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__)
157 #define DBG(...) ((void)0)
160 #define block_buf bb_common_bufsiz1
163 #if ENABLE_FEATURE_TAR_CREATE
166 ** writeTarFile(), writeFileToTarball(), and writeTarHeader() are
167 ** the only functions that deal with the HardLinkInfo structure.
168 ** Even these functions use the xxxHardLinkInfo() functions.
170 typedef struct HardLinkInfo {
171 struct HardLinkInfo *next; /* Next entry in list */
172 dev_t dev; /* Device number */
173 ino_t ino; /* Inode number */
174 // short linkCount; /* (Hard) Link Count */
175 char name[1]; /* Start of filename (must be last) */
178 /* Some info to be carried along when creating a new tarball */
179 typedef struct TarBallInfo {
180 int tarFd; /* Open-for-write file descriptor
182 int verboseFlag; /* Whether to print extra stuff or not */
183 const llist_t *excludeList; /* List of files to not include */
184 HardLinkInfo *hlInfoHead; /* Hard Link Tracking Information */
185 HardLinkInfo *hlInfo; /* Hard Link Info for the current file */
186 //TODO: save only st_dev + st_ino
187 struct stat tarFileStatBuf; /* Stat info for the tarball, letting
188 * us know the inode and device that the
189 * tarball lives, so we can avoid trying
190 * to include the tarball into itself */
193 /* A nice enum with all the possible tar file content types */
195 REGTYPE = '0', /* regular file */
196 REGTYPE0 = '\0', /* regular file (ancient bug compat) */
197 LNKTYPE = '1', /* hard link */
198 SYMTYPE = '2', /* symbolic link */
199 CHRTYPE = '3', /* character special */
200 BLKTYPE = '4', /* block special */
201 DIRTYPE = '5', /* directory */
202 FIFOTYPE = '6', /* FIFO special */
203 CONTTYPE = '7', /* reserved */
204 GNULONGLINK = 'K', /* GNU long (>100 chars) link name */
205 GNULONGNAME = 'L', /* GNU long (>100 chars) file name */
208 /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */
209 static void addHardLinkInfo(HardLinkInfo **hlInfoHeadPtr,
210 struct stat *statbuf,
211 const char *fileName)
213 /* Note: hlInfoHeadPtr can never be NULL! */
214 HardLinkInfo *hlInfo;
216 hlInfo = xmalloc(sizeof(HardLinkInfo) + strlen(fileName));
217 hlInfo->next = *hlInfoHeadPtr;
218 *hlInfoHeadPtr = hlInfo;
219 hlInfo->dev = statbuf->st_dev;
220 hlInfo->ino = statbuf->st_ino;
221 // hlInfo->linkCount = statbuf->st_nlink;
222 strcpy(hlInfo->name, fileName);
225 static void freeHardLinkInfo(HardLinkInfo **hlInfoHeadPtr)
227 HardLinkInfo *hlInfo;
228 HardLinkInfo *hlInfoNext;
231 hlInfo = *hlInfoHeadPtr;
233 hlInfoNext = hlInfo->next;
237 *hlInfoHeadPtr = NULL;
241 /* Might be faster (and bigger) if the dev/ino were stored in numeric order ;) */
242 static HardLinkInfo *findHardLinkInfo(HardLinkInfo *hlInfo, struct stat *statbuf)
245 if (statbuf->st_ino == hlInfo->ino
246 && statbuf->st_dev == hlInfo->dev
248 DBG("found hardlink:'%s'", hlInfo->name);
251 hlInfo = hlInfo->next;
256 /* Put an octal string into the specified buffer.
257 * The number is zero padded and possibly null terminated.
258 * Stores low-order bits only if whole value does not fit. */
259 static void putOctal(char *cp, int len, off_t value)
261 char tempBuffer[sizeof(off_t)*3 + 1];
262 char *tempString = tempBuffer;
265 width = sprintf(tempBuffer, "%0*"OFF_FMT"o", len, value);
266 tempString += (width - len);
268 /* If string has leading zeroes, we can drop one */
269 /* and field will have trailing '\0' */
270 /* (increases chances of compat with other tars) */
271 if (tempString[0] == '0')
274 /* Copy the string to the field */
275 memcpy(cp, tempString, len);
277 #define PUT_OCTAL(a, b) putOctal((a), sizeof(a), (b))
279 static void chksum_and_xwrite(int fd, struct tar_header_t* hp)
281 /* POSIX says that checksum is done on unsigned bytes
282 * (Sun and HP-UX gets it wrong... more details in
284 const unsigned char *cp;
287 strcpy(hp->magic, "ustar ");
289 /* Calculate and store the checksum (i.e., the sum of all of the bytes of
290 * the header). The checksum field must be filled with blanks for the
291 * calculation. The checksum field is formatted differently from the
292 * other fields: it has 6 digits, a null, then a space -- rather than
293 * digits, followed by a null like the other fields... */
294 memset(hp->chksum, ' ', sizeof(hp->chksum));
295 cp = (const unsigned char *) hp;
298 do { chksum += *cp++; } while (--size);
299 putOctal(hp->chksum, sizeof(hp->chksum)-1, chksum);
301 /* Now write the header out to disk */
302 xwrite(fd, hp, sizeof(*hp));
305 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
306 static void writeLongname(int fd, int type, const char *name, int dir)
308 static const struct {
309 char mode[8]; /* 100-107 */
310 char uid[8]; /* 108-115 */
311 char gid[8]; /* 116-123 */
312 char size[12]; /* 124-135 */
313 char mtime[12]; /* 136-147 */
321 struct tar_header_t header;
324 dir = !!dir; /* normalize: 0/1 */
325 size = strlen(name) + 1 + dir; /* GNU tar uses strlen+1 */
326 /* + dir: account for possible '/' */
328 memset(&header, 0, sizeof(header));
329 strcpy(header.name, "././@LongLink");
330 memcpy(header.mode, prefilled.mode, sizeof(prefilled));
331 PUT_OCTAL(header.size, size);
332 header.typeflag = type;
333 chksum_and_xwrite(fd, &header);
335 /* Write filename[/] and pad the block. */
336 /* dir=0: writes 'name<NUL>', pads */
337 /* dir=1: writes 'name', writes '/<NUL>', pads */
339 xwrite(fd, name, size - dir);
340 xwrite(fd, "/", dir);
341 size = (-size) & (TAR_BLOCK_SIZE-1);
342 memset(&header, 0, size);
343 xwrite(fd, &header, size);
347 /* Write out a tar header for the specified file/directory/whatever */
348 static int writeTarHeader(struct TarBallInfo *tbInfo,
349 const char *header_name, const char *fileName, struct stat *statbuf)
351 struct tar_header_t header;
353 memset(&header, 0, sizeof(header));
355 strncpy(header.name, header_name, sizeof(header.name));
357 /* POSIX says to mask mode with 07777. */
358 PUT_OCTAL(header.mode, statbuf->st_mode & 07777);
359 PUT_OCTAL(header.uid, statbuf->st_uid);
360 PUT_OCTAL(header.gid, statbuf->st_gid);
361 memset(header.size, '0', sizeof(header.size)-1); /* Regular file size is handled later */
362 /* users report that files with negative st_mtime cause trouble, so: */
363 PUT_OCTAL(header.mtime, statbuf->st_mtime >= 0 ? statbuf->st_mtime : 0);
365 /* Enter the user and group names */
366 safe_strncpy(header.uname, get_cached_username(statbuf->st_uid), sizeof(header.uname));
367 safe_strncpy(header.gname, get_cached_groupname(statbuf->st_gid), sizeof(header.gname));
369 if (tbInfo->hlInfo) {
370 /* This is a hard link */
371 header.typeflag = LNKTYPE;
372 strncpy(header.linkname, tbInfo->hlInfo->name,
373 sizeof(header.linkname));
374 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
375 /* Write out long linkname if needed */
376 if (header.linkname[sizeof(header.linkname)-1])
377 writeLongname(tbInfo->tarFd, GNULONGLINK,
378 tbInfo->hlInfo->name, 0);
380 } else if (S_ISLNK(statbuf->st_mode)) {
381 char *lpath = xmalloc_readlink_or_warn(fileName);
384 header.typeflag = SYMTYPE;
385 strncpy(header.linkname, lpath, sizeof(header.linkname));
386 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
387 /* Write out long linkname if needed */
388 if (header.linkname[sizeof(header.linkname)-1])
389 writeLongname(tbInfo->tarFd, GNULONGLINK, lpath, 0);
391 /* If it is larger than 100 bytes, bail out */
392 if (header.linkname[sizeof(header.linkname)-1]) {
394 bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported");
399 } else if (S_ISDIR(statbuf->st_mode)) {
400 header.typeflag = DIRTYPE;
401 /* Append '/' only if there is a space for it */
402 if (!header.name[sizeof(header.name)-1])
403 header.name[strlen(header.name)] = '/';
404 } else if (S_ISCHR(statbuf->st_mode)) {
405 header.typeflag = CHRTYPE;
406 PUT_OCTAL(header.devmajor, major(statbuf->st_rdev));
407 PUT_OCTAL(header.devminor, minor(statbuf->st_rdev));
408 } else if (S_ISBLK(statbuf->st_mode)) {
409 header.typeflag = BLKTYPE;
410 PUT_OCTAL(header.devmajor, major(statbuf->st_rdev));
411 PUT_OCTAL(header.devminor, minor(statbuf->st_rdev));
412 } else if (S_ISFIFO(statbuf->st_mode)) {
413 header.typeflag = FIFOTYPE;
414 } else if (S_ISREG(statbuf->st_mode)) {
415 /* header.size field is 12 bytes long */
416 /* Does octal-encoded size fit? */
417 uoff_t filesize = statbuf->st_size;
418 if (sizeof(filesize) <= 4
419 || filesize <= (uoff_t)0777777777777LL
421 PUT_OCTAL(header.size, filesize);
423 /* Does base256-encoded size fit?
424 * It always does unless off_t is wider than 64 bits.
426 else if (ENABLE_FEATURE_TAR_GNU_EXTENSIONS
427 #if ULLONG_MAX > 0xffffffffffffffffLL /* 2^64-1 */
428 && (filesize <= 0x3fffffffffffffffffffffffLL)
431 /* GNU tar uses "base-256 encoding" for very large numbers.
432 * Encoding is binary, with highest bit always set as a marker
433 * and sign in next-highest bit:
435 * bf ff .. ff - largest positive number
436 * ff ff .. ff - minus 1
437 * c0 00 .. 00 - smallest negative number
439 char *p8 = header.size + sizeof(header.size);
441 *--p8 = (uint8_t)filesize;
443 } while (p8 != header.size);
446 bb_error_msg_and_die("can't store file '%s' "
447 "of size %"OFF_FMT"u, aborting",
448 fileName, statbuf->st_size);
450 header.typeflag = REGTYPE;
452 bb_error_msg("%s: unknown file type", fileName);
456 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
457 /* Write out long name if needed */
458 /* (we, like GNU tar, output long linkname *before* long name) */
459 if (header.name[sizeof(header.name)-1])
460 writeLongname(tbInfo->tarFd, GNULONGNAME,
461 header_name, S_ISDIR(statbuf->st_mode));
464 /* Now write the header out to disk */
465 chksum_and_xwrite(tbInfo->tarFd, &header);
467 /* Now do the verbose thing (or not) */
468 if (tbInfo->verboseFlag) {
471 /* If archive goes to stdout, verbose goes to stderr */
472 if (tbInfo->tarFd == STDOUT_FILENO)
474 /* GNU "tar cvvf" prints "extended" listing a-la "ls -l" */
475 /* We don't have such excesses here: for us "v" == "vv" */
476 /* '/' is probably a GNUism */
477 fprintf(vbFd, "%s%s\n", header_name,
478 S_ISDIR(statbuf->st_mode) ? "/" : "");
484 #if ENABLE_FEATURE_TAR_FROM
485 static int exclude_file(const llist_t *excluded_files, const char *file)
487 while (excluded_files) {
488 if (excluded_files->data[0] == '/') {
489 if (fnmatch(excluded_files->data, file,
490 FNM_PATHNAME | FNM_LEADING_DIR) == 0)
495 for (p = file; p[0] != '\0'; p++) {
496 if ((p == file || p[-1] == '/')
498 && fnmatch(excluded_files->data, p,
499 FNM_PATHNAME | FNM_LEADING_DIR) == 0
505 excluded_files = excluded_files->link;
511 # define exclude_file(excluded_files, file) 0
514 static int FAST_FUNC writeFileToTarball(const char *fileName, struct stat *statbuf,
515 void *userData, int depth UNUSED_PARAM)
517 struct TarBallInfo *tbInfo = (struct TarBallInfo *) userData;
518 const char *header_name;
519 int inputFileFd = -1;
521 DBG("writeFileToTarball('%s')", fileName);
523 /* Strip leading '/' and such (must be before memorizing hardlink's name) */
524 header_name = strip_unsafe_prefix(fileName);
526 if (header_name[0] == '\0')
529 /* It is against the rules to archive a socket */
530 if (S_ISSOCK(statbuf->st_mode)) {
531 bb_error_msg("%s: socket ignored", fileName);
536 * Check to see if we are dealing with a hard link.
538 * Treat the first occurance of a given dev/inode as a file while
539 * treating any additional occurances as hard links. This is done
540 * by adding the file information to the HardLinkInfo linked list.
542 tbInfo->hlInfo = NULL;
543 if (!S_ISDIR(statbuf->st_mode) && statbuf->st_nlink > 1) {
544 DBG("'%s': st_nlink > 1", header_name);
545 tbInfo->hlInfo = findHardLinkInfo(tbInfo->hlInfoHead, statbuf);
546 if (tbInfo->hlInfo == NULL) {
547 DBG("'%s': addHardLinkInfo", header_name);
548 addHardLinkInfo(&tbInfo->hlInfoHead, statbuf, header_name);
552 /* It is a bad idea to store the archive we are in the process of creating,
553 * so check the device and inode to be sure that this particular file isn't
555 if (tbInfo->tarFileStatBuf.st_dev == statbuf->st_dev
556 && tbInfo->tarFileStatBuf.st_ino == statbuf->st_ino
558 bb_error_msg("%s: file is the archive; skipping", fileName);
562 if (exclude_file(tbInfo->excludeList, header_name))
565 #if !ENABLE_FEATURE_TAR_GNU_EXTENSIONS
566 if (strlen(header_name) >= NAME_SIZE) {
567 bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported");
572 /* Is this a regular file? */
573 if (tbInfo->hlInfo == NULL && S_ISREG(statbuf->st_mode)) {
574 /* open the file we want to archive, and make sure all is well */
575 inputFileFd = open_or_warn(fileName, O_RDONLY);
576 if (inputFileFd < 0) {
581 /* Add an entry to the tarball */
582 if (writeTarHeader(tbInfo, header_name, fileName, statbuf) == FALSE) {
586 /* If it was a regular file, write out the body */
587 if (inputFileFd >= 0) {
589 /* Write the file to the archive. */
590 /* We record size into header first, */
591 /* and then write out file. If file shrinks in between, */
592 /* tar will be corrupted. So we don't allow for that. */
593 /* NB: GNU tar 1.16 warns and pads with zeroes */
594 /* or even seeks back and updates header */
595 bb_copyfd_exact_size(inputFileFd, tbInfo->tarFd, statbuf->st_size);
597 ////readSize = bb_copyfd_size(inputFileFd, tbInfo->tarFd, statbuf->st_size);
598 ////if (readSize != statbuf->st_size && readSize >= 0) {
599 //// bb_error_msg_and_die("short read from %s, aborting", fileName);
602 /* Check that file did not grow in between? */
603 /* if (safe_read(inputFileFd, 1) == 1) warn but continue? */
607 /* Pad the file up to the tar block size */
608 /* (a few tricks here in the name of code size) */
609 readSize = (-(int)statbuf->st_size) & (TAR_BLOCK_SIZE-1);
610 memset(block_buf, 0, readSize);
611 xwrite(tbInfo->tarFd, block_buf, readSize);
617 #if SEAMLESS_COMPRESSION
618 /* Don't inline: vfork scares gcc and pessimizes code */
619 static void NOINLINE vfork_compressor(int tar_fd, const char *gzip)
623 // On Linux, vfork never unpauses parent early, although standard
624 // allows for that. Do we want to waste bytes checking for it?
625 # define WAIT_FOR_CHILD 0
626 volatile int vfork_exec_errno = 0;
627 struct fd_pair gzipDataPipe;
629 struct fd_pair gzipStatusPipe;
630 xpiped_pair(gzipStatusPipe);
632 xpiped_pair(gzipDataPipe);
634 signal(SIGPIPE, SIG_IGN); /* we only want EPIPE on errors */
640 /* NB: close _first_, then move fds! */
641 close(gzipDataPipe.wr);
643 close(gzipStatusPipe.rd);
644 /* gzipStatusPipe.wr will close only on exec -
645 * parent waits for this close to happen */
646 fcntl(gzipStatusPipe.wr, F_SETFD, FD_CLOEXEC);
648 xmove_fd(gzipDataPipe.rd, 0);
650 /* exec gzip/bzip2 program/applet */
651 BB_EXECLP(gzip, gzip, "-f", (char *)0);
652 vfork_exec_errno = errno;
657 xmove_fd(gzipDataPipe.wr, tar_fd);
658 close(gzipDataPipe.rd);
660 close(gzipStatusPipe.wr);
665 /* Wait until child execs (or fails to) */
666 n = full_read(gzipStatusPipe.rd, &buf, 1);
667 if (n < 0 /* && errno == EAGAIN */)
668 continue; /* try it again */
670 close(gzipStatusPipe.rd);
672 if (vfork_exec_errno) {
673 errno = vfork_exec_errno;
674 bb_perror_msg_and_die("can't execute '%s'", gzip);
677 #endif /* SEAMLESS_COMPRESSION */
680 #if !SEAMLESS_COMPRESSION
681 /* Do not pass gzip flag to writeTarFile() */
682 #define writeTarFile(tar_fd, verboseFlag, recurseFlags, include, exclude, gzip) \
683 writeTarFile(tar_fd, verboseFlag, recurseFlags, include, exclude)
685 /* gcc 4.2.1 inlines it, making code bigger */
686 static NOINLINE int writeTarFile(int tar_fd, int verboseFlag,
687 int recurseFlags, const llist_t *include,
688 const llist_t *exclude, const char *gzip)
690 int errorFlag = FALSE;
691 struct TarBallInfo tbInfo;
693 tbInfo.hlInfoHead = NULL;
694 tbInfo.tarFd = tar_fd;
695 tbInfo.verboseFlag = verboseFlag;
697 /* Store the stat info for the tarball's file, so
698 * can avoid including the tarball into itself.... */
699 xfstat(tbInfo.tarFd, &tbInfo.tarFileStatBuf, "can't stat tar file");
701 #if SEAMLESS_COMPRESSION
703 vfork_compressor(tbInfo.tarFd, gzip);
706 tbInfo.excludeList = exclude;
708 /* Read the directory/files and iterate over them one at a time */
710 if (!recursive_action(include->data, recurseFlags,
711 writeFileToTarball, writeFileToTarball, &tbInfo, 0)
715 include = include->link;
717 /* Write two empty blocks to the end of the archive */
718 memset(block_buf, 0, 2*TAR_BLOCK_SIZE);
719 xwrite(tbInfo.tarFd, block_buf, 2*TAR_BLOCK_SIZE);
721 /* To be pedantically correct, we would check if the tarball
722 * is smaller than 20 tar blocks, and pad it if it was smaller,
723 * but that isn't necessary for GNU tar interoperability, and
724 * so is considered a waste of space */
726 /* Close so the child process (if any) will exit */
729 /* Hang up the tools, close up shop, head home */
730 if (ENABLE_FEATURE_CLEAN_UP)
731 freeHardLinkInfo(&tbInfo.hlInfoHead);
734 bb_error_msg("error exit delayed from previous errors");
736 #if SEAMLESS_COMPRESSION
739 if (safe_waitpid(-1, &status, 0) == -1)
740 bb_perror_msg("waitpid");
741 else if (!WIFEXITED(status) || WEXITSTATUS(status))
742 /* gzip was killed or has exited with nonzero! */
748 #else /* !FEATURE_TAR_CREATE */
749 # define writeTarFile(...) 0
752 #if ENABLE_FEATURE_TAR_FROM
753 static llist_t *append_file_list_to_list(llist_t *list)
757 llist_t *newlist = NULL;
760 src_stream = xfopen_stdin(llist_pop(&list));
761 while ((line = xmalloc_fgetline(src_stream)) != NULL) {
762 /* kill trailing '/' unless the string is just "/" */
763 char *cp = last_char_is(line, '/');
766 llist_add_to_end(&newlist, line);
774 //usage:#define tar_trivial_usage
775 //usage: "-[" IF_FEATURE_TAR_CREATE("c") "xt"
776 //usage: IF_FEATURE_SEAMLESS_Z("Z")
777 //usage: IF_FEATURE_SEAMLESS_GZ("z")
778 //usage: IF_FEATURE_SEAMLESS_XZ("J")
779 //usage: IF_FEATURE_SEAMLESS_BZ2("j")
780 //usage: IF_FEATURE_SEAMLESS_LZMA("a")
781 //usage: IF_FEATURE_TAR_CREATE("h")
782 //usage: IF_FEATURE_TAR_NOPRESERVE_TIME("m")
784 //usage: IF_FEATURE_TAR_FROM("[-X FILE] [-T FILE] ")
785 //usage: "[-f TARFILE] [-C DIR] [FILE]..."
786 //usage:#define tar_full_usage "\n\n"
787 //usage: IF_FEATURE_TAR_CREATE("Create, extract, ")
788 //usage: IF_NOT_FEATURE_TAR_CREATE("Extract ")
789 //usage: "or list files from a tar file\n"
790 //usage: "\nOperation:"
791 //usage: IF_FEATURE_TAR_CREATE(
792 //usage: "\n c Create"
794 //usage: "\n x Extract"
796 //usage: "\n f Name of TARFILE ('-' for stdin/out)"
797 //usage: "\n C Change to DIR before operation"
798 //usage: "\n v Verbose"
799 //usage: IF_FEATURE_SEAMLESS_Z(
800 //usage: "\n Z (De)compress using compress"
802 //usage: IF_FEATURE_SEAMLESS_GZ(
803 //usage: "\n z (De)compress using gzip"
805 //usage: IF_FEATURE_SEAMLESS_XZ(
806 //usage: "\n J (De)compress using xz"
808 //usage: IF_FEATURE_SEAMLESS_BZ2(
809 //usage: "\n j (De)compress using bzip2"
811 //usage: IF_FEATURE_SEAMLESS_LZMA(
812 //usage: "\n a (De)compress using lzma"
814 //usage: "\n O Extract to stdout"
815 //usage: IF_FEATURE_TAR_CREATE(
816 //usage: "\n h Follow symlinks"
818 //usage: IF_FEATURE_TAR_NOPRESERVE_TIME(
819 //usage: "\n m Don't restore mtime"
821 //usage: IF_FEATURE_TAR_FROM(
822 //usage: IF_FEATURE_TAR_LONG_OPTIONS(
823 //usage: "\n exclude File to exclude"
825 //usage: "\n X File with names to exclude"
826 //usage: "\n T File with names to include"
829 //usage:#define tar_example_usage
830 //usage: "$ zcat /tmp/tarball.tar.gz | tar -xf -\n"
831 //usage: "$ tar -cf /tmp/tarball.tar /usr/local\n"
833 // Supported but aren't in --help:
835 // p same-permissions
839 // no-same-permissions
841 //IF_FEATURE_TAR_TO_COMMAND(
847 IF_FEATURE_TAR_CREATE( OPTBIT_CREATE ,)
848 IF_FEATURE_TAR_CREATE( OPTBIT_DEREFERENCE ,)
849 IF_FEATURE_SEAMLESS_BZ2( OPTBIT_BZIP2 ,)
850 IF_FEATURE_SEAMLESS_LZMA(OPTBIT_LZMA ,)
851 IF_FEATURE_TAR_FROM( OPTBIT_INCLUDE_FROM,)
852 IF_FEATURE_TAR_FROM( OPTBIT_EXCLUDE_FROM,)
853 IF_FEATURE_SEAMLESS_GZ( OPTBIT_GZIP ,)
854 IF_FEATURE_SEAMLESS_XZ( OPTBIT_XZ ,) // 16th bit
855 IF_FEATURE_SEAMLESS_Z( OPTBIT_COMPRESS ,)
856 IF_FEATURE_TAR_NOPRESERVE_TIME(OPTBIT_NOPRESERVE_TIME,)
857 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
859 IF_FEATURE_TAR_TO_COMMAND(OPTBIT_2COMMAND ,)
860 OPTBIT_NUMERIC_OWNER,
861 OPTBIT_NOPRESERVE_PERM,
864 OPT_TEST = 1 << 0, // t
865 OPT_EXTRACT = 1 << 1, // x
866 OPT_BASEDIR = 1 << 2, // C
867 OPT_TARNAME = 1 << 3, // f
868 OPT_2STDOUT = 1 << 4, // O
869 OPT_NOPRESERVE_OWNER = 1 << 5, // o == no-same-owner
871 OPT_VERBOSE = 1 << 7, // v
872 OPT_KEEP_OLD = 1 << 8, // k
873 OPT_CREATE = IF_FEATURE_TAR_CREATE( (1 << OPTBIT_CREATE )) + 0, // c
874 OPT_DEREFERENCE = IF_FEATURE_TAR_CREATE( (1 << OPTBIT_DEREFERENCE )) + 0, // h
875 OPT_BZIP2 = IF_FEATURE_SEAMLESS_BZ2( (1 << OPTBIT_BZIP2 )) + 0, // j
876 OPT_LZMA = IF_FEATURE_SEAMLESS_LZMA((1 << OPTBIT_LZMA )) + 0, // a
877 OPT_INCLUDE_FROM = IF_FEATURE_TAR_FROM( (1 << OPTBIT_INCLUDE_FROM)) + 0, // T
878 OPT_EXCLUDE_FROM = IF_FEATURE_TAR_FROM( (1 << OPTBIT_EXCLUDE_FROM)) + 0, // X
879 OPT_GZIP = IF_FEATURE_SEAMLESS_GZ( (1 << OPTBIT_GZIP )) + 0, // z
880 OPT_XZ = IF_FEATURE_SEAMLESS_XZ( (1 << OPTBIT_XZ )) + 0, // J
881 OPT_COMPRESS = IF_FEATURE_SEAMLESS_Z( (1 << OPTBIT_COMPRESS )) + 0, // Z
882 OPT_NOPRESERVE_TIME = IF_FEATURE_TAR_NOPRESERVE_TIME((1 << OPTBIT_NOPRESERVE_TIME)) + 0, // m
883 OPT_NORECURSION = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NORECURSION )) + 0, // no-recursion
884 OPT_2COMMAND = IF_FEATURE_TAR_TO_COMMAND( (1 << OPTBIT_2COMMAND )) + 0, // to-command
885 OPT_NUMERIC_OWNER = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NUMERIC_OWNER )) + 0, // numeric-owner
886 OPT_NOPRESERVE_PERM = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NOPRESERVE_PERM)) + 0, // no-same-permissions
887 OPT_OVERWRITE = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_OVERWRITE )) + 0, // overwrite
889 OPT_ANY_COMPRESS = (OPT_BZIP2 | OPT_LZMA | OPT_GZIP | OPT_XZ | OPT_COMPRESS),
891 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
892 static const char tar_longopts[] ALIGN1 =
893 "list\0" No_argument "t"
894 "extract\0" No_argument "x"
895 "directory\0" Required_argument "C"
896 "file\0" Required_argument "f"
897 "to-stdout\0" No_argument "O"
898 /* do not restore owner */
899 /* Note: GNU tar handles 'o' as no-same-owner only on extract,
900 * on create, 'o' is --old-archive. We do not support --old-archive. */
901 "no-same-owner\0" No_argument "o"
902 "same-permissions\0" No_argument "p"
903 "verbose\0" No_argument "v"
904 "keep-old\0" No_argument "k"
905 # if ENABLE_FEATURE_TAR_CREATE
906 "create\0" No_argument "c"
907 "dereference\0" No_argument "h"
909 # if ENABLE_FEATURE_SEAMLESS_BZ2
910 "bzip2\0" No_argument "j"
912 # if ENABLE_FEATURE_SEAMLESS_LZMA
913 "lzma\0" No_argument "a"
915 # if ENABLE_FEATURE_TAR_FROM
916 "files-from\0" Required_argument "T"
917 "exclude-from\0" Required_argument "X"
919 # if ENABLE_FEATURE_SEAMLESS_GZ
920 "gzip\0" No_argument "z"
922 # if ENABLE_FEATURE_SEAMLESS_XZ
923 "xz\0" No_argument "J"
925 # if ENABLE_FEATURE_SEAMLESS_Z
926 "compress\0" No_argument "Z"
928 # if ENABLE_FEATURE_TAR_NOPRESERVE_TIME
929 "touch\0" No_argument "m"
931 "no-recursion\0" No_argument "\xfa"
932 # if ENABLE_FEATURE_TAR_TO_COMMAND
933 "to-command\0" Required_argument "\xfb"
935 /* use numeric uid/gid from tar header, not textual */
936 "numeric-owner\0" No_argument "\xfc"
937 /* do not restore mode */
938 "no-same-permissions\0" No_argument "\xfd"
939 /* on unpack, open with O_TRUNC and !O_EXCL */
940 "overwrite\0" No_argument "\xfe"
941 /* --exclude takes next bit position in option mask, */
942 /* therefore we have to put it _after_ --no-same-permissions */
943 # if ENABLE_FEATURE_TAR_FROM
944 "exclude\0" Required_argument "\xff"
949 int tar_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
950 int tar_main(int argc UNUSED_PARAM, char **argv)
952 archive_handle_t *tar_handle;
953 char *base_dir = NULL;
954 const char *tar_filename = "-";
957 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
958 llist_t *excludes = NULL;
961 /* Initialise default values */
962 tar_handle = init_handle();
963 tar_handle->ah_flags = ARCHIVE_CREATE_LEADING_DIRS
964 | ARCHIVE_RESTORE_DATE
965 | ARCHIVE_UNLINK_OLD;
967 /* Apparently only root's tar preserves perms (see bug 3844) */
969 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_PERM;
971 /* Prepend '-' to the first argument if required */
972 opt_complementary = "--:" // first arg is options
973 "tt:vv:" // count -t,-v
974 IF_FEATURE_TAR_FROM("X::T::") // cumulative lists
975 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
976 "\xff::" // cumulative lists for --exclude
978 IF_FEATURE_TAR_CREATE("c:") "t:x:" // at least one of these is reqd
979 IF_FEATURE_TAR_CREATE("c--tx:t--cx:x--ct") // mutually exclusive
980 IF_NOT_FEATURE_TAR_CREATE("t--x:x--t"); // mutually exclusive
981 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
982 applet_long_options = tar_longopts;
985 if (argv[1] && argv[1][0] != '-') {
987 * 1st argument without dash handles options with parameters
988 * differently from dashed one: it takes *next argv[i]*
989 * as paramenter even if there are more chars in 1st argument:
990 * "tar fx TARFILE" - "x" is not taken as f's param
991 * but is interpreted as -x option
992 * "tar -xf TARFILE" - dashed equivalent of the above
993 * "tar -fx ..." - "x" is taken as f's param
994 * getopt32 wouldn't handle 1st command correctly.
995 * Unfortunately, people do use such commands.
996 * We massage argv[1] to work around it by moving 'f'
997 * to the end of the string.
998 * More contrived "tar fCx TARFILE DIR" still fails,
999 * but such commands are much less likely to be used.
1001 char *f = strchr(argv[1], 'f');
1003 while (f[1] != '\0') {
1011 opt = getopt32(argv,
1013 IF_FEATURE_TAR_CREATE( "ch" )
1014 IF_FEATURE_SEAMLESS_BZ2( "j" )
1015 IF_FEATURE_SEAMLESS_LZMA("a" )
1016 IF_FEATURE_TAR_FROM( "T:X:")
1017 IF_FEATURE_SEAMLESS_GZ( "z" )
1018 IF_FEATURE_SEAMLESS_XZ( "J" )
1019 IF_FEATURE_SEAMLESS_Z( "Z" )
1020 IF_FEATURE_TAR_NOPRESERVE_TIME("m")
1021 , &base_dir // -C dir
1022 , &tar_filename // -f filename
1023 IF_FEATURE_TAR_FROM(, &(tar_handle->accept)) // T
1024 IF_FEATURE_TAR_FROM(, &(tar_handle->reject)) // X
1025 IF_FEATURE_TAR_TO_COMMAND(, &(tar_handle->tar__to_command)) // --to-command
1026 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
1027 , &excludes // --exclude
1029 , &verboseFlag // combined count for -t and -v
1030 , &verboseFlag // combined count for -t and -v
1032 //bb_error_msg("opt:%08x", opt);
1035 if (verboseFlag) tar_handle->action_header = header_verbose_list;
1036 if (verboseFlag == 1) tar_handle->action_header = header_list;
1038 if (opt & OPT_EXTRACT)
1039 tar_handle->action_data = data_extract_all;
1041 if (opt & OPT_2STDOUT)
1042 tar_handle->action_data = data_extract_to_stdout;
1044 if (opt & OPT_2COMMAND) {
1045 putenv((char*)"TAR_FILETYPE=f");
1046 signal(SIGPIPE, SIG_IGN);
1047 tar_handle->action_data = data_extract_to_command;
1048 IF_FEATURE_TAR_TO_COMMAND(tar_handle->tar__to_command_shell = xstrdup(get_shell_name());)
1051 if (opt & OPT_KEEP_OLD)
1052 tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
1054 if (opt & OPT_NUMERIC_OWNER)
1055 tar_handle->ah_flags |= ARCHIVE_NUMERIC_OWNER;
1057 if (opt & OPT_NOPRESERVE_OWNER)
1058 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_OWNER;
1060 if (opt & OPT_NOPRESERVE_PERM)
1061 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_PERM;
1063 if (opt & OPT_OVERWRITE) {
1064 tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
1065 tar_handle->ah_flags |= ARCHIVE_O_TRUNC;
1068 if (opt & OPT_NOPRESERVE_TIME)
1069 tar_handle->ah_flags &= ~ARCHIVE_RESTORE_DATE;
1071 #if ENABLE_FEATURE_TAR_FROM
1072 tar_handle->reject = append_file_list_to_list(tar_handle->reject);
1073 # if ENABLE_FEATURE_TAR_LONG_OPTIONS
1074 /* Append excludes to reject */
1076 llist_t *next = excludes->link;
1077 excludes->link = tar_handle->reject;
1078 tar_handle->reject = excludes;
1082 tar_handle->accept = append_file_list_to_list(tar_handle->accept);
1085 /* Setup an array of filenames to work with */
1086 /* TODO: This is the same as in ar, make a separate function? */
1088 /* kill trailing '/' unless the string is just "/" */
1089 char *cp = last_char_is(*argv, '/');
1092 llist_add_to_end(&tar_handle->accept, *argv);
1096 if (tar_handle->accept || tar_handle->reject)
1097 tar_handle->filter = filter_accept_reject_list;
1099 /* Open the tar file */
1101 int tar_fd = STDIN_FILENO;
1102 int flags = O_RDONLY;
1104 if (opt & OPT_CREATE) {
1105 /* Make sure there is at least one file to tar up */
1106 if (tar_handle->accept == NULL)
1107 bb_error_msg_and_die("empty archive");
1109 tar_fd = STDOUT_FILENO;
1110 /* Mimicking GNU tar 1.15.1: */
1111 flags = O_WRONLY | O_CREAT | O_TRUNC;
1114 if (LONE_DASH(tar_filename)) {
1115 tar_handle->src_fd = tar_fd;
1116 tar_handle->seek = seek_by_read;
1118 if (ENABLE_FEATURE_TAR_AUTODETECT
1119 && flags == O_RDONLY
1120 && !(opt & OPT_ANY_COMPRESS)
1122 tar_handle->src_fd = open_zipped(tar_filename, /*fail_if_not_compressed:*/ 0);
1123 if (tar_handle->src_fd < 0)
1124 bb_perror_msg_and_die("can't open '%s'", tar_filename);
1126 tar_handle->src_fd = xopen(tar_filename, flags);
1134 //if (SEAMLESS_COMPRESSION)
1135 // /* We need to know whether child (gzip/bzip/etc) exits abnormally */
1136 // signal(SIGCHLD, check_errors_in_children);
1138 /* Create an archive */
1139 if (opt & OPT_CREATE) {
1140 #if SEAMLESS_COMPRESSION
1141 const char *zipMode = NULL;
1142 if (opt & OPT_COMPRESS)
1143 zipMode = "compress";
1146 if (opt & OPT_BZIP2)
1153 /* NB: writeTarFile() closes tar_handle->src_fd */
1154 return writeTarFile(tar_handle->src_fd, verboseFlag,
1155 (opt & OPT_DEREFERENCE ? ACTION_FOLLOWLINKS : 0)
1156 | (opt & OPT_NORECURSION ? 0 : ACTION_RECURSE),
1158 tar_handle->reject, zipMode);
1161 if (opt & OPT_ANY_COMPRESS) {
1162 USE_FOR_MMU(IF_DESKTOP(long long) int FAST_FUNC (*xformer)(transformer_state_t *xstate);)
1163 USE_FOR_NOMMU(const char *xformer_prog;)
1165 if (opt & OPT_COMPRESS)
1166 USE_FOR_MMU(xformer = unpack_Z_stream;)
1167 USE_FOR_NOMMU(xformer_prog = "uncompress";)
1169 USE_FOR_MMU(xformer = unpack_gz_stream;)
1170 USE_FOR_NOMMU(xformer_prog = "gunzip";)
1171 if (opt & OPT_BZIP2)
1172 USE_FOR_MMU(xformer = unpack_bz2_stream;)
1173 USE_FOR_NOMMU(xformer_prog = "bunzip2";)
1175 USE_FOR_MMU(xformer = unpack_lzma_stream;)
1176 USE_FOR_NOMMU(xformer_prog = "unlzma";)
1178 USE_FOR_MMU(xformer = unpack_xz_stream;)
1179 USE_FOR_NOMMU(xformer_prog = "unxz";)
1181 fork_transformer_with_sig(tar_handle->src_fd, xformer, xformer_prog);
1182 /* Can't lseek over pipes */
1183 tar_handle->seek = seek_by_read;
1184 /*tar_handle->offset = 0; - already is */
1187 /* Zero processed headers (== empty file) is not a valid tarball.
1188 * We (ab)use bb_got_signal as exitcode here,
1189 * because check_errors_in_children() uses _it_ as error indicator.
1191 bb_got_signal = EXIT_FAILURE;
1193 while (get_header_tar(tar_handle) == EXIT_SUCCESS)
1194 bb_got_signal = EXIT_SUCCESS; /* saw at least one header, good */
1196 /* Check that every file that should have been extracted was */
1197 while (tar_handle->accept) {
1198 if (!find_list_entry(tar_handle->reject, tar_handle->accept->data)
1199 && !find_list_entry(tar_handle->passed, tar_handle->accept->data)
1201 bb_error_msg_and_die("%s: not found in archive",
1202 tar_handle->accept->data);
1204 tar_handle->accept = tar_handle->accept->link;
1206 if (ENABLE_FEATURE_CLEAN_UP /* && tar_handle->src_fd != STDIN_FILENO */)
1207 close(tar_handle->src_fd);
1209 if (SEAMLESS_COMPRESSION || OPT_COMPRESS) {
1210 /* Set bb_got_signal to 1 if a child died with !0 exitcode */
1211 check_errors_in_children(0);
1214 return bb_got_signal;