Contents of /tags/mkinitrd-6_1_7/busybox/archival/tar.c
Parent Directory | Revision Log
Revision 909 -
(show annotations)
(download)
Fri Oct 16 13:40:14 2009 UTC (14 years, 11 months ago) by niro
File MIME type: text/plain
File size: 30879 byte(s)
Fri Oct 16 13:40:14 2009 UTC (14 years, 11 months ago) by niro
File MIME type: text/plain
File size: 30879 byte(s)
tagged 'mkinitrd-6_1_7'
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 the 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 tarball for details. |
24 | */ |
25 | |
26 | #include <fnmatch.h> |
27 | #include "libbb.h" |
28 | #include "unarchive.h" |
29 | |
30 | /* FIXME: Stop using this non-standard feature */ |
31 | #ifndef FNM_LEADING_DIR |
32 | #define FNM_LEADING_DIR 0 |
33 | #endif |
34 | |
35 | |
36 | #define block_buf bb_common_bufsiz1 |
37 | |
38 | |
39 | #if !ENABLE_FEATURE_SEAMLESS_GZ && !ENABLE_FEATURE_SEAMLESS_BZ2 |
40 | /* Do not pass gzip flag to writeTarFile() */ |
41 | #define writeTarFile(tar_fd, verboseFlag, dereferenceFlag, include, exclude, gzip) \ |
42 | writeTarFile(tar_fd, verboseFlag, dereferenceFlag, include, exclude) |
43 | #endif |
44 | |
45 | |
46 | #if ENABLE_FEATURE_TAR_CREATE |
47 | |
48 | /* Tar file constants */ |
49 | |
50 | #define TAR_BLOCK_SIZE 512 |
51 | |
52 | /* POSIX tar Header Block, from POSIX 1003.1-1990 */ |
53 | #define NAME_SIZE 100 |
54 | #define NAME_SIZE_STR "100" |
55 | typedef struct TarHeader TarHeader; |
56 | struct TarHeader { /* byte offset */ |
57 | char name[NAME_SIZE]; /* 0-99 */ |
58 | char mode[8]; /* 100-107 */ |
59 | char uid[8]; /* 108-115 */ |
60 | char gid[8]; /* 116-123 */ |
61 | char size[12]; /* 124-135 */ |
62 | char mtime[12]; /* 136-147 */ |
63 | char chksum[8]; /* 148-155 */ |
64 | char typeflag; /* 156-156 */ |
65 | char linkname[NAME_SIZE]; /* 157-256 */ |
66 | /* POSIX: "ustar" NUL "00" */ |
67 | /* GNU tar: "ustar " NUL */ |
68 | /* Normally it's defined as magic[6] followed by |
69 | * version[2], but we put them together to save code. |
70 | */ |
71 | char magic[8]; /* 257-264 */ |
72 | char uname[32]; /* 265-296 */ |
73 | char gname[32]; /* 297-328 */ |
74 | char devmajor[8]; /* 329-336 */ |
75 | char devminor[8]; /* 337-344 */ |
76 | char prefix[155]; /* 345-499 */ |
77 | char padding[12]; /* 500-512 (pad to exactly TAR_BLOCK_SIZE) */ |
78 | }; |
79 | |
80 | /* |
81 | ** writeTarFile(), writeFileToTarball(), and writeTarHeader() are |
82 | ** the only functions that deal with the HardLinkInfo structure. |
83 | ** Even these functions use the xxxHardLinkInfo() functions. |
84 | */ |
85 | typedef struct HardLinkInfo HardLinkInfo; |
86 | struct HardLinkInfo { |
87 | HardLinkInfo *next; /* Next entry in list */ |
88 | dev_t dev; /* Device number */ |
89 | ino_t ino; /* Inode number */ |
90 | short linkCount; /* (Hard) Link Count */ |
91 | char name[1]; /* Start of filename (must be last) */ |
92 | }; |
93 | |
94 | /* Some info to be carried along when creating a new tarball */ |
95 | typedef struct TarBallInfo TarBallInfo; |
96 | struct TarBallInfo { |
97 | int tarFd; /* Open-for-write file descriptor |
98 | * for the tarball */ |
99 | struct stat statBuf; /* Stat info for the tarball, letting |
100 | * us know the inode and device that the |
101 | * tarball lives, so we can avoid trying |
102 | * to include the tarball into itself */ |
103 | int verboseFlag; /* Whether to print extra stuff or not */ |
104 | const llist_t *excludeList; /* List of files to not include */ |
105 | HardLinkInfo *hlInfoHead; /* Hard Link Tracking Information */ |
106 | HardLinkInfo *hlInfo; /* Hard Link Info for the current file */ |
107 | }; |
108 | |
109 | /* A nice enum with all the possible tar file content types */ |
110 | enum TarFileType { |
111 | REGTYPE = '0', /* regular file */ |
112 | REGTYPE0 = '\0', /* regular file (ancient bug compat) */ |
113 | LNKTYPE = '1', /* hard link */ |
114 | SYMTYPE = '2', /* symbolic link */ |
115 | CHRTYPE = '3', /* character special */ |
116 | BLKTYPE = '4', /* block special */ |
117 | DIRTYPE = '5', /* directory */ |
118 | FIFOTYPE = '6', /* FIFO special */ |
119 | CONTTYPE = '7', /* reserved */ |
120 | GNULONGLINK = 'K', /* GNU long (>100 chars) link name */ |
121 | GNULONGNAME = 'L', /* GNU long (>100 chars) file name */ |
122 | }; |
123 | typedef enum TarFileType TarFileType; |
124 | |
125 | /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */ |
126 | static void addHardLinkInfo(HardLinkInfo **hlInfoHeadPtr, |
127 | struct stat *statbuf, |
128 | const char *fileName) |
129 | { |
130 | /* Note: hlInfoHeadPtr can never be NULL! */ |
131 | HardLinkInfo *hlInfo; |
132 | |
133 | hlInfo = xmalloc(sizeof(HardLinkInfo) + strlen(fileName)); |
134 | hlInfo->next = *hlInfoHeadPtr; |
135 | *hlInfoHeadPtr = hlInfo; |
136 | hlInfo->dev = statbuf->st_dev; |
137 | hlInfo->ino = statbuf->st_ino; |
138 | hlInfo->linkCount = statbuf->st_nlink; |
139 | strcpy(hlInfo->name, fileName); |
140 | } |
141 | |
142 | static void freeHardLinkInfo(HardLinkInfo **hlInfoHeadPtr) |
143 | { |
144 | HardLinkInfo *hlInfo; |
145 | HardLinkInfo *hlInfoNext; |
146 | |
147 | if (hlInfoHeadPtr) { |
148 | hlInfo = *hlInfoHeadPtr; |
149 | while (hlInfo) { |
150 | hlInfoNext = hlInfo->next; |
151 | free(hlInfo); |
152 | hlInfo = hlInfoNext; |
153 | } |
154 | *hlInfoHeadPtr = NULL; |
155 | } |
156 | } |
157 | |
158 | /* Might be faster (and bigger) if the dev/ino were stored in numeric order;) */ |
159 | static HardLinkInfo *findHardLinkInfo(HardLinkInfo *hlInfo, struct stat *statbuf) |
160 | { |
161 | while (hlInfo) { |
162 | if ((statbuf->st_ino == hlInfo->ino) && (statbuf->st_dev == hlInfo->dev)) |
163 | break; |
164 | hlInfo = hlInfo->next; |
165 | } |
166 | return hlInfo; |
167 | } |
168 | |
169 | /* Put an octal string into the specified buffer. |
170 | * The number is zero padded and possibly null terminated. |
171 | * Stores low-order bits only if whole value does not fit. */ |
172 | static void putOctal(char *cp, int len, off_t value) |
173 | { |
174 | char tempBuffer[sizeof(off_t)*3+1]; |
175 | char *tempString = tempBuffer; |
176 | int width; |
177 | |
178 | width = sprintf(tempBuffer, "%0*"OFF_FMT"o", len, value); |
179 | tempString += (width - len); |
180 | |
181 | /* If string has leading zeroes, we can drop one */ |
182 | /* and field will have trailing '\0' */ |
183 | /* (increases chances of compat with other tars) */ |
184 | if (tempString[0] == '0') |
185 | tempString++; |
186 | |
187 | /* Copy the string to the field */ |
188 | memcpy(cp, tempString, len); |
189 | } |
190 | #define PUT_OCTAL(a, b) putOctal((a), sizeof(a), (b)) |
191 | |
192 | static void chksum_and_xwrite(int fd, struct TarHeader* hp) |
193 | { |
194 | /* POSIX says that checksum is done on unsigned bytes |
195 | * (Sun and HP-UX gets it wrong... more details in |
196 | * GNU tar source) */ |
197 | const unsigned char *cp; |
198 | int chksum, size; |
199 | |
200 | strcpy(hp->magic, "ustar "); |
201 | |
202 | /* Calculate and store the checksum (i.e., the sum of all of the bytes of |
203 | * the header). The checksum field must be filled with blanks for the |
204 | * calculation. The checksum field is formatted differently from the |
205 | * other fields: it has 6 digits, a null, then a space -- rather than |
206 | * digits, followed by a null like the other fields... */ |
207 | memset(hp->chksum, ' ', sizeof(hp->chksum)); |
208 | cp = (const unsigned char *) hp; |
209 | chksum = 0; |
210 | size = sizeof(*hp); |
211 | do { chksum += *cp++; } while (--size); |
212 | putOctal(hp->chksum, sizeof(hp->chksum)-1, chksum); |
213 | |
214 | /* Now write the header out to disk */ |
215 | xwrite(fd, hp, sizeof(*hp)); |
216 | } |
217 | |
218 | #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS |
219 | static void writeLongname(int fd, int type, const char *name, int dir) |
220 | { |
221 | static const struct { |
222 | char mode[8]; /* 100-107 */ |
223 | char uid[8]; /* 108-115 */ |
224 | char gid[8]; /* 116-123 */ |
225 | char size[12]; /* 124-135 */ |
226 | char mtime[12]; /* 136-147 */ |
227 | } prefilled = { |
228 | "0000000", |
229 | "0000000", |
230 | "0000000", |
231 | "00000000000", |
232 | "00000000000", |
233 | }; |
234 | struct TarHeader header; |
235 | int size; |
236 | |
237 | dir = !!dir; /* normalize: 0/1 */ |
238 | size = strlen(name) + 1 + dir; /* GNU tar uses strlen+1 */ |
239 | /* + dir: account for possible '/' */ |
240 | |
241 | memset(&header, 0, sizeof(header)); |
242 | strcpy(header.name, "././@LongLink"); |
243 | memcpy(header.mode, prefilled.mode, sizeof(prefilled)); |
244 | PUT_OCTAL(header.size, size); |
245 | header.typeflag = type; |
246 | chksum_and_xwrite(fd, &header); |
247 | |
248 | /* Write filename[/] and pad the block. */ |
249 | /* dir=0: writes 'name<NUL>', pads */ |
250 | /* dir=1: writes 'name', writes '/<NUL>', pads */ |
251 | dir *= 2; |
252 | xwrite(fd, name, size - dir); |
253 | xwrite(fd, "/", dir); |
254 | size = (-size) & (TAR_BLOCK_SIZE-1); |
255 | memset(&header, 0, size); |
256 | xwrite(fd, &header, size); |
257 | } |
258 | #endif |
259 | |
260 | /* Write out a tar header for the specified file/directory/whatever */ |
261 | void BUG_tar_header_size(void); |
262 | static int writeTarHeader(struct TarBallInfo *tbInfo, |
263 | const char *header_name, const char *fileName, struct stat *statbuf) |
264 | { |
265 | struct TarHeader header; |
266 | |
267 | if (sizeof(header) != 512) |
268 | BUG_tar_header_size(); |
269 | |
270 | memset(&header, 0, sizeof(struct TarHeader)); |
271 | |
272 | strncpy(header.name, header_name, sizeof(header.name)); |
273 | |
274 | /* POSIX says to mask mode with 07777. */ |
275 | PUT_OCTAL(header.mode, statbuf->st_mode & 07777); |
276 | PUT_OCTAL(header.uid, statbuf->st_uid); |
277 | PUT_OCTAL(header.gid, statbuf->st_gid); |
278 | memset(header.size, '0', sizeof(header.size)-1); /* Regular file size is handled later */ |
279 | PUT_OCTAL(header.mtime, statbuf->st_mtime); |
280 | |
281 | /* Enter the user and group names */ |
282 | safe_strncpy(header.uname, get_cached_username(statbuf->st_uid), sizeof(header.uname)); |
283 | safe_strncpy(header.gname, get_cached_groupname(statbuf->st_gid), sizeof(header.gname)); |
284 | |
285 | if (tbInfo->hlInfo) { |
286 | /* This is a hard link */ |
287 | header.typeflag = LNKTYPE; |
288 | strncpy(header.linkname, tbInfo->hlInfo->name, |
289 | sizeof(header.linkname)); |
290 | #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS |
291 | /* Write out long linkname if needed */ |
292 | if (header.linkname[sizeof(header.linkname)-1]) |
293 | writeLongname(tbInfo->tarFd, GNULONGLINK, |
294 | tbInfo->hlInfo->name, 0); |
295 | #endif |
296 | } else if (S_ISLNK(statbuf->st_mode)) { |
297 | char *lpath = xmalloc_readlink_or_warn(fileName); |
298 | if (!lpath) |
299 | return FALSE; |
300 | header.typeflag = SYMTYPE; |
301 | strncpy(header.linkname, lpath, sizeof(header.linkname)); |
302 | #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS |
303 | /* Write out long linkname if needed */ |
304 | if (header.linkname[sizeof(header.linkname)-1]) |
305 | writeLongname(tbInfo->tarFd, GNULONGLINK, lpath, 0); |
306 | #else |
307 | /* If it is larger than 100 bytes, bail out */ |
308 | if (header.linkname[sizeof(header.linkname)-1]) { |
309 | free(lpath); |
310 | bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported"); |
311 | return FALSE; |
312 | } |
313 | #endif |
314 | free(lpath); |
315 | } else if (S_ISDIR(statbuf->st_mode)) { |
316 | header.typeflag = DIRTYPE; |
317 | /* Append '/' only if there is a space for it */ |
318 | if (!header.name[sizeof(header.name)-1]) |
319 | header.name[strlen(header.name)] = '/'; |
320 | } else if (S_ISCHR(statbuf->st_mode)) { |
321 | header.typeflag = CHRTYPE; |
322 | PUT_OCTAL(header.devmajor, major(statbuf->st_rdev)); |
323 | PUT_OCTAL(header.devminor, minor(statbuf->st_rdev)); |
324 | } else if (S_ISBLK(statbuf->st_mode)) { |
325 | header.typeflag = BLKTYPE; |
326 | PUT_OCTAL(header.devmajor, major(statbuf->st_rdev)); |
327 | PUT_OCTAL(header.devminor, minor(statbuf->st_rdev)); |
328 | } else if (S_ISFIFO(statbuf->st_mode)) { |
329 | header.typeflag = FIFOTYPE; |
330 | } else if (S_ISREG(statbuf->st_mode)) { |
331 | if (sizeof(statbuf->st_size) > 4 |
332 | && statbuf->st_size > (off_t)0777777777777LL |
333 | ) { |
334 | bb_error_msg_and_die("cannot store file '%s' " |
335 | "of size %"OFF_FMT"d, aborting", |
336 | fileName, statbuf->st_size); |
337 | } |
338 | header.typeflag = REGTYPE; |
339 | PUT_OCTAL(header.size, statbuf->st_size); |
340 | } else { |
341 | bb_error_msg("%s: unknown file type", fileName); |
342 | return FALSE; |
343 | } |
344 | |
345 | #if ENABLE_FEATURE_TAR_GNU_EXTENSIONS |
346 | /* Write out long name if needed */ |
347 | /* (we, like GNU tar, output long linkname *before* long name) */ |
348 | if (header.name[sizeof(header.name)-1]) |
349 | writeLongname(tbInfo->tarFd, GNULONGNAME, |
350 | header_name, S_ISDIR(statbuf->st_mode)); |
351 | #endif |
352 | |
353 | /* Now write the header out to disk */ |
354 | chksum_and_xwrite(tbInfo->tarFd, &header); |
355 | |
356 | /* Now do the verbose thing (or not) */ |
357 | if (tbInfo->verboseFlag) { |
358 | FILE *vbFd = stdout; |
359 | |
360 | if (tbInfo->tarFd == STDOUT_FILENO) /* If the archive goes to stdout, verbose to stderr */ |
361 | vbFd = stderr; |
362 | /* GNU "tar cvvf" prints "extended" listing a-la "ls -l" */ |
363 | /* We don't have such excesses here: for us "v" == "vv" */ |
364 | /* '/' is probably a GNUism */ |
365 | fprintf(vbFd, "%s%s\n", header_name, |
366 | S_ISDIR(statbuf->st_mode) ? "/" : ""); |
367 | } |
368 | |
369 | return TRUE; |
370 | } |
371 | |
372 | #if ENABLE_FEATURE_TAR_FROM |
373 | static int exclude_file(const llist_t *excluded_files, const char *file) |
374 | { |
375 | while (excluded_files) { |
376 | if (excluded_files->data[0] == '/') { |
377 | if (fnmatch(excluded_files->data, file, |
378 | FNM_PATHNAME | FNM_LEADING_DIR) == 0) |
379 | return 1; |
380 | } else { |
381 | const char *p; |
382 | |
383 | for (p = file; p[0] != '\0'; p++) { |
384 | if ((p == file || p[-1] == '/') && p[0] != '/' && |
385 | fnmatch(excluded_files->data, p, |
386 | FNM_PATHNAME | FNM_LEADING_DIR) == 0) |
387 | return 1; |
388 | } |
389 | } |
390 | excluded_files = excluded_files->link; |
391 | } |
392 | |
393 | return 0; |
394 | } |
395 | #else |
396 | #define exclude_file(excluded_files, file) 0 |
397 | #endif |
398 | |
399 | static int FAST_FUNC writeFileToTarball(const char *fileName, struct stat *statbuf, |
400 | void *userData, int depth UNUSED_PARAM) |
401 | { |
402 | struct TarBallInfo *tbInfo = (struct TarBallInfo *) userData; |
403 | const char *header_name; |
404 | int inputFileFd = -1; |
405 | |
406 | /* Strip leading '/' (must be before memorizing hardlink's name) */ |
407 | header_name = fileName; |
408 | while (header_name[0] == '/') { |
409 | static smallint warned; |
410 | |
411 | if (!warned) { |
412 | bb_error_msg("removing leading '/' from member names"); |
413 | warned = 1; |
414 | } |
415 | header_name++; |
416 | } |
417 | |
418 | if (header_name[0] == '\0') |
419 | return TRUE; |
420 | |
421 | /* It is against the rules to archive a socket */ |
422 | if (S_ISSOCK(statbuf->st_mode)) { |
423 | bb_error_msg("%s: socket ignored", fileName); |
424 | return TRUE; |
425 | } |
426 | |
427 | /* |
428 | * Check to see if we are dealing with a hard link. |
429 | * If so - |
430 | * Treat the first occurance of a given dev/inode as a file while |
431 | * treating any additional occurances as hard links. This is done |
432 | * by adding the file information to the HardLinkInfo linked list. |
433 | */ |
434 | tbInfo->hlInfo = NULL; |
435 | if (statbuf->st_nlink > 1) { |
436 | tbInfo->hlInfo = findHardLinkInfo(tbInfo->hlInfoHead, statbuf); |
437 | if (tbInfo->hlInfo == NULL) |
438 | addHardLinkInfo(&tbInfo->hlInfoHead, statbuf, header_name); |
439 | } |
440 | |
441 | /* It is a bad idea to store the archive we are in the process of creating, |
442 | * so check the device and inode to be sure that this particular file isn't |
443 | * the new tarball */ |
444 | if (tbInfo->statBuf.st_dev == statbuf->st_dev |
445 | && tbInfo->statBuf.st_ino == statbuf->st_ino |
446 | ) { |
447 | bb_error_msg("%s: file is the archive; skipping", fileName); |
448 | return TRUE; |
449 | } |
450 | |
451 | if (exclude_file(tbInfo->excludeList, header_name)) |
452 | return SKIP; |
453 | |
454 | #if !ENABLE_FEATURE_TAR_GNU_EXTENSIONS |
455 | if (strlen(header_name) >= NAME_SIZE) { |
456 | bb_error_msg("names longer than "NAME_SIZE_STR" chars not supported"); |
457 | return TRUE; |
458 | } |
459 | #endif |
460 | |
461 | /* Is this a regular file? */ |
462 | if (tbInfo->hlInfo == NULL && S_ISREG(statbuf->st_mode)) { |
463 | /* open the file we want to archive, and make sure all is well */ |
464 | inputFileFd = open_or_warn(fileName, O_RDONLY); |
465 | if (inputFileFd < 0) { |
466 | return FALSE; |
467 | } |
468 | } |
469 | |
470 | /* Add an entry to the tarball */ |
471 | if (writeTarHeader(tbInfo, header_name, fileName, statbuf) == FALSE) { |
472 | return FALSE; |
473 | } |
474 | |
475 | /* If it was a regular file, write out the body */ |
476 | if (inputFileFd >= 0) { |
477 | size_t readSize; |
478 | /* Write the file to the archive. */ |
479 | /* We record size into header first, */ |
480 | /* and then write out file. If file shrinks in between, */ |
481 | /* tar will be corrupted. So we don't allow for that. */ |
482 | /* NB: GNU tar 1.16 warns and pads with zeroes */ |
483 | /* or even seeks back and updates header */ |
484 | bb_copyfd_exact_size(inputFileFd, tbInfo->tarFd, statbuf->st_size); |
485 | ////off_t readSize; |
486 | ////readSize = bb_copyfd_size(inputFileFd, tbInfo->tarFd, statbuf->st_size); |
487 | ////if (readSize != statbuf->st_size && readSize >= 0) { |
488 | //// bb_error_msg_and_die("short read from %s, aborting", fileName); |
489 | ////} |
490 | |
491 | /* Check that file did not grow in between? */ |
492 | /* if (safe_read(inputFileFd, 1) == 1) warn but continue? */ |
493 | |
494 | close(inputFileFd); |
495 | |
496 | /* Pad the file up to the tar block size */ |
497 | /* (a few tricks here in the name of code size) */ |
498 | readSize = (-(int)statbuf->st_size) & (TAR_BLOCK_SIZE-1); |
499 | memset(block_buf, 0, readSize); |
500 | xwrite(tbInfo->tarFd, block_buf, readSize); |
501 | } |
502 | |
503 | return TRUE; |
504 | } |
505 | |
506 | #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2 |
507 | #if !(ENABLE_FEATURE_SEAMLESS_GZ && ENABLE_FEATURE_SEAMLESS_BZ2) |
508 | #define vfork_compressor(tar_fd, gzip) vfork_compressor(tar_fd) |
509 | #endif |
510 | /* Don't inline: vfork scares gcc and pessimizes code */ |
511 | static void NOINLINE vfork_compressor(int tar_fd, int gzip) |
512 | { |
513 | pid_t gzipPid; |
514 | #if ENABLE_FEATURE_SEAMLESS_GZ && ENABLE_FEATURE_SEAMLESS_BZ2 |
515 | const char *zip_exec = (gzip == 1) ? "gzip" : "bzip2"; |
516 | #elif ENABLE_FEATURE_SEAMLESS_GZ |
517 | const char *zip_exec = "gzip"; |
518 | #else /* only ENABLE_FEATURE_SEAMLESS_BZ2 */ |
519 | const char *zip_exec = "bzip2"; |
520 | #endif |
521 | // On Linux, vfork never unpauses parent early, although standard |
522 | // allows for that. Do we want to waste bytes checking for it? |
523 | #define WAIT_FOR_CHILD 0 |
524 | volatile int vfork_exec_errno = 0; |
525 | struct fd_pair gzipDataPipe; |
526 | #if WAIT_FOR_CHILD |
527 | struct fd_pair gzipStatusPipe; |
528 | xpiped_pair(gzipStatusPipe); |
529 | #endif |
530 | xpiped_pair(gzipDataPipe); |
531 | |
532 | signal(SIGPIPE, SIG_IGN); /* we only want EPIPE on errors */ |
533 | |
534 | #if defined(__GNUC__) && __GNUC__ |
535 | /* Avoid vfork clobbering */ |
536 | (void) &zip_exec; |
537 | #endif |
538 | |
539 | gzipPid = vfork(); |
540 | if (gzipPid < 0) |
541 | bb_perror_msg_and_die("vfork"); |
542 | |
543 | if (gzipPid == 0) { |
544 | /* child */ |
545 | /* NB: close _first_, then move fds! */ |
546 | close(gzipDataPipe.wr); |
547 | #if WAIT_FOR_CHILD |
548 | close(gzipStatusPipe.rd); |
549 | /* gzipStatusPipe.wr will close only on exec - |
550 | * parent waits for this close to happen */ |
551 | fcntl(gzipStatusPipe.wr, F_SETFD, FD_CLOEXEC); |
552 | #endif |
553 | xmove_fd(gzipDataPipe.rd, 0); |
554 | xmove_fd(tar_fd, 1); |
555 | /* exec gzip/bzip2 program/applet */ |
556 | BB_EXECLP(zip_exec, zip_exec, "-f", NULL); |
557 | vfork_exec_errno = errno; |
558 | _exit(EXIT_FAILURE); |
559 | } |
560 | |
561 | /* parent */ |
562 | xmove_fd(gzipDataPipe.wr, tar_fd); |
563 | close(gzipDataPipe.rd); |
564 | #if WAIT_FOR_CHILD |
565 | close(gzipStatusPipe.wr); |
566 | while (1) { |
567 | char buf; |
568 | int n; |
569 | |
570 | /* Wait until child execs (or fails to) */ |
571 | n = full_read(gzipStatusPipe.rd, &buf, 1); |
572 | if (n < 0 /* && errno == EAGAIN */) |
573 | continue; /* try it again */ |
574 | } |
575 | close(gzipStatusPipe.rd); |
576 | #endif |
577 | if (vfork_exec_errno) { |
578 | errno = vfork_exec_errno; |
579 | bb_perror_msg_and_die("cannot exec %s", zip_exec); |
580 | } |
581 | } |
582 | #endif /* ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2 */ |
583 | |
584 | |
585 | /* gcc 4.2.1 inlines it, making code bigger */ |
586 | static NOINLINE int writeTarFile(int tar_fd, int verboseFlag, |
587 | int dereferenceFlag, const llist_t *include, |
588 | const llist_t *exclude, int gzip) |
589 | { |
590 | int errorFlag = FALSE; |
591 | struct TarBallInfo tbInfo; |
592 | |
593 | tbInfo.hlInfoHead = NULL; |
594 | tbInfo.tarFd = tar_fd; |
595 | tbInfo.verboseFlag = verboseFlag; |
596 | |
597 | /* Store the stat info for the tarball's file, so |
598 | * can avoid including the tarball into itself.... */ |
599 | if (fstat(tbInfo.tarFd, &tbInfo.statBuf) < 0) |
600 | bb_perror_msg_and_die("cannot stat tar file"); |
601 | |
602 | #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2 |
603 | if (gzip) |
604 | vfork_compressor(tbInfo.tarFd, gzip); |
605 | #endif |
606 | |
607 | tbInfo.excludeList = exclude; |
608 | |
609 | /* Read the directory/files and iterate over them one at a time */ |
610 | while (include) { |
611 | if (!recursive_action(include->data, ACTION_RECURSE | |
612 | (dereferenceFlag ? ACTION_FOLLOWLINKS : 0), |
613 | writeFileToTarball, writeFileToTarball, &tbInfo, 0)) |
614 | { |
615 | errorFlag = TRUE; |
616 | } |
617 | include = include->link; |
618 | } |
619 | /* Write two empty blocks to the end of the archive */ |
620 | memset(block_buf, 0, 2*TAR_BLOCK_SIZE); |
621 | xwrite(tbInfo.tarFd, block_buf, 2*TAR_BLOCK_SIZE); |
622 | |
623 | /* To be pedantically correct, we would check if the tarball |
624 | * is smaller than 20 tar blocks, and pad it if it was smaller, |
625 | * but that isn't necessary for GNU tar interoperability, and |
626 | * so is considered a waste of space */ |
627 | |
628 | /* Close so the child process (if any) will exit */ |
629 | close(tbInfo.tarFd); |
630 | |
631 | /* Hang up the tools, close up shop, head home */ |
632 | if (ENABLE_FEATURE_CLEAN_UP) |
633 | freeHardLinkInfo(&tbInfo.hlInfoHead); |
634 | |
635 | if (errorFlag) |
636 | bb_error_msg("error exit delayed from previous errors"); |
637 | |
638 | #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2 |
639 | if (gzip) { |
640 | int status; |
641 | if (safe_waitpid(-1, &status, 0) == -1) |
642 | bb_perror_msg("waitpid"); |
643 | else if (!WIFEXITED(status) || WEXITSTATUS(status)) |
644 | /* gzip was killed or has exited with nonzero! */ |
645 | errorFlag = TRUE; |
646 | } |
647 | #endif |
648 | return errorFlag; |
649 | } |
650 | #else |
651 | int writeTarFile(int tar_fd, int verboseFlag, |
652 | int dereferenceFlag, const llist_t *include, |
653 | const llist_t *exclude, int gzip); |
654 | #endif /* FEATURE_TAR_CREATE */ |
655 | |
656 | #if ENABLE_FEATURE_TAR_FROM |
657 | static llist_t *append_file_list_to_list(llist_t *list) |
658 | { |
659 | FILE *src_stream; |
660 | char *line; |
661 | llist_t *newlist = NULL; |
662 | |
663 | while (list) { |
664 | src_stream = xfopen_for_read(llist_pop(&list)); |
665 | while ((line = xmalloc_fgetline(src_stream)) != NULL) { |
666 | /* kill trailing '/' unless the string is just "/" */ |
667 | char *cp = last_char_is(line, '/'); |
668 | if (cp > line) |
669 | *cp = '\0'; |
670 | llist_add_to(&newlist, line); |
671 | } |
672 | fclose(src_stream); |
673 | } |
674 | return newlist; |
675 | } |
676 | #else |
677 | #define append_file_list_to_list(x) 0 |
678 | #endif |
679 | |
680 | #if ENABLE_FEATURE_SEAMLESS_Z |
681 | static char FAST_FUNC get_header_tar_Z(archive_handle_t *archive_handle) |
682 | { |
683 | /* Can't lseek over pipes */ |
684 | archive_handle->seek = seek_by_read; |
685 | |
686 | /* do the decompression, and cleanup */ |
687 | if (xread_char(archive_handle->src_fd) != 0x1f |
688 | || xread_char(archive_handle->src_fd) != 0x9d |
689 | ) { |
690 | bb_error_msg_and_die("invalid magic"); |
691 | } |
692 | |
693 | open_transformer(archive_handle->src_fd, unpack_Z_stream, "uncompress"); |
694 | archive_handle->offset = 0; |
695 | while (get_header_tar(archive_handle) == EXIT_SUCCESS) |
696 | continue; |
697 | |
698 | /* Can only do one file at a time */ |
699 | return EXIT_FAILURE; |
700 | } |
701 | #else |
702 | #define get_header_tar_Z NULL |
703 | #endif |
704 | |
705 | #ifdef CHECK_FOR_CHILD_EXITCODE |
706 | /* Looks like it isn't needed - tar detects malformed (truncated) |
707 | * archive if e.g. bunzip2 fails */ |
708 | static int child_error; |
709 | |
710 | static void handle_SIGCHLD(int status) |
711 | { |
712 | /* Actually, 'status' is a signo. We reuse it for other needs */ |
713 | |
714 | /* Wait for any child without blocking */ |
715 | if (wait_any_nohang(&status) < 0) |
716 | /* wait failed?! I'm confused... */ |
717 | return; |
718 | |
719 | if (WIFEXITED(status) && WEXITSTATUS(status)==0) |
720 | /* child exited with 0 */ |
721 | return; |
722 | /* Cannot happen? |
723 | if (!WIFSIGNALED(status) && !WIFEXITED(status)) return; */ |
724 | child_error = 1; |
725 | } |
726 | #endif |
727 | |
728 | enum { |
729 | OPTBIT_KEEP_OLD = 7, |
730 | USE_FEATURE_TAR_CREATE( OPTBIT_CREATE ,) |
731 | USE_FEATURE_TAR_CREATE( OPTBIT_DEREFERENCE ,) |
732 | USE_FEATURE_SEAMLESS_BZ2( OPTBIT_BZIP2 ,) |
733 | USE_FEATURE_SEAMLESS_LZMA(OPTBIT_LZMA ,) |
734 | USE_FEATURE_TAR_FROM( OPTBIT_INCLUDE_FROM,) |
735 | USE_FEATURE_TAR_FROM( OPTBIT_EXCLUDE_FROM,) |
736 | USE_FEATURE_SEAMLESS_GZ( OPTBIT_GZIP ,) |
737 | USE_FEATURE_SEAMLESS_Z( OPTBIT_COMPRESS ,) |
738 | OPTBIT_NOPRESERVE_OWN, |
739 | OPTBIT_NOPRESERVE_PERM, |
740 | OPT_TEST = 1 << 0, // t |
741 | OPT_EXTRACT = 1 << 1, // x |
742 | OPT_BASEDIR = 1 << 2, // C |
743 | OPT_TARNAME = 1 << 3, // f |
744 | OPT_2STDOUT = 1 << 4, // O |
745 | OPT_P = 1 << 5, // p |
746 | OPT_VERBOSE = 1 << 6, // v |
747 | OPT_KEEP_OLD = 1 << 7, // k |
748 | OPT_CREATE = USE_FEATURE_TAR_CREATE( (1 << OPTBIT_CREATE )) + 0, // c |
749 | OPT_DEREFERENCE = USE_FEATURE_TAR_CREATE( (1 << OPTBIT_DEREFERENCE )) + 0, // h |
750 | OPT_BZIP2 = USE_FEATURE_SEAMLESS_BZ2( (1 << OPTBIT_BZIP2 )) + 0, // j |
751 | OPT_LZMA = USE_FEATURE_SEAMLESS_LZMA((1 << OPTBIT_LZMA )) + 0, // a |
752 | OPT_INCLUDE_FROM = USE_FEATURE_TAR_FROM( (1 << OPTBIT_INCLUDE_FROM)) + 0, // T |
753 | OPT_EXCLUDE_FROM = USE_FEATURE_TAR_FROM( (1 << OPTBIT_EXCLUDE_FROM)) + 0, // X |
754 | OPT_GZIP = USE_FEATURE_SEAMLESS_GZ( (1 << OPTBIT_GZIP )) + 0, // z |
755 | OPT_COMPRESS = USE_FEATURE_SEAMLESS_Z( (1 << OPTBIT_COMPRESS )) + 0, // Z |
756 | OPT_NOPRESERVE_OWN = 1 << OPTBIT_NOPRESERVE_OWN , // no-same-owner |
757 | OPT_NOPRESERVE_PERM = 1 << OPTBIT_NOPRESERVE_PERM, // no-same-permissions |
758 | }; |
759 | #if ENABLE_FEATURE_TAR_LONG_OPTIONS |
760 | static const char tar_longopts[] ALIGN1 = |
761 | "list\0" No_argument "t" |
762 | "extract\0" No_argument "x" |
763 | "directory\0" Required_argument "C" |
764 | "file\0" Required_argument "f" |
765 | "to-stdout\0" No_argument "O" |
766 | "same-permissions\0" No_argument "p" |
767 | "verbose\0" No_argument "v" |
768 | "keep-old\0" No_argument "k" |
769 | # if ENABLE_FEATURE_TAR_CREATE |
770 | "create\0" No_argument "c" |
771 | "dereference\0" No_argument "h" |
772 | # endif |
773 | # if ENABLE_FEATURE_SEAMLESS_BZ2 |
774 | "bzip2\0" No_argument "j" |
775 | # endif |
776 | # if ENABLE_FEATURE_SEAMLESS_LZMA |
777 | "lzma\0" No_argument "a" |
778 | # endif |
779 | # if ENABLE_FEATURE_TAR_FROM |
780 | "files-from\0" Required_argument "T" |
781 | "exclude-from\0" Required_argument "X" |
782 | # endif |
783 | # if ENABLE_FEATURE_SEAMLESS_GZ |
784 | "gzip\0" No_argument "z" |
785 | # endif |
786 | # if ENABLE_FEATURE_SEAMLESS_Z |
787 | "compress\0" No_argument "Z" |
788 | # endif |
789 | "no-same-owner\0" No_argument "\xfd" |
790 | "no-same-permissions\0" No_argument "\xfe" |
791 | /* --exclude takes next bit position in option mask, */ |
792 | /* therefore we have to either put it _after_ --no-same-perm */ |
793 | /* or add OPT[BIT]_EXCLUDE before OPT[BIT]_NOPRESERVE_OWN */ |
794 | # if ENABLE_FEATURE_TAR_FROM |
795 | "exclude\0" Required_argument "\xff" |
796 | # endif |
797 | ; |
798 | #endif |
799 | |
800 | int tar_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; |
801 | int tar_main(int argc UNUSED_PARAM, char **argv) |
802 | { |
803 | char FAST_FUNC (*get_header_ptr)(archive_handle_t *) = get_header_tar; |
804 | archive_handle_t *tar_handle; |
805 | char *base_dir = NULL; |
806 | const char *tar_filename = "-"; |
807 | unsigned opt; |
808 | int verboseFlag = 0; |
809 | #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM |
810 | llist_t *excludes = NULL; |
811 | #endif |
812 | |
813 | /* Initialise default values */ |
814 | tar_handle = init_handle(); |
815 | tar_handle->ah_flags = ARCHIVE_CREATE_LEADING_DIRS |
816 | | ARCHIVE_PRESERVE_DATE |
817 | | ARCHIVE_EXTRACT_UNCONDITIONAL; |
818 | |
819 | /* Apparently only root's tar preserves perms (see bug 3844) */ |
820 | if (getuid() != 0) |
821 | tar_handle->ah_flags |= ARCHIVE_NOPRESERVE_PERM; |
822 | |
823 | /* Prepend '-' to the first argument if required */ |
824 | opt_complementary = "--:" // first arg is options |
825 | "tt:vv:" // count -t,-v |
826 | "?:" // bail out with usage instead of error return |
827 | "X::T::" // cumulative lists |
828 | #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM |
829 | "\xff::" // cumulative lists for --exclude |
830 | #endif |
831 | USE_FEATURE_TAR_CREATE("c:") "t:x:" // at least one of these is reqd |
832 | USE_FEATURE_TAR_CREATE("c--tx:t--cx:x--ct") // mutually exclusive |
833 | SKIP_FEATURE_TAR_CREATE("t--x:x--t"); // mutually exclusive |
834 | #if ENABLE_FEATURE_TAR_LONG_OPTIONS |
835 | applet_long_options = tar_longopts; |
836 | #endif |
837 | opt = getopt32(argv, |
838 | "txC:f:Opvk" |
839 | USE_FEATURE_TAR_CREATE( "ch" ) |
840 | USE_FEATURE_SEAMLESS_BZ2( "j" ) |
841 | USE_FEATURE_SEAMLESS_LZMA("a" ) |
842 | USE_FEATURE_TAR_FROM( "T:X:") |
843 | USE_FEATURE_SEAMLESS_GZ( "z" ) |
844 | USE_FEATURE_SEAMLESS_Z( "Z" ) |
845 | , &base_dir // -C dir |
846 | , &tar_filename // -f filename |
847 | USE_FEATURE_TAR_FROM(, &(tar_handle->accept)) // T |
848 | USE_FEATURE_TAR_FROM(, &(tar_handle->reject)) // X |
849 | #if ENABLE_FEATURE_TAR_LONG_OPTIONS && ENABLE_FEATURE_TAR_FROM |
850 | , &excludes // --exclude |
851 | #endif |
852 | , &verboseFlag // combined count for -t and -v |
853 | , &verboseFlag // combined count for -t and -v |
854 | ); |
855 | argv += optind; |
856 | |
857 | if (verboseFlag) tar_handle->action_header = header_verbose_list; |
858 | if (verboseFlag == 1) tar_handle->action_header = header_list; |
859 | |
860 | if (opt & OPT_EXTRACT) |
861 | tar_handle->action_data = data_extract_all; |
862 | |
863 | if (opt & OPT_2STDOUT) |
864 | tar_handle->action_data = data_extract_to_stdout; |
865 | |
866 | if (opt & OPT_KEEP_OLD) |
867 | tar_handle->ah_flags &= ~ARCHIVE_EXTRACT_UNCONDITIONAL; |
868 | |
869 | if (opt & OPT_NOPRESERVE_OWN) |
870 | tar_handle->ah_flags |= ARCHIVE_NOPRESERVE_OWN; |
871 | |
872 | if (opt & OPT_NOPRESERVE_PERM) |
873 | tar_handle->ah_flags |= ARCHIVE_NOPRESERVE_PERM; |
874 | |
875 | if (opt & OPT_GZIP) |
876 | get_header_ptr = get_header_tar_gz; |
877 | |
878 | if (opt & OPT_BZIP2) |
879 | get_header_ptr = get_header_tar_bz2; |
880 | |
881 | if (opt & OPT_LZMA) |
882 | get_header_ptr = get_header_tar_lzma; |
883 | |
884 | if (opt & OPT_COMPRESS) |
885 | get_header_ptr = get_header_tar_Z; |
886 | |
887 | #if ENABLE_FEATURE_TAR_FROM |
888 | tar_handle->reject = append_file_list_to_list(tar_handle->reject); |
889 | #if ENABLE_FEATURE_TAR_LONG_OPTIONS |
890 | /* Append excludes to reject */ |
891 | while (excludes) { |
892 | llist_t *next = excludes->link; |
893 | excludes->link = tar_handle->reject; |
894 | tar_handle->reject = excludes; |
895 | excludes = next; |
896 | } |
897 | #endif |
898 | tar_handle->accept = append_file_list_to_list(tar_handle->accept); |
899 | #endif |
900 | |
901 | /* Setup an array of filenames to work with */ |
902 | /* TODO: This is the same as in ar, separate function ? */ |
903 | while (*argv) { |
904 | /* kill trailing '/' unless the string is just "/" */ |
905 | char *cp = last_char_is(*argv, '/'); |
906 | if (cp > *argv) |
907 | *cp = '\0'; |
908 | llist_add_to_end(&tar_handle->accept, *argv); |
909 | argv++; |
910 | } |
911 | |
912 | if (tar_handle->accept || tar_handle->reject) |
913 | tar_handle->filter = filter_accept_reject_list; |
914 | |
915 | /* Open the tar file */ |
916 | { |
917 | FILE *tar_stream; |
918 | int flags; |
919 | |
920 | if (opt & OPT_CREATE) { |
921 | /* Make sure there is at least one file to tar up. */ |
922 | if (tar_handle->accept == NULL) |
923 | bb_error_msg_and_die("empty archive"); |
924 | |
925 | tar_stream = stdout; |
926 | /* Mimicking GNU tar 1.15.1: */ |
927 | flags = O_WRONLY | O_CREAT | O_TRUNC; |
928 | } else { |
929 | tar_stream = stdin; |
930 | flags = O_RDONLY; |
931 | } |
932 | |
933 | if (LONE_DASH(tar_filename)) { |
934 | tar_handle->src_fd = fileno(tar_stream); |
935 | tar_handle->seek = seek_by_read; |
936 | } else { |
937 | if (ENABLE_FEATURE_TAR_AUTODETECT |
938 | && get_header_ptr == get_header_tar |
939 | && flags == O_RDONLY |
940 | ) { |
941 | tar_handle->src_fd = open_zipped(tar_filename); |
942 | if (tar_handle->src_fd < 0) |
943 | bb_perror_msg_and_die("can't open '%s'", tar_filename); |
944 | } else { |
945 | tar_handle->src_fd = xopen(tar_filename, flags); |
946 | } |
947 | } |
948 | } |
949 | |
950 | if (base_dir) |
951 | xchdir(base_dir); |
952 | |
953 | #ifdef CHECK_FOR_CHILD_EXITCODE |
954 | /* We need to know whether child (gzip/bzip/etc) exits abnormally */ |
955 | signal(SIGCHLD, handle_SIGCHLD); |
956 | #endif |
957 | |
958 | /* create an archive */ |
959 | if (opt & OPT_CREATE) { |
960 | #if ENABLE_FEATURE_SEAMLESS_GZ || ENABLE_FEATURE_SEAMLESS_BZ2 |
961 | int zipMode = 0; |
962 | if (ENABLE_FEATURE_SEAMLESS_GZ && (opt & OPT_GZIP)) |
963 | zipMode = 1; |
964 | if (ENABLE_FEATURE_SEAMLESS_BZ2 && (opt & OPT_BZIP2)) |
965 | zipMode = 2; |
966 | #endif |
967 | /* NB: writeTarFile() closes tar_handle->src_fd */ |
968 | return writeTarFile(tar_handle->src_fd, verboseFlag, opt & OPT_DEREFERENCE, |
969 | tar_handle->accept, |
970 | tar_handle->reject, zipMode); |
971 | } |
972 | |
973 | while (get_header_ptr(tar_handle) == EXIT_SUCCESS) |
974 | continue; |
975 | |
976 | /* Check that every file that should have been extracted was */ |
977 | while (tar_handle->accept) { |
978 | if (!find_list_entry(tar_handle->reject, tar_handle->accept->data) |
979 | && !find_list_entry(tar_handle->passed, tar_handle->accept->data) |
980 | ) { |
981 | bb_error_msg_and_die("%s: not found in archive", |
982 | tar_handle->accept->data); |
983 | } |
984 | tar_handle->accept = tar_handle->accept->link; |
985 | } |
986 | if (ENABLE_FEATURE_CLEAN_UP /* && tar_handle->src_fd != STDIN_FILENO */) |
987 | close(tar_handle->src_fd); |
988 | |
989 | return EXIT_SUCCESS; |
990 | } |