/** * 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_sha512.cl" #endif #define COMPARE_S "inc_comp_single.cl" #define COMPARE_M "inc_comp_multi.cl" typedef struct pbkdf2_sha512_tmp { u64 ipad[8]; u64 opad[8]; u64 dgst[16]; u64 out[16]; } pbkdf2_sha512_tmp_t; typedef struct pbkdf2_sha512 { u32 salt_buf[64]; } pbkdf2_sha512_t; DECLSPEC void hmac_sha512_run_V (u32x *w0, u32x *w1, u32x *w2, u32x *w3, u32x *w4, u32x *w5, u32x *w6, u32x *w7, u64x *ipad, u64x *opad, u64x *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]; sha512_transform_vector (w0, w1, w2, w3, w4, w5, w6, w7, digest); w0[0] = h32_from_64 (digest[0]); w0[1] = l32_from_64 (digest[0]); w0[2] = h32_from_64 (digest[1]); w0[3] = l32_from_64 (digest[1]); w1[0] = h32_from_64 (digest[2]); w1[1] = l32_from_64 (digest[2]); w1[2] = h32_from_64 (digest[3]); w1[3] = l32_from_64 (digest[3]); w2[0] = h32_from_64 (digest[4]); w2[1] = l32_from_64 (digest[4]); w2[2] = h32_from_64 (digest[5]); w2[3] = l32_from_64 (digest[5]); w3[0] = h32_from_64 (digest[6]); w3[1] = l32_from_64 (digest[6]); w3[2] = h32_from_64 (digest[7]); w3[3] = l32_from_64 (digest[7]); w4[0] = 0x80000000; w4[1] = 0; w4[2] = 0; w4[3] = 0; w5[0] = 0; w5[1] = 0; w5[2] = 0; w5[3] = 0; w6[0] = 0; w6[1] = 0; w6[2] = 0; w6[3] = 0; w7[0] = 0; w7[1] = 0; w7[2] = 0; w7[3] = (128 + 64) * 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]; sha512_transform_vector (w0, w1, w2, w3, w4, w5, w6, w7, digest); } KERNEL_FQ void m07100_init (KERN_ATTR_TMPS_ESALT (pbkdf2_sha512_tmp_t, pbkdf2_sha512_t)) { /** * base */ const u64 gid = get_global_id (0); if (gid >= GID_MAX) return; sha512_hmac_ctx_t sha512_hmac_ctx; sha512_hmac_init_global_swap (&sha512_hmac_ctx, pws[gid].i, pws[gid].pw_len); tmps[gid].ipad[0] = sha512_hmac_ctx.ipad.h[0]; tmps[gid].ipad[1] = sha512_hmac_ctx.ipad.h[1]; tmps[gid].ipad[2] = sha512_hmac_ctx.ipad.h[2]; tmps[gid].ipad[3] = sha512_hmac_ctx.ipad.h[3]; tmps[gid].ipad[4] = sha512_hmac_ctx.ipad.h[4]; tmps[gid].ipad[5] = sha512_hmac_ctx.ipad.h[5]; tmps[gid].ipad[6] = sha512_hmac_ctx.ipad.h[6]; tmps[gid].ipad[7] = sha512_hmac_ctx.ipad.h[7]; tmps[gid].opad[0] = sha512_hmac_ctx.opad.h[0]; tmps[gid].opad[1] = sha512_hmac_ctx.opad.h[1]; tmps[gid].opad[2] = sha512_hmac_ctx.opad.h[2]; tmps[gid].opad[3] = sha512_hmac_ctx.opad.h[3]; tmps[gid].opad[4] = sha512_hmac_ctx.opad.h[4]; tmps[gid].opad[5] = sha512_hmac_ctx.opad.h[5]; tmps[gid].opad[6] = sha512_hmac_ctx.opad.h[6]; tmps[gid].opad[7] = sha512_hmac_ctx.opad.h[7]; sha512_hmac_update_global_swap (&sha512_hmac_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf, salt_bufs[SALT_POS_HOST].salt_len); for (u32 i = 0, j = 1; i < 8; i += 8, j += 1) { sha512_hmac_ctx_t sha512_hmac_ctx2 = sha512_hmac_ctx; u32 w0[4]; u32 w1[4]; u32 w2[4]; u32 w3[4]; u32 w4[4]; u32 w5[4]; u32 w6[4]; u32 w7[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; w4[0] = 0; w4[1] = 0; w4[2] = 0; w4[3] = 0; w5[0] = 0; w5[1] = 0; w5[2] = 0; w5[3] = 0; w6[0] = 0; w6[1] = 0; w6[2] = 0; w6[3] = 0; w7[0] = 0; w7[1] = 0; w7[2] = 0; w7[3] = 0; sha512_hmac_update_128 (&sha512_hmac_ctx2, w0, w1, w2, w3, w4, w5, w6, w7, 4); sha512_hmac_final (&sha512_hmac_ctx2); tmps[gid].dgst[i + 0] = sha512_hmac_ctx2.opad.h[0]; tmps[gid].dgst[i + 1] = sha512_hmac_ctx2.opad.h[1]; tmps[gid].dgst[i + 2] = sha512_hmac_ctx2.opad.h[2]; tmps[gid].dgst[i + 3] = sha512_hmac_ctx2.opad.h[3]; tmps[gid].dgst[i + 4] = sha512_hmac_ctx2.opad.h[4]; tmps[gid].dgst[i + 5] = sha512_hmac_ctx2.opad.h[5]; tmps[gid].dgst[i + 6] = sha512_hmac_ctx2.opad.h[6]; tmps[gid].dgst[i + 7] = sha512_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 m07100_loop (KERN_ATTR_TMPS_ESALT (pbkdf2_sha512_tmp_t, pbkdf2_sha512_t)) { const u64 gid = get_global_id (0); if ((gid * VECT_SIZE) >= GID_MAX) return; u64x ipad[8]; u64x opad[8]; ipad[0] = pack64v (tmps, ipad, gid, 0); ipad[1] = pack64v (tmps, ipad, gid, 1); ipad[2] = pack64v (tmps, ipad, gid, 2); ipad[3] = pack64v (tmps, ipad, gid, 3); ipad[4] = pack64v (tmps, ipad, gid, 4); ipad[5] = pack64v (tmps, ipad, gid, 5); ipad[6] = pack64v (tmps, ipad, gid, 6); ipad[7] = pack64v (tmps, ipad, gid, 7); opad[0] = pack64v (tmps, opad, gid, 0); opad[1] = pack64v (tmps, opad, gid, 1); opad[2] = pack64v (tmps, opad, gid, 2); opad[3] = pack64v (tmps, opad, gid, 3); opad[4] = pack64v (tmps, opad, gid, 4); opad[5] = pack64v (tmps, opad, gid, 5); opad[6] = pack64v (tmps, opad, gid, 6); opad[7] = pack64v (tmps, opad, gid, 7); for (u32 i = 0; i < 8; i += 8) { u64x dgst[8]; u64x out[8]; dgst[0] = pack64v (tmps, dgst, gid, i + 0); dgst[1] = pack64v (tmps, dgst, gid, i + 1); dgst[2] = pack64v (tmps, dgst, gid, i + 2); dgst[3] = pack64v (tmps, dgst, gid, i + 3); dgst[4] = pack64v (tmps, dgst, gid, i + 4); dgst[5] = pack64v (tmps, dgst, gid, i + 5); dgst[6] = pack64v (tmps, dgst, gid, i + 6); dgst[7] = pack64v (tmps, dgst, gid, i + 7); out[0] = pack64v (tmps, out, gid, i + 0); out[1] = pack64v (tmps, out, gid, i + 1); out[2] = pack64v (tmps, out, gid, i + 2); out[3] = pack64v (tmps, out, gid, i + 3); out[4] = pack64v (tmps, out, gid, i + 4); out[5] = pack64v (tmps, out, gid, i + 5); out[6] = pack64v (tmps, out, gid, i + 6); out[7] = pack64v (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]; u32x w4[4]; u32x w5[4]; u32x w6[4]; u32x w7[4]; w0[0] = h32_from_64 (dgst[0]); w0[1] = l32_from_64 (dgst[0]); w0[2] = h32_from_64 (dgst[1]); w0[3] = l32_from_64 (dgst[1]); w1[0] = h32_from_64 (dgst[2]); w1[1] = l32_from_64 (dgst[2]); w1[2] = h32_from_64 (dgst[3]); w1[3] = l32_from_64 (dgst[3]); w2[0] = h32_from_64 (dgst[4]); w2[1] = l32_from_64 (dgst[4]); w2[2] = h32_from_64 (dgst[5]); w2[3] = l32_from_64 (dgst[5]); w3[0] = h32_from_64 (dgst[6]); w3[1] = l32_from_64 (dgst[6]); w3[2] = h32_from_64 (dgst[7]); w3[3] = l32_from_64 (dgst[7]); w4[0] = 0x80000000; w4[1] = 0; w4[2] = 0; w4[3] = 0; w5[0] = 0; w5[1] = 0; w5[2] = 0; w5[3] = 0; w6[0] = 0; w6[1] = 0; w6[2] = 0; w6[3] = 0; w7[0] = 0; w7[1] = 0; w7[2] = 0; w7[3] = (128 + 64) * 8; hmac_sha512_run_V (w0, w1, w2, w3, w4, w5, w6, w7, 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]; } unpack64v (tmps, dgst, gid, i + 0, dgst[0]); unpack64v (tmps, dgst, gid, i + 1, dgst[1]); unpack64v (tmps, dgst, gid, i + 2, dgst[2]); unpack64v (tmps, dgst, gid, i + 3, dgst[3]); unpack64v (tmps, dgst, gid, i + 4, dgst[4]); unpack64v (tmps, dgst, gid, i + 5, dgst[5]); unpack64v (tmps, dgst, gid, i + 6, dgst[6]); unpack64v (tmps, dgst, gid, i + 7, dgst[7]); unpack64v (tmps, out, gid, i + 0, out[0]); unpack64v (tmps, out, gid, i + 1, out[1]); unpack64v (tmps, out, gid, i + 2, out[2]); unpack64v (tmps, out, gid, i + 3, out[3]); unpack64v (tmps, out, gid, i + 4, out[4]); unpack64v (tmps, out, gid, i + 5, out[5]); unpack64v (tmps, out, gid, i + 6, out[6]); unpack64v (tmps, out, gid, i + 7, out[7]); } } KERNEL_FQ void m07100_comp (KERN_ATTR_TMPS_ESALT (pbkdf2_sha512_tmp_t, pbkdf2_sha512_t)) { /** * base */ const u64 gid = get_global_id (0); if (gid >= GID_MAX) return; const u64 lid = get_local_id (0); const u64 a = tmps[gid].out[0]; const u64 b = tmps[gid].out[1]; const u32 r0 = l32_from_64_S (a); const u32 r1 = h32_from_64_S (a); const u32 r2 = l32_from_64_S (b); const u32 r3 = h32_from_64_S (b); #define il_pos 0 #ifdef KERNEL_STATIC #include COMPARE_M #endif }