1 /* vfs_lookup.c 6.15 84/08/29 */ 2 3 #include "param.h" 4 #include "systm.h" 5 #include "inode.h" 6 #include "fs.h" 7 #include "mount.h" 8 #include "dir.h" 9 #include "user.h" 10 #include "buf.h" 11 #include "conf.h" 12 #include "uio.h" 13 #include "kernel.h" 14 15 struct buf *blkatoff(); 16 struct buf *freenamebuf; 17 int dirchk = 0; 18 19 /* 20 * Structures associated with name cacheing. 21 */ 22 #define NCHHASH 32 /* size of hash table */ 23 24 #if ((NCHHASH)&((NCHHASH)-1)) != 0 25 #define NHASH(h, i, d) ((unsigned)((h) + (i) + 13 * (int)(d)) % (NCHHASH)) 26 #else 27 #define NHASH(h, i, d) ((unsigned)((h) + (i) + 13 * (int)(d)) & ((NCHHASH)-1)) 28 #endif 29 30 union nchash { 31 union nchash *nch_head[2]; 32 struct nch *nch_chain[2]; 33 } nchash[NCHHASH]; 34 #define nch_forw nch_chain[0] 35 #define nch_back nch_chain[1] 36 37 struct nch *nchhead, **nchtail; /* LRU chain pointers */ 38 struct nchstats nchstats; /* cache effectiveness statistics */ 39 40 /* 41 * Convert a pathname into a pointer to a locked inode, 42 * with side effects usable in creating and removing files. 43 * This is a very central and rather complicated routine. 44 * 45 * The segflg defines whether the name is to be copied from user 46 * space or kernel space. 47 * 48 * The flag argument is (LOOKUP, CREATE, DELETE) depending on whether 49 * the name is to be (looked up, created, deleted). If flag has 50 * LOCKPARENT or'ed into it and the target of the pathname exists, 51 * namei returns both the target and its parent directory locked. 52 * If the file system is not maintained in a strict tree hierarchy, 53 * this can result in a deadlock situation. When creating and 54 * LOCKPARENT is specified, the target may not be ".". When deleting 55 * and LOCKPARENT is specified, the target may be ".", but the caller 56 * must check to insure it does an irele and iput instead of two iputs. 57 * 58 * The FOLLOW flag is set when symbolic links are to be followed 59 * when they occur at the end of the name translation process. 60 * 61 * Name caching works as follows: 62 * 63 * names found by directory scans are retained in a cache 64 * for future reference. It is managed LRU, so frequently 65 * used names will hang around. Cache is indexed by hash value 66 * obtained from (ino,dev,name) where ino & dev refer to the 67 * directory containing name. 68 * 69 * For simplicity (and economy of storage), names longer than 70 * some (small) maximum length are not cached, they occur 71 * infrequently in any case, and are almost never of interest. 72 * 73 * Upon reaching the last segment of a path, if the reference 74 * is for DELETE, or NOCACHE is set (rewrite), and the 75 * name is located in the cache, it will be dropped. 76 * 77 * We must be sure never to enter the name ".." into the cache 78 * because of the extremely kludgey way that rename() alters 79 * ".." in a situation like 80 * mv a/x b/x 81 * where x is a directory, and x/.. is the ".." in question. 82 * 83 * Overall outline of namei: 84 * 85 * copy in name 86 * get starting directory 87 * dirloop: 88 * check accessibility of directory 89 * dirloop2: 90 * copy next component of name to ndp->ni_dent 91 * handle degenerate case where name is null string 92 * look for name in cache, if found, then if at end of path 93 * and deleting or creating, drop it, else to haveino 94 * search for name in directory, to found or notfound 95 * notfound: 96 * if creating, return locked directory, leaving info on avail. slots 97 * else return error 98 * found: 99 * if at end of path and deleting, return information to allow delete 100 * if at end of path and rewriting (create and LOCKPARENT), lock target 101 * inode and return info to allow rewrite 102 * if .. and on mounted filesys, look in mount table for parent 103 * if not at end, if neither creating nor deleting, add name to cache 104 * haveino: 105 * if symbolic link, massage name in buffer and continue at dirloop 106 * if more components of name, do next level at dirloop 107 * return the answer as locked inode 108 * 109 * NOTE: (LOOKUP | LOCKPARENT) currently returns the parent inode, 110 * but unlocked. 111 */ 112 struct inode * 113 namei(ndp) 114 register struct nameidata *ndp; 115 { 116 register char *cp; /* pointer into pathname argument */ 117 /* these variables refer to things which must be freed or unlocked */ 118 register struct inode *dp = 0; /* the directory we are searching */ 119 register struct nch *ncp; /* cache slot for entry */ 120 register struct fs *fs; /* file system that directory is in */ 121 register struct buf *bp = 0; /* a buffer of directory entries */ 122 register struct direct *ep; /* the current directory entry */ 123 int entryoffsetinblock; /* offset of ep in bp's buffer */ 124 register struct buf *nbp; /* buffer storing path name argument */ 125 /* these variables hold information about the search for a slot */ 126 enum {NONE, COMPACT, FOUND} slotstatus; 127 int slotoffset = -1; /* offset of area with free space */ 128 int slotsize; /* size of area at slotoffset */ 129 int slotfreespace; /* amount of space free in slot */ 130 int slotneeded; /* size of the entry we're seeking */ 131 /* */ 132 int numdirpasses; /* strategy for directory search */ 133 int endsearch; /* offset to end directory search */ 134 int prevoff; /* ndp->ni_offset of previous entry */ 135 int nlink = 0; /* number of symbolic links taken */ 136 struct inode *pdp; /* saved dp during symlink work */ 137 int error, i; 138 int lockparent; 139 int docache; 140 unsigned hash; /* value of name hash for entry */ 141 union nchash *nhp; /* cache chain head for entry */ 142 int isdotdot; /* != 0 if current name is ".." */ 143 int flag; /* op ie, LOOKUP, CREATE, or DELETE */ 144 145 lockparent = ndp->ni_nameiop & LOCKPARENT; 146 docache = (ndp->ni_nameiop & NOCACHE) ^ NOCACHE; 147 flag = ndp->ni_nameiop &~ (LOCKPARENT|NOCACHE|FOLLOW); 148 if (flag == DELETE) 149 docache = 0; 150 /* 151 * Get a buffer for the name to be translated, and copy the 152 * name into the buffer. 153 */ 154 nbp = freenamebuf; 155 if (nbp == NULL) 156 nbp = geteblk(MAXPATHLEN); 157 else 158 freenamebuf = nbp->av_forw; 159 if (ndp->ni_segflg == UIO_SYSSPACE) 160 error = copystr(ndp->ni_dirp, nbp->b_un.b_addr, MAXPATHLEN, 161 (u_int *)0); 162 else 163 error = copyinstr(ndp->ni_dirp, nbp->b_un.b_addr, MAXPATHLEN, 164 (u_int *)0); 165 if (error) { 166 u.u_error = error; 167 goto bad; 168 } 169 170 /* 171 * Get starting directory. 172 */ 173 cp = nbp->b_un.b_addr; 174 if (*cp == '/') { 175 while (*cp == '/') 176 cp++; 177 if ((dp = u.u_rdir) == NULL) 178 dp = rootdir; 179 } else 180 dp = u.u_cdir; 181 fs = dp->i_fs; 182 ILOCK(dp); 183 dp->i_count++; 184 ndp->ni_pdir = (struct inode *)0xc0000000; /* illegal */ 185 186 /* 187 * We come to dirloop to search a new directory. 188 * The directory must be locked so that it can be 189 * iput, and fs must be already set to dp->i_fs. 190 */ 191 dirloop: 192 /* 193 * Check accessiblity of directory. 194 */ 195 if ((dp->i_mode&IFMT) != IFDIR) { 196 u.u_error = ENOTDIR; 197 goto bad; 198 } 199 if (access(dp, IEXEC)) 200 goto bad; 201 202 dirloop2: 203 /* 204 * Copy next component of name to ndp->ni_dent. 205 */ 206 hash = 0; 207 for (i = 0; *cp != 0 && *cp != '/'; cp++) { 208 if (i >= MAXNAMLEN) { 209 u.u_error = ENOENT; 210 goto bad; 211 } 212 if ((*cp&0377) == ('/'|0200) || (*cp&0200) && flag != DELETE) { 213 u.u_error = EPERM; 214 goto bad; 215 } 216 ndp->ni_dent.d_name[i++] = *cp; 217 hash += (unsigned char)*cp * i; 218 } 219 ndp->ni_dent.d_namlen = i; 220 ndp->ni_dent.d_name[i] = '\0'; 221 isdotdot = (i == 2 && 222 ndp->ni_dent.d_name[0] == '.' && ndp->ni_dent.d_name[1] == '.'); 223 224 /* 225 * Check for degenerate name (e.g. / or "") 226 * which is a way of talking about a directory, 227 * e.g. like "/." or ".". 228 */ 229 if (ndp->ni_dent.d_name[0] == '\0') { 230 if (flag != LOOKUP || lockparent) { 231 u.u_error = EISDIR; 232 goto bad; 233 } 234 nbp->av_forw = freenamebuf; 235 freenamebuf = nbp; 236 return (dp); 237 } 238 239 /* 240 * We now have a segment name to search for, and a directory to search. 241 * 242 * Before tediously performing a linear scan of the directory, 243 * check the name cache to see if the directory/name pair 244 * we are looking for is known already. We don't do this 245 * if the segment name is long, simply so the cache can avoid 246 * holding long names (which would either waste space, or 247 * add greatly to the complexity). 248 */ 249 if (ndp->ni_dent.d_namlen > NCHNAMLEN) { 250 nchstats.ncs_long++; 251 docache = 0; 252 } else { 253 nhp = &nchash[NHASH(hash, dp->i_number, dp->i_dev)]; 254 for (ncp = nhp->nch_forw; ncp != (struct nch *)nhp; 255 ncp = ncp->nc_forw) { 256 if (ncp->nc_ino == dp->i_number && 257 ncp->nc_dev == dp->i_dev && 258 ncp->nc_nlen == ndp->ni_dent.d_namlen && 259 !bcmp(ncp->nc_name, ndp->ni_dent.d_name, 260 ncp->nc_nlen)) 261 break; 262 } 263 264 if (ncp == (struct nch *)nhp) { 265 nchstats.ncs_miss++; 266 ncp = NULL; 267 } else { 268 if (ncp->nc_id != ncp->nc_ip->i_id) { 269 nchstats.ncs_falsehits++; 270 } else if (*cp == '\0' && !docache) { 271 nchstats.ncs_badhits++; 272 } else { 273 274 /* 275 * move this slot to end of LRU 276 * chain, if not already there 277 */ 278 if (ncp->nc_nxt) { 279 /* remove from LRU chain */ 280 *ncp->nc_prev = ncp->nc_nxt; 281 ncp->nc_nxt->nc_prev = ncp->nc_prev; 282 283 /* and replace at end of it */ 284 ncp->nc_nxt = NULL; 285 ncp->nc_prev = nchtail; 286 *nchtail = ncp; 287 nchtail = &ncp->nc_nxt; 288 } 289 290 /* 291 * Get the next inode in the path. 292 * See comment above other `IUNLOCK' code for 293 * an explaination of the locking protocol. 294 */ 295 pdp = dp; 296 dp = ncp->nc_ip; 297 if (dp == NULL) 298 panic("nami: null cache ino"); 299 if (pdp == dp) 300 dp->i_count++; 301 else { 302 if (isdotdot) { 303 IUNLOCK(pdp); 304 igrab(dp); 305 } else { 306 igrab(dp); 307 IUNLOCK(pdp); 308 } 309 } 310 311 /* 312 * Verify that the inode that we got 313 * did not change while we were waiting 314 * for it to be locked. 315 */ 316 if (ncp->nc_id != ncp->nc_ip->i_id) { 317 iput(dp); 318 ILOCK(pdp); 319 dp = pdp; 320 nchstats.ncs_falsehits++; 321 } else { 322 ndp->ni_dent.d_ino = dp->i_number; 323 /* ni_dent.d_reclen is garbage ... */ 324 nchstats.ncs_goodhits++; 325 goto haveino; 326 } 327 } 328 329 /* 330 * Last component and we are renaming or deleting, 331 * the cache entry is invalid, or otherwise don't 332 * want cache entry to exist. 333 */ 334 335 /* remove from LRU chain */ 336 *ncp->nc_prev = ncp->nc_nxt; 337 if (ncp->nc_nxt) 338 ncp->nc_nxt->nc_prev = ncp->nc_prev; 339 else 340 nchtail = ncp->nc_prev; 341 342 /* remove from hash chain */ 343 remque(ncp); 344 345 /* insert at head of LRU list (first to grab) */ 346 ncp->nc_nxt = nchhead; 347 ncp->nc_prev = &nchhead; 348 nchhead->nc_prev = &ncp->nc_nxt; 349 nchhead = ncp; 350 351 /* and make a dummy hash chain */ 352 ncp->nc_forw = ncp; 353 ncp->nc_back = ncp; 354 355 ncp = NULL; 356 } 357 } 358 359 /* 360 * Suppress search for slots unless creating 361 * file and at end of pathname, in which case 362 * we watch for a place to put the new file in 363 * case it doesn't already exist. 364 */ 365 slotstatus = FOUND; 366 if (flag == CREATE && *cp == 0) { 367 slotstatus = NONE; 368 slotfreespace = 0; 369 slotneeded = DIRSIZ(&ndp->ni_dent); 370 } 371 /* 372 * If this is the same directory that this process 373 * previously searched, pick up where we last left off. 374 * We cache only lookups as these are the most common 375 * and have the greatest payoff. Caching CREATE has little 376 * benefit as it usually must search the entire directory 377 * to determine that the entry does not exist. Caching the 378 * location of the last DELETE has not reduced profiling time 379 * and hence has been removed in the interest of simplicity. 380 */ 381 if (flag != LOOKUP || dp->i_number != u.u_ncache.nc_inumber || 382 dp->i_dev != u.u_ncache.nc_dev) { 383 ndp->ni_offset = 0; 384 numdirpasses = 1; 385 } else { 386 if ((dp->i_flag & ICHG) || dp->i_ctime >= u.u_ncache.nc_time) { 387 u.u_ncache.nc_prevoffset &= ~(DIRBLKSIZ - 1); 388 u.u_ncache.nc_time = time.tv_sec; 389 } 390 ndp->ni_offset = u.u_ncache.nc_prevoffset; 391 entryoffsetinblock = blkoff(fs, ndp->ni_offset); 392 if (entryoffsetinblock != 0) { 393 bp = blkatoff(dp, ndp->ni_offset, (char **)0); 394 if (bp == 0) 395 goto bad; 396 } 397 numdirpasses = 2; 398 nchstats.ncs_2passes++; 399 } 400 endsearch = roundup(dp->i_size, DIRBLKSIZ); 401 402 searchloop: 403 while (ndp->ni_offset < endsearch) { 404 /* 405 * If offset is on a block boundary, 406 * read the next directory block. 407 * Release previous if it exists. 408 */ 409 if (blkoff(fs, ndp->ni_offset) == 0) { 410 if (bp != NULL) 411 brelse(bp); 412 bp = blkatoff(dp, ndp->ni_offset, (char **)0); 413 if (bp == 0) 414 goto bad; 415 entryoffsetinblock = 0; 416 } 417 418 /* 419 * If still looking for a slot, and at a DIRBLKSIZE 420 * boundary, have to start looking for free space again. 421 */ 422 if (slotstatus == NONE && 423 (entryoffsetinblock&(DIRBLKSIZ-1)) == 0) { 424 slotoffset = -1; 425 slotfreespace = 0; 426 } 427 428 /* 429 * Get pointer to next entry. 430 * Full validation checks are slow, so we only check 431 * enough to insure forward progress through the 432 * directory. Complete checks can be run by patching 433 * "dirchk" to be true. 434 */ 435 ep = (struct direct *)(bp->b_un.b_addr + entryoffsetinblock); 436 if (ep->d_reclen <= 0 || 437 dirchk && dirbadentry(ep, entryoffsetinblock)) { 438 dirbad(dp, ndp->ni_offset, "mangled entry"); 439 i = DIRBLKSIZ - (entryoffsetinblock & (DIRBLKSIZ - 1)); 440 ndp->ni_offset += i; 441 entryoffsetinblock += i; 442 continue; 443 } 444 445 /* 446 * If an appropriate sized slot has not yet been found, 447 * check to see if one is available. Also accumulate space 448 * in the current block so that we can determine if 449 * compaction is viable. 450 */ 451 if (slotstatus != FOUND) { 452 int size = ep->d_reclen; 453 454 if (ep->d_ino != 0) 455 size -= DIRSIZ(ep); 456 if (size > 0) { 457 if (size >= slotneeded) { 458 slotstatus = FOUND; 459 slotoffset = ndp->ni_offset; 460 slotsize = ep->d_reclen; 461 } else if (slotstatus == NONE) { 462 slotfreespace += size; 463 if (slotoffset == -1) 464 slotoffset = ndp->ni_offset; 465 if (slotfreespace >= slotneeded) { 466 slotstatus = COMPACT; 467 slotsize = ndp->ni_offset + 468 ep->d_reclen - slotoffset; 469 } 470 } 471 } 472 } 473 474 /* 475 * Check for a name match. 476 */ 477 if (ep->d_ino) { 478 if (ep->d_namlen == ndp->ni_dent.d_namlen && 479 !bcmp(ndp->ni_dent.d_name, ep->d_name, 480 ep->d_namlen)) 481 goto found; 482 } 483 prevoff = ndp->ni_offset; 484 ndp->ni_offset += ep->d_reclen; 485 entryoffsetinblock += ep->d_reclen; 486 } 487 /* notfound: */ 488 /* 489 * If we started in the middle of the directory and failed 490 * to find our target, we must check the beginning as well. 491 */ 492 if (numdirpasses == 2) { 493 numdirpasses--; 494 ndp->ni_offset = 0; 495 endsearch = u.u_ncache.nc_prevoffset; 496 goto searchloop; 497 } 498 /* 499 * If creating, and at end of pathname and current 500 * directory has not been removed, then can consider 501 * allowing file to be created. 502 */ 503 if (flag == CREATE && *cp == 0 && dp->i_nlink != 0) { 504 /* 505 * Access for write is interpreted as allowing 506 * creation of files in the directory. 507 */ 508 if (access(dp, IWRITE)) 509 goto bad; 510 /* 511 * Return an indication of where the new directory 512 * entry should be put. If we didn't find a slot, 513 * then set ndp->ni_count to 0 indicating that the new 514 * slot belongs at the end of the directory. If we found 515 * a slot, then the new entry can be put in the range 516 * [ndp->ni_offset .. ndp->ni_offset + ndp->ni_count) 517 */ 518 if (slotstatus == NONE) { 519 ndp->ni_offset = roundup(dp->i_size, DIRBLKSIZ); 520 ndp->ni_count = 0; 521 } else { 522 ndp->ni_offset = slotoffset; 523 ndp->ni_count = slotsize; 524 } 525 dp->i_flag |= IUPD|ICHG; 526 if (bp) 527 brelse(bp); 528 nbp->av_forw = freenamebuf; 529 freenamebuf = nbp; 530 /* 531 * We return with the directory locked, so that 532 * the parameters we set up above will still be 533 * valid if we actually decide to do a direnter(). 534 * We return NULL to indicate that the entry doesn't 535 * currently exist, leaving a pointer to the (locked) 536 * directory inode in ndp->ni_pdir. 537 */ 538 ndp->ni_pdir = dp; 539 return (NULL); 540 } 541 u.u_error = ENOENT; 542 goto bad; 543 found: 544 if (numdirpasses == 2) 545 nchstats.ncs_pass2++; 546 /* 547 * Check that directory length properly reflects presence 548 * of this entry. 549 */ 550 if (entryoffsetinblock + DIRSIZ(ep) > dp->i_size) { 551 dirbad(dp, ndp->ni_offset, "i_size too small"); 552 dp->i_size = entryoffsetinblock + DIRSIZ(ep); 553 dp->i_flag |= IUPD|ICHG; 554 } 555 556 /* 557 * Found component in pathname. 558 * If the final component of path name, save information 559 * in the cache as to where the entry was found. 560 */ 561 if (*cp == '\0' && flag == LOOKUP) { 562 u.u_ncache.nc_prevoffset = ndp->ni_offset; 563 u.u_ncache.nc_inumber = dp->i_number; 564 u.u_ncache.nc_dev = dp->i_dev; 565 u.u_ncache.nc_time = time.tv_sec; 566 } 567 /* 568 * Save directory entry in ndp->ni_dent, 569 * and release directory buffer. 570 */ 571 bcopy((caddr_t)ep, (caddr_t)&ndp->ni_dent, (u_int)DIRSIZ(ep)); 572 brelse(bp); 573 bp = NULL; 574 575 /* 576 * If deleting, and at end of pathname, return 577 * parameters which can be used to remove file. 578 * If the lockparent flag isn't set, we return only 579 * the directory (in ndp->ni_pdir), otherwise we go 580 * on and lock the inode, being careful with ".". 581 */ 582 if (flag == DELETE && *cp == 0) { 583 /* 584 * Write access to directory required to delete files. 585 */ 586 if (access(dp, IWRITE)) 587 goto bad; 588 ndp->ni_pdir = dp; /* for dirremove() */ 589 /* 590 * Return pointer to current entry in ndp->ni_offset, 591 * and distance past previous entry (if there 592 * is a previous entry in this block) in ndp->ni_count. 593 * Save directory inode pointer in ndp->ni_pdir for dirremove(). 594 */ 595 if ((ndp->ni_offset&(DIRBLKSIZ-1)) == 0) 596 ndp->ni_count = 0; 597 else 598 ndp->ni_count = ndp->ni_offset - prevoff; 599 if (lockparent) { 600 if (dp->i_number == ndp->ni_dent.d_ino) 601 dp->i_count++; 602 else { 603 dp = iget(dp->i_dev, fs, ndp->ni_dent.d_ino); 604 if (dp == NULL) { 605 iput(ndp->ni_pdir); 606 goto bad; 607 } 608 /* 609 * If directory is "sticky", then user must own 610 * the directory, or the file in it, else he 611 * may not delete it (unless he's root). This 612 * implements append-only directories. 613 */ 614 if ((ndp->ni_pdir->i_mode & ISVTX) && 615 u.u_uid != 0 && 616 u.u_uid != ndp->ni_pdir->i_uid && 617 dp->i_uid != u.u_uid) { 618 iput(ndp->ni_pdir); 619 u.u_error = EPERM; 620 goto bad; 621 } 622 } 623 } 624 nbp->av_forw = freenamebuf; 625 freenamebuf = nbp; 626 return (dp); 627 } 628 629 /* 630 * Special handling for ".." allowing chdir out of mounted 631 * file system: indirect .. in root inode to reevaluate 632 * in directory file system was mounted on. 633 */ 634 if (isdotdot) { 635 if (dp == u.u_rdir) 636 ndp->ni_dent.d_ino = dp->i_number; 637 else if (ndp->ni_dent.d_ino == ROOTINO && 638 dp->i_number == ROOTINO) { 639 for (i = 1; i < NMOUNT; i++) 640 if (mount[i].m_bufp != NULL && 641 mount[i].m_dev == dp->i_dev) { 642 iput(dp); 643 dp = mount[i].m_inodp; 644 ILOCK(dp); 645 dp->i_count++; 646 fs = dp->i_fs; 647 cp -= 2; /* back over .. */ 648 goto dirloop2; 649 } 650 } 651 } 652 653 /* 654 * If rewriting (rename), return the inode and the 655 * information required to rewrite the present directory 656 * Must get inode of directory entry to verify it's a 657 * regular file, or empty directory. 658 */ 659 if ((flag == CREATE && lockparent) && *cp == 0) { 660 if (access(dp, IWRITE)) 661 goto bad; 662 ndp->ni_pdir = dp; /* for dirrewrite() */ 663 /* 664 * Careful about locking second inode. 665 * This can only occur if the target is ".". 666 */ 667 if (dp->i_number == ndp->ni_dent.d_ino) { 668 u.u_error = EISDIR; /* XXX */ 669 goto bad; 670 } 671 dp = iget(dp->i_dev, fs, ndp->ni_dent.d_ino); 672 if (dp == NULL) { 673 iput(ndp->ni_pdir); 674 goto bad; 675 } 676 nbp->av_forw = freenamebuf; 677 freenamebuf = nbp; 678 return (dp); 679 } 680 681 /* 682 * Check for symbolic link, which may require us to massage the 683 * name before we continue translation. We do not `iput' the 684 * directory because we may need it again if the symbolic link 685 * is relative to the current directory. Instead we save it 686 * unlocked as "pdp". We must get the target inode before unlocking 687 * the directory to insure that the inode will not be removed 688 * before we get it. We prevent deadlock by always fetching 689 * inodes from the root, moving down the directory tree. Thus 690 * when following backward pointers ".." we must unlock the 691 * parent directory before getting the requested directory. 692 * There is a potential race condition here if both the current 693 * and parent directories are removed before the `iget' for the 694 * inode associated with ".." returns. We hope that this occurs 695 * infrequently since we cannot avoid this race condition without 696 * implementing a sophisticated deadlock detection algorithm. 697 * Note also that this simple deadlock detection scheme will not 698 * work if the file system has any hard links other than ".." 699 * that point backwards in the directory structure. 700 */ 701 pdp = dp; 702 if (isdotdot) { 703 IUNLOCK(pdp); /* race to get the inode */ 704 dp = iget(dp->i_dev, fs, ndp->ni_dent.d_ino); 705 if (dp == NULL) 706 goto bad2; 707 } else if (dp->i_number == ndp->ni_dent.d_ino) { 708 dp->i_count++; /* we want ourself, ie "." */ 709 } else { 710 dp = iget(dp->i_dev, fs, ndp->ni_dent.d_ino); 711 IUNLOCK(pdp); 712 if (dp == NULL) 713 goto bad2; 714 } 715 716 /* 717 * insert name into cache (if we want it, and it isn't "." or "..") 718 * 719 * all other cases where making a cache entry would be wrong 720 * have already departed from the code sequence somewhere above. 721 */ 722 if (docache) { 723 if (ncp != NULL) 724 panic("nami: duplicating cache"); 725 726 /* 727 * free the cache slot at head of lru chain 728 */ 729 if (ncp = nchhead) { 730 /* remove from lru chain */ 731 *ncp->nc_prev = ncp->nc_nxt; 732 if (ncp->nc_nxt) 733 ncp->nc_nxt->nc_prev = ncp->nc_prev; 734 else 735 nchtail = ncp->nc_prev; 736 737 /* remove from old hash chain */ 738 remque(ncp); 739 740 /* grab the inode we just found */ 741 ncp->nc_ip = dp; 742 743 /* fill in cache info */ 744 ncp->nc_ino = pdp->i_number; /* parents inum */ 745 ncp->nc_dev = pdp->i_dev; /* & device */ 746 ncp->nc_idev = dp->i_dev; /* our device */ 747 ncp->nc_id = dp->i_id; /* identifier */ 748 ncp->nc_nlen = ndp->ni_dent.d_namlen; 749 bcopy(ndp->ni_dent.d_name, ncp->nc_name, ncp->nc_nlen); 750 751 /* link at end of lru chain */ 752 ncp->nc_nxt = NULL; 753 ncp->nc_prev = nchtail; 754 *nchtail = ncp; 755 nchtail = &ncp->nc_nxt; 756 757 /* and insert on hash chain */ 758 insque(ncp, nhp); 759 } 760 } 761 762 haveino: 763 fs = dp->i_fs; 764 765 /* 766 * Check for symbolic link 767 */ 768 if ((dp->i_mode & IFMT) == IFLNK && 769 ((ndp->ni_nameiop & FOLLOW) || *cp == '/')) { 770 u_int pathlen = strlen(cp) + 1; 771 772 if (dp->i_size + pathlen >= MAXPATHLEN - 1 || 773 ++nlink > MAXSYMLINKS) { 774 u.u_error = ELOOP; 775 goto bad2; 776 } 777 ovbcopy(cp, nbp->b_un.b_addr + dp->i_size, pathlen); 778 u.u_error = 779 rdwri(UIO_READ, dp, nbp->b_un.b_addr, (int)dp->i_size, 780 0, 1, (int *)0); 781 if (u.u_error) 782 goto bad2; 783 cp = nbp->b_un.b_addr; 784 iput(dp); 785 if (*cp == '/') { 786 irele(pdp); 787 while (*cp == '/') 788 cp++; 789 if ((dp = u.u_rdir) == NULL) 790 dp = rootdir; 791 ILOCK(dp); 792 dp->i_count++; 793 } else { 794 dp = pdp; 795 ILOCK(dp); 796 } 797 fs = dp->i_fs; 798 goto dirloop; 799 } 800 801 /* 802 * Not a symbolic link. If more pathname, 803 * continue at next component, else return. 804 */ 805 if (*cp == '/') { 806 while (*cp == '/') 807 cp++; 808 irele(pdp); 809 goto dirloop; 810 } 811 nbp->av_forw = freenamebuf; 812 freenamebuf = nbp; 813 if (lockparent) 814 ndp->ni_pdir = pdp; 815 else 816 irele(pdp); 817 return (dp); 818 bad2: 819 irele(pdp); 820 bad: 821 if (bp) 822 brelse(bp); 823 if (dp) 824 iput(dp); 825 nbp->av_forw = freenamebuf; 826 freenamebuf = nbp; 827 return (NULL); 828 } 829 830 831 dirbad(ip, offset, how) 832 struct inode *ip; 833 off_t offset; 834 char *how; 835 { 836 837 printf("%s: bad dir ino %d at offset %d: %s\n", 838 ip->i_fs->fs_fsmnt, ip->i_number, offset, how); 839 } 840 841 /* 842 * Do consistency checking on a directory entry: 843 * record length must be multiple of 4 844 * record length must not be non-negative 845 * entry must fit in rest of its DIRBLKSIZ block 846 * record must be large enough to contain entry 847 * name is not longer than MAXNAMLEN 848 * name must be as long as advertised, and null terminated 849 */ 850 dirbadentry(ep, entryoffsetinblock) 851 register struct direct *ep; 852 int entryoffsetinblock; 853 { 854 register int i; 855 856 if ((ep->d_reclen & 0x3) != 0 || ep->d_reclen <= 0 || 857 ep->d_reclen > DIRBLKSIZ - (entryoffsetinblock & (DIRBLKSIZ - 1)) || 858 ep->d_reclen < DIRSIZ(ep) || ep->d_namlen > MAXNAMLEN) 859 return (1); 860 for (i = 0; i < ep->d_namlen; i++) 861 if (ep->d_name[i] == '\0') 862 return (1); 863 return (ep->d_name[i]); 864 } 865 866 /* 867 * Write a directory entry after a call to namei, using the parameters 868 * which it left in the u. area. The argument ip is the inode which 869 * the new directory entry will refer to. The u. area field ndp->ni_pdir is 870 * a pointer to the directory to be written, which was left locked by 871 * namei. Remaining parameters (ndp->ni_offset, ndp->ni_count) indicate 872 * how the space for the new entry is to be gotten. 873 */ 874 direnter(ip, ndp) 875 struct inode *ip; 876 register struct nameidata *ndp; 877 { 878 register struct direct *ep, *nep; 879 struct buf *bp; 880 int loc, spacefree, error = 0; 881 u_int dsize; 882 int newentrysize; 883 char *dirbuf; 884 885 ndp->ni_dent.d_ino = ip->i_number; 886 newentrysize = DIRSIZ(&ndp->ni_dent); 887 if (ndp->ni_count == 0) { 888 /* 889 * If ndp->ni_count is 0, then namei could find no space in the 890 * directory. In this case ndp->ni_offset will be on a directory 891 * block boundary and we will write the new entry into a fresh 892 * block. 893 */ 894 if (ndp->ni_offset&(DIRBLKSIZ-1)) 895 panic("wdir: newblk"); 896 ndp->ni_dent.d_reclen = DIRBLKSIZ; 897 error = rdwri(UIO_WRITE, ndp->ni_pdir, (caddr_t)&ndp->ni_dent, 898 newentrysize, ndp->ni_offset, 1, (int *)0); 899 iput(ndp->ni_pdir); 900 return (error); 901 } 902 903 /* 904 * If ndp->ni_count is non-zero, then namei found space for the new 905 * entry in the range ndp->ni_offset to ndp->ni_offset + ndp->ni_count. 906 * in the directory. To use this space, we may have to compact 907 * the entries located there, by copying them together towards 908 * the beginning of the block, leaving the free space in 909 * one usable chunk at the end. 910 */ 911 912 /* 913 * Increase size of directory if entry eats into new space. 914 * This should never push the size past a new multiple of 915 * DIRBLKSIZE. 916 */ 917 if (ndp->ni_offset + ndp->ni_count > ndp->ni_pdir->i_size) 918 ndp->ni_pdir->i_size = ndp->ni_offset + ndp->ni_count; 919 920 /* 921 * Get the block containing the space for the new directory 922 * entry. Should return error by result instead of u.u_error. 923 */ 924 bp = blkatoff(ndp->ni_pdir, ndp->ni_offset, (char **)&dirbuf); 925 if (bp == 0) { 926 iput(ndp->ni_pdir); 927 return (u.u_error); 928 } 929 930 /* 931 * Find space for the new entry. In the simple case, the 932 * entry at offset base will have the space. If it does 933 * not, then namei arranged that compacting the region 934 * ndp->ni_offset to ndp->ni_offset+ndp->ni_count would yield the space. 935 */ 936 ep = (struct direct *)dirbuf; 937 dsize = DIRSIZ(ep); 938 spacefree = ep->d_reclen - dsize; 939 for (loc = ep->d_reclen; loc < ndp->ni_count; ) { 940 nep = (struct direct *)(dirbuf + loc); 941 if (ep->d_ino) { 942 /* trim the existing slot */ 943 ep->d_reclen = dsize; 944 ep = (struct direct *)((char *)ep + dsize); 945 } else { 946 /* overwrite; nothing there; header is ours */ 947 spacefree += dsize; 948 } 949 dsize = DIRSIZ(nep); 950 spacefree += nep->d_reclen - dsize; 951 loc += nep->d_reclen; 952 bcopy((caddr_t)nep, (caddr_t)ep, dsize); 953 } 954 /* 955 * Update the pointer fields in the previous entry (if any), 956 * copy in the new entry, and write out the block. 957 */ 958 if (ep->d_ino == 0) { 959 if (spacefree + dsize < newentrysize) 960 panic("wdir: compact1"); 961 ndp->ni_dent.d_reclen = spacefree + dsize; 962 } else { 963 if (spacefree < newentrysize) 964 panic("wdir: compact2"); 965 ndp->ni_dent.d_reclen = spacefree; 966 ep->d_reclen = dsize; 967 ep = (struct direct *)((char *)ep + dsize); 968 } 969 bcopy((caddr_t)&ndp->ni_dent, (caddr_t)ep, (u_int)newentrysize); 970 bwrite(bp); 971 ndp->ni_pdir->i_flag |= IUPD|ICHG; 972 iput(ndp->ni_pdir); 973 return (error); 974 } 975 976 /* 977 * Remove a directory entry after a call to namei, using the 978 * parameters which it left in the u. area. The u. entry 979 * ni_offset contains the offset into the directory of the 980 * entry to be eliminated. The ni_count field contains the 981 * size of the previous record in the directory. If this 982 * is 0, the first entry is being deleted, so we need only 983 * zero the inode number to mark the entry as free. If the 984 * entry isn't the first in the directory, we must reclaim 985 * the space of the now empty record by adding the record size 986 * to the size of the previous entry. 987 */ 988 dirremove(ndp) 989 register struct nameidata *ndp; 990 { 991 register struct inode *dp = ndp->ni_pdir; 992 register struct buf *bp; 993 struct direct *ep; 994 995 if (ndp->ni_count == 0) { 996 /* 997 * First entry in block: set d_ino to zero. 998 */ 999 ndp->ni_dent.d_ino = 0; 1000 (void) rdwri(UIO_WRITE, dp, (caddr_t)&ndp->ni_dent, 1001 (int)DIRSIZ(&ndp->ni_dent), ndp->ni_offset, 1, (int *)0); 1002 } else { 1003 /* 1004 * Collapse new free space into previous entry. 1005 */ 1006 bp = blkatoff(dp, (int)(ndp->ni_offset - ndp->ni_count), 1007 (char **)&ep); 1008 if (bp == 0) 1009 return (0); 1010 ep->d_reclen += ndp->ni_dent.d_reclen; 1011 bwrite(bp); 1012 dp->i_flag |= IUPD|ICHG; 1013 } 1014 return (1); 1015 } 1016 1017 /* 1018 * Rewrite an existing directory entry to point at the inode 1019 * supplied. The parameters describing the directory entry are 1020 * set up by a call to namei. 1021 */ 1022 dirrewrite(dp, ip, ndp) 1023 struct inode *dp, *ip; 1024 struct nameidata *ndp; 1025 { 1026 1027 ndp->ni_dent.d_ino = ip->i_number; 1028 u.u_error = rdwri(UIO_WRITE, dp, (caddr_t)&ndp->ni_dent, 1029 (int)DIRSIZ(&ndp->ni_dent), ndp->ni_offset, 1, (int *)0); 1030 iput(dp); 1031 } 1032 1033 /* 1034 * Return buffer with contents of block "offset" 1035 * from the beginning of directory "ip". If "res" 1036 * is non-zero, fill it in with a pointer to the 1037 * remaining space in the directory. 1038 */ 1039 struct buf * 1040 blkatoff(ip, offset, res) 1041 struct inode *ip; 1042 off_t offset; 1043 char **res; 1044 { 1045 register struct fs *fs = ip->i_fs; 1046 daddr_t lbn = lblkno(fs, offset); 1047 int base = blkoff(fs, offset); 1048 int bsize = blksize(fs, ip, lbn); 1049 daddr_t bn = fsbtodb(fs, bmap(ip, lbn, B_WRITE, base, bsize)); 1050 register struct buf *bp; 1051 1052 if (u.u_error) 1053 return (0); 1054 bp = bread(ip->i_dev, bn, bsize); 1055 if (bp->b_flags & B_ERROR) { 1056 brelse(bp); 1057 return (0); 1058 } 1059 if (res) 1060 *res = bp->b_un.b_addr + base; 1061 return (bp); 1062 } 1063 1064 /* 1065 * Check if a directory is empty or not. 1066 * Inode supplied must be locked. 1067 * 1068 * Using a struct dirtemplate here is not precisely 1069 * what we want, but better than using a struct direct. 1070 * 1071 * NB: does not handle corrupted directories. 1072 */ 1073 dirempty(ip, parentino) 1074 register struct inode *ip; 1075 ino_t parentino; 1076 { 1077 register off_t off; 1078 struct dirtemplate dbuf; 1079 register struct direct *dp = (struct direct *)&dbuf; 1080 int error, count; 1081 #define MINDIRSIZ (sizeof (struct dirtemplate) / 2) 1082 1083 for (off = 0; off < ip->i_size; off += dp->d_reclen) { 1084 error = rdwri(UIO_READ, ip, (caddr_t)dp, MINDIRSIZ, 1085 off, 1, &count); 1086 /* 1087 * Since we read MINDIRSIZ, residual must 1088 * be 0 unless we're at end of file. 1089 */ 1090 if (error || count != 0) 1091 return (0); 1092 /* skip empty entries */ 1093 if (dp->d_ino == 0) 1094 continue; 1095 /* accept only "." and ".." */ 1096 if (dp->d_namlen > 2) 1097 return (0); 1098 if (dp->d_name[0] != '.') 1099 return (0); 1100 /* 1101 * At this point d_namlen must be 1 or 2. 1102 * 1 implies ".", 2 implies ".." if second 1103 * char is also "." 1104 */ 1105 if (dp->d_namlen == 1) 1106 continue; 1107 if (dp->d_name[1] == '.' && dp->d_ino == parentino) 1108 continue; 1109 return (0); 1110 } 1111 return (1); 1112 } 1113 1114 /* 1115 * Check if source directory is in the path of the target directory. 1116 * Target is supplied locked, source is unlocked. 1117 * The target is always iput() before returning. 1118 */ 1119 checkpath(source, target) 1120 struct inode *source, *target; 1121 { 1122 struct dirtemplate dirbuf; 1123 register struct inode *ip; 1124 int error = 0; 1125 1126 ip = target; 1127 if (ip->i_number == source->i_number) { 1128 error = EEXIST; 1129 goto out; 1130 } 1131 if (ip->i_number == ROOTINO) 1132 goto out; 1133 1134 for (;;) { 1135 if ((ip->i_mode&IFMT) != IFDIR) { 1136 error = ENOTDIR; 1137 break; 1138 } 1139 error = rdwri(UIO_READ, ip, (caddr_t)&dirbuf, 1140 sizeof (struct dirtemplate), (off_t)0, 1, (int *)0); 1141 if (error != 0) 1142 break; 1143 if (dirbuf.dotdot_namlen != 2 || 1144 dirbuf.dotdot_name[0] != '.' || 1145 dirbuf.dotdot_name[1] != '.') { 1146 error = ENOTDIR; 1147 break; 1148 } 1149 if (dirbuf.dotdot_ino == source->i_number) { 1150 error = EINVAL; 1151 break; 1152 } 1153 if (dirbuf.dotdot_ino == ROOTINO) 1154 break; 1155 iput(ip); 1156 ip = iget(ip->i_dev, ip->i_fs, dirbuf.dotdot_ino); 1157 if (ip == NULL) { 1158 error = u.u_error; 1159 break; 1160 } 1161 } 1162 1163 out: 1164 if (error == ENOTDIR) 1165 printf("checkpath: .. not a directory\n"); 1166 if (ip != NULL) 1167 iput(ip); 1168 return (error); 1169 } 1170 1171 /* 1172 * Name cache initialization, from main() when we are booting 1173 */ 1174 nchinit() 1175 { 1176 register union nchash *nchp; 1177 register struct nch *ncp; 1178 1179 nchhead = 0; 1180 nchtail = &nchhead; 1181 1182 for (ncp = nch; ncp < &nch[nchsize]; ncp++) { 1183 ncp->nc_forw = ncp; /* hash chain */ 1184 ncp->nc_back = ncp; 1185 1186 ncp->nc_nxt = NULL; /* lru chain */ 1187 *nchtail = ncp; 1188 ncp->nc_prev = nchtail; 1189 nchtail = &ncp->nc_nxt; 1190 1191 /* all else is zero already */ 1192 } 1193 1194 for (nchp = nchash; nchp < &nchash[NCHHASH]; nchp++) { 1195 nchp->nch_head[0] = nchp; 1196 nchp->nch_head[1] = nchp; 1197 } 1198 } 1199 1200 /* 1201 * Cache flush, called when filesys is umounted to 1202 * remove entries that would now be invalid 1203 * 1204 * The line "nxtcp = nchhead" near the end is to avoid potential problems 1205 * if the cache lru chain is modified while we are dumping the 1206 * inode. This makes the algorithm O(n^2), but do you think I care? 1207 */ 1208 nchinval(dev) 1209 register dev_t dev; 1210 { 1211 register struct nch *ncp, *nxtcp; 1212 1213 for (ncp = nchhead; ncp; ncp = nxtcp) { 1214 nxtcp = ncp->nc_nxt; 1215 1216 if (ncp->nc_ip == NULL || 1217 (ncp->nc_idev != dev && ncp->nc_dev != dev)) 1218 continue; 1219 1220 /* free the resources we had */ 1221 ncp->nc_idev = NODEV; 1222 ncp->nc_dev = NODEV; 1223 ncp->nc_id = NULL; 1224 ncp->nc_ino = 0; 1225 ncp->nc_ip = NULL; 1226 1227 1228 /* remove the entry from its hash chain */ 1229 remque(ncp); 1230 /* and make a dummy one */ 1231 ncp->nc_forw = ncp; 1232 ncp->nc_back = ncp; 1233 1234 /* delete this entry from LRU chain */ 1235 *ncp->nc_prev = nxtcp; 1236 if (nxtcp) 1237 nxtcp->nc_prev = ncp->nc_prev; 1238 else 1239 nchtail = ncp->nc_prev; 1240 1241 /* cause rescan of list, it may have altered */ 1242 nxtcp = nchhead; 1243 /* put the now-free entry at head of LRU */ 1244 ncp->nc_nxt = nxtcp; 1245 ncp->nc_prev = &nchhead; 1246 nxtcp->nc_prev = &ncp->nc_nxt; 1247 nchhead = ncp; 1248 } 1249 } 1250