NOTE: can ony handle iovcnt == 1 839 */ 840 int 841 nfsm_uiotombuf(uiop, mq, siz, bpos) 842 struct uio *uiop; 843 struct mbuf **mq; 844 int siz; 845 caddr_t *bpos; 846 { 847 char *uiocp; 848 struct mbuf *mp, *mp2; 849 int xfer, left, mlen; 850 int uiosiz, clflg, rem; 851 char *cp; 852 853 #ifdef DIAGNOSTIC 854 if (uiop->uio_iovcnt != 1) 855 panic("nfsm_uiotombuf: iovcnt != 1"); 856 #endif 857 858 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 859 clflg = 1; 860 else 861 clflg = 0; 862 rem = nfsm_rndup(siz)-siz; 863 mp = mp2 = *mq; 864 while (siz > 0) { 865 left = uiop->uio_iov->iov_len; 866 uiocp = uiop->uio_iov->iov_base; 867 if (left > siz) 868 left = siz; 869 uiosiz = left; 870 while (left > 0) { 871 mlen = M_TRAILINGSPACE(mp); 872 if (mlen == 0) { 873 MGET(mp, M_WAIT, MT_DATA); 874 if (clflg) 875 MCLGET(mp, M_WAIT); 876 mp->m_len = 0; 877 mp2->m_next = mp; 878 mp2 = mp; 879 mlen = M_TRAILINGSPACE(mp); 880 } 881 xfer = (left > mlen) ? mlen : left; 882 #ifdef notdef 883 /* Not Yet.. */ 884 if (uiop->uio_iov->iov_op != NULL) 885 (*(uiop->uio_iov->iov_op)) 886 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 887 else 888 #endif 889 if (uiop->uio_segflg == UIO_SYSSPACE) 890 memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer); 891 else 892 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 893 mp->m_len += xfer; 894 left -= xfer; 895 uiocp += xfer; 896 uiop->uio_offset += xfer; 897 uiop->uio_resid -= xfer; 898 } 899 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 900 uiop->uio_iov->iov_len -= uiosiz; 901 siz -= uiosiz; 902 } 903 if (rem > 0) { 904 if (rem > M_TRAILINGSPACE(mp)) { 905 MGET(mp, M_WAIT, MT_DATA); 906 mp->m_len = 0; 907 mp2->m_next = mp; 908 } 909 cp = mtod(mp, caddr_t)+mp->m_len; 910 for (left = 0; left < rem; left++) 911 *cp++ = '\0'; 912 mp->m_len += rem; 913 *bpos = cp; 914 } else 915 *bpos = mtod(mp, caddr_t)+mp->m_len; 916 *mq = mp; 917 return (0); 918 } 919 920 /* 921 * Get at least "siz" bytes of correctly aligned data. 922 * When called the mbuf pointers are not necessarily correct, 923 * dsosp points to what ought to be in m_data and left contains 924 * what ought to be in m_len. 925 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 926 * cases. (The macros use the vars. dpos and dpos2) 927 */ 928 int 929 nfsm_disct(mdp, dposp, siz, left, cp2) 930 struct mbuf **mdp; 931 caddr_t *dposp; 932 int siz; 933 int left; 934 caddr_t *cp2; 935 { 936 struct mbuf *m1, *m2; 937 struct mbuf *havebuf = NULL; 938 caddr_t src = *dposp; 939 caddr_t dst; 940 int len; 941 942 #ifdef DEBUG 943 if (left < 0) 944 panic("nfsm_disct: left < 0"); 945 #endif 946 m1 = *mdp; 947 /* 948 * Skip through the mbuf chain looking for an mbuf with 949 * some data. If the first mbuf found has enough data 950 * and it is correctly aligned return it. 951 */ 952 while (left == 0) { 953 havebuf = m1; 954 *mdp = m1 = m1->m_next; 955 if (m1 == NULL) 956 return (EBADRPC); 957 src = mtod(m1, caddr_t); 958 left = m1->m_len; 959 /* 960 * If we start a new mbuf and it is big enough 961 * and correctly aligned just return it, don't 962 * do any pull up. 963 */ 964 if (left >= siz && nfsm_aligned(src)) { 965 *cp2 = src; 966 *dposp = src + siz; 967 return (0); 968 } 969 } 970 if (m1->m_flags & M_EXT) { 971 if (havebuf) { 972 /* If the first mbuf with data has external data 973 * and there is a previous empty mbuf use it 974 * to move the data into. 975 */ 976 m2 = m1; 977 *mdp = m1 = havebuf; 978 if (m1->m_flags & M_EXT) { 979 MEXTREMOVE(m1); 980 } 981 } else { 982 /* 983 * If the first mbuf has a external data 984 * and there is no previous empty mbuf 985 * allocate a new mbuf and move the external 986 * data to the new mbuf. Also make the first 987 * mbuf look empty. 988 */ 989 m2 = m_get(M_WAIT, MT_DATA); 990 m2->m_ext = m1->m_ext; 991 m2->m_data = src; 992 m2->m_len = left; 993 MCLADDREFERENCE(m1, m2); 994 MEXTREMOVE(m1); 995 m2->m_next = m1->m_next; 996 m1->m_next = m2; 997 } 998 m1->m_len = 0; 999 dst = m1->m_dat; 1000 } else { 1001 /* 1002 * If the first mbuf has no external data 1003 * move the data to the front of the mbuf. 1004 */ 1005 if ((dst = m1->m_dat) != src) 1006 memmove(dst, src, left); 1007 dst += left; 1008 m1->m_len = left; 1009 m2 = m1->m_next; 1010 } 1011 m1->m_flags &= ~M_PKTHDR; 1012 *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */ 1013 *dposp = mtod(m1, caddr_t) + siz; 1014 /* 1015 * Loop through mbufs pulling data up into first mbuf until 1016 * the first mbuf is full or there is no more data to 1017 * pullup. 1018 */ 1019 while ((len = (MLEN - m1->m_len)) != 0 && m2) { 1020 if ((len = min(len, m2->m_len)) != 0) 1021 memcpy(dst, m2->m_data, len); 1022 m1->m_len += len; 1023 dst += len; 1024 m2->m_data += len; 1025 m2->m_len -= len; 1026 m2 = m2->m_next; 1027 } 1028 if (m1->m_len < siz) 1029 return (EBADRPC); 1030 return (0); 1031 } 1032 1033 /* 1034 * Advance the position in the mbuf chain. 1035 */ 1036 int 1037 nfs_adv(mdp, dposp, offs, left) 1038 struct mbuf **mdp; 1039 caddr_t *dposp; 1040 int offs; 1041 int left; 1042 { 1043 struct mbuf *m; 1044 int s; 1045 1046 m = *mdp; 1047 s = left; 1048 while (s < offs) { 1049 offs -= s; 1050 m = m->m_next; 1051 if (m == NULL) 1052 return (EBADRPC); 1053 s = m->m_len; 1054 } 1055 *mdp = m; 1056 *dposp = mtod(m, caddr_t)+offs; 1057 return (0); 1058 } 1059 1060 /* 1061 * Copy a string into mbufs for the hard cases... 1062 */ 1063 int 1064 nfsm_strtmbuf(mb, bpos, cp, siz) 1065 struct mbuf **mb; 1066 char **bpos; 1067 const char *cp; 1068 long siz; 1069 { 1070 struct mbuf *m1 = NULL, *m2; 1071 long left, xfer, len, tlen; 1072 u_int32_t *tl; 1073 int putsize; 1074 1075 putsize = 1; 1076 m2 = *mb; 1077 left = M_TRAILINGSPACE(m2); 1078 if (left > 0) { 1079 tl = ((u_int32_t *)(*bpos)); 1080 *tl++ = txdr_unsigned(siz); 1081 putsize = 0; 1082 left -= NFSX_UNSIGNED; 1083 m2->m_len += NFSX_UNSIGNED; 1084 if (left > 0) { 1085 memcpy((caddr_t) tl, cp, left); 1086 siz -= left; 1087 cp += left; 1088 m2->m_len += left; 1089 left = 0; 1090 } 1091 } 1092 /* Loop around adding mbufs */ 1093 while (siz > 0) { 1094 MGET(m1, M_WAIT, MT_DATA); 1095 if (siz > MLEN) 1096 MCLGET(m1, M_WAIT); 1097 m1->m_len = NFSMSIZ(m1); 1098 m2->m_next = m1; 1099 m2 = m1; 1100 tl = mtod(m1, u_int32_t *); 1101 tlen = 0; 1102 if (putsize) { 1103 *tl++ = txdr_unsigned(siz); 1104 m1->m_len -= NFSX_UNSIGNED; 1105 tlen = NFSX_UNSIGNED; 1106 putsize = 0; 1107 } 1108 if (siz < m1->m_len) { 1109 len = nfsm_rndup(siz); 1110 xfer = siz; 1111 if (xfer < len) 1112 *(tl+(xfer>>2)) = 0; 1113 } else { 1114 xfer = len = m1->m_len; 1115 } 1116 memcpy((caddr_t) tl, cp, xfer); 1117 m1->m_len = len+tlen; 1118 siz -= xfer; 1119 cp += xfer; 1120 } 1121 *mb = m1; 1122 *bpos = mtod(m1, caddr_t)+m1->m_len; 1123 return (0); 1124 } 1125 1126 /* 1127 * Directory caching routines. They work as follows: 1128 * - a cache is maintained per VDIR nfsnode. 1129 * - for each offset cookie that is exported to userspace, and can 1130 * thus be thrown back at us as an offset to VOP_READDIR, store 1131 * information in the cache. 1132 * - cached are: 1133 * - cookie itself 1134 * - blocknumber (essentially just a search key in the buffer cache) 1135 * - entry number in block. 1136 * - offset cookie of block in which this entry is stored 1137 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1138 * - entries are looked up in a hash table 1139 * - also maintained is an LRU list of entries, used to determine 1140 * which ones to delete if the cache grows too large. 1141 * - if 32 <-> 64 translation mode is requested for a filesystem, 1142 * the cache also functions as a translation table 1143 * - in the translation case, invalidating the cache does not mean 1144 * flushing it, but just marking entries as invalid, except for 1145 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1146 * still be able to use the cache as a translation table. 1147 * - 32 bit cookies are uniquely created by combining the hash table 1148 * entry value, and one generation count per hash table entry, 1149 * incremented each time an entry is appended to the chain. 1150 * - the cache is invalidated each time a direcory is modified 1151 * - sanity checks are also done; if an entry in a block turns 1152 * out not to have a matching cookie, the cache is invalidated 1153 * and a new block starting from the wanted offset is fetched from 1154 * the server. 1155 * - directory entries as read from the server are extended to contain 1156 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1157 * the cache and exporting them to userspace through the cookie 1158 * argument to VOP_READDIR. 1159 */ 1160 1161 u_long 1162 nfs_dirhash(off) 1163 off_t off; 1164 { 1165 int i; 1166 char *cp = (char *)&off; 1167 u_long sum = 0L; 1168 1169 for (i = 0 ; i < sizeof (off); i++) 1170 sum += *cp++; 1171 1172 return sum; 1173 } 1174 1175 void 1176 nfs_initdircache(vp) 1177 struct vnode *vp; 1178 { 1179 struct nfsnode *np = VTONFS(vp); 1180 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1181 1182 np->n_dircachesize = 0; 1183 np->n_dblkno = 1; 1184 np->n_dircache = 1185 hashinit(NFS_DIRHASHSIZ, M_NFSDIROFF, M_WAITOK, &nfsdirhashmask); 1186 TAILQ_INIT(&np->n_dirchain); 1187 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1188 MALLOC(np->n_dirgens, unsigned *, 1189 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, 1190 M_WAITOK); 1191 memset((caddr_t)np->n_dirgens, 0, 1192 NFS_DIRHASHSIZ * sizeof (unsigned)); 1193 } 1194 } 1195 1196 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0}; 1197 1198 struct nfsdircache * 1199 nfs_searchdircache(vp, off, do32, hashent) 1200 struct vnode *vp; 1201 off_t off; 1202 int do32; 1203 int *hashent; 1204 { 1205 struct nfsdirhashhead *ndhp; 1206 struct nfsdircache *ndp = NULL; 1207 struct nfsnode *np = VTONFS(vp); 1208 unsigned ent; 1209 1210 /* 1211 * Zero is always a valid cookie. 1212 */ 1213 if (off == 0) 1214 return &dzero; 1215 1216 /* 1217 * We use a 32bit cookie as search key, directly reconstruct 1218 * the hashentry. Else use the hashfunction. 1219 */ 1220 if (do32) { 1221 ent = (u_int32_t)off >> 24; 1222 if (ent >= NFS_DIRHASHSIZ) 1223 return NULL; 1224 ndhp = &np->n_dircache[ent]; 1225 } else { 1226 ndhp = NFSDIRHASH(np, off); 1227 } 1228 1229 if (hashent) 1230 *hashent = (int)(ndhp - np->n_dircache); 1231 if (do32) { 1232 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) { 1233 if (ndp->dc_cookie32 == (u_int32_t)off) { 1234 /* 1235 * An invalidated entry will become the 1236 * start of a new block fetched from 1237 * the server. 1238 */ 1239 if (ndp->dc_blkno == -1) { 1240 ndp->dc_blkcookie = ndp->dc_cookie; 1241 ndp->dc_blkno = np->n_dblkno++; 1242 ndp->dc_entry = 0; 1243 } 1244 break; 1245 } 1246 } 1247 } else { 1248 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) 1249 if (ndp->dc_cookie == off) 1250 break; 1251 } 1252 return ndp; 1253 } 1254 1255 1256 struct nfsdircache * 1257 nfs_enterdircache(vp, off, blkoff, en, blkno) 1258 struct vnode *vp; 1259 off_t off, blkoff; 1260 daddr_t blkno; 1261 int en; 1262 { 1263 struct nfsnode *np = VTONFS(vp); 1264 struct nfsdirhashhead *ndhp; 1265 struct nfsdircache *ndp = NULL, *first; 1266 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1267 int hashent, gen, overwrite; 1268 1269 if (!np->n_dircache) 1270 /* 1271 * XXX would like to do this in nfs_nget but vtype 1272 * isn't known at that time. 1273 */ 1274 nfs_initdircache(vp); 1275 1276 /* 1277 * XXX refuse entries for offset 0. amd(8) erroneously sets 1278 * cookie 0 for the '.' entry, making this necessary. This 1279 * isn't so bad, as 0 is a special case anyway. 1280 */ 1281 if (off == 0) 1282 return &dzero; 1283 1284 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1285 1286 if (ndp && ndp->dc_blkno != -1) { 1287 /* 1288 * Overwriting an old entry. Check if it's the same. 1289 * If so, just return. If not, remove the old entry. 1290 */ 1291 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1292 return ndp; 1293 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1294 LIST_REMOVE(ndp, dc_hash); 1295 FREE(ndp, M_NFSDIROFF); 1296 ndp = 0; 1297 } 1298 1299 ndhp = &np->n_dircache[hashent]; 1300 1301 if (!ndp) { 1302 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF, 1303 M_WAITOK); 1304 overwrite = 0; 1305 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1306 /* 1307 * We're allocating a new entry, so bump the 1308 * generation number. 1309 */ 1310 gen = ++np->n_dirgens[hashent]; 1311 if (gen == 0) { 1312 np->n_dirgens[hashent]++; 1313 gen++; 1314 } 1315 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1316 } 1317 } else 1318 overwrite = 1; 1319 1320 /* 1321 * If the entry number is 0, we are at the start of a new block, so 1322 * allocate a new blocknumber. 1323 */ 1324 if (en == 0) 1325 ndp->dc_blkno = np->n_dblkno++; 1326 else 1327 ndp->dc_blkno = blkno; 1328 1329 ndp->dc_cookie = off; 1330 ndp->dc_blkcookie = blkoff; 1331 ndp->dc_entry = en; 1332 1333 if (overwrite) 1334 return ndp; 1335 1336 /* 1337 * If the maximum directory cookie cache size has been reached 1338 * for this node, take one off the front. The idea is that 1339 * directories are typically read front-to-back once, so that 1340 * the oldest entries can be thrown away without much performance 1341 * loss. 1342 */ 1343 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1344 first = np->n_dirchain.tqh_first; 1345 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain); 1346 LIST_REMOVE(first, dc_hash); 1347 FREE(first, M_NFSDIROFF); 1348 } else 1349 np->n_dircachesize++; 1350 1351 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1352 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1353 return ndp; 1354 } 1355 1356 void 1357 nfs_invaldircache(vp, forcefree) 1358 struct vnode *vp; 1359 int forcefree; 1360 { 1361 struct nfsnode *np = VTONFS(vp); 1362 struct nfsdircache *ndp = NULL; 1363 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1364 1365 #ifdef DIAGNOSTIC 1366 if (vp->v_type != VDIR) 1367 panic("nfs: invaldircache: not dir"); 1368 #endif 1369 1370 if (!np->n_dircache) 1371 return; 1372 1373 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1374 while ((ndp = np->n_dirchain.tqh_first)) { 1375 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1376 LIST_REMOVE(ndp, dc_hash); 1377 FREE(ndp, M_NFSDIROFF); 1378 } 1379 np->n_dircachesize = 0; 1380 if (forcefree && np->n_dirgens) { 1381 FREE(np->n_dirgens, M_NFSDIROFF); 1382 } 1383 } else { 1384 for (ndp = np->n_dirchain.tqh_first; ndp; 1385 ndp = ndp->dc_chain.tqe_next) 1386 ndp->dc_blkno = -1; 1387 } 1388 1389 np->n_dblkno = 1; 1390 } 1391 1392 /* 1393 * Called once before VFS init to initialize shared and 1394 * server-specific data structures. 1395 */ 1396 void 1397 nfs_init() 1398 { 1399 1400 #if !defined(alpha) && !defined(_LP64) && defined(DIAGNOSTIC) 1401 /* 1402 * Check to see if major data structures haven't bloated. 1403 */ 1404 if (sizeof (struct nfsnode) > NFS_NODEALLOC) { 1405 printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); 1406 printf("Try reducing NFS_SMALLFH\n"); 1407 } 1408 if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { 1409 printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); 1410 printf("Try reducing NFS_UIDHASHSIZ\n"); 1411 } 1412 if (sizeof (struct nfsuid) > NFS_UIDALLOC) { 1413 printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); 1414 printf("Try unionizing the nu_nickname and nu_flag fields\n"); 1415 } 1416 #endif 1417 1418 nfsrtt.pos = 0; 1419 rpc_vers = txdr_unsigned(RPC_VER2); 1420 rpc_call = txdr_unsigned(RPC_CALL); 1421 rpc_reply = txdr_unsigned(RPC_REPLY); 1422 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1423 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1424 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1425 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1426 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1427 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1428 nfs_prog = txdr_unsigned(NFS_PROG); 1429 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1430 nfs_true = txdr_unsigned(TRUE); 1431 nfs_false = txdr_unsigned(FALSE); 1432 nfs_xdrneg1 = txdr_unsigned(-1); 1433 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1434 if (nfs_ticks < 1) 1435 nfs_ticks = 1; 1436 #ifdef NFSSERVER 1437 nfsrv_init(0); /* Init server data structures */ 1438 nfsrv_initcache(); /* Init the server request cache */ 1439 #endif /* NFSSERVER */ 1440 1441 /* 1442 * Initialize the nqnfs data structures. 1443 */ 1444 if (nqnfsstarttime == 0) { 1445 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1446 + nqsrv_clockskew + nqsrv_writeslack; 1447 NQLOADNOVRAM(nqnfsstarttime); 1448 CIRCLEQ_INIT(&nqtimerhead); 1449 nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, M_WAITOK, &nqfhhash); 1450 } 1451 1452 /* 1453 * Initialize reply list and start timer 1454 */ 1455 TAILQ_INIT(&nfs_reqq); 1456 nfs_timer(NULL); 1457 } 1458 1459 #ifdef NFS 1460 /* 1461 * Called once at VFS init to initialize client-specific data structures. 1462 */ 1463 void 1464 nfs_vfs_init() 1465 { 1466 int i; 1467 1468 /* Ensure async daemons disabled */ 1469 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { 1470 nfs_iodwant[i] = (struct proc *)0; 1471 nfs_iodmount[i] = (struct nfsmount *)0; 1472 } 1473 nfs_nhinit(); /* Init the nfsnode table */ 1474 } 1475 1476 void 1477 nfs_vfs_done() 1478 { 1479 nfs_nhdone(); 1480 } 1481 1482 /* 1483 * Attribute cache routines. 1484 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1485 * that are on the mbuf list 1486 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1487 * error otherwise 1488 */ 1489 1490 /* 1491 * Load the attribute cache (that lives in the nfsnode entry) with 1492 * the values on the mbuf list and 1493 * Iff vap not NULL 1494 * copy the attributes to *vaper 1495 */ 1496 int 1497 nfsm_loadattrcache(vpp, mdp, dposp, vaper) 1498 struct vnode **vpp; 1499 struct mbuf **mdp; 1500 caddr_t *dposp; 1501 struct vattr *vaper; 1502 { 1503 int32_t t1; 1504 caddr_t cp2; 1505 int error = 0; 1506 struct mbuf *md; 1507 int v3 = NFS_ISV3(*vpp); 1508 1509 md = *mdp; 1510 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1511 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1512 if (error) 1513 return (error); 1514 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper); 1515 } 1516 1517 int 1518 nfs_loadattrcache(vpp, fp, vaper) 1519 struct vnode **vpp; 1520 struct nfs_fattr *fp; 1521 struct vattr *vaper; 1522 { 1523 struct vnode *vp = *vpp; 1524 struct vattr *vap; 1525 int v3 = NFS_ISV3(vp); 1526 enum vtype vtyp; 1527 u_short vmode; 1528 struct timespec mtime; 1529 struct vnode *nvp; 1530 int32_t rdev; 1531 struct nfsnode *np; 1532 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1533 1534 if (v3) { 1535 vtyp = nfsv3tov_type(fp->fa_type); 1536 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1537 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1538 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1539 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1540 } else { 1541 vtyp = nfsv2tov_type(fp->fa_type); 1542 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1543 if (vtyp == VNON || vtyp == VREG) 1544 vtyp = IFTOVT(vmode); 1545 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1546 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1547 1548 /* 1549 * Really ugly NFSv2 kludge. 1550 */ 1551 if (vtyp == VCHR && rdev == 0xffffffff) 1552 vtyp = VFIFO; 1553 } 1554 1555 /* 1556 * If v_type == VNON it is a new node, so fill in the v_type, 1557 * n_mtime fields. Check to see if it represents a special 1558 * device, and if so, check for a possible alias. Once the 1559 * correct vnode has been obtained, fill in the rest of the 1560 * information. 1561 */ 1562 np = VTONFS(vp); 1563 if (vp->v_type != vtyp) { 1564 vp->v_type = vtyp; 1565 if (vp->v_type == VFIFO) { 1566 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1567 vp->v_op = fifo_nfsv2nodeop_p; 1568 } 1569 if (vp->v_type == VCHR || vp->v_type == VBLK) { 1570 vp->v_op = spec_nfsv2nodeop_p; 1571 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1572 if (nvp) { 1573 /* 1574 * Discard unneeded vnode, but save its nfsnode. 1575 * Since the nfsnode does not have a lock, its 1576 * vnode lock has to be carried over. 1577 */ 1578 nvp->v_data = vp->v_data; 1579 vp->v_data = NULL; 1580 vp->v_op = spec_vnodeop_p; 1581 vput(vp); 1582 vgone(vp); 1583 /* 1584 * XXX When nfs starts locking, we need to 1585 * lock the new node here. 1586 */ 1587 /* 1588 * Reinitialize aliased node. 1589 */ 1590 np->n_vnode = nvp; 1591 *vpp = vp = nvp; 1592 } 1593 } 1594 np->n_mtime = mtime.tv_sec; 1595 } 1596 vap = np->n_vattr; 1597 vap->va_type = vtyp; 1598 vap->va_mode = vmode & ALLPERMS; 1599 vap->va_rdev = (dev_t)rdev; 1600 vap->va_mtime = mtime; 1601 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 1602 switch (vtyp) { 1603 case VDIR: 1604 vap->va_blocksize = NFS_DIRFRAGSIZ; 1605 break; 1606 case VBLK: 1607 vap->va_blocksize = BLKDEV_IOSIZE; 1608 break; 1609 case VCHR: 1610 vap->va_blocksize = MAXBSIZE; 1611 break; 1612 default: 1613 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1614 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1615 break; 1616 } 1617 if (v3) { 1618 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1619 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1620 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1621 vap->va_size = fxdr_hyper(&fp->fa3_size); 1622 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1623 vap->va_fileid = fxdr_unsigned(int32_t, 1624 fp->fa3_fileid.nfsuquad[1]); 1625 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1626 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); 1627 vap->va_flags = 0; 1628 vap->va_filerev = 0; 1629 } else { 1630 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1631 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1632 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1633 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1634 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1635 * NFS_FABLKSIZE; 1636 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1637 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1638 vap->va_flags = 0; 1639 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, 1640 fp->fa2_ctime.nfsv2_sec); 1641 vap->va_ctime.tv_nsec = 0; 1642 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1643 vap->va_filerev = 0; 1644 } 1645 if (vap->va_size != np->n_size) { 1646 if (vap->va_type == VREG) { 1647 if (np->n_flag & NMODIFIED) { 1648 if (vap->va_size < np->n_size) 1649 vap->va_size = np->n_size; 1650 else 1651 np->n_size = vap->va_size; 1652 } else 1653 np->n_size = vap->va_size; 1654 uvm_vnp_setsize(vp, np->n_size); 1655 } else 1656 np->n_size = vap->va_size; 1657 } 1658 np->n_attrstamp = time.tv_sec; 1659 if (vaper != NULL) { 1660 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1661 if (np->n_flag & NCHG) { 1662 if (np->n_flag & NACC) 1663 vaper->va_atime = np->n_atim; 1664 if (np->n_flag & NUPD) 1665 vaper->va_mtime = np->n_mtim; 1666 } 1667 } 1668 return (0); 1669 } 1670 1671 /* 1672 * Check the time stamp 1673 * If the cache is valid, copy contents to *vap and return 0 1674 * otherwise return an error 1675 */ 1676 int 1677 nfs_getattrcache(vp, vaper) 1678 struct vnode *vp; 1679 struct vattr *vaper; 1680 { 1681 struct nfsnode *np = VTONFS(vp); 1682 struct vattr *vap; 1683 1684 if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { 1685 nfsstats.attrcache_misses++; 1686 return (ENOENT); 1687 } 1688 nfsstats.attrcache_hits++; 1689 vap = np->n_vattr; 1690 if (vap->va_size != np->n_size) { 1691 if (vap->va_type == VREG) { 1692 if (np->n_flag & NMODIFIED) { 1693 if (vap->va_size < np->n_size) 1694 vap->va_size = np->n_size; 1695 else 1696 np->n_size = vap->va_size; 1697 } else 1698 np->n_size = vap->va_size; 1699 uvm_vnp_setsize(vp, np->n_size); 1700 } else 1701 np->n_size = vap->va_size; 1702 } 1703 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1704 if (np->n_flag & NCHG) { 1705 if (np->n_flag & NACC) 1706 vaper->va_atime = np->n_atim; 1707 if (np->n_flag & NUPD) 1708 vaper->va_mtime = np->n_mtim; 1709 } 1710 return (0); 1711 } 1712 1713 /* 1714 * Heuristic to see if the server XDR encodes directory cookies or not. 1715 * it is not supposed to, but a lot of servers may do this. Also, since 1716 * most/all servers will implement V2 as well, it is expected that they 1717 * may return just 32 bits worth of cookie information, so we need to 1718 * find out in which 32 bits this information is available. We do this 1719 * to avoid trouble with emulated binaries that can't handle 64 bit 1720 * directory offsets. 1721 */ 1722 1723 void 1724 nfs_cookieheuristic(vp, flagp, p, cred) 1725 struct vnode *vp; 1726 int *flagp; 1727 struct proc *p; 1728 struct ucred *cred; 1729 { 1730 struct uio auio; 1731 struct iovec aiov; 1732 caddr_t buf, cp; 1733 struct dirent *dp; 1734 off_t *cookies = NULL, *cop; 1735 int error, eof, nc, len; 1736 1737 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 1738 1739 aiov.iov_base = buf; 1740 aiov.iov_len = NFS_DIRFRAGSIZ; 1741 auio.uio_iov = &aiov; 1742 auio.uio_iovcnt = 1; 1743 auio.uio_rw = UIO_READ; 1744 auio.uio_segflg = UIO_SYSSPACE; 1745 auio.uio_procp = p; 1746 auio.uio_resid = NFS_DIRFRAGSIZ; 1747 auio.uio_offset = 0; 1748 1749 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 1750 1751 len = NFS_DIRFRAGSIZ - auio.uio_resid; 1752 if (error || len == 0) { 1753 FREE(buf, M_TEMP); 1754 if (cookies) 1755 FREE(cookies, M_TEMP); 1756 return; 1757 } 1758 1759 /* 1760 * Find the first valid entry and look at its offset cookie. 1761 */ 1762 1763 cp = buf; 1764 for (cop = cookies; len > 0; len -= dp->d_reclen) { 1765 dp = (struct dirent *)cp; 1766 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 1767 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 1768 *flagp |= NFSMNT_SWAPCOOKIE; 1769 nfs_invaldircache(vp, 0); 1770 nfs_vinvalbuf(vp, 0, cred, p, 1); 1771 } 1772 break; 1773 } 1774 cop++; 1775 cp += dp->d_reclen; 1776 } 1777 1778 FREE(buf, M_TEMP); 1779 FREE(cookies, M_TEMP); 1780 } 1781 #endif /* NFS */ 1782 1783 /* 1784 * Set up nameidata for a lookup() call and do it. 1785 * 1786 * If pubflag is set, this call is done for a lookup operation on the 1787 * public filehandle. In that case we allow crossing mountpoints and 1788 * absolute pathnames. However, the caller is expected to check that 1789 * the lookup result is within the public fs, and deny access if 1790 * it is not. 1791 */ 1792 int 1793 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) 1794 struct nameidata *ndp; 1795 fhandle_t *fhp; 1796 int len; 1797 struct nfssvc_sock *slp; 1798 struct mbuf *nam; 1799 struct mbuf **mdp; 1800 caddr_t *dposp; 1801 struct vnode **retdirp; 1802 struct proc *p; 1803 int kerbflag, pubflag; 1804 { 1805 int i, rem; 1806 struct mbuf *md; 1807 char *fromcp, *tocp, *cp; 1808 struct iovec aiov; 1809 struct uio auio; 1810 struct vnode *dp; 1811 int error, rdonly, linklen; 1812 struct componentname *cnp = &ndp->ni_cnd; 1813 1814 *retdirp = (struct vnode *)0; 1815 MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); 1816 /* 1817 * Copy the name from the mbuf list to ndp->ni_pnbuf 1818 * and set the various ndp fields appropriately. 1819 */ 1820 fromcp = *dposp; 1821 tocp = cnp->cn_pnbuf; 1822 md = *mdp; 1823 rem = mtod(md, caddr_t) + md->m_len - fromcp; 1824 for (i = 0; i < len; i++) { 1825 while (rem == 0) { 1826 md = md->m_next; 1827 if (md == NULL) { 1828 error = EBADRPC; 1829 goto out; 1830 } 1831 fromcp = mtod(md, caddr_t); 1832 rem = md->m_len; 1833 } 1834 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 1835 error = EACCES; 1836 goto out; 1837 } 1838 *tocp++ = *fromcp++; 1839 rem--; 1840 } 1841 *tocp = '\0'; 1842 *mdp = md; 1843 *dposp = fromcp; 1844 len = nfsm_rndup(len)-len; 1845 if (len > 0) { 1846 if (rem >= len) 1847 *dposp += len; 1848 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 1849 goto out; 1850 } 1851 1852 /* 1853 * Extract and set starting directory. 1854 */ 1855 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 1856 nam, &rdonly, kerbflag, pubflag); 1857 if (error) 1858 goto out; 1859 if (dp->v_type != VDIR) { 1860 vrele(dp); 1861 error = ENOTDIR; 1862 goto out; 1863 } 1864 1865 if (rdonly) 1866 cnp->cn_flags |= RDONLY; 1867 1868 *retdirp = dp; 1869 1870 if (pubflag) { 1871 /* 1872 * Oh joy. For WebNFS, handle those pesky '%' escapes, 1873 * and the 'native path' indicator. 1874 */ 1875 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); 1876 fromcp = cnp->cn_pnbuf; 1877 tocp = cp; 1878 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 1879 switch ((unsigned char)*fromcp) { 1880 case WEBNFS_NATIVE_CHAR: 1881 /* 1882 * 'Native' path for us is the same 1883 * as a path according to the NFS spec, 1884 * just skip the escape char. 1885 */ 1886 fromcp++; 1887 break; 1888 /* 1889 * More may be added in the future, range 0x80-0xff 1890 */ 1891 default: 1892 error = EIO; 1893 FREE(cp, M_NAMEI); 1894 goto out; 1895 } 1896 } 1897 /* 1898 * Translate the '%' escapes, URL-style. 1899 */ 1900 while (*fromcp != '\0') { 1901 if (*fromcp == WEBNFS_ESC_CHAR) { 1902 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 1903 fromcp++; 1904 *tocp++ = HEXSTRTOI(fromcp); 1905 fromcp += 2; 1906 continue; 1907 } else { 1908 error = ENOENT; 1909 FREE(cp, M_NAMEI); 1910 goto out; 1911 } 1912 } else 1913 *tocp++ = *fromcp++; 1914 } 1915 *tocp = '\0'; 1916 FREE(cnp->cn_pnbuf, M_NAMEI); 1917 cnp->cn_pnbuf = cp; 1918 } 1919 1920 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 1921 ndp->ni_segflg = UIO_SYSSPACE; 1922 1923 if (pubflag) { 1924 ndp->ni_rootdir = rootvnode; 1925 ndp->ni_loopcnt = 0; 1926 if (cnp->cn_pnbuf[0] == '/') 1927 dp = rootvnode; 1928 } else { 1929 cnp->cn_flags |= NOCROSSMOUNT; 1930 } 1931 1932 cnp->cn_proc = p; 1933 VREF(dp); 1934 1935 for (;;) { 1936 cnp->cn_nameptr = cnp->cn_pnbuf; 1937 ndp->ni_startdir = dp; 1938 /* 1939 * And call lookup() to do the real work 1940 */ 1941 error = lookup(ndp); 1942 if (error) 1943 break; 1944 /* 1945 * Check for encountering a symbolic link 1946 */ 1947 if ((cnp->cn_flags & ISSYMLINK) == 0) { 1948 if (cnp->cn_flags & (SAVENAME | SAVESTART)) { 1949 cnp->cn_flags |= HASBUF; 1950 return (0); 1951 } 1952 break; 1953 } else { 1954 if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) 1955 VOP_UNLOCK(ndp->ni_dvp, 0); 1956 if (!pubflag) { 1957 vrele(ndp->ni_dvp); 1958 vput(ndp->ni_vp); 1959 ndp->ni_vp = NULL; 1960 error = EINVAL; 1961 break; 1962 } 1963 1964 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 1965 error = ELOOP; 1966 break; 1967 } 1968 if (ndp->ni_pathlen > 1) 1969 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); 1970 else 1971 cp = cnp->cn_pnbuf; 1972 aiov.iov_base = cp; 1973 aiov.iov_len = MAXPATHLEN; 1974 auio.uio_iov = &aiov; 1975 auio.uio_iovcnt = 1; 1976 auio.uio_offset = 0; 1977 auio.uio_rw = UIO_READ; 1978 auio.uio_segflg = UIO_SYSSPACE; 1979 auio.uio_procp = (struct proc *)0; 1980 auio.uio_resid = MAXPATHLEN; 1981 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 1982 if (error) { 1983 badlink: 1984 if (ndp->ni_pathlen > 1) 1985 FREE(cp, M_NAMEI); 1986 break; 1987 } 1988 linklen = MAXPATHLEN - auio.uio_resid; 1989 if (linklen == 0) { 1990 error = ENOENT; 1991 goto badlink; 1992 } 1993 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 1994 error = ENAMETOOLONG; 1995 goto badlink; 1996 } 1997 if (ndp->ni_pathlen > 1) { 1998 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 1999 FREE(cnp->cn_pnbuf, M_NAMEI); 2000 cnp->cn_pnbuf = cp; 2001 } else 2002 cnp->cn_pnbuf[linklen] = '\0'; 2003 ndp->ni_pathlen += linklen; 2004 vput(ndp->ni_vp); 2005 dp = ndp->ni_dvp; 2006 /* 2007 * Check if root directory should replace current directory. 2008 */ 2009 if (cnp->cn_pnbuf[0] == '/') { 2010 vrele(dp); 2011 dp = ndp->ni_rootdir; 2012 VREF(dp); 2013 } 2014 } 2015 } 2016 out: 2017 FREE(cnp->cn_pnbuf, M_NAMEI); 2018 return (error); 2019 } 2020 2021 /* 2022 * A fiddled version of m_adj() that ensures null fill to a long 2023 * boundary and only trims off the back end 2024 */ 2025 void 2026 nfsm_adj(mp, len, nul) 2027 struct mbuf *mp; 2028 int len; 2029 int nul; 2030 { 2031 struct mbuf *m; 2032 int count, i; 2033 char *cp; 2034 2035 /* 2036 * Trim from tail. Scan the mbuf chain, 2037 * calculating its length and finding the last mbuf. 2038 * If the adjustment only affects this mbuf, then just 2039 * adjust and return. Otherwise, rescan and truncate 2040 * after the remaining size. 2041 */ 2042 count = 0; 2043 m = mp; 2044 for (;;) { 2045 count += m->m_len; 2046 if (m->m_next == (struct mbuf *)0) 2047 break; 2048 m = m->m_next; 2049 } 2050 if (m->m_len > len) { 2051 m->m_len -= len; 2052 if (nul > 0) { 2053 cp = mtod(m, caddr_t)+m->m_len-nul; 2054 for (i = 0; i < nul; i++) 2055 *cp++ = '\0'; 2056 } 2057 return; 2058 } 2059 count -= len; 2060 if (count < 0) 2061 count = 0; 2062 /* 2063 * Correct length for chain is "count". 2064 * Find the mbuf with last data, adjust its length, 2065 * and toss data from remaining mbufs on chain. 2066 */ 2067 for (m = mp; m; m = m->m_next) { 2068 if (m->m_len >= count) { 2069 m->m_len = count; 2070 if (nul > 0) { 2071 cp = mtod(m, caddr_t)+m->m_len-nul; 2072 for (i = 0; i < nul; i++) 2073 *cp++ = '\0'; 2074 } 2075 break; 2076 } 2077 count -= m->m_len; 2078 } 2079 for (m = m->m_next;m;m = m->m_next) 2080 m->m_len = 0; 2081 } 2082 2083 /* 2084 * Make these functions instead of macros, so that the kernel text size 2085 * doesn't get too big... 2086 */ 2087 void 2088 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2089 struct nfsrv_descript *nfsd; 2090 int before_ret; 2091 struct vattr *before_vap; 2092 int after_ret; 2093 struct vattr *after_vap; 2094 struct mbuf **mbp; 2095 char **bposp; 2096 { 2097 struct mbuf *mb = *mbp, *mb2; 2098 char *bpos = *bposp; 2099 u_int32_t *tl; 2100 2101 if (before_ret) { 2102 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2103 *tl = nfs_false; 2104 } else { 2105 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2106 *tl++ = nfs_true; 2107 txdr_hyper(before_vap->va_size, tl); 2108 tl += 2; 2109 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2110 tl += 2; 2111 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2112 } 2113 *bposp = bpos; 2114 *mbp = mb; 2115 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2116 } 2117 2118 void 2119 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2120 struct nfsrv_descript *nfsd; 2121 int after_ret; 2122 struct vattr *after_vap; 2123 struct mbuf **mbp; 2124 char **bposp; 2125 { 2126 struct mbuf *mb = *mbp, *mb2; 2127 char *bpos = *bposp; 2128 u_int32_t *tl; 2129 struct nfs_fattr *fp; 2130 2131 if (after_ret) { 2132 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2133 *tl = nfs_false; 2134 } else { 2135 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2136 *tl++ = nfs_true; 2137 fp = (struct nfs_fattr *)tl; 2138 nfsm_srvfattr(nfsd, after_vap, fp); 2139 } 2140 *mbp = mb; 2141 *bposp = bpos; 2142 } 2143 2144 void 2145 nfsm_srvfattr(nfsd, vap, fp) 2146 struct nfsrv_descript *nfsd; 2147 struct vattr *vap; 2148 struct nfs_fattr *fp; 2149 { 2150 2151 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2152 fp->fa_uid = txdr_unsigned(vap->va_uid); 2153 fp->fa_gid = txdr_unsigned(vap->va_gid); 2154 if (nfsd->nd_flag & ND_NFSV3) { 2155 fp->fa_type = vtonfsv3_type(vap->va_type); 2156 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2157 txdr_hyper(vap->va_size, &fp->fa3_size); 2158 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2159 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2160 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2161 fp->fa3_fsid.nfsuquad[0] = 0; 2162 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2163 fp->fa3_fileid.nfsuquad[0] = 0; 2164 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); 2165 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2166 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2167 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2168 } else { 2169 fp->fa_type = vtonfsv2_type(vap->va_type); 2170 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2171 fp->fa2_size = txdr_unsigned(vap->va_size); 2172 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2173 if (vap->va_type == VFIFO) 2174 fp->fa2_rdev = 0xffffffff; 2175 else 2176 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2177 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2178 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2179 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2180 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2181 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2182 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2183 } 2184 } 2185 2186 /* 2187 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2188 * - look up fsid in mount list (if not found ret error) 2189 * - get vp and export rights by calling VFS_FHTOVP() 2190 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2191 * - if not lockflag unlock it with VOP_UNLOCK() 2192 */ 2193 int 2194 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) 2195 fhandle_t *fhp; 2196 int lockflag; 2197 struct vnode **vpp; 2198 struct ucred *cred; 2199 struct nfssvc_sock *slp; 2200 struct mbuf *nam; 2201 int *rdonlyp; 2202 int kerbflag; 2203 { 2204 struct mount *mp; 2205 int i; 2206 struct ucred *credanon; 2207 int error, exflags; 2208 struct sockaddr_in *saddr; 2209 2210 *vpp = (struct vnode *)0; 2211 2212 if (nfs_ispublicfh(fhp)) { 2213 if (!pubflag || !nfs_pub.np_valid) 2214 return (ESTALE); 2215 fhp = &nfs_pub.np_handle; 2216 } 2217 2218 mp = vfs_getvfs(&fhp->fh_fsid); 2219 if (!mp) 2220 return (ESTALE); 2221 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); 2222 if (error) 2223 return (error); 2224 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2225 if (error) 2226 return (error); 2227 2228 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2229 saddr = mtod(nam, struct sockaddr_in *); 2230 if ((saddr->sin_family == AF_INET) && 2231 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2232 vput(*vpp); 2233 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2234 } 2235 #ifdef INET6 2236 if ((saddr->sin_family == AF_INET6) && 2237 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2238 vput(*vpp); 2239 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2240 } 2241 #endif 2242 } 2243 /* 2244 * Check/setup credentials. 2245 */ 2246 if (exflags & MNT_EXKERB) { 2247 if (!kerbflag) { 2248 vput(*vpp); 2249 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2250 } 2251 } else if (kerbflag) { 2252 vput(*vpp); 2253 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2254 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { 2255 cred->cr_uid = credanon->cr_uid; 2256 cred->cr_gid = credanon->cr_gid; 2257 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) 2258 cred->cr_groups[i] = credanon->cr_groups[i]; 2259 cred->cr_ngroups = i; 2260 } 2261 if (exflags & MNT_EXRDONLY) 2262 *rdonlyp = 1; 2263 else 2264 *rdonlyp = 0; 2265 if (!lockflag) 2266 VOP_UNLOCK(*vpp, 0); 2267 return (0); 2268 } 2269 2270 /* 2271 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2272 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has 2273 * transformed this to all zeroes in both cases, so check for it. 2274 */ 2275 int 2276 nfs_ispublicfh(fhp) 2277 fhandle_t *fhp; 2278 { 2279 char *cp = (char *)fhp; 2280 int i; 2281 2282 for (i = 0; i < NFSX_V3FH; i++) 2283 if (*cp++ != 0) 2284 return (FALSE); 2285 return (TRUE); 2286 } 2287 2288 /* 2289 * This function compares two net addresses by family and returns TRUE 2290 * if they are the same host. 2291 * If there is any doubt, return FALSE. 2292 * The AF_INET family is handled as a special case so that address mbufs 2293 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2294 */ 2295 int 2296 netaddr_match(family, haddr, nam) 2297 int family; 2298 union nethostaddr *haddr; 2299 struct mbuf *nam; 2300 { 2301 struct sockaddr_in *inetaddr; 2302 2303 switch (family) { 2304 case AF_INET: 2305 inetaddr = mtod(nam, struct sockaddr_in *); 2306 if (inetaddr->sin_family == AF_INET && 2307 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2308 return (1); 2309 break; 2310 #ifdef INET6 2311 case AF_INET6: 2312 { 2313 struct sockaddr_in6 *sin6_1, *sin6_2; 2314 2315 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2316 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2317 if (sin6_1->sin6_family == AF_INET6 && 2318 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2319 return 1; 2320 } 2321 #endif 2322 #ifdef ISO 2323 case AF_ISO: 2324 { 2325 struct sockaddr_iso *isoaddr1, *isoaddr2; 2326 2327 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2328 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2329 if (isoaddr1->siso_family == AF_ISO && 2330 isoaddr1->siso_nlen > 0 && 2331 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2332 SAME_ISOADDR(isoaddr1, isoaddr2)) 2333 return (1); 2334 break; 2335 } 2336 #endif /* ISO */ 2337 default: 2338 break; 2339 }; 2340 return (0); 2341 } 2342 2343 2344 /* 2345 * The write verifier has changed (probably due to a server reboot), so all 2346 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 2347 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 2348 * flag. Once done the new write verifier can be set for the mount point. 2349 */ 2350 void 2351 nfs_clearcommit(mp) 2352 struct mount *mp; 2353 { 2354 struct vnode *vp, *nvp; 2355 struct buf *bp, *nbp; 2356 int s; 2357 2358 s = splbio(); 2359 loop: 2360 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 2361 if (vp->v_mount != mp) /* Paranoia */ 2362 goto loop; 2363 nvp = vp->v_mntvnodes.le_next; 2364 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2365 nbp = bp->b_vnbufs.le_next; 2366 if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) 2367 == (B_DELWRI | B_NEEDCOMMIT)) 2368 bp->b_flags &= ~B_NEEDCOMMIT; 2369 } 2370 } 2371 splx(s); 2372 } 2373 2374 /* 2375 * Map errnos to NFS error numbers. For Version 3 also filter out error 2376 * numbers not specified for the associated procedure. 2377 */ 2378 int 2379 nfsrv_errmap(nd, err) 2380 struct nfsrv_descript *nd; 2381 int err; 2382 { 2383 short *defaulterrp, *errp; 2384 2385 if (nd->nd_flag & ND_NFSV3) { 2386 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2387 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2388 while (*++errp) { 2389 if (*errp == err) 2390 return (err); 2391 else if (*errp > err) 2392 break; 2393 } 2394 return ((int)*defaulterrp); 2395 } else 2396 return (err & 0xffff); 2397 } 2398 if (err <= ELAST) 2399 return ((int)nfsrv_v2errmap[err - 1]); 2400 return (NFSERR_IO); 2401 } 2402 2403 /* 2404 * Sort the group list in increasing numerical order. 2405 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort 2406 * that used to be here.) 2407 */ 2408 void 2409 nfsrvw_sort(list, num) 2410 gid_t *list; 2411 int num; 2412 { 2413 int i, j; 2414 gid_t v; 2415 2416 /* Insertion sort. */ 2417 for (i = 1; i < num; i++) { 2418 v = list[i]; 2419 /* find correct slot for value v, moving others up */ 2420 for (j = i; --j >= 0 && v < list[j];) 2421 list[j + 1] = list[j]; 2422 list[j + 1] = v; 2423 } 2424 } 2425 2426 /* 2427 * copy credentials making sure that the result can be compared with memcmp(). 2428 */ 2429 void 2430 nfsrv_setcred(incred, outcred) 2431 struct ucred *incred, *outcred; 2432 { 2433 int i; 2434 2435 memset((caddr_t)outcred, 0, sizeof (struct ucred)); 2436 outcred->cr_ref = 1; 2437 outcred->cr_uid = incred->cr_uid; 2438 outcred->cr_gid = incred->cr_gid; 2439 outcred->cr_ngroups = incred->cr_ngroups; 2440 for (i = 0; i < incred->cr_ngroups; i++) 2441 outcred->cr_groups[i] = incred->cr_groups[i]; 2442 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); 2443 } 2444
Indexes created Fri Oct 03 02:09:57 GMT 2025