1 /*-
2 * Copyright (c) 1992 Keith Muller.
3 * Copyright (c) 1992, 1993
4 * The Regents of the University of California. All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * Keith Muller of the University of California, San Diego.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#) Copyright (c) 1992, 1993 The Regents of the University of California. All rights reserved.
34 * @(#)pax.c 8.2 (Berkeley) 4/18/94
35 * $FreeBSD: src/bin/pax/pax.c,v 1.13.2.4 2002/11/07 15:29:53 imp Exp $
36 */
37
38 #include <sys/types.h>
39 #include <sys/stat.h>
40 #include <sys/time.h>
41 #include <sys/resource.h>
42 #include <err.h>
43 #include <errno.h>
44 #include <fcntl.h>
45 #include <locale.h>
46 #include <paths.h>
47 #include <signal.h>
48 #include <stdio.h>
49 #include <stdlib.h>
50 #include <string.h>
51 #include <unistd.h>
52 #include "pax.h"
53 #include "extern.h"
54 static int gen_init(void);
55
56 /*
57 * PAX main routines, general globals and some simple start up routines
58 */
59
60 /*
61 * Variables that can be accessed by any routine within pax
62 */
63 int act = DEFOP; /* read/write/append/copy */
64 FSUB *frmt = NULL; /* archive format type */
65 int cflag; /* match all EXCEPT pattern/file */
66 int cwdfd; /* starting cwd */
67 int dflag; /* directory member match only */
68 int iflag; /* interactive file/archive rename */
69 int kflag; /* do not overwrite existing files */
70 int lflag; /* use hard links when possible */
71 int nflag; /* select first archive member match */
72 int tflag; /* restore access time after read */
73 int uflag; /* ignore older modification time files */
74 int vflag; /* produce verbose output */
75 int Dflag; /* same as uflag except inode change time */
76 int Hflag; /* follow command line symlinks (write only) */
77 int Lflag; /* follow symlinks when writing */
78 int Xflag; /* archive files with same device id only */
79 int Yflag; /* same as Dflag except after name mode */
80 int Zflag; /* same as uflag except after name mode */
81 int vfpart; /* is partial verbose output in progress */
82 int patime = 1; /* preserve file access time */
83 int pmtime = 1; /* preserve file modification times */
84 int nodirs; /* do not create directories as needed */
85 int pmode; /* preserve file mode bits */
86 int pids; /* preserve file uid/gid */
87 int rmleadslash = 0; /* remove leading '/' from pathnames */
88 int exit_val; /* exit value */
89 int docrc; /* check/create file crc */
90 char *dirptr; /* destination dir in a copy */
91 char *argv0; /* root of argv[0] */
92 sigset_t s_mask; /* signal mask for cleanup critical sect */
93 FILE *listf; /* file pointer to print file list to */
94 char *tempfile; /* tempfile to use for mkstemp(3) */
95 char *tempbase; /* basename of tempfile to use for mkstemp(3) */
96
97 /*
98 * PAX - Portable Archive Interchange
99 *
100 * A utility to read, write, and write lists of the members of archive
101 * files and copy directory hierarchies. A variety of archive formats
102 * are supported (some are described in POSIX 1003.1 10.1):
103 *
104 * ustar - 10.1.1 extended tar interchange format
105 * cpio - 10.1.2 extended cpio interchange format
106 * tar - old BSD 4.3 tar format
107 * binary cpio - old cpio with binary header format
108 * sysVR4 cpio - with and without CRC
109 *
110 * This version is a superset of IEEE Std 1003.2b-d3
111 *
112 * Summary of Extensions to the IEEE Standard:
113 *
114 * 1 READ ENHANCEMENTS
115 * 1.1 Operations which read archives will continue to operate even when
116 * processing archives which may be damaged, truncated, or fail to meet
117 * format specs in several different ways. Damaged sections of archives
118 * are detected and avoided if possible. Attempts will be made to resync
119 * archive read operations even with badly damaged media.
120 * 1.2 Blocksize requirements are not strictly enforced on archive read.
121 * Tapes which have variable sized records can be read without errors.
122 * 1.3 The user can specify via the non-standard option flag -E if error
123 * resync operation should stop on a media error, try a specified number
124 * of times to correct, or try to correct forever.
125 * 1.4 Sparse files (lseek holes) stored on the archive (but stored with blocks
126 * of all zeros will be restored with holes appropriate for the target
127 * filesystem
128 * 1.5 The user is notified whenever something is found during archive
129 * read operations which violates spec (but the read will continue).
130 * 1.6 Multiple archive volumes can be read and may span over different
131 * archive devices
132 * 1.7 Rigidly restores all file attributes exactly as they are stored on the
133 * archive.
134 * 1.8 Modification change time ranges can be specified via multiple -T
135 * options. These allow a user to select files whose modification time
136 * lies within a specific time range.
137 * 1.9 Files can be selected based on owner (user name or uid) via one or more
138 * -U options.
139 * 1.10 Files can be selected based on group (group name or gid) via one o
140 * more -G options.
141 * 1.11 File modification time can be checked against existing file after
142 * name modification (-Z)
143 *
144 * 2 WRITE ENHANCEMENTS
145 * 2.1 Write operation will stop instead of allowing a user to create a flawed
146 * flawed archive (due to any problem).
147 * 2.2 Archives written by pax are forced to strictly conform to both the
148 * archive and pax the specific format specifications.
149 * 2.3 Blocking size and format is rigidly enforced on writes.
150 * 2.4 Formats which may exhibit header overflow problems (they have fields
151 * too small for large file systems, such as inode number storage), use
152 * routines designed to repair this problem. These techniques still
153 * conform to both pax and format specifications, but no longer truncate
154 * these fields. This removes any restrictions on using these archive
155 * formats on large file systems.
156 * 2.5 Multiple archive volumes can be written and may span over different
157 * archive devices
158 * 2.6 A archive volume record limit allows the user to specify the number
159 * of bytes stored on an archive volume. When reached the user is
160 * prompted for the next archive volume. This is specified with the
161 * non-standard -B flag. The limit is rounded up to the next blocksize.
162 * 2.7 All archive padding during write use zero filled sections. This makes
163 * it much easier to pull data out of flawed archive during read
164 * operations.
165 * 2.8 Access time reset with the -t applies to all file nodes (including
166 * directories).
167 * 2.9 Symbolic links can be followed with -L (optional in the spec).
168 * 2.10 Modification or inode change time ranges can be specified via
169 * multiple -T options. These allow a user to select files whose
170 * modification or inode change time lies within a specific time range.
171 * 2.11 Files can be selected based on owner (user name or uid) via one or more
172 * -U options.
173 * 2.12 Files can be selected based on group (group name or gid) via one o
174 * more -G options.
175 * 2.13 Symlinks which appear on the command line can be followed (without
176 * following other symlinks; -H flag)
177 *
178 * 3 COPY ENHANCEMENTS
179 * 3.1 Sparse files (lseek holes) can be copied without expanding the holes
180 * into zero filled blocks. The file copy is created with holes which are
181 * appropriate for the target filesystem
182 * 3.2 Access time as well as modification time on copied file trees can be
183 * preserved with the appropriate -p options.
184 * 3.3 Access time reset with the -t applies to all file nodes (including
185 * directories).
186 * 3.4 Symbolic links can be followed with -L (optional in the spec).
187 * 3.5 Modification or inode change time ranges can be specified via
188 * multiple -T options. These allow a user to select files whose
189 * modification or inode change time lies within a specific time range.
190 * 3.6 Files can be selected based on owner (user name or uid) via one or more
191 * -U options.
192 * 3.7 Files can be selected based on group (group name or gid) via one o
193 * more -G options.
194 * 3.8 Symlinks which appear on the command line can be followed (without
195 * following other symlinks; -H flag)
196 * 3.9 File inode change time can be checked against existing file before
197 * name modification (-D)
198 * 3.10 File inode change time can be checked against existing file after
199 * name modification (-Y)
200 * 3.11 File modification time can be checked against existing file after
201 * name modification (-Z)
202 *
203 * 4 GENERAL ENHANCEMENTS
204 * 4.1 Internal structure is designed to isolate format dependent and
205 * independent functions. Formats are selected via a format driver table.
206 * This encourages the addition of new archive formats by only having to
207 * write those routines which id, read and write the archive header.
208 */
209
210 /*
211 * main()
212 * parse options, set up and operate as specified by the user.
213 * any operational flaw will set exit_val to non-zero
214 * Return: 0 if ok, 1 otherwise
215 */
216
217 int
main(int argc,char * argv[])218 main(int argc, char *argv[])
219 {
220 const char *tmpdir;
221 size_t tdlen;
222
223 setlocale(LC_ALL, "");
224 listf = stderr;
225 /*
226 * Keep a reference to cwd, so we can always come back home.
227 */
228 cwdfd = open(".", O_RDONLY | O_CLOEXEC);
229 if (cwdfd < 0) {
230 syswarn(0, errno, "Can't open current working directory.");
231 return(exit_val);
232 }
233
234 /*
235 * Where should we put temporary files?
236 */
237 if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0')
238 tmpdir = _PATH_TMP;
239 tdlen = strlen(tmpdir);
240 while (tdlen > 0 && tmpdir[tdlen - 1] == '/')
241 tdlen--;
242 tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE));
243 if (tempfile == NULL) {
244 paxwarn(1, "Cannot allocate memory for temp file name.");
245 return(exit_val);
246 }
247 if (tdlen)
248 memcpy(tempfile, tmpdir, tdlen);
249 tempbase = tempfile + tdlen;
250 *tempbase++ = '/';
251
252 /*
253 * parse options, determine operational mode, general init
254 */
255 options(argc, argv);
256 if ((gen_init() < 0) || (tty_init() < 0))
257 return(exit_val);
258
259 /*
260 * select a primary operation mode
261 */
262 switch (act) {
263 case EXTRACT:
264 extract();
265 break;
266 case ARCHIVE:
267 archive();
268 break;
269 case APPND:
270 if (gzip_program != NULL)
271 err(1, "can not gzip while appending");
272 append();
273 break;
274 case COPY:
275 copy();
276 break;
277 default:
278 case LIST:
279 list();
280 break;
281 }
282 return(exit_val);
283 }
284
285 /*
286 * sig_cleanup()
287 * when interrupted we try to do whatever delayed processing we can.
288 * This is not critical, but we really ought to limit our damage when we
289 * are aborted by the user.
290 * Return:
291 * never....
292 */
293
294 void
sig_cleanup(int which_sig)295 sig_cleanup(int which_sig)
296 {
297 /*
298 * restore modes and times for any dirs we may have created
299 * or any dirs we may have read. Set vflag and vfpart so the user
300 * will clearly see the message on a line by itself.
301 */
302 vflag = vfpart = 1;
303 if (which_sig == SIGXCPU)
304 paxwarn(0, "Cpu time limit reached, cleaning up.");
305 else
306 paxwarn(0, "Signal caught, cleaning up.");
307
308 ar_close();
309 proc_dir();
310 if (tflag)
311 atdir_end();
312 exit(1);
313 }
314
315 /*
316 * setup_sig()
317 * set a signal to be caught, but only if it isn't being ignored already
318 */
319
320 static int
setup_sig(int sig,const struct sigaction * n_hand)321 setup_sig(int sig, const struct sigaction *n_hand)
322 {
323 struct sigaction o_hand;
324
325 if (sigaction(sig, NULL, &o_hand) < 0)
326 return (-1);
327
328 if (o_hand.sa_handler == SIG_IGN)
329 return (0);
330
331 return (sigaction(sig, n_hand, NULL));
332 }
333
334 /*
335 * gen_init()
336 * general setup routines. Not all are required, but they really help
337 * when dealing with a medium to large sized archives.
338 */
339
340 static int
gen_init(void)341 gen_init(void)
342 {
343 struct rlimit reslimit;
344 struct sigaction n_hand;
345
346 /*
347 * Really needed to handle large archives. We can run out of memory for
348 * internal tables really fast when we have a whole lot of files...
349 */
350 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
351 reslimit.rlim_cur = reslimit.rlim_max;
352 setrlimit(RLIMIT_DATA , &reslimit);
353 }
354
355 /*
356 * should file size limits be waived? if the os limits us, this is
357 * needed if we want to write a large archive
358 */
359 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
360 reslimit.rlim_cur = reslimit.rlim_max;
361 setrlimit(RLIMIT_FSIZE , &reslimit);
362 }
363
364 /*
365 * increase the size the stack can grow to
366 */
367 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
368 reslimit.rlim_cur = reslimit.rlim_max;
369 setrlimit(RLIMIT_STACK , &reslimit);
370 }
371
372 /*
373 * not really needed, but doesn't hurt
374 */
375 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
376 reslimit.rlim_cur = reslimit.rlim_max;
377 setrlimit(RLIMIT_RSS , &reslimit);
378 }
379
380 /*
381 * signal handling to reset stored directory times and modes. Since
382 * we deal with broken pipes via failed writes we ignore it. We also
383 * deal with any file size limit through failed writes. Cpu time
384 * limits are caught and a cleanup is forced.
385 */
386 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
387 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
388 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
389 (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
390 paxwarn(1, "Unable to set up signal mask");
391 return(-1);
392 }
393 memset(&n_hand, 0, sizeof n_hand);
394 n_hand.sa_mask = s_mask;
395 n_hand.sa_flags = 0;
396 n_hand.sa_handler = sig_cleanup;
397
398 if (setup_sig(SIGHUP, &n_hand) ||
399 setup_sig(SIGTERM, &n_hand) ||
400 setup_sig(SIGINT, &n_hand) ||
401 setup_sig(SIGQUIT, &n_hand) ||
402 setup_sig(SIGXCPU, &n_hand))
403 goto out;
404
405 n_hand.sa_handler = SIG_IGN;
406 if ((sigaction(SIGPIPE, &n_hand, NULL) < 0) ||
407 (sigaction(SIGXFSZ, &n_hand, NULL) < 0))
408 goto out;
409 return(0);
410
411 out:
412 syswarn(1, errno, "Unable to set up signal handler");
413 return(-1);
414 }
415