More -Wwrite-strings cleanup and make sure you can actually play it.
[dragonfly.git] / contrib / libarchive / archive_write_set_format_pax.c
blobbc0c8f5325c59e48afc9316839d8048ffbd12ac4
1 /*-
2 * Copyright (c) 2003-2004 Tim Kientzle
3 * All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer
10 * in this position and unchanged.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
18 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 #include "archive_platform.h"
28 __FBSDID("$FreeBSD: src/lib/libarchive/archive_write_set_format_pax.c,v 1.26 2005/04/23 17:46:51 kientzle Exp $");
30 #include <sys/stat.h>
31 #include <errno.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <unistd.h>
35 #include <wchar.h>
37 #include "archive.h"
38 #include "archive_entry.h"
39 #include "archive_private.h"
41 struct pax {
42 uint64_t entry_bytes_remaining;
43 uint64_t entry_padding;
44 struct archive_string pax_header;
45 char written;
48 static void add_pax_attr(struct archive_string *, const char *key,
49 const char *value);
50 static void add_pax_attr_int(struct archive_string *,
51 const char *key, int64_t value);
52 static void add_pax_attr_time(struct archive_string *,
53 const char *key, int64_t sec,
54 unsigned long nanos);
55 static void add_pax_attr_w(struct archive_string *,
56 const char *key, const wchar_t *wvalue);
57 static int archive_write_pax_data(struct archive *,
58 const void *, size_t);
59 static int archive_write_pax_finish(struct archive *);
60 static int archive_write_pax_finish_entry(struct archive *);
61 static int archive_write_pax_header(struct archive *,
62 struct archive_entry *);
63 static char *build_pax_attribute_name(char *dest, const char *src);
64 static char *build_ustar_entry_name(char *dest, const char *src, const char *insert);
65 static char *format_int(char *dest, int64_t);
66 static int write_nulls(struct archive *, size_t);
69 * Set output format to 'restricted pax' format.
71 * This is the same as normal 'pax', but tries to suppress
72 * the pax header whenever possible. This is the default for
73 * bsdtar, for instance.
75 int
76 archive_write_set_format_pax_restricted(struct archive *a)
78 int r;
79 r = archive_write_set_format_pax(a);
80 a->archive_format = ARCHIVE_FORMAT_TAR_PAX_RESTRICTED;
81 a->archive_format_name = "restricted POSIX pax interchange";
82 return (r);
86 * Set output format to 'pax' format.
88 int
89 archive_write_set_format_pax(struct archive *a)
91 struct pax *pax;
93 if (a->format_finish != NULL)
94 (a->format_finish)(a);
96 pax = malloc(sizeof(*pax));
97 if (pax == NULL) {
98 archive_set_error(a, ENOMEM, "Can't allocate pax data");
99 return (ARCHIVE_FATAL);
101 memset(pax, 0, sizeof(*pax));
102 a->format_data = pax;
104 a->pad_uncompressed = 1;
105 a->format_write_header = archive_write_pax_header;
106 a->format_write_data = archive_write_pax_data;
107 a->format_finish = archive_write_pax_finish;
108 a->format_finish_entry = archive_write_pax_finish_entry;
109 a->archive_format = ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE;
110 a->archive_format_name = "POSIX pax interchange";
111 return (ARCHIVE_OK);
115 * Note: This code assumes that 'nanos' has the same sign as 'sec',
116 * which implies that sec=-1, nanos=200000000 represents -1.2 seconds
117 * and not -0.8 seconds. This is a pretty pedantic point, as we're
118 * unlikely to encounter many real files created before Jan 1, 1970,
119 * much less ones with timestamps recorded to sub-second resolution.
121 static void
122 add_pax_attr_time(struct archive_string *as, const char *key,
123 int64_t sec, unsigned long nanos)
125 int digit, i;
126 char *t;
128 * Note that each byte contributes fewer than 3 base-10
129 * digits, so this will always be big enough.
131 char tmp[1 + 3*sizeof(sec) + 1 + 3*sizeof(nanos)];
133 tmp[sizeof(tmp) - 1] = 0;
134 t = tmp + sizeof(tmp) - 1;
136 /* Skip trailing zeros in the fractional part. */
137 for(digit = 0, i = 10; i > 0 && digit == 0; i--) {
138 digit = nanos % 10;
139 nanos /= 10;
142 /* Only format the fraction if it's non-zero. */
143 if (i > 0) {
144 while (i > 0) {
145 *--t = "0123456789"[digit];
146 digit = nanos % 10;
147 nanos /= 10;
148 i--;
150 *--t = '.';
152 t = format_int(t, sec);
154 add_pax_attr(as, key, t);
157 static char *
158 format_int(char *t, int64_t i)
160 int sign;
162 if (i < 0) {
163 sign = -1;
164 i = -i;
165 } else
166 sign = 1;
168 do {
169 *--t = "0123456789"[i % 10];
170 } while (i /= 10);
171 if (sign < 0)
172 *--t = '-';
173 return (t);
176 static void
177 add_pax_attr_int(struct archive_string *as, const char *key, int64_t value)
179 char tmp[1 + 3 * sizeof(value)];
181 tmp[sizeof(tmp) - 1] = 0;
182 add_pax_attr(as, key, format_int(tmp + sizeof(tmp) - 1, value));
185 static void
186 add_pax_attr_w(struct archive_string *as, const char *key, const wchar_t *wval)
188 int utf8len;
189 const wchar_t *wp;
190 unsigned long wc;
191 char *utf8_value, *p;
193 utf8len = 0;
194 for (wp = wval; *wp != L'\0'; ) {
195 wc = *wp++;
196 if (wc <= 0x7f)
197 utf8len++;
198 else if (wc <= 0x7ff)
199 utf8len += 2;
200 else if (wc <= 0xffff)
201 utf8len += 3;
202 else if (wc <= 0x1fffff)
203 utf8len += 4;
204 else if (wc <= 0x3ffffff)
205 utf8len += 5;
206 else if (wc <= 0x7fffffff)
207 utf8len += 6;
208 /* Ignore larger values; UTF-8 can't encode them. */
211 utf8_value = malloc(utf8len + 1);
212 for (wp = wval, p = utf8_value; *wp != L'\0'; ) {
213 wc = *wp++;
214 if (wc <= 0x7f) {
215 *p++ = (char)wc;
216 } else if (wc <= 0x7ff) {
217 p[0] = 0xc0 | ((wc >> 6) & 0x1f);
218 p[1] = 0x80 | (wc & 0x3f);
219 p += 2;
220 } else if (wc <= 0xffff) {
221 p[0] = 0xe0 | ((wc >> 12) & 0x0f);
222 p[1] = 0x80 | ((wc >> 6) & 0x3f);
223 p[2] = 0x80 | (wc & 0x3f);
224 p += 3;
225 } else if (wc <= 0x1fffff) {
226 p[0] = 0xf0 | ((wc >> 18) & 0x07);
227 p[1] = 0x80 | ((wc >> 12) & 0x3f);
228 p[2] = 0x80 | ((wc >> 6) & 0x3f);
229 p[3] = 0x80 | (wc & 0x3f);
230 p += 4;
231 } else if (wc <= 0x3ffffff) {
232 p[0] = 0xf8 | ((wc >> 24) & 0x03);
233 p[1] = 0x80 | ((wc >> 18) & 0x3f);
234 p[2] = 0x80 | ((wc >> 12) & 0x3f);
235 p[3] = 0x80 | ((wc >> 6) & 0x3f);
236 p[4] = 0x80 | (wc & 0x3f);
237 p += 5;
238 } else if (wc <= 0x7fffffff) {
239 p[0] = 0xfc | ((wc >> 30) & 0x01);
240 p[1] = 0x80 | ((wc >> 24) & 0x3f);
241 p[1] = 0x80 | ((wc >> 18) & 0x3f);
242 p[2] = 0x80 | ((wc >> 12) & 0x3f);
243 p[3] = 0x80 | ((wc >> 6) & 0x3f);
244 p[4] = 0x80 | (wc & 0x3f);
245 p += 6;
247 /* Ignore larger values; UTF-8 can't encode them. */
249 *p = '\0';
250 add_pax_attr(as, key, utf8_value);
251 free(utf8_value);
255 * Add a key/value attribute to the pax header. This function handles
256 * the length field and various other syntactic requirements.
258 static void
259 add_pax_attr(struct archive_string *as, const char *key, const char *value)
261 int digits, i, len, next_ten;
262 char tmp[1 + 3 * sizeof(int)]; /* < 3 base-10 digits per byte */
265 * PAX attributes have the following layout:
266 * <len> <space> <key> <=> <value> <nl>
268 len = 1 + strlen(key) + 1 + strlen(value) + 1;
271 * The <len> field includes the length of the <len> field, so
272 * computing the correct length is tricky. I start by
273 * counting the number of base-10 digits in 'len' and
274 * computing the next higher power of 10.
276 next_ten = 1;
277 digits = 0;
278 i = len;
279 while (i > 0) {
280 i = i / 10;
281 digits++;
282 next_ten = next_ten * 10;
285 * For example, if string without the length field is 99
286 * chars, then adding the 2 digit length "99" will force the
287 * total length past 100, requiring an extra digit. The next
288 * statement adjusts for this effect.
290 if (len + digits >= next_ten)
291 digits++;
293 /* Now, we have the right length so we can build the line. */
294 tmp[sizeof(tmp) - 1] = 0; /* Null-terminate the work area. */
295 archive_strcat(as, format_int(tmp + sizeof(tmp) - 1, len + digits));
296 archive_strappend_char(as, ' ');
297 archive_strcat(as, key);
298 archive_strappend_char(as, '=');
299 archive_strcat(as, value);
300 archive_strappend_char(as, '\n');
304 * TODO: Consider adding 'comment' and 'charset' fields to
305 * archive_entry so that clients can specify them. Also, consider
306 * adding generic key/value tags so clients can add arbitrary
307 * key/value data.
309 static int
310 archive_write_pax_header(struct archive *a,
311 struct archive_entry *entry_original)
313 struct archive_entry *entry_main;
314 const char *linkname, *p;
315 const char *hardlink;
316 const wchar_t *wp, *wp2;
317 const char *suffix_start;
318 int need_extension, r, ret;
319 struct pax *pax;
320 const struct stat *st_main, *st_original;
322 char paxbuff[512];
323 char ustarbuff[512];
324 char ustar_entry_name[256];
325 char pax_entry_name[256];
327 need_extension = 0;
328 pax = a->format_data;
329 pax->written = 1;
331 st_original = archive_entry_stat(entry_original);
333 hardlink = archive_entry_hardlink(entry_original);
335 /* Make sure this is a type of entry that we can handle here */
336 if (hardlink == NULL) {
337 switch (st_original->st_mode & S_IFMT) {
338 case S_IFREG:
339 case S_IFLNK:
340 case S_IFCHR:
341 case S_IFBLK:
342 case S_IFDIR:
343 case S_IFIFO:
344 break;
345 case S_IFSOCK:
346 archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
347 "tar format cannot archive socket");
348 return (ARCHIVE_WARN);
349 default:
350 archive_set_error(a, ARCHIVE_ERRNO_FILE_FORMAT,
351 "tar format cannot archive this (mode=0%lo)",
352 (unsigned long)st_original->st_mode);
353 return (ARCHIVE_WARN);
357 /* Copy entry so we can modify it as needed. */
358 entry_main = archive_entry_clone(entry_original);
359 archive_string_empty(&(pax->pax_header)); /* Blank our work area. */
360 st_main = archive_entry_stat(entry_main);
363 * Determining whether or not the name is too big is ugly
364 * because of the rules for dividing names between 'name' and
365 * 'prefix' fields. Here, I pick out the longest possible
366 * suffix, then test whether the remaining prefix is too long.
368 wp = archive_entry_pathname_w(entry_main);
369 p = archive_entry_pathname(entry_main);
370 if (strlen(p) <= 100) /* Short enough for just 'name' field */
371 suffix_start = p; /* Record a zero-length prefix */
372 else
373 /* Find the largest suffix that fits in 'name' field. */
374 suffix_start = strchr(p + strlen(p) - 100 - 1, '/');
376 /* Find non-ASCII character, if any. */
377 wp2 = wp;
378 while (*wp2 != L'\0' && *wp2 < 128)
379 wp2++;
382 * If name is too long, or has non-ASCII characters, add
383 * 'path' to pax extended attrs.
385 if (suffix_start == NULL || suffix_start - p > 155 || *wp2 != L'\0') {
386 add_pax_attr_w(&(pax->pax_header), "path", wp);
387 archive_entry_set_pathname(entry_main,
388 build_ustar_entry_name(ustar_entry_name, p, NULL));
389 need_extension = 1;
392 /* If link name is too long, add 'linkpath' to pax extended attrs. */
393 linkname = hardlink;
394 if (linkname == NULL)
395 linkname = archive_entry_symlink(entry_main);
397 if (linkname != NULL && strlen(linkname) > 100) {
398 add_pax_attr(&(pax->pax_header), "linkpath", linkname);
399 if (hardlink != NULL)
400 archive_entry_set_hardlink(entry_main,
401 "././@LongHardLink");
402 else
403 archive_entry_set_symlink(entry_main,
404 "././@LongSymLink");
405 need_extension = 1;
408 /* If file size is too large, add 'size' to pax extended attrs. */
409 if (st_main->st_size >= (((int64_t)1) << 33)) {
410 add_pax_attr_int(&(pax->pax_header), "size", st_main->st_size);
411 need_extension = 1;
414 /* If numeric GID is too large, add 'gid' to pax extended attrs. */
415 if (st_main->st_gid >= (1 << 18)) {
416 add_pax_attr_int(&(pax->pax_header), "gid", st_main->st_gid);
417 need_extension = 1;
420 /* If group name is too large, add 'gname' to pax extended attrs. */
421 /* TODO: If gname has non-ASCII characters, use pax attribute. */
422 p = archive_entry_gname(entry_main);
423 if (p != NULL && strlen(p) > 31) {
424 add_pax_attr(&(pax->pax_header), "gname", p);
425 archive_entry_set_gname(entry_main, NULL);
426 need_extension = 1;
429 /* If numeric UID is too large, add 'uid' to pax extended attrs. */
430 if (st_main->st_uid >= (1 << 18)) {
431 add_pax_attr_int(&(pax->pax_header), "uid", st_main->st_uid);
432 need_extension = 1;
435 /* If user name is too large, add 'uname' to pax extended attrs. */
436 /* TODO: If uname has non-ASCII characters, use pax attribute. */
437 p = archive_entry_uname(entry_main);
438 if (p != NULL && strlen(p) > 31) {
439 add_pax_attr(&(pax->pax_header), "uname", p);
440 archive_entry_set_uname(entry_main, NULL);
441 need_extension = 1;
445 * POSIX/SUSv3 doesn't provide a standard key for large device
446 * numbers. I use the same keys here that Joerg Schilling
447 * used for 'star.' (Which, somewhat confusingly, are called
448 * "devXXX" even though they code "rdev" values.) No doubt,
449 * other implementations use other keys. Note that there's no
450 * reason we can't write the same information into a number of
451 * different keys.
453 * Of course, this is only needed for block or char device entries.
455 if (S_ISBLK(st_main->st_mode) ||
456 S_ISCHR(st_main->st_mode)) {
458 * If rdevmajor is too large, add 'SCHILY.devmajor' to
459 * extended attributes.
461 dev_t rdevmajor, rdevminor;
462 rdevmajor = major(st_main->st_rdev);
463 rdevminor = minor(st_main->st_rdev);
464 if (rdevmajor >= (1 << 18)) {
465 add_pax_attr_int(&(pax->pax_header), "SCHILY.devmajor",
466 rdevmajor);
468 * Non-strict formatting below means we don't
469 * have to truncate here. Not truncating improves
470 * the chance that some more modern tar archivers
471 * (such as GNU tar 1.13) can restore the full
472 * value even if they don't understand the pax
473 * extended attributes. See my rant below about
474 * file size fields for additional details.
476 /* archive_entry_set_rdevmajor(entry_main,
477 rdevmajor & ((1 << 18) - 1)); */
478 need_extension = 1;
482 * If devminor is too large, add 'SCHILY.devminor' to
483 * extended attributes.
485 if (rdevminor >= (1 << 18)) {
486 add_pax_attr_int(&(pax->pax_header), "SCHILY.devminor",
487 rdevminor);
488 /* Truncation is not necessary here, either. */
489 /* archive_entry_set_rdevminor(entry_main,
490 rdevminor & ((1 << 18) - 1)); */
491 need_extension = 1;
496 * Technically, the mtime field in the ustar header can
497 * support 33 bits, but many platforms use signed 32-bit time
498 * values. The cutoff of 0x7fffffff here is a compromise.
499 * Yes, this check is duplicated just below; this helps to
500 * avoid writing an mtime attribute just to handle a
501 * high-resolution timestamp in "restricted pax" mode.
503 if (!need_extension &&
504 ((st_main->st_mtime < 0) || (st_main->st_mtime >= 0x7fffffff)))
505 need_extension = 1;
507 /* I use a star-compatible file flag attribute. */
508 p = archive_entry_fflags_text(entry_main);
509 if (!need_extension && p != NULL && *p != '\0')
510 need_extension = 1;
512 /* If there are non-trivial ACL entries, we need an extension. */
513 if (!need_extension && archive_entry_acl_count(entry_original,
514 ARCHIVE_ENTRY_ACL_TYPE_ACCESS) > 0)
515 need_extension = 1;
517 /* If there are non-trivial ACL entries, we need an extension. */
518 if (!need_extension && archive_entry_acl_count(entry_original,
519 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT) > 0)
520 need_extension = 1;
523 * The following items are handled differently in "pax
524 * restricted" format. In particular, in "pax restricted"
525 * format they won't be added unless need_extension is
526 * already set (we're already generating an extended header, so
527 * may as well include these).
529 if (a->archive_format != ARCHIVE_FORMAT_TAR_PAX_RESTRICTED ||
530 need_extension) {
532 if (st_main->st_mtime < 0 ||
533 st_main->st_mtime >= 0x7fffffff ||
534 ARCHIVE_STAT_MTIME_NANOS(st_main) != 0)
535 add_pax_attr_time(&(pax->pax_header), "mtime",
536 st_main->st_mtime,
537 ARCHIVE_STAT_MTIME_NANOS(st_main));
539 if (st_main->st_ctime != 0 ||
540 ARCHIVE_STAT_CTIME_NANOS(st_main) != 0)
541 add_pax_attr_time(&(pax->pax_header), "ctime",
542 st_main->st_ctime,
543 ARCHIVE_STAT_CTIME_NANOS(st_main));
545 if (st_main->st_atime != 0 ||
546 ARCHIVE_STAT_ATIME_NANOS(st_main) != 0)
547 add_pax_attr_time(&(pax->pax_header), "atime",
548 st_main->st_atime,
549 ARCHIVE_STAT_ATIME_NANOS(st_main));
551 /* I use a star-compatible file flag attribute. */
552 p = archive_entry_fflags_text(entry_main);
553 if (p != NULL && *p != '\0')
554 add_pax_attr(&(pax->pax_header), "SCHILY.fflags", p);
556 /* I use star-compatible ACL attributes. */
557 wp = archive_entry_acl_text_w(entry_original,
558 ARCHIVE_ENTRY_ACL_TYPE_ACCESS |
559 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID);
560 if (wp != NULL && *wp != L'\0')
561 add_pax_attr_w(&(pax->pax_header),
562 "SCHILY.acl.access", wp);
563 wp = archive_entry_acl_text_w(entry_original,
564 ARCHIVE_ENTRY_ACL_TYPE_DEFAULT |
565 ARCHIVE_ENTRY_ACL_STYLE_EXTRA_ID);
566 if (wp != NULL && *wp != L'\0')
567 add_pax_attr_w(&(pax->pax_header),
568 "SCHILY.acl.default", wp);
570 /* Include star-compatible metadata info. */
571 /* Note: "SCHILY.dev{major,minor}" are NOT the
572 * major/minor portions of "SCHILY.dev". */
573 add_pax_attr_int(&(pax->pax_header), "SCHILY.dev",
574 st_main->st_dev);
575 add_pax_attr_int(&(pax->pax_header), "SCHILY.ino",
576 st_main->st_ino);
577 add_pax_attr_int(&(pax->pax_header), "SCHILY.nlink",
578 st_main->st_nlink);
581 /* Only regular files have data. */
582 if (!S_ISREG(archive_entry_mode(entry_main)))
583 archive_entry_set_size(entry_main, 0);
586 * Pax-restricted does not store data for hardlinks, in order
587 * to improve compatibility with ustar.
589 if (a->archive_format != ARCHIVE_FORMAT_TAR_PAX_INTERCHANGE &&
590 hardlink != NULL)
591 archive_entry_set_size(entry_main, 0);
594 * XXX Full pax interchange format does permit a hardlink
595 * entry to have data associated with it. I'm not supporting
596 * that here because the client expects me to tell them whether
597 * or not this format expects data for hardlinks. If I
598 * don't check here, then every pax archive will end up with
599 * duplicated data for hardlinks. Someday, there may be
600 * need to select this behavior, in which case the following
601 * will need to be revisited. XXX
603 if (hardlink != NULL)
604 archive_entry_set_size(entry_main, 0);
606 /* Format 'ustar' header for main entry.
608 * The trouble with file size: If the reader can't understand
609 * the file size, they may not be able to locate the next
610 * entry and the rest of the archive is toast. Pax-compliant
611 * readers are supposed to ignore the file size in the main
612 * header, so the question becomes how to maximize portability
613 * for readers that don't support pax attribute extensions.
614 * For maximum compatibility, I permit numeric extensions in
615 * the main header so that the file size stored will always be
616 * correct, even if it's in a format that only some
617 * implementations understand. The technique used here is:
619 * a) If possible, follow the standard exactly. This handles
620 * files up to 8 gigabytes minus 1.
622 * b) If that fails, try octal but omit the field terminator.
623 * That handles files up to 64 gigabytes minus 1.
625 * c) Otherwise, use base-256 extensions. That handles files
626 * up to 2^63 in this implementation, with the potential to
627 * go up to 2^94. That should hold us for a while. ;-)
629 * The non-strict formatter uses similar logic for other
630 * numeric fields, though they're less critical.
632 __archive_write_format_header_ustar(a, ustarbuff, entry_main, -1, 0);
634 /* If we built any extended attributes, write that entry first. */
635 ret = ARCHIVE_OK;
636 if (archive_strlen(&(pax->pax_header)) > 0) {
637 struct stat st;
638 struct archive_entry *pax_attr_entry;
640 memset(&st, 0, sizeof(st));
641 pax_attr_entry = archive_entry_new();
642 p = archive_entry_pathname(entry_main);
643 archive_entry_set_pathname(pax_attr_entry,
644 build_pax_attribute_name(pax_entry_name, p));
645 st.st_size = archive_strlen(&(pax->pax_header));
646 st.st_uid = st_main->st_uid;
647 if (st.st_uid >= 1 << 18)
648 st.st_uid = (1 << 18) - 1;
649 st.st_gid = st_main->st_gid;
650 if (st.st_gid >= 1 << 18)
651 st.st_gid = (1 << 18) - 1;
652 st.st_mode = st_main->st_mode;
653 archive_entry_copy_stat(pax_attr_entry, &st);
655 archive_entry_set_uname(pax_attr_entry,
656 archive_entry_uname(entry_main));
657 archive_entry_set_gname(pax_attr_entry,
658 archive_entry_gname(entry_main));
660 ret = __archive_write_format_header_ustar(a, paxbuff,
661 pax_attr_entry, 'x', 1);
663 archive_entry_free(pax_attr_entry);
665 /* Note that the 'x' header shouldn't ever fail to format */
666 if (ret != 0) {
667 const char *msg = "archive_write_pax_header: "
668 "'x' header failed?! This can't happen.\n";
669 write(2, msg, strlen(msg));
670 exit(1);
672 r = (a->compression_write)(a, paxbuff, 512);
673 if (r != ARCHIVE_OK) {
674 pax->entry_bytes_remaining = 0;
675 pax->entry_padding = 0;
676 return (ARCHIVE_FATAL);
679 pax->entry_bytes_remaining = archive_strlen(&(pax->pax_header));
680 pax->entry_padding = 0x1ff & (- pax->entry_bytes_remaining);
682 r = (a->compression_write)(a, pax->pax_header.s,
683 archive_strlen(&(pax->pax_header)));
684 if (r != ARCHIVE_OK) {
685 /* If a write fails, we're pretty much toast. */
686 return (ARCHIVE_FATAL);
688 /* Pad out the end of the entry. */
689 r = write_nulls(a, pax->entry_padding);
690 if (r != ARCHIVE_OK) {
691 /* If a write fails, we're pretty much toast. */
692 return (ARCHIVE_FATAL);
694 pax->entry_bytes_remaining = pax->entry_padding = 0;
697 /* Write the header for main entry. */
698 r = (a->compression_write)(a, ustarbuff, 512);
699 if (r != ARCHIVE_OK)
700 return (r);
703 * Inform the client of the on-disk size we're using, so
704 * they can avoid unnecessarily writing a body for something
705 * that we're just going to ignore.
707 archive_entry_set_size(entry_original, archive_entry_size(entry_main));
708 pax->entry_bytes_remaining = archive_entry_size(entry_main);
709 pax->entry_padding = 0x1ff & (- pax->entry_bytes_remaining);
710 archive_entry_free(entry_main);
712 return (ret);
716 * We need a valid name for the regular 'ustar' entry. This routine
717 * tries to hack something more-or-less reasonable.
719 * The approach here tries to preserve leading dir names. We do so by
720 * working with four sections:
721 * 1) "prefix" directory names,
722 * 2) "suffix" directory names,
723 * 3) inserted dir name (optional),
724 * 4) filename.
726 * These sections must satisfy the following requirements:
727 * * Parts 1 & 2 together form an initial portion of the dir name.
728 * * Part 3 is specified by the caller. (It should not contain a leading
729 * or trailing '/'.)
730 * * Part 4 forms an initial portion of the base filename.
731 * * The filename must be <= 99 chars to fit the ustar 'name' field.
732 * * Parts 2, 3, 4 together must be <= 99 chars to fit the ustar 'name' fld.
733 * * Part 1 must be <= 155 chars to fit the ustar 'prefix' field.
734 * * If the original name ends in a '/', the new name must also end in a '/'
735 * * Trailing '/.' sequences may be stripped.
737 * Note: Recall that the ustar format does not store the '/' separating
738 * parts 1 & 2, but does store the '/' separating parts 2 & 3.
740 static char *
741 build_ustar_entry_name(char *dest, const char *src, const char *insert)
743 const char *prefix, *prefix_end;
744 const char *suffix, *suffix_end;
745 const char *filename, *filename_end;
746 char *p;
747 size_t s;
748 int need_slash = 0; /* Was there a trailing slash? */
749 size_t suffix_length = 99;
750 int insert_length;
752 /* Length of additional dir element to be added. */
753 if (insert == NULL)
754 insert_length = 0;
755 else
756 /* +2 here allows for '/' before and after the insert. */
757 insert_length = strlen(insert) + 2;
759 /* Step 0: Quick bailout in a common case. */
760 s = strlen(src);
761 if (s < 100 && insert == NULL) {
762 strcpy(dest, src);
763 return (dest);
766 /* Step 1: Locate filename and enforce the length restriction. */
767 filename_end = src + s;
768 /* Remove trailing '/' chars and '/.' pairs. */
769 for (;;) {
770 if (filename_end > src && filename_end[-1] == '/') {
771 filename_end --;
772 need_slash = 1; /* Remember to restore trailing '/'. */
773 continue;
775 if (filename_end > src + 1 && filename_end[-1] == '.'
776 && filename_end[-2] == '/') {
777 filename_end -= 2;
778 need_slash = 1; /* "foo/." will become "foo/" */
779 continue;
781 break;
783 if (need_slash)
784 suffix_length--;
785 /* Find start of filename. */
786 filename = filename_end - 1;
787 while ((filename > src) && (*filename != '/'))
788 filename --;
789 if ((*filename == '/') && (filename < filename_end - 1))
790 filename ++;
791 /* Adjust filename_end so that filename + insert fits in 99 chars. */
792 suffix_length -= insert_length;
793 if (filename_end > filename + suffix_length)
794 filename_end = filename + suffix_length;
795 /* Calculate max size for "suffix" section (#3 above). */
796 suffix_length -= filename_end - filename;
798 /* Step 2: Locate the "prefix" section of the dirname, including
799 * trailing '/'. */
800 prefix = src;
801 prefix_end = prefix + 155;
802 if (prefix_end > filename)
803 prefix_end = filename;
804 while (prefix_end > prefix && *prefix_end != '/')
805 prefix_end--;
806 if ((prefix_end < filename) && (*prefix_end == '/'))
807 prefix_end++;
809 /* Step 3: Locate the "suffix" section of the dirname,
810 * including trailing '/'. */
811 suffix = prefix_end;
812 suffix_end = suffix + suffix_length; /* Enforce limit. */
813 if (suffix_end > filename)
814 suffix_end = filename;
815 if (suffix_end < suffix)
816 suffix_end = suffix;
817 while (suffix_end > suffix && *suffix_end != '/')
818 suffix_end--;
819 if ((suffix_end < filename) && (*suffix_end == '/'))
820 suffix_end++;
822 /* Step 4: Build the new name. */
823 /* The OpenBSD strlcpy function is safer, but less portable. */
824 /* Rather than maintain two versions, just use the strncpy version. */
825 p = dest;
826 if (prefix_end > prefix) {
827 strncpy(p, prefix, prefix_end - prefix);
828 p += prefix_end - prefix;
830 if (suffix_end > suffix) {
831 strncpy(p, suffix, suffix_end - suffix);
832 p += suffix_end - suffix;
834 if (insert != NULL) {
835 if (prefix_end > prefix || suffix_end > suffix)
836 *p++ = '/';
837 strcpy(p, insert);
838 p += strlen(insert);
839 *p++ = '/';
841 strncpy(p, filename, filename_end - filename);
842 p += filename_end - filename;
843 if (need_slash)
844 *p++ = '/';
845 *p++ = '\0';
847 return (dest);
851 * The ustar header for the pax extended attributes must have a
852 * reasonable name: SUSv3 suggests 'dirname'/PaxHeader/'filename'
854 * Joerg Schiling has argued that this is unnecessary because, in practice,
855 * if the pax extended attributes get extracted as regular files, noone is
856 * going to bother reading those attributes to manually restore them.
857 * Based on this, 'star' uses /tmp/PaxHeader/'basename' as the ustar header
858 * name. This is a tempting argument, but I'm not entirely convinced.
859 * I'm also uncomfortable with the fact that "/tmp" is a Unix-ism.
861 * The following routine implements the SUSv3 recommendation, and is
862 * much simpler because build_ustar_entry_name() above already does
863 * most of the work (we just need to give it an extra path element to
864 * insert and handle a few pathological cases).
866 static char *
867 build_pax_attribute_name(char *dest, const char *src)
869 char *p;
871 /* Handle the null filename case. */
872 if (src == NULL || *src == '\0') {
873 strcpy(dest, "PaxHeader/blank");
874 return (dest);
877 /* Prune final '/' and other unwanted final elements. */
878 p = dest + strlen(dest);
879 for (;;) {
880 /* Ends in "/", remove the '/' */
881 if (p > dest && p[-1] == '/') {
882 *--p = '\0';
883 continue;
885 /* Ends in "/.", remove the '.' */
886 if (p > dest + 1 && p[-1] == '.'
887 && p[-2] == '/') {
888 *--p = '\0';
889 continue;
891 break;
894 /* Pathological case: After above, there was nothing left. */
895 if (p == dest) {
896 strcpy(dest, "/PaxHeader/rootdir");
897 return (dest);
900 /* Convert unadorned "." into "dot" */
901 if (*src == '.' && src[1] == '\0') {
902 strcpy(dest, "PaxHeader/currentdir");
903 return (dest);
906 /* General case: build a ustar-compatible name adding "/PaxHeader/". */
907 build_ustar_entry_name(dest, src, "PaxHeader");
909 return (dest);
912 /* Write two null blocks for the end of archive */
913 static int
914 archive_write_pax_finish(struct archive *a)
916 struct pax *pax;
917 int r;
919 r = ARCHIVE_OK;
920 pax = a->format_data;
921 if (pax->written && a->compression_write != NULL)
922 r = write_nulls(a, 512 * 2);
923 archive_string_free(&pax->pax_header);
924 free(pax);
925 a->format_data = NULL;
926 return (r);
929 static int
930 archive_write_pax_finish_entry(struct archive *a)
932 struct pax *pax;
933 int ret;
935 pax = a->format_data;
936 ret = write_nulls(a, pax->entry_bytes_remaining + pax->entry_padding);
937 pax->entry_bytes_remaining = pax->entry_padding = 0;
938 return (ret);
941 static int
942 write_nulls(struct archive *a, size_t padding)
944 int ret, to_write;
946 while (padding > 0) {
947 to_write = padding < a->null_length ? padding : a->null_length;
948 ret = (a->compression_write)(a, a->nulls, to_write);
949 if (ret != ARCHIVE_OK)
950 return (ret);
951 padding -= to_write;
953 return (ARCHIVE_OK);
956 static int
957 archive_write_pax_data(struct archive *a, const void *buff, size_t s)
959 struct pax *pax;
960 int ret;
962 pax = a->format_data;
963 pax->written = 1;
964 if (s > pax->entry_bytes_remaining)
965 s = pax->entry_bytes_remaining;
967 ret = (a->compression_write)(a, buff, s);
968 pax->entry_bytes_remaining -= s;
969 return (ret);