1 /* $NetBSD: pax.c,v 1.27 2003/06/23 13:06:53 grant 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. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 */ 39 40 #include <sys/cdefs.h> 41 #if defined(__COPYRIGHT) && !defined(lint) 42 __COPYRIGHT("@(#) Copyright (c) 1992, 1993\n\ 43 The Regents of the University of California. All rights reserved.\n"); 44 #endif /* not lint */ 45 46 #if defined(__RCSID) && !defined(lint) 47 #if 0 48 static char sccsid[] = "@(#)pax.c 8.2 (Berkeley) 4/18/94"; 49 #else 50 __RCSID("$NetBSD: pax.c,v 1.27 2003/06/23 13:06:53 grant Exp $"); 51 #endif 52 #endif /* not lint */ 53 54 #include <sys/types.h> 55 #include <sys/param.h> 56 #include <sys/stat.h> 57 #include <sys/time.h> 58 #include <sys/resource.h> 59 #include <stdio.h> 60 #include <fcntl.h> 61 #include <signal.h> 62 #include <unistd.h> 63 #include <stdlib.h> 64 #include <string.h> 65 #include <errno.h> 66 #include <paths.h> 67 #include "pax.h" 68 #include "extern.h" 69 static int gen_init(void); 70 71 /* 72 * PAX main routines, general globals and some simple start up routines 73 */ 74 75 /* 76 * Variables that can be accessed by any routine within pax 77 */ 78 int act = ERROR; /* read/write/append/copy */ 79 FSUB *frmt = NULL; /* archive format type */ 80 int cflag; /* match all EXCEPT pattern/file */ 81 int cwdfd; /* starting cwd */ 82 int dflag; /* directory member match only */ 83 int iflag; /* interactive file/archive rename */ 84 int jflag; /* pass through bzip2 */ 85 int kflag; /* do not overwrite existing files */ 86 int lflag; /* use hard links when possible */ 87 int nflag; /* select first archive member match */ 88 int tflag; /* restore access time after read */ 89 int uflag; /* ignore older modification time files */ 90 int vflag; /* produce verbose output */ 91 int zflag; /* use gzip */ 92 int Aflag; /* honor absolute path */ 93 int Dflag; /* same as uflag except inode change time */ 94 int Hflag; /* follow command line symlinks (write only) */ 95 int Lflag; /* follow symlinks when writing */ 96 int Mflag; /* treat stdin as an mtree(8) specfile */ 97 int Xflag; /* archive files with same device id only */ 98 int Yflag; /* same as Dflg except after name mode */ 99 int Zflag; /* same as uflg except after name mode */ 100 int vfpart; /* is partial verbose output in progress */ 101 int patime = 1; /* preserve file access time */ 102 int pmtime = 1; /* preserve file modification times */ 103 int nodirs; /* do not create directories as needed */ 104 int pfflags = 1; /* preserve file flags */ 105 int pmode; /* preserve file mode bits */ 106 int pids; /* preserve file uid/gid */ 107 int rmleadslash = 0; /* remove leading '/' from pathnames */ 108 int exit_val; /* exit value */ 109 int docrc; /* check/create file crc */ 110 char *dirptr; /* destination dir in a copy */ 111 char *ltmfrmt; /* -v locale time format (if any) */ 112 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 spceific 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 char *tmpdir; 246 size_t tdlen; 247 248 setprogname(argv[0]); 249 250 listf = stderr; 251 252 /* 253 * Keep a reference to cwd, so we can always come back home. 254 */ 255 cwdfd = open(".", O_RDONLY); 256 if (cwdfd < 0) { 257 syswarn(0, errno, "Can't open current working directory."); 258 return(exit_val); 259 } 260 261 /* 262 * Where should we put temporary files? 263 */ 264 if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0') 265 tmpdir = _PATH_TMP; 266 tdlen = strlen(tmpdir); 267 while(tdlen > 0 && tmpdir[tdlen - 1] == '/') 268 tdlen--; 269 tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE)); 270 if (tempfile == NULL) { 271 tty_warn(1, "Cannot allocate memory for temp file name."); 272 return(exit_val); 273 } 274 if (tdlen) 275 memcpy(tempfile, tmpdir, tdlen); 276 tempbase = tempfile + tdlen; 277 *tempbase++ = '/'; 278 279 /* 280 * parse options, determine operational mode, general init 281 */ 282 options(argc, argv); 283 if ((gen_init() < 0) || (tty_init() < 0)) 284 return(exit_val); 285 286 (void)time(&starttime); 287 #ifdef SIGINFO 288 (void)signal(SIGINFO, ar_summary); 289 #endif 290 /* 291 * select a primary operation mode 292 */ 293 switch(act) { 294 case EXTRACT: 295 extract(); 296 break; 297 case ARCHIVE: 298 archive(); 299 break; 300 case APPND: 301 if (gzip_program != NULL) 302 err(1, "cannot gzip while appending"); 303 append(); 304 /* 305 * Check if we tried to append on an empty file and 306 * turned into ARCHIVE mode. 307 */ 308 if (act == -ARCHIVE) { 309 act = ARCHIVE; 310 archive(); 311 } 312 break; 313 case COPY: 314 copy(); 315 break; 316 default: 317 case LIST: 318 list(); 319 break; 320 } 321 return(exit_val); 322 } 323 324 /* 325 * sig_cleanup() 326 * when interrupted we try to do whatever delayed processing we can. 327 * This is not critical, but we really ought to limit our damage when we 328 * are aborted by the user. 329 * Return: 330 * never.... 331 */ 332 333 void 334 sig_cleanup(int which_sig) 335 { 336 /* 337 * restore modes and times for any dirs we may have created 338 * or any dirs we may have read. Set vflag and vfpart so the user 339 * will clearly see the message on a line by itself. 340 */ 341 vflag = vfpart = 1; 342 if (which_sig == SIGXCPU) 343 tty_warn(0, "Cpu time limit reached, cleaning up."); 344 else 345 tty_warn(0, "Signal caught, cleaning up."); 346 347 ar_close(); 348 proc_dir(); 349 if (tflag) 350 atdir_end(); 351 exit(1); 352 } 353 354 /* 355 * gen_init() 356 * general setup routines. Not all are required, but they really help 357 * when dealing with a medium to large sized archives. 358 */ 359 360 static int 361 gen_init(void) 362 { 363 struct rlimit reslimit; 364 struct sigaction n_hand; 365 struct sigaction o_hand; 366 367 /* 368 * Really needed to handle large archives. We can run out of memory for 369 * internal tables really fast when we have a whole lot of files... 370 */ 371 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){ 372 reslimit.rlim_cur = reslimit.rlim_max; 373 (void)setrlimit(RLIMIT_DATA , &reslimit); 374 } 375 376 /* 377 * should file size limits be waived? if the os limits us, this is 378 * needed if we want to write a large archive 379 */ 380 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){ 381 reslimit.rlim_cur = reslimit.rlim_max; 382 (void)setrlimit(RLIMIT_FSIZE , &reslimit); 383 } 384 385 /* 386 * increase the size the stack can grow to 387 */ 388 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){ 389 reslimit.rlim_cur = reslimit.rlim_max; 390 (void)setrlimit(RLIMIT_STACK , &reslimit); 391 } 392 393 #ifdef RLIMIT_RSS 394 /* 395 * not really needed, but doesn't hurt 396 */ 397 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){ 398 reslimit.rlim_cur = reslimit.rlim_max; 399 (void)setrlimit(RLIMIT_RSS , &reslimit); 400 } 401 #endif 402 403 /* 404 * Handle posix locale 405 * 406 * set user defines time printing format for -v option 407 */ 408 ltmfrmt = getenv("LC_TIME"); 409 410 /* 411 * signal handling to reset stored directory times and modes. Since 412 * we deal with broken pipes via failed writes we ignore it. We also 413 * deal with any file size limit thorugh failed writes. Cpu time 414 * limits are caught and a cleanup is forced. 415 */ 416 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) || 417 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) || 418 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) || 419 (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) { 420 tty_warn(1, "Unable to set up signal mask"); 421 return(-1); 422 } 423 memset(&n_hand, 0, sizeof n_hand); 424 n_hand.sa_mask = s_mask; 425 n_hand.sa_flags = 0; 426 n_hand.sa_handler = sig_cleanup; 427 428 if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) && 429 (o_hand.sa_handler == SIG_IGN) && 430 (sigaction(SIGHUP, &o_hand, &o_hand) < 0)) 431 goto out; 432 433 if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) && 434 (o_hand.sa_handler == SIG_IGN) && 435 (sigaction(SIGTERM, &o_hand, &o_hand) < 0)) 436 goto out; 437 438 if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) && 439 (o_hand.sa_handler == SIG_IGN) && 440 (sigaction(SIGINT, &o_hand, &o_hand) < 0)) 441 goto out; 442 443 if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) && 444 (o_hand.sa_handler == SIG_IGN) && 445 (sigaction(SIGQUIT, &o_hand, &o_hand) < 0)) 446 goto out; 447 448 if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) && 449 (o_hand.sa_handler == SIG_IGN) && 450 (sigaction(SIGXCPU, &o_hand, &o_hand) < 0)) 451 goto out; 452 453 n_hand.sa_handler = SIG_IGN; 454 if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) || 455 (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0)) 456 goto out; 457 return(0); 458 459 out: 460 syswarn(1, errno, "Unable to set up signal handler"); 461 return(-1); 462 } 463