busybox/networking/tls_aesgcm.c
Denys Vlasenko 32ec5f1705 tls: AES-GCM: in GMULT, avoid memcpy, use one less variable in bit loop
function                                             old     new   delta
GMULT                                                168     159      -9

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2018-12-08 21:24:38 +01:00

201 lines
5.8 KiB
C

/*
* Copyright (C) 2018 Denys Vlasenko
*
* Licensed under GPLv2, see file LICENSE in this source tree.
*/
#include "tls.h"
typedef uint8_t byte;
typedef uint32_t word32;
#define XMEMSET memset
#define XMEMCPY memcpy
/* from wolfssl-3.15.3/wolfcrypt/src/aes.c */
static ALWAYS_INLINE void FlattenSzInBits(byte* buf, word32 sz)
{
/* Multiply the sz by 8 */
//bbox: these sizes are never even close to 2^32/8
// word32 szHi = (sz >> (8*sizeof(sz) - 3));
sz <<= 3;
/* copy over the words of the sz into the destination buffer */
// buf[0] = (szHi >> 24) & 0xff;
// buf[1] = (szHi >> 16) & 0xff;
// buf[2] = (szHi >> 8) & 0xff;
// buf[3] = szHi & 0xff;
*(uint32_t*)(buf + 0) = 0;
// buf[4] = (sz >> 24) & 0xff;
// buf[5] = (sz >> 16) & 0xff;
// buf[6] = (sz >> 8) & 0xff;
// buf[7] = sz & 0xff;
*(uint32_t*)(buf + 4) = SWAP_BE32(sz);
}
static void RIGHTSHIFTX(byte* x)
{
#define l ((unsigned long*)x)
#if 0
// Generic byte-at-a-time algorithm
int i;
byte carryIn = (x[15] & 0x01) ? 0xE1 : 0;
for (i = 0; i < AES_BLOCK_SIZE; i++) {
byte carryOut = (x[i] << 7); // zero, or 0x80
x[i] = (x[i] >> 1) ^ carryIn;
carryIn = carryOut;
}
#elif BB_BIG_ENDIAN
// Big-endian can shift-right in larger than byte chunks
// (we use the fact that 'x' is long-aligned)
unsigned long carryIn = (x[15] & 0x01)
? ((unsigned long)0xE1 << (LONG_BIT-8))
: 0;
# if ULONG_MAX <= 0xffffffff
int i;
for (i = 0; i < AES_BLOCK_SIZE/sizeof(long); i++) {
unsigned long carryOut = l[i] << (LONG_BIT-1); // zero, or 0x800..00
l[i] = (l[i] >> 1) ^ carryIn;
carryIn = carryOut;
}
# else
// 64-bit code: need to process only 2 words
unsigned long carryOut = l[0] << (LONG_BIT-1); // zero, or 0x800..00
l[0] = (l[0] >> 1) ^ carryIn;
l[1] = (l[1] >> 1) ^ carryOut;
# endif
#else /* LITTLE_ENDIAN */
// In order to use word-sized ops, little-endian needs to byteswap.
// On x86, code size increase is ~10 bytes compared to byte-by-byte.
unsigned long carryIn = (x[15] & 0x01)
? ((unsigned long)0xE1 << (LONG_BIT-8))
: 0;
# if ULONG_MAX <= 0xffffffff
int i;
for (i = 0; i < AES_BLOCK_SIZE/sizeof(long); i++) {
unsigned long ti = SWAP_BE32(l[i]);
unsigned long carryOut = ti << (LONG_BIT-1); // zero, or 0x800..00
ti = (ti >> 1) ^ carryIn;
l[i] = SWAP_BE32(ti);
carryIn = carryOut;
}
# else
// 64-bit code: need to process only 2 words
unsigned long tt = SWAP_BE64(l[0]);
unsigned long carryOut = tt << (LONG_BIT-1); // zero, or 0x800..00
tt = (tt >> 1) ^ carryIn; l[0] = SWAP_BE64(tt);
tt = SWAP_BE64(l[1]);
tt = (tt >> 1) ^ carryOut; l[1] = SWAP_BE64(tt);
# endif
#endif /* LITTLE_ENDIAN */
#undef l
}
// Caller guarantees X is aligned
static void GMULT(byte* X, byte* Y)
{
byte Z[AES_BLOCK_SIZE] ALIGNED_long;
//byte V[AES_BLOCK_SIZE] ALIGNED_long;
int i;
XMEMSET(Z, 0, AES_BLOCK_SIZE);
//XMEMCPY(V, X, AES_BLOCK_SIZE);
for (i = 0; i < AES_BLOCK_SIZE; i++) {
uint32_t y = 0x800000 | Y[i];
for (;;) { // for every bit in Y[i], from msb to lsb
if (y & 0x80) {
xorbuf_aligned_AES_BLOCK_SIZE(Z, X); // was V, not X
}
RIGHTSHIFTX(X); // was V, not X
y = y << 1;
if ((int32_t)y < 0) // if bit 0x80000000 set = if 8 iterations done
break;
}
}
XMEMCPY(X, Z, AES_BLOCK_SIZE);
}
//bbox:
// for TLS AES-GCM, a (which is AAD) is always 13 bytes long, and bbox code provides
// extra 3 zeroed bytes, making it a[16], or a[AES_BLOCK_SIZE].
// Resulting auth tag in s[] is also always AES_BLOCK_SIZE bytes.
//
// This allows some simplifications.
#define aSz 13
#define sSz AES_BLOCK_SIZE
void FAST_FUNC aesgcm_GHASH(byte* h,
const byte* a, //unsigned aSz,
const byte* c, unsigned cSz,
byte* s //, unsigned sSz
)
{
byte x[AES_BLOCK_SIZE] ALIGNED_long;
// byte scratch[AES_BLOCK_SIZE] ALIGNED_long;
unsigned blocks, partial;
//was: byte* h = aes->H;
//XMEMSET(x, 0, AES_BLOCK_SIZE);
/* Hash in A, the Additional Authentication Data */
// if (aSz != 0 && a != NULL) {
// blocks = aSz / AES_BLOCK_SIZE;
// partial = aSz % AES_BLOCK_SIZE;
// while (blocks--) {
//xorbuf(x, a, AES_BLOCK_SIZE);
XMEMCPY(x, a, AES_BLOCK_SIZE);// memcpy(x,a) = memset(x,0)+xorbuf(x,a)
GMULT(x, h);
// a += AES_BLOCK_SIZE;
// }
// if (partial != 0) {
// XMEMSET(scratch, 0, AES_BLOCK_SIZE);
// XMEMCPY(scratch, a, partial);
// xorbuf(x, scratch, AES_BLOCK_SIZE);
// GMULT(x, h);
// }
// }
/* Hash in C, the Ciphertext */
if (cSz != 0 /*&& c != NULL*/) {
blocks = cSz / AES_BLOCK_SIZE;
partial = cSz % AES_BLOCK_SIZE;
while (blocks--) {
if (BB_UNALIGNED_MEMACCESS_OK) // c is not guaranteed to be aligned
xorbuf_aligned_AES_BLOCK_SIZE(x, c);
else
xorbuf(x, c, AES_BLOCK_SIZE);
GMULT(x, h);
c += AES_BLOCK_SIZE;
}
if (partial != 0) {
//XMEMSET(scratch, 0, AES_BLOCK_SIZE);
//XMEMCPY(scratch, c, partial);
//xorbuf(x, scratch, AES_BLOCK_SIZE);
xorbuf(x, c, partial);//same result as above
GMULT(x, h);
}
}
/* Hash in the lengths of A and C in bits */
//FlattenSzInBits(&scratch[0], aSz);
//FlattenSzInBits(&scratch[8], cSz);
//xorbuf_aligned_AES_BLOCK_SIZE(x, scratch);
// simpler:
#define P32(v) ((uint32_t*)v)
//P32(x)[0] ^= 0;
P32(x)[1] ^= SWAP_BE32(aSz * 8);
//P32(x)[2] ^= 0;
P32(x)[3] ^= SWAP_BE32(cSz * 8);
#undef P32
GMULT(x, h);
/* Copy the result into s. */
XMEMCPY(s, x, sSz);
}