mirrors/openssl
0
0
Fork 0
mirror of https://github.com/openssl/openssl.git synced 2025-01-09 19:43:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

AES CTR-DRGB: performance improvement

Browse Source 
Optimize the the AES-based implementation of the CTR_DRBG
construction, see 10.2.1 in [1].
Due to the optimizations, the code may deviate (more) from the
pseudocode in [1], but it is functional equivalence being decisive
for compliance:

"All DRBG mechanisms and algorithms are described in this document
in pseudocode, which is intended to explain functionality.
The pseudocode is not intended to constrain real-world
implementations." [9 in [1]].

The following optimizations are done:

- Replace multiple plain AES encryptions by a single AES-ECB
  encryption of a corresponding pre-initialized buffer, where
  possible.
  This allows platform-specific AES-ECB support to
  be used and reduces the overhead of multiple EVP calls.

- Replace the generate operation loop (which is a counter
  increment followed by a plain AES encryption) by a
  loop which does a plain AES encryption followed by
  a counter increment. The latter loop is just a description
  of AES-CTR, so we replace it by a single AES-CTR
  encryption.
  This allows for platform-specific AES-CTR support to be used
  and reduces the overhead of multiple EVP calls.
  This change, that is, going from a pre- to a post- counter
  increment, requires the counter in the internal state
  to be kept at "+1" (compared to the pseudocode in [1])
  such that it is in the correct state, when a generate
  operation is called.
  That in turn also requires all other operations to be
  changed from pre- to post-increment to keep functional
  equivalence.

[1] NIST SP 800-90A Revision 1

Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10457)
This commit is contained in:
Patrick Steuer 2019-11-15 23:27:09 +01:00
parent 4b5371913e
commit 28bdbe1aaa
2 changed files with 139 additions and 82 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

215
crypto/rand/drbg_ctr.c
View File

@ -12,6 +12,7 @@
#include <openssl/crypto.h>
#include <openssl/err.h>
#include <openssl/rand.h>
#include "crypto/modes.h"
#include "internal/thread_once.h"
#include "rand_local.h"
@ -65,15 +66,15 @@ static void ctr_XOR(RAND_DRBG_CTR *ctr, const unsigned char *in, size_t inlen)
* Process a complete block using BCC algorithm of SP 800-90A 10.3.3
*/
__owur static int ctr_BCC_block(RAND_DRBG_CTR *ctr, unsigned char *out,
const unsigned char *in)
const unsigned char *in, int len)
{
int i, outlen = AES_BLOCK_SIZE;
for (i = 0; i < 16; i++)
for (i = 0; i < len; i++)
out[i] ^= in[i];
if (!EVP_CipherUpdate(ctr->ctx_df, out, &outlen, out, AES_BLOCK_SIZE)
|| outlen != AES_BLOCK_SIZE)
if (!EVP_CipherUpdate(ctr->ctx_df, out, &outlen, out, len)
|| outlen != len)
return 0;
return 1;
}
@ -84,12 +85,16 @@ __owur static int ctr_BCC_block(RAND_DRBG_CTR *ctr, unsigned char *out,
*/
__owur static int ctr_BCC_blocks(RAND_DRBG_CTR *ctr, const unsigned char *in)
{
if (!ctr_BCC_block(ctr, ctr->KX, in)
|| !ctr_BCC_block(ctr, ctr->KX + 16, in))
return 0;
if (ctr->keylen != 16 && !ctr_BCC_block(ctr, ctr->KX + 32, in))
return 0;
return 1;
unsigned char in_tmp[48];
unsigned char num_of_blk = 2;
memcpy(in_tmp, in, 16);
memcpy(in_tmp + 16, in, 16);
if (ctr->keylen != 16) {
memcpy(in_tmp + 32, in, 16);
num_of_blk = 3;
}
return ctr_BCC_block(ctr, ctr->KX, in_tmp, AES_BLOCK_SIZE * num_of_blk);
}
/*
@ -98,19 +103,14 @@ __owur static int ctr_BCC_blocks(RAND_DRBG_CTR *ctr, const unsigned char *in)
*/
__owur static int ctr_BCC_init(RAND_DRBG_CTR *ctr)
{
unsigned char bltmp[48] = {0};
unsigned char num_of_blk;
memset(ctr->KX, 0, 48);
memset(ctr->bltmp, 0, 16);
if (!ctr_BCC_block(ctr, ctr->KX, ctr->bltmp))
return 0;
ctr->bltmp[3] = 1;
if (!ctr_BCC_block(ctr, ctr->KX + 16, ctr->bltmp))
return 0;
if (ctr->keylen != 16) {
ctr->bltmp[3] = 2;
if (!ctr_BCC_block(ctr, ctr->KX + 32, ctr->bltmp))
return 0;
}
return 1;
num_of_blk = ctr->keylen == 16 ? 2 : 3;
bltmp[(AES_BLOCK_SIZE * 1) + 3] = 1;
bltmp[(AES_BLOCK_SIZE * 2) + 3] = 2;
return ctr_BCC_block(ctr, ctr->KX, bltmp, num_of_blk * AES_BLOCK_SIZE);
}
/*
@ -199,20 +199,20 @@ __owur static int ctr_df(RAND_DRBG_CTR *ctr,
|| !ctr_BCC_final(ctr))
return 0;
/* Set up key K */
if (!EVP_CipherInit_ex(ctr->ctx, ctr->cipher, NULL, ctr->KX, NULL, 1))
if (!EVP_CipherInit_ex(ctr->ctx_ecb, NULL, NULL, ctr->KX, NULL, -1))
return 0;
/* X follows key K */
if (!EVP_CipherUpdate(ctr->ctx, ctr->KX, &outlen, ctr->KX + ctr->keylen,
if (!EVP_CipherUpdate(ctr->ctx_ecb, ctr->KX, &outlen, ctr->KX + ctr->keylen,
AES_BLOCK_SIZE)
|| outlen != AES_BLOCK_SIZE)
return 0;
if (!EVP_CipherUpdate(ctr->ctx, ctr->KX + 16, &outlen, ctr->KX,
if (!EVP_CipherUpdate(ctr->ctx_ecb, ctr->KX + 16, &outlen, ctr->KX,
AES_BLOCK_SIZE)
|| outlen != AES_BLOCK_SIZE)
return 0;
if (ctr->keylen != 16)
if (!EVP_CipherUpdate(ctr->ctx, ctr->KX + 32, &outlen, ctr->KX + 16,
AES_BLOCK_SIZE)
if (!EVP_CipherUpdate(ctr->ctx_ecb, ctr->KX + 32, &outlen,
ctr->KX + 16, AES_BLOCK_SIZE)
|| outlen != AES_BLOCK_SIZE)
return 0;
return 1;
@ -231,31 +231,25 @@ __owur static int ctr_update(RAND_DRBG *drbg,
{
RAND_DRBG_CTR *ctr = &drbg->data.ctr;
int outlen = AES_BLOCK_SIZE;
unsigned char V_tmp[48], out[48];
unsigned char len;
/* correct key is already set up. */
memcpy(V_tmp, ctr->V, 16);
inc_128(ctr);
if (!EVP_CipherUpdate(ctr->ctx, ctr->K, &outlen, ctr->V, AES_BLOCK_SIZE)
|| outlen != AES_BLOCK_SIZE)
return 0;
/* If keylen longer than 128 bits need extra encrypt */
if (ctr->keylen != 16) {
memcpy(V_tmp + 16, ctr->V, 16);
if (ctr->keylen == 16) {
len = 32;
} else {
inc_128(ctr);
if (!EVP_CipherUpdate(ctr->ctx, ctr->K+16, &outlen, ctr->V,
AES_BLOCK_SIZE)
|| outlen != AES_BLOCK_SIZE)
return 0;
memcpy(V_tmp + 32, ctr->V, 16);
len = 48;
}
inc_128(ctr);
if (!EVP_CipherUpdate(ctr->ctx, ctr->V, &outlen, ctr->V, AES_BLOCK_SIZE)
|| outlen != AES_BLOCK_SIZE)
if (!EVP_CipherUpdate(ctr->ctx_ecb, out, &outlen, V_tmp, len)
|| outlen != len)
return 0;
/* If 192 bit key part of V is on end of K */
if (ctr->keylen == 24) {
memcpy(ctr->V + 8, ctr->V, 8);
memcpy(ctr->V, ctr->K + 24, 8);
}
memcpy(ctr->K, out, ctr->keylen);
memcpy(ctr->V, out + ctr->keylen, 16);
if ((drbg->flags & RAND_DRBG_FLAG_CTR_NO_DF) == 0) {
/* If no input reuse existing derived value */
@ -270,7 +264,8 @@ __owur static int ctr_update(RAND_DRBG *drbg,
ctr_XOR(ctr, in2, in2len);
}
if (!EVP_CipherInit_ex(ctr->ctx, ctr->cipher, NULL, ctr->K, NULL, 1))
if (!EVP_CipherInit_ex(ctr->ctx_ecb, NULL, NULL, ctr->K, NULL, -1)
|| !EVP_CipherInit_ex(ctr->ctx_ctr, NULL, NULL, ctr->K, NULL, -1))
return 0;
return 1;
}
@ -287,8 +282,10 @@ __owur static int drbg_ctr_instantiate(RAND_DRBG *drbg,
memset(ctr->K, 0, sizeof(ctr->K));
memset(ctr->V, 0, sizeof(ctr->V));
if (!EVP_CipherInit_ex(ctr->ctx, ctr->cipher, NULL, ctr->K, NULL, 1))
if (!EVP_CipherInit_ex(ctr->ctx_ecb, NULL, NULL, ctr->K, NULL, -1))
return 0;
inc_128(ctr);
if (!ctr_update(drbg, entropy, entropylen, pers, perslen, nonce, noncelen))
return 0;
return 1;
@ -298,20 +295,40 @@ __owur static int drbg_ctr_reseed(RAND_DRBG *drbg,
const unsigned char *entropy, size_t entropylen,
const unsigned char *adin, size_t adinlen)
{
RAND_DRBG_CTR *ctr = &drbg->data.ctr;
if (entropy == NULL)
return 0;
inc_128(ctr);
if (!ctr_update(drbg, entropy, entropylen, adin, adinlen, NULL, 0))
return 0;
return 1;
}
static void ctr96_inc(unsigned char *counter)
{
u32 n = 12, c = 1;
do {
--n;
c += counter[n];
counter[n] = (u8)c;
c >>= 8;
} while (n);
}
__owur static int drbg_ctr_generate(RAND_DRBG *drbg,
unsigned char *out, size_t outlen,
const unsigned char *adin, size_t adinlen)
{
RAND_DRBG_CTR *ctr = &drbg->data.ctr;
unsigned int ctr32, blocks;
int outl, buflen;
if (adin != NULL && adinlen != 0) {
inc_128(ctr);
if (!ctr_update(drbg, adin, adinlen, NULL, 0, NULL, 0))
return 0;
/* This means we reuse derived value */
@ -323,28 +340,51 @@ __owur static int drbg_ctr_generate(RAND_DRBG *drbg,
adinlen = 0;
}
for ( ; ; ) {
int outl = AES_BLOCK_SIZE;
inc_128(ctr);
if (outlen == 0) {
inc_128(ctr);
if (outlen < 16) {
/* Use K as temp space as it will be updated */
if (!EVP_CipherUpdate(ctr->ctx, ctr->K, &outl, ctr->V,
AES_BLOCK_SIZE)
|| outl != AES_BLOCK_SIZE)
return 0;
memcpy(out, ctr->K, outlen);
break;
}
if (!EVP_CipherUpdate(ctr->ctx, out, &outl, ctr->V, AES_BLOCK_SIZE)
|| outl != AES_BLOCK_SIZE)
if (!ctr_update(drbg, adin, adinlen, NULL, 0, NULL, 0))
return 0;
out += 16;
outlen -= 16;
if (outlen == 0)
break;
return 1;
}
memset(out, 0, outlen);
do {
if (!EVP_CipherInit_ex(ctr->ctx_ctr,
NULL, NULL, NULL, ctr->V, -1))
return 0;
/*-
* outlen has type size_t while EVP_CipherUpdate takes an
* int argument and thus cannot be guaranteed to process more
* than 2^31-1 bytes at a time. We process such huge generate
* requests in 2^30 byte chunks, which is the greatest multiple
* of AES block size lower than or equal to 2^31-1.
*/
buflen = outlen > (1U << 30) ? (1U << 30) : outlen;
blocks = (buflen + 15) / 16;
ctr32 = GETU32(ctr->V + 12) + blocks;
if (ctr32 < blocks) {
/* 32-bit counter overflow into V. */
blocks -= ctr32;
buflen = blocks * 16;
ctr32 = 0;
ctr96_inc(ctr->V);
}
PUTU32(ctr->V + 12, ctr32);
if (!EVP_CipherUpdate(ctr->ctx_ctr, out, &outl, out, buflen)
|| outl != buflen)
return 0;
out += buflen;
outlen -= buflen;
} while (outlen);
if (!ctr_update(drbg, adin, adinlen, NULL, 0, NULL, 0))
return 0;
return 1;
@ -352,9 +392,11 @@ __owur static int drbg_ctr_generate(RAND_DRBG *drbg,
static int drbg_ctr_uninstantiate(RAND_DRBG *drbg)
{
EVP_CIPHER_CTX_free(drbg->data.ctr.ctx);
EVP_CIPHER_CTX_free(drbg->data.ctr.ctx_ecb);
EVP_CIPHER_CTX_free(drbg->data.ctr.ctx_ctr);
EVP_CIPHER_CTX_free(drbg->data.ctr.ctx_df);
EVP_CIPHER_free(drbg->data.ctr.cipher);
EVP_CIPHER_free(drbg->data.ctr.cipher_ecb);
EVP_CIPHER_free(drbg->data.ctr.cipher_ctr);
OPENSSL_cleanse(&drbg->data.ctr, sizeof(drbg->data.ctr));
return 1;
}
@ -370,7 +412,8 @@ int drbg_ctr_init(RAND_DRBG *drbg)
{
RAND_DRBG_CTR *ctr = &drbg->data.ctr;
size_t keylen;
EVP_CIPHER *cipher = NULL;
EVP_CIPHER *cipher_ecb = NULL;
EVP_CIPHER *cipher_ctr = NULL;
switch (drbg->type) {
default:
@ -378,30 +421,41 @@ int drbg_ctr_init(RAND_DRBG *drbg)
return 0;
case NID_aes_128_ctr:
keylen = 16;
cipher = EVP_CIPHER_fetch(drbg->libctx, "AES-128-ECB", "");
cipher_ecb = EVP_CIPHER_fetch(drbg->libctx, "AES-128-ECB", "");
cipher_ctr = EVP_CIPHER_fetch(drbg->libctx, "AES-128-CTR", "");
break;
case NID_aes_192_ctr:
keylen = 24;
cipher = EVP_CIPHER_fetch(drbg->libctx, "AES-192-ECB", "");
cipher_ecb = EVP_CIPHER_fetch(drbg->libctx, "AES-192-ECB", "");
cipher_ctr = EVP_CIPHER_fetch(drbg->libctx, "AES-192-CTR", "");
break;
case NID_aes_256_ctr:
keylen = 32;
cipher = EVP_CIPHER_fetch(drbg->libctx, "AES-256-ECB", "");
cipher_ecb = EVP_CIPHER_fetch(drbg->libctx, "AES-256-ECB", "");
cipher_ctr = EVP_CIPHER_fetch(drbg->libctx, "AES-256-CTR", "");
break;
}
if (cipher == NULL)
if (cipher_ecb == NULL || cipher_ctr == NULL)
return 0;
EVP_CIPHER_free(ctr->cipher);
ctr->cipher = cipher;
drbg->meth = &drbg_ctr_meth;
EVP_CIPHER_free(ctr->cipher_ecb);
ctr->cipher_ecb = cipher_ecb;
EVP_CIPHER_free(ctr->cipher_ctr);
ctr->cipher_ctr = cipher_ctr;
ctr->keylen = keylen;
if (ctr->ctx == NULL)
ctr->ctx = EVP_CIPHER_CTX_new();
if (ctr->ctx == NULL)
if (ctr->ctx_ecb == NULL)
ctr->ctx_ecb = EVP_CIPHER_CTX_new();
if (ctr->ctx_ctr == NULL)
ctr->ctx_ctr = EVP_CIPHER_CTX_new();
if (ctr->ctx_ecb == NULL || ctr->ctx_ctr == NULL
|| !EVP_CipherInit_ex(ctr->ctx_ecb,
ctr->cipher_ecb, NULL, NULL, NULL, 1)
|| !EVP_CipherInit_ex(ctr->ctx_ctr,
ctr->cipher_ctr, NULL, NULL, NULL, 1))
return 0;
drbg->meth = &drbg_ctr_meth;
drbg->strength = keylen * 8;
drbg->seedlen = keylen + 16;
@ -419,7 +473,8 @@ int drbg_ctr_init(RAND_DRBG *drbg)
if (ctr->ctx_df == NULL)
return 0;
/* Set key schedule for df_key */
if (!EVP_CipherInit_ex(ctr->ctx_df, ctr->cipher, NULL, df_key, NULL, 1))
if (!EVP_CipherInit_ex(ctr->ctx_df,
ctr->cipher_ecb, NULL, df_key, NULL, 1))
return 0;
drbg->min_entropylen = ctr->keylen;

6
crypto/rand/rand_local.h
View File

@ -175,9 +175,11 @@ typedef struct rand_drbg_hmac_st {
* The state of a DRBG AES-CTR.
*/
typedef struct rand_drbg_ctr_st {
EVP_CIPHER_CTX *ctx;
EVP_CIPHER_CTX *ctx_ecb;
EVP_CIPHER_CTX *ctx_ctr;
EVP_CIPHER_CTX *ctx_df;
EVP_CIPHER *cipher;
EVP_CIPHER *cipher_ecb;
EVP_CIPHER *cipher_ctr;
size_t keylen;
unsigned char K[32];
unsigned char V[16];