1 /* $NetBSD: pax.c,v 1.17 2002/01/31 19:27:54 tv 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 1.17 2002/01/31 19:27:54 tv Exp $"); 51 #endif 52 #endif /* not lint */ 53 54 #include <stdio.h> 55 #include <sys/types.h> 56 #include <sys/param.h> 57 #include <sys/stat.h> 58 #include <sys/time.h> 59 #include <sys/resource.h> 60 #include <signal.h> 61 #include <unistd.h> 62 #include <stdlib.h> 63 #include <errno.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 = DEFOP; /* read/write/append/copy */ 76 FSUB *frmt = NULL; /* archive format type */ 77 int cflag; /* match all EXCEPT pattern/file */ 78 int dflag; /* directory member match only */ 79 int iflag; /* interactive file/archive rename */ 80 int kflag; /* do not overwrite existing files */ 81 int lflag; /* use hard links when possible */ 82 int nflag; /* select first archive member match */ 83 int tflag; /* restore access time after read */ 84 int uflag; /* ignore older modification time files */ 85 int vflag; /* produce verbose output */ 86 int zflag; /* use gzip */ 87 int Aflag; /* honor absolute path */ 88 int Dflag; /* same as uflag except inode change time */ 89 int Hflag; /* follow command line symlinks (write only) */ 90 int Lflag; /* follow symlinks when writing */ 91 int Mflag; /* treat stdin as an mtree(8) specfile */ 92 int Xflag; /* archive files with same device id only */ 93 int Yflag; /* same as Dflg except after name mode */ 94 int Zflag; /* same as uflg except after name mode */ 95 int vfpart; /* is partial verbose output in progress */ 96 int patime = 1; /* preserve file access time */ 97 int pmtime = 1; /* preserve file modification times */ 98 int pfflags = 1; /* preserve file flags */ 99 int pmode; /* preserve file mode bits */ 100 int pids; /* preserve file uid/gid */ 101 int exit_val; /* exit value */ 102 int docrc; /* check/create file crc */ 103 char *dirptr; /* destination dir in a copy */ 104 char *argv0; /* root of argv[0] */ 105 sigset_t s_mask; /* signal mask for cleanup critical sect */ 106 107 /* 108 * PAX - Portable Archive Interchange 109 * 110 * A utility to read, write, and write lists of the members of archive 111 * files and copy directory hierarchies. A variety of archive formats 112 * are supported (some are described in POSIX 1003.1 10.1): 113 * 114 * ustar - 10.1.1 extended tar interchange format 115 * cpio - 10.1.2 extended cpio interchange format 116 * tar - old BSD 4.3 tar format 117 * binary cpio - old cpio with binary header format 118 * sysVR4 cpio - with and without CRC 119 * 120 * This version is a superset of IEEE Std 1003.2b-d3 121 * 122 * Summary of Extensions to the IEEE Standard: 123 * 124 * 1 READ ENHANCEMENTS 125 * 1.1 Operations which read archives will continue to operate even when 126 * processing archives which may be damaged, truncated, or fail to meet 127 * format specs in several different ways. Damaged sections of archives 128 * are detected and avoided if possible. Attempts will be made to resync 129 * archive read operations even with badly damaged media. 130 * 1.2 Blocksize requirements are not strictly enforced on archive read. 131 * Tapes which have variable sized records can be read without errors. 132 * 1.3 The user can specify via the non-standard option flag -E if error 133 * resync operation should stop on a media error, try a specified number 134 * of times to correct, or try to correct forever. 135 * 1.4 Sparse files (lseek holes) stored on the archive (but stored with blocks 136 * of all zeros will be restored with holes appropriate for the target 137 * filesystem 138 * 1.5 The user is notified whenever something is found during archive 139 * read operations which violates spec (but the read will continue). 140 * 1.6 Multiple archive volumes can be read and may span over different 141 * archive devices 142 * 1.7 Rigidly restores all file attributes exactly as they are stored on the 143 * archive. 144 * 1.8 Modification change time ranges can be specified via multiple -T 145 * options. These allow a user to select files whose modification time 146 * lies within a specific time range. 147 * 1.9 Files can be selected based on owner (user name or uid) via one or more 148 * -U options. 149 * 1.10 Files can be selected based on group (group name or gid) via one o 150 * more -G options. 151 * 1.11 File modification time can be checked against existing file after 152 * name modification (-Z) 153 * 154 * 2 WRITE ENHANCEMENTS 155 * 2.1 Write operation will stop instead of allowing a user to create a flawed 156 * flawed archive (due to any problem). 157 * 2.2 Archives written by pax are forced to strictly conform to both the 158 * archive and pax the spceific format specifications. 159 * 2.3 Blocking size and format is rigidly enforced on writes. 160 * 2.4 Formats which may exhibit header overflow problems (they have fields 161 * too small for large file systems, such as inode number storage), use 162 * routines designed to repair this problem. These techniques still 163 * conform to both pax and format specifications, but no longer truncate 164 * these fields. This removes any restrictions on using these archive 165 * formats on large file systems. 166 * 2.5 Multiple archive volumes can be written and may span over different 167 * archive devices 168 * 2.6 A archive volume record limit allows the user to specify the number 169 * of bytes stored on an archive volume. When reached the user is 170 * prompted for the next archive volume. This is specified with the 171 * non-standard -B flag. The limit is rounded up to the next blocksize. 172 * 2.7 All archive padding during write use zero filled sections. This makes 173 * it much easier to pull data out of flawed archive during read 174 * operations. 175 * 2.8 Access time reset with the -t applies to all file nodes (including 176 * directories). 177 * 2.9 Symbolic links can be followed with -L (optional in the spec). 178 * 2.10 Modification or inode change time ranges can be specified via 179 * multiple -T options. These allow a user to select files whose 180 * modification or inode change time lies within a specific time range. 181 * 2.11 Files can be selected based on owner (user name or uid) via one or more 182 * -U options. 183 * 2.12 Files can be selected based on group (group name or gid) via one o 184 * more -G options. 185 * 2.13 Symlinks which appear on the command line can be followed (without 186 * following other symlinks; -H flag) 187 * 188 * 3 COPY ENHANCEMENTS 189 * 3.1 Sparse files (lseek holes) can be copied without expanding the holes 190 * into zero filled blocks. The file copy is created with holes which are 191 * appropriate for the target filesystem 192 * 3.2 Access time as well as modification time on copied file trees can be 193 * preserved with the appropriate -p options. 194 * 3.3 Access time reset with the -t applies to all file nodes (including 195 * directories). 196 * 3.4 Symbolic links can be followed with -L (optional in the spec). 197 * 3.5 Modification or inode change time ranges can be specified via 198 * multiple -T options. These allow a user to select files whose 199 * modification or inode change time lies within a specific time range. 200 * 3.6 Files can be selected based on owner (user name or uid) via one or more 201 * -U options. 202 * 3.7 Files can be selected based on group (group name or gid) via one o 203 * more -G options. 204 * 3.8 Symlinks which appear on the command line can be followed (without 205 * following other symlinks; -H flag) 206 * 3.9 File inode change time can be checked against existing file before 207 * name modification (-D) 208 * 3.10 File inode change time can be checked against existing file after 209 * name modification (-Y) 210 * 3.11 File modification time can be checked against existing file after 211 * name modification (-Z) 212 * 213 * 4 GENERAL ENHANCEMENTS 214 * 4.1 Internal structure is designed to isolate format dependent and 215 * independent functions. Formats are selected via a format driver table. 216 * This encourages the addition of new archive formats by only having to 217 * write those routines which id, read and write the archive header. 218 */ 219 220 /* 221 * main() 222 * parse options, set up and operate as specified by the user. 223 * any operational flaw will set exit_val to non-zero 224 * Return: 0 if ok, 1 otherwise 225 */ 226 227 int 228 main(int argc, char **argv) 229 { 230 /* 231 * parse options, determine operational mode, general init 232 */ 233 options(argc, argv); 234 if ((gen_init() < 0) || (tty_init() < 0)) 235 return(exit_val); 236 237 (void)time(&starttime); 238 #ifdef SIGINFO 239 (void)signal(SIGINFO, ar_summary); 240 #endif 241 /* 242 * select a primary operation mode 243 */ 244 switch(act) { 245 case EXTRACT: 246 extract(); 247 break; 248 case ARCHIVE: 249 archive(); 250 break; 251 case APPND: 252 append(); 253 break; 254 case COPY: 255 copy(); 256 break; 257 default: 258 case LIST: 259 list(); 260 break; 261 } 262 return(exit_val); 263 } 264 265 /* 266 * sig_cleanup() 267 * when interrupted we try to do whatever delayed processing we can. 268 * This is not critical, but we really ought to limit our damage when we 269 * are aborted by the user. 270 * Return: 271 * never.... 272 */ 273 274 void 275 sig_cleanup(int which_sig) 276 { 277 /* 278 * restore modes and times for any dirs we may have created 279 * or any dirs we may have read. Set vflag and vfpart so the user 280 * will clearly see the message on a line by itself. 281 */ 282 vflag = vfpart = 1; 283 if (which_sig == SIGXCPU) 284 tty_warn(0, "Cpu time limit reached, cleaning up."); 285 else 286 tty_warn(0, "Signal caught, cleaning up."); 287 288 ar_close(); 289 proc_dir(); 290 if (tflag) 291 atdir_end(); 292 exit(1); 293 /* NOTREACHED */ 294 } 295 296 /* 297 * gen_init() 298 * general setup routines. Not all are required, but they really help 299 * when dealing with a medium to large sized archives. 300 */ 301 302 static int 303 gen_init(void) 304 { 305 struct rlimit reslimit; 306 struct sigaction n_hand; 307 struct sigaction o_hand; 308 309 /* 310 * Really needed to handle large archives. We can run out of memory for 311 * internal tables really fast when we have a whole lot of files... 312 */ 313 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){ 314 reslimit.rlim_cur = reslimit.rlim_max; 315 (void)setrlimit(RLIMIT_DATA , &reslimit); 316 } 317 318 /* 319 * should file size limits be waived? if the os limits us, this is 320 * needed if we want to write a large archive 321 */ 322 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){ 323 reslimit.rlim_cur = reslimit.rlim_max; 324 (void)setrlimit(RLIMIT_FSIZE , &reslimit); 325 } 326 327 /* 328 * increase the size the stack can grow to 329 */ 330 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){ 331 reslimit.rlim_cur = reslimit.rlim_max; 332 (void)setrlimit(RLIMIT_STACK , &reslimit); 333 } 334 335 #ifdef RLIMIT_RSS 336 /* 337 * not really needed, but doesn't hurt 338 */ 339 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){ 340 reslimit.rlim_cur = reslimit.rlim_max; 341 (void)setrlimit(RLIMIT_RSS , &reslimit); 342 } 343 #endif 344 345 /* 346 * signal handling to reset stored directory times and modes. Since 347 * we deal with broken pipes via failed writes we ignore it. We also 348 * deal with any file size limit thorugh failed writes. Cpu time 349 * limits are caught and a cleanup is forced. 350 */ 351 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) || 352 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) || 353 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) || 354 (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) { 355 tty_warn(1, "Unable to set up signal mask"); 356 return(-1); 357 } 358 n_hand.sa_mask = s_mask; 359 n_hand.sa_flags = 0; 360 n_hand.sa_handler = sig_cleanup; 361 362 if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) && 363 (o_hand.sa_handler == SIG_IGN) && 364 (sigaction(SIGHUP, &o_hand, &o_hand) < 0)) 365 goto out; 366 367 if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) && 368 (o_hand.sa_handler == SIG_IGN) && 369 (sigaction(SIGTERM, &o_hand, &o_hand) < 0)) 370 goto out; 371 372 if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) && 373 (o_hand.sa_handler == SIG_IGN) && 374 (sigaction(SIGINT, &o_hand, &o_hand) < 0)) 375 goto out; 376 377 if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) && 378 (o_hand.sa_handler == SIG_IGN) && 379 (sigaction(SIGQUIT, &o_hand, &o_hand) < 0)) 380 goto out; 381 382 if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) && 383 (o_hand.sa_handler == SIG_IGN) && 384 (sigaction(SIGXCPU, &o_hand, &o_hand) < 0)) 385 goto out; 386 387 n_hand.sa_handler = SIG_IGN; 388 if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) || 389 (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0)) 390 goto out; 391 return(0); 392 393 out: 394 syswarn(1, errno, "Unable to set up signal handler"); 395 return(-1); 396 } 397