28
29#ifdef HAVE_SYS_STAT_H
30#include <sys/stat.h>
31#endif
32#ifdef HAVE_SYS_TYPES_H
33#include <sys/types.h> /* Linux doesn't define mode_t, etc. in sys/stat.h. */
34#endif
35#include <ctype.h>
36#ifdef HAVE_ERRNO_H
37#include <errno.h>
38#endif
39#ifdef HAVE_IO_H
40#include <io.h>
41#endif
42#ifdef HAVE_STDARG_H
43#include <stdarg.h>
44#endif
45#ifdef HAVE_STDINT_H
46#include <stdint.h>
47#endif
48#include <stdio.h>
49#ifdef HAVE_STDLIB_H
50#include <stdlib.h>
51#endif
52#ifdef HAVE_STRING_H
53#include <string.h>
54#endif
55#ifdef HAVE_WCTYPE_H
56#include <wctype.h>
57#else
58/* If we don't have wctype, we need to hack up some version of iswprint(). */
59#define iswprint isprint
60#endif
61
62#include "bsdtar.h"
63#include "err.h"
64#include "passphrase.h"
65
66static size_t bsdtar_expand_char(char *, size_t, char);
67static const char *strip_components(const char *path, int elements);
68
69#if defined(_WIN32) && !defined(__CYGWIN__)
70#define read _read
71#endif
72
73/* TODO: Hack up a version of mbtowc for platforms with no wide
74 * character support at all. I think the following might suffice,
75 * but it needs careful testing.
76 * #if !HAVE_MBTOWC
77 * #define mbtowc(wcp, p, n) ((*wcp = *p), 1)
78 * #endif
79 */
80
81/*
82 * Print a string, taking care with any non-printable characters.
83 *
84 * Note that we use a stack-allocated buffer to receive the formatted
85 * string if we can. This is partly performance (avoiding a call to
86 * malloc()), partly out of expedience (we have to call vsnprintf()
87 * before malloc() anyway to find out how big a buffer we need; we may
88 * as well point that first call at a small local buffer in case it
89 * works), but mostly for safety (so we can use this to print messages
90 * about out-of-memory conditions).
91 */
92
93void
94safe_fprintf(FILE *f, const char *fmt, ...)
95{
96 char fmtbuff_stack[256]; /* Place to format the printf() string. */
97 char outbuff[256]; /* Buffer for outgoing characters. */
98 char *fmtbuff_heap; /* If fmtbuff_stack is too small, we use malloc */
99 char *fmtbuff; /* Pointer to fmtbuff_stack or fmtbuff_heap. */
100 int fmtbuff_length;
101 int length, n;
102 va_list ap;
103 const char *p;
104 unsigned i;
105 wchar_t wc;
106 char try_wc;
107
108 /* Use a stack-allocated buffer if we can, for speed and safety. */
109 fmtbuff_heap = NULL;
110 fmtbuff_length = sizeof(fmtbuff_stack);
111 fmtbuff = fmtbuff_stack;
112
113 /* Try formatting into the stack buffer. */
114 va_start(ap, fmt);
115 length = vsnprintf(fmtbuff, fmtbuff_length, fmt, ap);
116 va_end(ap);
117
118 /* If the result was too large, allocate a buffer on the heap. */
119 while (length < 0 || length >= fmtbuff_length) {
120 if (length >= fmtbuff_length)
121 fmtbuff_length = length+1;
122 else if (fmtbuff_length < 8192)
123 fmtbuff_length *= 2;
124 else if (fmtbuff_length < 1000000)
125 fmtbuff_length += fmtbuff_length / 4;
126 else {
127 length = fmtbuff_length;
128 fmtbuff_heap[length-1] = '\0';
129 break;
130 }
131 free(fmtbuff_heap);
132 fmtbuff_heap = malloc(fmtbuff_length);
133
134 /* Reformat the result into the heap buffer if we can. */
135 if (fmtbuff_heap != NULL) {
136 fmtbuff = fmtbuff_heap;
137 va_start(ap, fmt);
138 length = vsnprintf(fmtbuff, fmtbuff_length, fmt, ap);
139 va_end(ap);
140 } else {
141 /* Leave fmtbuff pointing to the truncated
142 * string in fmtbuff_stack. */
143 fmtbuff = fmtbuff_stack;
144 length = sizeof(fmtbuff_stack) - 1;
145 break;
146 }
147 }
148
149 /* Note: mbrtowc() has a cleaner API, but mbtowc() seems a bit
150 * more portable, so we use that here instead. */
151 if (mbtowc(NULL, NULL, 1) == -1) { /* Reset the shift state. */
152 /* mbtowc() should never fail in practice, but
153 * handle the theoretical error anyway. */
154 free(fmtbuff_heap);
155 return;
156 }
157
158 /* Write data, expanding unprintable characters. */
159 p = fmtbuff;
160 i = 0;
161 try_wc = 1;
162 while (*p != '\0') {
163
164 /* Convert to wide char, test if the wide
165 * char is printable in the current locale. */
166 if (try_wc && (n = mbtowc(&wc, p, length)) != -1) {
167 length -= n;
168 if (iswprint(wc) && wc != L'\\') {
169 /* Printable, copy the bytes through. */
170 while (n-- > 0)
171 outbuff[i++] = *p++;
172 } else {
173 /* Not printable, format the bytes. */
174 while (n-- > 0)
175 i += (unsigned)bsdtar_expand_char(
176 outbuff, i, *p++);
177 }
178 } else {
179 /* After any conversion failure, don't bother
180 * trying to convert the rest. */
181 i += (unsigned)bsdtar_expand_char(outbuff, i, *p++);
182 try_wc = 0;
183 }
184
185 /* If our output buffer is full, dump it and keep going. */
186 if (i > (sizeof(outbuff) - 128)) {
187 outbuff[i] = '\0';
188 fprintf(f, "%s", outbuff);
189 i = 0;
190 }
191 }
192 outbuff[i] = '\0';
193 fprintf(f, "%s", outbuff);
194
195 /* If we allocated a heap-based formatting buffer, free it now. */
196 free(fmtbuff_heap);
197}
198
199/*
200 * Render an arbitrary sequence of bytes into printable ASCII characters.
201 */
202static size_t
203bsdtar_expand_char(char *buff, size_t offset, char c)
204{
205 size_t i = offset;
206
207 if (isprint((unsigned char)c) && c != '\\')
208 buff[i++] = c;
209 else {
210 buff[i++] = '\\';
211 switch (c) {
212 case '\a': buff[i++] = 'a'; break;
213 case '\b': buff[i++] = 'b'; break;
214 case '\f': buff[i++] = 'f'; break;
215 case '\n': buff[i++] = 'n'; break;
216#if '\r' != '\n'
217 /* On some platforms, \n and \r are the same. */
218 case '\r': buff[i++] = 'r'; break;
219#endif
220 case '\t': buff[i++] = 't'; break;
221 case '\v': buff[i++] = 'v'; break;
222 case '\\': buff[i++] = '\\'; break;
223 default:
224 sprintf(buff + i, "%03o", 0xFF & (int)c);
225 i += 3;
226 }
227 }
228
229 return (i - offset);
230}
231
232int
233yes(const char *fmt, ...)
234{
235 char buff[32];
236 char *p;
237 ssize_t l;
238
239 va_list ap;
240 va_start(ap, fmt);
241 vfprintf(stderr, fmt, ap);
242 va_end(ap);
243 fprintf(stderr, " (y/N)? ");
244 fflush(stderr);
245
246 l = read(2, buff, sizeof(buff) - 1);
247 if (l < 0) {
248 fprintf(stderr, "Keyboard read failed\n");
249 exit(1);
250 }
251 if (l == 0)
252 return (0);
253 buff[l] = 0;
254
255 for (p = buff; *p != '\0'; p++) {
256 if (isspace((unsigned char)*p))
257 continue;
258 switch(*p) {
259 case 'y': case 'Y':
260 return (1);
261 case 'n': case 'N':
262 return (0);
263 default:
264 return (0);
265 }
266 }
267
268 return (0);
269}
270
271/*-
272 * The logic here for -C <dir> attempts to avoid
273 * chdir() as long as possible. For example:
274 * "-C /foo -C /bar file" needs chdir("/bar") but not chdir("/foo")
275 * "-C /foo -C bar file" needs chdir("/foo/bar")
276 * "-C /foo -C bar /file1" does not need chdir()
277 * "-C /foo -C bar /file1 file2" needs chdir("/foo/bar") before file2
278 *
279 * The only correct way to handle this is to record a "pending" chdir
280 * request and combine multiple requests intelligently until we
281 * need to process a non-absolute file. set_chdir() adds the new dir
282 * to the pending list; do_chdir() actually executes any pending chdir.
283 *
284 * This way, programs that build tar command lines don't have to worry
285 * about -C with non-existent directories; such requests will only
286 * fail if the directory must be accessed.
287 *
288 */
289void
290set_chdir(struct bsdtar *bsdtar, const char *newdir)
291{
292#if defined(_WIN32) && !defined(__CYGWIN__)
293 if (newdir[0] == '/' || newdir[0] == '\\' ||
294 /* Detect this type, for example, "C:\" or "C:/" */
295 (((newdir[0] >= 'a' && newdir[0] <= 'z') ||
296 (newdir[0] >= 'A' && newdir[0] <= 'Z')) &&
297 newdir[1] == ':' && (newdir[2] == '/' || newdir[2] == '\\'))) {
298#else
299 if (newdir[0] == '/') {
300#endif
301 /* The -C /foo -C /bar case; dump first one. */
302 free(bsdtar->pending_chdir);
303 bsdtar->pending_chdir = NULL;
304 }
305 if (bsdtar->pending_chdir == NULL)
306 /* Easy case: no previously-saved dir. */
307 bsdtar->pending_chdir = strdup(newdir);
308 else {
309 /* The -C /foo -C bar case; concatenate */
310 char *old_pending = bsdtar->pending_chdir;
311 size_t old_len = strlen(old_pending);
312 bsdtar->pending_chdir = malloc(old_len + strlen(newdir) + 2);
313 if (old_pending[old_len - 1] == '/')
314 old_pending[old_len - 1] = '\0';
315 if (bsdtar->pending_chdir != NULL)
316 sprintf(bsdtar->pending_chdir, "%s/%s",
317 old_pending, newdir);
318 free(old_pending);
319 }
320 if (bsdtar->pending_chdir == NULL)
321 lafe_errc(1, errno, "No memory");
322}
323
324void
325do_chdir(struct bsdtar *bsdtar)
326{
327 if (bsdtar->pending_chdir == NULL)
328 return;
329
330 if (chdir(bsdtar->pending_chdir) != 0) {
331 lafe_errc(1, 0, "could not chdir to '%s'\n",
332 bsdtar->pending_chdir);
333 }
334 free(bsdtar->pending_chdir);
335 bsdtar->pending_chdir = NULL;
336}
337
338static const char *
339strip_components(const char *p, int elements)
340{
341 /* Skip as many elements as necessary. */
342 while (elements > 0) {
343 switch (*p++) {
344 case '/':
345#if defined(_WIN32) && !defined(__CYGWIN__)
346 case '\\': /* Support \ path sep on Windows ONLY. */
347#endif
348 elements--;
349 break;
350 case '\0':
351 /* Path is too short, skip it. */
352 return (NULL);
353 }
354 }
355
356 /* Skip any / characters. This handles short paths that have
357 * additional / termination. This also handles the case where
358 * the logic above stops in the middle of a duplicate //
359 * sequence (which would otherwise get converted to an
360 * absolute path). */
361 for (;;) {
362 switch (*p) {
363 case '/':
364#if defined(_WIN32) && !defined(__CYGWIN__)
365 case '\\': /* Support \ path sep on Windows ONLY. */
366#endif
367 ++p;
368 break;
369 case '\0':
370 return (NULL);
371 default:
372 return (p);
373 }
374 }
375}
376
377static void
378warn_strip_leading_char(struct bsdtar *bsdtar, const char *c)
379{
380 if (!bsdtar->warned_lead_slash) {
381 lafe_warnc(0,
382 "Removing leading '%c' from member names",
383 c[0]);
384 bsdtar->warned_lead_slash = 1;
385 }
386}
387
388static void
389warn_strip_drive_letter(struct bsdtar *bsdtar)
390{
391 if (!bsdtar->warned_lead_slash) {
392 lafe_warnc(0,
393 "Removing leading drive letter from "
394 "member names");
395 bsdtar->warned_lead_slash = 1;
396 }
397}
398
399/*
400 * Convert absolute path to non-absolute path by skipping leading
401 * absolute path prefixes.
402 */
403static const char*
404strip_absolute_path(struct bsdtar *bsdtar, const char *p)
405{
406 const char *rp;
407
408 /* Remove leading "//./" or "//?/" or "//?/UNC/"
409 * (absolute path prefixes used by Windows API) */
410 if ((p[0] == '/' || p[0] == '\\') &&
411 (p[1] == '/' || p[1] == '\\') &&
412 (p[2] == '.' || p[2] == '?') &&
413 (p[3] == '/' || p[3] == '\\'))
414 {
415 if (p[2] == '?' &&
416 (p[4] == 'U' || p[4] == 'u') &&
417 (p[5] == 'N' || p[5] == 'n') &&
418 (p[6] == 'C' || p[6] == 'c') &&
419 (p[7] == '/' || p[7] == '\\'))
420 p += 8;
421 else
422 p += 4;
423 warn_strip_drive_letter(bsdtar);
424 }
425
426 /* Remove multiple leading slashes and Windows drive letters. */
427 do {
428 rp = p;
429 if (((p[0] >= 'a' && p[0] <= 'z') ||
430 (p[0] >= 'A' && p[0] <= 'Z')) &&
431 p[1] == ':') {
432 p += 2;
433 warn_strip_drive_letter(bsdtar);
434 }
435
436 /* Remove leading "/../", "/./", "//", etc. */
437 while (p[0] == '/' || p[0] == '\\') {
438 if (p[1] == '.' &&
439 p[2] == '.' &&
440 (p[3] == '/' || p[3] == '\\')) {
441 p += 3; /* Remove "/..", leave "/" for next pass. */
442 } else if (p[1] == '.' &&
443 (p[2] == '/' || p[2] == '\\')) {
444 p += 2; /* Remove "/.", leave "/" for next pass. */
445 } else
446 p += 1; /* Remove "/". */
447 warn_strip_leading_char(bsdtar, rp);
448 }
449 } while (rp != p);
450
451 return (p);
452}
453
454/*
455 * Handle --strip-components and any future path-rewriting options.
456 * Returns non-zero if the pathname should not be extracted.
457 *
458 * Note: The rewrites are applied uniformly to pathnames and hardlink
459 * names but not to symlink bodies. This is deliberate: Symlink
460 * bodies are not necessarily filenames. Even when they are, they
461 * need to be interpreted relative to the directory containing them,
462 * so simple rewrites like this are rarely appropriate.
463 *
464 * TODO: Support pax-style regex path rewrites.
465 */
466int
467edit_pathname(struct bsdtar *bsdtar, struct archive_entry *entry)
468{
469 const char *name = archive_entry_pathname(entry);
470 const char *original_name = name;
471 const char *hardlinkname = archive_entry_hardlink(entry);
472 const char *original_hardlinkname = hardlinkname;
473#if defined(HAVE_REGEX_H) || defined(HAVE_PCREPOSIX_H)
474 char *subst_name;
475 int r;
476
477 /* Apply user-specified substitution to pathname. */
478 r = apply_substitution(bsdtar, name, &subst_name, 0, 0);
479 if (r == -1) {
480 lafe_warnc(0, "Invalid substitution, skipping entry");
481 return 1;
482 }
483 if (r == 1) {
484 archive_entry_copy_pathname(entry, subst_name);
485 if (*subst_name == '\0') {
486 free(subst_name);
487 return -1;
488 } else
489 free(subst_name);
490 name = archive_entry_pathname(entry);
491 original_name = name;
492 }
493
494 /* Apply user-specified substitution to hardlink target. */
495 if (hardlinkname != NULL) {
496 r = apply_substitution(bsdtar, hardlinkname, &subst_name, 0, 1);
497 if (r == -1) {
498 lafe_warnc(0, "Invalid substitution, skipping entry");
499 return 1;
500 }
501 if (r == 1) {
502 archive_entry_copy_hardlink(entry, subst_name);
503 free(subst_name);
504 }
505 hardlinkname = archive_entry_hardlink(entry);
506 original_hardlinkname = hardlinkname;
507 }
508
509 /* Apply user-specified substitution to symlink body. */
510 if (archive_entry_symlink(entry) != NULL) {
511 r = apply_substitution(bsdtar, archive_entry_symlink(entry), &subst_name, 1, 0);
512 if (r == -1) {
513 lafe_warnc(0, "Invalid substitution, skipping entry");
514 return 1;
515 }
516 if (r == 1) {
517 archive_entry_copy_symlink(entry, subst_name);
518 free(subst_name);
519 }
520 }
521#endif
522
523 /* Strip leading dir names as per --strip-components option. */
524 if (bsdtar->strip_components > 0) {
525 name = strip_components(name, bsdtar->strip_components);
526 if (name == NULL)
527 return (1);
528
529 if (hardlinkname != NULL) {
530 hardlinkname = strip_components(hardlinkname,
531 bsdtar->strip_components);
532 if (hardlinkname == NULL)
533 return (1);
534 }
535 }
536
537 if ((bsdtar->flags & OPTFLAG_ABSOLUTE_PATHS) == 0) {
538 /* By default, don't write or restore absolute pathnames. */
539 name = strip_absolute_path(bsdtar, name);
540 if (*name == '\0')
541 name = ".";
542
543 if (hardlinkname != NULL) {
544 hardlinkname = strip_absolute_path(bsdtar, hardlinkname);
545 if (*hardlinkname == '\0')
546 return (1);
547 }
548 } else {
549 /* Strip redundant leading '/' characters. */
550 while (name[0] == '/' && name[1] == '/')
551 name++;
552 }
553
554 /* Replace name in archive_entry. */
555 if (name != original_name) {
556 archive_entry_copy_pathname(entry, name);
557 }
558 if (hardlinkname != original_hardlinkname) {
559 archive_entry_copy_hardlink(entry, hardlinkname);
560 }
561 return (0);
562}
563
564/*
565 * It would be nice to just use printf() for formatting large numbers,
566 * but the compatibility problems are quite a headache. Hence the
567 * following simple utility function.
568 */
569const char *
570tar_i64toa(int64_t n0)
571{
572 static char buff[24];
573 uint64_t n = n0 < 0 ? -n0 : n0;
574 char *p = buff + sizeof(buff);
575
576 *--p = '\0';
577 do {
578 *--p = '0' + (int)(n % 10);
579 } while (n /= 10);
580 if (n0 < 0)
581 *--p = '-';
582 return p;
583}
584
585/*
586 * Like strcmp(), but try to be a little more aware of the fact that
587 * we're comparing two paths. Right now, it just handles leading
588 * "./" and trailing '/' specially, so that "a/b/" == "./a/b"
589 *
590 * TODO: Make this better, so that "./a//b/./c/" == "a/b/c"
591 * TODO: After this works, push it down into libarchive.
592 * TODO: Publish the path normalization routines in libarchive so
593 * that bsdtar can normalize paths and use fast strcmp() instead
594 * of this.
595 *
596 * Note: This is currently only used within write.c, so should
597 * not handle \ path separators.
598 */
599
600int
601pathcmp(const char *a, const char *b)
602{
603 /* Skip leading './' */
604 if (a[0] == '.' && a[1] == '/' && a[2] != '\0')
605 a += 2;
606 if (b[0] == '.' && b[1] == '/' && b[2] != '\0')
607 b += 2;
608 /* Find the first difference, or return (0) if none. */
609 while (*a == *b) {
610 if (*a == '\0')
611 return (0);
612 a++;
613 b++;
614 }
615 /*
616 * If one ends in '/' and the other one doesn't,
617 * they're the same.
618 */
619 if (a[0] == '/' && a[1] == '\0' && b[0] == '\0')
620 return (0);
621 if (a[0] == '\0' && b[0] == '/' && b[1] == '\0')
622 return (0);
623 /* They're really different, return the correct sign. */
624 return (*(const unsigned char *)a - *(const unsigned char *)b);
625}
626
627#define PPBUFF_SIZE 1024
628const char *
629passphrase_callback(struct archive *a, void *_client_data)
630{
631 struct bsdtar *bsdtar = (struct bsdtar *)_client_data;
632 (void)a; /* UNUSED */
633
634 if (bsdtar->ppbuff == NULL) {
635 bsdtar->ppbuff = malloc(PPBUFF_SIZE);
636 if (bsdtar->ppbuff == NULL)
637 lafe_errc(1, errno, "Out of memory");
638 }
639 return lafe_readpassphrase("Enter passphrase:",
640 bsdtar->ppbuff, PPBUFF_SIZE);
641}
642
643void
644passphrase_free(char *ppbuff)
645{
646 if (ppbuff != NULL) {
647 memset(ppbuff, 0, PPBUFF_SIZE);
648 free(ppbuff);
649 }
650}
651
652/*
653 * Display information about the current file.
654 *
655 * The format here roughly duplicates the output of 'ls -l'.
656 * This is based on SUSv2, where 'tar tv' is documented as
657 * listing additional information in an "unspecified format,"
658 * and 'pax -l' is documented as using the same format as 'ls -l'.
659 */
660void
661list_item_verbose(struct bsdtar *bsdtar, FILE *out, struct archive_entry *entry)
662{
663 char tmp[100];
664 size_t w;
665 const char *p;
666 const char *fmt;
667 time_t tim;
668 static time_t now;
| 28
29#ifdef HAVE_SYS_STAT_H
30#include <sys/stat.h>
31#endif
32#ifdef HAVE_SYS_TYPES_H
33#include <sys/types.h> /* Linux doesn't define mode_t, etc. in sys/stat.h. */
34#endif
35#include <ctype.h>
36#ifdef HAVE_ERRNO_H
37#include <errno.h>
38#endif
39#ifdef HAVE_IO_H
40#include <io.h>
41#endif
42#ifdef HAVE_STDARG_H
43#include <stdarg.h>
44#endif
45#ifdef HAVE_STDINT_H
46#include <stdint.h>
47#endif
48#include <stdio.h>
49#ifdef HAVE_STDLIB_H
50#include <stdlib.h>
51#endif
52#ifdef HAVE_STRING_H
53#include <string.h>
54#endif
55#ifdef HAVE_WCTYPE_H
56#include <wctype.h>
57#else
58/* If we don't have wctype, we need to hack up some version of iswprint(). */
59#define iswprint isprint
60#endif
61
62#include "bsdtar.h"
63#include "err.h"
64#include "passphrase.h"
65
66static size_t bsdtar_expand_char(char *, size_t, char);
67static const char *strip_components(const char *path, int elements);
68
69#if defined(_WIN32) && !defined(__CYGWIN__)
70#define read _read
71#endif
72
73/* TODO: Hack up a version of mbtowc for platforms with no wide
74 * character support at all. I think the following might suffice,
75 * but it needs careful testing.
76 * #if !HAVE_MBTOWC
77 * #define mbtowc(wcp, p, n) ((*wcp = *p), 1)
78 * #endif
79 */
80
81/*
82 * Print a string, taking care with any non-printable characters.
83 *
84 * Note that we use a stack-allocated buffer to receive the formatted
85 * string if we can. This is partly performance (avoiding a call to
86 * malloc()), partly out of expedience (we have to call vsnprintf()
87 * before malloc() anyway to find out how big a buffer we need; we may
88 * as well point that first call at a small local buffer in case it
89 * works), but mostly for safety (so we can use this to print messages
90 * about out-of-memory conditions).
91 */
92
93void
94safe_fprintf(FILE *f, const char *fmt, ...)
95{
96 char fmtbuff_stack[256]; /* Place to format the printf() string. */
97 char outbuff[256]; /* Buffer for outgoing characters. */
98 char *fmtbuff_heap; /* If fmtbuff_stack is too small, we use malloc */
99 char *fmtbuff; /* Pointer to fmtbuff_stack or fmtbuff_heap. */
100 int fmtbuff_length;
101 int length, n;
102 va_list ap;
103 const char *p;
104 unsigned i;
105 wchar_t wc;
106 char try_wc;
107
108 /* Use a stack-allocated buffer if we can, for speed and safety. */
109 fmtbuff_heap = NULL;
110 fmtbuff_length = sizeof(fmtbuff_stack);
111 fmtbuff = fmtbuff_stack;
112
113 /* Try formatting into the stack buffer. */
114 va_start(ap, fmt);
115 length = vsnprintf(fmtbuff, fmtbuff_length, fmt, ap);
116 va_end(ap);
117
118 /* If the result was too large, allocate a buffer on the heap. */
119 while (length < 0 || length >= fmtbuff_length) {
120 if (length >= fmtbuff_length)
121 fmtbuff_length = length+1;
122 else if (fmtbuff_length < 8192)
123 fmtbuff_length *= 2;
124 else if (fmtbuff_length < 1000000)
125 fmtbuff_length += fmtbuff_length / 4;
126 else {
127 length = fmtbuff_length;
128 fmtbuff_heap[length-1] = '\0';
129 break;
130 }
131 free(fmtbuff_heap);
132 fmtbuff_heap = malloc(fmtbuff_length);
133
134 /* Reformat the result into the heap buffer if we can. */
135 if (fmtbuff_heap != NULL) {
136 fmtbuff = fmtbuff_heap;
137 va_start(ap, fmt);
138 length = vsnprintf(fmtbuff, fmtbuff_length, fmt, ap);
139 va_end(ap);
140 } else {
141 /* Leave fmtbuff pointing to the truncated
142 * string in fmtbuff_stack. */
143 fmtbuff = fmtbuff_stack;
144 length = sizeof(fmtbuff_stack) - 1;
145 break;
146 }
147 }
148
149 /* Note: mbrtowc() has a cleaner API, but mbtowc() seems a bit
150 * more portable, so we use that here instead. */
151 if (mbtowc(NULL, NULL, 1) == -1) { /* Reset the shift state. */
152 /* mbtowc() should never fail in practice, but
153 * handle the theoretical error anyway. */
154 free(fmtbuff_heap);
155 return;
156 }
157
158 /* Write data, expanding unprintable characters. */
159 p = fmtbuff;
160 i = 0;
161 try_wc = 1;
162 while (*p != '\0') {
163
164 /* Convert to wide char, test if the wide
165 * char is printable in the current locale. */
166 if (try_wc && (n = mbtowc(&wc, p, length)) != -1) {
167 length -= n;
168 if (iswprint(wc) && wc != L'\\') {
169 /* Printable, copy the bytes through. */
170 while (n-- > 0)
171 outbuff[i++] = *p++;
172 } else {
173 /* Not printable, format the bytes. */
174 while (n-- > 0)
175 i += (unsigned)bsdtar_expand_char(
176 outbuff, i, *p++);
177 }
178 } else {
179 /* After any conversion failure, don't bother
180 * trying to convert the rest. */
181 i += (unsigned)bsdtar_expand_char(outbuff, i, *p++);
182 try_wc = 0;
183 }
184
185 /* If our output buffer is full, dump it and keep going. */
186 if (i > (sizeof(outbuff) - 128)) {
187 outbuff[i] = '\0';
188 fprintf(f, "%s", outbuff);
189 i = 0;
190 }
191 }
192 outbuff[i] = '\0';
193 fprintf(f, "%s", outbuff);
194
195 /* If we allocated a heap-based formatting buffer, free it now. */
196 free(fmtbuff_heap);
197}
198
199/*
200 * Render an arbitrary sequence of bytes into printable ASCII characters.
201 */
202static size_t
203bsdtar_expand_char(char *buff, size_t offset, char c)
204{
205 size_t i = offset;
206
207 if (isprint((unsigned char)c) && c != '\\')
208 buff[i++] = c;
209 else {
210 buff[i++] = '\\';
211 switch (c) {
212 case '\a': buff[i++] = 'a'; break;
213 case '\b': buff[i++] = 'b'; break;
214 case '\f': buff[i++] = 'f'; break;
215 case '\n': buff[i++] = 'n'; break;
216#if '\r' != '\n'
217 /* On some platforms, \n and \r are the same. */
218 case '\r': buff[i++] = 'r'; break;
219#endif
220 case '\t': buff[i++] = 't'; break;
221 case '\v': buff[i++] = 'v'; break;
222 case '\\': buff[i++] = '\\'; break;
223 default:
224 sprintf(buff + i, "%03o", 0xFF & (int)c);
225 i += 3;
226 }
227 }
228
229 return (i - offset);
230}
231
232int
233yes(const char *fmt, ...)
234{
235 char buff[32];
236 char *p;
237 ssize_t l;
238
239 va_list ap;
240 va_start(ap, fmt);
241 vfprintf(stderr, fmt, ap);
242 va_end(ap);
243 fprintf(stderr, " (y/N)? ");
244 fflush(stderr);
245
246 l = read(2, buff, sizeof(buff) - 1);
247 if (l < 0) {
248 fprintf(stderr, "Keyboard read failed\n");
249 exit(1);
250 }
251 if (l == 0)
252 return (0);
253 buff[l] = 0;
254
255 for (p = buff; *p != '\0'; p++) {
256 if (isspace((unsigned char)*p))
257 continue;
258 switch(*p) {
259 case 'y': case 'Y':
260 return (1);
261 case 'n': case 'N':
262 return (0);
263 default:
264 return (0);
265 }
266 }
267
268 return (0);
269}
270
271/*-
272 * The logic here for -C <dir> attempts to avoid
273 * chdir() as long as possible. For example:
274 * "-C /foo -C /bar file" needs chdir("/bar") but not chdir("/foo")
275 * "-C /foo -C bar file" needs chdir("/foo/bar")
276 * "-C /foo -C bar /file1" does not need chdir()
277 * "-C /foo -C bar /file1 file2" needs chdir("/foo/bar") before file2
278 *
279 * The only correct way to handle this is to record a "pending" chdir
280 * request and combine multiple requests intelligently until we
281 * need to process a non-absolute file. set_chdir() adds the new dir
282 * to the pending list; do_chdir() actually executes any pending chdir.
283 *
284 * This way, programs that build tar command lines don't have to worry
285 * about -C with non-existent directories; such requests will only
286 * fail if the directory must be accessed.
287 *
288 */
289void
290set_chdir(struct bsdtar *bsdtar, const char *newdir)
291{
292#if defined(_WIN32) && !defined(__CYGWIN__)
293 if (newdir[0] == '/' || newdir[0] == '\\' ||
294 /* Detect this type, for example, "C:\" or "C:/" */
295 (((newdir[0] >= 'a' && newdir[0] <= 'z') ||
296 (newdir[0] >= 'A' && newdir[0] <= 'Z')) &&
297 newdir[1] == ':' && (newdir[2] == '/' || newdir[2] == '\\'))) {
298#else
299 if (newdir[0] == '/') {
300#endif
301 /* The -C /foo -C /bar case; dump first one. */
302 free(bsdtar->pending_chdir);
303 bsdtar->pending_chdir = NULL;
304 }
305 if (bsdtar->pending_chdir == NULL)
306 /* Easy case: no previously-saved dir. */
307 bsdtar->pending_chdir = strdup(newdir);
308 else {
309 /* The -C /foo -C bar case; concatenate */
310 char *old_pending = bsdtar->pending_chdir;
311 size_t old_len = strlen(old_pending);
312 bsdtar->pending_chdir = malloc(old_len + strlen(newdir) + 2);
313 if (old_pending[old_len - 1] == '/')
314 old_pending[old_len - 1] = '\0';
315 if (bsdtar->pending_chdir != NULL)
316 sprintf(bsdtar->pending_chdir, "%s/%s",
317 old_pending, newdir);
318 free(old_pending);
319 }
320 if (bsdtar->pending_chdir == NULL)
321 lafe_errc(1, errno, "No memory");
322}
323
324void
325do_chdir(struct bsdtar *bsdtar)
326{
327 if (bsdtar->pending_chdir == NULL)
328 return;
329
330 if (chdir(bsdtar->pending_chdir) != 0) {
331 lafe_errc(1, 0, "could not chdir to '%s'\n",
332 bsdtar->pending_chdir);
333 }
334 free(bsdtar->pending_chdir);
335 bsdtar->pending_chdir = NULL;
336}
337
338static const char *
339strip_components(const char *p, int elements)
340{
341 /* Skip as many elements as necessary. */
342 while (elements > 0) {
343 switch (*p++) {
344 case '/':
345#if defined(_WIN32) && !defined(__CYGWIN__)
346 case '\\': /* Support \ path sep on Windows ONLY. */
347#endif
348 elements--;
349 break;
350 case '\0':
351 /* Path is too short, skip it. */
352 return (NULL);
353 }
354 }
355
356 /* Skip any / characters. This handles short paths that have
357 * additional / termination. This also handles the case where
358 * the logic above stops in the middle of a duplicate //
359 * sequence (which would otherwise get converted to an
360 * absolute path). */
361 for (;;) {
362 switch (*p) {
363 case '/':
364#if defined(_WIN32) && !defined(__CYGWIN__)
365 case '\\': /* Support \ path sep on Windows ONLY. */
366#endif
367 ++p;
368 break;
369 case '\0':
370 return (NULL);
371 default:
372 return (p);
373 }
374 }
375}
376
377static void
378warn_strip_leading_char(struct bsdtar *bsdtar, const char *c)
379{
380 if (!bsdtar->warned_lead_slash) {
381 lafe_warnc(0,
382 "Removing leading '%c' from member names",
383 c[0]);
384 bsdtar->warned_lead_slash = 1;
385 }
386}
387
388static void
389warn_strip_drive_letter(struct bsdtar *bsdtar)
390{
391 if (!bsdtar->warned_lead_slash) {
392 lafe_warnc(0,
393 "Removing leading drive letter from "
394 "member names");
395 bsdtar->warned_lead_slash = 1;
396 }
397}
398
399/*
400 * Convert absolute path to non-absolute path by skipping leading
401 * absolute path prefixes.
402 */
403static const char*
404strip_absolute_path(struct bsdtar *bsdtar, const char *p)
405{
406 const char *rp;
407
408 /* Remove leading "//./" or "//?/" or "//?/UNC/"
409 * (absolute path prefixes used by Windows API) */
410 if ((p[0] == '/' || p[0] == '\\') &&
411 (p[1] == '/' || p[1] == '\\') &&
412 (p[2] == '.' || p[2] == '?') &&
413 (p[3] == '/' || p[3] == '\\'))
414 {
415 if (p[2] == '?' &&
416 (p[4] == 'U' || p[4] == 'u') &&
417 (p[5] == 'N' || p[5] == 'n') &&
418 (p[6] == 'C' || p[6] == 'c') &&
419 (p[7] == '/' || p[7] == '\\'))
420 p += 8;
421 else
422 p += 4;
423 warn_strip_drive_letter(bsdtar);
424 }
425
426 /* Remove multiple leading slashes and Windows drive letters. */
427 do {
428 rp = p;
429 if (((p[0] >= 'a' && p[0] <= 'z') ||
430 (p[0] >= 'A' && p[0] <= 'Z')) &&
431 p[1] == ':') {
432 p += 2;
433 warn_strip_drive_letter(bsdtar);
434 }
435
436 /* Remove leading "/../", "/./", "//", etc. */
437 while (p[0] == '/' || p[0] == '\\') {
438 if (p[1] == '.' &&
439 p[2] == '.' &&
440 (p[3] == '/' || p[3] == '\\')) {
441 p += 3; /* Remove "/..", leave "/" for next pass. */
442 } else if (p[1] == '.' &&
443 (p[2] == '/' || p[2] == '\\')) {
444 p += 2; /* Remove "/.", leave "/" for next pass. */
445 } else
446 p += 1; /* Remove "/". */
447 warn_strip_leading_char(bsdtar, rp);
448 }
449 } while (rp != p);
450
451 return (p);
452}
453
454/*
455 * Handle --strip-components and any future path-rewriting options.
456 * Returns non-zero if the pathname should not be extracted.
457 *
458 * Note: The rewrites are applied uniformly to pathnames and hardlink
459 * names but not to symlink bodies. This is deliberate: Symlink
460 * bodies are not necessarily filenames. Even when they are, they
461 * need to be interpreted relative to the directory containing them,
462 * so simple rewrites like this are rarely appropriate.
463 *
464 * TODO: Support pax-style regex path rewrites.
465 */
466int
467edit_pathname(struct bsdtar *bsdtar, struct archive_entry *entry)
468{
469 const char *name = archive_entry_pathname(entry);
470 const char *original_name = name;
471 const char *hardlinkname = archive_entry_hardlink(entry);
472 const char *original_hardlinkname = hardlinkname;
473#if defined(HAVE_REGEX_H) || defined(HAVE_PCREPOSIX_H)
474 char *subst_name;
475 int r;
476
477 /* Apply user-specified substitution to pathname. */
478 r = apply_substitution(bsdtar, name, &subst_name, 0, 0);
479 if (r == -1) {
480 lafe_warnc(0, "Invalid substitution, skipping entry");
481 return 1;
482 }
483 if (r == 1) {
484 archive_entry_copy_pathname(entry, subst_name);
485 if (*subst_name == '\0') {
486 free(subst_name);
487 return -1;
488 } else
489 free(subst_name);
490 name = archive_entry_pathname(entry);
491 original_name = name;
492 }
493
494 /* Apply user-specified substitution to hardlink target. */
495 if (hardlinkname != NULL) {
496 r = apply_substitution(bsdtar, hardlinkname, &subst_name, 0, 1);
497 if (r == -1) {
498 lafe_warnc(0, "Invalid substitution, skipping entry");
499 return 1;
500 }
501 if (r == 1) {
502 archive_entry_copy_hardlink(entry, subst_name);
503 free(subst_name);
504 }
505 hardlinkname = archive_entry_hardlink(entry);
506 original_hardlinkname = hardlinkname;
507 }
508
509 /* Apply user-specified substitution to symlink body. */
510 if (archive_entry_symlink(entry) != NULL) {
511 r = apply_substitution(bsdtar, archive_entry_symlink(entry), &subst_name, 1, 0);
512 if (r == -1) {
513 lafe_warnc(0, "Invalid substitution, skipping entry");
514 return 1;
515 }
516 if (r == 1) {
517 archive_entry_copy_symlink(entry, subst_name);
518 free(subst_name);
519 }
520 }
521#endif
522
523 /* Strip leading dir names as per --strip-components option. */
524 if (bsdtar->strip_components > 0) {
525 name = strip_components(name, bsdtar->strip_components);
526 if (name == NULL)
527 return (1);
528
529 if (hardlinkname != NULL) {
530 hardlinkname = strip_components(hardlinkname,
531 bsdtar->strip_components);
532 if (hardlinkname == NULL)
533 return (1);
534 }
535 }
536
537 if ((bsdtar->flags & OPTFLAG_ABSOLUTE_PATHS) == 0) {
538 /* By default, don't write or restore absolute pathnames. */
539 name = strip_absolute_path(bsdtar, name);
540 if (*name == '\0')
541 name = ".";
542
543 if (hardlinkname != NULL) {
544 hardlinkname = strip_absolute_path(bsdtar, hardlinkname);
545 if (*hardlinkname == '\0')
546 return (1);
547 }
548 } else {
549 /* Strip redundant leading '/' characters. */
550 while (name[0] == '/' && name[1] == '/')
551 name++;
552 }
553
554 /* Replace name in archive_entry. */
555 if (name != original_name) {
556 archive_entry_copy_pathname(entry, name);
557 }
558 if (hardlinkname != original_hardlinkname) {
559 archive_entry_copy_hardlink(entry, hardlinkname);
560 }
561 return (0);
562}
563
564/*
565 * It would be nice to just use printf() for formatting large numbers,
566 * but the compatibility problems are quite a headache. Hence the
567 * following simple utility function.
568 */
569const char *
570tar_i64toa(int64_t n0)
571{
572 static char buff[24];
573 uint64_t n = n0 < 0 ? -n0 : n0;
574 char *p = buff + sizeof(buff);
575
576 *--p = '\0';
577 do {
578 *--p = '0' + (int)(n % 10);
579 } while (n /= 10);
580 if (n0 < 0)
581 *--p = '-';
582 return p;
583}
584
585/*
586 * Like strcmp(), but try to be a little more aware of the fact that
587 * we're comparing two paths. Right now, it just handles leading
588 * "./" and trailing '/' specially, so that "a/b/" == "./a/b"
589 *
590 * TODO: Make this better, so that "./a//b/./c/" == "a/b/c"
591 * TODO: After this works, push it down into libarchive.
592 * TODO: Publish the path normalization routines in libarchive so
593 * that bsdtar can normalize paths and use fast strcmp() instead
594 * of this.
595 *
596 * Note: This is currently only used within write.c, so should
597 * not handle \ path separators.
598 */
599
600int
601pathcmp(const char *a, const char *b)
602{
603 /* Skip leading './' */
604 if (a[0] == '.' && a[1] == '/' && a[2] != '\0')
605 a += 2;
606 if (b[0] == '.' && b[1] == '/' && b[2] != '\0')
607 b += 2;
608 /* Find the first difference, or return (0) if none. */
609 while (*a == *b) {
610 if (*a == '\0')
611 return (0);
612 a++;
613 b++;
614 }
615 /*
616 * If one ends in '/' and the other one doesn't,
617 * they're the same.
618 */
619 if (a[0] == '/' && a[1] == '\0' && b[0] == '\0')
620 return (0);
621 if (a[0] == '\0' && b[0] == '/' && b[1] == '\0')
622 return (0);
623 /* They're really different, return the correct sign. */
624 return (*(const unsigned char *)a - *(const unsigned char *)b);
625}
626
627#define PPBUFF_SIZE 1024
628const char *
629passphrase_callback(struct archive *a, void *_client_data)
630{
631 struct bsdtar *bsdtar = (struct bsdtar *)_client_data;
632 (void)a; /* UNUSED */
633
634 if (bsdtar->ppbuff == NULL) {
635 bsdtar->ppbuff = malloc(PPBUFF_SIZE);
636 if (bsdtar->ppbuff == NULL)
637 lafe_errc(1, errno, "Out of memory");
638 }
639 return lafe_readpassphrase("Enter passphrase:",
640 bsdtar->ppbuff, PPBUFF_SIZE);
641}
642
643void
644passphrase_free(char *ppbuff)
645{
646 if (ppbuff != NULL) {
647 memset(ppbuff, 0, PPBUFF_SIZE);
648 free(ppbuff);
649 }
650}
651
652/*
653 * Display information about the current file.
654 *
655 * The format here roughly duplicates the output of 'ls -l'.
656 * This is based on SUSv2, where 'tar tv' is documented as
657 * listing additional information in an "unspecified format,"
658 * and 'pax -l' is documented as using the same format as 'ls -l'.
659 */
660void
661list_item_verbose(struct bsdtar *bsdtar, FILE *out, struct archive_entry *entry)
662{
663 char tmp[100];
664 size_t w;
665 const char *p;
666 const char *fmt;
667 time_t tim;
668 static time_t now;
|
669
670 /*
671 * We avoid collecting the entire list in memory at once by
672 * listing things as we see them. However, that also means we can't
673 * just pre-compute the field widths. Instead, we start with guesses
674 * and just widen them as necessary. These numbers are completely
675 * arbitrary.
676 */
677 if (!bsdtar->u_width) {
678 bsdtar->u_width = 6;
679 bsdtar->gs_width = 13;
680 }
681 if (!now)
682 time(&now);
683 fprintf(out, "%s %d ",
684 archive_entry_strmode(entry),
685 archive_entry_nlink(entry));
686
687 /* Use uname if it's present, else uid. */
688 p = archive_entry_uname(entry);
689 if ((p == NULL) || (*p == '\0')) {
690 sprintf(tmp, "%lu ",
691 (unsigned long)archive_entry_uid(entry));
692 p = tmp;
693 }
694 w = strlen(p);
695 if (w > bsdtar->u_width)
696 bsdtar->u_width = w;
697 fprintf(out, "%-*s ", (int)bsdtar->u_width, p);
698
699 /* Use gname if it's present, else gid. */
700 p = archive_entry_gname(entry);
701 if (p != NULL && p[0] != '\0') {
702 fprintf(out, "%s", p);
703 w = strlen(p);
704 } else {
705 sprintf(tmp, "%lu",
706 (unsigned long)archive_entry_gid(entry));
707 w = strlen(tmp);
708 fprintf(out, "%s", tmp);
709 }
710
711 /*
712 * Print device number or file size, right-aligned so as to make
713 * total width of group and devnum/filesize fields be gs_width.
714 * If gs_width is too small, grow it.
715 */
716 if (archive_entry_filetype(entry) == AE_IFCHR
717 || archive_entry_filetype(entry) == AE_IFBLK) {
718 sprintf(tmp, "%lu,%lu",
719 (unsigned long)archive_entry_rdevmajor(entry),
720 (unsigned long)archive_entry_rdevminor(entry));
721 } else {
722 strcpy(tmp, tar_i64toa(archive_entry_size(entry)));
723 }
724 if (w + strlen(tmp) >= bsdtar->gs_width)
725 bsdtar->gs_width = w+strlen(tmp)+1;
726 fprintf(out, "%*s", (int)(bsdtar->gs_width - w), tmp);
727
728 /* Format the time using 'ls -l' conventions. */
729 tim = archive_entry_mtime(entry);
730#define HALF_YEAR (time_t)365 * 86400 / 2
731#if defined(_WIN32) && !defined(__CYGWIN__)
732#define DAY_FMT "%d" /* Windows' strftime function does not support %e format. */
733#else
734#define DAY_FMT "%e" /* Day number without leading zeros */
735#endif
736 if (tim < now - HALF_YEAR || tim > now + HALF_YEAR)
737 fmt = bsdtar->day_first ? DAY_FMT " %b %Y" : "%b " DAY_FMT " %Y";
738 else
739 fmt = bsdtar->day_first ? DAY_FMT " %b %H:%M" : "%b " DAY_FMT " %H:%M";
| 677
678 /*
679 * We avoid collecting the entire list in memory at once by
680 * listing things as we see them. However, that also means we can't
681 * just pre-compute the field widths. Instead, we start with guesses
682 * and just widen them as necessary. These numbers are completely
683 * arbitrary.
684 */
685 if (!bsdtar->u_width) {
686 bsdtar->u_width = 6;
687 bsdtar->gs_width = 13;
688 }
689 if (!now)
690 time(&now);
691 fprintf(out, "%s %d ",
692 archive_entry_strmode(entry),
693 archive_entry_nlink(entry));
694
695 /* Use uname if it's present, else uid. */
696 p = archive_entry_uname(entry);
697 if ((p == NULL) || (*p == '\0')) {
698 sprintf(tmp, "%lu ",
699 (unsigned long)archive_entry_uid(entry));
700 p = tmp;
701 }
702 w = strlen(p);
703 if (w > bsdtar->u_width)
704 bsdtar->u_width = w;
705 fprintf(out, "%-*s ", (int)bsdtar->u_width, p);
706
707 /* Use gname if it's present, else gid. */
708 p = archive_entry_gname(entry);
709 if (p != NULL && p[0] != '\0') {
710 fprintf(out, "%s", p);
711 w = strlen(p);
712 } else {
713 sprintf(tmp, "%lu",
714 (unsigned long)archive_entry_gid(entry));
715 w = strlen(tmp);
716 fprintf(out, "%s", tmp);
717 }
718
719 /*
720 * Print device number or file size, right-aligned so as to make
721 * total width of group and devnum/filesize fields be gs_width.
722 * If gs_width is too small, grow it.
723 */
724 if (archive_entry_filetype(entry) == AE_IFCHR
725 || archive_entry_filetype(entry) == AE_IFBLK) {
726 sprintf(tmp, "%lu,%lu",
727 (unsigned long)archive_entry_rdevmajor(entry),
728 (unsigned long)archive_entry_rdevminor(entry));
729 } else {
730 strcpy(tmp, tar_i64toa(archive_entry_size(entry)));
731 }
732 if (w + strlen(tmp) >= bsdtar->gs_width)
733 bsdtar->gs_width = w+strlen(tmp)+1;
734 fprintf(out, "%*s", (int)(bsdtar->gs_width - w), tmp);
735
736 /* Format the time using 'ls -l' conventions. */
737 tim = archive_entry_mtime(entry);
738#define HALF_YEAR (time_t)365 * 86400 / 2
739#if defined(_WIN32) && !defined(__CYGWIN__)
740#define DAY_FMT "%d" /* Windows' strftime function does not support %e format. */
741#else
742#define DAY_FMT "%e" /* Day number without leading zeros */
743#endif
744 if (tim < now - HALF_YEAR || tim > now + HALF_YEAR)
745 fmt = bsdtar->day_first ? DAY_FMT " %b %Y" : "%b " DAY_FMT " %Y";
746 else
747 fmt = bsdtar->day_first ? DAY_FMT " %b %H:%M" : "%b " DAY_FMT " %H:%M";
|