1 /* $NetBSD: pax.c,v 1.45 2008/07/20 00:52:40 lukem Exp $ */ 2 3 /*- 4 * Copyright (c) 1992 Keith Muller. 5 * Copyright (c) 1992, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * This code is derived from software contributed to Berkeley by 9 * Keith Muller of the University of California, San Diego. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 */ 35 36 #if HAVE_NBTOOL_CONFIG_H 37 #include "nbtool_config.h" 38 #endif 39 40 #include <sys/cdefs.h> 41 #if !defined(lint) 42 __COPYRIGHT("@(#) Copyright (c) 1992, 1993\ 43 The Regents of the University of California. All rights reserved."); 44 #if 0 45 static char sccsid[] = "@(#)pax.c 8.2 (Berkeley) 4/18/94"; 46 #else 47 __RCSID("$NetBSD: pax.c,v 1.45 2008/07/20 00:52:40 lukem Exp $"); 48 #endif 49 #endif /* not lint */ 50 51 #include <sys/types.h> 52 #include <sys/param.h> 53 #include <sys/stat.h> 54 #include <sys/time.h> 55 #include <sys/resource.h> 56 #include <stdio.h> 57 #include <fcntl.h> 58 #include <signal.h> 59 #include <unistd.h> 60 #include <stdlib.h> 61 #include <string.h> 62 #include <errno.h> 63 #include <paths.h> 64 #include <util.h> 65 #include "pax.h" 66 #include "extern.h" 67 static int gen_init(void); 68 69 /* 70 * PAX main routines, general globals and some simple start up routines 71 */ 72 73 /* 74 * Variables that can be accessed by any routine within pax 75 */ 76 int act = ERROR; /* read/write/append/copy */ 77 FSUB *frmt = NULL; /* archive format type */ 78 int cflag; /* match all EXCEPT pattern/file */ 79 int cwdfd = -1; /* starting cwd */ 80 int dflag; /* directory member match only */ 81 int iflag; /* interactive file/archive rename */ 82 int jflag; /* pass through bzip2 */ 83 int kflag; /* do not overwrite existing files */ 84 int lflag; /* use hard links when possible */ 85 int nflag; /* select first archive member match */ 86 int tflag; /* restore access time after read */ 87 int uflag; /* ignore older modification time files */ 88 int vflag; /* produce verbose output */ 89 int zflag; /* use gzip */ 90 int Aflag; /* honor absolute path */ 91 int Dflag; /* same as uflag except inode change time */ 92 int Hflag; /* follow command line symlinks (write only) */ 93 int Lflag; /* follow symlinks when writing */ 94 int Mflag; /* treat stdin as an mtree(8) specfile */ 95 int Vflag; /* produce somewhat verbose output (no listing) */ 96 int Xflag; /* archive files with same device id only */ 97 int Yflag; /* same as Dflg except after name mode */ 98 int Zflag; /* same as uflg except after name mode */ 99 int vfpart; /* is partial verbose output in progress */ 100 int patime = 1; /* preserve file access time */ 101 int pmtime = 1; /* preserve file modification times */ 102 int nodirs; /* do not create directories as needed */ 103 int pfflags = 1; /* preserve file flags */ 104 int pmode; /* preserve file mode bits */ 105 int pids; /* preserve file uid/gid */ 106 int rmleadslash = 0; /* remove leading '/' from pathnames */ 107 int exit_val; /* exit value */ 108 int docrc; /* check/create file crc */ 109 int to_stdout; /* extract to stdout */ 110 char *dirptr; /* destination dir in a copy */ 111 char *ltmfrmt; /* -v locale time format (if any) */ 112 const char *argv0; /* root of argv[0] */ 113 sigset_t s_mask; /* signal mask for cleanup critical sect */ 114 FILE *listf; /* file pointer to print file list to */ 115 char *tempfile; /* tempfile to use for mkstemp(3) */ 116 char *tempbase; /* basename of tempfile to use for mkstemp(3) */ 117 int forcelocal; /* force local operation even if the name 118 * contains a : 119 */ 120 int secure = 1; /* don't extract names that contain .. */ 121 122 /* 123 * PAX - Portable Archive Interchange 124 * 125 * A utility to read, write, and write lists of the members of archive 126 * files and copy directory hierarchies. A variety of archive formats 127 * are supported (some are described in POSIX 1003.1 10.1): 128 * 129 * ustar - 10.1.1 extended tar interchange format 130 * cpio - 10.1.2 extended cpio interchange format 131 * tar - old BSD 4.3 tar format 132 * binary cpio - old cpio with binary header format 133 * sysVR4 cpio - with and without CRC 134 * 135 * This version is a superset of IEEE Std 1003.2b-d3 136 * 137 * Summary of Extensions to the IEEE Standard: 138 * 139 * 1 READ ENHANCEMENTS 140 * 1.1 Operations which read archives will continue to operate even when 141 * processing archives which may be damaged, truncated, or fail to meet 142 * format specs in several different ways. Damaged sections of archives 143 * are detected and avoided if possible. Attempts will be made to resync 144 * archive read operations even with badly damaged media. 145 * 1.2 Blocksize requirements are not strictly enforced on archive read. 146 * Tapes which have variable sized records can be read without errors. 147 * 1.3 The user can specify via the non-standard option flag -E if error 148 * resync operation should stop on a media error, try a specified number 149 * of times to correct, or try to correct forever. 150 * 1.4 Sparse files (lseek holes) stored on the archive (but stored with blocks 151 * of all zeros will be restored with holes appropriate for the target 152 * filesystem 153 * 1.5 The user is notified whenever something is found during archive 154 * read operations which violates spec (but the read will continue). 155 * 1.6 Multiple archive volumes can be read and may span over different 156 * archive devices 157 * 1.7 Rigidly restores all file attributes exactly as they are stored on the 158 * archive. 159 * 1.8 Modification change time ranges can be specified via multiple -T 160 * options. These allow a user to select files whose modification time 161 * lies within a specific time range. 162 * 1.9 Files can be selected based on owner (user name or uid) via one or more 163 * -U options. 164 * 1.10 Files can be selected based on group (group name or gid) via one o 165 * more -G options. 166 * 1.11 File modification time can be checked against existing file after 167 * name modification (-Z) 168 * 169 * 2 WRITE ENHANCEMENTS 170 * 2.1 Write operation will stop instead of allowing a user to create a flawed 171 * flawed archive (due to any problem). 172 * 2.2 Archives written by pax are forced to strictly conform to both the 173 * archive and pax the specific format specifications. 174 * 2.3 Blocking size and format is rigidly enforced on writes. 175 * 2.4 Formats which may exhibit header overflow problems (they have fields 176 * too small for large file systems, such as inode number storage), use 177 * routines designed to repair this problem. These techniques still 178 * conform to both pax and format specifications, but no longer truncate 179 * these fields. This removes any restrictions on using these archive 180 * formats on large file systems. 181 * 2.5 Multiple archive volumes can be written and may span over different 182 * archive devices 183 * 2.6 A archive volume record limit allows the user to specify the number 184 * of bytes stored on an archive volume. When reached the user is 185 * prompted for the next archive volume. This is specified with the 186 * non-standard -B flag. The limit is rounded up to the next blocksize. 187 * 2.7 All archive padding during write use zero filled sections. This makes 188 * it much easier to pull data out of flawed archive during read 189 * operations. 190 * 2.8 Access time reset with the -t applies to all file nodes (including 191 * directories). 192 * 2.9 Symbolic links can be followed with -L (optional in the spec). 193 * 2.10 Modification or inode change time ranges can be specified via 194 * multiple -T options. These allow a user to select files whose 195 * modification or inode change time lies within a specific time range. 196 * 2.11 Files can be selected based on owner (user name or uid) via one or more 197 * -U options. 198 * 2.12 Files can be selected based on group (group name or gid) via one o 199 * more -G options. 200 * 2.13 Symlinks which appear on the command line can be followed (without 201 * following other symlinks; -H flag) 202 * 203 * 3 COPY ENHANCEMENTS 204 * 3.1 Sparse files (lseek holes) can be copied without expanding the holes 205 * into zero filled blocks. The file copy is created with holes which are 206 * appropriate for the target filesystem 207 * 3.2 Access time as well as modification time on copied file trees can be 208 * preserved with the appropriate -p options. 209 * 3.3 Access time reset with the -t applies to all file nodes (including 210 * directories). 211 * 3.4 Symbolic links can be followed with -L (optional in the spec). 212 * 3.5 Modification or inode change time ranges can be specified via 213 * multiple -T options. These allow a user to select files whose 214 * modification or inode change time lies within a specific time range. 215 * 3.6 Files can be selected based on owner (user name or uid) via one or more 216 * -U options. 217 * 3.7 Files can be selected based on group (group name or gid) via one o 218 * more -G options. 219 * 3.8 Symlinks which appear on the command line can be followed (without 220 * following other symlinks; -H flag) 221 * 3.9 File inode change time can be checked against existing file before 222 * name modification (-D) 223 * 3.10 File inode change time can be checked against existing file after 224 * name modification (-Y) 225 * 3.11 File modification time can be checked against existing file after 226 * name modification (-Z) 227 * 228 * 4 GENERAL ENHANCEMENTS 229 * 4.1 Internal structure is designed to isolate format dependent and 230 * independent functions. Formats are selected via a format driver table. 231 * This encourages the addition of new archive formats by only having to 232 * write those routines which id, read and write the archive header. 233 */ 234 235 /* 236 * main() 237 * parse options, set up and operate as specified by the user. 238 * any operational flaw will set exit_val to non-zero 239 * Return: 0 if ok, 1 otherwise 240 */ 241 242 int 243 main(int argc, char **argv) 244 { 245 const char *tmpdir; 246 size_t tdlen; 247 int rval; 248 249 setprogname(argv[0]); 250 251 listf = stderr; 252 253 /* 254 * parse options, determine operational mode 255 */ 256 options(argc, argv); 257 258 /* 259 * general init 260 */ 261 if ((gen_init() < 0) || (tty_init() < 0)) 262 return exit_val; 263 264 /* 265 * Keep a reference to cwd, so we can always come back home. 266 */ 267 cwdfd = open(".", O_RDONLY); 268 if (cwdfd < 0) { 269 syswarn(1, errno, "Can't open current working directory."); 270 return exit_val; 271 } 272 if (updatepath() == -1) 273 return exit_val; 274 275 /* 276 * Where should we put temporary files? 277 */ 278 if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0') 279 tmpdir = _PATH_TMP; 280 tdlen = strlen(tmpdir); 281 while(tdlen > 0 && tmpdir[tdlen - 1] == '/') 282 tdlen--; 283 tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE)); 284 if (tempfile == NULL) { 285 tty_warn(1, "Cannot allocate memory for temp file name."); 286 return exit_val; 287 } 288 if (tdlen) 289 memcpy(tempfile, tmpdir, tdlen); 290 tempbase = tempfile + tdlen; 291 *tempbase++ = '/'; 292 293 (void)time(&starttime); 294 #ifdef SIGINFO 295 (void)signal(SIGINFO, ar_summary); 296 #endif 297 /* 298 * select a primary operation mode 299 */ 300 switch (act) { 301 case EXTRACT: 302 rval = extract(); 303 break; 304 case ARCHIVE: 305 rval = archive(); 306 break; 307 case APPND: 308 if (gzip_program != NULL) 309 err(1, "cannot gzip while appending"); 310 rval = append(); 311 /* 312 * Check if we tried to append on an empty file and 313 * turned into ARCHIVE mode. 314 */ 315 if (act == -ARCHIVE) { 316 act = ARCHIVE; 317 rval = archive(); 318 } 319 break; 320 case COPY: 321 rval = copy(); 322 break; 323 default: 324 case LIST: 325 rval = list(); 326 break; 327 } 328 if (rval != 0) 329 exit_val = 1; 330 return exit_val; 331 } 332 333 /* 334 * sig_cleanup() 335 * when interrupted we try to do whatever delayed processing we can. 336 * This is not critical, but we really ought to limit our damage when we 337 * are aborted by the user. 338 * Return: 339 * never.... 340 */ 341 342 void 343 sig_cleanup(int which_sig) 344 { 345 /* 346 * restore modes and times for any dirs we may have created 347 * or any dirs we may have read. Set vflag and vfpart so the user 348 * will clearly see the message on a line by itself. 349 */ 350 vflag = vfpart = 1; 351 #ifdef SIGXCPU 352 if (which_sig == SIGXCPU) 353 tty_warn(1, "CPU time limit reached, cleaning up."); 354 else 355 #endif 356 tty_warn(1, "Signal caught, cleaning up."); 357 358 /* delete any open temporary file */ 359 if (xtmp_name) 360 (void)unlink(xtmp_name); 361 ar_close(); 362 proc_dir(); 363 if (tflag) 364 atdir_end(); 365 366 (void)raise_default_signal(which_sig); 367 exit(1); 368 } 369 370 /* 371 * gen_init() 372 * general setup routines. Not all are required, but they really help 373 * when dealing with a medium to large sized archives. 374 */ 375 376 static int 377 gen_init(void) 378 { 379 struct rlimit reslimit; 380 struct sigaction n_hand; 381 struct sigaction o_hand; 382 383 /* 384 * Really needed to handle large archives. We can run out of memory for 385 * internal tables really fast when we have a whole lot of files... 386 */ 387 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){ 388 reslimit.rlim_cur = reslimit.rlim_max; 389 (void)setrlimit(RLIMIT_DATA , &reslimit); 390 } 391 392 /* 393 * should file size limits be waived? if the os limits us, this is 394 * needed if we want to write a large archive 395 */ 396 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){ 397 reslimit.rlim_cur = reslimit.rlim_max; 398 (void)setrlimit(RLIMIT_FSIZE , &reslimit); 399 } 400 401 /* 402 * increase the size the stack can grow to 403 */ 404 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){ 405 reslimit.rlim_cur = reslimit.rlim_max; 406 (void)setrlimit(RLIMIT_STACK , &reslimit); 407 } 408 409 #ifdef RLIMIT_RSS 410 /* 411 * not really needed, but doesn't hurt 412 */ 413 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){ 414 reslimit.rlim_cur = reslimit.rlim_max; 415 (void)setrlimit(RLIMIT_RSS , &reslimit); 416 } 417 #endif 418 419 /* 420 * Handle posix locale 421 * 422 * set user defines time printing format for -v option 423 */ 424 ltmfrmt = getenv("LC_TIME"); 425 426 /* 427 * signal handling to reset stored directory times and modes. Since 428 * we deal with broken pipes via failed writes we ignore it. We also 429 * deal with any file size limit through failed writes. CPU time 430 * limits are caught and a cleanup is forced. 431 */ 432 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) || 433 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) || 434 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0)){ 435 tty_warn(1, "Unable to set up signal mask"); 436 return -1; 437 } 438 #ifdef SIGXCPU 439 if (sigaddset(&s_mask,SIGXCPU) < 0) { 440 tty_warn(1, "Unable to set up signal mask"); 441 return -1; 442 } 443 #endif 444 #ifdef SIGXFSZ 445 if (sigaddset(&s_mask,SIGXFSZ) < 0) { 446 tty_warn(1, "Unable to set up signal mask"); 447 return -1; 448 } 449 #endif 450 451 memset(&n_hand, 0, sizeof n_hand); 452 n_hand.sa_mask = s_mask; 453 n_hand.sa_flags = 0; 454 n_hand.sa_handler = sig_cleanup; 455 456 if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) && 457 (o_hand.sa_handler == SIG_IGN) && 458 (sigaction(SIGHUP, &o_hand, &o_hand) < 0)) 459 goto out; 460 461 if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) && 462 (o_hand.sa_handler == SIG_IGN) && 463 (sigaction(SIGTERM, &o_hand, &o_hand) < 0)) 464 goto out; 465 466 if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) && 467 (o_hand.sa_handler == SIG_IGN) && 468 (sigaction(SIGINT, &o_hand, &o_hand) < 0)) 469 goto out; 470 471 if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) && 472 (o_hand.sa_handler == SIG_IGN) && 473 (sigaction(SIGQUIT, &o_hand, &o_hand) < 0)) 474 goto out; 475 476 #ifdef SIGXCPU 477 if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) && 478 (o_hand.sa_handler == SIG_IGN) && 479 (sigaction(SIGXCPU, &o_hand, &o_hand) < 0)) 480 goto out; 481 #endif 482 n_hand.sa_handler = SIG_IGN; 483 if (sigaction(SIGPIPE, &n_hand, &o_hand) < 0) 484 goto out; 485 #ifdef SIGXFSZ 486 if (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0) 487 goto out; 488 #endif 489 return 0; 490 491 out: 492 syswarn(1, errno, "Unable to set up signal handler"); 493 return -1; 494 } 495