1 /* vi: set sw=4 ts=4: */
3 * Mini tar implementation for busybox
5 * Modifed to use common extraction code used by ar, cpio, dpkg-deb, dpkg
6 * Glenn McGrath <bug1@optushome.com.au>
8 * Note, that as of BusyBox-0.43, tar has been completely rewritten from the
9 * ground up. It still has remnents of the old code lying about, but it is
10 * very different now (i.e., cleaner, less global variables, etc.)
12 * Copyright (C) 1999,2000 by Lineo, inc. and Erik Andersen
13 * Copyright (C) 1999-2002 by Erik Andersen <andersee@debian.org>
15 * Based in part in the tar implementation in sash
16 * Copyright (c) 1999 by David I. Bell
17 * Permission is granted to use, distribute, or modify this source,
18 * provided that this copyright notice remains intact.
19 * Permission to distribute sash derived code under the GPL has been granted.
21 * Based in part on the tar implementation from busybox-0.28
22 * Copyright (C) 1995 Bruce Perens
23 * This is free software under the GNU General Public License.
25 * This program is free software; you can redistribute it and/or modify
26 * it under the terms of the GNU General Public License as published by
27 * the Free Software Foundation; either version 2 of the License, or
28 * (at your option) any later version.
30 * This program is distributed in the hope that it will be useful,
31 * but WITHOUT ANY WARRANTY; without even the implied warranty of
32 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
33 * General Public License for more details.
35 * You should have received a copy of the GNU General Public License
36 * along with this program; if not, write to the Free Software
37 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
52 #include <sys/socket.h>
53 #include "unarchive.h"
56 #ifdef CONFIG_FEATURE_TAR_CREATE
58 /* Tar file constants */
59 # define TAR_MAGIC "ustar" /* ustar and a null */
60 # define TAR_VERSION " " /* Be compatable with GNU tar format */
63 # define MAJOR(dev) (((dev)>>8)&0xff)
64 # define MINOR(dev) ((dev)&0xff)
67 static const int TAR_BLOCK_SIZE = 512;
68 static const int TAR_MAGIC_LEN = 6;
69 static const int TAR_VERSION_LEN = 2;
71 /* POSIX tar Header Block, from POSIX 1003.1-1990 */
72 enum { NAME_SIZE = 100 }; /* because gcc won't let me use 'static const int' */
73 struct TarHeader { /* byte offset */
74 char name[NAME_SIZE]; /* 0-99 */
75 char mode[8]; /* 100-107 */
76 char uid[8]; /* 108-115 */
77 char gid[8]; /* 116-123 */
78 char size[12]; /* 124-135 */
79 char mtime[12]; /* 136-147 */
80 char chksum[8]; /* 148-155 */
81 char typeflag; /* 156-156 */
82 char linkname[NAME_SIZE]; /* 157-256 */
83 char magic[6]; /* 257-262 */
84 char version[2]; /* 263-264 */
85 char uname[32]; /* 265-296 */
86 char gname[32]; /* 297-328 */
87 char devmajor[8]; /* 329-336 */
88 char devminor[8]; /* 337-344 */
89 char prefix[155]; /* 345-499 */
90 char padding[12]; /* 500-512 (pad to exactly the TAR_BLOCK_SIZE) */
92 typedef struct TarHeader TarHeader;
95 ** writeTarFile(), writeFileToTarball(), and writeTarHeader() are
96 ** the only functions that deal with the HardLinkInfo structure.
97 ** Even these functions use the xxxHardLinkInfo() functions.
99 typedef struct HardLinkInfo HardLinkInfo;
100 struct HardLinkInfo {
101 HardLinkInfo *next; /* Next entry in list */
102 dev_t dev; /* Device number */
103 ino_t ino; /* Inode number */
104 short linkCount; /* (Hard) Link Count */
105 char name[1]; /* Start of filename (must be last) */
108 /* Some info to be carried along when creating a new tarball */
110 char *fileName; /* File name of the tarball */
111 int tarFd; /* Open-for-write file descriptor
113 struct stat statBuf; /* Stat info for the tarball, letting
114 us know the inode and device that the
115 tarball lives, so we can avoid trying
116 to include the tarball into itself */
117 int verboseFlag; /* Whether to print extra stuff or not */
118 char **excludeList; /* List of files to not include */
119 HardLinkInfo *hlInfoHead; /* Hard Link Tracking Information */
120 HardLinkInfo *hlInfo; /* Hard Link Info for the current file */
122 typedef struct TarBallInfo TarBallInfo;
124 /* A nice enum with all the possible tar file content types */
126 REGTYPE = '0', /* regular file */
127 REGTYPE0 = '\0', /* regular file (ancient bug compat) */
128 LNKTYPE = '1', /* hard link */
129 SYMTYPE = '2', /* symbolic link */
130 CHRTYPE = '3', /* character special */
131 BLKTYPE = '4', /* block special */
132 DIRTYPE = '5', /* directory */
133 FIFOTYPE = '6', /* FIFO special */
134 CONTTYPE = '7', /* reserved */
135 GNULONGLINK = 'K', /* GNU long (>100 chars) link name */
136 GNULONGNAME = 'L', /* GNU long (>100 chars) file name */
138 typedef enum TarFileType TarFileType;
140 /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */
141 static inline void addHardLinkInfo(HardLinkInfo ** hlInfoHeadPtr, dev_t dev,
142 ino_t ino, short linkCount,
145 /* Note: hlInfoHeadPtr can never be NULL! */
146 HardLinkInfo *hlInfo;
149 (HardLinkInfo *) xmalloc(sizeof(HardLinkInfo) + strlen(name) + 1);
151 hlInfo->next = *hlInfoHeadPtr;
152 *hlInfoHeadPtr = hlInfo;
155 hlInfo->linkCount = linkCount;
156 strcpy(hlInfo->name, name);
161 static void freeHardLinkInfo(HardLinkInfo ** hlInfoHeadPtr)
163 HardLinkInfo *hlInfo = NULL;
164 HardLinkInfo *hlInfoNext = NULL;
167 hlInfo = *hlInfoHeadPtr;
169 hlInfoNext = hlInfo->next;
173 *hlInfoHeadPtr = NULL;
178 /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */
179 static inline HardLinkInfo *findHardLinkInfo(HardLinkInfo * hlInfo, dev_t dev,
183 if ((ino == hlInfo->ino) && (dev == hlInfo->dev))
185 hlInfo = hlInfo->next;
190 /* Put an octal string into the specified buffer.
191 * The number is zero and space padded and possibly null padded.
192 * Returns TRUE if successful. */
193 static int putOctal(char *cp, int len, long value)
197 char *tempString = tempBuffer;
199 /* Create a string of the specified length with an initial space,
200 * leading zeroes and the octal number, and a trailing null. */
201 sprintf(tempString, "%0*lo", len - 1, value);
203 /* If the string is too large, suppress the leading space. */
204 tempLength = strlen(tempString) + 1;
205 if (tempLength > len) {
210 /* If the string is still too large, suppress the trailing null. */
211 if (tempLength > len)
214 /* If the string is still too large, fail. */
215 if (tempLength > len)
218 /* Copy the string to the field. */
219 memcpy(cp, tempString, len);
224 /* Write out a tar header for the specified file/directory/whatever */
225 static inline int writeTarHeader(struct TarBallInfo *tbInfo,
226 const char *header_name,
227 const char *real_name, struct stat *statbuf)
230 struct TarHeader header;
231 const unsigned char *cp = (const unsigned char *) &header;
232 ssize_t size = sizeof(struct TarHeader);
234 memset(&header, 0, size);
236 strncpy(header.name, header_name, sizeof(header.name));
238 putOctal(header.mode, sizeof(header.mode), statbuf->st_mode);
239 putOctal(header.uid, sizeof(header.uid), statbuf->st_uid);
240 putOctal(header.gid, sizeof(header.gid), statbuf->st_gid);
241 putOctal(header.size, sizeof(header.size), 0); /* Regular file size is handled later */
242 putOctal(header.mtime, sizeof(header.mtime), statbuf->st_mtime);
243 strncpy(header.magic, TAR_MAGIC TAR_VERSION,
244 TAR_MAGIC_LEN + TAR_VERSION_LEN);
246 /* Enter the user and group names (default to root if it fails) */
247 my_getpwuid(header.uname, statbuf->st_uid);
249 strcpy(header.uname, "root");
250 my_getgrgid(header.gname, statbuf->st_gid);
252 strcpy(header.uname, "root");
254 if (tbInfo->hlInfo) {
255 /* This is a hard link */
256 header.typeflag = LNKTYPE;
257 strncpy(header.linkname, tbInfo->hlInfo->name,
258 sizeof(header.linkname));
259 } else if (S_ISLNK(statbuf->st_mode)) {
260 char *lpath = xreadlink(real_name);
262 if (!lpath) /* Already printed err msg inside xreadlink() */
264 header.typeflag = SYMTYPE;
265 strncpy(header.linkname, lpath, sizeof(header.linkname));
267 } else if (S_ISDIR(statbuf->st_mode)) {
268 header.typeflag = DIRTYPE;
269 strncat(header.name, "/", sizeof(header.name));
270 } else if (S_ISCHR(statbuf->st_mode)) {
271 header.typeflag = CHRTYPE;
272 putOctal(header.devmajor, sizeof(header.devmajor),
273 MAJOR(statbuf->st_rdev));
274 putOctal(header.devminor, sizeof(header.devminor),
275 MINOR(statbuf->st_rdev));
276 } else if (S_ISBLK(statbuf->st_mode)) {
277 header.typeflag = BLKTYPE;
278 putOctal(header.devmajor, sizeof(header.devmajor),
279 MAJOR(statbuf->st_rdev));
280 putOctal(header.devminor, sizeof(header.devminor),
281 MINOR(statbuf->st_rdev));
282 } else if (S_ISFIFO(statbuf->st_mode)) {
283 header.typeflag = FIFOTYPE;
284 } else if (S_ISREG(statbuf->st_mode)) {
285 header.typeflag = REGTYPE;
286 putOctal(header.size, sizeof(header.size), statbuf->st_size);
288 error_msg("%s: Unknown file type", real_name);
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(header.chksum, ' ', sizeof(header.chksum));
298 cp = (const unsigned char *) &header;
301 putOctal(header.chksum, 7, chksum);
303 /* Now write the header out to disk */
305 full_write(tbInfo->tarFd, (char *) &header,
306 sizeof(struct TarHeader))) < 0) {
307 error_msg(io_error, real_name, strerror(errno));
310 /* Pad the header up to the tar block size */
311 for (; size < TAR_BLOCK_SIZE; size++) {
312 write(tbInfo->tarFd, "\0", 1);
314 /* Now do the verbose thing (or not) */
316 if (tbInfo->verboseFlag) {
319 if (tbInfo->verboseFlag == 2) /* If the archive goes to stdout, verbose to stderr */
321 fprintf(vbFd, "%s\n", header.name);
327 # if defined CONFIG_FEATURE_TAR_EXCLUDE
328 static inline int exclude_file(char **excluded_files, const char *file)
332 if (excluded_files == NULL)
335 for (i = 0; excluded_files[i] != NULL; i++) {
336 if (excluded_files[i][0] == '/') {
337 if (fnmatch(excluded_files[i], file,
338 FNM_PATHNAME | FNM_LEADING_DIR) == 0)
343 for (p = file; p[0] != '\0'; p++) {
344 if ((p == file || p[-1] == '/') && p[0] != '/' &&
345 fnmatch(excluded_files[i], p,
346 FNM_PATHNAME | FNM_LEADING_DIR) == 0)
356 static int writeFileToTarball(const char *fileName, struct stat *statbuf,
359 struct TarBallInfo *tbInfo = (struct TarBallInfo *) userData;
360 const char *header_name;
363 ** Check to see if we are dealing with a hard link.
365 ** Treat the first occurance of a given dev/inode as a file while
366 ** treating any additional occurances as hard links. This is done
367 ** by adding the file information to the HardLinkInfo linked list.
369 tbInfo->hlInfo = NULL;
370 if (statbuf->st_nlink > 1) {
371 tbInfo->hlInfo = findHardLinkInfo(tbInfo->hlInfoHead, statbuf->st_dev,
373 if (tbInfo->hlInfo == NULL)
374 addHardLinkInfo(&tbInfo->hlInfoHead, statbuf->st_dev,
375 statbuf->st_ino, statbuf->st_nlink, fileName);
378 /* It is against the rules to archive a socket */
379 if (S_ISSOCK(statbuf->st_mode)) {
380 error_msg("%s: socket ignored", fileName);
384 /* It is a bad idea to store the archive we are in the process of creating,
385 * so check the device and inode to be sure that this particular file isn't
387 if (tbInfo->statBuf.st_dev == statbuf->st_dev &&
388 tbInfo->statBuf.st_ino == statbuf->st_ino) {
389 error_msg("%s: file is the archive; skipping", fileName);
393 header_name = fileName;
394 while (header_name[0] == '/') {
395 static int alreadyWarned = FALSE;
397 if (alreadyWarned == FALSE) {
398 error_msg("Removing leading '/' from member names");
399 alreadyWarned = TRUE;
404 if (strlen(fileName) >= NAME_SIZE) {
405 error_msg(name_longer_than_foo, NAME_SIZE);
409 if (header_name[0] == '\0')
412 # if defined CONFIG_FEATURE_TAR_EXCLUDE
413 if (exclude_file(tbInfo->excludeList, header_name)) {
416 # endif /* CONFIG_FEATURE_TAR_EXCLUDE */
418 if (writeTarHeader(tbInfo, header_name, fileName, statbuf) == FALSE) {
422 /* Now, if the file is a regular file, copy it out to the tarball */
423 if ((tbInfo->hlInfo == NULL)
424 && (S_ISREG(statbuf->st_mode))) {
427 ssize_t size = 0, readSize = 0;
429 /* open the file we want to archive, and make sure all is well */
430 if ((inputFileFd = open(fileName, O_RDONLY)) < 0) {
431 error_msg("%s: Cannot open: %s", fileName, strerror(errno));
435 /* write the file to the archive */
436 while ((size = full_read(inputFileFd, buffer, sizeof(buffer))) > 0) {
437 if (full_write(tbInfo->tarFd, buffer, size) != size) {
438 /* Output file seems to have a problem */
439 error_msg(io_error, fileName, strerror(errno));
445 error_msg(io_error, fileName, strerror(errno));
448 /* Pad the file up to the tar block size */
449 for (; (readSize % TAR_BLOCK_SIZE) != 0; readSize++) {
450 write(tbInfo->tarFd, "\0", 1);
458 static inline int writeTarFile(const char *tarName, int verboseFlag,
459 char **argv, char **excludeList, int gzip)
461 #ifdef CONFIG_FEATURE_TAR_GZIP
462 int gzipDataPipe[2] = { -1, -1 };
463 int gzipStatusPipe[2] = { -1, -1 };
467 int errorFlag = FALSE;
469 struct TarBallInfo tbInfo;
471 tbInfo.hlInfoHead = NULL;
473 /* Make sure there is at least one file to tar up. */
475 error_msg_and_die("Cowardly refusing to create an empty archive");
477 /* Open the tar file for writing. */
478 if (tarName == NULL) {
479 tbInfo.tarFd = fileno(stdout);
480 tbInfo.verboseFlag = verboseFlag ? 2 : 0;
482 tbInfo.tarFd = open(tarName, O_WRONLY | O_CREAT | O_TRUNC, 0644);
483 tbInfo.verboseFlag = verboseFlag ? 1 : 0;
486 if (tbInfo.tarFd < 0) {
487 perror_msg("Error opening '%s'", tarName);
488 freeHardLinkInfo(&tbInfo.hlInfoHead);
492 /* Store the stat info for the tarball's file, so
493 * can avoid including the tarball into itself.... */
494 if (fstat(tbInfo.tarFd, &tbInfo.statBuf) < 0)
495 error_msg_and_die(io_error, tarName, strerror(errno));
497 #ifdef CONFIG_FEATURE_TAR_GZIP
499 if (socketpair(AF_UNIX, SOCK_STREAM, 0, gzipDataPipe) < 0
500 || pipe(gzipStatusPipe) < 0)
501 perror_msg_and_die("Failed to create gzip pipe");
503 signal(SIGPIPE, SIG_IGN); /* we only want EPIPE on errors */
508 dup2(gzipDataPipe[0], 0);
509 close(gzipDataPipe[1]);
511 if (tbInfo.tarFd != 1);
512 dup2(tbInfo.tarFd, 1);
514 close(gzipStatusPipe[0]);
515 fcntl(gzipStatusPipe[1], F_SETFD, FD_CLOEXEC); /* close on exec shows sucess */
517 execl("/bin/gzip", "gzip", "-f", 0);
519 write(gzipStatusPipe[1], "", 1);
520 close(gzipStatusPipe[1]);
523 } else if (gzipPid > 0) {
524 close(gzipDataPipe[0]);
525 close(gzipStatusPipe[1]);
530 int n = read(gzipStatusPipe[0], &buf, 1);
533 error_msg_and_die("Could not exec gzip process"); /* socket was not closed => error */
534 else if ((n < 0) && (errno == EAGAIN || errno == EINTR))
535 continue; /* try it again */
538 close(gzipStatusPipe[0]);
540 tbInfo.tarFd = gzipDataPipe[1];
542 perror_msg_and_die("Failed to fork gzip process");
547 tbInfo.excludeList = excludeList;
549 /* Read the directory/files and iterate over them one at a time */
550 while (*argv != NULL) {
551 if (!recursive_action(*argv++, TRUE, FALSE, FALSE,
552 writeFileToTarball, writeFileToTarball,
557 /* Write two empty blocks to the end of the archive */
558 for (size = 0; size < (2 * TAR_BLOCK_SIZE); size++) {
559 write(tbInfo.tarFd, "\0", 1);
562 /* To be pedantically correct, we would check if the tarball
563 * is smaller than 20 tar blocks, and pad it if it was smaller,
564 * but that isn't necessary for GNU tar interoperability, and
565 * so is considered a waste of space */
567 /* Hang up the tools, close up shop, head home */
570 error_msg("Error exit delayed from previous errors");
572 freeHardLinkInfo(&tbInfo.hlInfoHead);
574 #ifdef CONFIG_FEATURE_TAR_GZIP
575 if (gzip && gzipPid) {
576 if (waitpid(gzipPid, NULL, 0) == -1)
577 printf("Couldnt wait ?");
583 #endif /* tar_create */
585 void append_file_to_list(const char *new_name, char ***list, int *list_count)
587 *list = realloc(*list, sizeof(char *) * (*list_count + 2));
588 (*list)[*list_count] = xstrdup(new_name);
590 (*list)[*list_count] = NULL;
593 void append_file_list_to_list(char *filename, char ***name_list,
599 src_stream = xfopen(filename, "r");
600 while ((line = get_line_from_file(src_stream)) != NULL) {
602 append_file_to_list(line, name_list, num_of_entries);
608 int tar_main(int argc, char **argv)
611 /* This is optional */
613 /* Require one and only one of these */
619 FILE *src_stream = NULL;
620 FILE *uncompressed_stream = NULL;
621 char **include_list = NULL;
622 char **exclude_list = NULL;
623 char *src_filename = NULL;
624 char *dst_prefix = NULL;
626 unsigned short untar_funct = 0;
627 unsigned short untar_funct_required = 0;
628 unsigned short extract_function = 0;
629 int include_list_count = 0;
631 #ifdef CONFIG_FEATURE_TAR_EXCLUDE
632 int exclude_list_count = 0;
634 #ifdef CONFIG_FEATURE_TAR_GZIP
643 /* Prepend '-' to the first argument if required */
644 if (argv[1][0] != '-') {
645 char *tmp = xmalloc(strlen(argv[1]) + 2);
648 strcpy(tmp + 1, argv[1]);
652 while ((opt = getopt(argc, argv, "ctxT:X:C:f:Opvz")) != -1) {
655 /* One and only one of these is required */
657 untar_funct_required |= untar_create;
660 untar_funct_required |= untar_list;
661 extract_function |= extract_list | extract_unconditional;
664 untar_funct_required |= untar_extract;
666 (extract_all_to_fs | extract_unconditional |
667 extract_create_leading_dirs);
670 /* These are optional */
671 /* Exclude or Include files listed in <filename> */
672 #ifdef CONFIG_FEATURE_TAR_EXCLUDE
674 append_file_list_to_list(optarg, &exclude_list,
675 &exclude_list_count);
679 /* by default a list is an include list */
680 append_file_list_to_list(optarg, &include_list,
681 &include_list_count);
684 case 'C': /* Change to dir <optarg> */
685 /* Make sure dst_prefix ends in a '/' */
686 dst_prefix = concat_path_file(optarg, "/");
688 case 'f': /* archive filename */
689 if (strcmp(optarg, "-") == 0) {
692 src_filename = xstrdup(optarg);
696 extract_function |= extract_to_stdout;
701 extract_function |= extract_verbose_list;
703 #ifdef CONFIG_FEATURE_TAR_GZIP
705 untar_funct |= untar_unzip;
713 /* Make sure the valid arguments were passed */
714 if (untar_funct_required == 0) {
715 error_msg_and_die("You must specify one of the `-ctx' options");
717 if ((untar_funct_required != untar_create) &&
718 (untar_funct_required != untar_extract) &&
719 (untar_funct_required != untar_list)) {
720 error_msg_and_die("You may not specify more than one `ctx' option.");
722 untar_funct |= untar_funct_required;
724 /* Setup an array of filenames to work with */
725 while (optind < argc) {
726 append_file_to_list(argv[optind], &include_list, &include_list_count);
729 if (extract_function & (extract_list | extract_all_to_fs)) {
730 if (dst_prefix == NULL) {
731 dst_prefix = xstrdup("./");
734 /* Setup the source of the tar data */
735 if (src_filename != NULL) {
736 src_stream = xfopen(src_filename, "r");
740 #ifdef CONFIG_FEATURE_TAR_GZIP
741 /* Get a binary tree of all the tar file headers */
742 if (untar_funct & untar_unzip) {
743 uncompressed_stream = gz_open(src_stream, &gunzip_pid);
745 #endif /* CONFIG_FEATURE_TAR_GZIP */
746 uncompressed_stream = src_stream;
748 /* extract or list archive */
749 unarchive(uncompressed_stream, stdout, &get_header_tar,
750 extract_function, dst_prefix, include_list, exclude_list);
751 fclose(uncompressed_stream);
753 #ifdef CONFIG_FEATURE_TAR_CREATE
754 /* create an archive */
755 else if (untar_funct & untar_create) {
756 int verboseFlag = FALSE;
757 int gzipFlag = FALSE;
759 #ifdef CONFIG_FEATURE_TAR_GZIP
760 if (untar_funct & untar_unzip)
763 #endif /* CONFIG_FEATURE_TAR_GZIP */
764 if (extract_function & extract_verbose_list)
767 writeTarFile(src_filename, verboseFlag, include_list, exclude_list,
770 #endif /* CONFIG_FEATURE_TAR_CREATE */
773 #ifdef CONFIG_FEATURE_TAR_GZIP
774 if (!(untar_funct & untar_create) && (untar_funct & untar_unzip)) {
777 gz_close(gunzip_pid);
779 #endif /* CONFIG_FEATURE_TAR_GZIP */
780 #ifdef CONFIG_FEATURE_CLEAN_UP
785 return (EXIT_SUCCESS);