xref: /freebsd/bin/pax/pax.c (revision b740c88bfb6453416926271c089262e7164dace3)
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  * 4. 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 
34 #if 0
35 #ifndef lint
36 static char const copyright[] =
37 "@(#) Copyright (c) 1992, 1993\n\
38 	The Regents of the University of California.  All rights reserved.\n";
39 #endif /* not lint */
40 
41 #ifndef lint
42 static char sccsid[] = "@(#)pax.c	8.2 (Berkeley) 4/18/94";
43 #endif /* not lint */
44 #endif
45 #include <sys/cdefs.h>
46 __FBSDID("$FreeBSD$");
47 
48 #include <sys/types.h>
49 #include <sys/stat.h>
50 #include <sys/time.h>
51 #include <sys/resource.h>
52 #include <err.h>
53 #include <errno.h>
54 #include <fcntl.h>
55 #include <locale.h>
56 #include <paths.h>
57 #include <signal.h>
58 #include <stdio.h>
59 #include <stdlib.h>
60 #include <string.h>
61 #include <unistd.h>
62 #include "pax.h"
63 #include "extern.h"
64 static int gen_init(void);
65 
66 /*
67  * PAX main routines, general globals and some simple start up routines
68  */
69 
70 /*
71  * Variables that can be accessed by any routine within pax
72  */
73 int	act = DEFOP;		/* read/write/append/copy */
74 FSUB	*frmt = NULL;		/* archive format type */
75 int	cflag;			/* match all EXCEPT pattern/file */
76 int	cwdfd;			/* starting cwd */
77 int	dflag;			/* directory member match only  */
78 int	iflag;			/* interactive file/archive rename */
79 int	kflag;			/* do not overwrite existing files */
80 int	lflag;			/* use hard links when possible */
81 int	nflag;			/* select first archive member match */
82 int	tflag;			/* restore access time after read */
83 int	uflag;			/* ignore older modification time files */
84 int	vflag;			/* produce verbose output */
85 int	Dflag;			/* same as uflag except inode change time */
86 int	Hflag;			/* follow command line symlinks (write only) */
87 int	Lflag;			/* follow symlinks when writing */
88 int	Oflag;			/* limit to single volume */
89 int	Xflag;			/* archive files with same device id only */
90 int	Yflag;			/* same as Dflg except after name mode */
91 int	Zflag;			/* same as uflg except after name mode */
92 int	vfpart;			/* is partial verbose output in progress */
93 int	patime = 1;		/* preserve file access time */
94 int	pmtime = 1;		/* preserve file modification times */
95 int	nodirs;			/* do not create directories as needed */
96 int	pmode;			/* preserve file mode bits */
97 int	pids;			/* preserve file uid/gid */
98 int	rmleadslash = 0;	/* remove leading '/' from pathnames */
99 int	exit_val;		/* exit value */
100 int	docrc;			/* check/create file crc */
101 char	*dirptr;		/* destination dir in a copy */
102 const	char *argv0;		/* root of argv[0] */
103 sigset_t s_mask;		/* signal mask for cleanup critical sect */
104 FILE	*listf;			/* file pointer to print file list to */
105 char	*tempfile;		/* tempfile to use for mkstemp(3) */
106 char	*tempbase;		/* basename of tempfile to use for mkstemp(3) */
107 
108 /*
109  *	PAX - Portable Archive Interchange
110  *
111  * 	A utility to read, write, and write lists of the members of archive
112  *	files and copy directory hierarchies. A variety of archive formats
113  *	are supported (some are described in POSIX 1003.1 10.1):
114  *
115  *		ustar - 10.1.1 extended tar interchange format
116  *		cpio  - 10.1.2 extended cpio interchange format
117  *		tar - old BSD 4.3 tar format
118  *		binary cpio - old cpio with binary header format
119  *		sysVR4 cpio -  with and without CRC
120  *
121  * This version is a superset of IEEE Std 1003.2b-d3
122  *
123  * Summary of Extensions to the IEEE Standard:
124  *
125  * 1	READ ENHANCEMENTS
126  * 1.1	Operations which read archives will continue to operate even when
127  *	processing archives which may be damaged, truncated, or fail to meet
128  *	format specs in several different ways. Damaged sections of archives
129  *	are detected and avoided if possible. Attempts will be made to resync
130  *	archive read operations even with badly damaged media.
131  * 1.2	Blocksize requirements are not strictly enforced on archive read.
132  *	Tapes which have variable sized records can be read without errors.
133  * 1.3	The user can specify via the non-standard option flag -E if error
134  *	resync operation should stop on a media error, try a specified number
135  *	of times to correct, or try to correct forever.
136  * 1.4	Sparse files (lseek holes) stored on the archive (but stored with blocks
137  *	of all zeros will be restored with holes appropriate for the target
138  *	file system
139  * 1.5	The user is notified whenever something is found during archive
140  *	read operations which violates spec (but the read will continue).
141  * 1.6	Multiple archive volumes can be read and may span over different
142  *	archive devices
143  * 1.7	Rigidly restores all file attributes exactly as they are stored on the
144  *	archive.
145  * 1.8	Modification change time ranges can be specified via multiple -T
146  *	options. These allow a user to select files whose modification time
147  *	lies within a specific time range.
148  * 1.9	Files can be selected based on owner (user name or uid) via one or more
149  *	-U options.
150  * 1.10	Files can be selected based on group (group name or gid) via one o
151  *	more -G options.
152  * 1.11	File modification time can be checked against existing file after
153  *	name modification (-Z)
154  *
155  * 2	WRITE ENHANCEMENTS
156  * 2.1	Write operation will stop instead of allowing a user to create a flawed
157  *	flawed archive (due to any problem).
158  * 2.2	Archives written by pax are forced to strictly conform to both the
159  *	archive and pax the specific format specifications.
160  * 2.3	Blocking size and format is rigidly enforced on writes.
161  * 2.4	Formats which may exhibit header overflow problems (they have fields
162  *	too small for large file systems, such as inode number storage), use
163  *	routines designed to repair this problem. These techniques still
164  *	conform to both pax and format specifications, but no longer truncate
165  *	these fields. This removes any restrictions on using these archive
166  *	formats on large file systems.
167  * 2.5	Multiple archive volumes can be written and may span over different
168  *	archive devices
169  * 2.6	A archive volume record limit allows the user to specify the number
170  *	of bytes stored on an archive volume. When reached the user is
171  *	prompted for the next archive volume. This is specified with the
172  *	non-standard -B flag. The limit is rounded up to the next blocksize.
173  * 2.7	All archive padding during write use zero filled sections. This makes
174  *	it much easier to pull data out of flawed archive during read
175  *	operations.
176  * 2.8	Access time reset with the -t applies to all file nodes (including
177  *	directories).
178  * 2.9	Symbolic links can be followed with -L (optional in the spec).
179  * 2.10	Modification or inode change time ranges can be specified via
180  *	multiple -T options. These allow a user to select files whose
181  *	modification or inode change time lies within a specific time range.
182  * 2.11	Files can be selected based on owner (user name or uid) via one or more
183  *	-U options.
184  * 2.12	Files can be selected based on group (group name or gid) via one o
185  *	more -G options.
186  * 2.13	Symlinks which appear on the command line can be followed (without
187  *	following other symlinks; -H flag)
188  *
189  * 3	COPY ENHANCEMENTS
190  * 3.1	Sparse files (lseek holes) can be copied without expanding the holes
191  *	into zero filled blocks. The file copy is created with holes which are
192  *	appropriate for the target file system
193  * 3.2	Access time as well as modification time on copied file trees can be
194  *	preserved with the appropriate -p options.
195  * 3.3	Access time reset with the -t applies to all file nodes (including
196  *	directories).
197  * 3.4	Symbolic links can be followed with -L (optional in the spec).
198  * 3.5	Modification or inode change time ranges can be specified via
199  *	multiple -T options. These allow a user to select files whose
200  *	modification or inode change time lies within a specific time range.
201  * 3.6	Files can be selected based on owner (user name or uid) via one or more
202  *	-U options.
203  * 3.7	Files can be selected based on group (group name or gid) via one o
204  *	more -G options.
205  * 3.8	Symlinks which appear on the command line can be followed (without
206  *	following other symlinks; -H flag)
207  * 3.9  File inode change time can be checked against existing file before
208  *	name modification (-D)
209  * 3.10 File inode change time can be checked against existing file after
210  *	name modification (-Y)
211  * 3.11	File modification time can be checked against existing file after
212  *	name modification (-Z)
213  *
214  * 4	GENERAL ENHANCEMENTS
215  * 4.1	Internal structure is designed to isolate format dependent and
216  *	independent functions. Formats are selected via a format driver table.
217  *	This encourages the addition of new archive formats by only having to
218  *	write those routines which id, read and write the archive header.
219  */
220 
221 /*
222  * main()
223  *	parse options, set up and operate as specified by the user.
224  *	any operational flaw will set exit_val to non-zero
225  * Return: 0 if ok, 1 otherwise
226  */
227 
228 int
229 main(int argc, char *argv[])
230 {
231 	const char *tmpdir;
232 	size_t tdlen;
233 
234 	(void) setlocale(LC_ALL, "");
235 	listf = stderr;
236 	/*
237 	 * Keep a reference to cwd, so we can always come back home.
238 	 */
239 	cwdfd = open(".", O_RDONLY);
240 	if (cwdfd < 0) {
241 		syswarn(0, errno, "Can't open current working directory.");
242 		return(exit_val);
243 	}
244 
245 	/*
246 	 * Where should we put temporary files?
247 	 */
248 	if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0')
249 		tmpdir = _PATH_TMP;
250 	tdlen = strlen(tmpdir);
251 	while(tdlen > 0 && tmpdir[tdlen - 1] == '/')
252 		tdlen--;
253 	tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE));
254 	if (tempfile == NULL) {
255 		paxwarn(1, "Cannot allocate memory for temp file name.");
256 		return(exit_val);
257 	}
258 	if (tdlen)
259 		memcpy(tempfile, tmpdir, tdlen);
260 	tempbase = tempfile + tdlen;
261 	*tempbase++ = '/';
262 
263 	/*
264 	 * parse options, determine operational mode, general init
265 	 */
266 	options(argc, argv);
267 	if ((gen_init() < 0) || (tty_init() < 0))
268 		return(exit_val);
269 
270 	/*
271 	 * select a primary operation mode
272 	 */
273 	switch(act) {
274 	case EXTRACT:
275 		extract();
276 		break;
277 	case ARCHIVE:
278 		archive();
279 		break;
280 	case APPND:
281 		if (gzip_program != NULL)
282 			err(1, "can not gzip while appending");
283 		append();
284 		break;
285 	case COPY:
286 		copy();
287 		break;
288 	default:
289 	case LIST:
290 		list();
291 		break;
292 	}
293 	return(exit_val);
294 }
295 
296 /*
297  * sig_cleanup()
298  *	when interrupted we try to do whatever delayed processing we can.
299  *	This is not critical, but we really ought to limit our damage when we
300  *	are aborted by the user.
301  * Return:
302  *	never....
303  */
304 
305 void
306 sig_cleanup(int which_sig)
307 {
308 	/*
309 	 * restore modes and times for any dirs we may have created
310 	 * or any dirs we may have read. Set vflag and vfpart so the user
311 	 * will clearly see the message on a line by itself.
312 	 */
313 	vflag = vfpart = 1;
314 	if (which_sig == SIGXCPU)
315 		paxwarn(0, "Cpu time limit reached, cleaning up.");
316 	else
317 		paxwarn(0, "Signal caught, cleaning up.");
318 
319 	ar_close();
320 	proc_dir();
321 	if (tflag)
322 		atdir_end();
323 	exit(1);
324 }
325 
326 /*
327  * gen_init()
328  *	general setup routines. Not all are required, but they really help
329  *	when dealing with a medium to large sized archives.
330  */
331 
332 static int
333 gen_init(void)
334 {
335 	struct rlimit reslimit;
336 	struct sigaction n_hand;
337 	struct sigaction o_hand;
338 
339 	/*
340 	 * Really needed to handle large archives. We can run out of memory for
341 	 * internal tables really fast when we have a whole lot of files...
342 	 */
343 	if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
344 		reslimit.rlim_cur = reslimit.rlim_max;
345 		(void)setrlimit(RLIMIT_DATA , &reslimit);
346 	}
347 
348 	/*
349 	 * should file size limits be waived? if the os limits us, this is
350 	 * needed if we want to write a large archive
351 	 */
352 	if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
353 		reslimit.rlim_cur = reslimit.rlim_max;
354 		(void)setrlimit(RLIMIT_FSIZE , &reslimit);
355 	}
356 
357 	/*
358 	 * increase the size the stack can grow to
359 	 */
360 	if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
361 		reslimit.rlim_cur = reslimit.rlim_max;
362 		(void)setrlimit(RLIMIT_STACK , &reslimit);
363 	}
364 
365 	/*
366 	 * not really needed, but doesn't hurt
367 	 */
368 	if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
369 		reslimit.rlim_cur = reslimit.rlim_max;
370 		(void)setrlimit(RLIMIT_RSS , &reslimit);
371 	}
372 
373 	/*
374 	 * signal handling to reset stored directory times and modes. Since
375 	 * we deal with broken pipes via failed writes we ignore it. We also
376 	 * deal with any file size limit thorough failed writes. Cpu time
377 	 * limits are caught and a cleanup is forced.
378 	 */
379 	if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
380 	    (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
381 	    (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
382 	    (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
383 		paxwarn(1, "Unable to set up signal mask");
384 		return(-1);
385 	}
386 	memset(&n_hand, 0, sizeof n_hand);
387 	n_hand.sa_mask = s_mask;
388 	n_hand.sa_flags = 0;
389 	n_hand.sa_handler = sig_cleanup;
390 
391 	if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) &&
392 	    (o_hand.sa_handler == SIG_IGN) &&
393 	    (sigaction(SIGHUP, &o_hand, &o_hand) < 0))
394 		goto out;
395 
396 	if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) &&
397 	    (o_hand.sa_handler == SIG_IGN) &&
398 	    (sigaction(SIGTERM, &o_hand, &o_hand) < 0))
399 		goto out;
400 
401 	if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) &&
402 	    (o_hand.sa_handler == SIG_IGN) &&
403 	    (sigaction(SIGINT, &o_hand, &o_hand) < 0))
404 		goto out;
405 
406 	if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) &&
407 	    (o_hand.sa_handler == SIG_IGN) &&
408 	    (sigaction(SIGQUIT, &o_hand, &o_hand) < 0))
409 		goto out;
410 
411 	if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) &&
412 	    (o_hand.sa_handler == SIG_IGN) &&
413 	    (sigaction(SIGXCPU, &o_hand, &o_hand) < 0))
414 		goto out;
415 
416 	n_hand.sa_handler = SIG_IGN;
417 	if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) ||
418 	    (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0))
419 		goto out;
420 	return(0);
421 
422     out:
423 	syswarn(1, errno, "Unable to set up signal handler");
424 	return(-1);
425 }
426