busybox/archival/libarchive/get_header_tar.c

455 lines
14 KiB
C
Raw Normal View History

/* vi: set sw=4 ts=4: */
/* Licensed under GPLv2 or later, see file LICENSE in this source tree.
*
* FIXME:
* In privileged mode if uname and gname map to a uid and gid then use the
* mapped value instead of the uid/gid values in tar header
*
* References:
* GNU tar and star man pages,
* Opengroup's ustar interchange format,
* http://www.opengroup.org/onlinepubs/007904975/utilities/pax.html
2001-10-25 19:48:08 +05:30
*/
#include "libbb.h"
#include "bb_archive.h"
2001-10-25 19:48:08 +05:30
typedef uint32_t aliased_uint32_t FIX_ALIASING;
typedef off_t aliased_off_t FIX_ALIASING;
/* NB: _DESTROYS_ str[len] character! */
static unsigned long long getOctal(char *str, int len)
{
unsigned long long v;
char *end;
/* NB: leading spaces are allowed. Using strtoull to handle that.
* The downside is that we accept e.g. "-123" too :(
*/
str[len] = '\0';
v = strtoull(str, &end, 8);
/* std: "Each numeric field is terminated by one or more
* <space> or NUL characters". We must support ' '! */
if (*end != '\0' && *end != ' ') {
int8_t first = str[0];
if (!(first & 0x80))
bb_error_msg_and_die("corrupted octal value in tar header");
/*
* GNU tar uses "base-256 encoding" for very large numbers.
* Encoding is binary, with highest bit always set as a marker
* and sign in next-highest bit:
* 80 00 .. 00 - zero
* bf ff .. ff - largest positive number
* ff ff .. ff - minus 1
* c0 00 .. 00 - smallest negative number
*
* Example of tar file with 8914993153 (0x213600001) byte file.
* Field starts at offset 7c:
* 00070 30 30 30 00 30 30 30 30 30 30 30 00 80 00 00 00 |000.0000000.....|
* 00080 00 00 00 02 13 60 00 01 31 31 31 32 30 33 33 36 |.....`..11120336|
*
* NB: tarballs with NEGATIVE unix times encoded that way were seen!
*/
/* Sign-extend 7bit 'first' to 64bit 'v' (that is, using 6th bit as sign): */
first <<= 1;
first >>= 1; /* now 7th bit = 6th bit */
v = first; /* sign-extend 8 bits to 64 */
while (--len != 0)
v = (v << 8) + (uint8_t) *++str;
}
return v;
}
#define GET_OCTAL(a) getOctal((a), sizeof(a))
/* "global" is 0 or 1 */
static void process_pax_hdr(archive_handle_t *archive_handle, unsigned sz, int global)
{
char *buf, *p;
unsigned blk_sz;
blk_sz = (sz + 511) & (~511);
p = buf = xmalloc(blk_sz + 1);
xread(archive_handle->src_fd, buf, blk_sz);
archive_handle->offset += blk_sz;
/* prevent bb_strtou from running off the buffer */
buf[sz] = '\0';
while (sz != 0) {
char *end, *value;
unsigned len;
/* Every record has this format: "LEN NAME=VALUE\n" */
len = bb_strtou(p, &end, 10);
/* expect errno to be EINVAL, because the character
* following the digits should be a space
*/
p += len;
sz -= len;
if (
/** (int)sz < 0 - not good enough for huge malicious VALUE of 2^32-1 */
(int)(sz|len) < 0 /* this works */
|| len == 0
|| errno != EINVAL
|| *end != ' '
) {
bb_error_msg("malformed extended header, skipped");
// More verbose version:
//bb_error_msg("malformed extended header at %"OFF_FMT"d, skipped",
// archive_handle->offset - (sz + len));
break;
}
/* overwrite the terminating newline with NUL
* (we do not bother to check that it *was* a newline)
*/
p[-1] = '\0';
value = end + 1;
#if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
libbb: introduce and use is_prefixed_with() function old new delta is_prefixed_with - 18 +18 complete_username 78 77 -1 man_main 737 735 -2 fsck_device 429 427 -2 unpack_ar_archive 80 76 -4 strip_unsafe_prefix 105 101 -4 singlemount 1054 1050 -4 rtc_adjtime_is_utc 90 86 -4 resolve_mount_spec 88 84 -4 parse_one_line 1029 1025 -4 parse_conf 1460 1456 -4 may_wakeup 83 79 -4 loadkmap_main 219 215 -4 get_irqs_from_stat 103 99 -4 get_header_cpio 913 909 -4 findfs_main 79 75 -4 fbsplash_main 1230 1226 -4 load_crontab 776 771 -5 expand_vars_to_list 1151 1146 -5 date_main 881 876 -5 skip_dev_pfx 30 24 -6 make_device 2199 2193 -6 complete_cmd_dir_file 773 767 -6 run_applet_and_exit 715 708 -7 uudecode_main 321 313 -8 pwdx_main 197 189 -8 execute 568 560 -8 i2cdetect_main 1186 1176 -10 procps_scan 1242 1230 -12 procps_read_smaps 1017 1005 -12 process_module 746 734 -12 patch_main 1903 1891 -12 nfsmount 3572 3560 -12 stack_machine 126 112 -14 process_timer_stats 449 435 -14 match_fstype 111 97 -14 do_ipaddr 1344 1330 -14 open_list_and_close 359 343 -16 get_header_tar 1795 1779 -16 prepend_new_eth_table 340 323 -17 fsck_main 1811 1794 -17 find_iface_state 56 38 -18 dnsd_main 1321 1303 -18 base_device 179 158 -21 find_keyword 104 82 -22 handle_incoming_and_exit 2785 2762 -23 parse_and_put_prompt 774 746 -28 modinfo 347 317 -30 find_action 204 171 -33 update_passwd 1470 1436 -34 ------------------------------------------------------------------------------ (add/remove: 1/0 grow/shrink: 0/49 up/down: 18/-540) Total: -522 bytes Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2015-03-12 22:18:34 +05:30
if (!global && is_prefixed_with(value, "path=")) {
value += sizeof("path=") - 1;
free(archive_handle->tar__longname);
archive_handle->tar__longname = xstrdup(value);
continue;
}
#endif
#if ENABLE_FEATURE_TAR_SELINUX
/* Scan for SELinux contexts, via "RHT.security.selinux" keyword.
* This is what Red Hat's patched version of tar uses.
*/
# define SELINUX_CONTEXT_KEYWORD "RHT.security.selinux"
libbb: introduce and use is_prefixed_with() function old new delta is_prefixed_with - 18 +18 complete_username 78 77 -1 man_main 737 735 -2 fsck_device 429 427 -2 unpack_ar_archive 80 76 -4 strip_unsafe_prefix 105 101 -4 singlemount 1054 1050 -4 rtc_adjtime_is_utc 90 86 -4 resolve_mount_spec 88 84 -4 parse_one_line 1029 1025 -4 parse_conf 1460 1456 -4 may_wakeup 83 79 -4 loadkmap_main 219 215 -4 get_irqs_from_stat 103 99 -4 get_header_cpio 913 909 -4 findfs_main 79 75 -4 fbsplash_main 1230 1226 -4 load_crontab 776 771 -5 expand_vars_to_list 1151 1146 -5 date_main 881 876 -5 skip_dev_pfx 30 24 -6 make_device 2199 2193 -6 complete_cmd_dir_file 773 767 -6 run_applet_and_exit 715 708 -7 uudecode_main 321 313 -8 pwdx_main 197 189 -8 execute 568 560 -8 i2cdetect_main 1186 1176 -10 procps_scan 1242 1230 -12 procps_read_smaps 1017 1005 -12 process_module 746 734 -12 patch_main 1903 1891 -12 nfsmount 3572 3560 -12 stack_machine 126 112 -14 process_timer_stats 449 435 -14 match_fstype 111 97 -14 do_ipaddr 1344 1330 -14 open_list_and_close 359 343 -16 get_header_tar 1795 1779 -16 prepend_new_eth_table 340 323 -17 fsck_main 1811 1794 -17 find_iface_state 56 38 -18 dnsd_main 1321 1303 -18 base_device 179 158 -21 find_keyword 104 82 -22 handle_incoming_and_exit 2785 2762 -23 parse_and_put_prompt 774 746 -28 modinfo 347 317 -30 find_action 204 171 -33 update_passwd 1470 1436 -34 ------------------------------------------------------------------------------ (add/remove: 1/0 grow/shrink: 0/49 up/down: 18/-540) Total: -522 bytes Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2015-03-12 22:18:34 +05:30
if (is_prefixed_with(value, SELINUX_CONTEXT_KEYWORD"=")) {
value += sizeof(SELINUX_CONTEXT_KEYWORD"=") - 1;
free(archive_handle->tar__sctx[global]);
archive_handle->tar__sctx[global] = xstrdup(value);
continue;
}
#endif
}
free(buf);
}
char FAST_FUNC get_header_tar(archive_handle_t *archive_handle)
2001-10-25 19:48:08 +05:30
{
2002-09-25 08:17:48 +05:30
file_header_t *file_header = archive_handle->file_header;
struct tar_header_t tar;
char *cp;
int i, sum_u, sum;
#if ENABLE_FEATURE_TAR_OLDSUN_COMPATIBILITY
int sum_s;
#endif
int parse_names;
/* Our "private data" */
#if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
# define p_longname (archive_handle->tar__longname)
# define p_linkname (archive_handle->tar__linkname)
#else
# define p_longname 0
# define p_linkname 0
#endif
#if ENABLE_FEATURE_TAR_GNU_EXTENSIONS || ENABLE_FEATURE_TAR_SELINUX
2006-12-26 23:47:42 +05:30
again:
#endif
2002-09-25 08:17:48 +05:30
/* Align header */
data_align(archive_handle, 512);
2002-09-25 08:17:48 +05:30
again_after_align:
#if ENABLE_DESKTOP || ENABLE_FEATURE_TAR_AUTODETECT
/* to prevent misdetection of bz2 sig */
*(aliased_uint32_t*)&tar = 0;
i = full_read(archive_handle->src_fd, &tar, 512);
/* If GNU tar sees EOF in above read, it says:
* "tar: A lone zero block at N", where N = kilobyte
* where EOF was met (not EOF block, actual EOF!),
* and exits with EXIT_SUCCESS.
* We will mimic exit(EXIT_SUCCESS), although we will not mimic
* the message and we don't check whether we indeed
* saw zero block directly before this. */
if (i == 0) {
bb_error_msg("short read");
/* this merely signals end of archive, not exit(1): */
return EXIT_FAILURE;
}
if (i != 512) {
IF_FEATURE_TAR_AUTODETECT(goto autodetect;)
bb_error_msg_and_die("short read");
}
#else
i = 512;
xread(archive_handle->src_fd, &tar, i);
#endif
archive_handle->offset += i;
2001-10-25 19:48:08 +05:30
2002-09-25 08:17:48 +05:30
/* If there is no filename its an empty header */
if (tar.name[0] == 0 && tar.prefix[0] == 0) {
if (archive_handle->tar__end) {
/* Second consecutive empty header - end of archive.
* Read until the end to empty the pipe from gz or bz2
*/
while (full_read(archive_handle->src_fd, &tar, 512) == 512)
continue;
return EXIT_FAILURE; /* "end of archive" */
}
archive_handle->tar__end = 1;
return EXIT_SUCCESS; /* "decoded one header" */
2001-10-25 19:48:08 +05:30
}
archive_handle->tar__end = 0;
2001-10-25 19:48:08 +05:30
/* Check header has valid magic, "ustar" is for the proper tar,
* five NULs are for the old tar format */
libbb: introduce and use is_prefixed_with() function old new delta is_prefixed_with - 18 +18 complete_username 78 77 -1 man_main 737 735 -2 fsck_device 429 427 -2 unpack_ar_archive 80 76 -4 strip_unsafe_prefix 105 101 -4 singlemount 1054 1050 -4 rtc_adjtime_is_utc 90 86 -4 resolve_mount_spec 88 84 -4 parse_one_line 1029 1025 -4 parse_conf 1460 1456 -4 may_wakeup 83 79 -4 loadkmap_main 219 215 -4 get_irqs_from_stat 103 99 -4 get_header_cpio 913 909 -4 findfs_main 79 75 -4 fbsplash_main 1230 1226 -4 load_crontab 776 771 -5 expand_vars_to_list 1151 1146 -5 date_main 881 876 -5 skip_dev_pfx 30 24 -6 make_device 2199 2193 -6 complete_cmd_dir_file 773 767 -6 run_applet_and_exit 715 708 -7 uudecode_main 321 313 -8 pwdx_main 197 189 -8 execute 568 560 -8 i2cdetect_main 1186 1176 -10 procps_scan 1242 1230 -12 procps_read_smaps 1017 1005 -12 process_module 746 734 -12 patch_main 1903 1891 -12 nfsmount 3572 3560 -12 stack_machine 126 112 -14 process_timer_stats 449 435 -14 match_fstype 111 97 -14 do_ipaddr 1344 1330 -14 open_list_and_close 359 343 -16 get_header_tar 1795 1779 -16 prepend_new_eth_table 340 323 -17 fsck_main 1811 1794 -17 find_iface_state 56 38 -18 dnsd_main 1321 1303 -18 base_device 179 158 -21 find_keyword 104 82 -22 handle_incoming_and_exit 2785 2762 -23 parse_and_put_prompt 774 746 -28 modinfo 347 317 -30 find_action 204 171 -33 update_passwd 1470 1436 -34 ------------------------------------------------------------------------------ (add/remove: 1/0 grow/shrink: 0/49 up/down: 18/-540) Total: -522 bytes Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2015-03-12 22:18:34 +05:30
if (!is_prefixed_with(tar.magic, "ustar")
&& (!ENABLE_FEATURE_TAR_OLDGNU_COMPATIBILITY
|| memcmp(tar.magic, "\0\0\0\0", 5) != 0)
) {
#if ENABLE_FEATURE_TAR_AUTODETECT
autodetect:
/* Two different causes for lseek() != 0:
* unseekable fd (would like to support that too, but...),
* or not first block (false positive, it's not .gz/.bz2!) */
if (lseek(archive_handle->src_fd, -i, SEEK_CUR) != 0)
goto err;
if (setup_unzip_on_fd(archive_handle->src_fd, /*fail_if_not_compressed:*/ 0) != 0)
err:
bb_error_msg_and_die("invalid tar magic");
archive_handle->offset = 0;
goto again_after_align;
#endif
bb_error_msg_and_die("invalid tar magic");
2001-10-25 19:48:08 +05:30
}
/* Do checksum on headers.
* POSIX says that checksum is done on unsigned bytes, but
2007-03-04 23:39:50 +05:30
* Sun and HP-UX gets it wrong... more details in
* GNU tar source. */
#if ENABLE_FEATURE_TAR_OLDSUN_COMPATIBILITY
sum_s = ' ' * sizeof(tar.chksum);
#endif
sum_u = ' ' * sizeof(tar.chksum);
2007-07-21 20:38:09 +05:30
for (i = 0; i < 148; i++) {
sum_u += ((unsigned char*)&tar)[i];
#if ENABLE_FEATURE_TAR_OLDSUN_COMPATIBILITY
sum_s += ((signed char*)&tar)[i];
#endif
2001-10-25 19:48:08 +05:30
}
2007-07-21 20:38:09 +05:30
for (i = 156; i < 512; i++) {
sum_u += ((unsigned char*)&tar)[i];
#if ENABLE_FEATURE_TAR_OLDSUN_COMPATIBILITY
sum_s += ((signed char*)&tar)[i];
#endif
2001-10-25 19:48:08 +05:30
}
/* This field does not need special treatment (getOctal) */
{
char *endp; /* gcc likes temp var for &endp */
sum = strtoul(tar.chksum, &endp, 8);
if ((*endp != '\0' && *endp != ' ')
|| (sum_u != sum IF_FEATURE_TAR_OLDSUN_COMPATIBILITY(&& sum_s != sum))
) {
bb_error_msg_and_die("invalid tar header checksum");
}
}
/* don't use xstrtoul, tar.chksum may have leading spaces */
sum = strtoul(tar.chksum, NULL, 8);
if (sum_u != sum IF_FEATURE_TAR_OLDSUN_COMPATIBILITY(&& sum_s != sum)) {
2006-10-08 18:19:22 +05:30
bb_error_msg_and_die("invalid tar header checksum");
2001-10-25 19:48:08 +05:30
}
/* 0 is reserved for high perf file, treat as normal file */
if (!tar.typeflag) tar.typeflag = '0';
parse_names = (tar.typeflag >= '0' && tar.typeflag <= '7');
/* getOctal trashes subsequent field, therefore we call it
* on fields in reverse order */
if (tar.devmajor[0]) {
char t = tar.prefix[0];
/* we trash prefix[0] here, but we DO need it later! */
unsigned minor = GET_OCTAL(tar.devminor);
unsigned major = GET_OCTAL(tar.devmajor);
file_header->device = makedev(major, minor);
tar.prefix[0] = t;
}
file_header->link_target = NULL;
if (!p_linkname && parse_names && tar.linkname[0]) {
file_header->link_target = xstrndup(tar.linkname, sizeof(tar.linkname));
/* FIXME: what if we have non-link object with link_target? */
/* Will link_target be free()ed? */
}
#if ENABLE_FEATURE_TAR_UNAME_GNAME
file_header->tar__uname = tar.uname[0] ? xstrndup(tar.uname, sizeof(tar.uname)) : NULL;
file_header->tar__gname = tar.gname[0] ? xstrndup(tar.gname, sizeof(tar.gname)) : NULL;
#endif
file_header->mtime = GET_OCTAL(tar.mtime);
file_header->size = GET_OCTAL(tar.size);
file_header->gid = GET_OCTAL(tar.gid);
file_header->uid = GET_OCTAL(tar.uid);
2004-02-20 08:04:42 +05:30
/* Set bits 0-11 of the files mode */
file_header->mode = 07777 & GET_OCTAL(tar.mode);
2004-02-20 08:04:42 +05:30
file_header->name = NULL;
if (!p_longname && parse_names) {
/* we trash mode[0] here, it's ok */
//tar.name[sizeof(tar.name)] = '\0'; - gcc 4.3.0 would complain
tar.mode[0] = '\0';
2006-11-26 05:17:32 +05:30
if (tar.prefix[0]) {
/* and padding[0] */
//tar.prefix[sizeof(tar.prefix)] = '\0'; - gcc 4.3.0 would complain
tar.padding[0] = '\0';
file_header->name = concat_path_file(tar.prefix, tar.name);
2006-11-26 05:17:32 +05:30
} else
file_header->name = xstrdup(tar.name);
}
2004-02-20 08:04:42 +05:30
/* Set bits 12-15 of the files mode */
/* (typeflag was not trashed because chksum does not use getOctal) */
switch (tar.typeflag) {
case '1': /* hardlink */
/* we mark hardlinks as regular files with zero size and a link name */
2004-02-20 08:04:42 +05:30
file_header->mode |= S_IFREG;
/* on size of link fields from star(4)
* ... For tar archives written by pre POSIX.1-1988
* implementations, the size field usually contains the size of
* the file and needs to be ignored as no data may follow this
* header type. For POSIX.1- 1988 compliant archives, the size
* field needs to be 0. For POSIX.1-2001 compliant archives,
* the size field may be non zero, indicating that file data is
* included in the archive.
* i.e; always assume this is zero for safety.
*/
goto size0;
case '7':
/* case 0: */
case '0':
#if ENABLE_FEATURE_TAR_OLDGNU_COMPATIBILITY
if (last_char_is(file_header->name, '/')) {
goto set_dir;
}
#endif
file_header->mode |= S_IFREG;
2002-08-22 23:17:09 +05:30
break;
case '2':
2002-09-25 08:17:48 +05:30
file_header->mode |= S_IFLNK;
/* have seen tarballs with size field containing
* the size of the link target's name */
size0:
file_header->size = 0;
2002-08-22 23:17:09 +05:30
break;
case '3':
2002-09-25 08:17:48 +05:30
file_header->mode |= S_IFCHR;
goto size0; /* paranoia */
case '4':
2002-09-25 08:17:48 +05:30
file_header->mode |= S_IFBLK;
goto size0;
case '5':
IF_FEATURE_TAR_OLDGNU_COMPATIBILITY(set_dir:)
2002-09-25 08:17:48 +05:30
file_header->mode |= S_IFDIR;
goto size0;
case '6':
2002-09-25 08:17:48 +05:30
file_header->mode |= S_IFIFO;
goto size0;
case 'g': /* pax global header */
case 'x': { /* pax extended header */
if ((uoff_t)file_header->size > 0xfffff) /* paranoia */
goto skip_ext_hdr;
process_pax_hdr(archive_handle, file_header->size, (tar.typeflag == 'g'));
goto again_after_align;
#if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
/* See http://www.gnu.org/software/tar/manual/html_node/Extensions.html */
case 'L':
/* free: paranoia: tar with several consecutive longnames */
free(p_longname);
/* For paranoia reasons we allocate extra NUL char */
p_longname = xzalloc(file_header->size + 1);
/* We read ASCIZ string, including NUL */
xread(archive_handle->src_fd, p_longname, file_header->size);
archive_handle->offset += file_header->size;
/* return get_header_tar(archive_handle); */
/* gcc 4.1.1 didn't optimize it into jump */
/* so we will do it ourself, this also saves stack */
goto again;
case 'K':
free(p_linkname);
p_linkname = xzalloc(file_header->size + 1);
xread(archive_handle->src_fd, p_linkname, file_header->size);
archive_handle->offset += file_header->size;
/* return get_header_tar(archive_handle); */
goto again;
/*
* case 'S': // Sparse file
* Was seen in the wild. Not supported (yet?).
* See https://www.gnu.org/software/tar/manual/html_section/tar_92.html
* for the format. (An "Old GNU Format" was seen, not PAX formats).
*/
// case 'D': /* GNU dump dir */
// case 'M': /* Continuation of multi volume archive */
// case 'N': /* Old GNU for names > 100 characters */
// case 'V': /* Volume header */
#endif
}
skip_ext_hdr:
{
off_t sz;
bb_error_msg("warning: skipping header '%c'", tar.typeflag);
sz = (file_header->size + 511) & ~(off_t)511;
archive_handle->offset += sz;
sz >>= 9; /* sz /= 512 but w/o contortions for signed div */
while (sz--)
xread(archive_handle->src_fd, &tar, 512);
/* return get_header_tar(archive_handle); */
goto again_after_align;
}
default:
bb_error_msg_and_die("unknown typeflag: 0x%x", tar.typeflag);
}
#if ENABLE_FEATURE_TAR_GNU_EXTENSIONS
if (p_longname) {
file_header->name = p_longname;
p_longname = NULL;
}
if (p_linkname) {
file_header->link_target = p_linkname;
p_linkname = NULL;
}
#endif
/* Everything up to and including last ".." component is stripped */
overlapping_strcpy(file_header->name, strip_unsafe_prefix(file_header->name));
//TODO: do the same for file_header->link_target?
/* Strip trailing '/' in directories */
/* Must be done after mode is set as '/' is used to check if it's a directory */
cp = last_char_is(file_header->name, '/');
if (archive_handle->filter(archive_handle) == EXIT_SUCCESS) {
archive_handle->action_header(/*archive_handle->*/ file_header);
/* Note that we kill the '/' only after action_header() */
/* (like GNU tar 1.15.1: verbose mode outputs "dir/dir/") */
if (cp)
*cp = '\0';
2002-09-25 08:17:48 +05:30
archive_handle->action_data(archive_handle);
if (archive_handle->accept || archive_handle->reject
|| (archive_handle->ah_flags & ARCHIVE_REMEMBER_NAMES)
) {
llist_add_to(&archive_handle->passed, file_header->name);
} else /* Caller isn't interested in list of unpacked files */
free(file_header->name);
2002-09-25 08:17:48 +05:30
} else {
data_skip(archive_handle);
free(file_header->name);
2002-09-25 08:17:48 +05:30
}
archive_handle->offset += file_header->size;
free(file_header->link_target);
/* Do not free(file_header->name)!
* It might be inserted in archive_handle->passed - see above */
#if ENABLE_FEATURE_TAR_UNAME_GNAME
free(file_header->tar__uname);
free(file_header->tar__gname);
#endif
return EXIT_SUCCESS; /* "decoded one header" */
}