nfs_subs.c revision 1.80
1 /* $NetBSD: nfs_subs.c,v 1.80 2000/08/03 06:15:05 thorpej 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 <netinet/in.h> 81 #ifdef ISO 82 #include <netiso/iso.h> 83 #endif 84 85 /* 86 * Data items converted to xdr at startup, since they are constant 87 * This is kinda hokey, but may save a little time doing byte swaps 88 */ 89 u_int32_t nfs_xdrneg1; 90 u_int32_t rpc_call, rpc_vers, rpc_reply, rpc_msgdenied, rpc_autherr, 91 rpc_mismatch, rpc_auth_unix, rpc_msgaccepted, 92 rpc_auth_kerb; 93 u_int32_t nfs_prog, nqnfs_prog, nfs_true, nfs_false; 94 95 /* And other global data */ 96 static u_int32_t nfs_xid = 0; 97 nfstype nfsv2_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFNON, 98 NFCHR, NFNON }; 99 nfstype nfsv3_type[9] = { NFNON, NFREG, NFDIR, NFBLK, NFCHR, NFLNK, NFSOCK, 100 NFFIFO, NFNON }; 101 enum vtype nv2tov_type[8] = { VNON, VREG, VDIR, VBLK, VCHR, VLNK, VNON, VNON }; 102 enum vtype nv3tov_type[8]={ VNON, VREG, VDIR, VBLK, VCHR, VLNK, VSOCK, VFIFO }; 103 int nfs_ticks; 104 105 /* NFS client/server stats. */ 106 struct nfsstats nfsstats; 107 108 /* 109 * Mapping of old NFS Version 2 RPC numbers to generic numbers. 110 */ 111 int nfsv3_procid[NFS_NPROCS] = { 112 NFSPROC_NULL, 113 NFSPROC_GETATTR, 114 NFSPROC_SETATTR, 115 NFSPROC_NOOP, 116 NFSPROC_LOOKUP, 117 NFSPROC_READLINK, 118 NFSPROC_READ, 119 NFSPROC_NOOP, 120 NFSPROC_WRITE, 121 NFSPROC_CREATE, 122 NFSPROC_REMOVE, 123 NFSPROC_RENAME, 124 NFSPROC_LINK, 125 NFSPROC_SYMLINK, 126 NFSPROC_MKDIR, 127 NFSPROC_RMDIR, 128 NFSPROC_READDIR, 129 NFSPROC_FSSTAT, 130 NFSPROC_NOOP, 131 NFSPROC_NOOP, 132 NFSPROC_NOOP, 133 NFSPROC_NOOP, 134 NFSPROC_NOOP, 135 NFSPROC_NOOP, 136 NFSPROC_NOOP, 137 NFSPROC_NOOP 138 }; 139 140 /* 141 * and the reverse mapping from generic to Version 2 procedure numbers 142 */ 143 int nfsv2_procid[NFS_NPROCS] = { 144 NFSV2PROC_NULL, 145 NFSV2PROC_GETATTR, 146 NFSV2PROC_SETATTR, 147 NFSV2PROC_LOOKUP, 148 NFSV2PROC_NOOP, 149 NFSV2PROC_READLINK, 150 NFSV2PROC_READ, 151 NFSV2PROC_WRITE, 152 NFSV2PROC_CREATE, 153 NFSV2PROC_MKDIR, 154 NFSV2PROC_SYMLINK, 155 NFSV2PROC_CREATE, 156 NFSV2PROC_REMOVE, 157 NFSV2PROC_RMDIR, 158 NFSV2PROC_RENAME, 159 NFSV2PROC_LINK, 160 NFSV2PROC_READDIR, 161 NFSV2PROC_NOOP, 162 NFSV2PROC_STATFS, 163 NFSV2PROC_NOOP, 164 NFSV2PROC_NOOP, 165 NFSV2PROC_NOOP, 166 NFSV2PROC_NOOP, 167 NFSV2PROC_NOOP, 168 NFSV2PROC_NOOP, 169 NFSV2PROC_NOOP, 170 }; 171 172 /* 173 * Maps errno values to nfs error numbers. 174 * Use NFSERR_IO as the catch all for ones not specifically defined in 175 * RFC 1094. 176 */ 177 static u_char nfsrv_v2errmap[ELAST] = { 178 NFSERR_PERM, NFSERR_NOENT, NFSERR_IO, NFSERR_IO, NFSERR_IO, 179 NFSERR_NXIO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 180 NFSERR_IO, NFSERR_IO, NFSERR_ACCES, NFSERR_IO, NFSERR_IO, 181 NFSERR_IO, NFSERR_EXIST, NFSERR_IO, NFSERR_NODEV, NFSERR_NOTDIR, 182 NFSERR_ISDIR, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 183 NFSERR_IO, NFSERR_FBIG, NFSERR_NOSPC, NFSERR_IO, NFSERR_ROFS, 184 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 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_NAMETOL, NFSERR_IO, NFSERR_IO, 191 NFSERR_NOTEMPTY, NFSERR_IO, NFSERR_IO, NFSERR_DQUOT, NFSERR_STALE, 192 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 193 NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, NFSERR_IO, 194 NFSERR_IO, NFSERR_IO, 195 }; 196 197 /* 198 * Maps errno values to nfs error numbers. 199 * Although it is not obvious whether or not NFS clients really care if 200 * a returned error value is in the specified list for the procedure, the 201 * safest thing to do is filter them appropriately. For Version 2, the 202 * X/Open XNFS document is the only specification that defines error values 203 * for each RPC (The RFC simply lists all possible error values for all RPCs), 204 * so I have decided to not do this for Version 2. 205 * The first entry is the default error return and the rest are the valid 206 * errors for that RPC in increasing numeric order. 207 */ 208 static short nfsv3err_null[] = { 209 0, 210 0, 211 }; 212 213 static short nfsv3err_getattr[] = { 214 NFSERR_IO, 215 NFSERR_IO, 216 NFSERR_STALE, 217 NFSERR_BADHANDLE, 218 NFSERR_SERVERFAULT, 219 0, 220 }; 221 222 static short nfsv3err_setattr[] = { 223 NFSERR_IO, 224 NFSERR_PERM, 225 NFSERR_IO, 226 NFSERR_ACCES, 227 NFSERR_INVAL, 228 NFSERR_NOSPC, 229 NFSERR_ROFS, 230 NFSERR_DQUOT, 231 NFSERR_STALE, 232 NFSERR_BADHANDLE, 233 NFSERR_NOT_SYNC, 234 NFSERR_SERVERFAULT, 235 0, 236 }; 237 238 static short nfsv3err_lookup[] = { 239 NFSERR_IO, 240 NFSERR_NOENT, 241 NFSERR_IO, 242 NFSERR_ACCES, 243 NFSERR_NOTDIR, 244 NFSERR_NAMETOL, 245 NFSERR_STALE, 246 NFSERR_BADHANDLE, 247 NFSERR_SERVERFAULT, 248 0, 249 }; 250 251 static short nfsv3err_access[] = { 252 NFSERR_IO, 253 NFSERR_IO, 254 NFSERR_STALE, 255 NFSERR_BADHANDLE, 256 NFSERR_SERVERFAULT, 257 0, 258 }; 259 260 static short nfsv3err_readlink[] = { 261 NFSERR_IO, 262 NFSERR_IO, 263 NFSERR_ACCES, 264 NFSERR_INVAL, 265 NFSERR_STALE, 266 NFSERR_BADHANDLE, 267 NFSERR_NOTSUPP, 268 NFSERR_SERVERFAULT, 269 0, 270 }; 271 272 static short nfsv3err_read[] = { 273 NFSERR_IO, 274 NFSERR_IO, 275 NFSERR_NXIO, 276 NFSERR_ACCES, 277 NFSERR_INVAL, 278 NFSERR_STALE, 279 NFSERR_BADHANDLE, 280 NFSERR_SERVERFAULT, 281 NFSERR_JUKEBOX, 282 0, 283 }; 284 285 static short nfsv3err_write[] = { 286 NFSERR_IO, 287 NFSERR_IO, 288 NFSERR_ACCES, 289 NFSERR_INVAL, 290 NFSERR_FBIG, 291 NFSERR_NOSPC, 292 NFSERR_ROFS, 293 NFSERR_DQUOT, 294 NFSERR_STALE, 295 NFSERR_BADHANDLE, 296 NFSERR_SERVERFAULT, 297 NFSERR_JUKEBOX, 298 0, 299 }; 300 301 static short nfsv3err_create[] = { 302 NFSERR_IO, 303 NFSERR_IO, 304 NFSERR_ACCES, 305 NFSERR_EXIST, 306 NFSERR_NOTDIR, 307 NFSERR_NOSPC, 308 NFSERR_ROFS, 309 NFSERR_NAMETOL, 310 NFSERR_DQUOT, 311 NFSERR_STALE, 312 NFSERR_BADHANDLE, 313 NFSERR_NOTSUPP, 314 NFSERR_SERVERFAULT, 315 0, 316 }; 317 318 static short nfsv3err_mkdir[] = { 319 NFSERR_IO, 320 NFSERR_IO, 321 NFSERR_ACCES, 322 NFSERR_EXIST, 323 NFSERR_NOTDIR, 324 NFSERR_NOSPC, 325 NFSERR_ROFS, 326 NFSERR_NAMETOL, 327 NFSERR_DQUOT, 328 NFSERR_STALE, 329 NFSERR_BADHANDLE, 330 NFSERR_NOTSUPP, 331 NFSERR_SERVERFAULT, 332 0, 333 }; 334 335 static short nfsv3err_symlink[] = { 336 NFSERR_IO, 337 NFSERR_IO, 338 NFSERR_ACCES, 339 NFSERR_EXIST, 340 NFSERR_NOTDIR, 341 NFSERR_NOSPC, 342 NFSERR_ROFS, 343 NFSERR_NAMETOL, 344 NFSERR_DQUOT, 345 NFSERR_STALE, 346 NFSERR_BADHANDLE, 347 NFSERR_NOTSUPP, 348 NFSERR_SERVERFAULT, 349 0, 350 }; 351 352 static short nfsv3err_mknod[] = { 353 NFSERR_IO, 354 NFSERR_IO, 355 NFSERR_ACCES, 356 NFSERR_EXIST, 357 NFSERR_NOTDIR, 358 NFSERR_NOSPC, 359 NFSERR_ROFS, 360 NFSERR_NAMETOL, 361 NFSERR_DQUOT, 362 NFSERR_STALE, 363 NFSERR_BADHANDLE, 364 NFSERR_NOTSUPP, 365 NFSERR_SERVERFAULT, 366 NFSERR_BADTYPE, 367 0, 368 }; 369 370 static short nfsv3err_remove[] = { 371 NFSERR_IO, 372 NFSERR_NOENT, 373 NFSERR_IO, 374 NFSERR_ACCES, 375 NFSERR_NOTDIR, 376 NFSERR_ROFS, 377 NFSERR_NAMETOL, 378 NFSERR_STALE, 379 NFSERR_BADHANDLE, 380 NFSERR_SERVERFAULT, 381 0, 382 }; 383 384 static short nfsv3err_rmdir[] = { 385 NFSERR_IO, 386 NFSERR_NOENT, 387 NFSERR_IO, 388 NFSERR_ACCES, 389 NFSERR_EXIST, 390 NFSERR_NOTDIR, 391 NFSERR_INVAL, 392 NFSERR_ROFS, 393 NFSERR_NAMETOL, 394 NFSERR_NOTEMPTY, 395 NFSERR_STALE, 396 NFSERR_BADHANDLE, 397 NFSERR_NOTSUPP, 398 NFSERR_SERVERFAULT, 399 0, 400 }; 401 402 static short nfsv3err_rename[] = { 403 NFSERR_IO, 404 NFSERR_NOENT, 405 NFSERR_IO, 406 NFSERR_ACCES, 407 NFSERR_EXIST, 408 NFSERR_XDEV, 409 NFSERR_NOTDIR, 410 NFSERR_ISDIR, 411 NFSERR_INVAL, 412 NFSERR_NOSPC, 413 NFSERR_ROFS, 414 NFSERR_MLINK, 415 NFSERR_NAMETOL, 416 NFSERR_NOTEMPTY, 417 NFSERR_DQUOT, 418 NFSERR_STALE, 419 NFSERR_BADHANDLE, 420 NFSERR_NOTSUPP, 421 NFSERR_SERVERFAULT, 422 0, 423 }; 424 425 static short nfsv3err_link[] = { 426 NFSERR_IO, 427 NFSERR_IO, 428 NFSERR_ACCES, 429 NFSERR_EXIST, 430 NFSERR_XDEV, 431 NFSERR_NOTDIR, 432 NFSERR_INVAL, 433 NFSERR_NOSPC, 434 NFSERR_ROFS, 435 NFSERR_MLINK, 436 NFSERR_NAMETOL, 437 NFSERR_DQUOT, 438 NFSERR_STALE, 439 NFSERR_BADHANDLE, 440 NFSERR_NOTSUPP, 441 NFSERR_SERVERFAULT, 442 0, 443 }; 444 445 static short nfsv3err_readdir[] = { 446 NFSERR_IO, 447 NFSERR_IO, 448 NFSERR_ACCES, 449 NFSERR_NOTDIR, 450 NFSERR_STALE, 451 NFSERR_BADHANDLE, 452 NFSERR_BAD_COOKIE, 453 NFSERR_TOOSMALL, 454 NFSERR_SERVERFAULT, 455 0, 456 }; 457 458 static short nfsv3err_readdirplus[] = { 459 NFSERR_IO, 460 NFSERR_IO, 461 NFSERR_ACCES, 462 NFSERR_NOTDIR, 463 NFSERR_STALE, 464 NFSERR_BADHANDLE, 465 NFSERR_BAD_COOKIE, 466 NFSERR_NOTSUPP, 467 NFSERR_TOOSMALL, 468 NFSERR_SERVERFAULT, 469 0, 470 }; 471 472 static short nfsv3err_fsstat[] = { 473 NFSERR_IO, 474 NFSERR_IO, 475 NFSERR_STALE, 476 NFSERR_BADHANDLE, 477 NFSERR_SERVERFAULT, 478 0, 479 }; 480 481 static short nfsv3err_fsinfo[] = { 482 NFSERR_STALE, 483 NFSERR_STALE, 484 NFSERR_BADHANDLE, 485 NFSERR_SERVERFAULT, 486 0, 487 }; 488 489 static short nfsv3err_pathconf[] = { 490 NFSERR_STALE, 491 NFSERR_STALE, 492 NFSERR_BADHANDLE, 493 NFSERR_SERVERFAULT, 494 0, 495 }; 496 497 static short nfsv3err_commit[] = { 498 NFSERR_IO, 499 NFSERR_IO, 500 NFSERR_STALE, 501 NFSERR_BADHANDLE, 502 NFSERR_SERVERFAULT, 503 0, 504 }; 505 506 static short *nfsrv_v3errmap[] = { 507 nfsv3err_null, 508 nfsv3err_getattr, 509 nfsv3err_setattr, 510 nfsv3err_lookup, 511 nfsv3err_access, 512 nfsv3err_readlink, 513 nfsv3err_read, 514 nfsv3err_write, 515 nfsv3err_create, 516 nfsv3err_mkdir, 517 nfsv3err_symlink, 518 nfsv3err_mknod, 519 nfsv3err_remove, 520 nfsv3err_rmdir, 521 nfsv3err_rename, 522 nfsv3err_link, 523 nfsv3err_readdir, 524 nfsv3err_readdirplus, 525 nfsv3err_fsstat, 526 nfsv3err_fsinfo, 527 nfsv3err_pathconf, 528 nfsv3err_commit, 529 }; 530 531 extern struct nfsrtt nfsrtt; 532 extern time_t nqnfsstarttime; 533 extern int nqsrv_clockskew; 534 extern int nqsrv_writeslack; 535 extern int nqsrv_maxlease; 536 extern int nqnfs_piggy[NFS_NPROCS]; 537 extern struct nfsnodehashhead *nfsnodehashtbl; 538 extern u_long nfsnodehash; 539 540 LIST_HEAD(nfsnodehashhead, nfsnode); 541 u_long nfsdirhashmask; 542 543 int nfs_webnamei __P((struct nameidata *, struct vnode *, struct proc *)); 544 545 /* 546 * Create the header for an rpc request packet 547 * The hsiz is the size of the rest of the nfs request header. 548 * (just used to decide if a cluster is a good idea) 549 */ 550 struct mbuf * 551 nfsm_reqh(vp, procid, hsiz, bposp) 552 struct vnode *vp; 553 u_long procid; 554 int hsiz; 555 caddr_t *bposp; 556 { 557 struct mbuf *mb; 558 u_int32_t *tl; 559 caddr_t bpos; 560 struct mbuf *mb2; 561 struct nfsmount *nmp; 562 int nqflag; 563 564 MGET(mb, M_WAIT, MT_DATA); 565 if (hsiz >= MINCLSIZE) 566 MCLGET(mb, M_WAIT); 567 mb->m_len = 0; 568 bpos = mtod(mb, caddr_t); 569 570 /* 571 * For NQNFS, add lease request. 572 */ 573 if (vp) { 574 nmp = VFSTONFS(vp->v_mount); 575 if (nmp->nm_flag & NFSMNT_NQNFS) { 576 nqflag = NQNFS_NEEDLEASE(vp, procid); 577 if (nqflag) { 578 nfsm_build(tl, u_int32_t *, 2*NFSX_UNSIGNED); 579 *tl++ = txdr_unsigned(nqflag); 580 *tl = txdr_unsigned(nmp->nm_leaseterm); 581 } else { 582 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 583 *tl = 0; 584 } 585 } 586 } 587 /* Finally, return values */ 588 *bposp = bpos; 589 return (mb); 590 } 591 592 /* 593 * Build the RPC header and fill in the authorization info. 594 * The authorization string argument is only used when the credentials 595 * come from outside of the kernel. 596 * Returns the head of the mbuf list. 597 */ 598 struct mbuf * 599 nfsm_rpchead(cr, nmflag, procid, auth_type, auth_len, auth_str, verf_len, 600 verf_str, mrest, mrest_len, mbp, xidp) 601 struct ucred *cr; 602 int nmflag; 603 int procid; 604 int auth_type; 605 int auth_len; 606 char *auth_str; 607 int verf_len; 608 char *verf_str; 609 struct mbuf *mrest; 610 int mrest_len; 611 struct mbuf **mbp; 612 u_int32_t *xidp; 613 { 614 struct mbuf *mb; 615 u_int32_t *tl; 616 caddr_t bpos; 617 int i; 618 struct mbuf *mreq, *mb2; 619 int siz, grpsiz, authsiz; 620 struct timeval tv; 621 static u_int32_t base; 622 623 authsiz = nfsm_rndup(auth_len); 624 MGETHDR(mb, M_WAIT, MT_DATA); 625 if ((authsiz + 10 * NFSX_UNSIGNED) >= MINCLSIZE) { 626 MCLGET(mb, M_WAIT); 627 } else if ((authsiz + 10 * NFSX_UNSIGNED) < MHLEN) { 628 MH_ALIGN(mb, authsiz + 10 * NFSX_UNSIGNED); 629 } else { 630 MH_ALIGN(mb, 8 * NFSX_UNSIGNED); 631 } 632 mb->m_len = 0; 633 mreq = mb; 634 bpos = mtod(mb, caddr_t); 635 636 /* 637 * First the RPC header. 638 */ 639 nfsm_build(tl, u_int32_t *, 8 * NFSX_UNSIGNED); 640 641 /* 642 * derive initial xid from system time 643 * XXX time is invalid if root not yet mounted 644 */ 645 if (!base && (rootvp)) { 646 microtime(&tv); 647 base = tv.tv_sec << 12; 648 nfs_xid = base; 649 } 650 /* 651 * Skip zero xid if it should ever happen. 652 */ 653 if (++nfs_xid == 0) 654 nfs_xid++; 655 656 *tl++ = *xidp = txdr_unsigned(nfs_xid); 657 *tl++ = rpc_call; 658 *tl++ = rpc_vers; 659 if (nmflag & NFSMNT_NQNFS) { 660 *tl++ = txdr_unsigned(NQNFS_PROG); 661 *tl++ = txdr_unsigned(NQNFS_VER3); 662 } else { 663 *tl++ = txdr_unsigned(NFS_PROG); 664 if (nmflag & NFSMNT_NFSV3) 665 *tl++ = txdr_unsigned(NFS_VER3); 666 else 667 *tl++ = txdr_unsigned(NFS_VER2); 668 } 669 if (nmflag & NFSMNT_NFSV3) 670 *tl++ = txdr_unsigned(procid); 671 else 672 *tl++ = txdr_unsigned(nfsv2_procid[procid]); 673 674 /* 675 * And then the authorization cred. 676 */ 677 *tl++ = txdr_unsigned(auth_type); 678 *tl = txdr_unsigned(authsiz); 679 switch (auth_type) { 680 case RPCAUTH_UNIX: 681 nfsm_build(tl, u_int32_t *, auth_len); 682 *tl++ = 0; /* stamp ?? */ 683 *tl++ = 0; /* NULL hostname */ 684 *tl++ = txdr_unsigned(cr->cr_uid); 685 *tl++ = txdr_unsigned(cr->cr_gid); 686 grpsiz = (auth_len >> 2) - 5; 687 *tl++ = txdr_unsigned(grpsiz); 688 for (i = 0; i < grpsiz; i++) 689 *tl++ = txdr_unsigned(cr->cr_groups[i]); 690 break; 691 case RPCAUTH_KERB4: 692 siz = auth_len; 693 while (siz > 0) { 694 if (M_TRAILINGSPACE(mb) == 0) { 695 MGET(mb2, M_WAIT, MT_DATA); 696 if (siz >= MINCLSIZE) 697 MCLGET(mb2, M_WAIT); 698 mb->m_next = mb2; 699 mb = mb2; 700 mb->m_len = 0; 701 bpos = mtod(mb, caddr_t); 702 } 703 i = min(siz, M_TRAILINGSPACE(mb)); 704 memcpy(bpos, auth_str, i); 705 mb->m_len += i; 706 auth_str += i; 707 bpos += i; 708 siz -= i; 709 } 710 if ((siz = (nfsm_rndup(auth_len) - auth_len)) > 0) { 711 for (i = 0; i < siz; i++) 712 *bpos++ = '\0'; 713 mb->m_len += siz; 714 } 715 break; 716 }; 717 718 /* 719 * And the verifier... 720 */ 721 nfsm_build(tl, u_int32_t *, 2 * NFSX_UNSIGNED); 722 if (verf_str) { 723 *tl++ = txdr_unsigned(RPCAUTH_KERB4); 724 *tl = txdr_unsigned(verf_len); 725 siz = verf_len; 726 while (siz > 0) { 727 if (M_TRAILINGSPACE(mb) == 0) { 728 MGET(mb2, M_WAIT, MT_DATA); 729 if (siz >= MINCLSIZE) 730 MCLGET(mb2, M_WAIT); 731 mb->m_next = mb2; 732 mb = mb2; 733 mb->m_len = 0; 734 bpos = mtod(mb, caddr_t); 735 } 736 i = min(siz, M_TRAILINGSPACE(mb)); 737 memcpy(bpos, verf_str, i); 738 mb->m_len += i; 739 verf_str += i; 740 bpos += i; 741 siz -= i; 742 } 743 if ((siz = (nfsm_rndup(verf_len) - verf_len)) > 0) { 744 for (i = 0; i < siz; i++) 745 *bpos++ = '\0'; 746 mb->m_len += siz; 747 } 748 } else { 749 *tl++ = txdr_unsigned(RPCAUTH_NULL); 750 *tl = 0; 751 } 752 mb->m_next = mrest; 753 mreq->m_pkthdr.len = authsiz + 10 * NFSX_UNSIGNED + mrest_len; 754 mreq->m_pkthdr.rcvif = (struct ifnet *)0; 755 *mbp = mb; 756 return (mreq); 757 } 758 759 /* 760 * copies mbuf chain to the uio scatter/gather list 761 */ 762 int 763 nfsm_mbuftouio(mrep, uiop, siz, dpos) 764 struct mbuf **mrep; 765 struct uio *uiop; 766 int siz; 767 caddr_t *dpos; 768 { 769 char *mbufcp, *uiocp; 770 int xfer, left, len; 771 struct mbuf *mp; 772 long uiosiz, rem; 773 int error = 0; 774 775 mp = *mrep; 776 mbufcp = *dpos; 777 len = mtod(mp, caddr_t)+mp->m_len-mbufcp; 778 rem = nfsm_rndup(siz)-siz; 779 while (siz > 0) { 780 if (uiop->uio_iovcnt <= 0 || uiop->uio_iov == NULL) 781 return (EFBIG); 782 left = uiop->uio_iov->iov_len; 783 uiocp = uiop->uio_iov->iov_base; 784 if (left > siz) 785 left = siz; 786 uiosiz = left; 787 while (left > 0) { 788 while (len == 0) { 789 mp = mp->m_next; 790 if (mp == NULL) 791 return (EBADRPC); 792 mbufcp = mtod(mp, caddr_t); 793 len = mp->m_len; 794 } 795 xfer = (left > len) ? len : left; 796 #ifdef notdef 797 /* Not Yet.. */ 798 if (uiop->uio_iov->iov_op != NULL) 799 (*(uiop->uio_iov->iov_op)) 800 (mbufcp, uiocp, xfer); 801 else 802 #endif 803 if (uiop->uio_segflg == UIO_SYSSPACE) 804 memcpy(uiocp, mbufcp, xfer); 805 else 806 copyout(mbufcp, uiocp, xfer); 807 left -= xfer; 808 len -= xfer; 809 mbufcp += xfer; 810 uiocp += xfer; 811 uiop->uio_offset += xfer; 812 uiop->uio_resid -= xfer; 813 } 814 if (uiop->uio_iov->iov_len <= siz) { 815 uiop->uio_iovcnt--; 816 uiop->uio_iov++; 817 } else { 818 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 819 uiop->uio_iov->iov_len -= uiosiz; 820 } 821 siz -= uiosiz; 822 } 823 *dpos = mbufcp; 824 *mrep = mp; 825 if (rem > 0) { 826 if (len < rem) 827 error = nfs_adv(mrep, dpos, rem, len); 828 else 829 *dpos += rem; 830 } 831 return (error); 832 } 833 834 /* 835 * copies a uio scatter/gather list to an mbuf chain. 836 * NOTE: can ony handle iovcnt == 1 837 */ 838 int 839 nfsm_uiotombuf(uiop, mq, siz, bpos) 840 struct uio *uiop; 841 struct mbuf **mq; 842 int siz; 843 caddr_t *bpos; 844 { 845 char *uiocp; 846 struct mbuf *mp, *mp2; 847 int xfer, left, mlen; 848 int uiosiz, clflg, rem; 849 char *cp; 850 851 #ifdef DIAGNOSTIC 852 if (uiop->uio_iovcnt != 1) 853 panic("nfsm_uiotombuf: iovcnt != 1"); 854 #endif 855 856 if (siz > MLEN) /* or should it >= MCLBYTES ?? */ 857 clflg = 1; 858 else 859 clflg = 0; 860 rem = nfsm_rndup(siz)-siz; 861 mp = mp2 = *mq; 862 while (siz > 0) { 863 left = uiop->uio_iov->iov_len; 864 uiocp = uiop->uio_iov->iov_base; 865 if (left > siz) 866 left = siz; 867 uiosiz = left; 868 while (left > 0) { 869 mlen = M_TRAILINGSPACE(mp); 870 if (mlen == 0) { 871 MGET(mp, M_WAIT, MT_DATA); 872 if (clflg) 873 MCLGET(mp, M_WAIT); 874 mp->m_len = 0; 875 mp2->m_next = mp; 876 mp2 = mp; 877 mlen = M_TRAILINGSPACE(mp); 878 } 879 xfer = (left > mlen) ? mlen : left; 880 #ifdef notdef 881 /* Not Yet.. */ 882 if (uiop->uio_iov->iov_op != NULL) 883 (*(uiop->uio_iov->iov_op)) 884 (uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 885 else 886 #endif 887 if (uiop->uio_segflg == UIO_SYSSPACE) 888 memcpy(mtod(mp, caddr_t)+mp->m_len, uiocp, xfer); 889 else 890 copyin(uiocp, mtod(mp, caddr_t)+mp->m_len, xfer); 891 mp->m_len += xfer; 892 left -= xfer; 893 uiocp += xfer; 894 uiop->uio_offset += xfer; 895 uiop->uio_resid -= xfer; 896 } 897 (caddr_t)uiop->uio_iov->iov_base += uiosiz; 898 uiop->uio_iov->iov_len -= uiosiz; 899 siz -= uiosiz; 900 } 901 if (rem > 0) { 902 if (rem > M_TRAILINGSPACE(mp)) { 903 MGET(mp, M_WAIT, MT_DATA); 904 mp->m_len = 0; 905 mp2->m_next = mp; 906 } 907 cp = mtod(mp, caddr_t)+mp->m_len; 908 for (left = 0; left < rem; left++) 909 *cp++ = '\0'; 910 mp->m_len += rem; 911 *bpos = cp; 912 } else 913 *bpos = mtod(mp, caddr_t)+mp->m_len; 914 *mq = mp; 915 return (0); 916 } 917 918 /* 919 * Get at least "siz" bytes of correctly aligned data. 920 * When called the mbuf pointers are not necessarily correct, 921 * dsosp points to what ought to be in m_data and left contains 922 * what ought to be in m_len. 923 * This is used by the macros nfsm_dissect and nfsm_dissecton for tough 924 * cases. (The macros use the vars. dpos and dpos2) 925 */ 926 int 927 nfsm_disct(mdp, dposp, siz, left, cp2) 928 struct mbuf **mdp; 929 caddr_t *dposp; 930 int siz; 931 int left; 932 caddr_t *cp2; 933 { 934 struct mbuf *m1, *m2; 935 struct mbuf *havebuf = NULL; 936 caddr_t src = *dposp; 937 caddr_t dst; 938 int len; 939 940 #ifdef DEBUG 941 if (left < 0) 942 panic("nfsm_disct: left < 0"); 943 #endif 944 m1 = *mdp; 945 /* 946 * Skip through the mbuf chain looking for an mbuf with 947 * some data. If the first mbuf found has enough data 948 * and it is correctly aligned return it. 949 */ 950 while (left == 0) { 951 havebuf = m1; 952 *mdp = m1 = m1->m_next; 953 if (m1 == NULL) 954 return (EBADRPC); 955 src = mtod(m1, caddr_t); 956 left = m1->m_len; 957 /* 958 * If we start a new mbuf and it is big enough 959 * and correctly aligned just return it, don't 960 * do any pull up. 961 */ 962 if (left >= siz && nfsm_aligned(src)) { 963 *cp2 = src; 964 *dposp = src + siz; 965 return (0); 966 } 967 } 968 if (m1->m_flags & M_EXT) { 969 if (havebuf) { 970 /* If the first mbuf with data has external data 971 * and there is a previous empty mbuf use it 972 * to move the data into. 973 */ 974 m2 = m1; 975 *mdp = m1 = havebuf; 976 if (m1->m_flags & M_EXT) { 977 MEXTREMOVE(m1); 978 } 979 } else { 980 /* 981 * If the first mbuf has a external data 982 * and there is no previous empty mbuf 983 * allocate a new mbuf and move the external 984 * data to the new mbuf. Also make the first 985 * mbuf look empty. 986 */ 987 m2 = m_get(M_WAIT, MT_DATA); 988 m2->m_ext = m1->m_ext; 989 m2->m_data = src; 990 m2->m_len = left; 991 MCLADDREFERENCE(m1, m2); 992 MEXTREMOVE(m1); 993 m2->m_next = m1->m_next; 994 m1->m_next = m2; 995 } 996 m1->m_len = 0; 997 dst = m1->m_dat; 998 } else { 999 /* 1000 * If the first mbuf has no external data 1001 * move the data to the front of the mbuf. 1002 */ 1003 if ((dst = m1->m_dat) != src) 1004 memmove(dst, src, left); 1005 dst += left; 1006 m1->m_len = left; 1007 m2 = m1->m_next; 1008 } 1009 m1->m_flags &= ~M_PKTHDR; 1010 *cp2 = m1->m_data = m1->m_dat; /* data is at beginning of buffer */ 1011 *dposp = mtod(m1, caddr_t) + siz; 1012 /* 1013 * Loop through mbufs pulling data up into first mbuf until 1014 * the first mbuf is full or there is no more data to 1015 * pullup. 1016 */ 1017 while ((len = (MLEN - m1->m_len)) != 0 && m2) { 1018 if ((len = min(len, m2->m_len)) != 0) 1019 memcpy(dst, m2->m_data, len); 1020 m1->m_len += len; 1021 dst += len; 1022 m2->m_data += len; 1023 m2->m_len -= len; 1024 m2 = m2->m_next; 1025 } 1026 if (m1->m_len < siz) 1027 return (EBADRPC); 1028 return (0); 1029 } 1030 1031 /* 1032 * Advance the position in the mbuf chain. 1033 */ 1034 int 1035 nfs_adv(mdp, dposp, offs, left) 1036 struct mbuf **mdp; 1037 caddr_t *dposp; 1038 int offs; 1039 int left; 1040 { 1041 struct mbuf *m; 1042 int s; 1043 1044 m = *mdp; 1045 s = left; 1046 while (s < offs) { 1047 offs -= s; 1048 m = m->m_next; 1049 if (m == NULL) 1050 return (EBADRPC); 1051 s = m->m_len; 1052 } 1053 *mdp = m; 1054 *dposp = mtod(m, caddr_t)+offs; 1055 return (0); 1056 } 1057 1058 /* 1059 * Copy a string into mbufs for the hard cases... 1060 */ 1061 int 1062 nfsm_strtmbuf(mb, bpos, cp, siz) 1063 struct mbuf **mb; 1064 char **bpos; 1065 const char *cp; 1066 long siz; 1067 { 1068 struct mbuf *m1 = NULL, *m2; 1069 long left, xfer, len, tlen; 1070 u_int32_t *tl; 1071 int putsize; 1072 1073 putsize = 1; 1074 m2 = *mb; 1075 left = M_TRAILINGSPACE(m2); 1076 if (left > 0) { 1077 tl = ((u_int32_t *)(*bpos)); 1078 *tl++ = txdr_unsigned(siz); 1079 putsize = 0; 1080 left -= NFSX_UNSIGNED; 1081 m2->m_len += NFSX_UNSIGNED; 1082 if (left > 0) { 1083 memcpy((caddr_t) tl, cp, left); 1084 siz -= left; 1085 cp += left; 1086 m2->m_len += left; 1087 left = 0; 1088 } 1089 } 1090 /* Loop around adding mbufs */ 1091 while (siz > 0) { 1092 MGET(m1, M_WAIT, MT_DATA); 1093 if (siz > MLEN) 1094 MCLGET(m1, M_WAIT); 1095 m1->m_len = NFSMSIZ(m1); 1096 m2->m_next = m1; 1097 m2 = m1; 1098 tl = mtod(m1, u_int32_t *); 1099 tlen = 0; 1100 if (putsize) { 1101 *tl++ = txdr_unsigned(siz); 1102 m1->m_len -= NFSX_UNSIGNED; 1103 tlen = NFSX_UNSIGNED; 1104 putsize = 0; 1105 } 1106 if (siz < m1->m_len) { 1107 len = nfsm_rndup(siz); 1108 xfer = siz; 1109 if (xfer < len) 1110 *(tl+(xfer>>2)) = 0; 1111 } else { 1112 xfer = len = m1->m_len; 1113 } 1114 memcpy((caddr_t) tl, cp, xfer); 1115 m1->m_len = len+tlen; 1116 siz -= xfer; 1117 cp += xfer; 1118 } 1119 *mb = m1; 1120 *bpos = mtod(m1, caddr_t)+m1->m_len; 1121 return (0); 1122 } 1123 1124 /* 1125 * Directory caching routines. They work as follows: 1126 * - a cache is maintained per VDIR nfsnode. 1127 * - for each offset cookie that is exported to userspace, and can 1128 * thus be thrown back at us as an offset to VOP_READDIR, store 1129 * information in the cache. 1130 * - cached are: 1131 * - cookie itself 1132 * - blocknumber (essentially just a search key in the buffer cache) 1133 * - entry number in block. 1134 * - offset cookie of block in which this entry is stored 1135 * - 32 bit cookie if NFSMNT_XLATECOOKIE is used. 1136 * - entries are looked up in a hash table 1137 * - also maintained is an LRU list of entries, used to determine 1138 * which ones to delete if the cache grows too large. 1139 * - if 32 <-> 64 translation mode is requested for a filesystem, 1140 * the cache also functions as a translation table 1141 * - in the translation case, invalidating the cache does not mean 1142 * flushing it, but just marking entries as invalid, except for 1143 * the <64bit cookie, 32bitcookie> pair which is still valid, to 1144 * still be able to use the cache as a translation table. 1145 * - 32 bit cookies are uniquely created by combining the hash table 1146 * entry value, and one generation count per hash table entry, 1147 * incremented each time an entry is appended to the chain. 1148 * - the cache is invalidated each time a direcory is modified 1149 * - sanity checks are also done; if an entry in a block turns 1150 * out not to have a matching cookie, the cache is invalidated 1151 * and a new block starting from the wanted offset is fetched from 1152 * the server. 1153 * - directory entries as read from the server are extended to contain 1154 * the 64bit and, optionally, the 32bit cookies, for sanity checking 1155 * the cache and exporting them to userspace through the cookie 1156 * argument to VOP_READDIR. 1157 */ 1158 1159 u_long 1160 nfs_dirhash(off) 1161 off_t off; 1162 { 1163 int i; 1164 char *cp = (char *)&off; 1165 u_long sum = 0L; 1166 1167 for (i = 0 ; i < sizeof (off); i++) 1168 sum += *cp++; 1169 1170 return sum; 1171 } 1172 1173 void 1174 nfs_initdircache(vp) 1175 struct vnode *vp; 1176 { 1177 struct nfsnode *np = VTONFS(vp); 1178 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1179 1180 np->n_dircachesize = 0; 1181 np->n_dblkno = 1; 1182 np->n_dircache = 1183 hashinit(NFS_DIRHASHSIZ, M_NFSDIROFF, M_WAITOK, &nfsdirhashmask); 1184 TAILQ_INIT(&np->n_dirchain); 1185 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1186 MALLOC(np->n_dirgens, unsigned *, 1187 NFS_DIRHASHSIZ * sizeof (unsigned), M_NFSDIROFF, 1188 M_WAITOK); 1189 memset((caddr_t)np->n_dirgens, 0, 1190 NFS_DIRHASHSIZ * sizeof (unsigned)); 1191 } 1192 } 1193 1194 static struct nfsdircache dzero = {0, 0, {0, 0}, {0, 0}, 0, 0, 0}; 1195 1196 struct nfsdircache * 1197 nfs_searchdircache(vp, off, do32, hashent) 1198 struct vnode *vp; 1199 off_t off; 1200 int do32; 1201 int *hashent; 1202 { 1203 struct nfsdirhashhead *ndhp; 1204 struct nfsdircache *ndp = NULL; 1205 struct nfsnode *np = VTONFS(vp); 1206 unsigned ent; 1207 1208 /* 1209 * Zero is always a valid cookie. 1210 */ 1211 if (off == 0) 1212 return &dzero; 1213 1214 /* 1215 * We use a 32bit cookie as search key, directly reconstruct 1216 * the hashentry. Else use the hashfunction. 1217 */ 1218 if (do32) { 1219 ent = (u_int32_t)off >> 24; 1220 if (ent >= NFS_DIRHASHSIZ) 1221 return NULL; 1222 ndhp = &np->n_dircache[ent]; 1223 } else { 1224 ndhp = NFSDIRHASH(np, off); 1225 } 1226 1227 if (hashent) 1228 *hashent = (int)(ndhp - np->n_dircache); 1229 if (do32) { 1230 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) { 1231 if (ndp->dc_cookie32 == (u_int32_t)off) { 1232 /* 1233 * An invalidated entry will become the 1234 * start of a new block fetched from 1235 * the server. 1236 */ 1237 if (ndp->dc_blkno == -1) { 1238 ndp->dc_blkcookie = ndp->dc_cookie; 1239 ndp->dc_blkno = np->n_dblkno++; 1240 ndp->dc_entry = 0; 1241 } 1242 break; 1243 } 1244 } 1245 } else { 1246 for (ndp = ndhp->lh_first; ndp; ndp = ndp->dc_hash.le_next) 1247 if (ndp->dc_cookie == off) 1248 break; 1249 } 1250 return ndp; 1251 } 1252 1253 1254 struct nfsdircache * 1255 nfs_enterdircache(vp, off, blkoff, en, blkno) 1256 struct vnode *vp; 1257 off_t off, blkoff; 1258 daddr_t blkno; 1259 int en; 1260 { 1261 struct nfsnode *np = VTONFS(vp); 1262 struct nfsdirhashhead *ndhp; 1263 struct nfsdircache *ndp = NULL, *first; 1264 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1265 int hashent, gen, overwrite; 1266 1267 if (!np->n_dircache) 1268 /* 1269 * XXX would like to do this in nfs_nget but vtype 1270 * isn't known at that time. 1271 */ 1272 nfs_initdircache(vp); 1273 1274 /* 1275 * XXX refuse entries for offset 0. amd(8) erroneously sets 1276 * cookie 0 for the '.' entry, making this necessary. This 1277 * isn't so bad, as 0 is a special case anyway. 1278 */ 1279 if (off == 0) 1280 return &dzero; 1281 1282 ndp = nfs_searchdircache(vp, off, 0, &hashent); 1283 1284 if (ndp && ndp->dc_blkno != -1) { 1285 /* 1286 * Overwriting an old entry. Check if it's the same. 1287 * If so, just return. If not, remove the old entry. 1288 */ 1289 if (ndp->dc_blkcookie == blkoff && ndp->dc_entry == en) 1290 return ndp; 1291 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1292 LIST_REMOVE(ndp, dc_hash); 1293 FREE(ndp, M_NFSDIROFF); 1294 ndp = 0; 1295 } 1296 1297 ndhp = &np->n_dircache[hashent]; 1298 1299 if (!ndp) { 1300 MALLOC(ndp, struct nfsdircache *, sizeof (*ndp), M_NFSDIROFF, 1301 M_WAITOK); 1302 overwrite = 0; 1303 if (nmp->nm_flag & NFSMNT_XLATECOOKIE) { 1304 /* 1305 * We're allocating a new entry, so bump the 1306 * generation number. 1307 */ 1308 gen = ++np->n_dirgens[hashent]; 1309 if (gen == 0) { 1310 np->n_dirgens[hashent]++; 1311 gen++; 1312 } 1313 ndp->dc_cookie32 = (hashent << 24) | (gen & 0xffffff); 1314 } 1315 } else 1316 overwrite = 1; 1317 1318 /* 1319 * If the entry number is 0, we are at the start of a new block, so 1320 * allocate a new blocknumber. 1321 */ 1322 if (en == 0) 1323 ndp->dc_blkno = np->n_dblkno++; 1324 else 1325 ndp->dc_blkno = blkno; 1326 1327 ndp->dc_cookie = off; 1328 ndp->dc_blkcookie = blkoff; 1329 ndp->dc_entry = en; 1330 1331 if (overwrite) 1332 return ndp; 1333 1334 /* 1335 * If the maximum directory cookie cache size has been reached 1336 * for this node, take one off the front. The idea is that 1337 * directories are typically read front-to-back once, so that 1338 * the oldest entries can be thrown away without much performance 1339 * loss. 1340 */ 1341 if (np->n_dircachesize == NFS_MAXDIRCACHE) { 1342 first = np->n_dirchain.tqh_first; 1343 TAILQ_REMOVE(&np->n_dirchain, first, dc_chain); 1344 LIST_REMOVE(first, dc_hash); 1345 FREE(first, M_NFSDIROFF); 1346 } else 1347 np->n_dircachesize++; 1348 1349 LIST_INSERT_HEAD(ndhp, ndp, dc_hash); 1350 TAILQ_INSERT_TAIL(&np->n_dirchain, ndp, dc_chain); 1351 return ndp; 1352 } 1353 1354 void 1355 nfs_invaldircache(vp, forcefree) 1356 struct vnode *vp; 1357 int forcefree; 1358 { 1359 struct nfsnode *np = VTONFS(vp); 1360 struct nfsdircache *ndp = NULL; 1361 struct nfsmount *nmp = VFSTONFS(vp->v_mount); 1362 1363 #ifdef DIAGNOSTIC 1364 if (vp->v_type != VDIR) 1365 panic("nfs: invaldircache: not dir"); 1366 #endif 1367 1368 if (!np->n_dircache) 1369 return; 1370 1371 if (!(nmp->nm_flag & NFSMNT_XLATECOOKIE) || forcefree) { 1372 while ((ndp = np->n_dirchain.tqh_first)) { 1373 TAILQ_REMOVE(&np->n_dirchain, ndp, dc_chain); 1374 LIST_REMOVE(ndp, dc_hash); 1375 FREE(ndp, M_NFSDIROFF); 1376 } 1377 np->n_dircachesize = 0; 1378 if (forcefree && np->n_dirgens) { 1379 FREE(np->n_dirgens, M_NFSDIROFF); 1380 } 1381 } else { 1382 for (ndp = np->n_dirchain.tqh_first; ndp; 1383 ndp = ndp->dc_chain.tqe_next) 1384 ndp->dc_blkno = -1; 1385 } 1386 1387 np->n_dblkno = 1; 1388 } 1389 1390 /* 1391 * Called once before VFS init to initialize shared and 1392 * server-specific data structures. 1393 */ 1394 void 1395 nfs_init() 1396 { 1397 1398 #if !defined(alpha) && !defined(_LP64) && defined(DIAGNOSTIC) 1399 /* 1400 * Check to see if major data structures haven't bloated. 1401 */ 1402 if (sizeof (struct nfsnode) > NFS_NODEALLOC) { 1403 printf("struct nfsnode bloated (> %dbytes)\n", NFS_NODEALLOC); 1404 printf("Try reducing NFS_SMALLFH\n"); 1405 } 1406 if (sizeof (struct nfssvc_sock) > NFS_SVCALLOC) { 1407 printf("struct nfssvc_sock bloated (> %dbytes)\n",NFS_SVCALLOC); 1408 printf("Try reducing NFS_UIDHASHSIZ\n"); 1409 } 1410 if (sizeof (struct nfsuid) > NFS_UIDALLOC) { 1411 printf("struct nfsuid bloated (> %dbytes)\n",NFS_UIDALLOC); 1412 printf("Try unionizing the nu_nickname and nu_flag fields\n"); 1413 } 1414 #endif 1415 1416 nfsrtt.pos = 0; 1417 rpc_vers = txdr_unsigned(RPC_VER2); 1418 rpc_call = txdr_unsigned(RPC_CALL); 1419 rpc_reply = txdr_unsigned(RPC_REPLY); 1420 rpc_msgdenied = txdr_unsigned(RPC_MSGDENIED); 1421 rpc_msgaccepted = txdr_unsigned(RPC_MSGACCEPTED); 1422 rpc_mismatch = txdr_unsigned(RPC_MISMATCH); 1423 rpc_autherr = txdr_unsigned(RPC_AUTHERR); 1424 rpc_auth_unix = txdr_unsigned(RPCAUTH_UNIX); 1425 rpc_auth_kerb = txdr_unsigned(RPCAUTH_KERB4); 1426 nfs_prog = txdr_unsigned(NFS_PROG); 1427 nqnfs_prog = txdr_unsigned(NQNFS_PROG); 1428 nfs_true = txdr_unsigned(TRUE); 1429 nfs_false = txdr_unsigned(FALSE); 1430 nfs_xdrneg1 = txdr_unsigned(-1); 1431 nfs_ticks = (hz * NFS_TICKINTVL + 500) / 1000; 1432 if (nfs_ticks < 1) 1433 nfs_ticks = 1; 1434 #ifdef NFSSERVER 1435 nfsrv_init(0); /* Init server data structures */ 1436 nfsrv_initcache(); /* Init the server request cache */ 1437 #endif /* NFSSERVER */ 1438 1439 /* 1440 * Initialize the nqnfs data structures. 1441 */ 1442 if (nqnfsstarttime == 0) { 1443 nqnfsstarttime = boottime.tv_sec + nqsrv_maxlease 1444 + nqsrv_clockskew + nqsrv_writeslack; 1445 NQLOADNOVRAM(nqnfsstarttime); 1446 CIRCLEQ_INIT(&nqtimerhead); 1447 nqfhhashtbl = hashinit(NQLCHSZ, M_NQLEASE, M_WAITOK, &nqfhhash); 1448 } 1449 1450 /* 1451 * Initialize reply list and start timer 1452 */ 1453 TAILQ_INIT(&nfs_reqq); 1454 nfs_timer(NULL); 1455 } 1456 1457 #ifdef NFS 1458 /* 1459 * Called once at VFS init to initialize client-specific data structures. 1460 */ 1461 void 1462 nfs_vfs_init() 1463 { 1464 int i; 1465 1466 /* Ensure async daemons disabled */ 1467 for (i = 0; i < NFS_MAXASYNCDAEMON; i++) { 1468 nfs_iodwant[i] = (struct proc *)0; 1469 nfs_iodmount[i] = (struct nfsmount *)0; 1470 } 1471 nfs_nhinit(); /* Init the nfsnode table */ 1472 } 1473 1474 void 1475 nfs_vfs_done() 1476 { 1477 nfs_nhdone(); 1478 } 1479 1480 /* 1481 * Attribute cache routines. 1482 * nfs_loadattrcache() - loads or updates the cache contents from attributes 1483 * that are on the mbuf list 1484 * nfs_getattrcache() - returns valid attributes if found in cache, returns 1485 * error otherwise 1486 */ 1487 1488 /* 1489 * Load the attribute cache (that lives in the nfsnode entry) with 1490 * the values on the mbuf list and 1491 * Iff vap not NULL 1492 * copy the attributes to *vaper 1493 */ 1494 int 1495 nfsm_loadattrcache(vpp, mdp, dposp, vaper) 1496 struct vnode **vpp; 1497 struct mbuf **mdp; 1498 caddr_t *dposp; 1499 struct vattr *vaper; 1500 { 1501 int32_t t1; 1502 caddr_t cp2; 1503 int error = 0; 1504 struct mbuf *md; 1505 int v3 = NFS_ISV3(*vpp); 1506 1507 md = *mdp; 1508 t1 = (mtod(md, caddr_t) + md->m_len) - *dposp; 1509 error = nfsm_disct(mdp, dposp, NFSX_FATTR(v3), t1, &cp2); 1510 if (error) 1511 return (error); 1512 return nfs_loadattrcache(vpp, (struct nfs_fattr *)cp2, vaper); 1513 } 1514 1515 int 1516 nfs_loadattrcache(vpp, fp, vaper) 1517 struct vnode **vpp; 1518 struct nfs_fattr *fp; 1519 struct vattr *vaper; 1520 { 1521 struct vnode *vp = *vpp; 1522 struct vattr *vap; 1523 int v3 = NFS_ISV3(vp); 1524 enum vtype vtyp; 1525 u_short vmode; 1526 struct timespec mtime; 1527 struct vnode *nvp; 1528 int32_t rdev; 1529 struct nfsnode *np; 1530 extern int (**spec_nfsv2nodeop_p) __P((void *)); 1531 1532 if (v3) { 1533 vtyp = nfsv3tov_type(fp->fa_type); 1534 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1535 rdev = makedev(fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata1), 1536 fxdr_unsigned(u_int32_t, fp->fa3_rdev.specdata2)); 1537 fxdr_nfsv3time(&fp->fa3_mtime, &mtime); 1538 } else { 1539 vtyp = nfsv2tov_type(fp->fa_type); 1540 vmode = fxdr_unsigned(u_short, fp->fa_mode); 1541 if (vtyp == VNON || vtyp == VREG) 1542 vtyp = IFTOVT(vmode); 1543 rdev = fxdr_unsigned(int32_t, fp->fa2_rdev); 1544 fxdr_nfsv2time(&fp->fa2_mtime, &mtime); 1545 1546 /* 1547 * Really ugly NFSv2 kludge. 1548 */ 1549 if (vtyp == VCHR && rdev == 0xffffffff) 1550 vtyp = VFIFO; 1551 } 1552 1553 /* 1554 * If v_type == VNON it is a new node, so fill in the v_type, 1555 * n_mtime fields. Check to see if it represents a special 1556 * device, and if so, check for a possible alias. Once the 1557 * correct vnode has been obtained, fill in the rest of the 1558 * information. 1559 */ 1560 np = VTONFS(vp); 1561 if (vp->v_type != vtyp) { 1562 vp->v_type = vtyp; 1563 if (vp->v_type == VFIFO) { 1564 extern int (**fifo_nfsv2nodeop_p) __P((void *)); 1565 vp->v_op = fifo_nfsv2nodeop_p; 1566 } 1567 if (vp->v_type == VCHR || vp->v_type == VBLK) { 1568 vp->v_op = spec_nfsv2nodeop_p; 1569 nvp = checkalias(vp, (dev_t)rdev, vp->v_mount); 1570 if (nvp) { 1571 /* 1572 * Discard unneeded vnode, but save its nfsnode. 1573 * Since the nfsnode does not have a lock, its 1574 * vnode lock has to be carried over. 1575 */ 1576 nvp->v_data = vp->v_data; 1577 vp->v_data = NULL; 1578 vp->v_op = spec_vnodeop_p; 1579 vput(vp); 1580 vgone(vp); 1581 /* 1582 * XXX When nfs starts locking, we need to 1583 * lock the new node here. 1584 */ 1585 /* 1586 * Reinitialize aliased node. 1587 */ 1588 np->n_vnode = nvp; 1589 *vpp = vp = nvp; 1590 } 1591 } 1592 np->n_mtime = mtime.tv_sec; 1593 } 1594 vap = np->n_vattr; 1595 vap->va_type = vtyp; 1596 vap->va_mode = vmode & ALLPERMS; 1597 vap->va_rdev = (dev_t)rdev; 1598 vap->va_mtime = mtime; 1599 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 1600 switch (vtyp) { 1601 case VDIR: 1602 vap->va_blocksize = NFS_DIRFRAGSIZ; 1603 break; 1604 case VBLK: 1605 vap->va_blocksize = BLKDEV_IOSIZE; 1606 break; 1607 case VCHR: 1608 vap->va_blocksize = MAXBSIZE; 1609 break; 1610 default: 1611 vap->va_blocksize = v3 ? vp->v_mount->mnt_stat.f_iosize : 1612 fxdr_unsigned(int32_t, fp->fa2_blocksize); 1613 break; 1614 } 1615 if (v3) { 1616 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1617 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1618 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1619 vap->va_size = fxdr_hyper(&fp->fa3_size); 1620 vap->va_bytes = fxdr_hyper(&fp->fa3_used); 1621 vap->va_fileid = fxdr_unsigned(int32_t, 1622 fp->fa3_fileid.nfsuquad[1]); 1623 fxdr_nfsv3time(&fp->fa3_atime, &vap->va_atime); 1624 fxdr_nfsv3time(&fp->fa3_ctime, &vap->va_ctime); 1625 vap->va_flags = 0; 1626 vap->va_filerev = 0; 1627 } else { 1628 vap->va_nlink = fxdr_unsigned(u_short, fp->fa_nlink); 1629 vap->va_uid = fxdr_unsigned(uid_t, fp->fa_uid); 1630 vap->va_gid = fxdr_unsigned(gid_t, fp->fa_gid); 1631 vap->va_size = fxdr_unsigned(u_int32_t, fp->fa2_size); 1632 vap->va_bytes = fxdr_unsigned(int32_t, fp->fa2_blocks) 1633 * NFS_FABLKSIZE; 1634 vap->va_fileid = fxdr_unsigned(int32_t, fp->fa2_fileid); 1635 fxdr_nfsv2time(&fp->fa2_atime, &vap->va_atime); 1636 vap->va_flags = 0; 1637 vap->va_ctime.tv_sec = fxdr_unsigned(u_int32_t, 1638 fp->fa2_ctime.nfsv2_sec); 1639 vap->va_ctime.tv_nsec = 0; 1640 vap->va_gen = fxdr_unsigned(u_int32_t,fp->fa2_ctime.nfsv2_usec); 1641 vap->va_filerev = 0; 1642 } 1643 if (vap->va_size != np->n_size) { 1644 if (vap->va_type == VREG) { 1645 if (np->n_flag & NMODIFIED) { 1646 if (vap->va_size < np->n_size) 1647 vap->va_size = np->n_size; 1648 else 1649 np->n_size = vap->va_size; 1650 } else 1651 np->n_size = vap->va_size; 1652 uvm_vnp_setsize(vp, np->n_size); 1653 } else 1654 np->n_size = vap->va_size; 1655 } 1656 np->n_attrstamp = time.tv_sec; 1657 if (vaper != NULL) { 1658 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(*vap)); 1659 if (np->n_flag & NCHG) { 1660 if (np->n_flag & NACC) 1661 vaper->va_atime = np->n_atim; 1662 if (np->n_flag & NUPD) 1663 vaper->va_mtime = np->n_mtim; 1664 } 1665 } 1666 return (0); 1667 } 1668 1669 /* 1670 * Check the time stamp 1671 * If the cache is valid, copy contents to *vap and return 0 1672 * otherwise return an error 1673 */ 1674 int 1675 nfs_getattrcache(vp, vaper) 1676 struct vnode *vp; 1677 struct vattr *vaper; 1678 { 1679 struct nfsnode *np = VTONFS(vp); 1680 struct vattr *vap; 1681 1682 if ((time.tv_sec - np->n_attrstamp) >= NFS_ATTRTIMEO(np)) { 1683 nfsstats.attrcache_misses++; 1684 return (ENOENT); 1685 } 1686 nfsstats.attrcache_hits++; 1687 vap = np->n_vattr; 1688 if (vap->va_size != np->n_size) { 1689 if (vap->va_type == VREG) { 1690 if (np->n_flag & NMODIFIED) { 1691 if (vap->va_size < np->n_size) 1692 vap->va_size = np->n_size; 1693 else 1694 np->n_size = vap->va_size; 1695 } else 1696 np->n_size = vap->va_size; 1697 uvm_vnp_setsize(vp, np->n_size); 1698 } else 1699 np->n_size = vap->va_size; 1700 } 1701 memcpy((caddr_t)vaper, (caddr_t)vap, sizeof(struct vattr)); 1702 if (np->n_flag & NCHG) { 1703 if (np->n_flag & NACC) 1704 vaper->va_atime = np->n_atim; 1705 if (np->n_flag & NUPD) 1706 vaper->va_mtime = np->n_mtim; 1707 } 1708 return (0); 1709 } 1710 1711 /* 1712 * Heuristic to see if the server XDR encodes directory cookies or not. 1713 * it is not supposed to, but a lot of servers may do this. Also, since 1714 * most/all servers will implement V2 as well, it is expected that they 1715 * may return just 32 bits worth of cookie information, so we need to 1716 * find out in which 32 bits this information is available. We do this 1717 * to avoid trouble with emulated binaries that can't handle 64 bit 1718 * directory offsets. 1719 */ 1720 1721 void 1722 nfs_cookieheuristic(vp, flagp, p, cred) 1723 struct vnode *vp; 1724 int *flagp; 1725 struct proc *p; 1726 struct ucred *cred; 1727 { 1728 struct uio auio; 1729 struct iovec aiov; 1730 caddr_t buf, cp; 1731 struct dirent *dp; 1732 off_t *cookies = NULL, *cop; 1733 int error, eof, nc, len; 1734 1735 MALLOC(buf, caddr_t, NFS_DIRFRAGSIZ, M_TEMP, M_WAITOK); 1736 1737 aiov.iov_base = buf; 1738 aiov.iov_len = NFS_DIRFRAGSIZ; 1739 auio.uio_iov = &aiov; 1740 auio.uio_iovcnt = 1; 1741 auio.uio_rw = UIO_READ; 1742 auio.uio_segflg = UIO_SYSSPACE; 1743 auio.uio_procp = p; 1744 auio.uio_resid = NFS_DIRFRAGSIZ; 1745 auio.uio_offset = 0; 1746 1747 error = VOP_READDIR(vp, &auio, cred, &eof, &cookies, &nc); 1748 1749 len = NFS_DIRFRAGSIZ - auio.uio_resid; 1750 if (error || len == 0) { 1751 FREE(buf, M_TEMP); 1752 if (cookies) 1753 free(cookies, M_TEMP); 1754 return; 1755 } 1756 1757 /* 1758 * Find the first valid entry and look at its offset cookie. 1759 */ 1760 1761 cp = buf; 1762 for (cop = cookies; len > 0; len -= dp->d_reclen) { 1763 dp = (struct dirent *)cp; 1764 if (dp->d_fileno != 0 && len >= dp->d_reclen) { 1765 if ((*cop >> 32) != 0 && (*cop & 0xffffffffLL) == 0) { 1766 *flagp |= NFSMNT_SWAPCOOKIE; 1767 nfs_invaldircache(vp, 0); 1768 nfs_vinvalbuf(vp, 0, cred, p, 1); 1769 } 1770 break; 1771 } 1772 cop++; 1773 cp += dp->d_reclen; 1774 } 1775 1776 FREE(buf, M_TEMP); 1777 free(cookies, M_TEMP); 1778 } 1779 #endif /* NFS */ 1780 1781 /* 1782 * Set up nameidata for a lookup() call and do it. 1783 * 1784 * If pubflag is set, this call is done for a lookup operation on the 1785 * public filehandle. In that case we allow crossing mountpoints and 1786 * absolute pathnames. However, the caller is expected to check that 1787 * the lookup result is within the public fs, and deny access if 1788 * it is not. 1789 */ 1790 int 1791 nfs_namei(ndp, fhp, len, slp, nam, mdp, dposp, retdirp, p, kerbflag, pubflag) 1792 struct nameidata *ndp; 1793 fhandle_t *fhp; 1794 int len; 1795 struct nfssvc_sock *slp; 1796 struct mbuf *nam; 1797 struct mbuf **mdp; 1798 caddr_t *dposp; 1799 struct vnode **retdirp; 1800 struct proc *p; 1801 int kerbflag, pubflag; 1802 { 1803 int i, rem; 1804 struct mbuf *md; 1805 char *fromcp, *tocp, *cp; 1806 struct iovec aiov; 1807 struct uio auio; 1808 struct vnode *dp; 1809 int error, rdonly, linklen; 1810 struct componentname *cnp = &ndp->ni_cnd; 1811 1812 *retdirp = (struct vnode *)0; 1813 MALLOC(cnp->cn_pnbuf, char *, len + 1, M_NAMEI, M_WAITOK); 1814 /* 1815 * Copy the name from the mbuf list to ndp->ni_pnbuf 1816 * and set the various ndp fields appropriately. 1817 */ 1818 fromcp = *dposp; 1819 tocp = cnp->cn_pnbuf; 1820 md = *mdp; 1821 rem = mtod(md, caddr_t) + md->m_len - fromcp; 1822 for (i = 0; i < len; i++) { 1823 while (rem == 0) { 1824 md = md->m_next; 1825 if (md == NULL) { 1826 error = EBADRPC; 1827 goto out; 1828 } 1829 fromcp = mtod(md, caddr_t); 1830 rem = md->m_len; 1831 } 1832 if (*fromcp == '\0' || (!pubflag && *fromcp == '/')) { 1833 error = EACCES; 1834 goto out; 1835 } 1836 *tocp++ = *fromcp++; 1837 rem--; 1838 } 1839 *tocp = '\0'; 1840 *mdp = md; 1841 *dposp = fromcp; 1842 len = nfsm_rndup(len)-len; 1843 if (len > 0) { 1844 if (rem >= len) 1845 *dposp += len; 1846 else if ((error = nfs_adv(mdp, dposp, len, rem)) != 0) 1847 goto out; 1848 } 1849 1850 /* 1851 * Extract and set starting directory. 1852 */ 1853 error = nfsrv_fhtovp(fhp, FALSE, &dp, ndp->ni_cnd.cn_cred, slp, 1854 nam, &rdonly, kerbflag, pubflag); 1855 if (error) 1856 goto out; 1857 if (dp->v_type != VDIR) { 1858 vrele(dp); 1859 error = ENOTDIR; 1860 goto out; 1861 } 1862 1863 if (rdonly) 1864 cnp->cn_flags |= RDONLY; 1865 1866 *retdirp = dp; 1867 1868 if (pubflag) { 1869 /* 1870 * Oh joy. For WebNFS, handle those pesky '%' escapes, 1871 * and the 'native path' indicator. 1872 */ 1873 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); 1874 fromcp = cnp->cn_pnbuf; 1875 tocp = cp; 1876 if ((unsigned char)*fromcp >= WEBNFS_SPECCHAR_START) { 1877 switch ((unsigned char)*fromcp) { 1878 case WEBNFS_NATIVE_CHAR: 1879 /* 1880 * 'Native' path for us is the same 1881 * as a path according to the NFS spec, 1882 * just skip the escape char. 1883 */ 1884 fromcp++; 1885 break; 1886 /* 1887 * More may be added in the future, range 0x80-0xff 1888 */ 1889 default: 1890 error = EIO; 1891 FREE(cp, M_NAMEI); 1892 goto out; 1893 } 1894 } 1895 /* 1896 * Translate the '%' escapes, URL-style. 1897 */ 1898 while (*fromcp != '\0') { 1899 if (*fromcp == WEBNFS_ESC_CHAR) { 1900 if (fromcp[1] != '\0' && fromcp[2] != '\0') { 1901 fromcp++; 1902 *tocp++ = HEXSTRTOI(fromcp); 1903 fromcp += 2; 1904 continue; 1905 } else { 1906 error = ENOENT; 1907 FREE(cp, M_NAMEI); 1908 goto out; 1909 } 1910 } else 1911 *tocp++ = *fromcp++; 1912 } 1913 *tocp = '\0'; 1914 FREE(cnp->cn_pnbuf, M_NAMEI); 1915 cnp->cn_pnbuf = cp; 1916 } 1917 1918 ndp->ni_pathlen = (tocp - cnp->cn_pnbuf) + 1; 1919 ndp->ni_segflg = UIO_SYSSPACE; 1920 1921 if (pubflag) { 1922 ndp->ni_rootdir = rootvnode; 1923 ndp->ni_loopcnt = 0; 1924 if (cnp->cn_pnbuf[0] == '/') 1925 dp = rootvnode; 1926 } else { 1927 cnp->cn_flags |= NOCROSSMOUNT; 1928 } 1929 1930 cnp->cn_proc = p; 1931 VREF(dp); 1932 1933 for (;;) { 1934 cnp->cn_nameptr = cnp->cn_pnbuf; 1935 ndp->ni_startdir = dp; 1936 /* 1937 * And call lookup() to do the real work 1938 */ 1939 error = lookup(ndp); 1940 if (error) 1941 break; 1942 /* 1943 * Check for encountering a symbolic link 1944 */ 1945 if ((cnp->cn_flags & ISSYMLINK) == 0) { 1946 if (cnp->cn_flags & (SAVENAME | SAVESTART)) { 1947 cnp->cn_flags |= HASBUF; 1948 return (0); 1949 } 1950 break; 1951 } else { 1952 if ((cnp->cn_flags & LOCKPARENT) && ndp->ni_pathlen == 1) 1953 VOP_UNLOCK(ndp->ni_dvp, 0); 1954 if (!pubflag) { 1955 vrele(ndp->ni_dvp); 1956 vput(ndp->ni_vp); 1957 ndp->ni_vp = NULL; 1958 error = EINVAL; 1959 break; 1960 } 1961 1962 if (ndp->ni_loopcnt++ >= MAXSYMLINKS) { 1963 error = ELOOP; 1964 break; 1965 } 1966 if (ndp->ni_pathlen > 1) 1967 MALLOC(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); 1968 else 1969 cp = cnp->cn_pnbuf; 1970 aiov.iov_base = cp; 1971 aiov.iov_len = MAXPATHLEN; 1972 auio.uio_iov = &aiov; 1973 auio.uio_iovcnt = 1; 1974 auio.uio_offset = 0; 1975 auio.uio_rw = UIO_READ; 1976 auio.uio_segflg = UIO_SYSSPACE; 1977 auio.uio_procp = (struct proc *)0; 1978 auio.uio_resid = MAXPATHLEN; 1979 error = VOP_READLINK(ndp->ni_vp, &auio, cnp->cn_cred); 1980 if (error) { 1981 badlink: 1982 if (ndp->ni_pathlen > 1) 1983 FREE(cp, M_NAMEI); 1984 break; 1985 } 1986 linklen = MAXPATHLEN - auio.uio_resid; 1987 if (linklen == 0) { 1988 error = ENOENT; 1989 goto badlink; 1990 } 1991 if (linklen + ndp->ni_pathlen >= MAXPATHLEN) { 1992 error = ENAMETOOLONG; 1993 goto badlink; 1994 } 1995 if (ndp->ni_pathlen > 1) { 1996 memcpy(cp + linklen, ndp->ni_next, ndp->ni_pathlen); 1997 FREE(cnp->cn_pnbuf, M_NAMEI); 1998 cnp->cn_pnbuf = cp; 1999 } else 2000 cnp->cn_pnbuf[linklen] = '\0'; 2001 ndp->ni_pathlen += linklen; 2002 vput(ndp->ni_vp); 2003 dp = ndp->ni_dvp; 2004 /* 2005 * Check if root directory should replace current directory. 2006 */ 2007 if (cnp->cn_pnbuf[0] == '/') { 2008 vrele(dp); 2009 dp = ndp->ni_rootdir; 2010 VREF(dp); 2011 } 2012 } 2013 } 2014 out: 2015 FREE(cnp->cn_pnbuf, M_NAMEI); 2016 return (error); 2017 } 2018 2019 /* 2020 * A fiddled version of m_adj() that ensures null fill to a long 2021 * boundary and only trims off the back end 2022 */ 2023 void 2024 nfsm_adj(mp, len, nul) 2025 struct mbuf *mp; 2026 int len; 2027 int nul; 2028 { 2029 struct mbuf *m; 2030 int count, i; 2031 char *cp; 2032 2033 /* 2034 * Trim from tail. Scan the mbuf chain, 2035 * calculating its length and finding the last mbuf. 2036 * If the adjustment only affects this mbuf, then just 2037 * adjust and return. Otherwise, rescan and truncate 2038 * after the remaining size. 2039 */ 2040 count = 0; 2041 m = mp; 2042 for (;;) { 2043 count += m->m_len; 2044 if (m->m_next == (struct mbuf *)0) 2045 break; 2046 m = m->m_next; 2047 } 2048 if (m->m_len > len) { 2049 m->m_len -= len; 2050 if (nul > 0) { 2051 cp = mtod(m, caddr_t)+m->m_len-nul; 2052 for (i = 0; i < nul; i++) 2053 *cp++ = '\0'; 2054 } 2055 return; 2056 } 2057 count -= len; 2058 if (count < 0) 2059 count = 0; 2060 /* 2061 * Correct length for chain is "count". 2062 * Find the mbuf with last data, adjust its length, 2063 * and toss data from remaining mbufs on chain. 2064 */ 2065 for (m = mp; m; m = m->m_next) { 2066 if (m->m_len >= count) { 2067 m->m_len = count; 2068 if (nul > 0) { 2069 cp = mtod(m, caddr_t)+m->m_len-nul; 2070 for (i = 0; i < nul; i++) 2071 *cp++ = '\0'; 2072 } 2073 break; 2074 } 2075 count -= m->m_len; 2076 } 2077 for (m = m->m_next;m;m = m->m_next) 2078 m->m_len = 0; 2079 } 2080 2081 /* 2082 * Make these functions instead of macros, so that the kernel text size 2083 * doesn't get too big... 2084 */ 2085 void 2086 nfsm_srvwcc(nfsd, before_ret, before_vap, after_ret, after_vap, mbp, bposp) 2087 struct nfsrv_descript *nfsd; 2088 int before_ret; 2089 struct vattr *before_vap; 2090 int after_ret; 2091 struct vattr *after_vap; 2092 struct mbuf **mbp; 2093 char **bposp; 2094 { 2095 struct mbuf *mb = *mbp, *mb2; 2096 char *bpos = *bposp; 2097 u_int32_t *tl; 2098 2099 if (before_ret) { 2100 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2101 *tl = nfs_false; 2102 } else { 2103 nfsm_build(tl, u_int32_t *, 7 * NFSX_UNSIGNED); 2104 *tl++ = nfs_true; 2105 txdr_hyper(before_vap->va_size, tl); 2106 tl += 2; 2107 txdr_nfsv3time(&(before_vap->va_mtime), tl); 2108 tl += 2; 2109 txdr_nfsv3time(&(before_vap->va_ctime), tl); 2110 } 2111 *bposp = bpos; 2112 *mbp = mb; 2113 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp); 2114 } 2115 2116 void 2117 nfsm_srvpostopattr(nfsd, after_ret, after_vap, mbp, bposp) 2118 struct nfsrv_descript *nfsd; 2119 int after_ret; 2120 struct vattr *after_vap; 2121 struct mbuf **mbp; 2122 char **bposp; 2123 { 2124 struct mbuf *mb = *mbp, *mb2; 2125 char *bpos = *bposp; 2126 u_int32_t *tl; 2127 struct nfs_fattr *fp; 2128 2129 if (after_ret) { 2130 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED); 2131 *tl = nfs_false; 2132 } else { 2133 nfsm_build(tl, u_int32_t *, NFSX_UNSIGNED + NFSX_V3FATTR); 2134 *tl++ = nfs_true; 2135 fp = (struct nfs_fattr *)tl; 2136 nfsm_srvfattr(nfsd, after_vap, fp); 2137 } 2138 *mbp = mb; 2139 *bposp = bpos; 2140 } 2141 2142 void 2143 nfsm_srvfattr(nfsd, vap, fp) 2144 struct nfsrv_descript *nfsd; 2145 struct vattr *vap; 2146 struct nfs_fattr *fp; 2147 { 2148 2149 fp->fa_nlink = txdr_unsigned(vap->va_nlink); 2150 fp->fa_uid = txdr_unsigned(vap->va_uid); 2151 fp->fa_gid = txdr_unsigned(vap->va_gid); 2152 if (nfsd->nd_flag & ND_NFSV3) { 2153 fp->fa_type = vtonfsv3_type(vap->va_type); 2154 fp->fa_mode = vtonfsv3_mode(vap->va_mode); 2155 txdr_hyper(vap->va_size, &fp->fa3_size); 2156 txdr_hyper(vap->va_bytes, &fp->fa3_used); 2157 fp->fa3_rdev.specdata1 = txdr_unsigned(major(vap->va_rdev)); 2158 fp->fa3_rdev.specdata2 = txdr_unsigned(minor(vap->va_rdev)); 2159 fp->fa3_fsid.nfsuquad[0] = 0; 2160 fp->fa3_fsid.nfsuquad[1] = txdr_unsigned(vap->va_fsid); 2161 fp->fa3_fileid.nfsuquad[0] = 0; 2162 fp->fa3_fileid.nfsuquad[1] = txdr_unsigned(vap->va_fileid); 2163 txdr_nfsv3time(&vap->va_atime, &fp->fa3_atime); 2164 txdr_nfsv3time(&vap->va_mtime, &fp->fa3_mtime); 2165 txdr_nfsv3time(&vap->va_ctime, &fp->fa3_ctime); 2166 } else { 2167 fp->fa_type = vtonfsv2_type(vap->va_type); 2168 fp->fa_mode = vtonfsv2_mode(vap->va_type, vap->va_mode); 2169 fp->fa2_size = txdr_unsigned(vap->va_size); 2170 fp->fa2_blocksize = txdr_unsigned(vap->va_blocksize); 2171 if (vap->va_type == VFIFO) 2172 fp->fa2_rdev = 0xffffffff; 2173 else 2174 fp->fa2_rdev = txdr_unsigned(vap->va_rdev); 2175 fp->fa2_blocks = txdr_unsigned(vap->va_bytes / NFS_FABLKSIZE); 2176 fp->fa2_fsid = txdr_unsigned(vap->va_fsid); 2177 fp->fa2_fileid = txdr_unsigned(vap->va_fileid); 2178 txdr_nfsv2time(&vap->va_atime, &fp->fa2_atime); 2179 txdr_nfsv2time(&vap->va_mtime, &fp->fa2_mtime); 2180 txdr_nfsv2time(&vap->va_ctime, &fp->fa2_ctime); 2181 } 2182 } 2183 2184 /* 2185 * nfsrv_fhtovp() - convert a fh to a vnode ptr (optionally locked) 2186 * - look up fsid in mount list (if not found ret error) 2187 * - get vp and export rights by calling VFS_FHTOVP() 2188 * - if cred->cr_uid == 0 or MNT_EXPORTANON set it to credanon 2189 * - if not lockflag unlock it with VOP_UNLOCK() 2190 */ 2191 int 2192 nfsrv_fhtovp(fhp, lockflag, vpp, cred, slp, nam, rdonlyp, kerbflag, pubflag) 2193 fhandle_t *fhp; 2194 int lockflag; 2195 struct vnode **vpp; 2196 struct ucred *cred; 2197 struct nfssvc_sock *slp; 2198 struct mbuf *nam; 2199 int *rdonlyp; 2200 int kerbflag; 2201 { 2202 struct mount *mp; 2203 int i; 2204 struct ucred *credanon; 2205 int error, exflags; 2206 struct sockaddr_in *saddr; 2207 2208 *vpp = (struct vnode *)0; 2209 2210 if (nfs_ispublicfh(fhp)) { 2211 if (!pubflag || !nfs_pub.np_valid) 2212 return (ESTALE); 2213 fhp = &nfs_pub.np_handle; 2214 } 2215 2216 mp = vfs_getvfs(&fhp->fh_fsid); 2217 if (!mp) 2218 return (ESTALE); 2219 error = VFS_CHECKEXP(mp, nam, &exflags, &credanon); 2220 if (error) 2221 return (error); 2222 error = VFS_FHTOVP(mp, &fhp->fh_fid, vpp); 2223 if (error) 2224 return (error); 2225 2226 if (!(exflags & (MNT_EXNORESPORT|MNT_EXPUBLIC))) { 2227 saddr = mtod(nam, struct sockaddr_in *); 2228 if ((saddr->sin_family == AF_INET) && 2229 ntohs(saddr->sin_port) >= IPPORT_RESERVED) { 2230 vput(*vpp); 2231 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2232 } 2233 #ifdef INET6 2234 if ((saddr->sin_family == AF_INET6) && 2235 ntohs(saddr->sin_port) >= IPV6PORT_RESERVED) { 2236 vput(*vpp); 2237 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2238 } 2239 #endif 2240 } 2241 /* 2242 * Check/setup credentials. 2243 */ 2244 if (exflags & MNT_EXKERB) { 2245 if (!kerbflag) { 2246 vput(*vpp); 2247 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2248 } 2249 } else if (kerbflag) { 2250 vput(*vpp); 2251 return (NFSERR_AUTHERR | AUTH_TOOWEAK); 2252 } else if (cred->cr_uid == 0 || (exflags & MNT_EXPORTANON)) { 2253 cred->cr_uid = credanon->cr_uid; 2254 cred->cr_gid = credanon->cr_gid; 2255 for (i = 0; i < credanon->cr_ngroups && i < NGROUPS; i++) 2256 cred->cr_groups[i] = credanon->cr_groups[i]; 2257 cred->cr_ngroups = i; 2258 } 2259 if (exflags & MNT_EXRDONLY) 2260 *rdonlyp = 1; 2261 else 2262 *rdonlyp = 0; 2263 if (!lockflag) 2264 VOP_UNLOCK(*vpp, 0); 2265 return (0); 2266 } 2267 2268 /* 2269 * WebNFS: check if a filehandle is a public filehandle. For v3, this 2270 * means a length of 0, for v2 it means all zeroes. nfsm_srvmtofh has 2271 * transformed this to all zeroes in both cases, so check for it. 2272 */ 2273 int 2274 nfs_ispublicfh(fhp) 2275 fhandle_t *fhp; 2276 { 2277 char *cp = (char *)fhp; 2278 int i; 2279 2280 for (i = 0; i < NFSX_V3FH; i++) 2281 if (*cp++ != 0) 2282 return (FALSE); 2283 return (TRUE); 2284 } 2285 2286 /* 2287 * This function compares two net addresses by family and returns TRUE 2288 * if they are the same host. 2289 * If there is any doubt, return FALSE. 2290 * The AF_INET family is handled as a special case so that address mbufs 2291 * don't need to be saved to store "struct in_addr", which is only 4 bytes. 2292 */ 2293 int 2294 netaddr_match(family, haddr, nam) 2295 int family; 2296 union nethostaddr *haddr; 2297 struct mbuf *nam; 2298 { 2299 struct sockaddr_in *inetaddr; 2300 2301 switch (family) { 2302 case AF_INET: 2303 inetaddr = mtod(nam, struct sockaddr_in *); 2304 if (inetaddr->sin_family == AF_INET && 2305 inetaddr->sin_addr.s_addr == haddr->had_inetaddr) 2306 return (1); 2307 break; 2308 #ifdef INET6 2309 case AF_INET6: 2310 { 2311 struct sockaddr_in6 *sin6_1, *sin6_2; 2312 2313 sin6_1 = mtod(nam, struct sockaddr_in6 *); 2314 sin6_2 = mtod(haddr->had_nam, struct sockaddr_in6 *); 2315 if (sin6_1->sin6_family == AF_INET6 && 2316 IN6_ARE_ADDR_EQUAL(&sin6_1->sin6_addr, &sin6_2->sin6_addr)) 2317 return 1; 2318 } 2319 #endif 2320 #ifdef ISO 2321 case AF_ISO: 2322 { 2323 struct sockaddr_iso *isoaddr1, *isoaddr2; 2324 2325 isoaddr1 = mtod(nam, struct sockaddr_iso *); 2326 isoaddr2 = mtod(haddr->had_nam, struct sockaddr_iso *); 2327 if (isoaddr1->siso_family == AF_ISO && 2328 isoaddr1->siso_nlen > 0 && 2329 isoaddr1->siso_nlen == isoaddr2->siso_nlen && 2330 SAME_ISOADDR(isoaddr1, isoaddr2)) 2331 return (1); 2332 break; 2333 } 2334 #endif /* ISO */ 2335 default: 2336 break; 2337 }; 2338 return (0); 2339 } 2340 2341 2342 /* 2343 * The write verifier has changed (probably due to a server reboot), so all 2344 * B_NEEDCOMMIT blocks will have to be written again. Since they are on the 2345 * dirty block list as B_DELWRI, all this takes is clearing the B_NEEDCOMMIT 2346 * flag. Once done the new write verifier can be set for the mount point. 2347 */ 2348 void 2349 nfs_clearcommit(mp) 2350 struct mount *mp; 2351 { 2352 struct vnode *vp, *nvp; 2353 struct buf *bp, *nbp; 2354 int s; 2355 2356 s = splbio(); 2357 loop: 2358 for (vp = mp->mnt_vnodelist.lh_first; vp; vp = nvp) { 2359 if (vp->v_mount != mp) /* Paranoia */ 2360 goto loop; 2361 nvp = vp->v_mntvnodes.le_next; 2362 for (bp = vp->v_dirtyblkhd.lh_first; bp; bp = nbp) { 2363 nbp = bp->b_vnbufs.le_next; 2364 if ((bp->b_flags & (B_BUSY | B_DELWRI | B_NEEDCOMMIT)) 2365 == (B_DELWRI | B_NEEDCOMMIT)) 2366 bp->b_flags &= ~B_NEEDCOMMIT; 2367 } 2368 } 2369 splx(s); 2370 } 2371 2372 /* 2373 * Map errnos to NFS error numbers. For Version 3 also filter out error 2374 * numbers not specified for the associated procedure. 2375 */ 2376 int 2377 nfsrv_errmap(nd, err) 2378 struct nfsrv_descript *nd; 2379 int err; 2380 { 2381 short *defaulterrp, *errp; 2382 2383 if (nd->nd_flag & ND_NFSV3) { 2384 if (nd->nd_procnum <= NFSPROC_COMMIT) { 2385 errp = defaulterrp = nfsrv_v3errmap[nd->nd_procnum]; 2386 while (*++errp) { 2387 if (*errp == err) 2388 return (err); 2389 else if (*errp > err) 2390 break; 2391 } 2392 return ((int)*defaulterrp); 2393 } else 2394 return (err & 0xffff); 2395 } 2396 if (err <= ELAST) 2397 return ((int)nfsrv_v2errmap[err - 1]); 2398 return (NFSERR_IO); 2399 } 2400 2401 /* 2402 * Sort the group list in increasing numerical order. 2403 * (Insertion sort by Chris Torek, who was grossed out by the bubble sort 2404 * that used to be here.) 2405 */ 2406 void 2407 nfsrvw_sort(list, num) 2408 gid_t *list; 2409 int num; 2410 { 2411 int i, j; 2412 gid_t v; 2413 2414 /* Insertion sort. */ 2415 for (i = 1; i < num; i++) { 2416 v = list[i]; 2417 /* find correct slot for value v, moving others up */ 2418 for (j = i; --j >= 0 && v < list[j];) 2419 list[j + 1] = list[j]; 2420 list[j + 1] = v; 2421 } 2422 } 2423 2424 /* 2425 * copy credentials making sure that the result can be compared with memcmp(). 2426 */ 2427 void 2428 nfsrv_setcred(incred, outcred) 2429 struct ucred *incred, *outcred; 2430 { 2431 int i; 2432 2433 memset((caddr_t)outcred, 0, sizeof (struct ucred)); 2434 outcred->cr_ref = 1; 2435 outcred->cr_uid = incred->cr_uid; 2436 outcred->cr_gid = incred->cr_gid; 2437 outcred->cr_ngroups = incred->cr_ngroups; 2438 for (i = 0; i < incred->cr_ngroups; i++) 2439 outcred->cr_groups[i] = incred->cr_groups[i]; 2440 nfsrvw_sort(outcred->cr_groups, outcred->cr_ngroups); 2441 } 2442
Indexes created Wed Sep 24 05:09:52 GMT 2025