1 /* vi: set sw=4 ts=4: */
3 * Mini tar implementation for busybox
5 * Modified to use common extraction code used by ar, cpio, dpkg-deb, dpkg
8 * Note, that as of BusyBox-0.43, tar has been completely rewritten from the
9 * ground up. It still has remnants of the old code lying about, but it is
10 * very different now (i.e., cleaner, less global variables, etc.)
12 * Copyright (C) 1999-2004 by Erik Andersen <andersen@codepoet.org>
14 * Based in part in the tar implementation in sash
15 * Copyright (c) 1999 by David I. Bell
16 * Permission is granted to use, distribute, or modify this source,
17 * provided that this copyright notice remains intact.
18 * Permission to distribute sash derived code under GPL has been granted.
20 * Based in part on the tar implementation from busybox-0.28
21 * Copyright (C) 1995 Bruce Perens
23 * Licensed under GPLv2 or later, see file LICENSE in this source tree.
25 /* TODO: security with -C DESTDIR option can be enhanced.
26 * Consider tar file created via:
27 * $ tar cvf bug.tar anything.txt
28 * $ ln -s /tmp symlink
29 * $ tar --append -f bug.tar symlink
32 * $ tar --append -f bug.tar symlink/evil.py
34 * This will result in an archive which contains:
35 * $ tar --list -f bug.tar
40 * Untarring it puts evil.py in '/tmp' even if the -C DESTDIR is given.
41 * This doesn't feel right, and IIRC GNU tar doesn't do that.
48 //config: tar is an archiving program. It's commonly used with gzip to
49 //config: create compressed archives. It's probably the most widely used
50 //config: UNIX archive program.
52 //config:config FEATURE_TAR_CREATE
53 //config: bool "Enable archive creation"
55 //config: depends on TAR
57 //config: If you enable this option you'll be able to create
58 //config: tar archives using the `-c' option.
60 //config:config FEATURE_TAR_AUTODETECT
61 //config: bool "Autodetect compressed tarballs"
63 //config: depends on TAR && (FEATURE_SEAMLESS_Z || FEATURE_SEAMLESS_GZ || FEATURE_SEAMLESS_BZ2 || FEATURE_SEAMLESS_LZMA || FEATURE_SEAMLESS_XZ)
65 //config: With this option tar can automatically detect compressed
66 //config: tarballs. Currently it works only on files (not pipes etc).
68 //config:config FEATURE_TAR_FROM
69 //config: bool "Enable -X (exclude from) and -T (include from) options)"
71 //config: depends on TAR
73 //config: If you enable this option you'll be able to specify
74 //config: a list of files to include or exclude from an archive.
76 //config:config FEATURE_TAR_OLDGNU_COMPATIBILITY
77 //config: bool "Support for old tar header format"
79 //config: depends on TAR || DPKG
81 //config: This option is required to unpack archives created in
82 //config: the old GNU format; help to kill this old format by
83 //config: repacking your ancient archives with the new format.
85 //config:config FEATURE_TAR_OLDSUN_COMPATIBILITY
86 //config: bool "Enable untarring of tarballs with checksums produced by buggy Sun tar"
88 //config: depends on TAR || DPKG
90 //config: This option is required to unpack archives created by some old
91 //config: version of Sun's tar (it was calculating checksum using signed
92 //config: arithmetic). It is said to be fixed in newer Sun tar, but "old"
93 //config: tarballs still exist.
95 //config:config FEATURE_TAR_GNU_EXTENSIONS
96 //config: bool "Support for GNU tar extensions (long filenames)"
98 //config: depends on TAR || DPKG
100 //config: With this option busybox supports GNU long filenames and
103 //config:config FEATURE_TAR_LONG_OPTIONS
104 //config: bool "Enable long options"
106 //config: depends on TAR && LONG_OPTS
108 //config: Enable use of long options, increases size by about 400 Bytes
110 //config:config FEATURE_TAR_TO_COMMAND
111 //config: bool "Support for writing to an external program"
113 //config: depends on TAR && FEATURE_TAR_LONG_OPTIONS
115 //config: If you enable this option you'll be able to instruct tar to send
116 //config: the contents of each extracted file to the standard input of an
117 //config: external program.
119 //config:config FEATURE_TAR_UNAME_GNAME
120 //config: bool "Enable use of user and group names"
122 //config: depends on TAR
124 //config: Enables use of user and group names in tar. This affects contents
125 //config: listings (-t) and preserving permissions when unpacking (-p).
126 //config: +200 bytes.
128 //config:config FEATURE_TAR_NOPRESERVE_TIME
129 //config: bool "Enable -m (do not preserve time) option"
131 //config: depends on TAR
133 //config: With this option busybox supports GNU tar -m
134 //config: (do not preserve time) option.
136 //config:config FEATURE_TAR_SELINUX
137 //config: bool "Support for extracting SELinux labels"
139 //config: depends on TAR && SELINUX
141 //config: With this option busybox supports restoring SELinux labels
142 //config: when extracting files from tar archives.
144 //applet:IF_TAR(APPLET(tar, BB_DIR_BIN, BB_SUID_DROP))
145 //kbuild:lib-$(CONFIG_TAR) += tar.o
149 #include "bb_archive.h"
150 /* FIXME: Stop using this non-standard feature */
151 #ifndef FNM_LEADING_DIR
152 # define FNM_LEADING_DIR 0
156 # define DBG(fmt, ...) bb_error_msg("%s: " fmt, __func__, ## __VA_ARGS__)
158 # define DBG(...) ((void)0)
160 #define DBG_OPTION_PARSING 0
163 #define block_buf bb_common_bufsiz1
166 #if ENABLE_FEATURE_TAR_CREATE
169 ** writeTarFile(), writeFileToTarball(), and writeTarHeader() are
170 ** the only functions that deal with the HardLinkInfo structure.
171 ** Even these functions use the xxxHardLinkInfo() functions.
173 typedef struct HardLinkInfo {
174 struct HardLinkInfo *next; /* Next entry in list */
175 dev_t dev; /* Device number */
176 ino_t ino; /* Inode number */
177 // short linkCount; /* (Hard) Link Count */
178 char name[1]; /* Start of filename (must be last) */
181 /* Some info to be carried along when creating a new tarball */
182 typedef struct TarBallInfo {
183 int tarFd; /* Open-for-write file descriptor
185 int verboseFlag; /* Whether to print extra stuff or not */
186 const llist_t *excludeList; /* List of files to not include */
187 HardLinkInfo *hlInfoHead; /* Hard Link Tracking Information */
188 HardLinkInfo *hlInfo; /* Hard Link Info for the current file */
189 //TODO: save only st_dev + st_ino
190 struct stat tarFileStatBuf; /* Stat info for the tarball, letting
191 * us know the inode and device that the
192 * tarball lives, so we can avoid trying
193 * to include the tarball into itself */
196 /* A nice enum with all the possible tar file content types */
198 REGTYPE = '0', /* regular file */
199 REGTYPE0 = '\0', /* regular file (ancient bug compat) */
200 LNKTYPE = '1', /* hard link */
201 SYMTYPE = '2', /* symbolic link */
202 CHRTYPE = '3', /* character special */
203 BLKTYPE = '4', /* block special */
204 DIRTYPE = '5', /* directory */
205 FIFOTYPE = '6', /* FIFO special */
206 CONTTYPE = '7', /* reserved */
207 GNULONGLINK = 'K', /* GNU long (>100 chars) link name */
208 GNULONGNAME = 'L', /* GNU long (>100 chars) file name */
211 /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */
212 static void addHardLinkInfo(HardLinkInfo **hlInfoHeadPtr,
213 struct stat *statbuf,
214 const char *fileName)
216 /* Note: hlInfoHeadPtr can never be NULL! */
217 HardLinkInfo *hlInfo;
219 hlInfo = xmalloc(sizeof(HardLinkInfo) + strlen(fileName));
220 hlInfo->next = *hlInfoHeadPtr;
221 *hlInfoHeadPtr = hlInfo;
222 hlInfo->dev = statbuf->st_dev;
223 hlInfo->ino = statbuf->st_ino;
224 // hlInfo->linkCount = statbuf->st_nlink;
225 strcpy(hlInfo->name, fileName);
228 static void freeHardLinkInfo(HardLinkInfo **hlInfoHeadPtr)
230 HardLinkInfo *hlInfo;
231 HardLinkInfo *hlInfoNext;
234 hlInfo = *hlInfoHeadPtr;
236 hlInfoNext = hlInfo->next;
240 *hlInfoHeadPtr = NULL;
244 /* Might be faster (and bigger) if the dev/ino were stored in numeric order ;) */
245 static HardLinkInfo *findHardLinkInfo(HardLinkInfo *hlInfo, struct stat *statbuf)
248 if (statbuf->st_ino == hlInfo->ino
249 && statbuf->st_dev == hlInfo->dev
251 DBG("found hardlink:'%s'", hlInfo->name);
254 hlInfo = hlInfo->next;
259 /* Put an octal string into the specified buffer.
260 * The number is zero padded and possibly null terminated.
261 * Stores low-order bits only if whole value does not fit. */
262 static void putOctal(char *cp, int len, off_t value)
264 char tempBuffer[sizeof(off_t)*3 + 1];
265 char *tempString = tempBuffer;
268 width = sprintf(tempBuffer, "%0*"OFF_FMT"o", len, value);
269 tempString += (width - len);
271 /* If string has leading zeroes, we can drop one */
272 /* and field will have trailing '\0' */
273 /* (increases chances of compat with other tars) */
274 if (tempString[0] == '0')
277 /* Copy the string to the field */
278 memcpy(cp, tempString, len);
280 #define PUT_OCTAL(a, b) putOctal((a), sizeof(a), (b))
282 static void chksum_and_xwrite(int fd, struct tar_header_t* hp)
284 /* POSIX says that checksum is done on unsigned bytes
285 * (Sun and HP-UX gets it wrong... more details in
287 const unsigned char *cp;
290 strcpy(hp->magic, "ustar ");
292 /* Calculate and store the checksum (i.e., the sum of all of the bytes of
293 * the header). The checksum field must be filled with blanks for the
294 * calculation. The checksum field is formatted differently from the
295 * other fields: it has 6 digits, a null, then a space -- rather than
296 * digits, followed by a null like the other fields... */
297 memset(hp->chksum, ' ', sizeof(hp->chksum));
298 cp = (const unsigned char *) hp;
301 do { chksum += *cp++; } while (--size);
302 putOctal(hp->chksum, sizeof(hp->chksum)-1, chksum);
304 /* Now write the header out to disk */
305 xwrite(fd, hp, sizeof(*hp));
308 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
309 static void writeLongname(int fd, int type, const char *name, int dir)
311 static const struct {
312 char mode[8]; /* 100-107 */
313 char uid[8]; /* 108-115 */
314 char gid[8]; /* 116-123 */
315 char size[12]; /* 124-135 */
316 char mtime[12]; /* 136-147 */
324 struct tar_header_t header;
327 dir = !!dir; /* normalize: 0/1 */
328 size = strlen(name) + 1 + dir; /* GNU tar uses strlen+1 */
329 /* + dir: account for possible '/' */
331 memset(&header, 0, sizeof(header));
332 strcpy(header.name, "././@LongLink");
333 memcpy(header.mode, prefilled.mode, sizeof(prefilled));
334 PUT_OCTAL(header.size, size);
335 header.typeflag = type;
336 chksum_and_xwrite(fd, &header);
338 /* Write filename[/] and pad the block. */
339 /* dir=0: writes 'name<NUL>', pads */
340 /* dir=1: writes 'name', writes '/<NUL>', pads */
342 xwrite(fd, name, size - dir);
343 xwrite(fd, "/", dir);
344 size = (-size) & (TAR_BLOCK_SIZE-1);
345 memset(&header, 0, size);
346 xwrite(fd, &header, size);
350 /* Write out a tar header for the specified file/directory/whatever */
351 static int writeTarHeader(struct TarBallInfo *tbInfo,
352 const char *header_name, const char *fileName, struct stat *statbuf)
354 struct tar_header_t header;
356 memset(&header, 0, sizeof(header));
358 strncpy(header.name, header_name, sizeof(header.name));
360 /* POSIX says to mask mode with 07777. */
361 PUT_OCTAL(header.mode, statbuf->st_mode & 07777);
362 PUT_OCTAL(header.uid, statbuf->st_uid);
363 PUT_OCTAL(header.gid, statbuf->st_gid);
364 memset(header.size, '0', sizeof(header.size)-1); /* Regular file size is handled later */
365 /* users report that files with negative st_mtime cause trouble, so: */
366 PUT_OCTAL(header.mtime, statbuf->st_mtime >= 0 ? statbuf->st_mtime : 0);
368 /* Enter the user and group names */
369 safe_strncpy(header.uname, get_cached_username(statbuf->st_uid), sizeof(header.uname));
370 safe_strncpy(header.gname, get_cached_groupname(statbuf->st_gid), sizeof(header.gname));
372 if (tbInfo->hlInfo) {
373 /* This is a hard link */
374 header.typeflag = LNKTYPE;
375 strncpy(header.linkname, tbInfo->hlInfo->name,
376 sizeof(header.linkname));
377 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
378 /* Write out long linkname if needed */
379 if (header.linkname[sizeof(header.linkname)-1])
380 writeLongname(tbInfo->tarFd, GNULONGLINK,
381 tbInfo->hlInfo->name, 0);
383 } else if (S_ISLNK(statbuf->st_mode)) {
384 char *lpath = xmalloc_readlink_or_warn(fileName);
387 header.typeflag = SYMTYPE;
388 strncpy(header.linkname, lpath, sizeof(header.linkname));
389 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
390 /* Write out long linkname if needed */
391 if (header.linkname[sizeof(header.linkname)-1])
392 writeLongname(tbInfo->tarFd, GNULONGLINK, lpath, 0);
394 /* If it is larger than 100 bytes, bail out */
395 if (header.linkname[sizeof(header.linkname)-1]) {
397 bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported");
402 } else if (S_ISDIR(statbuf->st_mode)) {
403 header.typeflag = DIRTYPE;
404 /* Append '/' only if there is a space for it */
405 if (!header.name[sizeof(header.name)-1])
406 header.name[strlen(header.name)] = '/';
407 } else if (S_ISCHR(statbuf->st_mode)) {
408 header.typeflag = CHRTYPE;
409 PUT_OCTAL(header.devmajor, major(statbuf->st_rdev));
410 PUT_OCTAL(header.devminor, minor(statbuf->st_rdev));
411 } else if (S_ISBLK(statbuf->st_mode)) {
412 header.typeflag = BLKTYPE;
413 PUT_OCTAL(header.devmajor, major(statbuf->st_rdev));
414 PUT_OCTAL(header.devminor, minor(statbuf->st_rdev));
415 } else if (S_ISFIFO(statbuf->st_mode)) {
416 header.typeflag = FIFOTYPE;
417 } else if (S_ISREG(statbuf->st_mode)) {
418 /* header.size field is 12 bytes long */
419 /* Does octal-encoded size fit? */
420 uoff_t filesize = statbuf->st_size;
421 if (sizeof(filesize) <= 4
422 || filesize <= (uoff_t)0777777777777LL
424 PUT_OCTAL(header.size, filesize);
426 /* Does base256-encoded size fit?
427 * It always does unless off_t is wider than 64 bits.
429 else if (ENABLE_FEATURE_TAR_GNU_EXTENSIONS
430 #if ULLONG_MAX > 0xffffffffffffffffLL /* 2^64-1 */
431 && (filesize <= 0x3fffffffffffffffffffffffLL)
434 /* GNU tar uses "base-256 encoding" for very large numbers.
435 * Encoding is binary, with highest bit always set as a marker
436 * and sign in next-highest bit:
438 * bf ff .. ff - largest positive number
439 * ff ff .. ff - minus 1
440 * c0 00 .. 00 - smallest negative number
442 char *p8 = header.size + sizeof(header.size);
444 *--p8 = (uint8_t)filesize;
446 } while (p8 != header.size);
449 bb_error_msg_and_die("can't store file '%s' "
450 "of size %"OFF_FMT"u, aborting",
451 fileName, statbuf->st_size);
453 header.typeflag = REGTYPE;
455 bb_error_msg("%s: unknown file type", fileName);
459 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
460 /* Write out long name if needed */
461 /* (we, like GNU tar, output long linkname *before* long name) */
462 if (header.name[sizeof(header.name)-1])
463 writeLongname(tbInfo->tarFd, GNULONGNAME,
464 header_name, S_ISDIR(statbuf->st_mode));
467 /* Now write the header out to disk */
468 chksum_and_xwrite(tbInfo->tarFd, &header);
470 /* Now do the verbose thing (or not) */
471 if (tbInfo->verboseFlag) {
474 /* If archive goes to stdout, verbose goes to stderr */
475 if (tbInfo->tarFd == STDOUT_FILENO)
477 /* GNU "tar cvvf" prints "extended" listing a-la "ls -l" */
478 /* We don't have such excesses here: for us "v" == "vv" */
479 /* '/' is probably a GNUism */
480 fprintf(vbFd, "%s%s\n", header_name,
481 S_ISDIR(statbuf->st_mode) ? "/" : "");
487 #if ENABLE_FEATURE_TAR_FROM
488 static int exclude_file(const llist_t *excluded_files, const char *file)
490 while (excluded_files) {
491 if (excluded_files->data[0] == '/') {
492 if (fnmatch(excluded_files->data, file,
493 FNM_PATHNAME | FNM_LEADING_DIR) == 0)
498 for (p = file; p[0] != '\0'; p++) {
499 if ((p == file || p[-1] == '/')
501 && fnmatch(excluded_files->data, p,
502 FNM_PATHNAME | FNM_LEADING_DIR) == 0
508 excluded_files = excluded_files->link;
514 # define exclude_file(excluded_files, file) 0
517 static int FAST_FUNC writeFileToTarball(const char *fileName, struct stat *statbuf,
518 void *userData, int depth UNUSED_PARAM)
520 struct TarBallInfo *tbInfo = (struct TarBallInfo *) userData;
521 const char *header_name;
522 int inputFileFd = -1;
524 DBG("writeFileToTarball('%s')", fileName);
526 /* Strip leading '/' and such (must be before memorizing hardlink's name) */
527 header_name = strip_unsafe_prefix(fileName);
529 if (header_name[0] == '\0')
532 /* It is against the rules to archive a socket */
533 if (S_ISSOCK(statbuf->st_mode)) {
534 bb_error_msg("%s: socket ignored", fileName);
539 * Check to see if we are dealing with a hard link.
541 * Treat the first occurance of a given dev/inode as a file while
542 * treating any additional occurances as hard links. This is done
543 * by adding the file information to the HardLinkInfo linked list.
545 tbInfo->hlInfo = NULL;
546 if (!S_ISDIR(statbuf->st_mode) && statbuf->st_nlink > 1) {
547 DBG("'%s': st_nlink > 1", header_name);
548 tbInfo->hlInfo = findHardLinkInfo(tbInfo->hlInfoHead, statbuf);
549 if (tbInfo->hlInfo == NULL) {
550 DBG("'%s': addHardLinkInfo", header_name);
551 addHardLinkInfo(&tbInfo->hlInfoHead, statbuf, header_name);
555 /* It is a bad idea to store the archive we are in the process of creating,
556 * so check the device and inode to be sure that this particular file isn't
558 if (tbInfo->tarFileStatBuf.st_dev == statbuf->st_dev
559 && tbInfo->tarFileStatBuf.st_ino == statbuf->st_ino
561 bb_error_msg("%s: file is the archive; skipping", fileName);
565 if (exclude_file(tbInfo->excludeList, header_name))
568 #if !ENABLE_FEATURE_TAR_GNU_EXTENSIONS
569 if (strlen(header_name) >= NAME_SIZE) {
570 bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported");
575 /* Is this a regular file? */
576 if (tbInfo->hlInfo == NULL && S_ISREG(statbuf->st_mode)) {
577 /* open the file we want to archive, and make sure all is well */
578 inputFileFd = open_or_warn(fileName, O_RDONLY);
579 if (inputFileFd < 0) {
584 /* Add an entry to the tarball */
585 if (writeTarHeader(tbInfo, header_name, fileName, statbuf) == FALSE) {
589 /* If it was a regular file, write out the body */
590 if (inputFileFd >= 0) {
592 /* Write the file to the archive. */
593 /* We record size into header first, */
594 /* and then write out file. If file shrinks in between, */
595 /* tar will be corrupted. So we don't allow for that. */
596 /* NB: GNU tar 1.16 warns and pads with zeroes */
597 /* or even seeks back and updates header */
598 bb_copyfd_exact_size(inputFileFd, tbInfo->tarFd, statbuf->st_size);
600 ////readSize = bb_copyfd_size(inputFileFd, tbInfo->tarFd, statbuf->st_size);
601 ////if (readSize != statbuf->st_size && readSize >= 0) {
602 //// bb_error_msg_and_die("short read from %s, aborting", fileName);
605 /* Check that file did not grow in between? */
606 /* if (safe_read(inputFileFd, 1) == 1) warn but continue? */
610 /* Pad the file up to the tar block size */
611 /* (a few tricks here in the name of code size) */
612 readSize = (-(int)statbuf->st_size) & (TAR_BLOCK_SIZE-1);
613 memset(block_buf, 0, readSize);
614 xwrite(tbInfo->tarFd, block_buf, readSize);
620 #if SEAMLESS_COMPRESSION
621 /* Don't inline: vfork scares gcc and pessimizes code */
622 static void NOINLINE vfork_compressor(int tar_fd, const char *gzip)
626 // On Linux, vfork never unpauses parent early, although standard
627 // allows for that. Do we want to waste bytes checking for it?
628 # define WAIT_FOR_CHILD 0
629 volatile int vfork_exec_errno = 0;
630 struct fd_pair gzipDataPipe;
632 struct fd_pair gzipStatusPipe;
633 xpiped_pair(gzipStatusPipe);
635 xpiped_pair(gzipDataPipe);
637 signal(SIGPIPE, SIG_IGN); /* we only want EPIPE on errors */
643 /* NB: close _first_, then move fds! */
644 close(gzipDataPipe.wr);
646 close(gzipStatusPipe.rd);
647 /* gzipStatusPipe.wr will close only on exec -
648 * parent waits for this close to happen */
649 fcntl(gzipStatusPipe.wr, F_SETFD, FD_CLOEXEC);
651 xmove_fd(gzipDataPipe.rd, 0);
653 /* exec gzip/bzip2 program/applet */
654 BB_EXECLP(gzip, gzip, "-f", (char *)0);
655 vfork_exec_errno = errno;
660 xmove_fd(gzipDataPipe.wr, tar_fd);
661 close(gzipDataPipe.rd);
663 close(gzipStatusPipe.wr);
668 /* Wait until child execs (or fails to) */
669 n = full_read(gzipStatusPipe.rd, &buf, 1);
670 if (n < 0 /* && errno == EAGAIN */)
671 continue; /* try it again */
673 close(gzipStatusPipe.rd);
675 if (vfork_exec_errno) {
676 errno = vfork_exec_errno;
677 bb_perror_msg_and_die("can't execute '%s'", gzip);
680 #endif /* SEAMLESS_COMPRESSION */
683 #if !SEAMLESS_COMPRESSION
684 /* Do not pass gzip flag to writeTarFile() */
685 #define writeTarFile(tar_fd, verboseFlag, recurseFlags, include, exclude, gzip) \
686 writeTarFile(tar_fd, verboseFlag, recurseFlags, include, exclude)
688 /* gcc 4.2.1 inlines it, making code bigger */
689 static NOINLINE int writeTarFile(int tar_fd, int verboseFlag,
690 int recurseFlags, const llist_t *include,
691 const llist_t *exclude, const char *gzip)
693 int errorFlag = FALSE;
694 struct TarBallInfo tbInfo;
696 tbInfo.hlInfoHead = NULL;
697 tbInfo.tarFd = tar_fd;
698 tbInfo.verboseFlag = verboseFlag;
700 /* Store the stat info for the tarball's file, so
701 * can avoid including the tarball into itself.... */
702 xfstat(tbInfo.tarFd, &tbInfo.tarFileStatBuf, "can't stat tar file");
704 #if SEAMLESS_COMPRESSION
706 vfork_compressor(tbInfo.tarFd, gzip);
709 tbInfo.excludeList = exclude;
711 /* Read the directory/files and iterate over them one at a time */
713 if (!recursive_action(include->data, recurseFlags,
714 writeFileToTarball, writeFileToTarball, &tbInfo, 0)
718 include = include->link;
720 /* Write two empty blocks to the end of the archive */
721 memset(block_buf, 0, 2*TAR_BLOCK_SIZE);
722 xwrite(tbInfo.tarFd, block_buf, 2*TAR_BLOCK_SIZE);
724 /* To be pedantically correct, we would check if the tarball
725 * is smaller than 20 tar blocks, and pad it if it was smaller,
726 * but that isn't necessary for GNU tar interoperability, and
727 * so is considered a waste of space */
729 /* Close so the child process (if any) will exit */
732 /* Hang up the tools, close up shop, head home */
733 if (ENABLE_FEATURE_CLEAN_UP)
734 freeHardLinkInfo(&tbInfo.hlInfoHead);
737 bb_error_msg("error exit delayed from previous errors");
739 #if SEAMLESS_COMPRESSION
742 if (safe_waitpid(-1, &status, 0) == -1)
743 bb_perror_msg("waitpid");
744 else if (!WIFEXITED(status) || WEXITSTATUS(status))
745 /* gzip was killed or has exited with nonzero! */
751 #else /* !FEATURE_TAR_CREATE */
752 # define writeTarFile(...) 0
755 #if ENABLE_FEATURE_TAR_FROM
756 static llist_t *append_file_list_to_list(llist_t *list)
760 llist_t *newlist = NULL;
763 src_stream = xfopen_stdin(llist_pop(&list));
764 while ((line = xmalloc_fgetline(src_stream)) != NULL) {
765 /* kill trailing '/' unless the string is just "/" */
766 char *cp = last_char_is(line, '/');
769 llist_add_to_end(&newlist, line);
777 //usage:#define tar_trivial_usage
778 //usage: "-[" IF_FEATURE_TAR_CREATE("c") "xt"
779 //usage: IF_FEATURE_SEAMLESS_Z("Z")
780 //usage: IF_FEATURE_SEAMLESS_GZ("z")
781 //usage: IF_FEATURE_SEAMLESS_XZ("J")
782 //usage: IF_FEATURE_SEAMLESS_BZ2("j")
783 //usage: IF_FEATURE_SEAMLESS_LZMA("a")
784 //usage: IF_FEATURE_TAR_CREATE("h")
785 //usage: IF_FEATURE_TAR_NOPRESERVE_TIME("m")
787 //usage: IF_FEATURE_TAR_FROM("[-X FILE] [-T FILE] ")
788 //usage: "[-f TARFILE] [-C DIR] [FILE]..."
789 //usage:#define tar_full_usage "\n\n"
790 //usage: IF_FEATURE_TAR_CREATE("Create, extract, ")
791 //usage: IF_NOT_FEATURE_TAR_CREATE("Extract ")
792 //usage: "or list files from a tar file\n"
793 //usage: "\nOperation:"
794 //usage: IF_FEATURE_TAR_CREATE(
795 //usage: "\n c Create"
797 //usage: "\n x Extract"
799 //usage: "\n f Name of TARFILE ('-' for stdin/out)"
800 //usage: "\n C Change to DIR before operation"
801 //usage: "\n v Verbose"
802 //usage: IF_FEATURE_SEAMLESS_Z(
803 //usage: "\n Z (De)compress using compress"
805 //usage: IF_FEATURE_SEAMLESS_GZ(
806 //usage: "\n z (De)compress using gzip"
808 //usage: IF_FEATURE_SEAMLESS_XZ(
809 //usage: "\n J (De)compress using xz"
811 //usage: IF_FEATURE_SEAMLESS_BZ2(
812 //usage: "\n j (De)compress using bzip2"
814 //usage: IF_FEATURE_SEAMLESS_LZMA(
815 //usage: "\n a (De)compress using lzma"
817 //usage: "\n O Extract to stdout"
818 //usage: IF_FEATURE_TAR_CREATE(
819 //usage: "\n h Follow symlinks"
821 //usage: IF_FEATURE_TAR_NOPRESERVE_TIME(
822 //usage: "\n m Don't restore mtime"
824 //usage: IF_FEATURE_TAR_FROM(
825 //usage: IF_FEATURE_TAR_LONG_OPTIONS(
826 //usage: "\n exclude File to exclude"
828 //usage: "\n X File with names to exclude"
829 //usage: "\n T File with names to include"
832 //usage:#define tar_example_usage
833 //usage: "$ zcat /tmp/tarball.tar.gz | tar -xf -\n"
834 //usage: "$ tar -cf /tmp/tarball.tar /usr/local\n"
836 // Supported but aren't in --help:
838 // p same-permissions
842 // no-same-permissions
844 //IF_FEATURE_TAR_TO_COMMAND(
850 IF_FEATURE_TAR_CREATE( OPTBIT_CREATE ,)
851 IF_FEATURE_TAR_CREATE( OPTBIT_DEREFERENCE ,)
852 IF_FEATURE_SEAMLESS_BZ2( OPTBIT_BZIP2 ,)
853 IF_FEATURE_SEAMLESS_LZMA(OPTBIT_LZMA ,)
854 IF_FEATURE_TAR_FROM( OPTBIT_INCLUDE_FROM,)
855 IF_FEATURE_TAR_FROM( OPTBIT_EXCLUDE_FROM,)
856 IF_FEATURE_SEAMLESS_GZ( OPTBIT_GZIP ,)
857 IF_FEATURE_SEAMLESS_XZ( OPTBIT_XZ ,) // 16th bit
858 IF_FEATURE_SEAMLESS_Z( OPTBIT_COMPRESS ,)
859 IF_FEATURE_TAR_NOPRESERVE_TIME(OPTBIT_NOPRESERVE_TIME,)
860 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
861 OPTBIT_STRIP_COMPONENTS,
863 IF_FEATURE_TAR_TO_COMMAND(OPTBIT_2COMMAND ,)
864 OPTBIT_NUMERIC_OWNER,
865 OPTBIT_NOPRESERVE_PERM,
868 OPT_TEST = 1 << 0, // t
869 OPT_EXTRACT = 1 << 1, // x
870 OPT_BASEDIR = 1 << 2, // C
871 OPT_TARNAME = 1 << 3, // f
872 OPT_2STDOUT = 1 << 4, // O
873 OPT_NOPRESERVE_OWNER = 1 << 5, // o == no-same-owner
875 OPT_VERBOSE = 1 << 7, // v
876 OPT_KEEP_OLD = 1 << 8, // k
877 OPT_CREATE = IF_FEATURE_TAR_CREATE( (1 << OPTBIT_CREATE )) + 0, // c
878 OPT_DEREFERENCE = IF_FEATURE_TAR_CREATE( (1 << OPTBIT_DEREFERENCE )) + 0, // h
879 OPT_BZIP2 = IF_FEATURE_SEAMLESS_BZ2( (1 << OPTBIT_BZIP2 )) + 0, // j
880 OPT_LZMA = IF_FEATURE_SEAMLESS_LZMA((1 << OPTBIT_LZMA )) + 0, // a
881 OPT_INCLUDE_FROM = IF_FEATURE_TAR_FROM( (1 << OPTBIT_INCLUDE_FROM)) + 0, // T
882 OPT_EXCLUDE_FROM = IF_FEATURE_TAR_FROM( (1 << OPTBIT_EXCLUDE_FROM)) + 0, // X
883 OPT_GZIP = IF_FEATURE_SEAMLESS_GZ( (1 << OPTBIT_GZIP )) + 0, // z
884 OPT_XZ = IF_FEATURE_SEAMLESS_XZ( (1 << OPTBIT_XZ )) + 0, // J
885 OPT_COMPRESS = IF_FEATURE_SEAMLESS_Z( (1 << OPTBIT_COMPRESS )) + 0, // Z
886 OPT_NOPRESERVE_TIME = IF_FEATURE_TAR_NOPRESERVE_TIME((1 << OPTBIT_NOPRESERVE_TIME)) + 0, // m
887 OPT_STRIP_COMPONENTS = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_STRIP_COMPONENTS)) + 0, // strip-components
888 OPT_NORECURSION = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NORECURSION )) + 0, // no-recursion
889 OPT_2COMMAND = IF_FEATURE_TAR_TO_COMMAND( (1 << OPTBIT_2COMMAND )) + 0, // to-command
890 OPT_NUMERIC_OWNER = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NUMERIC_OWNER )) + 0, // numeric-owner
891 OPT_NOPRESERVE_PERM = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NOPRESERVE_PERM)) + 0, // no-same-permissions
892 OPT_OVERWRITE = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_OVERWRITE )) + 0, // overwrite
894 OPT_ANY_COMPRESS = (OPT_BZIP2 | OPT_LZMA | OPT_GZIP | OPT_XZ | OPT_COMPRESS),
896 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
897 static const char tar_longopts[] ALIGN1 =
898 "list\0" No_argument "t"
899 "extract\0" No_argument "x"
900 "directory\0" Required_argument "C"
901 "file\0" Required_argument "f"
902 "to-stdout\0" No_argument "O"
903 /* do not restore owner */
904 /* Note: GNU tar handles 'o' as no-same-owner only on extract,
905 * on create, 'o' is --old-archive. We do not support --old-archive. */
906 "no-same-owner\0" No_argument "o"
907 "same-permissions\0" No_argument "p"
908 "verbose\0" No_argument "v"
909 "keep-old\0" No_argument "k"
910 # if ENABLE_FEATURE_TAR_CREATE
911 "create\0" No_argument "c"
912 "dereference\0" No_argument "h"
914 # if ENABLE_FEATURE_SEAMLESS_BZ2
915 "bzip2\0" No_argument "j"
917 # if ENABLE_FEATURE_SEAMLESS_LZMA
918 "lzma\0" No_argument "a"
920 # if ENABLE_FEATURE_TAR_FROM
921 "files-from\0" Required_argument "T"
922 "exclude-from\0" Required_argument "X"
924 # if ENABLE_FEATURE_SEAMLESS_GZ
925 "gzip\0" No_argument "z"
927 # if ENABLE_FEATURE_SEAMLESS_XZ
928 "xz\0" No_argument "J"
930 # if ENABLE_FEATURE_SEAMLESS_Z
931 "compress\0" No_argument "Z"
933 # if ENABLE_FEATURE_TAR_NOPRESERVE_TIME
934 "touch\0" No_argument "m"
936 "strip-components\0" Required_argument "\xf9"
937 "no-recursion\0" No_argument "\xfa"
938 # if ENABLE_FEATURE_TAR_TO_COMMAND
939 "to-command\0" Required_argument "\xfb"
941 /* use numeric uid/gid from tar header, not textual */
942 "numeric-owner\0" No_argument "\xfc"
943 /* do not restore mode */
944 "no-same-permissions\0" No_argument "\xfd"
945 /* on unpack, open with O_TRUNC and !O_EXCL */
946 "overwrite\0" No_argument "\xfe"
947 /* --exclude takes next bit position in option mask, */
948 /* therefore we have to put it _after_ --no-same-permissions */
949 # if ENABLE_FEATURE_TAR_FROM
950 "exclude\0" Required_argument "\xff"
955 int tar_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
956 int tar_main(int argc UNUSED_PARAM, char **argv)
958 archive_handle_t *tar_handle;
959 char *base_dir = NULL;
960 const char *tar_filename = "-";
963 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
964 llist_t *excludes = NULL;
967 /* Initialise default values */
968 tar_handle = init_handle();
969 tar_handle->ah_flags = ARCHIVE_CREATE_LEADING_DIRS
970 | ARCHIVE_RESTORE_DATE
971 | ARCHIVE_UNLINK_OLD;
973 /* Apparently only root's tar preserves perms (see bug 3844) */
975 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_PERM;
977 /* Prepend '-' to the first argument if required */
978 opt_complementary = "--:" // first arg is options
979 "tt:vv:" // count -t,-v
980 IF_FEATURE_TAR_FROM("X::T::") // cumulative lists
981 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
982 "\xff::" // --exclude=PATTERN is a list
984 IF_FEATURE_TAR_CREATE("c:") "t:x:" // at least one of these is reqd
985 IF_FEATURE_TAR_CREATE("c--tx:t--cx:x--ct") // mutually exclusive
986 IF_NOT_FEATURE_TAR_CREATE("t--x:x--t") // mutually exclusive
987 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
988 ":\xf9+" // --strip-components=NUM
991 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
992 applet_long_options = tar_longopts;
995 if (argv[1] && argv[1][0] != '-') {
997 * 1st argument without dash handles options with parameters
998 * differently from dashed one: it takes *next argv[i]*
999 * as paramenter even if there are more chars in 1st argument:
1000 * "tar fx TARFILE" - "x" is not taken as f's param
1001 * but is interpreted as -x option
1002 * "tar -xf TARFILE" - dashed equivalent of the above
1003 * "tar -fx ..." - "x" is taken as f's param
1004 * getopt32 wouldn't handle 1st command correctly.
1005 * Unfortunately, people do use such commands.
1006 * We massage argv[1] to work around it by moving 'f'
1007 * to the end of the string.
1008 * More contrived "tar fCx TARFILE DIR" still fails,
1009 * but such commands are much less likely to be used.
1011 char *f = strchr(argv[1], 'f');
1013 while (f[1] != '\0') {
1021 opt = getopt32(argv,
1023 IF_FEATURE_TAR_CREATE( "ch" )
1024 IF_FEATURE_SEAMLESS_BZ2( "j" )
1025 IF_FEATURE_SEAMLESS_LZMA("a" )
1026 IF_FEATURE_TAR_FROM( "T:X:")
1027 IF_FEATURE_SEAMLESS_GZ( "z" )
1028 IF_FEATURE_SEAMLESS_XZ( "J" )
1029 IF_FEATURE_SEAMLESS_Z( "Z" )
1030 IF_FEATURE_TAR_NOPRESERVE_TIME("m")
1031 IF_FEATURE_TAR_LONG_OPTIONS("\xf9:") // --strip-components
1032 , &base_dir // -C dir
1033 , &tar_filename // -f filename
1034 IF_FEATURE_TAR_FROM(, &(tar_handle->accept)) // T
1035 IF_FEATURE_TAR_FROM(, &(tar_handle->reject)) // X
1036 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
1037 , &tar_handle->tar__strip_components // --strip-components
1039 IF_FEATURE_TAR_TO_COMMAND(, &(tar_handle->tar__to_command)) // --to-command
1040 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
1041 , &excludes // --exclude
1043 , &verboseFlag // combined count for -t and -v
1044 , &verboseFlag // combined count for -t and -v
1046 #if DBG_OPTION_PARSING
1047 bb_error_msg("opt: 0x%08x", opt);
1048 # define showopt(o) bb_error_msg("opt & %s(%x): %x", #o, o, opt & o);
1050 showopt(OPT_EXTRACT );
1051 showopt(OPT_BASEDIR );
1052 showopt(OPT_TARNAME );
1053 showopt(OPT_2STDOUT );
1054 showopt(OPT_NOPRESERVE_OWNER);
1056 showopt(OPT_VERBOSE );
1057 showopt(OPT_KEEP_OLD );
1058 showopt(OPT_CREATE );
1059 showopt(OPT_DEREFERENCE );
1060 showopt(OPT_BZIP2 );
1062 showopt(OPT_INCLUDE_FROM );
1063 showopt(OPT_EXCLUDE_FROM );
1066 showopt(OPT_COMPRESS );
1067 showopt(OPT_NOPRESERVE_TIME );
1068 showopt(OPT_STRIP_COMPONENTS);
1069 showopt(OPT_NORECURSION );
1070 showopt(OPT_2COMMAND );
1071 showopt(OPT_NUMERIC_OWNER );
1072 showopt(OPT_NOPRESERVE_PERM );
1073 showopt(OPT_OVERWRITE );
1074 showopt(OPT_ANY_COMPRESS );
1075 bb_error_msg("base_dir:'%s'", base_dir);
1076 bb_error_msg("tar_filename:'%s'", tar_filename);
1077 bb_error_msg("verboseFlag:%d", verboseFlag);
1078 bb_error_msg("tar_handle->tar__to_command:'%s'", tar_handle->tar__to_command);
1079 bb_error_msg("tar_handle->tar__strip_components:%u", tar_handle->tar__strip_components);
1086 tar_handle->action_header = header_verbose_list;
1087 if (verboseFlag == 1)
1088 tar_handle->action_header = header_list;
1090 if (opt & OPT_EXTRACT)
1091 tar_handle->action_data = data_extract_all;
1093 if (opt & OPT_2STDOUT)
1094 tar_handle->action_data = data_extract_to_stdout;
1096 if (opt & OPT_2COMMAND) {
1097 putenv((char*)"TAR_FILETYPE=f");
1098 signal(SIGPIPE, SIG_IGN);
1099 tar_handle->action_data = data_extract_to_command;
1100 IF_FEATURE_TAR_TO_COMMAND(tar_handle->tar__to_command_shell = xstrdup(get_shell_name());)
1103 if (opt & OPT_KEEP_OLD)
1104 tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
1106 if (opt & OPT_NUMERIC_OWNER)
1107 tar_handle->ah_flags |= ARCHIVE_NUMERIC_OWNER;
1109 if (opt & OPT_NOPRESERVE_OWNER)
1110 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_OWNER;
1112 if (opt & OPT_NOPRESERVE_PERM)
1113 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_PERM;
1115 if (opt & OPT_OVERWRITE) {
1116 tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
1117 tar_handle->ah_flags |= ARCHIVE_O_TRUNC;
1120 if (opt & OPT_NOPRESERVE_TIME)
1121 tar_handle->ah_flags &= ~ARCHIVE_RESTORE_DATE;
1123 #if ENABLE_FEATURE_TAR_FROM
1124 tar_handle->reject = append_file_list_to_list(tar_handle->reject);
1125 # if ENABLE_FEATURE_TAR_LONG_OPTIONS
1126 /* Append excludes to reject */
1128 llist_t *next = excludes->link;
1129 excludes->link = tar_handle->reject;
1130 tar_handle->reject = excludes;
1134 tar_handle->accept = append_file_list_to_list(tar_handle->accept);
1137 /* Setup an array of filenames to work with */
1138 /* TODO: This is the same as in ar, make a separate function? */
1140 /* kill trailing '/' unless the string is just "/" */
1141 char *cp = last_char_is(*argv, '/');
1144 llist_add_to_end(&tar_handle->accept, *argv);
1148 if (tar_handle->accept || tar_handle->reject)
1149 tar_handle->filter = filter_accept_reject_list;
1151 /* Open the tar file */
1153 int tar_fd = STDIN_FILENO;
1154 int flags = O_RDONLY;
1156 if (opt & OPT_CREATE) {
1157 /* Make sure there is at least one file to tar up */
1158 if (tar_handle->accept == NULL)
1159 bb_error_msg_and_die("empty archive");
1161 tar_fd = STDOUT_FILENO;
1162 /* Mimicking GNU tar 1.15.1: */
1163 flags = O_WRONLY | O_CREAT | O_TRUNC;
1166 if (LONE_DASH(tar_filename)) {
1167 tar_handle->src_fd = tar_fd;
1168 tar_handle->seek = seek_by_read;
1170 if (ENABLE_FEATURE_TAR_AUTODETECT
1171 && flags == O_RDONLY
1172 && !(opt & OPT_ANY_COMPRESS)
1174 tar_handle->src_fd = open_zipped(tar_filename, /*fail_if_not_compressed:*/ 0);
1175 if (tar_handle->src_fd < 0)
1176 bb_perror_msg_and_die("can't open '%s'", tar_filename);
1178 tar_handle->src_fd = xopen(tar_filename, flags);
1186 //if (SEAMLESS_COMPRESSION)
1187 // /* We need to know whether child (gzip/bzip/etc) exits abnormally */
1188 // signal(SIGCHLD, check_errors_in_children);
1190 /* Create an archive */
1191 if (opt & OPT_CREATE) {
1192 #if SEAMLESS_COMPRESSION
1193 const char *zipMode = NULL;
1194 if (opt & OPT_COMPRESS)
1195 zipMode = "compress";
1198 if (opt & OPT_BZIP2)
1205 /* NB: writeTarFile() closes tar_handle->src_fd */
1206 return writeTarFile(tar_handle->src_fd, verboseFlag,
1207 (opt & OPT_DEREFERENCE ? ACTION_FOLLOWLINKS : 0)
1208 | (opt & OPT_NORECURSION ? 0 : ACTION_RECURSE),
1210 tar_handle->reject, zipMode);
1213 if (opt & OPT_ANY_COMPRESS) {
1214 USE_FOR_MMU(IF_DESKTOP(long long) int FAST_FUNC (*xformer)(transformer_state_t *xstate);)
1215 USE_FOR_NOMMU(const char *xformer_prog;)
1217 if (opt & OPT_COMPRESS)
1218 USE_FOR_MMU(xformer = unpack_Z_stream;)
1219 USE_FOR_NOMMU(xformer_prog = "uncompress";)
1221 USE_FOR_MMU(xformer = unpack_gz_stream;)
1222 USE_FOR_NOMMU(xformer_prog = "gunzip";)
1223 if (opt & OPT_BZIP2)
1224 USE_FOR_MMU(xformer = unpack_bz2_stream;)
1225 USE_FOR_NOMMU(xformer_prog = "bunzip2";)
1227 USE_FOR_MMU(xformer = unpack_lzma_stream;)
1228 USE_FOR_NOMMU(xformer_prog = "unlzma";)
1230 USE_FOR_MMU(xformer = unpack_xz_stream;)
1231 USE_FOR_NOMMU(xformer_prog = "unxz";)
1233 fork_transformer_with_sig(tar_handle->src_fd, xformer, xformer_prog);
1234 /* Can't lseek over pipes */
1235 tar_handle->seek = seek_by_read;
1236 /*tar_handle->offset = 0; - already is */
1239 /* Zero processed headers (== empty file) is not a valid tarball.
1240 * We (ab)use bb_got_signal as exitcode here,
1241 * because check_errors_in_children() uses _it_ as error indicator.
1243 bb_got_signal = EXIT_FAILURE;
1245 while (get_header_tar(tar_handle) == EXIT_SUCCESS)
1246 bb_got_signal = EXIT_SUCCESS; /* saw at least one header, good */
1248 /* Check that every file that should have been extracted was */
1249 while (tar_handle->accept) {
1250 if (!find_list_entry(tar_handle->reject, tar_handle->accept->data)
1251 && !find_list_entry(tar_handle->passed, tar_handle->accept->data)
1253 bb_error_msg_and_die("%s: not found in archive",
1254 tar_handle->accept->data);
1256 tar_handle->accept = tar_handle->accept->link;
1258 if (ENABLE_FEATURE_CLEAN_UP /* && tar_handle->src_fd != STDIN_FILENO */)
1259 close(tar_handle->src_fd);
1261 if (SEAMLESS_COMPRESSION || OPT_COMPRESS) {
1262 /* Set bb_got_signal to 1 if a child died with !0 exitcode */
1263 check_errors_in_children(0);
1266 return bb_got_signal;