/** * Author......: See docs/credits.txt * License.....: MIT */ #define NEW_SIMD_CODE #define XSTR(x) #x #define STR(x) XSTR(x) #ifdef KERNEL_STATIC #include STR(INCLUDE_PATH/inc_vendor.h) #include STR(INCLUDE_PATH/inc_types.h) #include STR(INCLUDE_PATH/inc_platform.cl) #include STR(INCLUDE_PATH/inc_common.cl) #include STR(INCLUDE_PATH/inc_simd.cl) #include STR(INCLUDE_PATH/inc_hash_sha256.cl) #include STR(INCLUDE_PATH/inc_cipher_aes.cl) #endif #define COMPARE_M STR(INCLUDE_PATH/inc_comp_multi.cl) typedef struct vbox { u32 salt1_buf[64]; u32 salt1_len; u32 aes_key_len; u32 enc_pass_buf[128]; u32 salt2_buf[64]; u32 salt2_len; } vbox_t; typedef struct pbkdf2_sha256_tmp { u32 ipad[8]; u32 opad[8]; u32 dgst[64]; u32 out[64]; } pbkdf2_sha256_tmp_t; DECLSPEC void hmac_sha256_run_V (PRIVATE_AS u32x *w0, PRIVATE_AS u32x *w1, PRIVATE_AS u32x *w2, PRIVATE_AS u32x *w3, PRIVATE_AS u32x *ipad, PRIVATE_AS u32x *opad, PRIVATE_AS u32x *digest) { digest[0] = ipad[0]; digest[1] = ipad[1]; digest[2] = ipad[2]; digest[3] = ipad[3]; digest[4] = ipad[4]; digest[5] = ipad[5]; digest[6] = ipad[6]; digest[7] = ipad[7]; sha256_transform_vector (w0, w1, w2, w3, digest); w0[0] = digest[0]; w0[1] = digest[1]; w0[2] = digest[2]; w0[3] = digest[3]; w1[0] = digest[4]; w1[1] = digest[5]; w1[2] = digest[6]; w1[3] = digest[7]; w2[0] = 0x80000000; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = (64 + 32) * 8; digest[0] = opad[0]; digest[1] = opad[1]; digest[2] = opad[2]; digest[3] = opad[3]; digest[4] = opad[4]; digest[5] = opad[5]; digest[6] = opad[6]; digest[7] = opad[7]; sha256_transform_vector (w0, w1, w2, w3, digest); } DECLSPEC void xts_mul2 (PRIVATE_AS u32 *in, PRIVATE_AS u32 *out) { const u32 c = in[3] >> 31; out[3] = (in[3] << 1) | (in[2] >> 31); out[2] = (in[2] << 1) | (in[1] >> 31); out[1] = (in[1] << 1) | (in[0] >> 31); out[0] = (in[0] << 1); out[0] ^= c * 0x87; } KERNEL_FQ void m27600_init (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, vbox_t)) { /** * base */ const u64 gid = get_global_id (0); if (gid >= GID_CNT) return; sha256_hmac_ctx_t sha256_hmac_ctx; sha256_hmac_init_global_swap (&sha256_hmac_ctx, pws[gid].i, pws[gid].pw_len); tmps[gid].ipad[0] = sha256_hmac_ctx.ipad.h[0]; tmps[gid].ipad[1] = sha256_hmac_ctx.ipad.h[1]; tmps[gid].ipad[2] = sha256_hmac_ctx.ipad.h[2]; tmps[gid].ipad[3] = sha256_hmac_ctx.ipad.h[3]; tmps[gid].ipad[4] = sha256_hmac_ctx.ipad.h[4]; tmps[gid].ipad[5] = sha256_hmac_ctx.ipad.h[5]; tmps[gid].ipad[6] = sha256_hmac_ctx.ipad.h[6]; tmps[gid].ipad[7] = sha256_hmac_ctx.ipad.h[7]; tmps[gid].opad[0] = sha256_hmac_ctx.opad.h[0]; tmps[gid].opad[1] = sha256_hmac_ctx.opad.h[1]; tmps[gid].opad[2] = sha256_hmac_ctx.opad.h[2]; tmps[gid].opad[3] = sha256_hmac_ctx.opad.h[3]; tmps[gid].opad[4] = sha256_hmac_ctx.opad.h[4]; tmps[gid].opad[5] = sha256_hmac_ctx.opad.h[5]; tmps[gid].opad[6] = sha256_hmac_ctx.opad.h[6]; tmps[gid].opad[7] = sha256_hmac_ctx.opad.h[7]; sha256_hmac_update_global_swap (&sha256_hmac_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].salt1_buf, esalt_bufs[DIGESTS_OFFSET_HOST].salt1_len); for (u32 i = 0, j = 1; i < 16; i += 8, j += 1) { sha256_hmac_ctx_t sha256_hmac_ctx2 = sha256_hmac_ctx; u32 w0[4]; u32 w1[4]; u32 w2[4]; u32 w3[4]; w0[0] = j; 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] = 0; sha256_hmac_update_64 (&sha256_hmac_ctx2, w0, w1, w2, w3, 4); sha256_hmac_final (&sha256_hmac_ctx2); tmps[gid].dgst[i + 0] = sha256_hmac_ctx2.opad.h[0]; tmps[gid].dgst[i + 1] = sha256_hmac_ctx2.opad.h[1]; tmps[gid].dgst[i + 2] = sha256_hmac_ctx2.opad.h[2]; tmps[gid].dgst[i + 3] = sha256_hmac_ctx2.opad.h[3]; tmps[gid].dgst[i + 4] = sha256_hmac_ctx2.opad.h[4]; tmps[gid].dgst[i + 5] = sha256_hmac_ctx2.opad.h[5]; tmps[gid].dgst[i + 6] = sha256_hmac_ctx2.opad.h[6]; tmps[gid].dgst[i + 7] = sha256_hmac_ctx2.opad.h[7]; tmps[gid].out[i + 0] = tmps[gid].dgst[i + 0]; tmps[gid].out[i + 1] = tmps[gid].dgst[i + 1]; tmps[gid].out[i + 2] = tmps[gid].dgst[i + 2]; tmps[gid].out[i + 3] = tmps[gid].dgst[i + 3]; tmps[gid].out[i + 4] = tmps[gid].dgst[i + 4]; tmps[gid].out[i + 5] = tmps[gid].dgst[i + 5]; tmps[gid].out[i + 6] = tmps[gid].dgst[i + 6]; tmps[gid].out[i + 7] = tmps[gid].dgst[i + 7]; } } KERNEL_FQ void m27600_loop (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, vbox_t)) { const u64 gid = get_global_id (0); if ((gid * VECT_SIZE) >= GID_CNT) return; u32x ipad[8]; u32x opad[8]; ipad[0] = packv (tmps, ipad, gid, 0); ipad[1] = packv (tmps, ipad, gid, 1); ipad[2] = packv (tmps, ipad, gid, 2); ipad[3] = packv (tmps, ipad, gid, 3); ipad[4] = packv (tmps, ipad, gid, 4); ipad[5] = packv (tmps, ipad, gid, 5); ipad[6] = packv (tmps, ipad, gid, 6); ipad[7] = packv (tmps, ipad, gid, 7); opad[0] = packv (tmps, opad, gid, 0); opad[1] = packv (tmps, opad, gid, 1); opad[2] = packv (tmps, opad, gid, 2); opad[3] = packv (tmps, opad, gid, 3); opad[4] = packv (tmps, opad, gid, 4); opad[5] = packv (tmps, opad, gid, 5); opad[6] = packv (tmps, opad, gid, 6); opad[7] = packv (tmps, opad, gid, 7); for (u32 i = 0; i < 16; i += 8) { u32x dgst[8]; u32x out[8]; dgst[0] = packv (tmps, dgst, gid, i + 0); dgst[1] = packv (tmps, dgst, gid, i + 1); dgst[2] = packv (tmps, dgst, gid, i + 2); dgst[3] = packv (tmps, dgst, gid, i + 3); dgst[4] = packv (tmps, dgst, gid, i + 4); dgst[5] = packv (tmps, dgst, gid, i + 5); dgst[6] = packv (tmps, dgst, gid, i + 6); dgst[7] = packv (tmps, dgst, gid, i + 7); out[0] = packv (tmps, out, gid, i + 0); out[1] = packv (tmps, out, gid, i + 1); out[2] = packv (tmps, out, gid, i + 2); out[3] = packv (tmps, out, gid, i + 3); out[4] = packv (tmps, out, gid, i + 4); out[5] = packv (tmps, out, gid, i + 5); out[6] = packv (tmps, out, gid, i + 6); out[7] = packv (tmps, out, gid, i + 7); for (u32 j = 0; j < LOOP_CNT; j++) { u32x w0[4]; u32x w1[4]; u32x w2[4]; u32x w3[4]; w0[0] = dgst[0]; w0[1] = dgst[1]; w0[2] = dgst[2]; w0[3] = dgst[3]; w1[0] = dgst[4]; w1[1] = dgst[5]; w1[2] = dgst[6]; w1[3] = dgst[7]; w2[0] = 0x80000000; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = (64 + 32) * 8; hmac_sha256_run_V (w0, w1, w2, w3, ipad, opad, dgst); out[0] ^= dgst[0]; out[1] ^= dgst[1]; out[2] ^= dgst[2]; out[3] ^= dgst[3]; out[4] ^= dgst[4]; out[5] ^= dgst[5]; out[6] ^= dgst[6]; out[7] ^= dgst[7]; } unpackv (tmps, dgst, gid, i + 0, dgst[0]); unpackv (tmps, dgst, gid, i + 1, dgst[1]); unpackv (tmps, dgst, gid, i + 2, dgst[2]); unpackv (tmps, dgst, gid, i + 3, dgst[3]); unpackv (tmps, dgst, gid, i + 4, dgst[4]); unpackv (tmps, dgst, gid, i + 5, dgst[5]); unpackv (tmps, dgst, gid, i + 6, dgst[6]); unpackv (tmps, dgst, gid, i + 7, dgst[7]); unpackv (tmps, out, gid, i + 0, out[0]); unpackv (tmps, out, gid, i + 1, out[1]); unpackv (tmps, out, gid, i + 2, out[2]); unpackv (tmps, out, gid, i + 3, out[3]); unpackv (tmps, out, gid, i + 4, out[4]); unpackv (tmps, out, gid, i + 5, out[5]); unpackv (tmps, out, gid, i + 6, out[6]); unpackv (tmps, out, gid, i + 7, out[7]); } } KERNEL_FQ void m27600_init2 (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, vbox_t)) { const u64 gid = get_global_id (0); const u64 lid = get_local_id (0); const u64 lsz = get_local_size (0); /** * aes shared */ #ifdef REAL_SHM LOCAL_VK u32 s_td0[256]; LOCAL_VK u32 s_td1[256]; LOCAL_VK u32 s_td2[256]; LOCAL_VK u32 s_td3[256]; LOCAL_VK u32 s_td4[256]; LOCAL_VK u32 s_te0[256]; LOCAL_VK u32 s_te1[256]; LOCAL_VK u32 s_te2[256]; LOCAL_VK u32 s_te3[256]; LOCAL_VK u32 s_te4[256]; for (u32 i = lid; i < 256; i += lsz) { s_td0[i] = td0[i]; s_td1[i] = td1[i]; s_td2[i] = td2[i]; s_td3[i] = td3[i]; s_td4[i] = td4[i]; s_te0[i] = te0[i]; s_te1[i] = te1[i]; s_te2[i] = te2[i]; s_te3[i] = te3[i]; s_te4[i] = te4[i]; } SYNC_THREADS (); #else CONSTANT_AS u32a *s_td0 = td0; CONSTANT_AS u32a *s_td1 = td1; CONSTANT_AS u32a *s_td2 = td2; CONSTANT_AS u32a *s_td3 = td3; CONSTANT_AS u32a *s_td4 = td4; CONSTANT_AS u32a *s_te0 = te0; CONSTANT_AS u32a *s_te1 = te1; CONSTANT_AS u32a *s_te2 = te2; CONSTANT_AS u32a *s_te3 = te3; CONSTANT_AS u32a *s_te4 = te4; #endif if (gid >= GID_CNT) return; /** * aes */ u32 ukey1[8]; u32 ukey2[8]; ukey1[0] = tmps[gid].out[ 0]; ukey1[1] = tmps[gid].out[ 1]; ukey1[2] = tmps[gid].out[ 2]; ukey1[3] = tmps[gid].out[ 3]; ukey1[4] = tmps[gid].out[ 4]; ukey1[5] = tmps[gid].out[ 5]; ukey1[6] = tmps[gid].out[ 6]; ukey1[7] = tmps[gid].out[ 7]; ukey2[0] = tmps[gid].out[ 8]; ukey2[1] = tmps[gid].out[ 9]; ukey2[2] = tmps[gid].out[10]; ukey2[3] = tmps[gid].out[11]; ukey2[4] = tmps[gid].out[12]; ukey2[5] = tmps[gid].out[13]; ukey2[6] = tmps[gid].out[14]; ukey2[7] = tmps[gid].out[15]; ukey1[0] = hc_swap32_S (ukey1[0]); ukey1[1] = hc_swap32_S (ukey1[1]); ukey1[2] = hc_swap32_S (ukey1[2]); ukey1[3] = hc_swap32_S (ukey1[3]); ukey1[4] = hc_swap32_S (ukey1[4]); ukey1[5] = hc_swap32_S (ukey1[5]); ukey1[6] = hc_swap32_S (ukey1[6]); ukey1[7] = hc_swap32_S (ukey1[7]); ukey2[0] = hc_swap32_S (ukey2[0]); ukey2[1] = hc_swap32_S (ukey2[1]); ukey2[2] = hc_swap32_S (ukey2[2]); ukey2[3] = hc_swap32_S (ukey2[3]); ukey2[4] = hc_swap32_S (ukey2[4]); ukey2[5] = hc_swap32_S (ukey2[5]); ukey2[6] = hc_swap32_S (ukey2[6]); ukey2[7] = hc_swap32_S (ukey2[7]); u32 ks[60]; u32 aes_decrypt[16]; // first block u32 in[4]; in[0] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[0]; in[1] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[1]; in[2] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[2]; in[3] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[3]; u32 out[4]; out[0] = in[0]; out[1] = in[1]; out[2] = in[2]; out[3] = in[3]; u32 S[4] = { 0 }; // tweak, 16 x 0x00 u32 T[4] = { 0 }; aes256_set_encrypt_key (ks, ukey2, s_te0, s_te1, s_te2, s_te3); aes256_encrypt (ks, S, T, s_te0, s_te1, s_te2, s_te3, s_te4); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes256_set_decrypt_key (ks, ukey1, s_te0, s_te1, s_te2, s_te3, s_td0, s_td1, s_td2, s_td3); aes256_decrypt (ks, out, out, s_td0, s_td1, s_td2, s_td3, s_td4); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes_decrypt[0] = out[0]; aes_decrypt[1] = out[1]; aes_decrypt[2] = out[2]; aes_decrypt[3] = out[3]; // next in[0] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[4]; in[1] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[5]; in[2] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[6]; in[3] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[7]; out[0] = in[0]; out[1] = in[1]; out[2] = in[2]; out[3] = in[3]; xts_mul2 (T, T); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes256_decrypt (ks, out, out, s_td0, s_td1, s_td2, s_td3, s_td4); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes_decrypt[4] = out[0]; aes_decrypt[5] = out[1]; aes_decrypt[6] = out[2]; aes_decrypt[7] = out[3]; // next in[0] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[8]; in[1] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[9]; in[2] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[10]; in[3] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[11]; out[0] = in[0]; out[1] = in[1]; out[2] = in[2]; out[3] = in[3]; xts_mul2 (T, T); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes256_decrypt (ks, out, out, s_td0, s_td1, s_td2, s_td3, s_td4); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes_decrypt[8] = out[0]; aes_decrypt[9] = out[1]; aes_decrypt[10] = out[2]; aes_decrypt[11] = out[3]; // next in[0] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[12]; in[1] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[13]; in[2] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[14]; in[3] = esalt_bufs[DIGESTS_OFFSET_HOST].enc_pass_buf[15]; out[0] = in[0]; out[1] = in[1]; out[2] = in[2]; out[3] = in[3]; xts_mul2 (T, T); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes256_decrypt (ks, out, out, s_td0, s_td1, s_td2, s_td3, s_td4); out[0] ^= T[0]; out[1] ^= T[1]; out[2] ^= T[2]; out[3] ^= T[3]; aes_decrypt[12] = out[0]; aes_decrypt[13] = out[1]; aes_decrypt[14] = out[2]; aes_decrypt[15] = out[3]; // pbkdf2-hmac-sha256 u32 w0[4]; u32 w1[4]; u32 w2[4]; u32 w3[4]; w0[0] = hc_swap32_S (aes_decrypt[ 0]); w0[1] = hc_swap32_S (aes_decrypt[ 1]); w0[2] = hc_swap32_S (aes_decrypt[ 2]); w0[3] = hc_swap32_S (aes_decrypt[ 3]); w1[0] = hc_swap32_S (aes_decrypt[ 4]); w1[1] = hc_swap32_S (aes_decrypt[ 5]); w1[2] = hc_swap32_S (aes_decrypt[ 6]); w1[3] = hc_swap32_S (aes_decrypt[ 7]); w2[0] = hc_swap32_S (aes_decrypt[ 8]); w2[1] = hc_swap32_S (aes_decrypt[ 9]); w2[2] = hc_swap32_S (aes_decrypt[10]); w2[3] = hc_swap32_S (aes_decrypt[11]); w3[0] = hc_swap32_S (aes_decrypt[12]); w3[1] = hc_swap32_S (aes_decrypt[13]); w3[2] = hc_swap32_S (aes_decrypt[14]); w3[3] = hc_swap32_S (aes_decrypt[15]); // init2 sha256_hmac_ctx_t sha256_hmac_ctx; sha256_hmac_init_64 (&sha256_hmac_ctx, w0, w1, w2, w3); tmps[gid].ipad[0] = sha256_hmac_ctx.ipad.h[0]; tmps[gid].ipad[1] = sha256_hmac_ctx.ipad.h[1]; tmps[gid].ipad[2] = sha256_hmac_ctx.ipad.h[2]; tmps[gid].ipad[3] = sha256_hmac_ctx.ipad.h[3]; tmps[gid].ipad[4] = sha256_hmac_ctx.ipad.h[4]; tmps[gid].ipad[5] = sha256_hmac_ctx.ipad.h[5]; tmps[gid].ipad[6] = sha256_hmac_ctx.ipad.h[6]; tmps[gid].ipad[7] = sha256_hmac_ctx.ipad.h[7]; tmps[gid].opad[0] = sha256_hmac_ctx.opad.h[0]; tmps[gid].opad[1] = sha256_hmac_ctx.opad.h[1]; tmps[gid].opad[2] = sha256_hmac_ctx.opad.h[2]; tmps[gid].opad[3] = sha256_hmac_ctx.opad.h[3]; tmps[gid].opad[4] = sha256_hmac_ctx.opad.h[4]; tmps[gid].opad[5] = sha256_hmac_ctx.opad.h[5]; tmps[gid].opad[6] = sha256_hmac_ctx.opad.h[6]; tmps[gid].opad[7] = sha256_hmac_ctx.opad.h[7]; sha256_hmac_update_global_swap (&sha256_hmac_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].salt2_buf, esalt_bufs[DIGESTS_OFFSET_HOST].salt2_len); for (u32 i = 0, j = 1; i < 8; i += 8, j += 1) { sha256_hmac_ctx_t sha256_hmac_ctx2 = sha256_hmac_ctx; w0[0] = j; 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] = 0; sha256_hmac_update_64 (&sha256_hmac_ctx2, w0, w1, w2, w3, 4); sha256_hmac_final (&sha256_hmac_ctx2); tmps[gid].dgst[i + 0] = sha256_hmac_ctx2.opad.h[0]; tmps[gid].dgst[i + 1] = sha256_hmac_ctx2.opad.h[1]; tmps[gid].dgst[i + 2] = sha256_hmac_ctx2.opad.h[2]; tmps[gid].dgst[i + 3] = sha256_hmac_ctx2.opad.h[3]; tmps[gid].dgst[i + 4] = sha256_hmac_ctx2.opad.h[4]; tmps[gid].dgst[i + 5] = sha256_hmac_ctx2.opad.h[5]; tmps[gid].dgst[i + 6] = sha256_hmac_ctx2.opad.h[6]; tmps[gid].dgst[i + 7] = sha256_hmac_ctx2.opad.h[7]; tmps[gid].out[i + 0] = tmps[gid].dgst[i + 0]; tmps[gid].out[i + 1] = tmps[gid].dgst[i + 1]; tmps[gid].out[i + 2] = tmps[gid].dgst[i + 2]; tmps[gid].out[i + 3] = tmps[gid].dgst[i + 3]; tmps[gid].out[i + 4] = tmps[gid].dgst[i + 4]; tmps[gid].out[i + 5] = tmps[gid].dgst[i + 5]; tmps[gid].out[i + 6] = tmps[gid].dgst[i + 6]; tmps[gid].out[i + 7] = tmps[gid].dgst[i + 7]; } } KERNEL_FQ void m27600_loop2 (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, vbox_t)) { const u64 gid = get_global_id (0); if ((gid * VECT_SIZE) >= GID_CNT) return; u32x ipad[8]; u32x opad[8]; ipad[0] = packv (tmps, ipad, gid, 0); ipad[1] = packv (tmps, ipad, gid, 1); ipad[2] = packv (tmps, ipad, gid, 2); ipad[3] = packv (tmps, ipad, gid, 3); ipad[4] = packv (tmps, ipad, gid, 4); ipad[5] = packv (tmps, ipad, gid, 5); ipad[6] = packv (tmps, ipad, gid, 6); ipad[7] = packv (tmps, ipad, gid, 7); opad[0] = packv (tmps, opad, gid, 0); opad[1] = packv (tmps, opad, gid, 1); opad[2] = packv (tmps, opad, gid, 2); opad[3] = packv (tmps, opad, gid, 3); opad[4] = packv (tmps, opad, gid, 4); opad[5] = packv (tmps, opad, gid, 5); opad[6] = packv (tmps, opad, gid, 6); opad[7] = packv (tmps, opad, gid, 7); for (u32 i = 0; i < 8; i += 8) { u32x dgst[8]; u32x out[8]; dgst[0] = packv (tmps, dgst, gid, i + 0); dgst[1] = packv (tmps, dgst, gid, i + 1); dgst[2] = packv (tmps, dgst, gid, i + 2); dgst[3] = packv (tmps, dgst, gid, i + 3); dgst[4] = packv (tmps, dgst, gid, i + 4); dgst[5] = packv (tmps, dgst, gid, i + 5); dgst[6] = packv (tmps, dgst, gid, i + 6); dgst[7] = packv (tmps, dgst, gid, i + 7); out[0] = packv (tmps, out, gid, i + 0); out[1] = packv (tmps, out, gid, i + 1); out[2] = packv (tmps, out, gid, i + 2); out[3] = packv (tmps, out, gid, i + 3); out[4] = packv (tmps, out, gid, i + 4); out[5] = packv (tmps, out, gid, i + 5); out[6] = packv (tmps, out, gid, i + 6); out[7] = packv (tmps, out, gid, i + 7); for (u32 j = 0; j < LOOP_CNT; j++) { u32x w0[4]; u32x w1[4]; u32x w2[4]; u32x w3[4]; w0[0] = dgst[0]; w0[1] = dgst[1]; w0[2] = dgst[2]; w0[3] = dgst[3]; w1[0] = dgst[4]; w1[1] = dgst[5]; w1[2] = dgst[6]; w1[3] = dgst[7]; w2[0] = 0x80000000; w2[1] = 0; w2[2] = 0; w2[3] = 0; w3[0] = 0; w3[1] = 0; w3[2] = 0; w3[3] = (64 + 32) * 8; hmac_sha256_run_V (w0, w1, w2, w3, ipad, opad, dgst); out[0] ^= dgst[0]; out[1] ^= dgst[1]; out[2] ^= dgst[2]; out[3] ^= dgst[3]; out[4] ^= dgst[4]; out[5] ^= dgst[5]; out[6] ^= dgst[6]; out[7] ^= dgst[7]; } unpackv (tmps, dgst, gid, i + 0, dgst[0]); unpackv (tmps, dgst, gid, i + 1, dgst[1]); unpackv (tmps, dgst, gid, i + 2, dgst[2]); unpackv (tmps, dgst, gid, i + 3, dgst[3]); unpackv (tmps, dgst, gid, i + 4, dgst[4]); unpackv (tmps, dgst, gid, i + 5, dgst[5]); unpackv (tmps, dgst, gid, i + 6, dgst[6]); unpackv (tmps, dgst, gid, i + 7, dgst[7]); unpackv (tmps, out, gid, i + 0, out[0]); unpackv (tmps, out, gid, i + 1, out[1]); unpackv (tmps, out, gid, i + 2, out[2]); unpackv (tmps, out, gid, i + 3, out[3]); unpackv (tmps, out, gid, i + 4, out[4]); unpackv (tmps, out, gid, i + 5, out[5]); unpackv (tmps, out, gid, i + 6, out[6]); unpackv (tmps, out, gid, i + 7, out[7]); } } KERNEL_FQ void m27600_comp (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, vbox_t)) { const u64 gid = get_global_id (0); if (gid >= GID_CNT) return; const u32 r0 = tmps[gid].out[0]; const u32 r1 = tmps[gid].out[1]; const u32 r2 = tmps[gid].out[2]; const u32 r3 = tmps[gid].out[3]; #define il_pos 0 #ifdef KERNEL_STATIC #include COMPARE_M #endif }