ebaa965c051007129ec29ce543370d7516ba8a69
[oweals/busybox.git] / archival / tar.c
1 /* vi: set sw=4 ts=4: */
2 /*
3  * Mini tar implementation for busybox
4  *
5  * Modified to use common extraction code used by ar, cpio, dpkg-deb, dpkg
6  *  by Glenn McGrath
7  *
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.)
11  *
12  * Copyright (C) 1999-2004 by Erik Andersen <andersen@codepoet.org>
13  *
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.
19  *
20  * Based in part on the tar implementation from busybox-0.28
21  *  Copyright (C) 1995 Bruce Perens
22  *
23  * Licensed under GPLv2 or later, see file LICENSE in this source tree.
24  */
25
26 #include <fnmatch.h>
27 #include "libbb.h"
28 #include "archive.h"
29 /* FIXME: Stop using this non-standard feature */
30 #ifndef FNM_LEADING_DIR
31 # define FNM_LEADING_DIR 0
32 #endif
33
34
35 //#define DBG(fmt, ...) bb_error_msg("%s: " fmt, __func__, ## __VA_ARGS__)
36 #define DBG(...) ((void)0)
37
38
39 #define block_buf bb_common_bufsiz1
40
41
42 #if !ENABLE_FEATURE_SEAMLESS_GZ && !ENABLE_FEATURE_SEAMLESS_BZ2
43 /* Do not pass gzip flag to writeTarFile() */
44 #define writeTarFile(tar_fd, verboseFlag, dereferenceFlag, include, exclude, gzip) \
45         writeTarFile(tar_fd, verboseFlag, dereferenceFlag, include, exclude)
46 #endif
47
48
49 #if ENABLE_FEATURE_TAR_CREATE
50
51 /*
52 ** writeTarFile(), writeFileToTarball(), and writeTarHeader() are
53 ** the only functions that deal with the HardLinkInfo structure.
54 ** Even these functions use the xxxHardLinkInfo() functions.
55 */
56 typedef struct HardLinkInfo {
57         struct HardLinkInfo *next; /* Next entry in list */
58         dev_t dev;                 /* Device number */
59         ino_t ino;                 /* Inode number */
60 //      short linkCount;           /* (Hard) Link Count */
61         char name[1];              /* Start of filename (must be last) */
62 } HardLinkInfo;
63
64 /* Some info to be carried along when creating a new tarball */
65 typedef struct TarBallInfo {
66         int tarFd;                      /* Open-for-write file descriptor
67                                          * for the tarball */
68         int verboseFlag;                /* Whether to print extra stuff or not */
69         const llist_t *excludeList;     /* List of files to not include */
70         HardLinkInfo *hlInfoHead;       /* Hard Link Tracking Information */
71         HardLinkInfo *hlInfo;           /* Hard Link Info for the current file */
72 //TODO: save only st_dev + st_ino
73         struct stat tarFileStatBuf;     /* Stat info for the tarball, letting
74                                          * us know the inode and device that the
75                                          * tarball lives, so we can avoid trying
76                                          * to include the tarball into itself */
77 } TarBallInfo;
78
79 /* A nice enum with all the possible tar file content types */
80 enum {
81         REGTYPE = '0',          /* regular file */
82         REGTYPE0 = '\0',        /* regular file (ancient bug compat) */
83         LNKTYPE = '1',          /* hard link */
84         SYMTYPE = '2',          /* symbolic link */
85         CHRTYPE = '3',          /* character special */
86         BLKTYPE = '4',          /* block special */
87         DIRTYPE = '5',          /* directory */
88         FIFOTYPE = '6',         /* FIFO special */
89         CONTTYPE = '7',         /* reserved */
90         GNULONGLINK = 'K',      /* GNU long (>100 chars) link name */
91         GNULONGNAME = 'L',      /* GNU long (>100 chars) file name */
92 };
93
94 /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */
95 static void addHardLinkInfo(HardLinkInfo **hlInfoHeadPtr,
96                                         struct stat *statbuf,
97                                         const char *fileName)
98 {
99         /* Note: hlInfoHeadPtr can never be NULL! */
100         HardLinkInfo *hlInfo;
101
102         hlInfo = xmalloc(sizeof(HardLinkInfo) + strlen(fileName));
103         hlInfo->next = *hlInfoHeadPtr;
104         *hlInfoHeadPtr = hlInfo;
105         hlInfo->dev = statbuf->st_dev;
106         hlInfo->ino = statbuf->st_ino;
107 //      hlInfo->linkCount = statbuf->st_nlink;
108         strcpy(hlInfo->name, fileName);
109 }
110
111 static void freeHardLinkInfo(HardLinkInfo **hlInfoHeadPtr)
112 {
113         HardLinkInfo *hlInfo;
114         HardLinkInfo *hlInfoNext;
115
116         if (hlInfoHeadPtr) {
117                 hlInfo = *hlInfoHeadPtr;
118                 while (hlInfo) {
119                         hlInfoNext = hlInfo->next;
120                         free(hlInfo);
121                         hlInfo = hlInfoNext;
122                 }
123                 *hlInfoHeadPtr = NULL;
124         }
125 }
126
127 /* Might be faster (and bigger) if the dev/ino were stored in numeric order ;) */
128 static HardLinkInfo *findHardLinkInfo(HardLinkInfo *hlInfo, struct stat *statbuf)
129 {
130         while (hlInfo) {
131                 if (statbuf->st_ino == hlInfo->ino
132                  && statbuf->st_dev == hlInfo->dev
133                 ) {
134                         DBG("found hardlink:'%s'", hlInfo->name);
135                         break;
136                 }
137                 hlInfo = hlInfo->next;
138         }
139         return hlInfo;
140 }
141
142 /* Put an octal string into the specified buffer.
143  * The number is zero padded and possibly null terminated.
144  * Stores low-order bits only if whole value does not fit. */
145 static void putOctal(char *cp, int len, off_t value)
146 {
147         char tempBuffer[sizeof(off_t)*3 + 1];
148         char *tempString = tempBuffer;
149         int width;
150
151         width = sprintf(tempBuffer, "%0*"OFF_FMT"o", len, value);
152         tempString += (width - len);
153
154         /* If string has leading zeroes, we can drop one */
155         /* and field will have trailing '\0' */
156         /* (increases chances of compat with other tars) */
157         if (tempString[0] == '0')
158                 tempString++;
159
160         /* Copy the string to the field */
161         memcpy(cp, tempString, len);
162 }
163 #define PUT_OCTAL(a, b) putOctal((a), sizeof(a), (b))
164
165 static void chksum_and_xwrite(int fd, struct tar_header_t* hp)
166 {
167         /* POSIX says that checksum is done on unsigned bytes
168          * (Sun and HP-UX gets it wrong... more details in
169          * GNU tar source) */
170         const unsigned char *cp;
171         int chksum, size;
172
173         strcpy(hp->magic, "ustar  ");
174
175         /* Calculate and store the checksum (i.e., the sum of all of the bytes of
176          * the header).  The checksum field must be filled with blanks for the
177          * calculation.  The checksum field is formatted differently from the
178          * other fields: it has 6 digits, a null, then a space -- rather than
179          * digits, followed by a null like the other fields... */
180         memset(hp->chksum, ' ', sizeof(hp->chksum));
181         cp = (const unsigned char *) hp;
182         chksum = 0;
183         size = sizeof(*hp);
184         do { chksum += *cp++; } while (--size);
185         putOctal(hp->chksum, sizeof(hp->chksum)-1, chksum);
186
187         /* Now write the header out to disk */
188         xwrite(fd, hp, sizeof(*hp));
189 }
190
191 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
192 static void writeLongname(int fd, int type, const char *name, int dir)
193 {
194         static const struct {
195                 char mode[8];             /* 100-107 */
196                 char uid[8];              /* 108-115 */
197                 char gid[8];              /* 116-123 */
198                 char size[12];            /* 124-135 */
199                 char mtime[12];           /* 136-147 */
200         } prefilled = {
201                 "0000000",
202                 "0000000",
203                 "0000000",
204                 "00000000000",
205                 "00000000000",
206         };
207         struct tar_header_t header;
208         int size;
209
210         dir = !!dir; /* normalize: 0/1 */
211         size = strlen(name) + 1 + dir; /* GNU tar uses strlen+1 */
212         /* + dir: account for possible '/' */
213
214         memset(&header, 0, sizeof(header));
215         strcpy(header.name, "././@LongLink");
216         memcpy(header.mode, prefilled.mode, sizeof(prefilled));
217         PUT_OCTAL(header.size, size);
218         header.typeflag = type;
219         chksum_and_xwrite(fd, &header);
220
221         /* Write filename[/] and pad the block. */
222         /* dir=0: writes 'name<NUL>', pads */
223         /* dir=1: writes 'name', writes '/<NUL>', pads */
224         dir *= 2;
225         xwrite(fd, name, size - dir);
226         xwrite(fd, "/", dir);
227         size = (-size) & (TAR_BLOCK_SIZE-1);
228         memset(&header, 0, size);
229         xwrite(fd, &header, size);
230 }
231 #endif
232
233 /* Write out a tar header for the specified file/directory/whatever */
234 static int writeTarHeader(struct TarBallInfo *tbInfo,
235                 const char *header_name, const char *fileName, struct stat *statbuf)
236 {
237         struct tar_header_t header;
238
239         memset(&header, 0, sizeof(header));
240
241         strncpy(header.name, header_name, sizeof(header.name));
242
243         /* POSIX says to mask mode with 07777. */
244         PUT_OCTAL(header.mode, statbuf->st_mode & 07777);
245         PUT_OCTAL(header.uid, statbuf->st_uid);
246         PUT_OCTAL(header.gid, statbuf->st_gid);
247         memset(header.size, '0', sizeof(header.size)-1); /* Regular file size is handled later */
248         PUT_OCTAL(header.mtime, statbuf->st_mtime);
249
250         /* Enter the user and group names */
251         safe_strncpy(header.uname, get_cached_username(statbuf->st_uid), sizeof(header.uname));
252         safe_strncpy(header.gname, get_cached_groupname(statbuf->st_gid), sizeof(header.gname));
253
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 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
260                 /* Write out long linkname if needed */
261                 if (header.linkname[sizeof(header.linkname)-1])
262                         writeLongname(tbInfo->tarFd, GNULONGLINK,
263                                         tbInfo->hlInfo->name, 0);
264 #endif
265         } else if (S_ISLNK(statbuf->st_mode)) {
266                 char *lpath = xmalloc_readlink_or_warn(fileName);
267                 if (!lpath)
268                         return FALSE;
269                 header.typeflag = SYMTYPE;
270                 strncpy(header.linkname, lpath, sizeof(header.linkname));
271 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
272                 /* Write out long linkname if needed */
273                 if (header.linkname[sizeof(header.linkname)-1])
274                         writeLongname(tbInfo->tarFd, GNULONGLINK, lpath, 0);
275 #else
276                 /* If it is larger than 100 bytes, bail out */
277                 if (header.linkname[sizeof(header.linkname)-1]) {
278                         free(lpath);
279                         bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported");
280                         return FALSE;
281                 }
282 #endif
283                 free(lpath);
284         } else if (S_ISDIR(statbuf->st_mode)) {
285                 header.typeflag = DIRTYPE;
286                 /* Append '/' only if there is a space for it */
287                 if (!header.name[sizeof(header.name)-1])
288                         header.name[strlen(header.name)] = '/';
289         } else if (S_ISCHR(statbuf->st_mode)) {
290                 header.typeflag = CHRTYPE;
291                 PUT_OCTAL(header.devmajor, major(statbuf->st_rdev));
292                 PUT_OCTAL(header.devminor, minor(statbuf->st_rdev));
293         } else if (S_ISBLK(statbuf->st_mode)) {
294                 header.typeflag = BLKTYPE;
295                 PUT_OCTAL(header.devmajor, major(statbuf->st_rdev));
296                 PUT_OCTAL(header.devminor, minor(statbuf->st_rdev));
297         } else if (S_ISFIFO(statbuf->st_mode)) {
298                 header.typeflag = FIFOTYPE;
299         } else if (S_ISREG(statbuf->st_mode)) {
300                 if (sizeof(statbuf->st_size) > 4
301                  && statbuf->st_size > (off_t)0777777777777LL
302                 ) {
303                         bb_error_msg_and_die("can't store file '%s' "
304                                 "of size %"OFF_FMT"u, aborting",
305                                 fileName, statbuf->st_size);
306                 }
307                 header.typeflag = REGTYPE;
308                 PUT_OCTAL(header.size, statbuf->st_size);
309         } else {
310                 bb_error_msg("%s: unknown file type", fileName);
311                 return FALSE;
312         }
313
314 #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
315         /* Write out long name if needed */
316         /* (we, like GNU tar, output long linkname *before* long name) */
317         if (header.name[sizeof(header.name)-1])
318                 writeLongname(tbInfo->tarFd, GNULONGNAME,
319                                 header_name, S_ISDIR(statbuf->st_mode));
320 #endif
321
322         /* Now write the header out to disk */
323         chksum_and_xwrite(tbInfo->tarFd, &header);
324
325         /* Now do the verbose thing (or not) */
326         if (tbInfo->verboseFlag) {
327                 FILE *vbFd = stdout;
328
329                 /* If archive goes to stdout, verbose goes to stderr */
330                 if (tbInfo->tarFd == STDOUT_FILENO)
331                         vbFd = stderr;
332                 /* GNU "tar cvvf" prints "extended" listing a-la "ls -l" */
333                 /* We don't have such excesses here: for us "v" == "vv" */
334                 /* '/' is probably a GNUism */
335                 fprintf(vbFd, "%s%s\n", header_name,
336                                 S_ISDIR(statbuf->st_mode) ? "/" : "");
337         }
338
339         return TRUE;
340 }
341
342 #if ENABLE_FEATURE_TAR_FROM
343 static int exclude_file(const llist_t *excluded_files, const char *file)
344 {
345         while (excluded_files) {
346                 if (excluded_files->data[0] == '/') {
347                         if (fnmatch(excluded_files->data, file,
348                                         FNM_PATHNAME | FNM_LEADING_DIR) == 0)
349                                 return 1;
350                 } else {
351                         const char *p;
352
353                         for (p = file; p[0] != '\0'; p++) {
354                                 if ((p == file || p[-1] == '/')
355                                  && p[0] != '/'
356                                  && fnmatch(excluded_files->data, p,
357                                                 FNM_PATHNAME | FNM_LEADING_DIR) == 0
358                                 ) {
359                                         return 1;
360                                 }
361                         }
362                 }
363                 excluded_files = excluded_files->link;
364         }
365
366         return 0;
367 }
368 #else
369 # define exclude_file(excluded_files, file) 0
370 #endif
371
372 static int FAST_FUNC writeFileToTarball(const char *fileName, struct stat *statbuf,
373                         void *userData, int depth UNUSED_PARAM)
374 {
375         struct TarBallInfo *tbInfo = (struct TarBallInfo *) userData;
376         const char *header_name;
377         int inputFileFd = -1;
378
379         DBG("writeFileToTarball('%s')", fileName);
380
381         /* Strip leading '/' (must be before memorizing hardlink's name) */
382         header_name = fileName;
383         while (header_name[0] == '/') {
384                 static smallint warned;
385
386                 if (!warned) {
387                         bb_error_msg("removing leading '/' from member names");
388                         warned = 1;
389                 }
390                 header_name++;
391         }
392
393         if (header_name[0] == '\0')
394                 return TRUE;
395
396         /* It is against the rules to archive a socket */
397         if (S_ISSOCK(statbuf->st_mode)) {
398                 bb_error_msg("%s: socket ignored", fileName);
399                 return TRUE;
400         }
401
402         /*
403          * Check to see if we are dealing with a hard link.
404          * If so -
405          * Treat the first occurance of a given dev/inode as a file while
406          * treating any additional occurances as hard links.  This is done
407          * by adding the file information to the HardLinkInfo linked list.
408          */
409         tbInfo->hlInfo = NULL;
410         if (!S_ISDIR(statbuf->st_mode) && statbuf->st_nlink > 1) {
411                 DBG("'%s': st_nlink > 1", header_name);
412                 tbInfo->hlInfo = findHardLinkInfo(tbInfo->hlInfoHead, statbuf);
413                 if (tbInfo->hlInfo == NULL) {
414                         DBG("'%s': addHardLinkInfo", header_name);
415                         addHardLinkInfo(&tbInfo->hlInfoHead, statbuf, header_name);
416                 }
417         }
418
419         /* It is a bad idea to store the archive we are in the process of creating,
420          * so check the device and inode to be sure that this particular file isn't
421          * the new tarball */
422         if (tbInfo->tarFileStatBuf.st_dev == statbuf->st_dev
423          && tbInfo->tarFileStatBuf.st_ino == statbuf->st_ino
424         ) {
425                 bb_error_msg("%s: file is the archive; skipping", fileName);
426                 return TRUE;
427         }
428
429         if (exclude_file(tbInfo->excludeList, header_name))
430                 return SKIP;
431
432 #if !ENABLE_FEATURE_TAR_GNU_EXTENSIONS
433         if (strlen(header_name) >= NAME_SIZE) {
434                 bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported");
435                 return TRUE;
436         }
437 #endif
438
439         /* Is this a regular file? */
440         if (tbInfo->hlInfo == NULL && S_ISREG(statbuf->st_mode)) {
441                 /* open the file we want to archive, and make sure all is well */
442                 inputFileFd = open_or_warn(fileName, O_RDONLY);
443                 if (inputFileFd < 0) {
444                         return FALSE;
445                 }
446         }
447
448         /* Add an entry to the tarball */
449         if (writeTarHeader(tbInfo, header_name, fileName, statbuf) == FALSE) {
450                 return FALSE;
451         }
452
453         /* If it was a regular file, write out the body */
454         if (inputFileFd >= 0) {
455                 size_t readSize;
456                 /* Write the file to the archive. */
457                 /* We record size into header first, */
458                 /* and then write out file. If file shrinks in between, */
459                 /* tar will be corrupted. So we don't allow for that. */
460                 /* NB: GNU tar 1.16 warns and pads with zeroes */
461                 /* or even seeks back and updates header */
462                 bb_copyfd_exact_size(inputFileFd, tbInfo->tarFd, statbuf->st_size);
463                 ////off_t readSize;
464                 ////readSize = bb_copyfd_size(inputFileFd, tbInfo->tarFd, statbuf->st_size);
465                 ////if (readSize != statbuf->st_size && readSize >= 0) {
466                 ////    bb_error_msg_and_die("short read from %s, aborting", fileName);
467                 ////}
468
469                 /* Check that file did not grow in between? */
470                 /* if (safe_read(inputFileFd, 1) == 1) warn but continue? */
471
472                 close(inputFileFd);
473
474                 /* Pad the file up to the tar block size */
475                 /* (a few tricks here in the name of code size) */
476                 readSize = (-(int)statbuf->st_size) & (TAR_BLOCK_SIZE-1);
477                 memset(block_buf, 0, readSize);
478                 xwrite(tbInfo->tarFd, block_buf, readSize);
479         }
480
481         return TRUE;
482 }
483
484 #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2
485 # if !(ENABLE_FEATURE_SEAMLESS_GZ && ENABLE_FEATURE_SEAMLESS_BZ2)
486 #  define vfork_compressor(tar_fd, gzip) vfork_compressor(tar_fd)
487 # endif
488 /* Don't inline: vfork scares gcc and pessimizes code */
489 static void NOINLINE vfork_compressor(int tar_fd, int gzip)
490 {
491         pid_t gzipPid;
492 # if ENABLE_FEATURE_SEAMLESS_GZ && ENABLE_FEATURE_SEAMLESS_BZ2
493         const char *zip_exec = (gzip == 1) ? "gzip" : "bzip2";
494 # elif ENABLE_FEATURE_SEAMLESS_GZ
495         const char *zip_exec = "gzip";
496 # else /* only ENABLE_FEATURE_SEAMLESS_BZ2 */
497         const char *zip_exec = "bzip2";
498 # endif
499         // On Linux, vfork never unpauses parent early, although standard
500         // allows for that. Do we want to waste bytes checking for it?
501 # define WAIT_FOR_CHILD 0
502         volatile int vfork_exec_errno = 0;
503         struct fd_pair gzipDataPipe;
504 # if WAIT_FOR_CHILD
505         struct fd_pair gzipStatusPipe;
506         xpiped_pair(gzipStatusPipe);
507 # endif
508         xpiped_pair(gzipDataPipe);
509
510         signal(SIGPIPE, SIG_IGN); /* we only want EPIPE on errors */
511
512 # if defined(__GNUC__) && __GNUC__
513         /* Avoid vfork clobbering */
514         (void) &zip_exec;
515 # endif
516
517         gzipPid = xvfork();
518
519         if (gzipPid == 0) {
520                 /* child */
521                 /* NB: close _first_, then move fds! */
522                 close(gzipDataPipe.wr);
523 # if WAIT_FOR_CHILD
524                 close(gzipStatusPipe.rd);
525                 /* gzipStatusPipe.wr will close only on exec -
526                  * parent waits for this close to happen */
527                 fcntl(gzipStatusPipe.wr, F_SETFD, FD_CLOEXEC);
528 # endif
529                 xmove_fd(gzipDataPipe.rd, 0);
530                 xmove_fd(tar_fd, 1);
531                 /* exec gzip/bzip2 program/applet */
532                 BB_EXECLP(zip_exec, zip_exec, "-f", NULL);
533                 vfork_exec_errno = errno;
534                 _exit(EXIT_FAILURE);
535         }
536
537         /* parent */
538         xmove_fd(gzipDataPipe.wr, tar_fd);
539         close(gzipDataPipe.rd);
540 # if WAIT_FOR_CHILD
541         close(gzipStatusPipe.wr);
542         while (1) {
543                 char buf;
544                 int n;
545
546                 /* Wait until child execs (or fails to) */
547                 n = full_read(gzipStatusPipe.rd, &buf, 1);
548                 if (n < 0 /* && errno == EAGAIN */)
549                         continue;       /* try it again */
550         }
551         close(gzipStatusPipe.rd);
552 # endif
553         if (vfork_exec_errno) {
554                 errno = vfork_exec_errno;
555                 bb_perror_msg_and_die("can't execute '%s'", zip_exec);
556         }
557 }
558 #endif /* ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2 */
559
560
561 /* gcc 4.2.1 inlines it, making code bigger */
562 static NOINLINE int writeTarFile(int tar_fd, int verboseFlag,
563         int dereferenceFlag, const llist_t *include,
564         const llist_t *exclude, int gzip)
565 {
566         int errorFlag = FALSE;
567         struct TarBallInfo tbInfo;
568
569         tbInfo.hlInfoHead = NULL;
570         tbInfo.tarFd = tar_fd;
571         tbInfo.verboseFlag = verboseFlag;
572
573         /* Store the stat info for the tarball's file, so
574          * can avoid including the tarball into itself....  */
575         xfstat(tbInfo.tarFd, &tbInfo.tarFileStatBuf, "can't stat tar file");
576
577 #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2
578         if (gzip)
579                 vfork_compressor(tbInfo.tarFd, gzip);
580 #endif
581
582         tbInfo.excludeList = exclude;
583
584         /* Read the directory/files and iterate over them one at a time */
585         while (include) {
586                 if (!recursive_action(include->data, ACTION_RECURSE |
587                                 (dereferenceFlag ? ACTION_FOLLOWLINKS : 0),
588                                 writeFileToTarball, writeFileToTarball, &tbInfo, 0)
589                 ) {
590                         errorFlag = TRUE;
591                 }
592                 include = include->link;
593         }
594         /* Write two empty blocks to the end of the archive */
595         memset(block_buf, 0, 2*TAR_BLOCK_SIZE);
596         xwrite(tbInfo.tarFd, block_buf, 2*TAR_BLOCK_SIZE);
597
598         /* To be pedantically correct, we would check if the tarball
599          * is smaller than 20 tar blocks, and pad it if it was smaller,
600          * but that isn't necessary for GNU tar interoperability, and
601          * so is considered a waste of space */
602
603         /* Close so the child process (if any) will exit */
604         close(tbInfo.tarFd);
605
606         /* Hang up the tools, close up shop, head home */
607         if (ENABLE_FEATURE_CLEAN_UP)
608                 freeHardLinkInfo(&tbInfo.hlInfoHead);
609
610         if (errorFlag)
611                 bb_error_msg("error exit delayed from previous errors");
612
613 #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2
614         if (gzip) {
615                 int status;
616                 if (safe_waitpid(-1, &status, 0) == -1)
617                         bb_perror_msg("waitpid");
618                 else if (!WIFEXITED(status) || WEXITSTATUS(status))
619                         /* gzip was killed or has exited with nonzero! */
620                         errorFlag = TRUE;
621         }
622 #endif
623         return errorFlag;
624 }
625 #else
626 int writeTarFile(int tar_fd, int verboseFlag,
627         int dereferenceFlag, const llist_t *include,
628         const llist_t *exclude, int gzip);
629 #endif /* FEATURE_TAR_CREATE */
630
631 #if ENABLE_FEATURE_TAR_FROM
632 static llist_t *append_file_list_to_list(llist_t *list)
633 {
634         FILE *src_stream;
635         char *line;
636         llist_t *newlist = NULL;
637
638         while (list) {
639                 src_stream = xfopen_for_read(llist_pop(&list));
640                 while ((line = xmalloc_fgetline(src_stream)) != NULL) {
641                         /* kill trailing '/' unless the string is just "/" */
642                         char *cp = last_char_is(line, '/');
643                         if (cp > line)
644                                 *cp = '\0';
645                         llist_add_to(&newlist, line);
646                 }
647                 fclose(src_stream);
648         }
649         return newlist;
650 }
651 #else
652 # define append_file_list_to_list(x) 0
653 #endif
654
655 #if ENABLE_FEATURE_SEAMLESS_Z
656 static char FAST_FUNC get_header_tar_Z(archive_handle_t *archive_handle)
657 {
658         /* Can't lseek over pipes */
659         archive_handle->seek = seek_by_read;
660
661         /* do the decompression, and cleanup */
662         if (xread_char(archive_handle->src_fd) != 0x1f
663          || xread_char(archive_handle->src_fd) != 0x9d
664         ) {
665                 bb_error_msg_and_die("invalid magic");
666         }
667
668         open_transformer(archive_handle->src_fd, unpack_Z_stream, "uncompress");
669         archive_handle->offset = 0;
670         while (get_header_tar(archive_handle) == EXIT_SUCCESS)
671                 continue;
672
673         /* Can only do one file at a time */
674         return EXIT_FAILURE;
675 }
676 #else
677 # define get_header_tar_Z NULL
678 #endif
679
680 #ifdef CHECK_FOR_CHILD_EXITCODE
681 /* Looks like it isn't needed - tar detects malformed (truncated)
682  * archive if e.g. bunzip2 fails */
683 static int child_error;
684
685 static void handle_SIGCHLD(int status)
686 {
687         /* Actually, 'status' is a signo. We reuse it for other needs */
688
689         /* Wait for any child without blocking */
690         if (wait_any_nohang(&status) < 0)
691                 /* wait failed?! I'm confused... */
692                 return;
693
694         if (WIFEXITED(status) && WEXITSTATUS(status) == 0)
695                 /* child exited with 0 */
696                 return;
697         /* Cannot happen?
698         if (!WIFSIGNALED(status) && !WIFEXITED(status)) return; */
699         child_error = 1;
700 }
701 #endif
702
703 //usage:#define tar_trivial_usage
704 //usage:       "-[" IF_FEATURE_TAR_CREATE("c") "xt" IF_FEATURE_SEAMLESS_GZ("z")
705 //usage:        IF_FEATURE_SEAMLESS_BZ2("j") IF_FEATURE_SEAMLESS_LZMA("a")
706 //usage:        IF_FEATURE_SEAMLESS_Z("Z") IF_FEATURE_TAR_NOPRESERVE_TIME("m") "vO] "
707 //usage:        IF_FEATURE_TAR_FROM("[-X FILE] ")
708 //usage:       "[-f TARFILE] [-C DIR] [FILE]..."
709 //usage:#define tar_full_usage "\n\n"
710 //usage:        IF_FEATURE_TAR_CREATE("Create, extract, ")
711 //usage:        IF_NOT_FEATURE_TAR_CREATE("Extract ")
712 //usage:        "or list files from a tar file\n"
713 //usage:     "\nOperation:"
714 //usage:        IF_FEATURE_TAR_CREATE(
715 //usage:     "\n        c       Create"
716 //usage:        )
717 //usage:     "\n        x       Extract"
718 //usage:     "\n        t       List"
719 //usage:     "\nOptions:"
720 //usage:     "\n        f       Name of TARFILE ('-' for stdin/out)"
721 //usage:     "\n        C       Change to DIR before operation"
722 //usage:     "\n        v       Verbose"
723 //usage:        IF_FEATURE_SEAMLESS_GZ(
724 //usage:     "\n        z       (De)compress using gzip"
725 //usage:        )
726 //usage:        IF_FEATURE_SEAMLESS_BZ2(
727 //usage:     "\n        j       (De)compress using bzip2"
728 //usage:        )
729 //usage:        IF_FEATURE_SEAMLESS_LZMA(
730 //usage:     "\n        a       (De)compress using lzma"
731 //usage:        )
732 //usage:        IF_FEATURE_SEAMLESS_Z(
733 //usage:     "\n        Z       (De)compress using compress"
734 //usage:        )
735 //usage:     "\n        O       Extract to stdout"
736 //usage:        IF_FEATURE_TAR_CREATE(
737 //usage:     "\n        h       Follow symlinks"
738 //usage:        )
739 //usage:        IF_FEATURE_TAR_NOPRESERVE_TIME(
740 //usage:     "\n        m       Don't restore mtime"
741 //usage:        )
742 //usage:        IF_FEATURE_TAR_FROM(
743 //usage:        IF_FEATURE_TAR_LONG_OPTIONS(
744 //usage:     "\n        exclude File to exclude"
745 //usage:        )
746 //usage:     "\n        X       File with names to exclude"
747 //usage:     "\n        T       File with names to include"
748 //usage:        )
749 //usage:
750 //usage:#define tar_example_usage
751 //usage:       "$ zcat /tmp/tarball.tar.gz | tar -xf -\n"
752 //usage:       "$ tar -cf /tmp/tarball.tar /usr/local\n"
753
754 // Supported but aren't in --help:
755 //      o       no-same-owner
756 //      p       same-permissions
757 //      k       keep-old
758 //      numeric-owner
759 //      no-same-permissions
760 //      overwrite
761 //IF_FEATURE_TAR_TO_COMMAND(
762 //      to-command
763 //)
764
765 enum {
766         OPTBIT_KEEP_OLD = 8,
767         IF_FEATURE_TAR_CREATE(   OPTBIT_CREATE      ,)
768         IF_FEATURE_TAR_CREATE(   OPTBIT_DEREFERENCE ,)
769         IF_FEATURE_SEAMLESS_BZ2( OPTBIT_BZIP2       ,)
770         IF_FEATURE_SEAMLESS_LZMA(OPTBIT_LZMA        ,)
771         IF_FEATURE_TAR_FROM(     OPTBIT_INCLUDE_FROM,)
772         IF_FEATURE_TAR_FROM(     OPTBIT_EXCLUDE_FROM,)
773         IF_FEATURE_SEAMLESS_GZ(  OPTBIT_GZIP        ,)
774         IF_FEATURE_SEAMLESS_Z(   OPTBIT_COMPRESS    ,) // 16th bit
775         IF_FEATURE_TAR_NOPRESERVE_TIME(OPTBIT_NOPRESERVE_TIME,)
776 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
777         IF_FEATURE_TAR_TO_COMMAND(OPTBIT_2COMMAND   ,)
778         OPTBIT_NUMERIC_OWNER,
779         OPTBIT_NOPRESERVE_PERM,
780         OPTBIT_OVERWRITE,
781 #endif
782         OPT_TEST         = 1 << 0, // t
783         OPT_EXTRACT      = 1 << 1, // x
784         OPT_BASEDIR      = 1 << 2, // C
785         OPT_TARNAME      = 1 << 3, // f
786         OPT_2STDOUT      = 1 << 4, // O
787         OPT_NOPRESERVE_OWNER = 1 << 5, // o == no-same-owner
788         OPT_P            = 1 << 6, // p
789         OPT_VERBOSE      = 1 << 7, // v
790         OPT_KEEP_OLD     = 1 << 8, // k
791         OPT_CREATE       = IF_FEATURE_TAR_CREATE(   (1 << OPTBIT_CREATE      )) + 0, // c
792         OPT_DEREFERENCE  = IF_FEATURE_TAR_CREATE(   (1 << OPTBIT_DEREFERENCE )) + 0, // h
793         OPT_BZIP2        = IF_FEATURE_SEAMLESS_BZ2( (1 << OPTBIT_BZIP2       )) + 0, // j
794         OPT_LZMA         = IF_FEATURE_SEAMLESS_LZMA((1 << OPTBIT_LZMA        )) + 0, // a
795         OPT_INCLUDE_FROM = IF_FEATURE_TAR_FROM(     (1 << OPTBIT_INCLUDE_FROM)) + 0, // T
796         OPT_EXCLUDE_FROM = IF_FEATURE_TAR_FROM(     (1 << OPTBIT_EXCLUDE_FROM)) + 0, // X
797         OPT_GZIP         = IF_FEATURE_SEAMLESS_GZ(  (1 << OPTBIT_GZIP        )) + 0, // z
798         OPT_COMPRESS     = IF_FEATURE_SEAMLESS_Z(   (1 << OPTBIT_COMPRESS    )) + 0, // Z
799         OPT_NOPRESERVE_TIME = IF_FEATURE_TAR_NOPRESERVE_TIME((1 << OPTBIT_NOPRESERVE_TIME)) + 0, // m
800         OPT_2COMMAND        = IF_FEATURE_TAR_TO_COMMAND(  (1 << OPTBIT_2COMMAND       )) + 0, // to-command
801         OPT_NUMERIC_OWNER   = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NUMERIC_OWNER  )) + 0, // numeric-owner
802         OPT_NOPRESERVE_PERM = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_NOPRESERVE_PERM)) + 0, // no-same-permissions
803         OPT_OVERWRITE       = IF_FEATURE_TAR_LONG_OPTIONS((1 << OPTBIT_OVERWRITE      )) + 0, // overwrite
804 };
805 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
806 static const char tar_longopts[] ALIGN1 =
807         "list\0"                No_argument       "t"
808         "extract\0"             No_argument       "x"
809         "directory\0"           Required_argument "C"
810         "file\0"                Required_argument "f"
811         "to-stdout\0"           No_argument       "O"
812         /* do not restore owner */
813         /* Note: GNU tar handles 'o' as no-same-owner only on extract,
814          * on create, 'o' is --old-archive. We do not support --old-archive. */
815         "no-same-owner\0"       No_argument       "o"
816         "same-permissions\0"    No_argument       "p"
817         "verbose\0"             No_argument       "v"
818         "keep-old\0"            No_argument       "k"
819 # if ENABLE_FEATURE_TAR_CREATE
820         "create\0"              No_argument       "c"
821         "dereference\0"         No_argument       "h"
822 # endif
823 # if ENABLE_FEATURE_SEAMLESS_BZ2
824         "bzip2\0"               No_argument       "j"
825 # endif
826 # if ENABLE_FEATURE_SEAMLESS_LZMA
827         "lzma\0"                No_argument       "a"
828 # endif
829 # if ENABLE_FEATURE_TAR_FROM
830         "files-from\0"          Required_argument "T"
831         "exclude-from\0"        Required_argument "X"
832 # endif
833 # if ENABLE_FEATURE_SEAMLESS_GZ
834         "gzip\0"                No_argument       "z"
835 # endif
836 # if ENABLE_FEATURE_SEAMLESS_Z
837         "compress\0"            No_argument       "Z"
838 # endif
839 # if ENABLE_FEATURE_TAR_NOPRESERVE_TIME
840         "touch\0"               No_argument       "m"
841 # endif
842 # if ENABLE_FEATURE_TAR_TO_COMMAND
843         "to-command\0"          Required_argument "\xfb"
844 # endif
845         /* use numeric uid/gid from tar header, not textual */
846         "numeric-owner\0"       No_argument       "\xfc"
847         /* do not restore mode */
848         "no-same-permissions\0" No_argument       "\xfd"
849         /* on unpack, open with O_TRUNC and !O_EXCL */
850         "overwrite\0"           No_argument       "\xfe"
851         /* --exclude takes next bit position in option mask, */
852         /* therefore we have to put it _after_ --no-same-permissions */
853 # if ENABLE_FEATURE_TAR_FROM
854         "exclude\0"             Required_argument "\xff"
855 # endif
856         ;
857 #endif
858
859 int tar_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
860 int tar_main(int argc UNUSED_PARAM, char **argv)
861 {
862         char FAST_FUNC (*get_header_ptr)(archive_handle_t *) = get_header_tar;
863         archive_handle_t *tar_handle;
864         char *base_dir = NULL;
865         const char *tar_filename = "-";
866         unsigned opt;
867         int verboseFlag = 0;
868 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
869         llist_t *excludes = NULL;
870 #endif
871
872         /* Initialise default values */
873         tar_handle = init_handle();
874         tar_handle->ah_flags = ARCHIVE_CREATE_LEADING_DIRS
875                              | ARCHIVE_RESTORE_DATE
876                              | ARCHIVE_UNLINK_OLD;
877
878         /* Apparently only root's tar preserves perms (see bug 3844) */
879         if (getuid() != 0)
880                 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_PERM;
881
882         /* Prepend '-' to the first argument if required */
883         opt_complementary = "--:" // first arg is options
884                 "tt:vv:" // count -t,-v
885                 "X::T::" // cumulative lists
886 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
887                 "\xff::" // cumulative lists for --exclude
888 #endif
889                 IF_FEATURE_TAR_CREATE("c:") "t:x:" // at least one of these is reqd
890                 IF_FEATURE_TAR_CREATE("c--tx:t--cx:x--ct") // mutually exclusive
891                 IF_NOT_FEATURE_TAR_CREATE("t--x:x--t"); // mutually exclusive
892 #if ENABLE_FEATURE_TAR_LONG_OPTIONS
893         applet_long_options = tar_longopts;
894 #endif
895 #if ENABLE_DESKTOP
896         if (argv[1] && argv[1][0] != '-') {
897                 /* Compat:
898                  * 1st argument without dash handles options with parameters
899                  * differently from dashed one: it takes *next argv[i]*
900                  * as paramenter even if there are more chars in 1st argument:
901                  *  "tar fx TARFILE" - "x" is not taken as f's param
902                  *  but is interpreted as -x option
903                  *  "tar -xf TARFILE" - dashed equivalent of the above
904                  *  "tar -fx ..." - "x" is taken as f's param
905                  * getopt32 wouldn't handle 1st command correctly.
906                  * Unfortunately, people do use such commands.
907                  * We massage argv[1] to work around it by moving 'f'
908                  * to the end of the string.
909                  * More contrived "tar fCx TARFILE DIR" still fails,
910                  * but such commands are much less likely to be used.
911                  */
912                 char *f = strchr(argv[1], 'f');
913                 if (f) {
914                         while (f[1] != '\0') {
915                                 *f = f[1];
916                                 f++;
917                         }
918                         *f = 'f';
919                 }
920         }
921 #endif
922         opt = getopt32(argv,
923                 "txC:f:Oopvk"
924                 IF_FEATURE_TAR_CREATE(   "ch"  )
925                 IF_FEATURE_SEAMLESS_BZ2( "j"   )
926                 IF_FEATURE_SEAMLESS_LZMA("a"   )
927                 IF_FEATURE_TAR_FROM(     "T:X:")
928                 IF_FEATURE_SEAMLESS_GZ(  "z"   )
929                 IF_FEATURE_SEAMLESS_Z(   "Z"   )
930                 IF_FEATURE_TAR_NOPRESERVE_TIME("m")
931                 , &base_dir // -C dir
932                 , &tar_filename // -f filename
933                 IF_FEATURE_TAR_FROM(, &(tar_handle->accept)) // T
934                 IF_FEATURE_TAR_FROM(, &(tar_handle->reject)) // X
935                 IF_FEATURE_TAR_TO_COMMAND(, &(tar_handle->tar__to_command)) // --to-command
936 #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM
937                 , &excludes // --exclude
938 #endif
939                 , &verboseFlag // combined count for -t and -v
940                 , &verboseFlag // combined count for -t and -v
941                 );
942         //bb_error_msg("opt:%08x", opt);
943         argv += optind;
944
945         if (verboseFlag) tar_handle->action_header = header_verbose_list;
946         if (verboseFlag == 1) tar_handle->action_header = header_list;
947
948         if (opt & OPT_EXTRACT)
949                 tar_handle->action_data = data_extract_all;
950
951         if (opt & OPT_2STDOUT)
952                 tar_handle->action_data = data_extract_to_stdout;
953
954         if (opt & OPT_2COMMAND) {
955                 putenv((char*)"TAR_FILETYPE=f");
956                 signal(SIGPIPE, SIG_IGN);
957                 tar_handle->action_data = data_extract_to_command;
958         }
959
960         if (opt & OPT_KEEP_OLD)
961                 tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
962
963         if (opt & OPT_NUMERIC_OWNER)
964                 tar_handle->ah_flags |= ARCHIVE_NUMERIC_OWNER;
965
966         if (opt & OPT_NOPRESERVE_OWNER)
967                 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_OWNER;
968
969         if (opt & OPT_NOPRESERVE_PERM)
970                 tar_handle->ah_flags |= ARCHIVE_DONT_RESTORE_PERM;
971
972         if (opt & OPT_OVERWRITE) {
973                 tar_handle->ah_flags &= ~ARCHIVE_UNLINK_OLD;
974                 tar_handle->ah_flags |= ARCHIVE_O_TRUNC;
975         }
976
977         if (opt & OPT_GZIP)
978                 get_header_ptr = get_header_tar_gz;
979
980         if (opt & OPT_BZIP2)
981                 get_header_ptr = get_header_tar_bz2;
982
983         if (opt & OPT_LZMA)
984                 get_header_ptr = get_header_tar_lzma;
985
986         if (opt & OPT_COMPRESS)
987                 get_header_ptr = get_header_tar_Z;
988
989         if (opt & OPT_NOPRESERVE_TIME)
990                 tar_handle->ah_flags &= ~ARCHIVE_RESTORE_DATE;
991
992 #if ENABLE_FEATURE_TAR_FROM
993         tar_handle->reject = append_file_list_to_list(tar_handle->reject);
994 # if ENABLE_FEATURE_TAR_LONG_OPTIONS
995         /* Append excludes to reject */
996         while (excludes) {
997                 llist_t *next = excludes->link;
998                 excludes->link = tar_handle->reject;
999                 tar_handle->reject = excludes;
1000                 excludes = next;
1001         }
1002 # endif
1003         tar_handle->accept = append_file_list_to_list(tar_handle->accept);
1004 #endif
1005
1006         /* Setup an array of filenames to work with */
1007         /* TODO: This is the same as in ar, make a separate function? */
1008         while (*argv) {
1009                 /* kill trailing '/' unless the string is just "/" */
1010                 char *cp = last_char_is(*argv, '/');
1011                 if (cp > *argv)
1012                         *cp = '\0';
1013                 llist_add_to_end(&tar_handle->accept, *argv);
1014                 argv++;
1015         }
1016
1017         if (tar_handle->accept || tar_handle->reject)
1018                 tar_handle->filter = filter_accept_reject_list;
1019
1020         /* Open the tar file */
1021         {
1022                 int tar_fd = STDIN_FILENO;
1023                 int flags = O_RDONLY;
1024
1025                 if (opt & OPT_CREATE) {
1026                         /* Make sure there is at least one file to tar up */
1027                         if (tar_handle->accept == NULL)
1028                                 bb_error_msg_and_die("empty archive");
1029
1030                         tar_fd = STDOUT_FILENO;
1031                         /* Mimicking GNU tar 1.15.1: */
1032                         flags = O_WRONLY | O_CREAT | O_TRUNC;
1033                 }
1034
1035                 if (LONE_DASH(tar_filename)) {
1036                         tar_handle->src_fd = tar_fd;
1037                         tar_handle->seek = seek_by_read;
1038                 } else {
1039                         if (ENABLE_FEATURE_TAR_AUTODETECT && flags == O_RDONLY) {
1040                                 get_header_ptr = get_header_tar;
1041                                 tar_handle->src_fd = open_zipped(tar_filename);
1042                                 if (tar_handle->src_fd < 0)
1043                                         bb_perror_msg_and_die("can't open '%s'", tar_filename);
1044                         } else {
1045                                 tar_handle->src_fd = xopen(tar_filename, flags);
1046                         }
1047                 }
1048         }
1049
1050         if (base_dir)
1051                 xchdir(base_dir);
1052
1053 #ifdef CHECK_FOR_CHILD_EXITCODE
1054         /* We need to know whether child (gzip/bzip/etc) exits abnormally */
1055         signal(SIGCHLD, handle_SIGCHLD);
1056 #endif
1057
1058         /* Create an archive */
1059         if (opt & OPT_CREATE) {
1060 #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2
1061                 int zipMode = 0;
1062                 if (ENABLE_FEATURE_SEAMLESS_GZ && (opt & OPT_GZIP))
1063                         zipMode = 1;
1064                 if (ENABLE_FEATURE_SEAMLESS_BZ2 && (opt & OPT_BZIP2))
1065                         zipMode = 2;
1066 #endif
1067                 /* NB: writeTarFile() closes tar_handle->src_fd */
1068                 return writeTarFile(tar_handle->src_fd, verboseFlag, opt & OPT_DEREFERENCE,
1069                                 tar_handle->accept,
1070                                 tar_handle->reject, zipMode);
1071         }
1072
1073         while (get_header_ptr(tar_handle) == EXIT_SUCCESS)
1074                 continue;
1075
1076         /* Check that every file that should have been extracted was */
1077         while (tar_handle->accept) {
1078                 if (!find_list_entry(tar_handle->reject, tar_handle->accept->data)
1079                  && !find_list_entry(tar_handle->passed, tar_handle->accept->data)
1080                 ) {
1081                         bb_error_msg_and_die("%s: not found in archive",
1082                                 tar_handle->accept->data);
1083                 }
1084                 tar_handle->accept = tar_handle->accept->link;
1085         }
1086         if (ENABLE_FEATURE_CLEAN_UP /* && tar_handle->src_fd != STDIN_FILENO */)
1087                 close(tar_handle->src_fd);
1088
1089         return EXIT_SUCCESS;
1090 }