| 1 | /* Create simple DB database from textual input. |
| 2 | Copyright (C) 1996-2021 Free Software Foundation, Inc. |
| 3 | This file is part of the GNU C Library. |
| 4 | Contributed by Ulrich Drepper <drepper@cygnus.com>, 1996. |
| 5 | |
| 6 | The GNU C Library is free software; you can redistribute it and/or |
| 7 | modify it under the terms of the GNU Lesser General Public |
| 8 | License as published by the Free Software Foundation; either |
| 9 | version 2.1 of the License, or (at your option) any later version. |
| 10 | |
| 11 | The GNU C Library is distributed in the hope that it will be useful, |
| 12 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 14 | Lesser General Public License for more details. |
| 15 | |
| 16 | You should have received a copy of the GNU Lesser General Public |
| 17 | License along with the GNU C Library; if not, see |
| 18 | <https://www.gnu.org/licenses/>. */ |
| 19 | |
| 20 | #include <argp.h> |
| 21 | #include <assert.h> |
| 22 | #include <ctype.h> |
| 23 | #include <errno.h> |
| 24 | #include <error.h> |
| 25 | #include <fcntl.h> |
| 26 | #include <inttypes.h> |
| 27 | #include <libintl.h> |
| 28 | #include <locale.h> |
| 29 | #include <search.h> |
| 30 | #include <stdbool.h> |
| 31 | #include <stdio.h> |
| 32 | #include <stdlib.h> |
| 33 | #include <string.h> |
| 34 | #include <unistd.h> |
| 35 | #include <stdint.h> |
| 36 | #include <sys/mman.h> |
| 37 | #include <sys/param.h> |
| 38 | #include <sys/stat.h> |
| 39 | #include <sys/uio.h> |
| 40 | #include "nss_db/nss_db.h" |
| 41 | #include <libc-diag.h> |
| 42 | |
| 43 | /* Get libc version number. */ |
| 44 | #include "../version.h" |
| 45 | |
| 46 | /* The hashing function we use. */ |
| 47 | #include "../intl/hash-string.h" |
| 48 | |
| 49 | /* SELinux support. */ |
| 50 | #ifdef HAVE_SELINUX |
| 51 | # include <selinux/selinux.h> |
| 52 | #endif |
| 53 | |
| 54 | #ifndef MAP_POPULATE |
| 55 | # define MAP_POPULATE 0 |
| 56 | #endif |
| 57 | |
| 58 | #define PACKAGE _libc_intl_domainname |
| 59 | |
| 60 | /* List of data bases. */ |
| 61 | struct database |
| 62 | { |
| 63 | char dbid; |
| 64 | bool ; |
| 65 | struct database *next; |
| 66 | void *entries; |
| 67 | size_t nentries; |
| 68 | size_t nhashentries; |
| 69 | stridx_t *hashtable; |
| 70 | size_t keystrlen; |
| 71 | stridx_t *keyidxtab; |
| 72 | char *keystrtab; |
| 73 | } *databases; |
| 74 | static size_t ndatabases; |
| 75 | static size_t nhashentries_total; |
| 76 | static size_t valstrlen; |
| 77 | static void *valstrtree; |
| 78 | static char *valstrtab; |
| 79 | static size_t ; |
| 80 | |
| 81 | /* Database entry. */ |
| 82 | struct dbentry |
| 83 | { |
| 84 | stridx_t validx; |
| 85 | uint32_t hashval; |
| 86 | char str[0]; |
| 87 | }; |
| 88 | |
| 89 | /* Stored string entry. */ |
| 90 | struct valstrentry |
| 91 | { |
| 92 | stridx_t idx; |
| 93 | bool ; |
| 94 | char str[0]; |
| 95 | }; |
| 96 | |
| 97 | |
| 98 | /* True if any entry has been added. */ |
| 99 | static bool any_dbentry; |
| 100 | |
| 101 | /* If non-zero convert key to lower case. */ |
| 102 | static int to_lowercase; |
| 103 | |
| 104 | /* If non-zero print content of input file, one entry per line. */ |
| 105 | static int do_undo; |
| 106 | |
| 107 | /* If non-zero do not print informational messages. */ |
| 108 | static int be_quiet; |
| 109 | |
| 110 | /* Name of output file. */ |
| 111 | static const char *output_name; |
| 112 | |
| 113 | /* Name and version of program. */ |
| 114 | static void print_version (FILE *stream, struct argp_state *state); |
| 115 | void (*argp_program_version_hook) (FILE *, struct argp_state *) = print_version; |
| 116 | |
| 117 | /* Definitions of arguments for argp functions. */ |
| 118 | static const struct argp_option options[] = |
| 119 | { |
| 120 | { "fold-case" , 'f', NULL, 0, N_("Convert key to lower case" ) }, |
| 121 | { "output" , 'o', N_("NAME" ), 0, N_("Write output to file NAME" ) }, |
| 122 | { "quiet" , 'q', NULL, 0, |
| 123 | N_("Do not print messages while building database" ) }, |
| 124 | { "undo" , 'u', NULL, 0, |
| 125 | N_("Print content of database file, one entry a line" ) }, |
| 126 | { "generated" , 'g', N_("CHAR" ), 0, |
| 127 | N_("Generated line not part of iteration" ) }, |
| 128 | { NULL, 0, NULL, 0, NULL } |
| 129 | }; |
| 130 | |
| 131 | /* Short description of program. */ |
| 132 | static const char doc[] = N_("Create simple database from textual input." ); |
| 133 | |
| 134 | /* Strings for arguments in help texts. */ |
| 135 | static const char args_doc[] = N_("\ |
| 136 | INPUT-FILE OUTPUT-FILE\n-o OUTPUT-FILE INPUT-FILE\n-u INPUT-FILE" ); |
| 137 | |
| 138 | /* Prototype for option handler. */ |
| 139 | static error_t parse_opt (int key, char *arg, struct argp_state *state); |
| 140 | |
| 141 | /* Function to print some extra text in the help message. */ |
| 142 | static char *more_help (int key, const char *text, void *input); |
| 143 | |
| 144 | /* Data structure to communicate with argp functions. */ |
| 145 | static struct argp argp = |
| 146 | { |
| 147 | options, parse_opt, args_doc, doc, NULL, more_help |
| 148 | }; |
| 149 | |
| 150 | |
| 151 | /* List of databases which are not part of the iteration table. */ |
| 152 | static struct db_option |
| 153 | { |
| 154 | char dbid; |
| 155 | struct db_option *next; |
| 156 | } *db_options; |
| 157 | |
| 158 | |
| 159 | /* Prototypes for local functions. */ |
| 160 | static int process_input (FILE *input, const char *inname, |
| 161 | int to_lowercase, int be_quiet); |
| 162 | static int print_database (int fd); |
| 163 | static void compute_tables (void); |
| 164 | static int write_output (int fd); |
| 165 | |
| 166 | /* SELinux support. */ |
| 167 | #ifdef HAVE_SELINUX |
| 168 | /* Set the SELinux file creation context for the given file. */ |
| 169 | static void set_file_creation_context (const char *outname, mode_t mode); |
| 170 | static void reset_file_creation_context (void); |
| 171 | #else |
| 172 | # define set_file_creation_context(_outname,_mode) |
| 173 | # define reset_file_creation_context() |
| 174 | #endif |
| 175 | |
| 176 | |
| 177 | /* External functions. */ |
| 178 | #include <programs/xmalloc.h> |
| 179 | |
| 180 | |
| 181 | int |
| 182 | main (int argc, char *argv[]) |
| 183 | { |
| 184 | const char *input_name; |
| 185 | FILE *input_file; |
| 186 | int remaining; |
| 187 | int mode = 0644; |
| 188 | |
| 189 | /* Set locale via LC_ALL. */ |
| 190 | setlocale (LC_ALL, "" ); |
| 191 | |
| 192 | /* Set the text message domain. */ |
| 193 | textdomain (_libc_intl_domainname); |
| 194 | |
| 195 | /* Initialize local variables. */ |
| 196 | input_name = NULL; |
| 197 | |
| 198 | /* Parse and process arguments. */ |
| 199 | argp_parse (&argp, argc, argv, 0, &remaining, NULL); |
| 200 | |
| 201 | /* Determine file names. */ |
| 202 | if (do_undo || output_name != NULL) |
| 203 | { |
| 204 | if (remaining + 1 != argc) |
| 205 | { |
| 206 | wrong_arguments: |
| 207 | error (0, 0, gettext ("wrong number of arguments" )); |
| 208 | argp_help (&argp, stdout, ARGP_HELP_SEE, |
| 209 | program_invocation_short_name); |
| 210 | exit (1); |
| 211 | } |
| 212 | input_name = argv[remaining]; |
| 213 | } |
| 214 | else |
| 215 | { |
| 216 | if (remaining + 2 != argc) |
| 217 | goto wrong_arguments; |
| 218 | |
| 219 | input_name = argv[remaining++]; |
| 220 | output_name = argv[remaining]; |
| 221 | } |
| 222 | |
| 223 | /* Special handling if we are asked to print the database. */ |
| 224 | if (do_undo) |
| 225 | { |
| 226 | int fd = open (input_name, O_RDONLY); |
| 227 | if (fd == -1) |
| 228 | error (EXIT_FAILURE, errno, gettext ("cannot open database file `%s'" ), |
| 229 | input_name); |
| 230 | |
| 231 | int status = print_database (fd); |
| 232 | |
| 233 | close (fd); |
| 234 | |
| 235 | return status; |
| 236 | } |
| 237 | |
| 238 | /* Open input file. */ |
| 239 | if (strcmp (input_name, "-" ) == 0 || strcmp (input_name, "/dev/stdin" ) == 0) |
| 240 | input_file = stdin; |
| 241 | else |
| 242 | { |
| 243 | struct stat64 st; |
| 244 | |
| 245 | input_file = fopen64 (input_name, "r" ); |
| 246 | if (input_file == NULL) |
| 247 | error (EXIT_FAILURE, errno, gettext ("cannot open input file `%s'" ), |
| 248 | input_name); |
| 249 | |
| 250 | /* Get the access rights from the source file. The output file should |
| 251 | have the same. */ |
| 252 | if (fstat64 (fileno (input_file), &st) >= 0) |
| 253 | mode = st.st_mode & ACCESSPERMS; |
| 254 | } |
| 255 | |
| 256 | /* Start the real work. */ |
| 257 | int status = process_input (input_file, input_name, to_lowercase, be_quiet); |
| 258 | |
| 259 | /* Close files. */ |
| 260 | if (input_file != stdin) |
| 261 | fclose (input_file); |
| 262 | |
| 263 | /* No need to continue when we did not read the file successfully. */ |
| 264 | if (status != EXIT_SUCCESS) |
| 265 | return status; |
| 266 | |
| 267 | /* Bail out if nothing is to be done. */ |
| 268 | if (!any_dbentry) |
| 269 | { |
| 270 | if (be_quiet) |
| 271 | return EXIT_SUCCESS; |
| 272 | else |
| 273 | error (EXIT_SUCCESS, 0, gettext ("no entries to be processed" )); |
| 274 | } |
| 275 | |
| 276 | /* Compute hash and string tables. */ |
| 277 | compute_tables (); |
| 278 | |
| 279 | /* Open output file. This must not be standard output so we don't |
| 280 | handle "-" and "/dev/stdout" special. */ |
| 281 | char *tmp_output_name; |
| 282 | if (asprintf (&tmp_output_name, "%s.XXXXXX" , output_name) == -1) |
| 283 | error (EXIT_FAILURE, errno, gettext ("cannot create temporary file name" )); |
| 284 | |
| 285 | set_file_creation_context (output_name, mode); |
| 286 | int fd = mkstemp (tmp_output_name); |
| 287 | reset_file_creation_context (); |
| 288 | if (fd == -1) |
| 289 | error (EXIT_FAILURE, errno, gettext ("cannot create temporary file" )); |
| 290 | |
| 291 | status = write_output (fd); |
| 292 | |
| 293 | if (status == EXIT_SUCCESS) |
| 294 | { |
| 295 | struct stat64 st; |
| 296 | |
| 297 | if (fstat64 (fd, &st) == 0) |
| 298 | { |
| 299 | if ((st.st_mode & ACCESSPERMS) != mode) |
| 300 | /* We ignore problems with changing the mode. */ |
| 301 | fchmod (fd, mode); |
| 302 | } |
| 303 | else |
| 304 | { |
| 305 | error (0, errno, gettext ("cannot stat newly created file" )); |
| 306 | status = EXIT_FAILURE; |
| 307 | } |
| 308 | } |
| 309 | |
| 310 | close (fd); |
| 311 | |
| 312 | if (status == EXIT_SUCCESS) |
| 313 | { |
| 314 | if (rename (tmp_output_name, output_name) != 0) |
| 315 | { |
| 316 | error (0, errno, gettext ("cannot rename temporary file" )); |
| 317 | status = EXIT_FAILURE; |
| 318 | goto do_unlink; |
| 319 | } |
| 320 | } |
| 321 | else |
| 322 | do_unlink: |
| 323 | unlink (tmp_output_name); |
| 324 | |
| 325 | return status; |
| 326 | } |
| 327 | |
| 328 | |
| 329 | /* Handle program arguments. */ |
| 330 | static error_t |
| 331 | parse_opt (int key, char *arg, struct argp_state *state) |
| 332 | { |
| 333 | struct db_option *newp; |
| 334 | |
| 335 | switch (key) |
| 336 | { |
| 337 | case 'f': |
| 338 | to_lowercase = 1; |
| 339 | break; |
| 340 | case 'o': |
| 341 | output_name = arg; |
| 342 | break; |
| 343 | case 'q': |
| 344 | be_quiet = 1; |
| 345 | break; |
| 346 | case 'u': |
| 347 | do_undo = 1; |
| 348 | break; |
| 349 | case 'g': |
| 350 | newp = xmalloc (sizeof (*newp)); |
| 351 | newp->dbid = arg[0]; |
| 352 | newp->next = db_options; |
| 353 | db_options = newp; |
| 354 | break; |
| 355 | default: |
| 356 | return ARGP_ERR_UNKNOWN; |
| 357 | } |
| 358 | return 0; |
| 359 | } |
| 360 | |
| 361 | |
| 362 | static char * |
| 363 | more_help (int key, const char *text, void *input) |
| 364 | { |
| 365 | char *tp = NULL; |
| 366 | switch (key) |
| 367 | { |
| 368 | case ARGP_KEY_HELP_EXTRA: |
| 369 | /* We print some extra information. */ |
| 370 | if (asprintf (&tp, gettext ("\ |
| 371 | For bug reporting instructions, please see:\n\ |
| 372 | %s.\n" ), REPORT_BUGS_TO) < 0) |
| 373 | return NULL; |
| 374 | return tp; |
| 375 | default: |
| 376 | break; |
| 377 | } |
| 378 | return (char *) text; |
| 379 | } |
| 380 | |
| 381 | /* Print the version information. */ |
| 382 | static void |
| 383 | print_version (FILE *stream, struct argp_state *state) |
| 384 | { |
| 385 | fprintf (stream, "makedb %s%s\n" , PKGVERSION, VERSION); |
| 386 | fprintf (stream, gettext ("\ |
| 387 | Copyright (C) %s Free Software Foundation, Inc.\n\ |
| 388 | This is free software; see the source for copying conditions. There is NO\n\ |
| 389 | warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n\ |
| 390 | " ), "2021" ); |
| 391 | fprintf (stream, gettext ("Written by %s.\n" ), "Ulrich Drepper" ); |
| 392 | } |
| 393 | |
| 394 | |
| 395 | static int |
| 396 | dbentry_compare (const void *p1, const void *p2) |
| 397 | { |
| 398 | const struct dbentry *d1 = (const struct dbentry *) p1; |
| 399 | const struct dbentry *d2 = (const struct dbentry *) p2; |
| 400 | |
| 401 | if (d1->hashval != d2->hashval) |
| 402 | return d1->hashval < d2->hashval ? -1 : 1; |
| 403 | |
| 404 | return strcmp (d1->str, d2->str); |
| 405 | } |
| 406 | |
| 407 | |
| 408 | static int |
| 409 | valstr_compare (const void *p1, const void *p2) |
| 410 | { |
| 411 | const struct valstrentry *d1 = (const struct valstrentry *) p1; |
| 412 | const struct valstrentry *d2 = (const struct valstrentry *) p2; |
| 413 | |
| 414 | return strcmp (d1->str, d2->str); |
| 415 | } |
| 416 | |
| 417 | |
| 418 | static int |
| 419 | process_input (FILE *input, const char *inname, int to_lowercase, int be_quiet) |
| 420 | { |
| 421 | char *line; |
| 422 | size_t linelen; |
| 423 | int status; |
| 424 | size_t linenr; |
| 425 | |
| 426 | line = NULL; |
| 427 | linelen = 0; |
| 428 | status = EXIT_SUCCESS; |
| 429 | linenr = 0; |
| 430 | |
| 431 | struct database *last_database = NULL; |
| 432 | |
| 433 | while (!feof_unlocked (input)) |
| 434 | { |
| 435 | ssize_t n = getline (&line, &linelen, input); |
| 436 | if (n < 0) |
| 437 | /* This means end of file or some bug. */ |
| 438 | break; |
| 439 | if (n == 0) |
| 440 | /* Short read. Probably interrupted system call. */ |
| 441 | continue; |
| 442 | |
| 443 | ++linenr; |
| 444 | |
| 445 | if (line[n - 1] == '\n') |
| 446 | /* Remove trailing newline. */ |
| 447 | line[--n] = '\0'; |
| 448 | |
| 449 | char *cp = line; |
| 450 | while (isspace (*cp)) |
| 451 | ++cp; |
| 452 | |
| 453 | if (*cp == '#' || *cp == '\0') |
| 454 | /* First non-space character in line '#': it's a comment. |
| 455 | Also go to the next line if it is empty except for whitespaces. */ |
| 456 | continue; |
| 457 | |
| 458 | /* Skip over the character indicating the database so that it is not |
| 459 | affected by TO_LOWERCASE. */ |
| 460 | char *key = cp++; |
| 461 | while (*cp != '\0' && !isspace (*cp)) |
| 462 | { |
| 463 | if (to_lowercase) |
| 464 | *cp = tolower (*cp); |
| 465 | ++cp; |
| 466 | } |
| 467 | |
| 468 | if (*cp == '\0') |
| 469 | /* It's a line without a value field. */ |
| 470 | continue; |
| 471 | |
| 472 | *cp++ = '\0'; |
| 473 | size_t keylen = cp - key; |
| 474 | |
| 475 | while (isspace (*cp)) |
| 476 | ++cp; |
| 477 | |
| 478 | char *data = cp; |
| 479 | size_t datalen = (&line[n] - cp) + 1; |
| 480 | |
| 481 | /* Find the database. */ |
| 482 | if (last_database == NULL || last_database->dbid != key[0]) |
| 483 | { |
| 484 | last_database = databases; |
| 485 | while (last_database != NULL && last_database->dbid != key[0]) |
| 486 | last_database = last_database->next; |
| 487 | |
| 488 | if (last_database == NULL) |
| 489 | { |
| 490 | last_database = xmalloc (sizeof (*last_database)); |
| 491 | last_database->dbid = key[0]; |
| 492 | last_database->extra_string = false; |
| 493 | last_database->next = databases; |
| 494 | last_database->entries = NULL; |
| 495 | last_database->nentries = 0; |
| 496 | last_database->keystrlen = 0; |
| 497 | databases = last_database; |
| 498 | |
| 499 | struct db_option *runp = db_options; |
| 500 | while (runp != NULL) |
| 501 | if (runp->dbid == key[0]) |
| 502 | { |
| 503 | last_database->extra_string = true; |
| 504 | break; |
| 505 | } |
| 506 | else |
| 507 | runp = runp->next; |
| 508 | } |
| 509 | } |
| 510 | |
| 511 | /* Skip the database selector. */ |
| 512 | ++key; |
| 513 | --keylen; |
| 514 | |
| 515 | /* Store the data. */ |
| 516 | struct valstrentry *nentry = xmalloc (sizeof (struct valstrentry) |
| 517 | + datalen); |
| 518 | if (last_database->extra_string) |
| 519 | nentry->idx = extrastrlen; |
| 520 | else |
| 521 | nentry->idx = valstrlen; |
| 522 | nentry->extra_string = last_database->extra_string; |
| 523 | memcpy (nentry->str, data, datalen); |
| 524 | |
| 525 | struct valstrentry **fdata = tsearch (nentry, &valstrtree, |
| 526 | valstr_compare); |
| 527 | if (fdata == NULL) |
| 528 | error (EXIT_FAILURE, errno, gettext ("cannot create search tree" )); |
| 529 | |
| 530 | if (*fdata != nentry) |
| 531 | { |
| 532 | /* We can reuse a string. */ |
| 533 | free (nentry); |
| 534 | nentry = *fdata; |
| 535 | } |
| 536 | else |
| 537 | if (last_database->extra_string) |
| 538 | extrastrlen += datalen; |
| 539 | else |
| 540 | valstrlen += datalen; |
| 541 | |
| 542 | /* Store the key. */ |
| 543 | struct dbentry *newp = xmalloc (sizeof (struct dbentry) + keylen); |
| 544 | newp->validx = nentry->idx; |
| 545 | newp->hashval = __hash_string (key); |
| 546 | memcpy (newp->str, key, keylen); |
| 547 | |
| 548 | struct dbentry **found = tsearch (newp, &last_database->entries, |
| 549 | dbentry_compare); |
| 550 | if (found == NULL) |
| 551 | error (EXIT_FAILURE, errno, gettext ("cannot create search tree" )); |
| 552 | |
| 553 | if (*found != newp) |
| 554 | { |
| 555 | free (newp); |
| 556 | if (!be_quiet) |
| 557 | error_at_line (0, 0, inname, linenr, gettext ("duplicate key" )); |
| 558 | continue; |
| 559 | } |
| 560 | |
| 561 | ++last_database->nentries; |
| 562 | last_database->keystrlen += keylen; |
| 563 | |
| 564 | any_dbentry = true; |
| 565 | } |
| 566 | |
| 567 | if (ferror_unlocked (input)) |
| 568 | { |
| 569 | error (0, 0, gettext ("problems while reading `%s'" ), inname); |
| 570 | status = EXIT_FAILURE; |
| 571 | } |
| 572 | |
| 573 | return status; |
| 574 | } |
| 575 | |
| 576 | |
| 577 | static void |
| 578 | copy_valstr (const void *nodep, const VISIT which, const int depth) |
| 579 | { |
| 580 | if (which != leaf && which != postorder) |
| 581 | return; |
| 582 | |
| 583 | const struct valstrentry *p = *(const struct valstrentry **) nodep; |
| 584 | |
| 585 | strcpy (valstrtab + (p->extra_string ? valstrlen : 0) + p->idx, p->str); |
| 586 | } |
| 587 | |
| 588 | |
| 589 | /* Determine if the candidate is prime by using a modified trial division |
| 590 | algorithm. The candidate must be both odd and greater than 4. */ |
| 591 | static int |
| 592 | is_prime (size_t candidate) |
| 593 | { |
| 594 | size_t divn = 3; |
| 595 | size_t sq = divn * divn; |
| 596 | |
| 597 | assert (candidate > 4 && candidate % 2 != 0); |
| 598 | |
| 599 | while (sq < candidate && candidate % divn != 0) |
| 600 | { |
| 601 | ++divn; |
| 602 | sq += 4 * divn; |
| 603 | ++divn; |
| 604 | } |
| 605 | |
| 606 | return candidate % divn != 0; |
| 607 | } |
| 608 | |
| 609 | |
| 610 | static size_t |
| 611 | next_prime (size_t seed) |
| 612 | { |
| 613 | /* Make sure that we're always greater than 4. */ |
| 614 | seed = (seed + 4) | 1; |
| 615 | |
| 616 | while (!is_prime (seed)) |
| 617 | seed += 2; |
| 618 | |
| 619 | return seed; |
| 620 | } |
| 621 | |
| 622 | |
| 623 | static void |
| 624 | compute_tables (void) |
| 625 | { |
| 626 | valstrtab = xmalloc (roundup (valstrlen + extrastrlen, sizeof (stridx_t))); |
| 627 | while ((valstrlen + extrastrlen) % sizeof (stridx_t) != 0) |
| 628 | valstrtab[valstrlen++] = '\0'; |
| 629 | twalk (valstrtree, copy_valstr); |
| 630 | |
| 631 | static struct database *db; |
| 632 | for (db = databases; db != NULL; db = db->next) |
| 633 | if (db->nentries != 0) |
| 634 | { |
| 635 | ++ndatabases; |
| 636 | |
| 637 | /* We simply use an odd number large than twice the number of |
| 638 | elements to store in the hash table for the size. This gives |
| 639 | enough efficiency. */ |
| 640 | #define TEST_RANGE 30 |
| 641 | size_t nhashentries_min = next_prime (db->nentries < TEST_RANGE |
| 642 | ? db->nentries |
| 643 | : db->nentries * 2 - TEST_RANGE); |
| 644 | size_t nhashentries_max = MAX (nhashentries_min, db->nentries * 4); |
| 645 | size_t nhashentries_best = nhashentries_min; |
| 646 | size_t chainlength_best = db->nentries; |
| 647 | |
| 648 | db->hashtable = xmalloc (2 * nhashentries_max * sizeof (stridx_t) |
| 649 | + db->keystrlen); |
| 650 | db->keyidxtab = db->hashtable + nhashentries_max; |
| 651 | db->keystrtab = (char *) (db->keyidxtab + nhashentries_max); |
| 652 | |
| 653 | static size_t max_chainlength; |
| 654 | static char *wp; |
| 655 | static size_t nhashentries; |
| 656 | static bool copy_string; |
| 657 | |
| 658 | void add_key(const void *nodep, const VISIT which, const int depth) |
| 659 | { |
| 660 | if (which != leaf && which != postorder) |
| 661 | return; |
| 662 | |
| 663 | const struct dbentry *dbe = *(const struct dbentry **) nodep; |
| 664 | |
| 665 | ptrdiff_t stridx; |
| 666 | if (copy_string) |
| 667 | { |
| 668 | stridx = wp - db->keystrtab; |
| 669 | wp = stpcpy (wp, dbe->str) + 1; |
| 670 | } |
| 671 | else |
| 672 | stridx = 0; |
| 673 | |
| 674 | size_t hidx = dbe->hashval % nhashentries; |
| 675 | size_t hval2 = 1 + dbe->hashval % (nhashentries - 2); |
| 676 | size_t chainlength = 0; |
| 677 | |
| 678 | while (db->hashtable[hidx] != ~((stridx_t) 0)) |
| 679 | { |
| 680 | ++chainlength; |
| 681 | if ((hidx += hval2) >= nhashentries) |
| 682 | hidx -= nhashentries; |
| 683 | } |
| 684 | |
| 685 | db->hashtable[hidx] = ((db->extra_string ? valstrlen : 0) |
| 686 | + dbe->validx); |
| 687 | db->keyidxtab[hidx] = stridx; |
| 688 | |
| 689 | max_chainlength = MAX (max_chainlength, chainlength); |
| 690 | } |
| 691 | |
| 692 | copy_string = false; |
| 693 | nhashentries = nhashentries_min; |
| 694 | for (size_t cnt = 0; cnt < TEST_RANGE; ++cnt) |
| 695 | { |
| 696 | memset (db->hashtable, '\xff', nhashentries * sizeof (stridx_t)); |
| 697 | |
| 698 | max_chainlength = 0; |
| 699 | wp = db->keystrtab; |
| 700 | |
| 701 | twalk (db->entries, add_key); |
| 702 | |
| 703 | if (max_chainlength == 0) |
| 704 | { |
| 705 | /* No need to look further, this is as good as it gets. */ |
| 706 | nhashentries_best = nhashentries; |
| 707 | break; |
| 708 | } |
| 709 | |
| 710 | if (max_chainlength < chainlength_best) |
| 711 | { |
| 712 | chainlength_best = max_chainlength; |
| 713 | nhashentries_best = nhashentries; |
| 714 | } |
| 715 | |
| 716 | nhashentries = next_prime (nhashentries + 1); |
| 717 | if (nhashentries > nhashentries_max) |
| 718 | break; |
| 719 | } |
| 720 | |
| 721 | /* Recompute the best table again, this time fill in the strings. */ |
| 722 | nhashentries = nhashentries_best; |
| 723 | memset (db->hashtable, '\xff', |
| 724 | 2 * nhashentries_max * sizeof (stridx_t)); |
| 725 | copy_string = true; |
| 726 | wp = db->keystrtab; |
| 727 | |
| 728 | twalk (db->entries, add_key); |
| 729 | |
| 730 | db->nhashentries = nhashentries_best; |
| 731 | nhashentries_total += nhashentries_best; |
| 732 | } |
| 733 | } |
| 734 | |
| 735 | |
| 736 | static int |
| 737 | write_output (int fd) |
| 738 | { |
| 739 | struct nss_db_header *; |
| 740 | uint64_t file_offset = (sizeof (struct nss_db_header) |
| 741 | + (ndatabases * sizeof (header->dbs[0]))); |
| 742 | header = alloca (file_offset); |
| 743 | |
| 744 | header->magic = NSS_DB_MAGIC; |
| 745 | header->ndbs = ndatabases; |
| 746 | header->valstroffset = file_offset; |
| 747 | header->valstrlen = valstrlen; |
| 748 | |
| 749 | size_t filled_dbs = 0; |
| 750 | size_t iov_nelts = 2 + ndatabases * 3; |
| 751 | struct iovec iov[iov_nelts]; |
| 752 | iov[0].iov_base = header; |
| 753 | iov[0].iov_len = file_offset; |
| 754 | |
| 755 | iov[1].iov_base = valstrtab; |
| 756 | iov[1].iov_len = valstrlen + extrastrlen; |
| 757 | file_offset += iov[1].iov_len; |
| 758 | |
| 759 | size_t keydataoffset = file_offset + nhashentries_total * sizeof (stridx_t); |
| 760 | for (struct database *db = databases; db != NULL; db = db->next) |
| 761 | if (db->entries != NULL) |
| 762 | { |
| 763 | assert (file_offset % sizeof (stridx_t) == 0); |
| 764 | assert (filled_dbs < ndatabases); |
| 765 | |
| 766 | header->dbs[filled_dbs].id = db->dbid; |
| 767 | memset (header->dbs[filled_dbs].pad, '\0', |
| 768 | sizeof (header->dbs[0].pad)); |
| 769 | header->dbs[filled_dbs].hashsize = db->nhashentries; |
| 770 | |
| 771 | iov[2 + filled_dbs].iov_base = db->hashtable; |
| 772 | iov[2 + filled_dbs].iov_len = db->nhashentries * sizeof (stridx_t); |
| 773 | header->dbs[filled_dbs].hashoffset = file_offset; |
| 774 | file_offset += iov[2 + filled_dbs].iov_len; |
| 775 | |
| 776 | iov[2 + ndatabases + filled_dbs * 2].iov_base = db->keyidxtab; |
| 777 | iov[2 + ndatabases + filled_dbs * 2].iov_len |
| 778 | = db->nhashentries * sizeof (stridx_t); |
| 779 | header->dbs[filled_dbs].keyidxoffset = keydataoffset; |
| 780 | keydataoffset += iov[2 + ndatabases + filled_dbs * 2].iov_len; |
| 781 | |
| 782 | iov[3 + ndatabases + filled_dbs * 2].iov_base = db->keystrtab; |
| 783 | iov[3 + ndatabases + filled_dbs * 2].iov_len = db->keystrlen; |
| 784 | header->dbs[filled_dbs].keystroffset = keydataoffset; |
| 785 | keydataoffset += iov[3 + ndatabases + filled_dbs * 2].iov_len; |
| 786 | |
| 787 | ++filled_dbs; |
| 788 | } |
| 789 | |
| 790 | assert (filled_dbs == ndatabases); |
| 791 | assert (file_offset == (iov[0].iov_len + iov[1].iov_len |
| 792 | + nhashentries_total * sizeof (stridx_t))); |
| 793 | header->allocate = file_offset; |
| 794 | |
| 795 | #if __GNUC_PREREQ (10, 0) && !__GNUC_PREREQ (11, 0) |
| 796 | DIAG_PUSH_NEEDS_COMMENT; |
| 797 | /* Avoid GCC 10 false positive warning: specified size exceeds maximum |
| 798 | object size. */ |
| 799 | DIAG_IGNORE_NEEDS_COMMENT (10, "-Wstringop-overflow" ); |
| 800 | #endif |
| 801 | |
| 802 | assert (iov_nelts <= INT_MAX); |
| 803 | if (writev (fd, iov, iov_nelts) != keydataoffset) |
| 804 | { |
| 805 | error (0, errno, gettext ("failed to write new database file" )); |
| 806 | return EXIT_FAILURE; |
| 807 | } |
| 808 | |
| 809 | #if __GNUC_PREREQ (10, 0) && !__GNUC_PREREQ (11, 0) |
| 810 | DIAG_POP_NEEDS_COMMENT; |
| 811 | #endif |
| 812 | |
| 813 | return EXIT_SUCCESS; |
| 814 | } |
| 815 | |
| 816 | |
| 817 | static int |
| 818 | print_database (int fd) |
| 819 | { |
| 820 | struct stat64 st; |
| 821 | if (fstat64 (fd, &st) != 0) |
| 822 | error (EXIT_FAILURE, errno, gettext ("cannot stat database file" )); |
| 823 | |
| 824 | const struct nss_db_header * = mmap (NULL, st.st_size, PROT_READ, |
| 825 | MAP_PRIVATE|MAP_POPULATE, fd, 0); |
| 826 | if (header == MAP_FAILED) |
| 827 | error (EXIT_FAILURE, errno, gettext ("cannot map database file" )); |
| 828 | |
| 829 | if (header->magic != NSS_DB_MAGIC) |
| 830 | error (EXIT_FAILURE, 0, gettext ("file not a database file" )); |
| 831 | |
| 832 | const char *valstrtab = (const char *) header + header->valstroffset; |
| 833 | |
| 834 | for (unsigned int dbidx = 0; dbidx < header->ndbs; ++dbidx) |
| 835 | { |
| 836 | const stridx_t *stridxtab |
| 837 | = ((const stridx_t *) ((const char *) header |
| 838 | + header->dbs[dbidx].keyidxoffset)); |
| 839 | const char *keystrtab |
| 840 | = (const char *) header + header->dbs[dbidx].keystroffset; |
| 841 | const stridx_t *hashtab |
| 842 | = (const stridx_t *) ((const char *) header |
| 843 | + header->dbs[dbidx].hashoffset); |
| 844 | |
| 845 | for (uint32_t hidx = 0; hidx < header->dbs[dbidx].hashsize; ++hidx) |
| 846 | if (hashtab[hidx] != ~((stridx_t) 0)) |
| 847 | printf ("%c%s %s\n" , |
| 848 | header->dbs[dbidx].id, |
| 849 | keystrtab + stridxtab[hidx], |
| 850 | valstrtab + hashtab[hidx]); |
| 851 | } |
| 852 | |
| 853 | return EXIT_SUCCESS; |
| 854 | } |
| 855 | |
| 856 | |
| 857 | #ifdef HAVE_SELINUX |
| 858 | |
| 859 | /* security_context_t and matchpathcon (along with several other symbols) were |
| 860 | marked as deprecated by the SELinux API starting from version 3.1. We use |
| 861 | them here, but should eventually switch to the newer API. */ |
| 862 | DIAG_PUSH_NEEDS_COMMENT |
| 863 | DIAG_IGNORE_NEEDS_COMMENT (10, "-Wdeprecated-declarations" ); |
| 864 | |
| 865 | static void |
| 866 | set_file_creation_context (const char *outname, mode_t mode) |
| 867 | { |
| 868 | static int enabled; |
| 869 | static int enforcing; |
| 870 | security_context_t ctx; |
| 871 | |
| 872 | /* Check if SELinux is enabled, and remember. */ |
| 873 | if (enabled == 0) |
| 874 | enabled = is_selinux_enabled () ? 1 : -1; |
| 875 | if (enabled < 0) |
| 876 | return; |
| 877 | |
| 878 | /* Check if SELinux is enforcing, and remember. */ |
| 879 | if (enforcing == 0) |
| 880 | enforcing = security_getenforce () ? 1 : -1; |
| 881 | |
| 882 | /* Determine the context which the file should have. */ |
| 883 | ctx = NULL; |
| 884 | if (matchpathcon (outname, S_IFREG | mode, &ctx) == 0 && ctx != NULL) |
| 885 | { |
| 886 | if (setfscreatecon (ctx) != 0) |
| 887 | error (enforcing > 0 ? EXIT_FAILURE : 0, 0, |
| 888 | gettext ("cannot set file creation context for `%s'" ), |
| 889 | outname); |
| 890 | |
| 891 | freecon (ctx); |
| 892 | } |
| 893 | } |
| 894 | DIAG_POP_NEEDS_COMMENT |
| 895 | |
| 896 | static void |
| 897 | reset_file_creation_context (void) |
| 898 | { |
| 899 | setfscreatecon (NULL); |
| 900 | } |
| 901 | #endif |
| 902 | |