Loading...
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 | /* * fs/dcache.c * * Complete reimplementation * (C) 1997 Thomas Schoebel-Theuer */ /* The new dcache is exclusively called from the VFS, not from * the specific fs'es any more. Despite having the same name as in the * old code, it has less to do with it. * * It serves many purposes: * * 1) Any inode that has been retrieved with lookup() and is in use * (i_count>0), has access to its full absolute path name, by going * to inode->i_dentry and then recursively following the entry->d_parent * chain. Use d_path() as predefined method for that. * You may find out the corresponding inode belonging to * a dentry by calling d_inode(). This can be used as an easy way for * determining .. and its absolute pathname, an old UNIX problem that * deserved a solution for a long time. * Note that hardlinked inodes may have multiple dentries assigned to * (via the d_next chain), reflecting multiple alias pathnames. * * 2) If not disabled by filesystem types specifying FS_NO_DCACHE, * the dentries of unused (aged) inodes are retained for speeding up * lookup()s, by allowing hashed inquiry starting from the dentry of * the parent directory. * * 3) It can remeber so-called "negative entries", that is dentries for * pathnames that are known to *not* exist, so unneccessary repeated * lookup()s for non-existant names can be saved. * * 4) It provides a means for keeping deleted files (inode->i_nlink==0) * accessible in the so-called *basket*. Inodes in the basket have been * removed with unlink() while being in use (i_count>0), so they would * normally use up space on the disk and be accessile through their * filedescriptor, but would not be accessible for lookup() any more. * The basket simply keeps such files in the dcache (for potential * dcache lookup) until they are either eventually removed completely, * or transferred to the second-level basket, the so-called *ibasket*. * The ibasket is implemented in the new inode code, on request of * filesystem types that have the flag FS_IBASKET set, and proliferates * the unlinked files when i_count has gone to zero, at least as long * as there is space on the disk and enough inodes remain available * and no umount() has started. * * 5) Preliminary dentries can be added by readdir(). While normal dentries * directly point to the inode via u.d_inode only the inode number is * known from readdir(), but not more. They can be converted to * normal dentries by using d_inode(). */ /* * Notes on the allocation strategy: * * The dcache is a full slave cache of the inodes. Whenever an inode * is cleared, all the dentries associated with it will recursively * disappear. dentries have no own reference counting; this has to * be obeyed for SMP. * If directories could go out of inode cache while * successors are alive, this would interrupt the d_parent chain of * the live successors. To prevent this without using zombies, all * directories are thus prevented from __iput() as long as successors * are alive. */ #include <linux/config.h> #include <linux/string.h> #include <linux/mm.h> #include <linux/fs.h> #include <linux/dalloc.h> #include <linux/dlists.h> /* this should be removed after the beta phase */ /* #define DEBUG */ /*#undef DEBUG*/ /* #define DEBUG_DDIR_COUNT */ #define D_HASHSIZE 64 /* local flags for d_flag */ #define D_DIR 32 #define D_HASHED 64 #define D_ZOMBIE 128 #define D_PRELIMINARY 256 #define D_INC_DDIR 512 /* local flags for d_del() */ #define D_RECURSIVE 4 #define D_NO_FREE 8 /* adjust these constants if you know a probability distribution ... */ #define D_SMALL 16 #define D_MEDIUM 64 #define D_LARGE 256 #define D_HUGE D_MAXLEN #define BASE_DHEADER(x) (struct dheader*)((unsigned long)(x) & ~(PAGE_SIZE-1)) #define BYTE_ADD(x,n) (void*)((char*)(x) + (n)) #define BYTE_SUB(x,n) (void*)((char*)(x) - (n)) /* This is for global allocation of dentries. Remove this when * converting to SLAB. */ struct dheader { struct dentry * emptylist; short free, maxfree; struct dheader * next; struct dheader * prev; }; struct anchors { struct dheader * free; /* each contains at least 1 empty dentry */ struct dheader * full; /* all the used up ones */ struct dheader * dir_free; struct dheader * dir_full; }; /* This is only used for directory dentries. Think of it as an extension * of the dentry. * It is defined as separate struct, so it uses up space only * where necessary. */ struct ddir { struct dentry * dd_hashtable[D_HASHSIZE]; struct dentry * dd_neglist; struct dentry * dd_basketlist; struct dentry * dd_zombielist; unsigned short dd_alloced; /* # d_alloc()ed, but not yet d_add()ed */ unsigned short dd_hashed; /* # of entries in hashtable */ unsigned short dd_true_hashed; /* # non-preliminaries in hashtable */ unsigned short dd_negs; /* # of negative entries */ }; DEF_INSERT(header,struct dheader,next,prev) DEF_REMOVE(header,struct dheader,next,prev) DEF_INSERT(alias,struct dentry,d_next,d_prev) DEF_REMOVE(alias,struct dentry,d_next,d_prev) DEF_INSERT(hash,struct dentry,d_hash_next,d_hash_prev) DEF_REMOVE(hash,struct dentry,d_hash_next,d_hash_prev) DEF_INSERT(basket,struct dentry,d_basket_next,d_basket_prev) DEF_REMOVE(basket,struct dentry,d_basket_next,d_basket_prev) static struct anchors anchors[4]; struct dentry * the_root = NULL; unsigned long name_cache_init(unsigned long mem_start, unsigned long mem_end) { memset(anchors, 0, sizeof(anchors)); return mem_start; } #ifdef DEBUG /* throw this away after the beta phase */ /*************************************************************************/ extern void xcheck(char * txt, struct inode * p); static int x_alloc = 0; static int x_freed = 0; static int x_free = 0; static void * tst[20000]; static int cnt = 0; static void ins(void* ptr) { extern int inodes_stat; tst[cnt++] = ptr; if(cnt % 1000 == 0) printk("------%d allocated: %d: %d %d %d\n", inodes_stat, cnt, x_alloc, x_freed, x_free); if(cnt>=20000) panic("stop"); } #if 0 static inline int search(void* ptr) { int i; for(i = cnt-1; i>=0; i--) if(tst[i] == ptr) return i; return -1; } #define TST(n,x) if(search(x)<0) printk("%s bad ptr %p line %d\n", n, x, __LINE__) #else #define TST(n,x) /*nothing*/ #endif void LOG(char * txt, struct dentry * entry) { static int count = 0; if(entry) { TST(txt,entry); } if(count) { count--; printk("%s: entry=%p\n", txt, entry); } } #ifdef DEBUG_DDIR_COUNT static struct ddir * d_dir(struct dentry * entry); void recursive_test(struct dentry * entry) { int i; struct ddir * ddir = d_dir(entry); int sons = 0; if(ddir->dd_zombielist) sons++; for(i=0; i < D_HASHSIZE; i++) { struct dentry ** base = &ddir->dd_hashtable[i]; struct dentry * tmp = *base; if(tmp) do { TST("__clear",tmp); if(!(tmp->d_flag & D_HASHED)) { printk("VFS: dcache entry not hashed!\n"); printpath(*base); printk("\n"); printpath(tmp); } if(!(tmp->d_flag & D_PRELIMINARY)) sons++; if(tmp->d_flag & D_DIR) recursive_test(tmp); tmp = tmp->d_hash_next; } while(tmp && tmp != *base); } if(!sons && !(entry->d_flag & D_PRELIMINARY) && entry->u.d_inode) { struct inode * inode = entry->u.d_inode; if(!atomic_read(&inode->i_count)) { if(!(inode->i_status & 1/*ST_AGED*/)) { printpath(entry); printk(" is not aged!\n"); } if(inode->i_ddir_count) { printpath(entry); printk(" has ddir_count blockage!\n"); } } } } #else #define recursive_test(e) /*nothing*/ #endif #else #define TST(n,x) /*nothing*/ #define LOG(n,x) /*nothing*/ #define xcheck(t,i) /*nothing*/ #define recursive_test(e) /*nothing*/ /*****************************************************************************/ #endif void printpath(struct dentry * entry) { if(!IS_ROOT(entry)) printpath(entry->d_parent); printk("/%s", entry->d_name); } static inline long has_sons(struct ddir * ddir) { return ((ddir->dd_alloced | ddir->dd_hashed) || ddir->dd_neglist || ddir->dd_basketlist || ddir->dd_zombielist); } static inline int has_true_sons(struct ddir * ddir) { return (ddir->dd_alloced | ddir->dd_true_hashed); } /* Only hold the i_ddir_count pseudo refcount when neccessary (i.e. when * they have true_sons), to prevent keeping too much dir inodes in use. */ static inline void inc_ddir(struct dentry * entry, struct inode * inode) { if(!(entry->d_flag & D_INC_DDIR)) { entry->d_flag |= D_INC_DDIR; #ifdef DEBUG if(inode->i_ddir_count) { printpath(entry); printk(" ddir_count=%d\n", inode->i_ddir_count); } #endif inode->i_ddir_count++; _get_inode(inode); } } static inline blocking void dec_ddir(struct dentry * entry, struct inode * inode) { if(entry->d_flag & D_INC_DDIR) { entry->d_flag &= ~D_INC_DDIR; inode->i_ddir_count--; if(!inode->i_ddir_count) __iput(inode); } } /* Do not inline this many times. */ static void d_panic(void) { panic("VFS: dcache directory corruption"); } static inline struct ddir * d_dir(struct dentry * entry) { struct ddir * res = BYTE_SUB(entry, sizeof(struct ddir)); if(!(entry->d_flag & D_DIR)) d_panic(); #ifdef DEBUG if(!entry) panic("entry NULL!"); if(BASE_DHEADER(res) != BASE_DHEADER(entry)) printk("Scheisse!!!\n"); #endif return res; } static /*inline*/ struct dheader * dinit(int isdir, int size) { struct dheader * res = (struct dheader*)__get_free_page(GFP_KERNEL); int restlen = PAGE_SIZE - sizeof(struct dheader); struct dentry * ptr = BYTE_ADD(res, sizeof(struct dheader)); if(!res) return NULL; memset(res, 0, sizeof(struct dheader)); if(isdir) { ptr = BYTE_ADD(ptr, sizeof(struct ddir)); size += sizeof(struct ddir); } if(BASE_DHEADER(ptr) != res) panic("Bad kernel page alignment"); size += sizeof(struct dentry) - D_MAXLEN; res->emptylist = NULL; res->free = 0; while(restlen >= size) { #ifdef DEBUG ins(ptr); if(BASE_DHEADER(ptr) != res) panic("Wrong dinit!"); #endif ptr->d_next = res->emptylist; res->emptylist = ptr; ptr = BYTE_ADD(ptr, size); res->free++; restlen -= size; } res->maxfree = res->free; return res; } static /*inline*/ struct dentry * __dalloc(struct anchors * anchor, struct dentry * parent, int isdir, int len, int size) { struct dheader ** free = isdir ? &anchor->dir_free : &anchor->free; struct dheader ** full = isdir ? &anchor->dir_full : &anchor->full; struct dheader * base = *free; struct dentry * res; if(!base) { base = dinit(isdir, size); if(!base) return NULL; insert_header(free, base); } base->free--; res = base->emptylist; if(!(base->emptylist = res->d_next)) { remove_header(free, base); insert_header(full, base); } memset(res, 0, sizeof(struct dentry) - D_MAXLEN); if(isdir) { res->d_flag = D_DIR; memset(d_dir(res), 0, sizeof(struct ddir)); } res->d_len = len; res->d_parent = parent; if(parent) { struct ddir * pdir = d_dir(parent); #ifdef DEBUG if(pdir->dd_alloced > 1 && !IS_ROOT(parent)) { printpath(parent); printk(" dd_alloced=%d\n", pdir->dd_alloced); } #endif pdir->dd_alloced++; } #ifdef DEBUG x_alloc++; #endif return res; } struct dentry * d_alloc(struct dentry * parent, int len, int isdir) { int i, size; #ifdef DEBUG if(the_root) recursive_test(the_root); LOG("d_alloc", parent); #endif if(len >= D_MEDIUM) { if(len >= D_LARGE) { i = 3; size = D_HUGE; } else { i = 2; size = D_LARGE; } } else if(len >= D_SMALL) { i = 1; size = D_MEDIUM; } else { i = 0; size = D_SMALL; } return __dalloc(&anchors[i], parent, isdir, len, size); } extern blocking struct dentry * d_alloc_root(struct inode * root_inode) { struct dentry * res = the_root; if(res) { d_del(res, D_NO_CLEAR_INODE); /* invalidate everything beyond */ } else { struct ddir * ddir; the_root = res = d_alloc(NULL, 0, 1); LOG("d_alloc_root", res); res->d_parent = res; res->d_name[0]='\0'; ddir = d_dir(res); ddir->dd_alloced = 999; /* protect from deletion */ } insert_alias(&root_inode->i_dentry, res); root_inode->i_dent_count++; root_inode->i_ddir_count++; res->u.d_inode = root_inode; return res; } static inline unsigned long d_hash(char first, char last) { return ((unsigned long)first ^ ((unsigned long)last << 4)) & (D_HASHSIZE-1); } static inline struct dentry ** d_base_entry(struct ddir * pdir, struct dentry * entry) { return &pdir->dd_hashtable[d_hash(entry->d_name[0], entry->d_name[entry->d_len-1])]; } static inline struct dentry ** d_base_qstr(struct ddir * pdir, struct qstr * s1, struct qstr * s2) { unsigned long hash; if(s2 && s2->len) { hash = d_hash(s1->name[0], s2->name[s2->len-1]); } else { hash = d_hash(s1->name[0], s1->name[s1->len-1]); } return &pdir->dd_hashtable[hash]; } static /*inline*/ blocking void recursive_clear(struct ddir * ddir, int flags) { int i, retry = 0; flags = (flags | D_RECURSIVE) & ~D_NO_CLEAR_INODE; again: /* Clear those separately that are not in the hashtable. */ while(ddir->dd_zombielist) d_del(ddir->dd_zombielist, flags); if(!ddir->dd_hashed) return; /* shortcut */ for(i=0; i < D_HASHSIZE; i++) { struct dentry ** base = &ddir->dd_hashtable[i]; struct dentry * tmp; while((tmp = *base)) { #ifdef DEBUG TST("__clear",tmp); if(!(tmp->d_flag & D_HASHED)) { printk("VFS: dcache entry not hashed!\n"); printpath(*base); printk("\n"); printpath(tmp); } #endif /* printk("["); */ d_del(tmp, flags); /* printk("]"); */ } } /* New entries may have been added during blocking. */ if(ddir->dd_hashed && ++retry < 10) goto again; #ifdef DEBUG if(ddir->dd_hashed || ddir->dd_true_hashed) { printk("remained %d/%d in hashtable\n", ddir->dd_true_hashed, ddir->dd_hashed); } #endif } static /*inline*/ blocking void _d_remove_from_parent(struct dentry * entry, struct ddir * pdir, struct inode * inode, int flags) { if(entry->d_flag & D_HASHED) { struct dentry ** base = d_base_entry(pdir, entry); remove_hash(base, entry); entry->d_flag &= ~D_HASHED; pdir->dd_hashed--; if(!(entry->d_flag & D_PRELIMINARY)) { pdir->dd_true_hashed--; if(!inode) { #ifdef DEBUG if(!entry->d_next || !entry->d_prev) { printpath(entry); printk(" flags=%x d_flag=%x negs=%d " "hashed=%d\n", flags, entry->d_flag, pdir->dd_negs, pdir->dd_hashed); } #endif remove_alias(&pdir->dd_neglist, entry); pdir->dd_negs--; } } } else if(!(entry->d_flag & D_ZOMBIE)) { #ifdef DEBUG if(!pdir->dd_alloced) printk("dd_alloced is 0!\n"); #endif pdir->dd_alloced--; } if(entry->d_flag & D_BASKET) { remove_basket(&pdir->dd_basketlist, entry); entry->d_flag &= ~D_BASKET; } } /* Theoretically, zombies should never or extremely seldom appear, * so this code is nearly superfluous. * A way to get zombies is while using inodes (i_count>0), unlink() * them as well as rmdir() the parent dir => the parent dir becomes a zombie. * Zombies are *not* in the hashtable, because somebody could re-creat() * that filename in it's parent dir again. * Besides coding errors during beta phase, when forcing an umount() * (e.g. at shutdown time), inodes could be in use such that the parent * dir is cleared, resulting also in zombies. */ static /*inline*/ void _d_handle_zombie(struct dentry * entry, struct ddir * ddir, struct ddir * pdir) { if(entry->d_flag & D_DIR) { if(entry->d_flag & D_ZOMBIE) { if(!has_sons(ddir)) { entry->d_flag &= ~D_ZOMBIE; remove_hash(&pdir->dd_zombielist, entry); if(!pdir->dd_zombielist && (entry->d_parent->d_flag & D_ZOMBIE)) { d_del(entry->d_parent, D_NORMAL); } } } else if(has_sons(ddir)) { entry->d_flag |= D_ZOMBIE; insert_hash(&pdir->dd_zombielist, entry); /* Not sure when this message should show up... */ if(!IS_ROOT(entry)) { printk("VFS: clearing dcache directory " "with successors\n"); #ifdef DEBUG printpath(entry); printk(" d_flag=%x alloced=%d negs=%d hashed=%d " "basket=%p zombies=%p\n", entry->d_flag, ddir->dd_alloced, ddir->dd_negs, ddir->dd_hashed, ddir->dd_basketlist, ddir->dd_zombielist); #endif } } } } static /*inline*/ blocking void _d_del(struct dentry * entry, struct anchors * anchor, int flags) { struct dheader ** free; struct dheader ** full; struct dheader * base = BASE_DHEADER(entry); struct ddir * ddir = NULL; struct ddir * pdir; struct inode * inode = entry->d_flag & D_PRELIMINARY ? NULL : entry->u.d_inode; #ifdef DEBUG if(inode) xcheck("_d_del", inode); #endif if(!entry->d_parent) { printk("VFS: dcache parent is NULL\n"); return; } if(entry->d_flag & D_DIR) { free = &anchor->dir_free; full = &anchor->dir_full; } else { free = &anchor->free; full = &anchor->full; } pdir = d_dir(entry->d_parent); if(!IS_ROOT(entry)) _d_remove_from_parent(entry, pdir, inode, flags); /* This may block, be careful! _d_remove_from_parent() is * thus called before. * * And there is also another problem, for FS_NO_DCACHE filesystems, * when a recursive_clear hits a directory and performs a dec_ddir() * on that parent (what 'entry' is right now) this could call __iput(). * This is ok for the most part, but FS_NO_DCACHE will cause a * redundant call to d_del() for what is now 'entry'. We do not want * this because this disturbs our state. So we use a special flag * to notify __iput() of this situation. -DaveM */ if(entry->d_flag & D_DIR) { entry->d_flag |= D_DDELIP; ddir = d_dir(entry); recursive_clear(ddir, flags); entry->d_flag &= ~D_DDELIP; } if(IS_ROOT(entry)) return; if(!has_true_sons(pdir)) dec_ddir(entry->d_parent, entry->d_parent->u.d_inode); if(flags & D_NO_FREE) { /* Make it re-d_add()able */ pdir->dd_alloced++; entry->d_flag &= D_DIR; } else _d_handle_zombie(entry, ddir, pdir); entry->u.d_inode = NULL; if(inode) { remove_alias(&inode->i_dentry, entry); inode->i_dent_count--; if (entry->d_flag & D_DIR) dec_ddir(entry, inode); if(!(flags & D_NO_CLEAR_INODE) && !(atomic_read(&inode->i_count) + inode->i_ddir_count + inode->i_dent_count)) { #ifdef DEBUG printk("#"); #endif /* This may block also. */ _clear_inode(inode, 0, 0); } } if(!(flags & D_NO_FREE) && !(entry->d_flag & D_ZOMBIE)) { base->free++; if(base->free == base->maxfree) { #ifndef DEBUG remove_header(free, base); free_page((unsigned long)base); goto done; #endif } entry->d_next = base->emptylist; base->emptylist = entry; if(!entry->d_next) { remove_header(full, base); insert_header(free, base); } #ifdef DEBUG x_freed++; #endif } #ifndef DEBUG done: #else x_free++; #endif } blocking void d_del(struct dentry * entry, int flags) { int i; if(!entry) return; LOG("d_clear", entry); if(entry->d_len >= D_MEDIUM) { if(entry->d_len >= D_LARGE) { i = 3; } else { i = 2; } } else if(entry->d_len >= D_SMALL) { i = 1; } else { i = 0; } _d_del(entry, &anchors[i], flags); } static inline struct dentry * __dlookup(struct dentry ** base, struct qstr * name, struct qstr * appendix) { struct dentry * tmp = *base; if(tmp && name->len) { int totallen = name->len; if(appendix) totallen += appendix->len; do { if(tmp->d_len == totallen && !(tmp->d_flag & D_DUPLICATE) && !strncmp(tmp->d_name, name->name, name->len) && (!appendix || !strncmp(tmp->d_name+name->len, appendix->name, appendix->len))) return tmp; tmp = tmp->d_hash_next; } while(tmp != *base); } return NULL; } struct dentry * d_lookup(struct inode * dir, struct qstr * name, struct qstr * appendix) { if(dir->i_dentry) { struct ddir * ddir = d_dir(dir->i_dentry); struct dentry ** base = d_base_qstr(ddir, name, appendix); return __dlookup(base, name, appendix); } return NULL; } static /*inline*/ blocking void _d_insert_to_parent(struct dentry * entry, struct ddir * pdir, struct inode * inode, struct qstr * ininame, int flags) { struct dentry ** base; struct dentry * parent = entry->d_parent; #ifdef DEBUG if(!pdir->dd_alloced) printk("dd_alloced is 0!\n"); #endif base = d_base_qstr(pdir, ininame, NULL); if(!(flags & (D_NOCHECKDUP|D_DUPLICATE)) && __dlookup(base, ininame, NULL)) { d_del(entry, D_NO_CLEAR_INODE); return; } if(entry->d_flag & D_HASHED) { printk("VFS: dcache entry is already hashed\n"); return; } if(!(flags & D_PRELIMINARY)) pdir->dd_true_hashed++; pdir->dd_hashed++; insert_hash(base, entry); entry->d_flag |= D_HASHED; pdir->dd_alloced--; if(flags & D_BASKET) insert_basket(&pdir->dd_basketlist, entry); #ifdef DEBUG if(inode && inode->i_dentry && (entry->d_flag & D_DIR)) { struct dentry * tmp = inode->i_dentry; printk("Auweia inode=%p entry=%p (%p %p %s)\n", inode, entry, parent->u.d_inode, parent, parent->d_name); printk("entry path="); printpath(entry); printk("\n"); do { TST("auweia",tmp); printk("alias path="); printpath(tmp); printk("\n"); tmp = tmp->d_next; } while(tmp != inode->i_dentry); printk("\n"); } #endif if(has_true_sons(pdir)) inc_ddir(parent, parent->u.d_inode); if(!inode && !(flags & D_PRELIMINARY)) { insert_alias(&pdir->dd_neglist, entry); pdir->dd_negs++; /* Don't allow the negative list to grow too much ... */ /* XXX There is a bad bug here. We remove the dentry from * XXX the alias list, but later on recursive_clear() and * XXX d_del() have no way of knowing this, they thus try * XXX to remove it again from the alias linked list and we * XXX get a crash. -DaveM */ #if 0 /* FIXME */ while(pdir->dd_negs > (pdir->dd_true_hashed >> 1) + 5) { struct dentry *removal = pdir->dd_neglist->d_prev; remove_alias(&pdir->dd_neglist, removal); pdir->dd_negs--; } #endif } } blocking void d_add(struct dentry * entry, struct inode * inode, struct qstr * ininame, int flags) { struct dentry * parent = entry->d_parent; struct qstr dummy; struct ddir * pdir; #ifdef DEBUG if(inode) xcheck("d_add", inode); if(IS_ROOT(entry)) { printk("VFS: d_add for root dentry "); printpath(entry); printk(" -> "); if(ininame) printk("%s", ininame->name); printk("\n"); return; } if(!parent) panic("d_add with parent==NULL"); LOG("d_add", entry); #endif if(ininame) { if(ininame->len != entry->d_len) { printk("VFS: d_add with wrong string length"); entry->d_len = ininame->len; /* kludge */ } memcpy(entry->d_name, ininame->name, ininame->len); entry->d_name[ininame->len] = '\0'; } else { dummy.name = entry->d_name; dummy.len = entry->d_len; ininame = &dummy; } if(entry->d_flag & D_HASHED) printk("VFS: d_add of already added dcache entry\n"); pdir = d_dir(parent); _d_insert_to_parent(entry, pdir, inode, ininame, flags); entry->d_flag |= flags; if(inode && !(flags & D_PRELIMINARY)) { if(entry->d_flag & D_DIR) { if(inode->i_dentry) { printk("VFS: creating dcache directory alias\n"); return; } } insert_alias(&inode->i_dentry, entry); inode->i_dent_count++; } entry->u.d_inode = inode; } blocking struct dentry * d_entry(struct dentry * parent, struct qstr * name, struct inode * inode) { struct ddir * pdir = d_dir(parent); struct dentry ** base = d_base_qstr(pdir, name, NULL); struct dentry * found = __dlookup(base, name, NULL); if(!found) { int isdir = (inode && S_ISDIR(inode->i_mode)); found = d_alloc(parent, name->len, isdir); if(found) { d_add(found, inode, name, isdir ? (D_DIR|D_NOCHECKDUP) : D_NOCHECKDUP); } else printk("VFS: problem with d_alloc\n"); } return found; } blocking void d_entry_preliminary(struct dentry * parent, struct qstr * name, unsigned long ino) { struct ddir * pdir = d_dir(parent); struct dentry ** base = d_base_qstr(pdir, name, NULL); struct dentry * found = __dlookup(base, name, NULL); if(!found && ino) { struct dentry * new = d_alloc(parent, name->len, 0); if(new) { d_add(new, NULL, name, D_PRELIMINARY|D_NOCHECKDUP); new->u.d_ino = ino; } else printk("VFS: problem with d_alloc\n"); } } blocking void d_move(struct dentry * entry, struct inode * newdir, struct qstr * newname, struct qstr * newapp) { struct ddir tmp; struct dentry * new; struct inode * inode; int len; int flags; if(!entry) return; inode = entry->u.d_inode; flags = entry->d_flag; if((flags & D_PRELIMINARY) || !inode) { if(!(flags & D_PRELIMINARY)) printk("VFS: trying to move negative dcache entry\n"); d_del(entry, D_NO_CLEAR_INODE); return; } #if 0 printk("d_move %p '%s' -> '%s%s' dent_count=%d\n", inode, entry->d_name, newname->name, newapp ? newapp->name : "", inode->i_dent_count); #endif if(flags & D_ZOMBIE) { printk("VFS: moving zombie entry\n"); } if(flags & D_DIR) { struct ddir * ddir = d_dir(entry); memcpy(&tmp, ddir, sizeof(struct ddir)); /* Simulate empty dir for d_del(). */ memset(ddir, 0, sizeof(struct ddir)); } len = newname->len; if(newapp) { len += newapp->len; flags |= D_BASKET; } else flags &= ~D_BASKET; new = d_alloc(newdir->i_dentry, len, flags & D_DIR); memcpy(new->d_name, newname->name, newname->len); if(newapp) memcpy(new->d_name+newname->len, newapp->name, newapp->len); new->d_name[len] = '\0'; d_del(entry, D_NO_CLEAR_INODE); d_add(new, inode, NULL, flags & (D_DIR|D_BASKET)); if(flags & D_DIR) { struct ddir * ddir = d_dir(new); memcpy(ddir, &tmp, sizeof(struct ddir)); } } int d_path(struct dentry * entry, struct inode * chroot, char * buf) { if(IS_ROOT(entry) || (chroot && entry->u.d_inode == chroot && !(entry->d_flag & D_PRELIMINARY))) { *buf = '/'; return 1; } else { int len = d_path(entry->d_parent, chroot, buf); buf += len; if(len > 1) { *buf++ = '/'; len++; } memcpy(buf, entry->d_name, entry->d_len); return len + entry->d_len; } } struct dentry * d_basket(struct dentry * dir_entry) { if(dir_entry && (dir_entry->d_flag & D_DIR)) { struct ddir * ddir = d_dir(dir_entry); return ddir->dd_basketlist; } else return NULL; } int d_isbasket(struct dentry * entry) { return entry->d_flag & D_BASKET; } blocking struct inode * d_inode(struct dentry ** changing_entry) { struct dentry * entry = *changing_entry; struct inode * inode; #ifdef CONFIG_DCACHE_PRELOAD if(entry->d_flag & D_PRELIMINARY) { struct qstr name = { entry->d_name, entry->d_len }; struct ddir * pdir = d_dir(entry->d_parent); struct dentry ** base = d_base_qstr(pdir, &name, NULL); struct dentry * found; unsigned long ino; struct inode * dir = entry->d_parent->u.d_inode; TST("d_inode",entry); ino = entry->u.d_ino; if(!dir) d_panic(); /* Prevent concurrent d_lookup()s or d_inode()s before * giving up vfs_lock. This just removes from the parent, * but does not deallocate it. */ /* !!!!!!! Aiee, here is an unresolved race if somebody * unlink()s the inode during the iget(). The problem is * that we need to synchronize externally. Proposed solution: * put a rw_lock (read-mode) on the parent dir for each * iget(), lookup() and so on, and a write-mode lock for * everything that changes the dir (e.g. unlink()), and do * this consistently everywhere in the generic VFS (not in * the concrete filesystems). This should kill similar * races everywhere, with a single clean concept. * Later, the synchronization stuff can be cleaned out * of the concrete fs'es. */ d_del(entry, D_NO_CLEAR_INODE|D_NO_FREE); vfs_unlock(); /* This circumvents the normal lookup() of pathnames. * Therefore, preliminary entries must not be used * (see FS_NO_DCACHE and FS_NO_PRELIM) if the fs does not * permit fetching *valid* inodes with plain iget(). */ inode = __iget(dir->i_sb, ino, 0); vfs_lock(); if(!inode) { printk("VFS: preliminary dcache entry was invalid\n"); *changing_entry = NULL; return NULL; } xcheck("d_inode iget()", inode); if((found = __dlookup(base, &name, NULL))) { d_del(entry, D_NO_CLEAR_INODE); *changing_entry = found; } else if(S_ISDIR(inode->i_mode)) { struct dentry * new = d_alloc(entry->d_parent, entry->d_len, 1); if(new) d_add(new, inode, &name, D_DIR); *changing_entry = new; /* Finally deallocate old entry. */ d_del(entry, D_NO_CLEAR_INODE); } else { /* Re-insert to the parent, but now as normal dentry. */ d_add(entry, inode, NULL, 0); } return inode; } #endif inode = entry->u.d_inode; if(inode) { #ifdef DEBUG xcheck("d_inode", inode); #endif iinc_zero(inode); } return inode; } |