1 /* $NetBSD: pax.c,v 1.37 2005/04/24 01:45:04 christos 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\n\ 43 The Regents of the University of California. All rights reserved.\n"); 44 #if 0 45 static char sccsid[] = "@(#)pax.c 8.2 (Berkeley) 4/18/94"; 46 #else 47 __RCSID("$NetBSD: pax.c,v 1.37 2005/04/24 01:45:04 christos 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 "pax.h" 65 #include "extern.h" 66 static int gen_init(void); 67 68 /* 69 * PAX main routines, general globals and some simple start up routines 70 */ 71 72 /* 73 * Variables that can be accessed by any routine within pax 74 */ 75 int act = ERROR; /* read/write/append/copy */ 76 FSUB *frmt = NULL; /* archive format type */ 77 int cflag; /* match all EXCEPT pattern/file */ 78 int cwdfd = -1; /* starting cwd */ 79 int dflag; /* directory member match only */ 80 int iflag; /* interactive file/archive rename */ 81 int jflag; /* pass through bzip2 */ 82 int kflag; /* do not overwrite existing files */ 83 int lflag; /* use hard links when possible */ 84 int nflag; /* select first archive member match */ 85 int tflag; /* restore access time after read */ 86 int uflag; /* ignore older modification time files */ 87 int vflag; /* produce verbose output */ 88 int zflag; /* use gzip */ 89 int Aflag; /* honor absolute path */ 90 int Dflag; /* same as uflag except inode change time */ 91 int Hflag; /* follow command line symlinks (write only) */ 92 int Lflag; /* follow symlinks when writing */ 93 int Mflag; /* treat stdin as an mtree(8) specfile */ 94 int Xflag; /* archive files with same device id only */ 95 int Yflag; /* same as Dflg except after name mode */ 96 int Zflag; /* same as uflg except after name mode */ 97 int vfpart; /* is partial verbose output in progress */ 98 int patime = 1; /* preserve file access time */ 99 int pmtime = 1; /* preserve file modification times */ 100 int nodirs; /* do not create directories as needed */ 101 int pfflags = 1; /* preserve file flags */ 102 int pmode; /* preserve file mode bits */ 103 int pids; /* preserve file uid/gid */ 104 int rmleadslash = 0; /* remove leading '/' from pathnames */ 105 int exit_val; /* exit value */ 106 int docrc; /* check/create file crc */ 107 int to_stdout; /* extract to stdout */ 108 char *dirptr; /* destination dir in a copy */ 109 char *ltmfrmt; /* -v locale time format (if any) */ 110 const char *argv0; /* root of argv[0] */ 111 sigset_t s_mask; /* signal mask for cleanup critical sect */ 112 FILE *listf; /* file pointer to print file list to */ 113 char *tempfile; /* tempfile to use for mkstemp(3) */ 114 char *tempbase; /* basename of tempfile to use for mkstemp(3) */ 115 int forcelocal; /* force local operation even if the name 116 * contains a : 117 */ 118 int secure = 1; /* don't extract names that contain .. */ 119 120 /* 121 * PAX - Portable Archive Interchange 122 * 123 * A utility to read, write, and write lists of the members of archive 124 * files and copy directory hierarchies. A variety of archive formats 125 * are supported (some are described in POSIX 1003.1 10.1): 126 * 127 * ustar - 10.1.1 extended tar interchange format 128 * cpio - 10.1.2 extended cpio interchange format 129 * tar - old BSD 4.3 tar format 130 * binary cpio - old cpio with binary header format 131 * sysVR4 cpio - with and without CRC 132 * 133 * This version is a superset of IEEE Std 1003.2b-d3 134 * 135 * Summary of Extensions to the IEEE Standard: 136 * 137 * 1 READ ENHANCEMENTS 138 * 1.1 Operations which read archives will continue to operate even when 139 * processing archives which may be damaged, truncated, or fail to meet 140 * format specs in several different ways. Damaged sections of archives 141 * are detected and avoided if possible. Attempts will be made to resync 142 * archive read operations even with badly damaged media. 143 * 1.2 Blocksize requirements are not strictly enforced on archive read. 144 * Tapes which have variable sized records can be read without errors. 145 * 1.3 The user can specify via the non-standard option flag -E if error 146 * resync operation should stop on a media error, try a specified number 147 * of times to correct, or try to correct forever. 148 * 1.4 Sparse files (lseek holes) stored on the archive (but stored with blocks 149 * of all zeros will be restored with holes appropriate for the target 150 * filesystem 151 * 1.5 The user is notified whenever something is found during archive 152 * read operations which violates spec (but the read will continue). 153 * 1.6 Multiple archive volumes can be read and may span over different 154 * archive devices 155 * 1.7 Rigidly restores all file attributes exactly as they are stored on the 156 * archive. 157 * 1.8 Modification change time ranges can be specified via multiple -T 158 * options. These allow a user to select files whose modification time 159 * lies within a specific time range. 160 * 1.9 Files can be selected based on owner (user name or uid) via one or more 161 * -U options. 162 * 1.10 Files can be selected based on group (group name or gid) via one o 163 * more -G options. 164 * 1.11 File modification time can be checked against existing file after 165 * name modification (-Z) 166 * 167 * 2 WRITE ENHANCEMENTS 168 * 2.1 Write operation will stop instead of allowing a user to create a flawed 169 * flawed archive (due to any problem). 170 * 2.2 Archives written by pax are forced to strictly conform to both the 171 * archive and pax the spceific format specifications. 172 * 2.3 Blocking size and format is rigidly enforced on writes. 173 * 2.4 Formats which may exhibit header overflow problems (they have fields 174 * too small for large file systems, such as inode number storage), use 175 * routines designed to repair this problem. These techniques still 176 * conform to both pax and format specifications, but no longer truncate 177 * these fields. This removes any restrictions on using these archive 178 * formats on large file systems. 179 * 2.5 Multiple archive volumes can be written and may span over different 180 * archive devices 181 * 2.6 A archive volume record limit allows the user to specify the number 182 * of bytes stored on an archive volume. When reached the user is 183 * prompted for the next archive volume. This is specified with the 184 * non-standard -B flag. The limit is rounded up to the next blocksize. 185 * 2.7 All archive padding during write use zero filled sections. This makes 186 * it much easier to pull data out of flawed archive during read 187 * operations. 188 * 2.8 Access time reset with the -t applies to all file nodes (including 189 * directories). 190 * 2.9 Symbolic links can be followed with -L (optional in the spec). 191 * 2.10 Modification or inode change time ranges can be specified via 192 * multiple -T options. These allow a user to select files whose 193 * modification or inode change time lies within a specific time range. 194 * 2.11 Files can be selected based on owner (user name or uid) via one or more 195 * -U options. 196 * 2.12 Files can be selected based on group (group name or gid) via one o 197 * more -G options. 198 * 2.13 Symlinks which appear on the command line can be followed (without 199 * following other symlinks; -H flag) 200 * 201 * 3 COPY ENHANCEMENTS 202 * 3.1 Sparse files (lseek holes) can be copied without expanding the holes 203 * into zero filled blocks. The file copy is created with holes which are 204 * appropriate for the target filesystem 205 * 3.2 Access time as well as modification time on copied file trees can be 206 * preserved with the appropriate -p options. 207 * 3.3 Access time reset with the -t applies to all file nodes (including 208 * directories). 209 * 3.4 Symbolic links can be followed with -L (optional in the spec). 210 * 3.5 Modification or inode change time ranges can be specified via 211 * multiple -T options. These allow a user to select files whose 212 * modification or inode change time lies within a specific time range. 213 * 3.6 Files can be selected based on owner (user name or uid) via one or more 214 * -U options. 215 * 3.7 Files can be selected based on group (group name or gid) via one o 216 * more -G options. 217 * 3.8 Symlinks which appear on the command line can be followed (without 218 * following other symlinks; -H flag) 219 * 3.9 File inode change time can be checked against existing file before 220 * name modification (-D) 221 * 3.10 File inode change time can be checked against existing file after 222 * name modification (-Y) 223 * 3.11 File modification time can be checked against existing file after 224 * name modification (-Z) 225 * 226 * 4 GENERAL ENHANCEMENTS 227 * 4.1 Internal structure is designed to isolate format dependent and 228 * independent functions. Formats are selected via a format driver table. 229 * This encourages the addition of new archive formats by only having to 230 * write those routines which id, read and write the archive header. 231 */ 232 233 /* 234 * main() 235 * parse options, set up and operate as specified by the user. 236 * any operational flaw will set exit_val to non-zero 237 * Return: 0 if ok, 1 otherwise 238 */ 239 240 int 241 main(int argc, char **argv) 242 { 243 const char *tmpdir; 244 size_t tdlen; 245 246 setprogname(argv[0]); 247 248 listf = stderr; 249 250 /* 251 * parse options, determine operational mode 252 */ 253 options(argc, argv); 254 255 /* 256 * general init 257 */ 258 if ((gen_init() < 0) || (tty_init() < 0)) 259 return(exit_val); 260 261 /* 262 * Keep a reference to cwd, so we can always come back home. 263 */ 264 cwdfd = open(".", O_RDONLY); 265 if (cwdfd < 0) { 266 syswarn(0, errno, "Can't open current working directory."); 267 return(exit_val); 268 } 269 if (updatepath() == -1) 270 return(exit_val); 271 272 /* 273 * Where should we put temporary files? 274 */ 275 if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0') 276 tmpdir = _PATH_TMP; 277 tdlen = strlen(tmpdir); 278 while(tdlen > 0 && tmpdir[tdlen - 1] == '/') 279 tdlen--; 280 tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE)); 281 if (tempfile == NULL) { 282 tty_warn(1, "Cannot allocate memory for temp file name."); 283 return(exit_val); 284 } 285 if (tdlen) 286 memcpy(tempfile, tmpdir, tdlen); 287 tempbase = tempfile + tdlen; 288 *tempbase++ = '/'; 289 290 (void)time(&starttime); 291 #ifdef SIGINFO 292 (void)signal(SIGINFO, ar_summary); 293 #endif 294 /* 295 * select a primary operation mode 296 */ 297 switch(act) { 298 case EXTRACT: 299 extract(); 300 break; 301 case ARCHIVE: 302 archive(); 303 break; 304 case APPND: 305 if (gzip_program != NULL) 306 err(1, "cannot gzip while appending"); 307 append(); 308 /* 309 * Check if we tried to append on an empty file and 310 * turned into ARCHIVE mode. 311 */ 312 if (act == -ARCHIVE) { 313 act = ARCHIVE; 314 archive(); 315 } 316 break; 317 case COPY: 318 copy(); 319 break; 320 default: 321 case LIST: 322 list(); 323 break; 324 } 325 return(exit_val); 326 } 327 328 /* 329 * sig_cleanup() 330 * when interrupted we try to do whatever delayed processing we can. 331 * This is not critical, but we really ought to limit our damage when we 332 * are aborted by the user. 333 * Return: 334 * never.... 335 */ 336 337 void 338 sig_cleanup(int which_sig) 339 { 340 /* 341 * restore modes and times for any dirs we may have created 342 * or any dirs we may have read. Set vflag and vfpart so the user 343 * will clearly see the message on a line by itself. 344 */ 345 vflag = vfpart = 1; 346 #ifdef SIGXCPU 347 if (which_sig == SIGXCPU) 348 tty_warn(0, "CPU time limit reached, cleaning up."); 349 else 350 #endif 351 tty_warn(0, "Signal caught, cleaning up."); 352 353 /* delete any open temporary file */ 354 if (xtmp_name) 355 (void)unlink(xtmp_name); 356 ar_close(); 357 proc_dir(); 358 if (tflag) 359 atdir_end(); 360 exit(1); 361 } 362 363 /* 364 * gen_init() 365 * general setup routines. Not all are required, but they really help 366 * when dealing with a medium to large sized archives. 367 */ 368 369 static int 370 gen_init(void) 371 { 372 struct rlimit reslimit; 373 struct sigaction n_hand; 374 struct sigaction o_hand; 375 376 /* 377 * Really needed to handle large archives. We can run out of memory for 378 * internal tables really fast when we have a whole lot of files... 379 */ 380 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){ 381 reslimit.rlim_cur = reslimit.rlim_max; 382 (void)setrlimit(RLIMIT_DATA , &reslimit); 383 } 384 385 /* 386 * should file size limits be waived? if the os limits us, this is 387 * needed if we want to write a large archive 388 */ 389 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){ 390 reslimit.rlim_cur = reslimit.rlim_max; 391 (void)setrlimit(RLIMIT_FSIZE , &reslimit); 392 } 393 394 /* 395 * increase the size the stack can grow to 396 */ 397 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){ 398 reslimit.rlim_cur = reslimit.rlim_max; 399 (void)setrlimit(RLIMIT_STACK , &reslimit); 400 } 401 402 #ifdef RLIMIT_RSS 403 /* 404 * not really needed, but doesn't hurt 405 */ 406 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){ 407 reslimit.rlim_cur = reslimit.rlim_max; 408 (void)setrlimit(RLIMIT_RSS , &reslimit); 409 } 410 #endif 411 412 /* 413 * Handle posix locale 414 * 415 * set user defines time printing format for -v option 416 */ 417 ltmfrmt = getenv("LC_TIME"); 418 419 /* 420 * signal handling to reset stored directory times and modes. Since 421 * we deal with broken pipes via failed writes we ignore it. We also 422 * deal with any file size limit thorugh failed writes. CPU time 423 * limits are caught and a cleanup is forced. 424 */ 425 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) || 426 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) || 427 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0)){ 428 tty_warn(1, "Unable to set up signal mask"); 429 return(-1); 430 } 431 #ifdef SIGXCPU 432 if (sigaddset(&s_mask,SIGXCPU) < 0) { 433 tty_warn(1, "Unable to set up signal mask"); 434 return(-1); 435 } 436 #endif 437 #ifdef SIGXFSZ 438 if (sigaddset(&s_mask,SIGXFSZ) < 0) { 439 tty_warn(1, "Unable to set up signal mask"); 440 return(-1); 441 } 442 #endif 443 444 memset(&n_hand, 0, sizeof n_hand); 445 n_hand.sa_mask = s_mask; 446 n_hand.sa_flags = 0; 447 n_hand.sa_handler = sig_cleanup; 448 449 if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) && 450 (o_hand.sa_handler == SIG_IGN) && 451 (sigaction(SIGHUP, &o_hand, &o_hand) < 0)) 452 goto out; 453 454 if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) && 455 (o_hand.sa_handler == SIG_IGN) && 456 (sigaction(SIGTERM, &o_hand, &o_hand) < 0)) 457 goto out; 458 459 if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) && 460 (o_hand.sa_handler == SIG_IGN) && 461 (sigaction(SIGINT, &o_hand, &o_hand) < 0)) 462 goto out; 463 464 if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) && 465 (o_hand.sa_handler == SIG_IGN) && 466 (sigaction(SIGQUIT, &o_hand, &o_hand) < 0)) 467 goto out; 468 469 #ifdef SIGXCPU 470 if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) && 471 (o_hand.sa_handler == SIG_IGN) && 472 (sigaction(SIGXCPU, &o_hand, &o_hand) < 0)) 473 goto out; 474 #endif 475 n_hand.sa_handler = SIG_IGN; 476 if (sigaction(SIGPIPE, &n_hand, &o_hand) < 0) 477 goto out; 478 #ifdef SIGXFSZ 479 if (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0) 480 goto out; 481 #endif 482 return(0); 483 484 out: 485 syswarn(1, errno, "Unable to set up signal handler"); 486 return(-1); 487 } 488