DeforaNetworks/hashcat
3
0
Fork 0
Code Releases Activity

Merge pull request #2906 from hashcat/master

Browse Source 
Backports
This commit is contained in:
Jens Steube
2021-07-27 09:48:56 +02:00
committed by GitHub
parent b66c093c17 e5ac568afb
commit f6de3e61e0
16 changed files with 1378 additions and 123 deletions
Download Patch File Download Diff File Expand all files Collapse all files
OpenCL/inc_ecc_secp256k1.cl
+6 -2
View File
@@ -124,7 +124,9 @@ DECLSPEC u32 sub (u32 *r, const u32 *a, const u32 *b)
: "r"(a[0]), "r"(a[1]), "r"(a[2]), "r"(a[3]), "r"(a[4]), "r"(a[5]), "r"(a[6]), "r"(a[7]),
"r"(b[0]), "r"(b[1]), "r"(b[2]), "r"(b[3]), "r"(b[4]), "r"(b[5]), "r"(b[6]), "r"(b[7])
);
#elif (defined IS_AMD || defined IS_HIP) && HAS_VSUB == 1 && HAS_VSUBB == 1
// HIP doesnt support these so we stick to OpenCL (aka IS_AMD) - is also faster without asm
//#elif (defined IS_AMD || defined IS_HIP) && HAS_VSUB == 1 && HAS_VSUBB == 1
#elif 0
__asm__ __volatile__
(
"V_SUB_U32 %0, %9, %17;"
@@ -176,7 +178,9 @@ DECLSPEC u32 add (u32 *r, const u32 *a, const u32 *b)
: "r"(a[0]), "r"(a[1]), "r"(a[2]), "r"(a[3]), "r"(a[4]), "r"(a[5]), "r"(a[6]), "r"(a[7]),
"r"(b[0]), "r"(b[1]), "r"(b[2]), "r"(b[3]), "r"(b[4]), "r"(b[5]), "r"(b[6]), "r"(b[7])
);
#elif (defined IS_AMD || defined IS_HIP) && HAS_VADD == 1 && HAS_VADDC == 1
// HIP doesnt support these so we stick to OpenCL (aka IS_AMD) - is also faster without asm
//#elif (defined IS_AMD || defined IS_HIP) && HAS_VSUB == 1 && HAS_VSUBB == 1
#elif 0
__asm__ __volatile__
(
"V_ADD_U32 %0, %9, %17;"
OpenCL/m09000-pure.cl
+105 -36
View File
@@ -310,6 +310,51 @@ CONSTANT_VK u32a c_pbox[18] =
0x9216d5d9, 0x8979fb1b
};
// Yes, works only with CUDA atm
#ifdef DYNAMIC_LOCAL
#define BCRYPT_AVOID_BANK_CONFLICTS
#endif
#ifdef BCRYPT_AVOID_BANK_CONFLICTS
// access pattern: minimize bank ID based on thread ID but thread ID is not saved from computation
#define KEY32(lid,key) (((key) * FIXED_LOCAL_SIZE) + (lid))
DECLSPEC u32 GET_KEY32 (LOCAL_AS u32 *S, const u64 key)
{
const u64 lid = get_local_id (0);
return S[KEY32 (lid, key)];
}
DECLSPEC void SET_KEY32 (LOCAL_AS u32 *S, const u64 key, const u32 val)
{
const u64 lid = get_local_id (0);
S[KEY32 (lid, key)] = val;
}
#undef KEY32
#else
// access pattern: linear access with S offset already set to right offset based on thread ID saving it from compuation
// makes sense if there are not thread ID's (for instance on CPU)
DECLSPEC inline u32 GET_KEY32 (LOCAL_AS u32 *S, const u64 key)
{
return S[key];
}
DECLSPEC inline void SET_KEY32 (LOCAL_AS u32 *S, const u64 key, const u32 val)
{
S[key] = val;
}
#endif
#define BF_ROUND(L,R,N) \
{ \
u32 tmp; \
@@ -319,10 +364,10 @@ CONSTANT_VK u32a c_pbox[18] =
const u32 r2 = unpack_v8b_from_v32_S ((L)); \
const u32 r3 = unpack_v8a_from_v32_S ((L)); \
\
tmp = S0[r0]; \
tmp += S1[r1]; \
tmp ^= S2[r2]; \
tmp += S3[r3]; \
tmp = GET_KEY32 (S0, r0); \
tmp += GET_KEY32 (S1, r1); \
tmp ^= GET_KEY32 (S2, r2); \
tmp += GET_KEY32 (S3, r3); \
\
(R) ^= tmp ^ P[(N)]; \
}
@@ -357,6 +402,10 @@ CONSTANT_VK u32a c_pbox[18] =
L ^= P[17]; \
}
#ifdef DYNAMIC_LOCAL
extern __shared__ u32 S[];
#endif
KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m09000_init (KERN_ATTR_TMPS (pwsafe2_tmp_t))
{
/**
@@ -471,22 +520,33 @@ KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m09000_init (KERN_ATTR_TMPS
P[i] = c_pbox[i];
}
#ifdef DYNAMIC_LOCAL
// from host
#else
LOCAL_VK u32 S0_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S1_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S2_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S3_all[FIXED_LOCAL_SIZE][256];
#endif
#ifdef BCRYPT_AVOID_BANK_CONFLICTS
LOCAL_AS u32 *S0 = S + (FIXED_LOCAL_SIZE * 256 * 0);
LOCAL_AS u32 *S1 = S + (FIXED_LOCAL_SIZE * 256 * 1);
LOCAL_AS u32 *S2 = S + (FIXED_LOCAL_SIZE * 256 * 2);
LOCAL_AS u32 *S3 = S + (FIXED_LOCAL_SIZE * 256 * 3);
#else
LOCAL_AS u32 *S0 = S0_all[lid];
LOCAL_AS u32 *S1 = S1_all[lid];
LOCAL_AS u32 *S2 = S2_all[lid];
LOCAL_AS u32 *S3 = S3_all[lid];
#endif
for (u32 i = 0; i < 256; i++)
{
S0[i] = c_sbox0[i];
S1[i] = c_sbox1[i];
S2[i] = c_sbox2[i];
S3[i] = c_sbox3[i];
SET_KEY32 (S0, i, c_sbox0[i]);
SET_KEY32 (S1, i, c_sbox1[i]);
SET_KEY32 (S2, i, c_sbox2[i]);
SET_KEY32 (S3, i, c_sbox3[i]);
}
for (u32 i = 0; i < 18; i++)
@@ -509,59 +569,59 @@ KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m09000_init (KERN_ATTR_TMPS
{
BF_ENCRYPT (L0, R0);
S0[i + 0] = L0;
S0[i + 1] = R0;
SET_KEY32 (S0, i + 0, L0);
SET_KEY32 (S0, i + 1, R0);
BF_ENCRYPT (L0, R0);
S0[i + 2] = L0;
S0[i + 3] = R0;
SET_KEY32 (S0, i + 2, L0);
SET_KEY32 (S0, i + 3, R0);
}
for (u32 i = 0; i < 256; i += 4)
{
BF_ENCRYPT (L0, R0);
S1[i + 0] = L0;
S1[i + 1] = R0;
SET_KEY32 (S1, i + 0, L0);
SET_KEY32 (S1, i + 1, R0);
BF_ENCRYPT (L0, R0);
S1[i + 2] = L0;
S1[i + 3] = R0;
SET_KEY32 (S1, i + 2, L0);
SET_KEY32 (S1, i + 3, R0);
}
for (u32 i = 0; i < 256; i += 4)
{
BF_ENCRYPT (L0, R0);
S2[i + 0] = L0;
S2[i + 1] = R0;
SET_KEY32 (S2, i + 0, L0);
SET_KEY32 (S2, i + 1, R0);
BF_ENCRYPT (L0, R0);
S2[i + 2] = L0;
S2[i + 3] = R0;
SET_KEY32 (S2, i + 2, L0);
SET_KEY32 (S2, i + 3, R0);
}
for (u32 i = 0; i < 256; i += 4)
{
BF_ENCRYPT (L0, R0);
S3[i + 0] = L0;
S3[i + 1] = R0;
SET_KEY32 (S3, i + 0, L0);
SET_KEY32 (S3, i + 1, R0);
BF_ENCRYPT (L0, R0);
S3[i + 2] = L0;
S3[i + 3] = R0;
SET_KEY32 (S3, i + 2, L0);
SET_KEY32 (S3, i + 3, R0);
}
// store
tmps[gid].digest[0] = salt_buf[0];
tmps[gid].digest[1] = salt_buf[1];
// store
for (u32 i = 0; i < 18; i++)
{
tmps[gid].P[i] = P[i];
@@ -569,10 +629,10 @@ KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m09000_init (KERN_ATTR_TMPS
for (u32 i = 0; i < 256; i++)
{
tmps[gid].S0[i] = S0[i];
tmps[gid].S1[i] = S1[i];
tmps[gid].S2[i] = S2[i];
tmps[gid].S3[i] = S3[i];
tmps[gid].S0[i] = GET_KEY32 (S0, i);
tmps[gid].S1[i] = GET_KEY32 (S1, i);
tmps[gid].S2[i] = GET_KEY32 (S2, i);
tmps[gid].S3[i] = GET_KEY32 (S3, i);
}
}
@@ -602,22 +662,33 @@ KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m09000_loop (KERN_ATTR_TMPS
P[i] = tmps[gid].P[i];
}
#ifdef DYNAMIC_LOCAL
// from host
#else
LOCAL_VK u32 S0_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S1_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S2_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S3_all[FIXED_LOCAL_SIZE][256];
#endif
#ifdef BCRYPT_AVOID_BANK_CONFLICTS
LOCAL_AS u32 *S0 = S + (FIXED_LOCAL_SIZE * 256 * 0);
LOCAL_AS u32 *S1 = S + (FIXED_LOCAL_SIZE * 256 * 1);
LOCAL_AS u32 *S2 = S + (FIXED_LOCAL_SIZE * 256 * 2);
LOCAL_AS u32 *S3 = S + (FIXED_LOCAL_SIZE * 256 * 3);
#else
LOCAL_AS u32 *S0 = S0_all[lid];
LOCAL_AS u32 *S1 = S1_all[lid];
LOCAL_AS u32 *S2 = S2_all[lid];
LOCAL_AS u32 *S3 = S3_all[lid];
#endif
for (u32 i = 0; i < 256; i++)
{
S0[i] = tmps[gid].S0[i];
S1[i] = tmps[gid].S1[i];
S2[i] = tmps[gid].S2[i];
S3[i] = tmps[gid].S3[i];
SET_KEY32 (S0, i, tmps[gid].S0[i]);
SET_KEY32 (S1, i, tmps[gid].S1[i]);
SET_KEY32 (S2, i, tmps[gid].S2[i]);
SET_KEY32 (S3, i, tmps[gid].S3[i]);
}
// loop
@@ -630,8 +701,6 @@ KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m09000_loop (KERN_ATTR_TMPS
BF_ENCRYPT (L0, R0);
}
// store
tmps[gid].digest[0] = L0;
tmps[gid].digest[1] = R0;
}
OpenCL/m18600-pure.cl
+89 -29
View File
@@ -319,6 +319,51 @@ CONSTANT_VK u32a c_pbox[18] =
0x9216d5d9, 0x8979fb1b
};
// Yes, works only with CUDA atm
#ifdef DYNAMIC_LOCAL
#define BCRYPT_AVOID_BANK_CONFLICTS
#endif
#ifdef BCRYPT_AVOID_BANK_CONFLICTS
// access pattern: minimize bank ID based on thread ID but thread ID is not saved from computation
#define KEY32(lid,key) (((key) * FIXED_LOCAL_SIZE_COMP) + (lid))
DECLSPEC u32 GET_KEY32 (LOCAL_AS u32 *S, const u64 key)
{
const u64 lid = get_local_id (0);
return S[KEY32 (lid, key)];
}
DECLSPEC void SET_KEY32 (LOCAL_AS u32 *S, const u64 key, const u32 val)
{
const u64 lid = get_local_id (0);
S[KEY32 (lid, key)] = val;
}
#undef KEY32
#else
// access pattern: linear access with S offset already set to right offset based on thread ID saving it from compuation
// makes sense if there are not thread ID's (for instance on CPU)
DECLSPEC inline u32 GET_KEY32 (LOCAL_AS u32 *S, const u64 key)
{
return S[key];
}
DECLSPEC inline void SET_KEY32 (LOCAL_AS u32 *S, const u64 key, const u32 val)
{
S[key] = val;
}
#endif
#define BF_ROUND(L,R,N) \
{ \
u32 tmp; \
@@ -328,10 +373,10 @@ CONSTANT_VK u32a c_pbox[18] =
const u32 r2 = unpack_v8b_from_v32_S ((L)); \
const u32 r3 = unpack_v8a_from_v32_S ((L)); \
\
tmp = S0[r0]; \
tmp += S1[r1]; \
tmp ^= S2[r2]; \
tmp += S3[r3]; \
tmp = GET_KEY32 (S0, r0); \
tmp += GET_KEY32 (S1, r1); \
tmp ^= GET_KEY32 (S2, r2); \
tmp += GET_KEY32 (S3, r3); \
\
(R) ^= tmp ^ P[(N)]; \
}
@@ -366,6 +411,10 @@ CONSTANT_VK u32a c_pbox[18] =
L ^= P[17]; \
}
#ifdef DYNAMIC_LOCAL
extern __shared__ u32 S[];
#endif
DECLSPEC void hmac_sha1_run_V (u32x *w0, u32x *w1, u32x *w2, u32x *w3, u32x *ipad, u32x *opad, u32x *digest)
{
digest[0] = ipad[0];
@@ -586,7 +635,7 @@ KERNEL_FQ void m18600_loop (KERN_ATTR_TMPS_ESALT (odf11_tmp_t, odf11_t))
}
}
KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m18600_comp (KERN_ATTR_TMPS_ESALT (odf11_tmp_t, odf11_t))
KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE_COMP) m18600_comp (KERN_ATTR_TMPS_ESALT (odf11_tmp_t, odf11_t))
{
const u64 gid = get_global_id (0);
const u64 lid = get_local_id (0);
@@ -616,22 +665,33 @@ KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m18600_comp (KERN_ATTR_TMPS_
P[i] = c_pbox[i] ^ ukey[i % 4];
}
LOCAL_VK u32 S0_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S1_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S2_all[FIXED_LOCAL_SIZE][256];
LOCAL_VK u32 S3_all[FIXED_LOCAL_SIZE][256];
#ifdef DYNAMIC_LOCAL
// from host
#else
LOCAL_VK u32 S0_all[FIXED_LOCAL_SIZE_COMP][256];
LOCAL_VK u32 S1_all[FIXED_LOCAL_SIZE_COMP][256];
LOCAL_VK u32 S2_all[FIXED_LOCAL_SIZE_COMP][256];
LOCAL_VK u32 S3_all[FIXED_LOCAL_SIZE_COMP][256];
#endif
#ifdef BCRYPT_AVOID_BANK_CONFLICTS
LOCAL_AS u32 *S0 = S + (FIXED_LOCAL_SIZE_COMP * 256 * 0);
LOCAL_AS u32 *S1 = S + (FIXED_LOCAL_SIZE_COMP * 256 * 1);
LOCAL_AS u32 *S2 = S + (FIXED_LOCAL_SIZE_COMP * 256 * 2);
LOCAL_AS u32 *S3 = S + (FIXED_LOCAL_SIZE_COMP * 256 * 3);
#else
LOCAL_AS u32 *S0 = S0_all[lid];
LOCAL_AS u32 *S1 = S1_all[lid];
LOCAL_AS u32 *S2 = S2_all[lid];
LOCAL_AS u32 *S3 = S3_all[lid];
#endif
for (u32 i = 0; i < 256; i++)
{
S0[i] = c_sbox0[i];
S1[i] = c_sbox1[i];
S2[i] = c_sbox2[i];
S3[i] = c_sbox3[i];
SET_KEY32 (S0, i, c_sbox0[i]);
SET_KEY32 (S1, i, c_sbox1[i]);
SET_KEY32 (S2, i, c_sbox2[i]);
SET_KEY32 (S3, i, c_sbox3[i]);
}
u32 L0 = 0;
@@ -649,52 +709,52 @@ KERNEL_FQ void FIXED_THREAD_COUNT(FIXED_LOCAL_SIZE) m18600_comp (KERN_ATTR_TMPS_
{
BF_ENCRYPT (L0, R0);
S0[i + 0] = L0;
S0[i + 1] = R0;
SET_KEY32 (S0, i + 0, L0);
SET_KEY32 (S0, i + 1, R0);
BF_ENCRYPT (L0, R0);
S0[i + 2] = L0;
S0[i + 3] = R0;
SET_KEY32 (S0, i + 2, L0);
SET_KEY32 (S0, i + 3, R0);
}
for (u32 i = 0; i < 256; i += 4)
{
BF_ENCRYPT (L0, R0);
S1[i + 0] = L0;
S1[i + 1] = R0;
SET_KEY32 (S1, i + 0, L0);
SET_KEY32 (S1, i + 1, R0);
BF_ENCRYPT (L0, R0);
S1[i + 2] = L0;
S1[i + 3] = R0;
SET_KEY32 (S1, i + 2, L0);
SET_KEY32 (S1, i + 3, R0);
}
for (u32 i = 0; i < 256; i += 4)
{
BF_ENCRYPT (L0, R0);
S2[i + 0] = L0;
S2[i + 1] = R0;
SET_KEY32 (S2, i + 0, L0);
SET_KEY32 (S2, i + 1, R0);
BF_ENCRYPT (L0, R0);
S2[i + 2] = L0;
S2[i + 3] = R0;
SET_KEY32 (S2, i + 2, L0);
SET_KEY32 (S2, i + 3, R0);
}
for (u32 i = 0; i < 256; i += 4)
{
BF_ENCRYPT (L0, R0);
S3[i + 0] = L0;
S3[i + 1] = R0;
SET_KEY32 (S3, i + 0, L0);
SET_KEY32 (S3, i + 1, R0);
BF_ENCRYPT (L0, R0);
S3[i + 2] = L0;
S3[i + 3] = R0;
SET_KEY32 (S3, i + 2, L0);
SET_KEY32 (S3, i + 3, R0);
}
GLOBAL_AS const odf11_t *es = &esalt_bufs[DIGESTS_OFFSET];
OpenCL/m25000-pure.cl
+590
View File
@@ -0,0 +1,590 @@
/**
* Author......: See docs/credits.txt
* License.....: MIT
*/
//#define NEW_SIMD_CODE
#ifdef KERNEL_STATIC
#include "inc_vendor.h"
#include "inc_types.h"
#include "inc_platform.cl"
#include "inc_common.cl"
#include "inc_simd.cl"
#include "inc_hash_md5.cl"
#include "inc_hash_sha1.cl"
#endif
#define COMPARE_S "inc_comp_single.cl"
#define COMPARE_M "inc_comp_multi.cl"
#define SNMPV3_SALT_MAX 1500
#define SNMPV3_ENGINEID_MAX 34
#define SNMPV3_MSG_AUTH_PARAMS_LEN 12
#define SNMPV3_ROUNDS 1048576
#define SNMPV3_MAX_PW_LENGTH 64
#define SNMPV3_TMP_ELEMS 4096 // 4096 = (256 (max pw length) * 64) / sizeof (u32)
#define SNMPV3_HASH_ELEMS_MD5 4
#define SNMPV3_HASH_ELEMS_SHA1 8 // 8 = aligned 5
#define SNMPV3_MAX_SALT_ELEMS 512 // 512 * 4 = 2048 > 1500, also has to be multiple of 64
#define SNMPV3_MAX_ENGINE_ELEMS 16 // 16 * 4 = 64 > 32, also has to be multiple of 64
#define SNMPV3_MAX_PNUM_ELEMS 4 // 4 * 4 = 16 > 9
typedef struct hmac_md5_tmp
{
u32 tmp_md5[SNMPV3_TMP_ELEMS];
u32 tmp_sha1[SNMPV3_TMP_ELEMS];
u32 h_md5[SNMPV3_HASH_ELEMS_MD5];
u32 h_sha1[SNMPV3_HASH_ELEMS_SHA1];
} hmac_md5_tmp_t;
typedef struct snmpv3
{
u32 salt_buf[SNMPV3_MAX_SALT_ELEMS];
u32 salt_len;
u32 engineID_buf[SNMPV3_MAX_ENGINE_ELEMS];
u32 engineID_len;
u32 packet_number[SNMPV3_MAX_PNUM_ELEMS];
} snmpv3_t;
KERNEL_FQ void m25000_init (KERN_ATTR_TMPS_ESALT (hmac_md5_tmp_t, snmpv3_t))
{
/**
* modifier
*/
const u64 gid = get_global_id (0);
if (gid >= gid_max) return;
/**
* base
*/
const u32 pw_len = pws[gid].pw_len;
u32 w[64] = { 0 };
for (u32 i = 0, idx = 0; i < pw_len; i += 4, idx += 1)
{
w[idx] = pws[gid].i[idx];
}
u8 *src_ptr = (u8 *) w;
// password 64 times, also swapped
u32 dst_buf[16];
u8 *dst_ptr = (u8 *) dst_buf;
int tmp_idx = 0;
for (int i = 0; i < 64; i++)
{
for (int j = 0; j < pw_len; j++)
{
const int dst_idx = tmp_idx & 63;
dst_ptr[dst_idx] = src_ptr[j];
// write to global memory every time 64 byte are written into cache
if (dst_idx == 63)
{
const int tmp_idx4 = (tmp_idx - 63) / 4;
// md5
tmps[gid].tmp_md5[tmp_idx4 + 0] = dst_buf[ 0];
tmps[gid].tmp_md5[tmp_idx4 + 1] = dst_buf[ 1];
tmps[gid].tmp_md5[tmp_idx4 + 2] = dst_buf[ 2];
tmps[gid].tmp_md5[tmp_idx4 + 3] = dst_buf[ 3];
tmps[gid].tmp_md5[tmp_idx4 + 4] = dst_buf[ 4];
tmps[gid].tmp_md5[tmp_idx4 + 5] = dst_buf[ 5];
tmps[gid].tmp_md5[tmp_idx4 + 6] = dst_buf[ 6];
tmps[gid].tmp_md5[tmp_idx4 + 7] = dst_buf[ 7];
tmps[gid].tmp_md5[tmp_idx4 + 8] = dst_buf[ 8];
tmps[gid].tmp_md5[tmp_idx4 + 9] = dst_buf[ 9];
tmps[gid].tmp_md5[tmp_idx4 + 10] = dst_buf[10];
tmps[gid].tmp_md5[tmp_idx4 + 11] = dst_buf[11];
tmps[gid].tmp_md5[tmp_idx4 + 12] = dst_buf[12];
tmps[gid].tmp_md5[tmp_idx4 + 13] = dst_buf[13];
tmps[gid].tmp_md5[tmp_idx4 + 14] = dst_buf[14];
tmps[gid].tmp_md5[tmp_idx4 + 15] = dst_buf[15];
// sha1
tmps[gid].tmp_sha1[tmp_idx4 + 0] = hc_swap32_S (dst_buf[ 0]);
tmps[gid].tmp_sha1[tmp_idx4 + 1] = hc_swap32_S (dst_buf[ 1]);
tmps[gid].tmp_sha1[tmp_idx4 + 2] = hc_swap32_S (dst_buf[ 2]);
tmps[gid].tmp_sha1[tmp_idx4 + 3] = hc_swap32_S (dst_buf[ 3]);
tmps[gid].tmp_sha1[tmp_idx4 + 4] = hc_swap32_S (dst_buf[ 4]);
tmps[gid].tmp_sha1[tmp_idx4 + 5] = hc_swap32_S (dst_buf[ 5]);
tmps[gid].tmp_sha1[tmp_idx4 + 6] = hc_swap32_S (dst_buf[ 6]);
tmps[gid].tmp_sha1[tmp_idx4 + 7] = hc_swap32_S (dst_buf[ 7]);
tmps[gid].tmp_sha1[tmp_idx4 + 8] = hc_swap32_S (dst_buf[ 8]);
tmps[gid].tmp_sha1[tmp_idx4 + 9] = hc_swap32_S (dst_buf[ 9]);
tmps[gid].tmp_sha1[tmp_idx4 + 10] = hc_swap32_S (dst_buf[10]);
tmps[gid].tmp_sha1[tmp_idx4 + 11] = hc_swap32_S (dst_buf[11]);
tmps[gid].tmp_sha1[tmp_idx4 + 12] = hc_swap32_S (dst_buf[12]);
tmps[gid].tmp_sha1[tmp_idx4 + 13] = hc_swap32_S (dst_buf[13]);
tmps[gid].tmp_sha1[tmp_idx4 + 14] = hc_swap32_S (dst_buf[14]);
tmps[gid].tmp_sha1[tmp_idx4 + 15] = hc_swap32_S (dst_buf[15]);
}
tmp_idx++;
}
}
// hash md5
tmps[gid].h_md5[0] = MD5M_A;
tmps[gid].h_md5[1] = MD5M_B;
tmps[gid].h_md5[2] = MD5M_C;
tmps[gid].h_md5[3] = MD5M_D;
// hash sha1
tmps[gid].h_sha1[0] = SHA1M_A;
tmps[gid].h_sha1[1] = SHA1M_B;
tmps[gid].h_sha1[2] = SHA1M_C;
tmps[gid].h_sha1[3] = SHA1M_D;
tmps[gid].h_sha1[4] = SHA1M_E;
}
KERNEL_FQ void m25000_loop (KERN_ATTR_TMPS_ESALT (hmac_md5_tmp_t, snmpv3_t))
{
/**
* base
*/
const u64 gid = get_global_id (0);
if (gid >= gid_max) return;
u32 h_md5[4];
h_md5[0] = tmps[gid].h_md5[0];
h_md5[1] = tmps[gid].h_md5[1];
h_md5[2] = tmps[gid].h_md5[2];
h_md5[3] = tmps[gid].h_md5[3];
u32 h_sha1[5];
h_sha1[0] = tmps[gid].h_sha1[0];
h_sha1[1] = tmps[gid].h_sha1[1];
h_sha1[2] = tmps[gid].h_sha1[2];
h_sha1[3] = tmps[gid].h_sha1[3];
h_sha1[4] = tmps[gid].h_sha1[4];
const u32 pw_len = pws[gid].pw_len;
const int pw_len64 = pw_len * 64;
#define SNMPV3_TMP_ELEMS_OPT 1024 // 1024 = (64 max pw length * 64) / sizeof (u32)
// for pw length > 64 we use global memory reads
if (pw_len < 64)
{
u32 tmp_shared[SNMPV3_TMP_ELEMS_OPT];
// md5
for (int i = 0; i < pw_len64 / 4; i++)
{
tmp_shared[i] = tmps[gid].tmp_md5[i];
}
for (int i = 0, j = loop_pos; i < loop_cnt; i += 64, j += 64)
{
const int idx = (j % pw_len64) / 4; // the optimization trick is to be able to do this
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
w0[0] = tmp_shared[idx + 0];
w0[1] = tmp_shared[idx + 1];
w0[2] = tmp_shared[idx + 2];
w0[3] = tmp_shared[idx + 3];
w1[0] = tmp_shared[idx + 4];
w1[1] = tmp_shared[idx + 5];
w1[2] = tmp_shared[idx + 6];
w1[3] = tmp_shared[idx + 7];
w2[0] = tmp_shared[idx + 8];
w2[1] = tmp_shared[idx + 9];
w2[2] = tmp_shared[idx + 10];
w2[3] = tmp_shared[idx + 11];
w3[0] = tmp_shared[idx + 12];
w3[1] = tmp_shared[idx + 13];
w3[2] = tmp_shared[idx + 14];
w3[3] = tmp_shared[idx + 15];
md5_transform (w0, w1, w2, w3, h_md5);
}
// sha1
for (int i = 0; i < pw_len64 / 4; i++)
{
tmp_shared[i] = tmps[gid].tmp_sha1[i];
}
for (int i = 0, j = loop_pos; i < loop_cnt; i += 64, j += 64)
{
const int idx = (j % pw_len64) / 4; // the optimization trick is to be able to do this
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
w0[0] = tmp_shared[idx + 0];
w0[1] = tmp_shared[idx + 1];
w0[2] = tmp_shared[idx + 2];
w0[3] = tmp_shared[idx + 3];
w1[0] = tmp_shared[idx + 4];
w1[1] = tmp_shared[idx + 5];
w1[2] = tmp_shared[idx + 6];
w1[3] = tmp_shared[idx + 7];
w2[0] = tmp_shared[idx + 8];
w2[1] = tmp_shared[idx + 9];
w2[2] = tmp_shared[idx + 10];
w2[3] = tmp_shared[idx + 11];
w3[0] = tmp_shared[idx + 12];
w3[1] = tmp_shared[idx + 13];
w3[2] = tmp_shared[idx + 14];
w3[3] = tmp_shared[idx + 15];
sha1_transform (w0, w1, w2, w3, h_sha1);
}
}
else
{
for (int i = 0, j = loop_pos; i < loop_cnt; i += 64, j += 64)
{
const int idx = (j % pw_len64) / 4; // the optimization trick is to be able to do this
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
// md5
w0[0] = tmps[gid].tmp_md5[idx + 0];
w0[1] = tmps[gid].tmp_md5[idx + 1];
w0[2] = tmps[gid].tmp_md5[idx + 2];
w0[3] = tmps[gid].tmp_md5[idx + 3];
w1[0] = tmps[gid].tmp_md5[idx + 4];
w1[1] = tmps[gid].tmp_md5[idx + 5];
w1[2] = tmps[gid].tmp_md5[idx + 6];
w1[3] = tmps[gid].tmp_md5[idx + 7];
w2[0] = tmps[gid].tmp_md5[idx + 8];
w2[1] = tmps[gid].tmp_md5[idx + 9];
w2[2] = tmps[gid].tmp_md5[idx + 10];
w2[3] = tmps[gid].tmp_md5[idx + 11];
w3[0] = tmps[gid].tmp_md5[idx + 12];
w3[1] = tmps[gid].tmp_md5[idx + 13];
w3[2] = tmps[gid].tmp_md5[idx + 14];
w3[3] = tmps[gid].tmp_md5[idx + 15];
md5_transform (w0, w1, w2, w3, h_md5);
// sha1
w0[0] = tmps[gid].tmp_sha1[idx + 0];
w0[1] = tmps[gid].tmp_sha1[idx + 1];
w0[2] = tmps[gid].tmp_sha1[idx + 2];
w0[3] = tmps[gid].tmp_sha1[idx + 3];
w1[0] = tmps[gid].tmp_sha1[idx + 4];
w1[1] = tmps[gid].tmp_sha1[idx + 5];
w1[2] = tmps[gid].tmp_sha1[idx + 6];
w1[3] = tmps[gid].tmp_sha1[idx + 7];
w2[0] = tmps[gid].tmp_sha1[idx + 8];
w2[1] = tmps[gid].tmp_sha1[idx + 9];
w2[2] = tmps[gid].tmp_sha1[idx + 10];
w2[3] = tmps[gid].tmp_sha1[idx + 11];
w3[0] = tmps[gid].tmp_sha1[idx + 12];
w3[1] = tmps[gid].tmp_sha1[idx + 13];
w3[2] = tmps[gid].tmp_sha1[idx + 14];
w3[3] = tmps[gid].tmp_sha1[idx + 15];
sha1_transform (w0, w1, w2, w3, h_sha1);
}
}
tmps[gid].h_md5[0] = h_md5[0];
tmps[gid].h_md5[1] = h_md5[1];
tmps[gid].h_md5[2] = h_md5[2];
tmps[gid].h_md5[3] = h_md5[3];
tmps[gid].h_sha1[0] = h_sha1[0];
tmps[gid].h_sha1[1] = h_sha1[1];
tmps[gid].h_sha1[2] = h_sha1[2];
tmps[gid].h_sha1[3] = h_sha1[3];
tmps[gid].h_sha1[4] = h_sha1[4];
}
KERNEL_FQ void m25000_comp (KERN_ATTR_TMPS_ESALT (hmac_md5_tmp_t, snmpv3_t))
{
/**
* modifier
*/
const u64 gid = get_global_id (0);
if (gid >= gid_max) return;
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
// md5
w0[0] = 0x00000080;
w0[1] = 0;
w0[2] = 0;
w0[3] = 0;
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[3] = 0;
w2[0] = 0;
w2[1] = 0;
w2[2] = 0;
w2[3] = 0;
w3[0] = 0;
w3[1] = 0;
w3[2] = 1048576 * 8;
w3[3] = 0;
u32 h_md5[4];
h_md5[0] = tmps[gid].h_md5[0];
h_md5[1] = tmps[gid].h_md5[1];
h_md5[2] = tmps[gid].h_md5[2];
h_md5[3] = tmps[gid].h_md5[3];
md5_transform (w0, w1, w2, w3, h_md5);
// sha1
w0[0] = 0x80000000;
w0[1] = 0;
w0[2] = 0;
w0[3] = 0;
w1[0] = 0;
w1[1] = 0;
w1[2] = 0;
w1[3] = 0;
w2[0] = 0;
w2[1] = 0;
w2[2] = 0;
w2[3] = 0;
w3[0] = 0;
w3[1] = 0;
w3[2] = 0;
w3[3] = 1048576 * 8;
u32 h_sha1[5];
h_sha1[0] = tmps[gid].h_sha1[0];
h_sha1[1] = tmps[gid].h_sha1[1];
h_sha1[2] = tmps[gid].h_sha1[2];
h_sha1[3] = tmps[gid].h_sha1[3];
h_sha1[4] = tmps[gid].h_sha1[4];
sha1_transform (w0, w1, w2, w3, h_sha1);
md5_ctx_t md5_ctx;
sha1_ctx_t sha1_ctx;
md5_init (&md5_ctx);
sha1_init (&sha1_ctx);
u32 w[16];
// md5
w[ 0] = h_md5[0];
w[ 1] = h_md5[1];
w[ 2] = h_md5[2];
w[ 3] = h_md5[3];
w[ 4] = 0;
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
md5_update (&md5_ctx, w, 16);
// sha1
w[ 0] = h_sha1[0];
w[ 1] = h_sha1[1];
w[ 2] = h_sha1[2];
w[ 3] = h_sha1[3];
w[ 4] = h_sha1[4];
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
sha1_update (&sha1_ctx, w, 20);
// engineID
md5_update_global (&md5_ctx, esalt_bufs[DIGESTS_OFFSET].engineID_buf, esalt_bufs[DIGESTS_OFFSET].engineID_len);
sha1_update_global_swap (&sha1_ctx, esalt_bufs[DIGESTS_OFFSET].engineID_buf, esalt_bufs[DIGESTS_OFFSET].engineID_len);
// md5
w[ 0] = h_md5[0];
w[ 1] = h_md5[1];
w[ 2] = h_md5[2];
w[ 3] = h_md5[3];
w[ 4] = 0;
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
md5_update (&md5_ctx, w, 16);
// sha1
w[ 0] = h_sha1[0];
w[ 1] = h_sha1[1];
w[ 2] = h_sha1[2];
w[ 3] = h_sha1[3];
w[ 4] = h_sha1[4];
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
sha1_update (&sha1_ctx, w, 20);
md5_final (&md5_ctx);
sha1_final (&sha1_ctx);
// md5
w[ 0] = md5_ctx.h[0];
w[ 1] = md5_ctx.h[1];
w[ 2] = md5_ctx.h[2];
w[ 3] = md5_ctx.h[3];
w[ 4] = 0;
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
md5_hmac_ctx_t md5_hmac_ctx;
md5_hmac_init (&md5_hmac_ctx, w, 16);
md5_hmac_update_global (&md5_hmac_ctx, esalt_bufs[DIGESTS_OFFSET].salt_buf, esalt_bufs[DIGESTS_OFFSET].salt_len);
md5_hmac_final (&md5_hmac_ctx);
{
const u32 r0 = hc_swap32_S (md5_hmac_ctx.opad.h[DGST_R0]);
const u32 r1 = hc_swap32_S (md5_hmac_ctx.opad.h[DGST_R1]);
const u32 r2 = hc_swap32_S (md5_hmac_ctx.opad.h[DGST_R2]);
const u32 r3 = 0;
#define il_pos 0
#ifdef KERNEL_STATIC
#include COMPARE_M
#endif
}
// sha1
w[ 0] = sha1_ctx.h[0];
w[ 1] = sha1_ctx.h[1];
w[ 2] = sha1_ctx.h[2];
w[ 3] = sha1_ctx.h[3];
w[ 4] = sha1_ctx.h[4];
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
sha1_hmac_ctx_t sha1_hmac_ctx;
sha1_hmac_init (&sha1_hmac_ctx, w, 20);
sha1_hmac_update_global_swap (&sha1_hmac_ctx, esalt_bufs[DIGESTS_OFFSET].salt_buf, esalt_bufs[DIGESTS_OFFSET].salt_len);
sha1_hmac_final (&sha1_hmac_ctx);
{
const u32 r0 = sha1_hmac_ctx.opad.h[DGST_R0];
const u32 r1 = sha1_hmac_ctx.opad.h[DGST_R1];
const u32 r2 = sha1_hmac_ctx.opad.h[DGST_R2];
const u32 r3 = 0;
#define il_pos 0
#ifdef KERNEL_STATIC
#include COMPARE_M
#endif
}
}
OpenCL/m25100-pure.cl
+2 -2
View File
@@ -18,8 +18,8 @@
#define COMPARE_M "inc_comp_multi.cl"
#define SNMPV3_SALT_MAX 1500
#define SNMPV3_ENGINEID_MAX 32
#define SNMPV3_MSG_AUTH_PARAMS_MAX 12
#define SNMPV3_ENGINEID_MAX 34
#define SNMPV3_MSG_AUTH_PARAMS_LEN 12
#define SNMPV3_ROUNDS 1048576
#define SNMPV3_MAX_PW_LENGTH 64
OpenCL/m25200-pure.cl
+8 -8
View File
@@ -18,13 +18,17 @@
#define COMPARE_M "inc_comp_multi.cl"
#define SNMPV3_SALT_MAX 1500
#define SNMPV3_ENGINEID_MAX 32
#define SNMPV3_MSG_AUTH_PARAMS_MAX 12
#define SNMPV3_ENGINEID_MAX 34
#define SNMPV3_MSG_AUTH_PARAMS_LEN 12
#define SNMPV3_ROUNDS 1048576
#define SNMPV3_MAX_PW_LENGTH 64
#define SNMPV3_TMP_ELEMS 4096 // 4096 = (256 (max pw length) * 64) / sizeof (u32)
#define SNMPV3_HASH_ELEMS 8 // 8 = aligned 5
#define SNMPV3_TMP_ELEMS 4096 // 4096 = (256 (max pw length) * 64) / sizeof (u32)
#define SNMPV3_HASH_ELEMS 8 // 8 = aligned 5
#define SNMPV3_MAX_SALT_ELEMS 512 // 512 * 4 = 2048 > 1500, also has to be multiple of 64
#define SNMPV3_MAX_ENGINE_ELEMS 16 // 16 * 4 = 64 > 32, also has to be multiple of 64
#define SNMPV3_MAX_PNUM_ELEMS 4 // 4 * 4 = 16 > 9
typedef struct hmac_sha1_tmp
{
@@ -33,10 +37,6 @@ typedef struct hmac_sha1_tmp
} hmac_sha1_tmp_t;
#define SNMPV3_MAX_SALT_ELEMS 512 // 512 * 4 = 2048 > 1500, also has to be multiple of 64
#define SNMPV3_MAX_ENGINE_ELEMS 16 // 16 * 4 = 64 > 32, also has to be multiple of 64
#define SNMPV3_MAX_PNUM_ELEMS 4 // 4 * 4 = 16 > 9
typedef struct snmpv3
{
u32 salt_buf[SNMPV3_MAX_SALT_ELEMS];