1 /* $OpenBSD: mansearch.c,v 1.50 2016/07/09 15:23:36 schwarze Exp $ */ 2 /* 3 * Copyright (c) 2012 Kristaps Dzonsons <kristaps@bsd.lv> 4 * Copyright (c) 2013, 2014, 2015, 2016 Ingo Schwarze <schwarze@openbsd.org> 5 * 6 * Permission to use, copy, modify, and distribute this software for any 7 * purpose with or without fee is hereby granted, provided that the above 8 * copyright notice and this permission notice appear in all copies. 9 * 10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES 11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR 13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 17 */ 18 #include "config.h" 19 20 #include <sys/mman.h> 21 #include <sys/types.h> 22 23 #include <assert.h> 24 #if HAVE_ERR 25 #include <err.h> 26 #endif 27 #include <errno.h> 28 #include <fcntl.h> 29 #include <glob.h> 30 #include <limits.h> 31 #include <regex.h> 32 #include <stdio.h> 33 #include <stdint.h> 34 #include <stddef.h> 35 #include <stdlib.h> 36 #include <string.h> 37 #include <unistd.h> 38 39 #include "mandoc.h" 40 #include "mandoc_aux.h" 41 #include "mandoc_ohash.h" 42 #include "manconf.h" 43 #include "mansearch.h" 44 #include "dbm.h" 45 46 struct expr { 47 /* Used for terms: */ 48 struct dbm_match match; /* Match type and expression. */ 49 uint64_t bits; /* Type mask. */ 50 /* Used for OR and AND groups: */ 51 struct expr *next; /* Next child in the parent group. */ 52 struct expr *child; /* First child in this group. */ 53 enum { EXPR_TERM, EXPR_OR, EXPR_AND } type; 54 }; 55 56 const char *const mansearch_keynames[KEY_MAX] = { 57 "arch", "sec", "Xr", "Ar", "Fa", "Fl", "Dv", "Fn", 58 "Ic", "Pa", "Cm", "Li", "Em", "Cd", "Va", "Ft", 59 "Tn", "Er", "Ev", "Sy", "Sh", "In", "Ss", "Ox", 60 "An", "Mt", "St", "Bx", "At", "Nx", "Fx", "Lk", 61 "Ms", "Bsx", "Dx", "Rs", "Vt", "Lb", "Nm", "Nd" 62 }; 63 64 65 static struct ohash *manmerge(struct expr *, struct ohash *); 66 static struct ohash *manmerge_term(struct expr *, struct ohash *); 67 static struct ohash *manmerge_or(struct expr *, struct ohash *); 68 static struct ohash *manmerge_and(struct expr *, struct ohash *); 69 static char *buildnames(const struct dbm_page *); 70 static char *buildoutput(size_t, int32_t); 71 static size_t lstlen(const char *); 72 static void lstcat(char *, size_t *, const char *); 73 static int lstmatch(const char *, const char *); 74 static struct expr *exprcomp(const struct mansearch *, 75 int, char *[], int *); 76 static struct expr *expr_and(const struct mansearch *, 77 int, char *[], int *); 78 static struct expr *exprterm(const struct mansearch *, 79 int, char *[], int *); 80 static void exprfree(struct expr *); 81 static int manpage_compare(const void *, const void *); 82 83 84 int 85 mansearch(const struct mansearch *search, 86 const struct manpaths *paths, 87 int argc, char *argv[], 88 struct manpage **res, size_t *sz) 89 { 90 char buf[PATH_MAX]; 91 struct dbm_res *rp; 92 struct expr *e; 93 struct dbm_page *page; 94 struct manpage *mpage; 95 struct ohash *htab; 96 size_t cur, i, maxres, outkey; 97 unsigned int slot; 98 int argi, chdir_status, getcwd_status, im; 99 100 argi = 0; 101 if ((e = exprcomp(search, argc, argv, &argi)) == NULL) { 102 *sz = 0; 103 return 0; 104 } 105 106 cur = maxres = 0; 107 *res = NULL; 108 109 outkey = KEY_Nd; 110 if (search->outkey != NULL) 111 for (im = 0; im < KEY_MAX; im++) 112 if (0 == strcasecmp(search->outkey, 113 mansearch_keynames[im])) { 114 outkey = im; 115 break; 116 } 117 118 /* 119 * Remember the original working directory, if possible. 120 * This will be needed if the second or a later directory 121 * is given as a relative path. 122 * Do not error out if the current directory is not 123 * searchable: Maybe it won't be needed after all. 124 */ 125 126 if (getcwd(buf, PATH_MAX) == NULL) { 127 getcwd_status = 0; 128 (void)strlcpy(buf, strerror(errno), sizeof(buf)); 129 } else 130 getcwd_status = 1; 131 132 /* 133 * Loop over the directories (containing databases) for us to 134 * search. 135 * Don't let missing/bad databases/directories phase us. 136 * In each, try to open the resident database and, if it opens, 137 * scan it for our match expression. 138 */ 139 140 chdir_status = 0; 141 for (i = 0; i < paths->sz; i++) { 142 if (chdir_status && paths->paths[i][0] != '/') { 143 if ( ! getcwd_status) { 144 warnx("%s: getcwd: %s", paths->paths[i], buf); 145 continue; 146 } else if (chdir(buf) == -1) { 147 warn("%s", buf); 148 continue; 149 } 150 } 151 if (chdir(paths->paths[i]) == -1) { 152 warn("%s", paths->paths[i]); 153 continue; 154 } 155 chdir_status = 1; 156 157 if (dbm_open(MANDOC_DB) == -1) { 158 warn("%s/%s", paths->paths[i], MANDOC_DB); 159 continue; 160 } 161 162 if ((htab = manmerge(e, NULL)) == NULL) { 163 dbm_close(); 164 continue; 165 } 166 167 for (rp = ohash_first(htab, &slot); rp != NULL; 168 rp = ohash_next(htab, &slot)) { 169 page = dbm_page_get(rp->page); 170 171 if (lstmatch(search->sec, page->sect) == 0 || 172 lstmatch(search->arch, page->arch) == 0) 173 continue; 174 175 if (cur + 1 > maxres) { 176 maxres += 1024; 177 *res = mandoc_reallocarray(*res, 178 maxres, sizeof(**res)); 179 } 180 mpage = *res + cur; 181 mandoc_asprintf(&mpage->file, "%s/%s", 182 paths->paths[i], page->file + 1); 183 mpage->names = buildnames(page); 184 mpage->output = (int)outkey == KEY_Nd ? 185 mandoc_strdup(page->desc) : 186 buildoutput(outkey, page->addr); 187 mpage->ipath = i; 188 mpage->bits = rp->bits; 189 mpage->sec = *page->sect - '0'; 190 if (mpage->sec < 0 || mpage->sec > 9) 191 mpage->sec = 10; 192 mpage->form = *page->file; 193 free(rp); 194 cur++; 195 } 196 ohash_delete(htab); 197 free(htab); 198 dbm_close(); 199 200 /* 201 * In man(1) mode, prefer matches in earlier trees 202 * over matches in later trees. 203 */ 204 205 if (cur && search->firstmatch) 206 break; 207 } 208 qsort(*res, cur, sizeof(struct manpage), manpage_compare); 209 if (chdir_status && getcwd_status && chdir(buf) == -1) 210 warn("%s", buf); 211 exprfree(e); 212 *sz = cur; 213 return 1; 214 } 215 216 /* 217 * Merge the results for the expression tree rooted at e 218 * into the the result list htab. 219 */ 220 static struct ohash * 221 manmerge(struct expr *e, struct ohash *htab) 222 { 223 switch (e->type) { 224 case EXPR_TERM: 225 return manmerge_term(e, htab); 226 case EXPR_OR: 227 return manmerge_or(e->child, htab); 228 case EXPR_AND: 229 return manmerge_and(e->child, htab); 230 default: 231 abort(); 232 } 233 } 234 235 static struct ohash * 236 manmerge_term(struct expr *e, struct ohash *htab) 237 { 238 struct dbm_res res, *rp; 239 uint64_t ib; 240 unsigned int slot; 241 int im; 242 243 if (htab == NULL) { 244 htab = mandoc_malloc(sizeof(*htab)); 245 mandoc_ohash_init(htab, 4, offsetof(struct dbm_res, page)); 246 } 247 248 for (im = 0, ib = 1; im < KEY_MAX; im++, ib <<= 1) { 249 if ((e->bits & ib) == 0) 250 continue; 251 252 switch (ib) { 253 case TYPE_arch: 254 dbm_page_byarch(&e->match); 255 break; 256 case TYPE_sec: 257 dbm_page_bysect(&e->match); 258 break; 259 case TYPE_Nm: 260 dbm_page_byname(&e->match); 261 break; 262 case TYPE_Nd: 263 dbm_page_bydesc(&e->match); 264 break; 265 default: 266 dbm_page_bymacro(im - 2, &e->match); 267 break; 268 } 269 270 /* 271 * When hashing for deduplication, use the unique 272 * page ID itself instead of a hash function; 273 * that is quite efficient. 274 */ 275 276 for (;;) { 277 res = dbm_page_next(); 278 if (res.page == -1) 279 break; 280 slot = ohash_lookup_memory(htab, 281 (char *)&res, sizeof(res.page), res.page); 282 if ((rp = ohash_find(htab, slot)) != NULL) { 283 rp->bits |= res.bits; 284 continue; 285 } 286 rp = mandoc_malloc(sizeof(*rp)); 287 *rp = res; 288 ohash_insert(htab, slot, rp); 289 } 290 } 291 return htab; 292 } 293 294 static struct ohash * 295 manmerge_or(struct expr *e, struct ohash *htab) 296 { 297 while (e != NULL) { 298 htab = manmerge(e, htab); 299 e = e->next; 300 } 301 return htab; 302 } 303 304 static struct ohash * 305 manmerge_and(struct expr *e, struct ohash *htab) 306 { 307 struct ohash *hand, *h1, *h2; 308 struct dbm_res *res; 309 unsigned int slot1, slot2; 310 311 /* Evaluate the first term of the AND clause. */ 312 313 hand = manmerge(e, NULL); 314 315 while ((e = e->next) != NULL) { 316 317 /* Evaluate the next term and prepare for ANDing. */ 318 319 h2 = manmerge(e, NULL); 320 if (ohash_entries(h2) < ohash_entries(hand)) { 321 h1 = h2; 322 h2 = hand; 323 } else 324 h1 = hand; 325 hand = mandoc_malloc(sizeof(*hand)); 326 mandoc_ohash_init(hand, 4, offsetof(struct dbm_res, page)); 327 328 /* Keep all pages that are in both result sets. */ 329 330 for (res = ohash_first(h1, &slot1); res != NULL; 331 res = ohash_next(h1, &slot1)) { 332 if (ohash_find(h2, ohash_lookup_memory(h2, 333 (char *)res, sizeof(res->page), 334 res->page)) == NULL) 335 free(res); 336 else 337 ohash_insert(hand, ohash_lookup_memory(hand, 338 (char *)res, sizeof(res->page), 339 res->page), res); 340 } 341 342 /* Discard the merged results. */ 343 344 for (res = ohash_first(h2, &slot2); res != NULL; 345 res = ohash_next(h2, &slot2)) 346 free(res); 347 ohash_delete(h2); 348 free(h2); 349 ohash_delete(h1); 350 free(h1); 351 } 352 353 /* Merge the result of the AND into htab. */ 354 355 if (htab == NULL) 356 return hand; 357 358 for (res = ohash_first(hand, &slot1); res != NULL; 359 res = ohash_next(hand, &slot1)) { 360 slot2 = ohash_lookup_memory(htab, 361 (char *)res, sizeof(res->page), res->page); 362 if (ohash_find(htab, slot2) == NULL) 363 ohash_insert(htab, slot2, res); 364 else 365 free(res); 366 } 367 368 /* Discard the merged result. */ 369 370 ohash_delete(hand); 371 free(hand); 372 return htab; 373 } 374 375 void 376 mansearch_free(struct manpage *res, size_t sz) 377 { 378 size_t i; 379 380 for (i = 0; i < sz; i++) { 381 free(res[i].file); 382 free(res[i].names); 383 free(res[i].output); 384 } 385 free(res); 386 } 387 388 static int 389 manpage_compare(const void *vp1, const void *vp2) 390 { 391 const struct manpage *mp1, *mp2; 392 int diff; 393 394 mp1 = vp1; 395 mp2 = vp2; 396 return (diff = mp2->bits - mp1->bits) ? diff : 397 (diff = mp1->sec - mp2->sec) ? diff : 398 strcasecmp(mp1->names, mp2->names); 399 } 400 401 static char * 402 buildnames(const struct dbm_page *page) 403 { 404 char *buf; 405 size_t i, sz; 406 407 sz = lstlen(page->name) + 1 + lstlen(page->sect) + 408 (page->arch == NULL ? 0 : 1 + lstlen(page->arch)) + 2; 409 buf = mandoc_malloc(sz); 410 i = 0; 411 lstcat(buf, &i, page->name); 412 buf[i++] = '('; 413 lstcat(buf, &i, page->sect); 414 if (page->arch != NULL) { 415 buf[i++] = '/'; 416 lstcat(buf, &i, page->arch); 417 } 418 buf[i++] = ')'; 419 buf[i++] = '\0'; 420 assert(i == sz); 421 return buf; 422 } 423 424 /* 425 * Count the buffer space needed to print the NUL-terminated 426 * list of NUL-terminated strings, when printing two separator 427 * characters between strings. 428 */ 429 static size_t 430 lstlen(const char *cp) 431 { 432 size_t sz; 433 434 for (sz = 0;; sz++) { 435 if (cp[0] == '\0') { 436 if (cp[1] == '\0') 437 break; 438 sz++; 439 } else if (cp[0] < ' ') 440 sz--; 441 cp++; 442 } 443 return sz; 444 } 445 446 /* 447 * Print the NUL-terminated list of NUL-terminated strings 448 * into the buffer, seperating strings with a comma and a blank. 449 */ 450 static void 451 lstcat(char *buf, size_t *i, const char *cp) 452 { 453 for (;;) { 454 if (cp[0] == '\0') { 455 if (cp[1] == '\0') 456 break; 457 buf[(*i)++] = ','; 458 buf[(*i)++] = ' '; 459 } else if (cp[0] >= ' ') 460 buf[(*i)++] = cp[0]; 461 cp++; 462 } 463 } 464 465 /* 466 * Return 1 if the string *want occurs in any of the strings 467 * in the NUL-terminated string list *have, or 0 otherwise. 468 * If either argument is NULL or empty, assume no filtering 469 * is desired and return 1. 470 */ 471 static int 472 lstmatch(const char *want, const char *have) 473 { 474 if (want == NULL || have == NULL || *have == '\0') 475 return 1; 476 while (*have != '\0') { 477 if (strcasestr(have, want) != NULL) 478 return 1; 479 have = strchr(have, '\0') + 1; 480 } 481 return 0; 482 } 483 484 /* 485 * Build a list of values taken by the macro im 486 * in the manual page with big-endian address addr. 487 */ 488 static char * 489 buildoutput(size_t im, int32_t addr) 490 { 491 const char *oldoutput, *sep; 492 char *output, *newoutput, *value; 493 494 output = NULL; 495 dbm_macro_bypage(im - 2, addr); 496 while ((value = dbm_macro_next()) != NULL) { 497 if (output == NULL) { 498 oldoutput = ""; 499 sep = ""; 500 } else { 501 oldoutput = output; 502 sep = " # "; 503 } 504 mandoc_asprintf(&newoutput, "%s%s%s", oldoutput, sep, value); 505 free(output); 506 output = newoutput; 507 } 508 return output; 509 } 510 511 /* 512 * Compile a set of string tokens into an expression. 513 * Tokens in "argv" are assumed to be individual expression atoms (e.g., 514 * "(", "foo=bar", etc.). 515 */ 516 static struct expr * 517 exprcomp(const struct mansearch *search, int argc, char *argv[], int *argi) 518 { 519 struct expr *parent, *child; 520 int needterm, nested; 521 522 if ((nested = *argi) == argc) 523 return NULL; 524 needterm = 1; 525 parent = child = NULL; 526 while (*argi < argc) { 527 if (strcmp(")", argv[*argi]) == 0) { 528 if (needterm) 529 warnx("missing term " 530 "before closing parenthesis"); 531 needterm = 0; 532 if (nested) 533 break; 534 warnx("ignoring unmatched right parenthesis"); 535 ++*argi; 536 continue; 537 } 538 if (strcmp("-o", argv[*argi]) == 0) { 539 if (needterm) { 540 if (*argi > 0) 541 warnx("ignoring -o after %s", 542 argv[*argi - 1]); 543 else 544 warnx("ignoring initial -o"); 545 } 546 needterm = 1; 547 ++*argi; 548 continue; 549 } 550 needterm = 0; 551 if (child == NULL) { 552 child = expr_and(search, argc, argv, argi); 553 continue; 554 } 555 if (parent == NULL) { 556 parent = mandoc_calloc(1, sizeof(*parent)); 557 parent->type = EXPR_OR; 558 parent->next = NULL; 559 parent->child = child; 560 } 561 child->next = expr_and(search, argc, argv, argi); 562 child = child->next; 563 } 564 if (needterm && *argi) 565 warnx("ignoring trailing %s", argv[*argi - 1]); 566 return parent == NULL ? child : parent; 567 } 568 569 static struct expr * 570 expr_and(const struct mansearch *search, int argc, char *argv[], int *argi) 571 { 572 struct expr *parent, *child; 573 int needterm; 574 575 needterm = 1; 576 parent = child = NULL; 577 while (*argi < argc) { 578 if (strcmp(")", argv[*argi]) == 0) { 579 if (needterm) 580 warnx("missing term " 581 "before closing parenthesis"); 582 needterm = 0; 583 break; 584 } 585 if (strcmp("-o", argv[*argi]) == 0) 586 break; 587 if (strcmp("-a", argv[*argi]) == 0) { 588 if (needterm) { 589 if (*argi > 0) 590 warnx("ignoring -a after %s", 591 argv[*argi - 1]); 592 else 593 warnx("ignoring initial -a"); 594 } 595 needterm = 1; 596 ++*argi; 597 continue; 598 } 599 if (needterm == 0) 600 break; 601 if (child == NULL) { 602 child = exprterm(search, argc, argv, argi); 603 if (child != NULL) 604 needterm = 0; 605 continue; 606 } 607 needterm = 0; 608 if (parent == NULL) { 609 parent = mandoc_calloc(1, sizeof(*parent)); 610 parent->type = EXPR_AND; 611 parent->next = NULL; 612 parent->child = child; 613 } 614 child->next = exprterm(search, argc, argv, argi); 615 if (child->next != NULL) { 616 child = child->next; 617 needterm = 0; 618 } 619 } 620 if (needterm && *argi) 621 warnx("ignoring trailing %s", argv[*argi - 1]); 622 return parent == NULL ? child : parent; 623 } 624 625 static struct expr * 626 exprterm(const struct mansearch *search, int argc, char *argv[], int *argi) 627 { 628 char errbuf[BUFSIZ]; 629 struct expr *e; 630 char *key, *val; 631 uint64_t iterbit; 632 int cs, i, irc; 633 634 if (strcmp("(", argv[*argi]) == 0) { 635 ++*argi; 636 e = exprcomp(search, argc, argv, argi); 637 if (*argi < argc) { 638 assert(strcmp(")", argv[*argi]) == 0); 639 ++*argi; 640 } else 641 warnx("unclosed parenthesis"); 642 return e; 643 } 644 645 e = mandoc_calloc(1, sizeof(*e)); 646 e->type = EXPR_TERM; 647 e->bits = 0; 648 e->next = NULL; 649 e->child = NULL; 650 651 if (search->argmode == ARG_NAME) { 652 e->bits = TYPE_Nm; 653 e->match.type = DBM_EXACT; 654 e->match.str = argv[(*argi)++]; 655 return e; 656 } 657 658 /* 659 * Separate macro keys from search string. 660 * If needed, request regular expression handling. 661 */ 662 663 if (search->argmode == ARG_WORD) { 664 e->bits = TYPE_Nm; 665 e->match.type = DBM_REGEX; 666 #if HAVE_REWB_BSD 667 mandoc_asprintf(&val, "[[:<:]]%s[[:>:]]", argv[*argi]); 668 #elif HAVE_REWB_SYSV 669 mandoc_asprintf(&val, "\\<%s\\>", argv[*argi]); 670 #else 671 mandoc_asprintf(&val, 672 "(^|[^a-zA-Z01-9_])%s([^a-zA-Z01-9_]|$)", argv[*argi]); 673 #endif 674 cs = 0; 675 } else if ((val = strpbrk(argv[*argi], "=~")) == NULL) { 676 e->bits = TYPE_Nm | TYPE_Nd; 677 e->match.type = DBM_SUB; 678 e->match.str = argv[*argi]; 679 } else { 680 if (val == argv[*argi]) 681 e->bits = TYPE_Nm | TYPE_Nd; 682 if (*val == '=') { 683 e->match.type = DBM_SUB; 684 e->match.str = val + 1; 685 } else 686 e->match.type = DBM_REGEX; 687 *val++ = '\0'; 688 if (strstr(argv[*argi], "arch") != NULL) 689 cs = 0; 690 } 691 692 /* Compile regular expressions. */ 693 694 if (e->match.type == DBM_REGEX) { 695 e->match.re = mandoc_malloc(sizeof(*e->match.re)); 696 irc = regcomp(e->match.re, val, 697 REG_EXTENDED | REG_NOSUB | (cs ? 0 : REG_ICASE)); 698 if (irc) { 699 regerror(irc, e->match.re, errbuf, sizeof(errbuf)); 700 warnx("regcomp /%s/: %s", val, errbuf); 701 } 702 if (search->argmode == ARG_WORD) 703 free(val); 704 if (irc) { 705 free(e->match.re); 706 free(e); 707 ++*argi; 708 return NULL; 709 } 710 } 711 712 if (e->bits) { 713 ++*argi; 714 return e; 715 } 716 717 /* 718 * Parse out all possible fields. 719 * If the field doesn't resolve, bail. 720 */ 721 722 while (NULL != (key = strsep(&argv[*argi], ","))) { 723 if ('\0' == *key) 724 continue; 725 for (i = 0, iterbit = 1; i < KEY_MAX; i++, iterbit <<= 1) { 726 if (0 == strcasecmp(key, mansearch_keynames[i])) { 727 e->bits |= iterbit; 728 break; 729 } 730 } 731 if (i == KEY_MAX) { 732 if (strcasecmp(key, "any")) 733 warnx("treating unknown key " 734 "\"%s\" as \"any\"", key); 735 e->bits |= ~0ULL; 736 } 737 } 738 739 ++*argi; 740 return e; 741 } 742 743 static void 744 exprfree(struct expr *e) 745 { 746 if (e->next != NULL) 747 exprfree(e->next); 748 if (e->child != NULL) 749 exprfree(e->child); 750 free(e); 751 } 752