emo/busybox
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

tar: support GNU 256-bit encoding in all numeric fields

Browse Source 
function                                             old     new   delta
getOctal                                              63     125     +62
get_header_tar                                      1572    1496     -76
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/1 up/down: 62/-76)            Total: -14 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This commit is contained in:
Denys Vlasenko 2010-05-06 20:08:14 +02:00
parent c7efd6441d
commit 7b48eb4372
1 changed files with 31 additions and 80 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

103
archival/libunarchive/get_header_tar.c
View File

@ -18,87 +18,45 @@ typedef uint32_t aliased_uint32_t FIX_ALIASING;
typedef off_t aliased_off_t FIX_ALIASING; typedef off_t aliased_off_t FIX_ALIASING;
/* /* NB: _DESTROYS_ str[len] character! */
* GNU tar uses "base-256 encoding" for very large numbers (>8 billion). 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 * Encoding is binary, with highest bit always set as a marker
* and sign in next-highest bit: * and sign in next-highest bit:
* 80 00 .. 00 - zero * 80 00 .. 00 - zero
* bf ff .. ff - largest positive number * bf ff .. ff - largest positive number
* ff ff .. ff - minus 1 * ff ff .. ff - minus 1
* c0 00 .. 00 - smallest negative number * c0 00 .. 00 - smallest negative number
*
* We expect it only in size field, where negative numbers don't make sense.
*/
static off_t getBase256_len12(const char *str)
{
off_t value;
int len;
/* if (*str & 0x40) error; - caller prevents this */
if (sizeof(off_t) >= 12) {
/* Probably 128-bit (16 byte) off_t. Can be optimized. */
len = 12;
value = *str++ & 0x3f;
while (--len)
value = (value << 8) + (unsigned char) *str++;
return value;
}
#ifdef CHECK_FOR_OVERFLOW
/* Can be optimized to eat 32-bit chunks */
char c = *str++ & 0x3f;
len = 12;
while (1) {
if (c)
bb_error_msg_and_die("overflow in base-256 encoded file size");
if (--len == sizeof(off_t))
break;
c = *str++;
}
#else
str += (12 - sizeof(off_t));
#endif
/* Now str points to sizeof(off_t) least significant bytes.
* *
* Example of tar file with 8914993153 (0x213600001) byte file. * Example of tar file with 8914993153 (0x213600001) byte file.
* Field starts at offset 7c: * Field starts at offset 7c:
* 00070 30 30 30 00 30 30 30 30 30 30 30 00 80 00 00 00 |000.0000000.....| * 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| * 00080 00 00 00 02 13 60 00 01 31 31 31 32 30 33 33 36 |.....`..11120336|
* *
* str is at offset 80 or 84 now (64-bit or 32-bit off_t). * NB: tarballs with NEGATIVE unix times encoded that way were seen!
* We (ab)use the fact that value happens to be aligned,
* and fetch it in one go:
*/ */
if (sizeof(off_t) == 8) { v = first;
value = *(aliased_off_t*)str; /* Sign-extend using 6th bit: */
value = SWAP_BE64(value); v <<= sizeof(unsigned long long)*8 - 7;
} else if (sizeof(off_t) == 4) { v = (long long)v >> (sizeof(unsigned long long)*8 - 7);
value = *(aliased_off_t*)str; while (--len != 0)
value = SWAP_BE32(value); v = (v << 8) + (unsigned char) *str++;
} else {
value = 0;
len = sizeof(off_t);
while (--len)
value = (value << 8) + (unsigned char) *str++;
} }
return value;
}
/* NB: _DESTROYS_ str[len] character! */
static unsigned long long getOctal(char *str, int len)
{
unsigned long long v;
/* 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, &str, 8);
/* std: "Each numeric field is terminated by one or more
* <space> or NUL characters". We must support ' '! */
if (*str != '\0' && *str != ' ')
bb_error_msg_and_die("corrupted octal value in tar header");
return v; return v;
} }
#define GET_OCTAL(a) getOctal((a), sizeof(a)) #define GET_OCTAL(a) getOctal((a), sizeof(a))
@ -358,15 +316,8 @@ char FAST_FUNC get_header_tar(archive_handle_t *archive_handle)
file_header->tar__uname = tar.uname[0] ? xstrndup(tar.uname, sizeof(tar.uname)) : NULL; 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; file_header->tar__gname = tar.gname[0] ? xstrndup(tar.gname, sizeof(tar.gname)) : NULL;
#endif #endif
/* mtime: rudimentally handle GNU tar's "base256 encoding" file_header->mtime = GET_OCTAL(tar.mtime);
* People report tarballs with NEGATIVE unix times encoded that way */ file_header->size = GET_OCTAL(tar.size);
file_header->mtime = (tar.mtime[0] & 0x80) /* base256? */
? 0 /* bogus */
: GET_OCTAL(tar.mtime);
/* size: handle GNU tar's "base256 encoding" */
file_header->size = (tar.size[0] & 0xc0) == 0x80 /* positive base256? */
? getBase256_len12(tar.size)
: GET_OCTAL(tar.size);
file_header->gid = GET_OCTAL(tar.gid); file_header->gid = GET_OCTAL(tar.gid);
file_header->uid = GET_OCTAL(tar.uid); file_header->uid = GET_OCTAL(tar.uid);
/* Set bits 0-11 of the files mode */ /* Set bits 0-11 of the files mode */