Replace base{32,64}_decode() with table-driven implementations. The new

code is less strict about padding, thus ensuring compatibility with
implementations which do not understand padding, such as MIME::Base32.


git-svn-id: svn+ssh://svn.openpam.org/svn/openpam/trunk@771 185d5e19-27fe-0310-9dcf-9bff6b9f3609
This commit is contained in:
Dag-Erling Smørgrav 2014-03-06 17:54:58 +00:00
parent 4645bc1762
commit 17144e7a5f
2 changed files with 136 additions and 176 deletions

View File

@ -37,14 +37,36 @@
#include <errno.h> #include <errno.h>
#include <stdint.h> #include <stdint.h>
#include <stdio.h>
#include <security/oath.h> #include <security/oath.h>
static const char b32[] = static const char b32enc[] =
"ABCDEFGHIJKLMNOPQRSTUVWXYZ234567"; "ABCDEFGHIJKLMNOPQRSTUVWXYZ234567";
static const uint8_t b32dec[256] = {
['A'] = 0, ['B'] = 1, ['C'] = 2, ['D'] = 3,
['E'] = 4, ['F'] = 5, ['G'] = 6, ['H'] = 7,
['I'] = 8, ['J'] = 9, ['K'] = 10, ['L'] = 11,
['M'] = 12, ['N'] = 13, ['O'] = 14, ['P'] = 15,
['Q'] = 16, ['R'] = 17, ['S'] = 18, ['T'] = 19,
['U'] = 20, ['V'] = 21, ['W'] = 22, ['X'] = 23,
['Y'] = 24, ['Z'] = 25,
['a'] = 0, ['b'] = 1, ['c'] = 2, ['d'] = 3,
['e'] = 4, ['f'] = 5, ['g'] = 6, ['h'] = 7,
['i'] = 8, ['j'] = 9, ['k'] = 10, ['l'] = 11,
['m'] = 12, ['n'] = 13, ['o'] = 14, ['p'] = 15,
['q'] = 16, ['r'] = 17, ['s'] = 18, ['t'] = 19,
['u'] = 20, ['v'] = 21, ['w'] = 22, ['x'] = 23,
['y'] = 24, ['z'] = 25,
['2'] = 26, ['3'] = 27, ['4'] = 28, ['5'] = 29,
['6'] = 30, ['7'] = 31,
};
/* /*
* Encode data in RFC 3548 base 32 representation. The target buffer must * Encode data in RFC 4648 base 32 representation. The target buffer must
* have room for base32_enclen(len) characters and a terminating NUL. * have room for base32_enclen(len) characters and a terminating NUL.
*/ */
int int
@ -52,8 +74,10 @@ base32_enc(const uint8_t *in, size_t ilen, char *out, size_t *olen)
{ {
uint64_t bits; uint64_t bits;
if (*olen <= base32_enclen(ilen)) if (*olen <= base32_enclen(ilen)) {
errno = ENOSPC;
return (-1); return (-1);
}
*olen = 0; *olen = 0;
while (ilen >= 5) { while (ilen >= 5) {
bits = 0; bits = 0;
@ -64,14 +88,14 @@ base32_enc(const uint8_t *in, size_t ilen, char *out, size_t *olen)
bits |= (uint64_t)in[4]; bits |= (uint64_t)in[4];
ilen -= 5; ilen -= 5;
in += 5; in += 5;
out[0] = b32[bits >> 35 & 0x1f]; out[0] = b32enc[bits >> 35 & 0x1f];
out[1] = b32[bits >> 30 & 0x1f]; out[1] = b32enc[bits >> 30 & 0x1f];
out[2] = b32[bits >> 25 & 0x1f]; out[2] = b32enc[bits >> 25 & 0x1f];
out[3] = b32[bits >> 20 & 0x1f]; out[3] = b32enc[bits >> 20 & 0x1f];
out[4] = b32[bits >> 15 & 0x1f]; out[4] = b32enc[bits >> 15 & 0x1f];
out[5] = b32[bits >> 10 & 0x1f]; out[5] = b32enc[bits >> 10 & 0x1f];
out[6] = b32[bits >> 5 & 0x1f]; out[6] = b32enc[bits >> 5 & 0x1f];
out[7] = b32[bits & 0x1f]; out[7] = b32enc[bits & 0x1f];
*olen += 8; *olen += 8;
out += 8; out += 8;
} }
@ -87,13 +111,13 @@ base32_enc(const uint8_t *in, size_t ilen, char *out, size_t *olen)
case 1: case 1:
bits |= (uint64_t)in[0] << 32; bits |= (uint64_t)in[0] << 32;
} }
out[0] = b32[bits >> 35 & 0x1f]; out[0] = b32enc[bits >> 35 & 0x1f];
out[1] = b32[bits >> 30 & 0x1f]; out[1] = b32enc[bits >> 30 & 0x1f];
out[2] = ilen > 1 ? b32[bits >> 25 & 0x1f] : '='; out[2] = ilen > 1 ? b32enc[bits >> 25 & 0x1f] : '=';
out[3] = ilen > 1 ? b32[bits >> 20 & 0x1f] : '='; out[3] = ilen > 1 ? b32enc[bits >> 20 & 0x1f] : '=';
out[4] = ilen > 2 ? b32[bits >> 15 & 0x1f] : '='; out[4] = ilen > 2 ? b32enc[bits >> 15 & 0x1f] : '=';
out[5] = ilen > 3 ? b32[bits >> 10 & 0x1f] : '='; out[5] = ilen > 3 ? b32enc[bits >> 10 & 0x1f] : '=';
out[6] = ilen > 3 ? b32[bits >> 5 & 0x1f] : '='; out[6] = ilen > 3 ? b32enc[bits >> 5 & 0x1f] : '=';
out[7] = '='; out[7] = '=';
*olen += 8; *olen += 8;
out += 8; out += 8;
@ -104,10 +128,15 @@ base32_enc(const uint8_t *in, size_t ilen, char *out, size_t *olen)
} }
/* /*
* Decode data in RFC 2548 base 32 representation, stopping at the * Decode data in RFC 4648 base 32 representation, stopping at the
* terminating NUL, the first invalid (non-base32, non-whitespace) * terminating NUL, the first invalid (non-base32, non-whitespace)
* character or after len characters, whichever comes first. * character or after len characters, whichever comes first.
* *
* Padding is handled sloppily: any padding character following the data
* is silently consumed. This not only simplifies the code but ensures
* compatibility with implementations which do not emit or understand
* padding.
*
* The olen argument is used by the caller to pass the size of the buffer * The olen argument is used by the caller to pass the size of the buffer
* and by base32_dec() to return the amount of data successfully decoded. * and by base32_dec() to return the amount of data successfully decoded.
* If the buffer is too small, base32_dec() discards the excess data, but * If the buffer is too small, base32_dec() discards the excess data, but
@ -117,90 +146,40 @@ int
base32_dec(const char *in, size_t ilen, uint8_t *out, size_t *olen) base32_dec(const char *in, size_t ilen, uint8_t *out, size_t *olen)
{ {
size_t len; size_t len;
uint64_t bits; int bits, shift, padding;
int shift;
for (len = 0, bits = 0, shift = 40; ilen && *in && *in != '='; --ilen, ++in) { for (bits = shift = padding = len = 0; ilen && *in; --ilen, ++in) {
if (*in == ' ' || *in == '\t' || *in == '\r' || *in == '\n') { if (*in == ' ' || *in == '\t' || *in == '\r' || *in == '\n' ||
(padding && *in == '=')) {
/* consume */
continue; continue;
} else if (*in >= 'A' && *in <= 'Z') { } else if (!padding && b32dec[(int)*in]) {
shift -= 5; /* shift into accumulator */
bits |= (uint64_t)(*in - 'A') << shift; shift += 5;
} else if (*in >= 'a' && *in <= 'z') { bits = bits << 5 | b32dec[(int)*in];
shift -= 5; } else if (!padding && shift && *in == '=') {
bits |= (uint64_t)(*in - 'a') << shift; /* final byte */
} else if (*in >= '2' && *in <= '7') { shift = 0;
shift -= 5; padding = 1;
bits |= (uint64_t)(*in - '2' + 26) << shift;
} else if (*in == '=') {
/* handled below */
break;
} else { } else {
goto bad; /* error */
*olen = 0;
errno = EINVAL;
return (-1);
} }
if (shift == 0) { if (shift >= 8) {
if ((len += 5) <= *olen) { /* output accumulated byte */
*out++ = (bits >> 32) & 0xff; shift -= 8;
*out++ = (bits >> 24) & 0xff; if (len++ < *olen)
*out++ = (bits >> 16) & 0xff; *out++ = (bits >> shift) & 0xff;
*out++ = (bits >> 8) & 0xff;
*out++ = bits & 0xff;
}
bits = 0;
shift = 40;
} }
} }
if (ilen && *in == '=' && /* report decoded length */
(shift == 30 || shift == 20 || shift == 15 || shift == 5)) {
/*
* Padding:
*
* 00 8 AA====== 30
* 00 00 16 AAAA==== 20
* 00 00 00 24 AAAAA=== 15
* 00 00 00 00 32 AAAAAAA= 5
*
* XXX We should check that the last few bits before the
* padding starts are zero.
*/
switch (shift) {
case 5:
if (++len <= *olen)
*out++ = (bits >> 32) & 0xff;
bits <<= 8;
case 15:
if (++len <= *olen)
*out++ = (bits >> 32) & 0xff;
bits <<= 8;
case 20:
if (++len <= *olen)
*out++ = (bits >> 32) & 0xff;
bits <<= 8;
case 30:
if (++len <= *olen)
*out++ = (bits >> 32) & 0xff;
bits <<= 8;
break;
default:
goto bad;
}
/* consume remaining padding and whitespace */
for (; ilen && *in; --ilen, ++in) {
if (*in == ' ' || *in == '\t' || *in == '\r' || *in == '\n')
continue;
else if (*in == '=' && shift)
shift -= 5;
else
goto bad;
}
}
if (ilen)
goto bad;
*olen = len; *olen = len;
if (len > *olen) if (len > *olen) {
/* overflow */
errno = ENOSPC;
return (-1); return (-1);
}
return (0); return (0);
bad:
*olen = 0;
return (-1);
} }

View File

@ -40,13 +40,32 @@
#include <security/oath.h> #include <security/oath.h>
static const char b64[] = static const char b64enc[] =
"ABCDEFGHIJKLMNOPQRSTUVWXYZ" "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
"abcdefghijklmnopqrstuvwxyz" "abcdefghijklmnopqrstuvwxyz"
"0123456789+/"; "0123456789+/";
static const uint8_t b64dec[256] = {
['A'] = 0, ['B'] = 1, ['C'] = 2, ['D'] = 3,
['E'] = 4, ['F'] = 5, ['G'] = 6, ['H'] = 7,
['I'] = 8, ['J'] = 9, ['K'] = 10, ['L'] = 11,
['M'] = 12, ['N'] = 13, ['O'] = 14, ['P'] = 15,
['Q'] = 16, ['R'] = 17, ['S'] = 18, ['T'] = 19,
['U'] = 20, ['V'] = 21, ['W'] = 22, ['X'] = 23,
['Y'] = 24, ['Z'] = 25, ['a'] = 26, ['b'] = 27,
['c'] = 28, ['d'] = 29, ['e'] = 30, ['f'] = 31,
['g'] = 32, ['h'] = 33, ['i'] = 34, ['j'] = 35,
['k'] = 36, ['l'] = 37, ['m'] = 38, ['n'] = 39,
['o'] = 40, ['p'] = 41, ['q'] = 42, ['r'] = 43,
['s'] = 44, ['t'] = 45, ['u'] = 46, ['v'] = 47,
['w'] = 48, ['x'] = 49, ['y'] = 50, ['z'] = 51,
['0'] = 52, ['1'] = 53, ['2'] = 54, ['3'] = 55,
['4'] = 56, ['5'] = 57, ['6'] = 58, ['7'] = 59,
['8'] = 60, ['9'] = 61, ['+'] = 62, ['/'] = 63,
};
/* /*
* Encode data in RFC 3548 base 64 representation. The target buffer must * Encode data in RFC 4648 base 64 representation. The target buffer must
* have room for base64_enclen(len) characters and a terminating NUL. * have room for base64_enclen(len) characters and a terminating NUL.
*/ */
int int
@ -64,10 +83,10 @@ base64_enc(const uint8_t *in, size_t ilen, char *out, size_t *olen)
bits |= (uint32_t)in[2]; bits |= (uint32_t)in[2];
ilen -= 3; ilen -= 3;
in += 3; in += 3;
out[0] = b64[bits >> 18 & 0x3f]; out[0] = b64enc[bits >> 18 & 0x3f];
out[1] = b64[bits >> 12 & 0x3f]; out[1] = b64enc[bits >> 12 & 0x3f];
out[2] = b64[bits >> 6 & 0x3f]; out[2] = b64enc[bits >> 6 & 0x3f];
out[3] = b64[bits & 0x3f]; out[3] = b64enc[bits & 0x3f];
*olen += 4; *olen += 4;
out += 4; out += 4;
} }
@ -79,9 +98,9 @@ base64_enc(const uint8_t *in, size_t ilen, char *out, size_t *olen)
case 1: case 1:
bits |= (uint32_t)in[0] << 16; bits |= (uint32_t)in[0] << 16;
} }
out[0] = b64[bits >> 18 & 0x3f]; out[0] = b64enc[bits >> 18 & 0x3f];
out[1] = b64[bits >> 12 & 0x3f]; out[1] = b64enc[bits >> 12 & 0x3f];
out[2] = ilen > 1 ? b64[bits >> 6 & 0x3f] : '='; out[2] = ilen > 1 ? b64enc[bits >> 6 & 0x3f] : '=';
out[3] = '='; out[3] = '=';
*olen += 4; *olen += 4;
out += 4; out += 4;
@ -92,10 +111,15 @@ base64_enc(const uint8_t *in, size_t ilen, char *out, size_t *olen)
} }
/* /*
* Decode data in RFC 2548 base 64 representation, stopping at the * Decode data in RFC 4648 base 64 representation, stopping at the
* terminating NUL, the first invalid (non-base64, non-whitespace) * terminating NUL, the first invalid (non-base64, non-whitespace)
* character or after len characters, whichever comes first. * character or after len characters, whichever comes first.
* *
* Padding is handled sloppily: any padding character following the data
* is silently consumed. This not only simplifies the code but ensures
* compatibility with implementations which do not emit or understand
* padding.
*
* The olen argument is used by the caller to pass the size of the buffer * The olen argument is used by the caller to pass the size of the buffer
* and by base64_dec() to return the amount of data successfully decoded. * and by base64_dec() to return the amount of data successfully decoded.
* If the buffer is too small, base64_dec() discards the excess data, but * If the buffer is too small, base64_dec() discards the excess data, but
@ -105,83 +129,40 @@ int
base64_dec(const char *in, size_t ilen, uint8_t *out, size_t *olen) base64_dec(const char *in, size_t ilen, uint8_t *out, size_t *olen)
{ {
size_t len; size_t len;
uint32_t bits; int bits, shift, padding;
int shift;
for (len = 0, bits = 0, shift = 24; ilen && *in; --ilen, ++in) { for (bits = shift = padding = len = 0; ilen && *in; --ilen, ++in) {
if (*in == ' ' || *in == '\t' || *in == '\r' || *in == '\n') { if (*in == ' ' || *in == '\t' || *in == '\r' || *in == '\n' ||
(padding && *in == '=')) {
/* consume */
continue; continue;
} else if (*in >= 'A' && *in <= 'Z') { } else if (!padding && b64dec[(int)*in]) {
shift -= 6; /* shift into accumulator */
bits |= (uint32_t)(*in - 'A') << shift; shift += 6;
} else if (*in >= 'a' && *in <= 'z') { bits = bits << 6 | b64dec[(int)*in];
shift -= 6; } else if (!padding && shift && *in == '=') {
bits |= (uint32_t)(*in - 'a' + 26) << shift; /* final byte */
} else if (*in >= '0' && *in <= '9') { shift = 0;
shift -= 6; padding = 1;
bits |= (uint32_t)(*in - '0' + 52) << shift;
} else if (*in == '+') {
shift -= 6;
bits |= (uint32_t)62 << shift;
} else if (*in == '/') {
shift -= 6;
bits |= (uint32_t)63 << shift;
} else if (*in == '=') {
/* handled below */
break;
} else { } else {
goto bad; /* error */
*olen = 0;
errno = EINVAL;
return (-1);
} }
if (shift == 0) { if (shift >= 8) {
if ((len += 3) <= *olen) { /* output accumulated byte */
*out++ = (bits >> 16) & 0xff; shift -= 8;
*out++ = (bits >> 8) & 0xff; if (len++ < *olen)
*out++ = bits & 0xff; *out++ = (bits >> shift) & 0xff;
}
bits = 0;
shift = 24;
} }
} }
if (ilen && *in == '=' && (shift == 12 || shift == 6)) { /* report decoded length */
/*
* Padding:
*
* 00 8 AA== 12
* 00 00 16 AAA= 6
*
* XXX We should check that the last few bits before the
* padding starts are zero.
*/
switch (shift) {
case 6:
if (++len <= *olen)
*out++ = (bits >> 16) & 0xff;
bits <<= 8;
case 12:
if (++len <= *olen)
*out++ = (bits >> 16) & 0xff;
bits <<= 8;
break;
default:
goto bad;
}
/* consume remaining padding and whitespace */
for (; ilen && *in; --ilen, ++in) {
if (*in == ' ' || *in == '\t' || *in == '\r' || *in == '\n')
continue;
else if (*in == '=' && shift)
shift -= 6;
else
goto bad;
}
}
if (ilen)
goto bad;
*olen = len; *olen = len;
if (len > *olen) if (len > *olen) {
/* overflow */
errno = ENOSPC;
return (-1); return (-1);
}
return (0); return (0);
bad:
*olen = 0;
return (-1);
} }