Page MenuHome GnuPG
Files F35313458

No OneTemporary
Actions

View Options

diff --git a/cipher/camellia-gfni-avx512-amd64.S b/cipher/camellia-gfni-avx512-amd64.S
index 643eed3e..22ae43d9 100644
--- a/cipher/camellia-gfni-avx512-amd64.S
+++ b/cipher/camellia-gfni-avx512-amd64.S
@@ -1,1634 +1,1868 @@
/* camellia-gfni-avx512-amd64.S - GFNI/AVX512 implementation of Camellia
*
- * Copyright (C) 2022-2023 Jussi Kivilinna <jussi.kivilinna@iki.fi>
+ * Copyright (C) 2022-2023,2025 Jussi Kivilinna <jussi.kivilinna@iki.fi>
*
* This file is part of Libgcrypt.
*
* Libgcrypt is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of
* the License, or (at your option) any later version.
*
* Libgcrypt is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include <config.h>
#ifdef __x86_64
#if (defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS)) && \
defined(ENABLE_GFNI_SUPPORT) && defined(ENABLE_AVX512_SUPPORT)
#include "asm-common-amd64.h"
#define CAMELLIA_TABLE_BYTE_LEN 272
/* struct CAMELLIA_context: */
#define key_table 0
#define key_bitlength CAMELLIA_TABLE_BYTE_LEN
/* register macros */
#define CTX %rdi
#define RIO %r8
/**********************************************************************
helper macros
**********************************************************************/
#define zmm0_x xmm0
#define zmm1_x xmm1
#define zmm2_x xmm2
#define zmm3_x xmm3
#define zmm4_x xmm4
#define zmm5_x xmm5
#define zmm6_x xmm6
#define zmm7_x xmm7
#define zmm8_x xmm8
#define zmm9_x xmm9
#define zmm10_x xmm10
#define zmm11_x xmm11
#define zmm12_x xmm12
#define zmm13_x xmm13
#define zmm14_x xmm14
#define zmm15_x xmm15
#define zmm0_y ymm0
#define zmm1_y ymm1
#define zmm2_y ymm2
#define zmm3_y ymm3
#define zmm4_y ymm4
#define zmm5_y ymm5
#define zmm6_y ymm6
#define zmm7_y ymm7
#define zmm8_y ymm8
#define zmm9_y ymm9
#define zmm10_y ymm10
#define zmm11_y ymm11
#define zmm12_y ymm12
#define zmm13_y ymm13
#define zmm14_y ymm14
#define zmm15_y ymm15
#define mem_ab_0 %zmm16
#define mem_ab_1 %zmm17
#define mem_ab_2 %zmm31
#define mem_ab_3 %zmm18
#define mem_ab_4 %zmm19
#define mem_ab_5 %zmm20
#define mem_ab_6 %zmm21
#define mem_ab_7 %zmm22
#define mem_cd_0 %zmm23
#define mem_cd_1 %zmm24
#define mem_cd_2 %zmm30
#define mem_cd_3 %zmm25
#define mem_cd_4 %zmm26
#define mem_cd_5 %zmm27
#define mem_cd_6 %zmm28
#define mem_cd_7 %zmm29
#define clear_vec4(v0,v1,v2,v3) \
vpxord v0, v0, v0; \
vpxord v1, v1, v1; \
vpxord v2, v2, v2; \
vpxord v3, v3, v3
#define clear_zmm16_zmm31() \
clear_vec4(%ymm16, %ymm20, %ymm24, %ymm28); \
clear_vec4(%ymm17, %ymm21, %ymm25, %ymm29); \
clear_vec4(%ymm18, %ymm22, %ymm26, %ymm30); \
clear_vec4(%ymm19, %ymm23, %ymm27, %ymm31)
#define clear_regs() \
vzeroall; \
clear_zmm16_zmm31()
/**********************************************************************
GFNI helper macros and constants
**********************************************************************/
#define BV8(a0,a1,a2,a3,a4,a5,a6,a7) \
( (((a0) & 1) << 0) | \
(((a1) & 1) << 1) | \
(((a2) & 1) << 2) | \
(((a3) & 1) << 3) | \
(((a4) & 1) << 4) | \
(((a5) & 1) << 5) | \
(((a6) & 1) << 6) | \
(((a7) & 1) << 7) )
#define BM8X8(l0,l1,l2,l3,l4,l5,l6,l7) \
( ((l7) << (0 * 8)) | \
((l6) << (1 * 8)) | \
((l5) << (2 * 8)) | \
((l4) << (3 * 8)) | \
((l3) << (4 * 8)) | \
((l2) << (5 * 8)) | \
((l1) << (6 * 8)) | \
((l0) << (7 * 8)) )
/* Pre-filters and post-filters constants for Camellia sboxes s1, s2, s3 and s4.
* See http://urn.fi/URN:NBN:fi:oulu-201305311409, pages 43-48.
*
* Pre-filters are directly from above source, "θ₁"/"θ₄". Post-filters are
* combination of function "A" (AES SubBytes affine transformation) and
* "ψ₁"/"ψ₂"/"ψ₃".
*/
/* Constant from "θ₁(x)" and "θ₄(x)" functions. */
#define pre_filter_constant_s1234 BV8(1, 0, 1, 0, 0, 0, 1, 0)
/* Constant from "ψ₁(A(x))" function: */
#define post_filter_constant_s14 BV8(0, 1, 1, 1, 0, 1, 1, 0)
/* Constant from "ψ₂(A(x))" function: */
#define post_filter_constant_s2 BV8(0, 0, 1, 1, 1, 0, 1, 1)
/* Constant from "ψ₃(A(x))" function: */
#define post_filter_constant_s3 BV8(1, 1, 1, 0, 1, 1, 0, 0)
/**********************************************************************
64-way parallel camellia
**********************************************************************/
/* roundsm64 (GFNI/AVX512 version)
* IN:
* x0..x7: byte-sliced AB state
* mem_cd: register pointer storing CD state
* key: index for key material
* OUT:
* x0..x7: new byte-sliced CD state
*/
#define roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, t0, t1, t2, t3, t4, t5, \
t6, t7, mem_cd, key) \
/* \
* S-function with AES subbytes \
*/ \
vpbroadcastq .Lpre_filter_bitmatrix_s123 rRIP, t5; \
vpbroadcastq .Lpre_filter_bitmatrix_s4 rRIP, t2; \
vpbroadcastq .Lpost_filter_bitmatrix_s14 rRIP, t4; \
vpbroadcastq .Lpost_filter_bitmatrix_s2 rRIP, t3; \
vpbroadcastq .Lpost_filter_bitmatrix_s3 rRIP, t6; \
\
/* prefilter sboxes */ \
vgf2p8affineqb $(pre_filter_constant_s1234), t5, x0, x0; \
vgf2p8affineqb $(pre_filter_constant_s1234), t5, x7, x7; \
vgf2p8affineqb $(pre_filter_constant_s1234), t2, x3, x3; \
vgf2p8affineqb $(pre_filter_constant_s1234), t2, x6, x6; \
vgf2p8affineqb $(pre_filter_constant_s1234), t5, x2, x2; \
vgf2p8affineqb $(pre_filter_constant_s1234), t5, x5, x5; \
vgf2p8affineqb $(pre_filter_constant_s1234), t5, x1, x1; \
vgf2p8affineqb $(pre_filter_constant_s1234), t5, x4, x4; \
\
/* sbox GF8 inverse + postfilter sboxes 1 and 4 */ \
vgf2p8affineinvqb $(post_filter_constant_s14), t4, x0, x0; \
vgf2p8affineinvqb $(post_filter_constant_s14), t4, x7, x7; \
vgf2p8affineinvqb $(post_filter_constant_s14), t4, x3, x3; \
vgf2p8affineinvqb $(post_filter_constant_s14), t4, x6, x6; \
\
/* sbox GF8 inverse + postfilter sbox 3 */ \
vgf2p8affineinvqb $(post_filter_constant_s3), t6, x2, x2; \
vgf2p8affineinvqb $(post_filter_constant_s3), t6, x5, x5; \
\
/* sbox GF8 inverse + postfilter sbox 2 */ \
vgf2p8affineinvqb $(post_filter_constant_s2), t3, x1, x1; \
vgf2p8affineinvqb $(post_filter_constant_s2), t3, x4, x4; \
\
vpbroadcastb 7+key, t7; \
vpbroadcastb 6+key, t6; \
\
/* P-function */ \
vpxorq x5, x0, x0; \
vpxorq x6, x1, x1; \
vpxorq x7, x2, x2; \
vpxorq x4, x3, x3; \
\
vpbroadcastb 5+key, t5; \
vpbroadcastb 4+key, t4; \
\
vpxorq x2, x4, x4; \
vpxorq x3, x5, x5; \
vpxorq x0, x6, x6; \
vpxorq x1, x7, x7; \
\
vpbroadcastb 3+key, t3; \
vpbroadcastb 2+key, t2; \
\
vpxorq x7, x0, x0; \
vpxorq x4, x1, x1; \
vpxorq x5, x2, x2; \
vpxorq x6, x3, x3; \
\
vpbroadcastb 1+key, t1; \
vpbroadcastb 0+key, t0; \
\
vpxorq x3, x4, x4; \
vpxorq x0, x5, x5; \
vpxorq x1, x6, x6; \
vpxorq x2, x7, x7; /* note: high and low parts swapped */ \
\
/* Add key material and result to CD (x becomes new CD) */ \
\
vpternlogq $0x96, mem_cd##_4, t7, x0; \
vpternlogq $0x96, mem_cd##_5, t6, x1; \
vpternlogq $0x96, mem_cd##_6, t5, x2; \
vpternlogq $0x96, mem_cd##_7, t4, x3; \
vpternlogq $0x96, mem_cd##_0, t3, x4; \
vpternlogq $0x96, mem_cd##_1, t2, x5; \
vpternlogq $0x96, mem_cd##_2, t1, x6; \
vpternlogq $0x96, mem_cd##_3, t0, x7;
/*
* IN/OUT:
* x0..x7: byte-sliced AB state preloaded
* mem_ab: byte-sliced AB state in memory
* mem_cb: byte-sliced CD state in memory
*/
#define two_roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, i, dir, store_ab) \
roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_cd, (key_table + (i) * 8)(CTX)); \
\
vmovdqu64 x0, mem_cd##_4; \
vmovdqu64 x1, mem_cd##_5; \
vmovdqu64 x2, mem_cd##_6; \
vmovdqu64 x3, mem_cd##_7; \
vmovdqu64 x4, mem_cd##_0; \
vmovdqu64 x5, mem_cd##_1; \
vmovdqu64 x6, mem_cd##_2; \
vmovdqu64 x7, mem_cd##_3; \
\
roundsm64(x4, x5, x6, x7, x0, x1, x2, x3, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, (key_table + ((i) + (dir)) * 8)(CTX)); \
\
store_ab(x0, x1, x2, x3, x4, x5, x6, x7, mem_ab);
#define dummy_store(x0, x1, x2, x3, x4, x5, x6, x7, mem_ab) /* do nothing */
#define store_ab_state(x0, x1, x2, x3, x4, x5, x6, x7, mem_ab) \
/* Store new AB state */ \
vmovdqu64 x4, mem_ab##_4; \
vmovdqu64 x5, mem_ab##_5; \
vmovdqu64 x6, mem_ab##_6; \
vmovdqu64 x7, mem_ab##_7; \
vmovdqu64 x0, mem_ab##_0; \
vmovdqu64 x1, mem_ab##_1; \
vmovdqu64 x2, mem_ab##_2; \
vmovdqu64 x3, mem_ab##_3;
#define enc_rounds64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, i) \
two_roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, (i) + 2, 1, store_ab_state); \
two_roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, (i) + 4, 1, store_ab_state); \
two_roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, (i) + 6, 1, dummy_store);
#define dec_rounds64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, i) \
two_roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, (i) + 7, -1, store_ab_state); \
two_roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, (i) + 5, -1, store_ab_state); \
two_roundsm64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, (i) + 3, -1, dummy_store);
/*
* IN:
* v0..3: byte-sliced 32-bit integers
* OUT:
* v0..3: (IN << 1)
* t0, t1, t2, zero: (IN >> 7)
*/
#define rol32_1_64(v0, v1, v2, v3, t0, t1, t2, t3, right_shift_by_7) \
vgf2p8affineqb $0, right_shift_by_7, v0, t0; \
vpaddb v0, v0, v0; \
\
vgf2p8affineqb $0, right_shift_by_7, v1, t1; \
vpaddb v1, v1, v1; \
\
vgf2p8affineqb $0, right_shift_by_7, v2, t2; \
vpaddb v2, v2, v2; \
\
vgf2p8affineqb $0, right_shift_by_7, v3, t3; \
vpaddb v3, v3, v3;
/*
* IN:
* r: byte-sliced AB state in memory
* l: byte-sliced CD state in memory
* OUT:
* x0..x7: new byte-sliced CD state
*/
#define fls64(l, l0, l1, l2, l3, l4, l5, l6, l7, r, t0, t1, t2, t3, tt0, \
tt1, tt2, tt3, kll, klr, krl, krr, tmp) \
/* \
* t0 = kll; \
* t0 &= ll; \
* lr ^= rol32(t0, 1); \
*/ \
vpbroadcastq .Lright_shift_by_7 rRIP, tmp; \
vpbroadcastb 0+kll, t3; \
vpbroadcastb 1+kll, t2; \
vpbroadcastb 2+kll, t1; \
vpbroadcastb 3+kll, t0; \
\
vpandq l0, t0, t0; \
vpandq l1, t1, t1; \
vpandq l2, t2, t2; \
vpandq l3, t3, t3; \
\
rol32_1_64(t3, t2, t1, t0, tt0, tt1, tt2, tt3, tmp); \
\
vpternlogq $0x96, tt2, t0, l4; \
vmovdqu64 l4, l##_4; \
vpternlogq $0x96, tt1, t1, l5; \
vmovdqu64 l5, l##_5; \
vpternlogq $0x96, tt0, t2, l6; \
vmovdqu64 l6, l##_6; \
vpternlogq $0x96, tt3, t3, l7; \
vmovdqu64 l7, l##_7; \
\
/* \
* t2 = krr; \
* t2 |= rr; \
* rl ^= t2; \
*/ \
\
vpbroadcastb 0+krr, t3; \
vpbroadcastb 1+krr, t2; \
vpbroadcastb 2+krr, t1; \
vpbroadcastb 3+krr, t0; \
\
vpternlogq $0x1e, r##_4, t0, r##_0; \
vpternlogq $0x1e, r##_5, t1, r##_1; \
vpternlogq $0x1e, r##_6, t2, r##_2; \
vpternlogq $0x1e, r##_7, t3, r##_3; \
\
/* \
* t2 = krl; \
* t2 &= rl; \
* rr ^= rol32(t2, 1); \
*/ \
vpbroadcastb 0+krl, t3; \
vpbroadcastb 1+krl, t2; \
vpbroadcastb 2+krl, t1; \
vpbroadcastb 3+krl, t0; \
\
vpandq r##_0, t0, t0; \
vpandq r##_1, t1, t1; \
vpandq r##_2, t2, t2; \
vpandq r##_3, t3, t3; \
\
rol32_1_64(t3, t2, t1, t0, tt0, tt1, tt2, tt3, tmp); \
\
vpternlogq $0x96, tt2, t0, r##_4; \
vpternlogq $0x96, tt1, t1, r##_5; \
vpternlogq $0x96, tt0, t2, r##_6; \
vpternlogq $0x96, tt3, t3, r##_7; \
\
/* \
* t0 = klr; \
* t0 |= lr; \
* ll ^= t0; \
*/ \
\
vpbroadcastb 0+klr, t3; \
vpbroadcastb 1+klr, t2; \
vpbroadcastb 2+klr, t1; \
vpbroadcastb 3+klr, t0; \
\
vpternlogq $0x1e, l4, t0, l0; \
vmovdqu64 l0, l##_0; \
vpternlogq $0x1e, l5, t1, l1; \
vmovdqu64 l1, l##_1; \
vpternlogq $0x1e, l6, t2, l2; \
vmovdqu64 l2, l##_2; \
vpternlogq $0x1e, l7, t3, l3; \
vmovdqu64 l3, l##_3;
#define transpose_4x4(x0, x1, x2, x3, t1, t2) \
vpunpckhdq x1, x0, t2; \
vpunpckldq x1, x0, x0; \
\
vpunpckldq x3, x2, t1; \
vpunpckhdq x3, x2, x2; \
\
vpunpckhqdq t1, x0, x1; \
vpunpcklqdq t1, x0, x0; \
\
vpunpckhqdq x2, t2, x3; \
vpunpcklqdq x2, t2, x2;
#define byteslice_16x16b_fast(a0, b0, c0, d0, a1, b1, c1, d1, a2, b2, c2, d2, \
a3, b3, c3, d3, st0, st1) \
transpose_4x4(a0, a1, a2, a3, st0, st1); \
transpose_4x4(b0, b1, b2, b3, st0, st1); \
\
transpose_4x4(c0, c1, c2, c3, st0, st1); \
transpose_4x4(d0, d1, d2, d3, st0, st1); \
\
vbroadcasti64x2 .Lshufb_16x16b rRIP, st0; \
vpshufb st0, a0, a0; \
vpshufb st0, a1, a1; \
vpshufb st0, a2, a2; \
vpshufb st0, a3, a3; \
vpshufb st0, b0, b0; \
vpshufb st0, b1, b1; \
vpshufb st0, b2, b2; \
vpshufb st0, b3, b3; \
vpshufb st0, c0, c0; \
vpshufb st0, c1, c1; \
vpshufb st0, c2, c2; \
vpshufb st0, c3, c3; \
vpshufb st0, d0, d0; \
vpshufb st0, d1, d1; \
vpshufb st0, d2, d2; \
vpshufb st0, d3, d3; \
\
transpose_4x4(a0, b0, c0, d0, st0, st1); \
transpose_4x4(a1, b1, c1, d1, st0, st1); \
\
transpose_4x4(a2, b2, c2, d2, st0, st1); \
transpose_4x4(a3, b3, c3, d3, st0, st1); \
/* does not adjust output bytes inside vectors */
/* load blocks to registers and apply pre-whitening */
#define inpack64_pre(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, rio, key) \
vpbroadcastq key, x0; \
vpshufb .Lpack_bswap rRIP, x0, x0; \
\
vpxorq 0 * 64(rio), x0, y7; \
vpxorq 1 * 64(rio), x0, y6; \
vpxorq 2 * 64(rio), x0, y5; \
vpxorq 3 * 64(rio), x0, y4; \
vpxorq 4 * 64(rio), x0, y3; \
vpxorq 5 * 64(rio), x0, y2; \
vpxorq 6 * 64(rio), x0, y1; \
vpxorq 7 * 64(rio), x0, y0; \
vpxorq 8 * 64(rio), x0, x7; \
vpxorq 9 * 64(rio), x0, x6; \
vpxorq 10 * 64(rio), x0, x5; \
vpxorq 11 * 64(rio), x0, x4; \
vpxorq 12 * 64(rio), x0, x3; \
vpxorq 13 * 64(rio), x0, x2; \
vpxorq 14 * 64(rio), x0, x1; \
vpxorq 15 * 64(rio), x0, x0;
/* byteslice pre-whitened blocks and store to temporary memory */
#define inpack64_post(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, mem_ab, mem_cd, tmp0, tmp1) \
byteslice_16x16b_fast(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, \
y4, y5, y6, y7, tmp0, tmp1); \
\
vmovdqu64 x0, mem_ab##_0; \
vmovdqu64 x1, mem_ab##_1; \
vmovdqu64 x2, mem_ab##_2; \
vmovdqu64 x3, mem_ab##_3; \
vmovdqu64 x4, mem_ab##_4; \
vmovdqu64 x5, mem_ab##_5; \
vmovdqu64 x6, mem_ab##_6; \
vmovdqu64 x7, mem_ab##_7; \
vmovdqu64 y0, mem_cd##_0; \
vmovdqu64 y1, mem_cd##_1; \
vmovdqu64 y2, mem_cd##_2; \
vmovdqu64 y3, mem_cd##_3; \
vmovdqu64 y4, mem_cd##_4; \
vmovdqu64 y5, mem_cd##_5; \
vmovdqu64 y6, mem_cd##_6; \
vmovdqu64 y7, mem_cd##_7;
/* de-byteslice, apply post-whitening and store blocks */
#define outunpack64(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, \
y5, y6, y7, key, tmp0, tmp1) \
byteslice_16x16b_fast(y0, y4, x0, x4, y1, y5, x1, x5, y2, y6, x2, x6, \
y3, y7, x3, x7, tmp0, tmp1); \
\
vpbroadcastq key, tmp0; \
vpshufb .Lpack_bswap rRIP, tmp0, tmp0; \
\
vpxorq tmp0, y7, y7; \
vpxorq tmp0, y6, y6; \
vpxorq tmp0, y5, y5; \
vpxorq tmp0, y4, y4; \
vpxorq tmp0, y3, y3; \
vpxorq tmp0, y2, y2; \
vpxorq tmp0, y1, y1; \
vpxorq tmp0, y0, y0; \
vpxorq tmp0, x7, x7; \
vpxorq tmp0, x6, x6; \
vpxorq tmp0, x5, x5; \
vpxorq tmp0, x4, x4; \
vpxorq tmp0, x3, x3; \
vpxorq tmp0, x2, x2; \
vpxorq tmp0, x1, x1; \
vpxorq tmp0, x0, x0;
#define write_output(x0, x1, x2, x3, x4, x5, x6, x7, y0, y1, y2, y3, y4, y5, \
y6, y7, rio) \
vmovdqu64 x0, 0 * 64(rio); \
vmovdqu64 x1, 1 * 64(rio); \
vmovdqu64 x2, 2 * 64(rio); \
vmovdqu64 x3, 3 * 64(rio); \
vmovdqu64 x4, 4 * 64(rio); \
vmovdqu64 x5, 5 * 64(rio); \
vmovdqu64 x6, 6 * 64(rio); \
vmovdqu64 x7, 7 * 64(rio); \
vmovdqu64 y0, 8 * 64(rio); \
vmovdqu64 y1, 9 * 64(rio); \
vmovdqu64 y2, 10 * 64(rio); \
vmovdqu64 y3, 11 * 64(rio); \
vmovdqu64 y4, 12 * 64(rio); \
vmovdqu64 y5, 13 * 64(rio); \
vmovdqu64 y6, 14 * 64(rio); \
vmovdqu64 y7, 15 * 64(rio);
SECTION_RODATA
#define SHUFB_BYTES(idx) \
0 + (idx), 4 + (idx), 8 + (idx), 12 + (idx)
_gcry_camellia_gfni_avx512__constants:
ELF(.type _gcry_camellia_gfni_avx512__constants,@object;)
.align 64
.Lpack_bswap:
.long 0x00010203, 0x04050607, 0x80808080, 0x80808080
.long 0x00010203, 0x04050607, 0x80808080, 0x80808080
.long 0x00010203, 0x04050607, 0x80808080, 0x80808080
.long 0x00010203, 0x04050607, 0x80808080, 0x80808080
.Lcounter0123_lo:
.quad 0, 0
.quad 1, 0
.quad 2, 0
.quad 3, 0
.align 16
.Lcounter4444_lo:
.quad 4, 0
.Lcounter8888_lo:
.quad 8, 0
.Lcounter16161616_lo:
.quad 16, 0
.Lcounter1111_hi:
.quad 0, 1
.Lshufb_16x16b:
.byte SHUFB_BYTES(0), SHUFB_BYTES(1), SHUFB_BYTES(2), SHUFB_BYTES(3)
/* For CTR-mode IV byteswap */
.Lbswap128_mask:
.byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
/* Pre-filters and post-filters bit-matrixes for Camellia sboxes s1, s2, s3
* and s4.
* See http://urn.fi/URN:NBN:fi:oulu-201305311409, pages 43-48.
*
* Pre-filters are directly from above source, "θ₁"/"θ₄". Post-filters are
* combination of function "A" (AES SubBytes affine transformation) and
* "ψ₁"/"ψ₂"/"ψ₃".
*/
/* Bit-matrix from "θ₁(x)" function: */
.Lpre_filter_bitmatrix_s123:
.quad BM8X8(BV8(1, 1, 1, 0, 1, 1, 0, 1),
BV8(0, 0, 1, 1, 0, 0, 1, 0),
BV8(1, 1, 0, 1, 0, 0, 0, 0),
BV8(1, 0, 1, 1, 0, 0, 1, 1),
BV8(0, 0, 0, 0, 1, 1, 0, 0),
BV8(1, 0, 1, 0, 0, 1, 0, 0),
BV8(0, 0, 1, 0, 1, 1, 0, 0),
BV8(1, 0, 0, 0, 0, 1, 1, 0))
/* Bit-matrix from "θ₄(x)" function: */
.Lpre_filter_bitmatrix_s4:
.quad BM8X8(BV8(1, 1, 0, 1, 1, 0, 1, 1),
BV8(0, 1, 1, 0, 0, 1, 0, 0),
BV8(1, 0, 1, 0, 0, 0, 0, 1),
BV8(0, 1, 1, 0, 0, 1, 1, 1),
BV8(0, 0, 0, 1, 1, 0, 0, 0),
BV8(0, 1, 0, 0, 1, 0, 0, 1),
BV8(0, 1, 0, 1, 1, 0, 0, 0),
BV8(0, 0, 0, 0, 1, 1, 0, 1))
/* Bit-matrix from "ψ₁(A(x))" function: */
.Lpost_filter_bitmatrix_s14:
.quad BM8X8(BV8(0, 0, 0, 0, 0, 0, 0, 1),
BV8(0, 1, 1, 0, 0, 1, 1, 0),
BV8(1, 0, 1, 1, 1, 1, 1, 0),
BV8(0, 0, 0, 1, 1, 0, 1, 1),
BV8(1, 0, 0, 0, 1, 1, 1, 0),
BV8(0, 1, 0, 1, 1, 1, 1, 0),
BV8(0, 1, 1, 1, 1, 1, 1, 1),
BV8(0, 0, 0, 1, 1, 1, 0, 0))
/* Bit-matrix from "ψ₂(A(x))" function: */
.Lpost_filter_bitmatrix_s2:
.quad BM8X8(BV8(0, 0, 0, 1, 1, 1, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 1),
BV8(0, 1, 1, 0, 0, 1, 1, 0),
BV8(1, 0, 1, 1, 1, 1, 1, 0),
BV8(0, 0, 0, 1, 1, 0, 1, 1),
BV8(1, 0, 0, 0, 1, 1, 1, 0),
BV8(0, 1, 0, 1, 1, 1, 1, 0),
BV8(0, 1, 1, 1, 1, 1, 1, 1))
/* Bit-matrix from "ψ₃(A(x))" function: */
.Lpost_filter_bitmatrix_s3:
.quad BM8X8(BV8(0, 1, 1, 0, 0, 1, 1, 0),
BV8(1, 0, 1, 1, 1, 1, 1, 0),
BV8(0, 0, 0, 1, 1, 0, 1, 1),
BV8(1, 0, 0, 0, 1, 1, 1, 0),
BV8(0, 1, 0, 1, 1, 1, 1, 0),
BV8(0, 1, 1, 1, 1, 1, 1, 1),
BV8(0, 0, 0, 1, 1, 1, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 1))
/* Bit-matrix for right shifting uint8_t values in vector by 7. */
.Lright_shift_by_7:
.quad BM8X8(BV8(0, 0, 0, 0, 0, 0, 0, 1),
BV8(0, 0, 0, 0, 0, 0, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 0),
BV8(0, 0, 0, 0, 0, 0, 0, 0))
/* CTR byte addition constants */
.align 64
.Lbige_addb_0_1:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1
.Lbige_addb_2_3:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3
.Lbige_addb_4_5:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5
.Lbige_addb_6_7:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 6
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7
.Lbige_addb_8_9:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 9
.Lbige_addb_10_11:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 10
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 11
.Lbige_addb_12_13:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 12
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 13
.Lbige_addb_14_15:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 14
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 15
.Lbige_addb_16:
.byte 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 16
+.align 16
+/* Shuffling constants for AVX512+GFNI 1-way variant. */
+.Lsp1mask_swap32_gfni:
+ .byte 0xff, 0x04, 0x04, 0x04, 0xff, 0x04, 0x04, 0x04
+ .byte 0xff, 0x03, 0x03, 0x03, 0x03, 0xff, 0xff, 0x03
+.Lsp2mask_swap32_gfni:
+ .byte 0x07, 0x07, 0x07, 0xff, 0x07, 0x07, 0x07, 0xff
+ .byte 0x02, 0x02, 0x02, 0xff, 0xff, 0xff, 0x02, 0x02
+.Lsp3mask_swap32_gfni:
+ .byte 0x06, 0x06, 0xff, 0x06, 0x06, 0x06, 0xff, 0x06
+ .byte 0x01, 0x01, 0xff, 0x01, 0xff, 0x01, 0x01, 0xff
+.Lsp4mask_swap32_gfni:
+ .byte 0x00, 0xff, 0x00, 0x00, 0x00, 0x00, 0xff, 0xff
+ .byte 0x05, 0xff, 0x05, 0x05, 0x05, 0xff, 0x05, 0x05
+
ELF(.size _gcry_camellia_gfni_avx512__constants,.-_gcry_camellia_gfni_avx512__constants;)
.text
.align 16
ELF(.type __camellia_gfni_avx512_enc_blk64,@function;)
__camellia_gfni_avx512_enc_blk64:
/* input:
* %rdi: ctx, CTX
* %r8d: 24 for 16 byte key, 32 for larger
* %zmm0..%zmm15: 64 plaintext blocks
* output:
* %zmm0..%zmm15: 64 encrypted blocks, order swapped:
* 7, 8, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8
*/
CFI_STARTPROC();
leaq (-8 * 8)(CTX, %r8, 8), %r8;
inpack64_post(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
%zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15, mem_ab, mem_cd, %zmm30, %zmm31);
.align 8
.Lenc_loop:
enc_rounds64(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
%zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15, mem_ab, mem_cd, 0);
cmpq %r8, CTX;
je .Lenc_done;
leaq (8 * 8)(CTX), CTX;
fls64(mem_ab, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
mem_cd, %zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15,
((key_table) + 0)(CTX),
((key_table) + 4)(CTX),
((key_table) + 8)(CTX),
((key_table) + 12)(CTX),
%zmm31);
jmp .Lenc_loop;
.align 8
.Lenc_done:
/* load CD for output */
vmovdqu64 mem_cd_0, %zmm8;
vmovdqu64 mem_cd_1, %zmm9;
vmovdqu64 mem_cd_2, %zmm10;
vmovdqu64 mem_cd_3, %zmm11;
vmovdqu64 mem_cd_4, %zmm12;
vmovdqu64 mem_cd_5, %zmm13;
vmovdqu64 mem_cd_6, %zmm14;
vmovdqu64 mem_cd_7, %zmm15;
outunpack64(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
%zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15, ((key_table) + 8 * 8)(%r8), %zmm30, %zmm31);
ret_spec_stop;
CFI_ENDPROC();
ELF(.size __camellia_gfni_avx512_enc_blk64,.-__camellia_gfni_avx512_enc_blk64;)
.align 16
ELF(.type __camellia_gfni_avx512_dec_blk64,@function;)
__camellia_gfni_avx512_dec_blk64:
/* input:
* %rdi: ctx, CTX
* %r8d: 24 for 16 byte key, 32 for larger
* %zmm0..%zmm15: 64 encrypted blocks
* output:
* %zmm0..%zmm15: 64 plaintext blocks, order swapped:
* 7, 8, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8
*/
CFI_STARTPROC();
movq %r8, %rcx;
movq CTX, %r8
leaq (-8 * 8)(CTX, %rcx, 8), CTX;
inpack64_post(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
%zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15, mem_ab, mem_cd, %zmm30, %zmm31);
.align 8
.Ldec_loop:
dec_rounds64(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
%zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15, mem_ab, mem_cd, 0);
cmpq %r8, CTX;
je .Ldec_done;
fls64(mem_ab, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
mem_cd, %zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15,
((key_table) + 8)(CTX),
((key_table) + 12)(CTX),
((key_table) + 0)(CTX),
((key_table) + 4)(CTX),
%zmm31);
leaq (-8 * 8)(CTX), CTX;
jmp .Ldec_loop;
.align 8
.Ldec_done:
/* load CD for output */
vmovdqu64 mem_cd_0, %zmm8;
vmovdqu64 mem_cd_1, %zmm9;
vmovdqu64 mem_cd_2, %zmm10;
vmovdqu64 mem_cd_3, %zmm11;
vmovdqu64 mem_cd_4, %zmm12;
vmovdqu64 mem_cd_5, %zmm13;
vmovdqu64 mem_cd_6, %zmm14;
vmovdqu64 mem_cd_7, %zmm15;
outunpack64(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
%zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15, (key_table)(CTX), %zmm30, %zmm31);
ret_spec_stop;
CFI_ENDPROC();
ELF(.size __camellia_gfni_avx512_dec_blk64,.-__camellia_gfni_avx512_dec_blk64;)
#define add_le128(out, in, lo_counter, hi_counter1) \
vpaddq lo_counter, in, out; \
vpcmpuq $1, lo_counter, out, %k1; \
kaddb %k1, %k1, %k1; \
vpaddq hi_counter1, out, out{%k1};
.align 16
.globl _gcry_camellia_gfni_avx512_ctr_enc
ELF(.type _gcry_camellia_gfni_avx512_ctr_enc,@function;)
_gcry_camellia_gfni_avx512_ctr_enc:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (64 blocks)
* %rdx: src (64 blocks)
* %rcx: iv (big endian, 128bit)
*/
CFI_STARTPROC();
spec_stop_avx512;
cmpl $128, key_bitlength(CTX);
movl $32, %r8d;
movl $24, %eax;
cmovel %eax, %r8d; /* max */
cmpb $(0x100 - 64), 15(%rcx);
jbe .Lctr_byteadd;
vbroadcasti64x2 .Lbswap128_mask rRIP, %zmm19;
vmovdqa64 .Lcounter0123_lo rRIP, %zmm21;
vbroadcasti64x2 .Lcounter4444_lo rRIP, %zmm22;
vbroadcasti64x2 .Lcounter8888_lo rRIP, %zmm23;
vbroadcasti64x2 .Lcounter16161616_lo rRIP, %zmm24;
vbroadcasti64x2 .Lcounter1111_hi rRIP, %zmm25;
/* load IV and byteswap */
movq 8(%rcx), %r11;
movq (%rcx), %r10;
bswapq %r11;
bswapq %r10;
vbroadcasti64x2 (%rcx), %zmm0;
vpshufb %zmm19, %zmm0, %zmm0;
/* check need for handling 64-bit overflow and carry */
cmpq $(0xffffffffffffffff - 64), %r11;
ja .Lload_ctr_carry;
/* construct IVs */
vpaddq %zmm21, %zmm0, %zmm15; /* +0:+1:+2:+3 */
vpaddq %zmm22, %zmm15, %zmm14; /* +4:+5:+6:+7 */
vpaddq %zmm23, %zmm15, %zmm13; /* +8:+9:+10:+11 */
vpaddq %zmm23, %zmm14, %zmm12; /* +12:+13:+14:+15 */
vpaddq %zmm24, %zmm15, %zmm11; /* +16... */
vpaddq %zmm24, %zmm14, %zmm10; /* +20... */
vpaddq %zmm24, %zmm13, %zmm9; /* +24... */
vpaddq %zmm24, %zmm12, %zmm8; /* +28... */
vpaddq %zmm24, %zmm11, %zmm7; /* +32... */
vpaddq %zmm24, %zmm10, %zmm6; /* +36... */
vpaddq %zmm24, %zmm9, %zmm5; /* +40... */
vpaddq %zmm24, %zmm8, %zmm4; /* +44... */
vpaddq %zmm24, %zmm7, %zmm3; /* +48... */
vpaddq %zmm24, %zmm6, %zmm2; /* +52... */
vpaddq %zmm24, %zmm5, %zmm1; /* +56... */
vpaddq %zmm24, %zmm4, %zmm0; /* +60... */
jmp .Lload_ctr_done;
.align 4
.Lload_ctr_carry:
/* construct IVs */
add_le128(%zmm15, %zmm0, %zmm21, %zmm25); /* +0:+1:+2:+3 */
add_le128(%zmm14, %zmm15, %zmm22, %zmm25); /* +4:+5:+6:+7 */
add_le128(%zmm13, %zmm15, %zmm23, %zmm25); /* +8:+9:+10:+11 */
add_le128(%zmm12, %zmm14, %zmm23, %zmm25); /* +12:+13:+14:+15 */
add_le128(%zmm11, %zmm15, %zmm24, %zmm25); /* +16... */
add_le128(%zmm10, %zmm14, %zmm24, %zmm25); /* +20... */
add_le128(%zmm9, %zmm13, %zmm24, %zmm25); /* +24... */
add_le128(%zmm8, %zmm12, %zmm24, %zmm25); /* +28... */
add_le128(%zmm7, %zmm11, %zmm24, %zmm25); /* +32... */
add_le128(%zmm6, %zmm10, %zmm24, %zmm25); /* +36... */
add_le128(%zmm5, %zmm9, %zmm24, %zmm25); /* +40... */
add_le128(%zmm4, %zmm8, %zmm24, %zmm25); /* +44... */
add_le128(%zmm3, %zmm7, %zmm24, %zmm25); /* +48... */
add_le128(%zmm2, %zmm6, %zmm24, %zmm25); /* +52... */
add_le128(%zmm1, %zmm5, %zmm24, %zmm25); /* +56... */
add_le128(%zmm0, %zmm4, %zmm24, %zmm25); /* +60... */
kxorq %k1, %k1, %k1;
.align 4
.Lload_ctr_done:
vbroadcasti64x2 .Lpack_bswap rRIP, %zmm17;
vpbroadcastq (key_table)(CTX), %zmm16;
vpshufb %zmm17, %zmm16, %zmm16;
/* Byte-swap IVs and update counter. */
addq $64, %r11;
adcq $0, %r10;
vpshufb %zmm19, %zmm15, %zmm15;
vpshufb %zmm19, %zmm14, %zmm14;
vpshufb %zmm19, %zmm13, %zmm13;
vpshufb %zmm19, %zmm12, %zmm12;
vpshufb %zmm19, %zmm11, %zmm11;
vpshufb %zmm19, %zmm10, %zmm10;
vpshufb %zmm19, %zmm9, %zmm9;
vpshufb %zmm19, %zmm8, %zmm8;
bswapq %r11;
bswapq %r10;
vpshufb %zmm19, %zmm7, %zmm7;
vpshufb %zmm19, %zmm6, %zmm6;
vpshufb %zmm19, %zmm5, %zmm5;
vpshufb %zmm19, %zmm4, %zmm4;
vpshufb %zmm19, %zmm3, %zmm3;
vpshufb %zmm19, %zmm2, %zmm2;
vpshufb %zmm19, %zmm1, %zmm1;
vpshufb %zmm19, %zmm0, %zmm0;
movq %r11, 8(%rcx);
movq %r10, (%rcx);
.align 16
.Lctr_inpack64_pre:
/* inpack64_pre: */
vpxorq %zmm0, %zmm16, %zmm0;
vpxorq %zmm1, %zmm16, %zmm1;
vpxorq %zmm2, %zmm16, %zmm2;
vpxorq %zmm3, %zmm16, %zmm3;
vpxorq %zmm4, %zmm16, %zmm4;
vpxorq %zmm5, %zmm16, %zmm5;
vpxorq %zmm6, %zmm16, %zmm6;
vpxorq %zmm7, %zmm16, %zmm7;
vpxorq %zmm8, %zmm16, %zmm8;
vpxorq %zmm9, %zmm16, %zmm9;
vpxorq %zmm10, %zmm16, %zmm10;
vpxorq %zmm11, %zmm16, %zmm11;
vpxorq %zmm12, %zmm16, %zmm12;
vpxorq %zmm13, %zmm16, %zmm13;
vpxorq %zmm14, %zmm16, %zmm14;
vpxorq %zmm15, %zmm16, %zmm15;
call __camellia_gfni_avx512_enc_blk64;
vpxorq 0 * 64(%rdx), %zmm7, %zmm7;
vpxorq 1 * 64(%rdx), %zmm6, %zmm6;
vpxorq 2 * 64(%rdx), %zmm5, %zmm5;
vpxorq 3 * 64(%rdx), %zmm4, %zmm4;
vpxorq 4 * 64(%rdx), %zmm3, %zmm3;
vpxorq 5 * 64(%rdx), %zmm2, %zmm2;
vpxorq 6 * 64(%rdx), %zmm1, %zmm1;
vpxorq 7 * 64(%rdx), %zmm0, %zmm0;
vpxorq 8 * 64(%rdx), %zmm15, %zmm15;
vpxorq 9 * 64(%rdx), %zmm14, %zmm14;
vpxorq 10 * 64(%rdx), %zmm13, %zmm13;
vpxorq 11 * 64(%rdx), %zmm12, %zmm12;
vpxorq 12 * 64(%rdx), %zmm11, %zmm11;
vpxorq 13 * 64(%rdx), %zmm10, %zmm10;
vpxorq 14 * 64(%rdx), %zmm9, %zmm9;
vpxorq 15 * 64(%rdx), %zmm8, %zmm8;
write_output(%zmm7, %zmm6, %zmm5, %zmm4, %zmm3, %zmm2, %zmm1, %zmm0,
%zmm15, %zmm14, %zmm13, %zmm12, %zmm11, %zmm10, %zmm9,
%zmm8, %rsi);
clear_regs();
ret_spec_stop;
.align 16
.Lctr_byteadd_full_ctr_carry:
movq 8(%rcx), %r11;
movq (%rcx), %r10;
bswapq %r11;
bswapq %r10;
addq $64, %r11;
adcq $0, %r10;
bswapq %r11;
bswapq %r10;
movq %r11, 8(%rcx);
movq %r10, (%rcx);
jmp .Lctr_byteadd_zmm;
.align 16
.Lctr_byteadd:
vbroadcasti64x2 (%rcx), %zmm12;
je .Lctr_byteadd_full_ctr_carry;
addb $64, 15(%rcx);
.Lctr_byteadd_zmm:
vbroadcasti64x2 .Lbige_addb_16 rRIP, %zmm16;
vmovdqa64 .Lbige_addb_0_1 rRIP, %zmm17;
vmovdqa64 .Lbige_addb_4_5 rRIP, %zmm18;
vmovdqa64 .Lbige_addb_8_9 rRIP, %zmm19;
vmovdqa64 .Lbige_addb_12_13 rRIP, %zmm20;
vpaddb %zmm16, %zmm12, %zmm8;
vpaddb %zmm17, %zmm12, %zmm15;
vpaddb %zmm18, %zmm12, %zmm14;
vpaddb %zmm19, %zmm12, %zmm13;
vpaddb %zmm20, %zmm12, %zmm12;
vpaddb %zmm16, %zmm8, %zmm4;
vpaddb %zmm17, %zmm8, %zmm11;
vpaddb %zmm18, %zmm8, %zmm10;
vpaddb %zmm19, %zmm8, %zmm9;
vpaddb %zmm20, %zmm8, %zmm8;
vpaddb %zmm16, %zmm4, %zmm0;
vpaddb %zmm17, %zmm4, %zmm7;
vpaddb %zmm18, %zmm4, %zmm6;
vpaddb %zmm19, %zmm4, %zmm5;
vpaddb %zmm20, %zmm4, %zmm4;
vpaddb %zmm17, %zmm0, %zmm3;
vpaddb %zmm18, %zmm0, %zmm2;
vpaddb %zmm19, %zmm0, %zmm1;
vpaddb %zmm20, %zmm0, %zmm0;
vbroadcasti64x2 .Lpack_bswap rRIP, %zmm17
vpbroadcastq (key_table)(CTX), %zmm16;
vpshufb %zmm17, %zmm16, %zmm16;
jmp .Lctr_inpack64_pre;
CFI_ENDPROC();
ELF(.size _gcry_camellia_gfni_avx512_ctr_enc,.-_gcry_camellia_gfni_avx512_ctr_enc;)
.align 16
.globl _gcry_camellia_gfni_avx512_cbc_dec
ELF(.type _gcry_camellia_gfni_avx512_cbc_dec,@function;)
_gcry_camellia_gfni_avx512_cbc_dec:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (64 blocks)
* %rdx: src (64 blocks)
* %rcx: iv
*/
CFI_STARTPROC();
spec_stop_avx512;
movq %rcx, %r9;
cmpl $128, key_bitlength(CTX);
movl $32, %r8d;
movl $24, %eax;
cmovel %eax, %r8d; /* max */
inpack64_pre(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %zmm7,
%zmm8, %zmm9, %zmm10, %zmm11, %zmm12, %zmm13, %zmm14,
%zmm15, %rdx, (key_table)(CTX, %r8, 8));
call __camellia_gfni_avx512_dec_blk64;
/* XOR output with IV */
vmovdqu64 (%r9), %xmm16;
vinserti64x2 $1, (0 * 16)(%rdx), %ymm16, %ymm16;
vinserti64x4 $1, (1 * 16)(%rdx), %zmm16, %zmm16;
vpxorq %zmm16, %zmm7, %zmm7;
vpxorq (0 * 64 + 48)(%rdx), %zmm6, %zmm6;
vpxorq (1 * 64 + 48)(%rdx), %zmm5, %zmm5;
vpxorq (2 * 64 + 48)(%rdx), %zmm4, %zmm4;
vpxorq (3 * 64 + 48)(%rdx), %zmm3, %zmm3;
vpxorq (4 * 64 + 48)(%rdx), %zmm2, %zmm2;
vpxorq (5 * 64 + 48)(%rdx), %zmm1, %zmm1;
vpxorq (6 * 64 + 48)(%rdx), %zmm0, %zmm0;
vpxorq (7 * 64 + 48)(%rdx), %zmm15, %zmm15;
vpxorq (8 * 64 + 48)(%rdx), %zmm14, %zmm14;
vpxorq (9 * 64 + 48)(%rdx), %zmm13, %zmm13;
vpxorq (10 * 64 + 48)(%rdx), %zmm12, %zmm12;
vpxorq (11 * 64 + 48)(%rdx), %zmm11, %zmm11;
vpxorq (12 * 64 + 48)(%rdx), %zmm10, %zmm10;
vpxorq (13 * 64 + 48)(%rdx), %zmm9, %zmm9;
vpxorq (14 * 64 + 48)(%rdx), %zmm8, %zmm8;
vmovdqu64 (15 * 64 + 48)(%rdx), %xmm16;
write_output(%zmm7, %zmm6, %zmm5, %zmm4, %zmm3, %zmm2, %zmm1, %zmm0,
%zmm15, %zmm14, %zmm13, %zmm12, %zmm11, %zmm10, %zmm9,
%zmm8, %rsi);
/* store new IV */
vmovdqu64 %xmm16, (0)(%r9);
clear_regs();
ret_spec_stop;
CFI_ENDPROC();
ELF(.size _gcry_camellia_gfni_avx512_cbc_dec,.-_gcry_camellia_gfni_avx512_cbc_dec;)
.align 16
.globl _gcry_camellia_gfni_avx512_cfb_dec
ELF(.type _gcry_camellia_gfni_avx512_cfb_dec,@function;)
_gcry_camellia_gfni_avx512_cfb_dec:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (32 blocks)
* %rdx: src (32 blocks)
* %rcx: iv
*/
CFI_STARTPROC();
spec_stop_avx512;
cmpl $128, key_bitlength(CTX);
movl $32, %r8d;
movl $24, %eax;
cmovel %eax, %r8d; /* max */
/* inpack64_pre: */
vpbroadcastq (key_table)(CTX), %zmm0;
vpshufb .Lpack_bswap rRIP, %zmm0, %zmm0;
vmovdqu64 (%rcx), %xmm15;
vinserti64x2 $1, (%rdx), %ymm15, %ymm15;
vinserti64x4 $1, 16(%rdx), %zmm15, %zmm15;
vpxorq %zmm15, %zmm0, %zmm15;
vpxorq (0 * 64 + 48)(%rdx), %zmm0, %zmm14;
vpxorq (1 * 64 + 48)(%rdx), %zmm0, %zmm13;
vpxorq (2 * 64 + 48)(%rdx), %zmm0, %zmm12;
vpxorq (3 * 64 + 48)(%rdx), %zmm0, %zmm11;
vpxorq (4 * 64 + 48)(%rdx), %zmm0, %zmm10;
vpxorq (5 * 64 + 48)(%rdx), %zmm0, %zmm9;
vpxorq (6 * 64 + 48)(%rdx), %zmm0, %zmm8;
vpxorq (7 * 64 + 48)(%rdx), %zmm0, %zmm7;
vpxorq (8 * 64 + 48)(%rdx), %zmm0, %zmm6;
vpxorq (9 * 64 + 48)(%rdx), %zmm0, %zmm5;
vpxorq (10 * 64 + 48)(%rdx), %zmm0, %zmm4;
vpxorq (11 * 64 + 48)(%rdx), %zmm0, %zmm3;
vpxorq (12 * 64 + 48)(%rdx), %zmm0, %zmm2;
vpxorq (13 * 64 + 48)(%rdx), %zmm0, %zmm1;
vpxorq (14 * 64 + 48)(%rdx), %zmm0, %zmm0;
vmovdqu64 (15 * 64 + 48)(%rdx), %xmm16;
vmovdqu64 %xmm16, (%rcx); /* store new IV */
call __camellia_gfni_avx512_enc_blk64;
vpxorq 0 * 64(%rdx), %zmm7, %zmm7;
vpxorq 1 * 64(%rdx), %zmm6, %zmm6;
vpxorq 2 * 64(%rdx), %zmm5, %zmm5;
vpxorq 3 * 64(%rdx), %zmm4, %zmm4;
vpxorq 4 * 64(%rdx), %zmm3, %zmm3;
vpxorq 5 * 64(%rdx), %zmm2, %zmm2;
vpxorq 6 * 64(%rdx), %zmm1, %zmm1;
vpxorq 7 * 64(%rdx), %zmm0, %zmm0;
vpxorq 8 * 64(%rdx), %zmm15, %zmm15;
vpxorq 9 * 64(%rdx), %zmm14, %zmm14;
vpxorq 10 * 64(%rdx), %zmm13, %zmm13;
vpxorq 11 * 64(%rdx), %zmm12, %zmm12;
vpxorq 12 * 64(%rdx), %zmm11, %zmm11;
vpxorq 13 * 64(%rdx), %zmm10, %zmm10;
vpxorq 14 * 64(%rdx), %zmm9, %zmm9;
vpxorq 15 * 64(%rdx), %zmm8, %zmm8;
write_output(%zmm7, %zmm6, %zmm5, %zmm4, %zmm3, %zmm2, %zmm1, %zmm0,
%zmm15, %zmm14, %zmm13, %zmm12, %zmm11, %zmm10, %zmm9,
%zmm8, %rsi);
clear_regs();
ret_spec_stop;
CFI_ENDPROC();
ELF(.size _gcry_camellia_gfni_avx512_cfb_dec,.-_gcry_camellia_gfni_avx512_cfb_dec;)
.align 16
.globl _gcry_camellia_gfni_avx512_ocb_enc
ELF(.type _gcry_camellia_gfni_avx512_ocb_enc,@function;)
_gcry_camellia_gfni_avx512_ocb_enc:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (64 blocks)
* %rdx: src (64 blocks)
* %rcx: offset
* %r8 : checksum
* %r9 : L pointers (void *L[64])
*/
CFI_STARTPROC();
spec_stop_avx512;
pushq %r12;
CFI_PUSH(%r12);
pushq %r13;
CFI_PUSH(%r13);
pushq %r14;
CFI_PUSH(%r14);
pushq %r15;
CFI_PUSH(%r15);
pushq %rbx;
CFI_PUSH(%rbx);
vmovdqu64 (%rcx), %xmm30;
/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
/* Checksum_i = Checksum_{i-1} xor P_i */
/* C_i = Offset_i xor ENCIPHER(K, P_i xor Offset_i) */
#define OCB_INPUT(n, l0reg, l1reg, l2reg, l3reg, zreg, zplain) \
vmovdqu64 (n * 64)(%rdx), zplain; \
vpxorq (l0reg), %xmm30, %xmm16; \
vpxorq (l1reg), %xmm16, %xmm30; \
vinserti64x2 $1, %xmm30, %ymm16, %ymm16; \
vpxorq (l2reg), %xmm30, %xmm30; \
vinserti64x2 $2, %xmm30, %zmm16, %zmm16; \
vpxorq (l3reg), %xmm30, %xmm30; \
vinserti64x2 $3, %xmm30, %zmm16, %zmm16; \
vpxorq zplain, %zmm16, zreg; \
vmovdqu64 %zmm16, (n * 64)(%rsi);
#define OCB_LOAD_PTRS(n) \
movq ((n * 4 * 8) + (0 * 8))(%r9), %r10; \
movq ((n * 4 * 8) + (1 * 8))(%r9), %r11; \
movq ((n * 4 * 8) + (2 * 8))(%r9), %r12; \
movq ((n * 4 * 8) + (3 * 8))(%r9), %r13; \
movq ((n * 4 * 8) + (4 * 8))(%r9), %r14; \
movq ((n * 4 * 8) + (5 * 8))(%r9), %r15; \
movq ((n * 4 * 8) + (6 * 8))(%r9), %rax; \
movq ((n * 4 * 8) + (7 * 8))(%r9), %rbx;
OCB_LOAD_PTRS(0);
OCB_INPUT(0, %r10, %r11, %r12, %r13, %zmm15, %zmm20);
OCB_INPUT(1, %r14, %r15, %rax, %rbx, %zmm14, %zmm21);
OCB_LOAD_PTRS(2);
OCB_INPUT(2, %r10, %r11, %r12, %r13, %zmm13, %zmm22);
vpternlogq $0x96, %zmm20, %zmm21, %zmm22;
OCB_INPUT(3, %r14, %r15, %rax, %rbx, %zmm12, %zmm23);
OCB_LOAD_PTRS(4);
OCB_INPUT(4, %r10, %r11, %r12, %r13, %zmm11, %zmm24);
OCB_INPUT(5, %r14, %r15, %rax, %rbx, %zmm10, %zmm25);
vpternlogq $0x96, %zmm23, %zmm24, %zmm25;
OCB_LOAD_PTRS(6);
OCB_INPUT(6, %r10, %r11, %r12, %r13, %zmm9, %zmm20);
OCB_INPUT(7, %r14, %r15, %rax, %rbx, %zmm8, %zmm21);
OCB_LOAD_PTRS(8);
OCB_INPUT(8, %r10, %r11, %r12, %r13, %zmm7, %zmm26);
vpternlogq $0x96, %zmm20, %zmm21, %zmm26;
OCB_INPUT(9, %r14, %r15, %rax, %rbx, %zmm6, %zmm23);
OCB_LOAD_PTRS(10);
OCB_INPUT(10, %r10, %r11, %r12, %r13, %zmm5, %zmm24);
OCB_INPUT(11, %r14, %r15, %rax, %rbx, %zmm4, %zmm27);
vpternlogq $0x96, %zmm23, %zmm24, %zmm27;
OCB_LOAD_PTRS(12);
OCB_INPUT(12, %r10, %r11, %r12, %r13, %zmm3, %zmm20);
OCB_INPUT(13, %r14, %r15, %rax, %rbx, %zmm2, %zmm21);
OCB_LOAD_PTRS(14);
OCB_INPUT(14, %r10, %r11, %r12, %r13, %zmm1, %zmm23);
vpternlogq $0x96, %zmm20, %zmm21, %zmm23;
OCB_INPUT(15, %r14, %r15, %rax, %rbx, %zmm0, %zmm24);
#undef OCB_LOAD_PTRS
#undef OCB_INPUT
vpbroadcastq (key_table)(CTX), %zmm16;
vpshufb .Lpack_bswap rRIP, %zmm16, %zmm16;
vpternlogq $0x96, %zmm24, %zmm22, %zmm25;
vpternlogq $0x96, %zmm26, %zmm27, %zmm23;
vpxorq %zmm25, %zmm23, %zmm20;
vextracti64x4 $1, %zmm20, %ymm21;
vpxorq %ymm21, %ymm20, %ymm20;
vextracti64x2 $1, %ymm20, %xmm21;
vpternlogq $0x96, (%r8), %xmm21, %xmm20;
vmovdqu64 %xmm30, (%rcx);
vmovdqu64 %xmm20, (%r8);
cmpl $128, key_bitlength(CTX);
movl $32, %r8d;
movl $24, %eax;
cmovel %eax, %r8d; /* max */
/* inpack64_pre: */
vpxorq %zmm0, %zmm16, %zmm0;
vpxorq %zmm1, %zmm16, %zmm1;
vpxorq %zmm2, %zmm16, %zmm2;
vpxorq %zmm3, %zmm16, %zmm3;
vpxorq %zmm4, %zmm16, %zmm4;
vpxorq %zmm5, %zmm16, %zmm5;
vpxorq %zmm6, %zmm16, %zmm6;
vpxorq %zmm7, %zmm16, %zmm7;
vpxorq %zmm8, %zmm16, %zmm8;
vpxorq %zmm9, %zmm16, %zmm9;
vpxorq %zmm10, %zmm16, %zmm10;
vpxorq %zmm11, %zmm16, %zmm11;
vpxorq %zmm12, %zmm16, %zmm12;
vpxorq %zmm13, %zmm16, %zmm13;
vpxorq %zmm14, %zmm16, %zmm14;
vpxorq %zmm15, %zmm16, %zmm15;
call __camellia_gfni_avx512_enc_blk64;
vpxorq 0 * 64(%rsi), %zmm7, %zmm7;
vpxorq 1 * 64(%rsi), %zmm6, %zmm6;
vpxorq 2 * 64(%rsi), %zmm5, %zmm5;
vpxorq 3 * 64(%rsi), %zmm4, %zmm4;
vpxorq 4 * 64(%rsi), %zmm3, %zmm3;
vpxorq 5 * 64(%rsi), %zmm2, %zmm2;
vpxorq 6 * 64(%rsi), %zmm1, %zmm1;
vpxorq 7 * 64(%rsi), %zmm0, %zmm0;
vpxorq 8 * 64(%rsi), %zmm15, %zmm15;
vpxorq 9 * 64(%rsi), %zmm14, %zmm14;
vpxorq 10 * 64(%rsi), %zmm13, %zmm13;
vpxorq 11 * 64(%rsi), %zmm12, %zmm12;
vpxorq 12 * 64(%rsi), %zmm11, %zmm11;
vpxorq 13 * 64(%rsi), %zmm10, %zmm10;
vpxorq 14 * 64(%rsi), %zmm9, %zmm9;
vpxorq 15 * 64(%rsi), %zmm8, %zmm8;
write_output(%zmm7, %zmm6, %zmm5, %zmm4, %zmm3, %zmm2, %zmm1, %zmm0,
%zmm15, %zmm14, %zmm13, %zmm12, %zmm11, %zmm10, %zmm9,
%zmm8, %rsi);
clear_regs();
popq %rbx;
CFI_RESTORE(%rbx);
popq %r15;
CFI_RESTORE(%r15);
popq %r14;
CFI_RESTORE(%r14);
popq %r13;
CFI_RESTORE(%r12);
popq %r12;
CFI_RESTORE(%r13);
ret_spec_stop;
CFI_ENDPROC();
ELF(.size _gcry_camellia_gfni_avx512_ocb_enc,.-_gcry_camellia_gfni_avx512_ocb_enc;)
.align 16
.globl _gcry_camellia_gfni_avx512_ocb_dec
ELF(.type _gcry_camellia_gfni_avx512_ocb_dec,@function;)
_gcry_camellia_gfni_avx512_ocb_dec:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (64 blocks)
* %rdx: src (64 blocks)
* %rcx: offset
* %r8 : checksum
* %r9 : L pointers (void *L[64])
*/
CFI_STARTPROC();
spec_stop_avx512;
pushq %r12;
CFI_PUSH(%r12);
pushq %r13;
CFI_PUSH(%r13);
pushq %r14;
CFI_PUSH(%r14);
pushq %r15;
CFI_PUSH(%r15);
pushq %rbx;
CFI_PUSH(%rbx);
pushq %r8;
CFI_PUSH(%r8);
vmovdqu64 (%rcx), %xmm30;
/* Offset_i = Offset_{i-1} xor L_{ntz(i)} */
/* C_i = Offset_i xor DECIPHER(K, P_i xor Offset_i) */
#define OCB_INPUT(n, l0reg, l1reg, l2reg, l3reg, zreg) \
vpxorq (l0reg), %xmm30, %xmm16; \
vpxorq (l1reg), %xmm16, %xmm30; \
vinserti64x2 $1, %xmm30, %ymm16, %ymm16; \
vpxorq (l2reg), %xmm30, %xmm30; \
vinserti64x2 $2, %xmm30, %zmm16, %zmm16; \
vpxorq (l3reg), %xmm30, %xmm30; \
vinserti64x2 $3, %xmm30, %zmm16, %zmm16; \
vpxorq (n * 64)(%rdx), %zmm16, zreg; \
vmovdqu64 %zmm16, (n * 64)(%rsi);
#define OCB_LOAD_PTRS(n) \
movq ((n * 4 * 8) + (0 * 8))(%r9), %r10; \
movq ((n * 4 * 8) + (1 * 8))(%r9), %r11; \
movq ((n * 4 * 8) + (2 * 8))(%r9), %r12; \
movq ((n * 4 * 8) + (3 * 8))(%r9), %r13; \
movq ((n * 4 * 8) + (4 * 8))(%r9), %r14; \
movq ((n * 4 * 8) + (5 * 8))(%r9), %r15; \
movq ((n * 4 * 8) + (6 * 8))(%r9), %rax; \
movq ((n * 4 * 8) + (7 * 8))(%r9), %rbx;
OCB_LOAD_PTRS(0);
OCB_INPUT(0, %r10, %r11, %r12, %r13, %zmm15);
OCB_INPUT(1, %r14, %r15, %rax, %rbx, %zmm14);
OCB_LOAD_PTRS(2);
OCB_INPUT(2, %r10, %r11, %r12, %r13, %zmm13);
OCB_INPUT(3, %r14, %r15, %rax, %rbx, %zmm12);
OCB_LOAD_PTRS(4);
OCB_INPUT(4, %r10, %r11, %r12, %r13, %zmm11);
OCB_INPUT(5, %r14, %r15, %rax, %rbx, %zmm10);
OCB_LOAD_PTRS(6);
OCB_INPUT(6, %r10, %r11, %r12, %r13, %zmm9);
OCB_INPUT(7, %r14, %r15, %rax, %rbx, %zmm8);
OCB_LOAD_PTRS(8);
OCB_INPUT(8, %r10, %r11, %r12, %r13, %zmm7);
OCB_INPUT(9, %r14, %r15, %rax, %rbx, %zmm6);
OCB_LOAD_PTRS(10);
OCB_INPUT(10, %r10, %r11, %r12, %r13, %zmm5);
OCB_INPUT(11, %r14, %r15, %rax, %rbx, %zmm4);
OCB_LOAD_PTRS(12);
OCB_INPUT(12, %r10, %r11, %r12, %r13, %zmm3);
OCB_INPUT(13, %r14, %r15, %rax, %rbx, %zmm2);
OCB_LOAD_PTRS(14);
OCB_INPUT(14, %r10, %r11, %r12, %r13, %zmm1);
OCB_INPUT(15, %r14, %r15, %rax, %rbx, %zmm0);
#undef OCB_LOAD_PTRS
#undef OCB_INPUT
vmovdqu64 %xmm30, (%rcx);
cmpl $128, key_bitlength(CTX);
movl $32, %r8d;
movl $24, %eax;
cmovel %eax, %r8d; /* max */
vpbroadcastq (key_table)(CTX, %r8, 8), %zmm16;
vpshufb .Lpack_bswap rRIP, %zmm16, %zmm16;
/* inpack64_pre: */
vpxorq %zmm0, %zmm16, %zmm0;
vpxorq %zmm1, %zmm16, %zmm1;
vpxorq %zmm2, %zmm16, %zmm2;
vpxorq %zmm3, %zmm16, %zmm3;
vpxorq %zmm4, %zmm16, %zmm4;
vpxorq %zmm5, %zmm16, %zmm5;
vpxorq %zmm6, %zmm16, %zmm6;
vpxorq %zmm7, %zmm16, %zmm7;
vpxorq %zmm8, %zmm16, %zmm8;
vpxorq %zmm9, %zmm16, %zmm9;
vpxorq %zmm10, %zmm16, %zmm10;
vpxorq %zmm11, %zmm16, %zmm11;
vpxorq %zmm12, %zmm16, %zmm12;
vpxorq %zmm13, %zmm16, %zmm13;
vpxorq %zmm14, %zmm16, %zmm14;
vpxorq %zmm15, %zmm16, %zmm15;
call __camellia_gfni_avx512_dec_blk64;
vpxorq 0 * 64(%rsi), %zmm7, %zmm7;
vpxorq 1 * 64(%rsi), %zmm6, %zmm6;
vpxorq 2 * 64(%rsi), %zmm5, %zmm5;
vpxorq 3 * 64(%rsi), %zmm4, %zmm4;
vpxorq 4 * 64(%rsi), %zmm3, %zmm3;
vpxorq 5 * 64(%rsi), %zmm2, %zmm2;
vpxorq 6 * 64(%rsi), %zmm1, %zmm1;
vpxorq 7 * 64(%rsi), %zmm0, %zmm0;
vpxorq 8 * 64(%rsi), %zmm15, %zmm15;
vpxorq 9 * 64(%rsi), %zmm14, %zmm14;
vpxorq 10 * 64(%rsi), %zmm13, %zmm13;
vpxorq 11 * 64(%rsi), %zmm12, %zmm12;
vpxorq 12 * 64(%rsi), %zmm11, %zmm11;
vpxorq 13 * 64(%rsi), %zmm10, %zmm10;
vpxorq 14 * 64(%rsi), %zmm9, %zmm9;
vpxorq 15 * 64(%rsi), %zmm8, %zmm8;
write_output(%zmm7, %zmm6, %zmm5, %zmm4, %zmm3, %zmm2, %zmm1, %zmm0,
%zmm15, %zmm14, %zmm13, %zmm12, %zmm11, %zmm10, %zmm9,
%zmm8, %rsi);
popq %r8;
CFI_RESTORE(%r8);
/* Checksum_i = Checksum_{i-1} xor C_i */
vpternlogq $0x96, %zmm7, %zmm6, %zmm5;
vpternlogq $0x96, %zmm4, %zmm3, %zmm2;
vpternlogq $0x96, %zmm1, %zmm0, %zmm15;
vpternlogq $0x96, %zmm14, %zmm13, %zmm12;
vpternlogq $0x96, %zmm11, %zmm10, %zmm9;
vpternlogq $0x96, %zmm5, %zmm2, %zmm15;
vpternlogq $0x96, %zmm12, %zmm9, %zmm8;
vpxorq %zmm15, %zmm8, %zmm8;
vextracti64x4 $1, %zmm8, %ymm0;
vpxor %ymm0, %ymm8, %ymm8;
vextracti128 $1, %ymm8, %xmm0;
vpternlogq $0x96, (%r8), %xmm0, %xmm8;
vmovdqu64 %xmm8, (%r8);
clear_regs();
popq %rbx;
CFI_RESTORE(%rbx);
popq %r15;
CFI_RESTORE(%r15);
popq %r14;
CFI_RESTORE(%r14);
popq %r13;
CFI_RESTORE(%r12);
popq %r12;
CFI_RESTORE(%r13);
ret_spec_stop;
CFI_ENDPROC();
ELF(.size _gcry_camellia_gfni_avx512_ocb_dec,.-_gcry_camellia_gfni_avx512_ocb_dec;)
.align 16
.globl _gcry_camellia_gfni_avx512_enc_blk64
ELF(.type _gcry_camellia_gfni_avx512_enc_blk64,@function;)
_gcry_camellia_gfni_avx512_enc_blk64:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (64 blocks)
* %rdx: src (64 blocks)
*/
CFI_STARTPROC();
spec_stop_avx512;
cmpl $128, key_bitlength(CTX);
movl $32, %r8d;
movl $24, %eax;
cmovel %eax, %r8d; /* max */
xorl %eax, %eax;
vpbroadcastq (key_table)(CTX), %zmm0;
vpshufb .Lpack_bswap rRIP, %zmm0, %zmm0;
vpxorq (0) * 64(%rdx), %zmm0, %zmm15;
vpxorq (1) * 64(%rdx), %zmm0, %zmm14;
vpxorq (2) * 64(%rdx), %zmm0, %zmm13;
vpxorq (3) * 64(%rdx), %zmm0, %zmm12;
vpxorq (4) * 64(%rdx), %zmm0, %zmm11;
vpxorq (5) * 64(%rdx), %zmm0, %zmm10;
vpxorq (6) * 64(%rdx), %zmm0, %zmm9;
vpxorq (7) * 64(%rdx), %zmm0, %zmm8;
vpxorq (8) * 64(%rdx), %zmm0, %zmm7;
vpxorq (9) * 64(%rdx), %zmm0, %zmm6;
vpxorq (10) * 64(%rdx), %zmm0, %zmm5;
vpxorq (11) * 64(%rdx), %zmm0, %zmm4;
vpxorq (12) * 64(%rdx), %zmm0, %zmm3;
vpxorq (13) * 64(%rdx), %zmm0, %zmm2;
vpxorq (14) * 64(%rdx), %zmm0, %zmm1;
vpxorq (15) * 64(%rdx), %zmm0, %zmm0;
call __camellia_gfni_avx512_enc_blk64;
vmovdqu64 %zmm7, (0) * 64(%rsi);
vmovdqu64 %zmm6, (1) * 64(%rsi);
vmovdqu64 %zmm5, (2) * 64(%rsi);
vmovdqu64 %zmm4, (3) * 64(%rsi);
vmovdqu64 %zmm3, (4) * 64(%rsi);
vmovdqu64 %zmm2, (5) * 64(%rsi);
vmovdqu64 %zmm1, (6) * 64(%rsi);
vmovdqu64 %zmm0, (7) * 64(%rsi);
vmovdqu64 %zmm15, (8) * 64(%rsi);
vmovdqu64 %zmm14, (9) * 64(%rsi);
vmovdqu64 %zmm13, (10) * 64(%rsi);
vmovdqu64 %zmm12, (11) * 64(%rsi);
vmovdqu64 %zmm11, (12) * 64(%rsi);
vmovdqu64 %zmm10, (13) * 64(%rsi);
vmovdqu64 %zmm9, (14) * 64(%rsi);
vmovdqu64 %zmm8, (15) * 64(%rsi);
clear_regs();
ret_spec_stop;
CFI_ENDPROC();
ELF(.size _gcry_camellia_gfni_avx512_enc_blk64,.-_gcry_camellia_gfni_avx512_enc_blk64;)
.align 16
.globl _gcry_camellia_gfni_avx512_dec_blk64
ELF(.type _gcry_camellia_gfni_avx512_dec_blk64,@function;)
_gcry_camellia_gfni_avx512_dec_blk64:
/* input:
* %rdi: ctx, CTX
* %rsi: dst (64 blocks)
* %rdx: src (64 blocks)
*/
CFI_STARTPROC();
spec_stop_avx512;
cmpl $128, key_bitlength(CTX);
movl $32, %r8d;
movl $24, %eax;
cmovel %eax, %r8d; /* max */
xorl %eax, %eax;
vpbroadcastq (key_table)(CTX, %r8, 8), %zmm0;
vpshufb .Lpack_bswap rRIP, %zmm0, %zmm0;
vpxorq (0) * 64(%rdx), %zmm0, %zmm15;
vpxorq (1) * 64(%rdx), %zmm0, %zmm14;
vpxorq (2) * 64(%rdx), %zmm0, %zmm13;
vpxorq (3) * 64(%rdx), %zmm0, %zmm12;
vpxorq (4) * 64(%rdx), %zmm0, %zmm11;
vpxorq (5) * 64(%rdx), %zmm0, %zmm10;
vpxorq (6) * 64(%rdx), %zmm0, %zmm9;
vpxorq (7) * 64(%rdx), %zmm0, %zmm8;
vpxorq (8) * 64(%rdx), %zmm0, %zmm7;
vpxorq (9) * 64(%rdx), %zmm0, %zmm6;
vpxorq (10) * 64(%rdx), %zmm0, %zmm5;
vpxorq (11) * 64(%rdx), %zmm0, %zmm4;
vpxorq (12) * 64(%rdx), %zmm0, %zmm3;
vpxorq (13) * 64(%rdx), %zmm0, %zmm2;
vpxorq (14) * 64(%rdx), %zmm0, %zmm1;
vpxorq (15) * 64(%rdx), %zmm0, %zmm0;
call __camellia_gfni_avx512_dec_blk64;
vmovdqu64 %zmm7, (0) * 64(%rsi);
vmovdqu64 %zmm6, (1) * 64(%rsi);
vmovdqu64 %zmm5, (2) * 64(%rsi);
vmovdqu64 %zmm4, (3) * 64(%rsi);
vmovdqu64 %zmm3, (4) * 64(%rsi);
vmovdqu64 %zmm2, (5) * 64(%rsi);
vmovdqu64 %zmm1, (6) * 64(%rsi);
vmovdqu64 %zmm0, (7) * 64(%rsi);
vmovdqu64 %zmm15, (8) * 64(%rsi);
vmovdqu64 %zmm14, (9) * 64(%rsi);
vmovdqu64 %zmm13, (10) * 64(%rsi);
vmovdqu64 %zmm12, (11) * 64(%rsi);
vmovdqu64 %zmm11, (12) * 64(%rsi);
vmovdqu64 %zmm10, (13) * 64(%rsi);
vmovdqu64 %zmm9, (14) * 64(%rsi);
vmovdqu64 %zmm8, (15) * 64(%rsi);
clear_regs();
ret_spec_stop;
CFI_ENDPROC();
ELF(.size _gcry_camellia_gfni_avx512_dec_blk64,.-_gcry_camellia_gfni_avx512_dec_blk64;)
+/**********************************************************************
+ 1-block non-parallel camellia
+ **********************************************************************/
+
+/* Camellia F function, AVX512+GFNI version */
+#define camellia_f_gfni(ab, cd, x1, x2, x3, x4, \
+ pre_filter_bitmatrix_s123, pre_filter_bitmatrix_s4, \
+ post_filter_bitmatrix_s14, post_filter_bitmatrix_s2, \
+ post_filter_bitmatrix_s3, sp1mask, sp2mask, sp3mask, \
+ sp4mask) \
+ /* camellia sboxes s1, s2, s3, s4 */ \
+ vgf2p8affineqb $(pre_filter_constant_s1234), \
+ pre_filter_bitmatrix_s4, ab, x4; \
+ vgf2p8affineinvqb $(post_filter_constant_s14), \
+ post_filter_bitmatrix_s14, x4, x4; \
+ vgf2p8affineqb $(pre_filter_constant_s1234), \
+ pre_filter_bitmatrix_s123, ab, x1; \
+ vgf2p8affineinvqb $(post_filter_constant_s2), \
+ post_filter_bitmatrix_s2, x1, x2; \
+ vgf2p8affineinvqb $(post_filter_constant_s3), \
+ post_filter_bitmatrix_s3, x1, x3; \
+ vgf2p8affineinvqb $(post_filter_constant_s14), \
+ post_filter_bitmatrix_s14, x1, x1; \
+ \
+ /* permutation */ \
+ vpshufb sp4mask, x4, x4; \
+ vpshufb sp2mask, x2, x2; \
+ vpshufb sp3mask, x3, x3; \
+ vpshufb sp1mask, x1, x1; \
+ vpxor x4, x2, x2; \
+ vpternlogd $0x96, x3, x1, x2; \
+ vpsrldq $8, x2, x1; \
+ \
+ /* output xor */ \
+ vpternlogd $0x96, x2, x1, cd;
+
+#define preload_camellia_f_consts() \
+ vmovq .Lpre_filter_bitmatrix_s123 rRIP, %xmm14; \
+ vmovq .Lpre_filter_bitmatrix_s4 rRIP, %xmm13; \
+ vmovq .Lpost_filter_bitmatrix_s14 rRIP, %xmm12; \
+ vmovq .Lpost_filter_bitmatrix_s2 rRIP, %xmm11; \
+ vmovq .Lpost_filter_bitmatrix_s3 rRIP, %xmm10; \
+ vmovdqa .Lsp1mask_swap32_gfni rRIP, %xmm9; \
+ vmovdqa .Lsp2mask_swap32_gfni rRIP, %xmm8; \
+ vmovdqa .Lsp3mask_swap32_gfni rRIP, %xmm7; \
+ vmovdqa .Lsp4mask_swap32_gfni rRIP, %xmm6;
+
+#define do_camellia_f(ab, cd, t0, t1, t2, t3) \
+ camellia_f_gfni(ab, cd, t0, t1, t2, t3, \
+ %xmm14, %xmm13, %xmm12, %xmm11, %xmm10, \
+ %xmm9, %xmm8, %xmm7, %xmm6);
+
+#define preload_camellia_key_consts() \
+ kxnorb %k1, %k1, %k1; \
+ vmovdqa .Lpack_bswap rRIP, %xmm15; \
+ kshiftrb $7, %k1, %k1;
+
+#define add_roundkey_blk1(cd, key) \
+ vpxorq key, cd, cd{%k1}{z};
+
+#define do_fls_blk1(ll, lr, rl, rr, t0, t1, kll, klr, krl, krr) \
+ vpternlogd $0x1E, krr, rr, rl{%k1}{z}; \
+ vpandd kll, ll, t0{%k1}{z}; \
+ vpandd krl, rl, t1{%k1}{z}; \
+ vprold $1, t0, t0; \
+ vprold $1, t1, t1; \
+ vpxor t0, lr, lr; \
+ vpxor t1, rr, rr; \
+ vpternlogd $0x1E, klr, lr, ll{%k1}{z};
+
+#define clear_regs_blk1() \
+ kxorq %k1, %k1, %k1; \
+ vzeroall;
+
+#define roundsm_blk1(ab, cd, t0, t1, t2, t3, key) \
+ add_roundkey_blk1(cd, key); \
+ do_camellia_f(ab, cd, t0, t1, t2, t3);
+
+#define two_roundsm_blk1(ab, cd, t0, t1, t2, t3, i, dir, ctx) \
+ roundsm_blk1(ab, cd, t0, t1, t2, t3, \
+ (key_table + (i) * 8)(ctx)); \
+ roundsm_blk1(cd, ab, t0, t1, t2, t3, \
+ (key_table + ((i) + (dir)) * 8)(ctx));
+
+#define enc_rounds_blk1(ab, cd, t0, t1, t2, t3, i, ctx) \
+ two_roundsm_blk1(ab, cd, t0, t1, t2, t3, (i) + 2, 1, ctx); \
+ two_roundsm_blk1(ab, cd, t0, t1, t2, t3, (i) + 4, 1, ctx); \
+ two_roundsm_blk1(ab, cd, t0, t1, t2, t3, (i) + 6, 1, ctx);
+
+#define dec_rounds_blk1(ab, cd, t0, t1, t2, t3, i, ctx) \
+ two_roundsm_blk1(ab, cd, t0, t1, t2, t3, (i) + 7, -1, ctx); \
+ two_roundsm_blk1(ab, cd, t0, t1, t2, t3, (i) + 5, -1, ctx); \
+ two_roundsm_blk1(ab, cd, t0, t1, t2, t3, (i) + 3, -1, ctx);
+
+#define fls_blk1(ll_lr, rl_rr, tmp_lr, tmp_rr, t0, t1, kll, klr, krl, krr) \
+ vpsrlq $32, rl_rr, tmp_rr; \
+ vpsrlq $32, ll_lr, tmp_lr; \
+ do_fls_blk1(ll_lr, tmp_lr, rl_rr, tmp_rr, t0, t1, kll, klr, krl, krr); \
+ vpunpckldq tmp_lr, ll_lr, ll_lr; \
+ vpunpckldq tmp_rr, rl_rr, rl_rr;
+
+#define inpack_blk1(ab, cd, src, key) \
+ vmovq 0(src), ab; \
+ vmovq 8(src), cd; \
+ vpshufb %xmm15, ab, ab; \
+ vpshufb %xmm15, cd, cd; \
+ add_roundkey_blk1(ab, key);
+
+#define outunpack_blk1(ab, cd, dst, key) \
+ add_roundkey_blk1(cd, key); \
+ vpshufb %xmm15, ab, ab; \
+ vpshufb %xmm15, cd, cd; \
+ vmovq ab, 8(dst); \
+ vmovq cd, 0(dst);
+
+.align 16
+.globl _gcry_camellia_gfni_avx512_enc_blk1
+ELF(.type _gcry_camellia_gfni_avx512_enc_blk1,@function;)
+
+_gcry_camellia_gfni_avx512_enc_blk1:
+ /* input:
+ * %rdi: ctx, CTX
+ * %rsi: dst (16 bytes)
+ * %rdx: src (16 bytes)
+ */
+ CFI_STARTPROC();
+ spec_stop_avx512;
+
+ preload_camellia_key_consts();
+
+ cmpl $128, key_bitlength(CTX);
+ movl $32, %r8d;
+ movl $24, %eax;
+ cmovel %eax, %r8d; /* max */
+
+ inpack_blk1(%xmm0, %xmm1, %rdx, (key_table)(CTX));
+
+ preload_camellia_f_consts();
+
+ leaq (-8 * 8)(CTX, %r8, 8), %r8;
+
+.align 16
+.Lenc_loop_blk1:
+ enc_rounds_blk1(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, 0, CTX);
+
+ cmpq %r8, CTX;
+ je .Lenc_done_blk1;
+ leaq (8 * 8)(CTX), CTX;
+
+ fls_blk1(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5,
+ ((key_table) + 0)(CTX),
+ ((key_table) + 4)(CTX),
+ ((key_table) + 8)(CTX),
+ ((key_table) + 12)(CTX));
+
+ jmp .Lenc_loop_blk1;
+
+.align 16
+.Lenc_done_blk1:
+ outunpack_blk1(%xmm0, %xmm1, %rsi, ((key_table) + 8 * 8)(CTX));
+
+ clear_regs_blk1();
+
+ ret_spec_stop;
+ CFI_ENDPROC();
+ELF(.size _gcry_camellia_gfni_avx512_enc_blk1,.-_gcry_camellia_gfni_avx512_enc_blk1;)
+
+.align 16
+.globl _gcry_camellia_gfni_avx512_dec_blk1
+ELF(.type _gcry_camellia_gfni_avx512_dec_blk1,@function;)
+
+_gcry_camellia_gfni_avx512_dec_blk1:
+ /* input:
+ * %rdi: ctx, CTX
+ * %rsi: dst (16 bytes)
+ * %rdx: src (16 bytes)
+ */
+ CFI_STARTPROC();
+ spec_stop_avx512;
+
+ preload_camellia_key_consts();
+
+ cmpl $128, key_bitlength(CTX);
+ movl $32, %r8d;
+ movl $24, %eax;
+ cmovel %eax, %r8d; /* max */
+
+ inpack_blk1(%xmm0, %xmm1, %rdx, (key_table)(CTX, %r8, 8));
+
+ preload_camellia_f_consts();
+
+ leaq (-8 * 8)(CTX, %r8, 8), %rax;
+
+.align 16
+.Ldec_loop_blk1:
+ dec_rounds_blk1(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, 0, %rax);
+
+ cmpq CTX, %rax;
+ je .Ldec_done_blk1;
+
+ fls_blk1(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5,
+ ((key_table) + 8)(%rax),
+ ((key_table) + 12)(%rax),
+ ((key_table) + 0)(%rax),
+ ((key_table) + 4)(%rax));
+
+ leaq (-8 * 8)(%rax), %rax;
+ jmp .Ldec_loop_blk1;
+
+.align 16
+.Ldec_done_blk1:
+ outunpack_blk1(%xmm0, %xmm1, %rsi, (key_table)(CTX));
+
+ clear_regs_blk1();
+
+ ret_spec_stop;
+ CFI_ENDPROC();
+ELF(.size _gcry_camellia_gfni_avx512_dec_blk1,.-_gcry_camellia_gfni_avx512_dec_blk1;)
+
#endif /* defined(ENABLE_GFNI_SUPPORT) && defined(ENABLE_AVX512_SUPPORT) */
#endif /* __x86_64 */
diff --git a/cipher/camellia-glue.c b/cipher/camellia-glue.c
index 5051a305..78ff22b9 100644
--- a/cipher/camellia-glue.c
+++ b/cipher/camellia-glue.c
@@ -1,1843 +1,1869 @@
/* camellia-glue.c - Glue for the Camellia cipher
* Copyright (C) 2007 Free Software Foundation, Inc.
*
* This file is part of Libgcrypt.
*
* Libgcrypt is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of
* the License, or (at your option) any later version.
*
* Libgcrypt is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this program; if not, see <https://www.gnu.org/licenses/>.
* SPDX-License-Identifier: LGPL-2.1-or-later
*/
/* I put all the libgcrypt-specific stuff in this file to keep the
camellia.c/camellia.h files exactly as provided by NTT. If they
update their code, this should make it easier to bring the changes
in. - dshaw
There is one small change which needs to be done: Include the
following code at the top of camellia.h: */
#if 0
/* To use Camellia with libraries it is often useful to keep the name
* space of the library clean. The following macro is thus useful:
*
* #define CAMELLIA_EXT_SYM_PREFIX foo_
*
* This prefixes all external symbols with "foo_".
*/
#ifdef HAVE_CONFIG_H
#include <config.h>
#endif
#ifdef CAMELLIA_EXT_SYM_PREFIX
#define CAMELLIA_PREFIX1(x,y) x ## y
#define CAMELLIA_PREFIX2(x,y) CAMELLIA_PREFIX1(x,y)
#define CAMELLIA_PREFIX(x) CAMELLIA_PREFIX2(CAMELLIA_EXT_SYM_PREFIX,x)
#define Camellia_Ekeygen CAMELLIA_PREFIX(Camellia_Ekeygen)
#define Camellia_EncryptBlock CAMELLIA_PREFIX(Camellia_EncryptBlock)
#define Camellia_DecryptBlock CAMELLIA_PREFIX(Camellia_DecryptBlock)
#define camellia_decrypt128 CAMELLIA_PREFIX(camellia_decrypt128)
#define camellia_decrypt256 CAMELLIA_PREFIX(camellia_decrypt256)
#define camellia_encrypt128 CAMELLIA_PREFIX(camellia_encrypt128)
#define camellia_encrypt256 CAMELLIA_PREFIX(camellia_encrypt256)
#define camellia_setup128 CAMELLIA_PREFIX(camellia_setup128)
#define camellia_setup192 CAMELLIA_PREFIX(camellia_setup192)
#define camellia_setup256 CAMELLIA_PREFIX(camellia_setup256)
#endif /*CAMELLIA_EXT_SYM_PREFIX*/
#endif /* Code sample. */
#include <config.h>
#include "types.h"
#include "g10lib.h"
#include "cipher.h"
#include "camellia.h"
#include "bufhelp.h"
#include "cipher-internal.h"
#include "bulkhelp.h"
/* Helper macro to force alignment to 16 bytes. */
#ifdef HAVE_GCC_ATTRIBUTE_ALIGNED
# define ATTR_ALIGNED_16 __attribute__ ((aligned (16)))
#else
# define ATTR_ALIGNED_16
#endif
/* USE_AESNI inidicates whether to compile with Intel AES-NI/AVX code. */
#undef USE_AESNI_AVX
#if defined(ENABLE_AESNI_SUPPORT) && defined(ENABLE_AVX_SUPPORT)
# if defined(__x86_64__) && (defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS))
# define USE_AESNI_AVX 1
# endif
#endif
/* USE_AESNI_AVX2 inidicates whether to compile with Intel AES-NI/AVX2 code. */
#undef USE_AESNI_AVX2
#if defined(ENABLE_AESNI_SUPPORT) && defined(ENABLE_AVX2_SUPPORT)
# if defined(__x86_64__) && (defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS))
# define USE_AESNI_AVX2 1
# endif
#endif
/* USE_VAES_AVX2 inidicates whether to compile with Intel VAES/AVX2 code. */
#undef USE_VAES_AVX2
#if defined(USE_AESNI_AVX2) && defined(HAVE_GCC_INLINE_ASM_VAES_VPCLMUL)
# define USE_VAES_AVX2 1
#endif
/* USE_GFNI_AVX2 inidicates whether to compile with Intel GFNI/AVX2 code. */
#undef USE_GFNI_AVX2
#if defined(USE_AESNI_AVX2) && defined(ENABLE_GFNI_SUPPORT)
# define USE_GFNI_AVX2 1
#endif
/* USE_GFNI_AVX512 inidicates whether to compile with Intel GFNI/AVX512 code. */
#undef USE_GFNI_AVX512
#if defined(USE_GFNI_AVX2) && defined(ENABLE_AVX512_SUPPORT)
# define USE_GFNI_AVX512 1
#endif
/* USE_PPC_CRYPTO indicates whether to enable PowerPC vector crypto
* accelerated code. */
#undef USE_PPC_CRYPTO
#if !defined(WORDS_BIGENDIAN) && defined(ENABLE_PPC_CRYPTO_SUPPORT) && \
defined(HAVE_COMPATIBLE_CC_PPC_ALTIVEC) && \
defined(HAVE_GCC_INLINE_ASM_PPC_ALTIVEC) && \
(SIZEOF_UNSIGNED_LONG == 8) && (__GNUC__ >= 4)
# define USE_PPC_CRYPTO 1
#endif
/* USE_AARCH64_CE indicates whether to enable ARMv8/CE accelerated code. */
#undef USE_AARCH64_CE
#if defined(__AARCH64EL__) && \
defined(HAVE_COMPATIBLE_GCC_AARCH64_PLATFORM_AS) && \
defined(HAVE_GCC_INLINE_ASM_AARCH64_CRYPTO) && \
defined(HAVE_COMPATIBLE_CC_AARCH64_NEON_INTRINSICS) && \
(__GNUC__ >= 4)
# define USE_AARCH64_CE 1
#endif
typedef struct
{
KEY_TABLE_TYPE keytable;
int keybitlength;
#ifdef USE_AESNI_AVX
unsigned int use_aesni_avx:1; /* AES-NI/AVX implementation shall be used. */
#endif /*USE_AESNI_AVX*/
#ifdef USE_AESNI_AVX2
unsigned int use_avx2:1; /* If any of AVX2 implementation is enabled. */
unsigned int use_aesni_avx2:1;/* AES-NI/AVX2 implementation shall be used. */
unsigned int use_vaes_avx2:1; /* VAES/AVX2 implementation shall be used. */
unsigned int use_gfni_avx2:1; /* GFNI/AVX2 implementation shall be used. */
unsigned int use_gfni_avx512:1; /* GFNI/AVX512 implementation shall be used. */
#endif /*USE_AESNI_AVX2*/
#ifdef USE_PPC_CRYPTO
unsigned int use_ppc:1;
unsigned int use_ppc8:1;
unsigned int use_ppc9:1;
#endif /*USE_PPC_CRYPTO*/
#ifdef USE_AARCH64_CE
unsigned int use_aarch64ce:1;
#endif /*USE_AARCH64_CE*/
} CAMELLIA_context;
/* Assembly implementations use SystemV ABI, ABI conversion and additional
* stack to store XMM6-XMM15 needed on Win64. */
#undef ASM_FUNC_ABI
#undef ASM_EXTRA_STACK
#if defined(USE_AESNI_AVX) || defined(USE_AESNI_AVX2)
# ifdef HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS
# define ASM_FUNC_ABI __attribute__((sysv_abi))
# define ASM_EXTRA_STACK (10 * 16)
# else
# define ASM_FUNC_ABI
# define ASM_EXTRA_STACK 0
# endif
#endif
#ifdef USE_AESNI_AVX
/* Assembler implementations of Camellia using AES-NI and AVX. Process data
in 16 blocks same time.
*/
extern void _gcry_camellia_aesni_avx_ctr_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *ctr) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_cbc_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_cfb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_ocb_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[16]) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_ocb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[16]) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_ocb_auth(CAMELLIA_context *ctx,
const unsigned char *abuf,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[16]) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_keygen(CAMELLIA_context *ctx,
const unsigned char *key,
unsigned int keylen) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_ecb_enc(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in)
ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx_ecb_dec(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in)
ASM_FUNC_ABI;
static const int avx_burn_stack_depth = 16 * CAMELLIA_BLOCK_SIZE + 16 +
2 * sizeof(void *) + ASM_EXTRA_STACK;
#endif
#ifdef USE_AESNI_AVX2
/* Assembler implementations of Camellia using AES-NI and AVX2. Process data
in 32 blocks same time.
*/
extern void _gcry_camellia_aesni_avx2_ctr_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *ctr) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx2_cbc_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx2_cfb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx2_ocb_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx2_ocb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx2_ocb_auth(CAMELLIA_context *ctx,
const unsigned char *abuf,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx2_enc_blk1_32(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned int nblocks)
ASM_FUNC_ABI;
extern void _gcry_camellia_aesni_avx2_dec_blk1_32(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned int nblocks)
ASM_FUNC_ABI;
static const int avx2_burn_stack_depth = 32 * CAMELLIA_BLOCK_SIZE + 16 +
2 * sizeof(void *) + ASM_EXTRA_STACK;
#endif
#ifdef USE_VAES_AVX2
/* Assembler implementations of Camellia using VAES and AVX2. Process data
in 32 blocks same time.
*/
extern void _gcry_camellia_vaes_avx2_ctr_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *ctr) ASM_FUNC_ABI;
extern void _gcry_camellia_vaes_avx2_cbc_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_vaes_avx2_cfb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_vaes_avx2_ocb_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_vaes_avx2_ocb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_vaes_avx2_ocb_auth(CAMELLIA_context *ctx,
const unsigned char *abuf,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_vaes_avx2_enc_blk1_32(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned int nblocks)
ASM_FUNC_ABI;
extern void _gcry_camellia_vaes_avx2_dec_blk1_32(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned int nblocks)
ASM_FUNC_ABI;
#endif
#ifdef USE_GFNI_AVX2
/* Assembler implementations of Camellia using GFNI and AVX2. Process data
in 32 blocks same time.
*/
extern void _gcry_camellia_gfni_avx2_ctr_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *ctr) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx2_cbc_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx2_cfb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx2_ocb_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx2_ocb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx2_ocb_auth(CAMELLIA_context *ctx,
const unsigned char *abuf,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx2_enc_blk1_32(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned int nblocks)
ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx2_dec_blk1_32(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned int nblocks)
ASM_FUNC_ABI;
#endif
#ifdef USE_GFNI_AVX512
/* Assembler implementations of Camellia using GFNI and AVX512. Process data
in 64 blocks same time.
*/
extern void _gcry_camellia_gfni_avx512_ctr_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *ctr) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx512_cbc_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx512_cfb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *iv) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx512_ocb_enc(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx512_ocb_dec(CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in,
unsigned char *offset,
unsigned char *checksum,
const u64 Ls[32]) ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx512_enc_blk64(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in)
ASM_FUNC_ABI;
extern void _gcry_camellia_gfni_avx512_dec_blk64(const CAMELLIA_context *ctx,
unsigned char *out,
const unsigned char *in)
ASM_FUNC_ABI;
+extern void _gcry_camellia_gfni_avx512_enc_blk1(const CAMELLIA_context *ctx,
+ unsigned char *out,
+ const unsigned char *in)
+ ASM_FUNC_ABI;
+
+extern void _gcry_camellia_gfni_avx512_dec_blk1(const CAMELLIA_context *ctx,
+ unsigned char *out,
+ const unsigned char *in)
+ ASM_FUNC_ABI;
+
/* Stack not used by AVX512 implementation. */
static const int avx512_burn_stack_depth = 0;
#endif
#ifdef USE_PPC_CRYPTO
extern void _gcry_camellia_ppc8_encrypt_blk16(const void *key_table,
void *out,
const void *in,
int key_length);
extern void _gcry_camellia_ppc8_decrypt_blk16(const void *key_table,
void *out,
const void *in,
int key_length);
extern void _gcry_camellia_ppc9_encrypt_blk16(const void *key_table,
void *out,
const void *in,
int key_length);
extern void _gcry_camellia_ppc9_decrypt_blk16(const void *key_table,
void *out,
const void *in,
int key_length);
extern void _gcry_camellia_ppc8_keygen(void *key_table, const void *vkey,
unsigned int keylen);
extern void _gcry_camellia_ppc9_keygen(void *key_table, const void *vkey,
unsigned int keylen);
void camellia_ppc_enc_blk16(const CAMELLIA_context *ctx, unsigned char *out,
const unsigned char *in)
{
if (ctx->use_ppc9)
_gcry_camellia_ppc9_encrypt_blk16 (ctx->keytable, out, in,
ctx->keybitlength / 8);
else
_gcry_camellia_ppc8_encrypt_blk16 (ctx->keytable, out, in,
ctx->keybitlength / 8);
}
void camellia_ppc_dec_blk16(const CAMELLIA_context *ctx, unsigned char *out,
const unsigned char *in)
{
if (ctx->use_ppc9)
_gcry_camellia_ppc9_decrypt_blk16 (ctx->keytable, out, in,
ctx->keybitlength / 8);
else
_gcry_camellia_ppc8_decrypt_blk16 (ctx->keytable, out, in,
ctx->keybitlength / 8);
}
static const int ppc_burn_stack_depth = 16 * CAMELLIA_BLOCK_SIZE + 16 +
2 * sizeof(void *);
#endif /*USE_PPC_CRYPTO*/
#ifdef USE_AARCH64_CE
extern void _gcry_camellia_aarch64ce_encrypt_blk16(const void *key_table,
void *out, const void *in,
int key_length);
extern void _gcry_camellia_aarch64ce_decrypt_blk16(const void *key_table,
void *out, const void *in,
int key_length);
extern void _gcry_camellia_aarch64ce_keygen(void *key_table, const void *vkey,
unsigned int keylen);
void camellia_aarch64ce_enc_blk16(const CAMELLIA_context *ctx,
unsigned char *out, const unsigned char *in)
{
_gcry_camellia_aarch64ce_encrypt_blk16 (ctx->keytable, out, in,
ctx->keybitlength / 8);
}
void camellia_aarch64ce_dec_blk16(const CAMELLIA_context *ctx,
unsigned char *out, const unsigned char *in)
{
_gcry_camellia_aarch64ce_decrypt_blk16 (ctx->keytable, out, in,
ctx->keybitlength / 8);
}
static const int aarch64ce_burn_stack_depth = 16 * CAMELLIA_BLOCK_SIZE + 16 +
2 * sizeof(void *);
#endif /*USE_AARCH64_CE*/
static const char *selftest(void);
static void _gcry_camellia_ctr_enc (void *context, unsigned char *ctr,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks);
static void _gcry_camellia_cbc_dec (void *context, unsigned char *iv,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks);
static void _gcry_camellia_cfb_dec (void *context, unsigned char *iv,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks);
static void _gcry_camellia_xts_crypt (void *context, unsigned char *tweak,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks, int encrypt);
static void _gcry_camellia_ecb_crypt (void *context, void *outbuf_arg,
const void *inbuf_arg, size_t nblocks,
int encrypt);
static void _gcry_camellia_ctr32le_enc (void *context, unsigned char *ctr,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks);
static size_t _gcry_camellia_ocb_crypt (gcry_cipher_hd_t c, void *outbuf_arg,
const void *inbuf_arg, size_t nblocks,
int encrypt);
static size_t _gcry_camellia_ocb_auth (gcry_cipher_hd_t c, const void *abuf_arg,
size_t nblocks);
static gcry_err_code_t
camellia_setkey(void *c, const byte *key, unsigned keylen,
cipher_bulk_ops_t *bulk_ops)
{
CAMELLIA_context *ctx=c;
static int initialized=0;
static const char *selftest_failed=NULL;
unsigned int hwf = _gcry_get_hw_features ();
(void)hwf;
if(keylen!=16 && keylen!=24 && keylen!=32)
return GPG_ERR_INV_KEYLEN;
if(!initialized)
{
initialized=1;
selftest_failed=selftest();
if(selftest_failed)
log_error("%s\n",selftest_failed);
}
if(selftest_failed)
return GPG_ERR_SELFTEST_FAILED;
#ifdef USE_AESNI_AVX
ctx->use_aesni_avx = (hwf & HWF_INTEL_AESNI) && (hwf & HWF_INTEL_AVX);
#endif
#ifdef USE_AESNI_AVX2
ctx->use_aesni_avx2 = (hwf & HWF_INTEL_AESNI) && (hwf & HWF_INTEL_AVX2);
ctx->use_vaes_avx2 = 0;
ctx->use_gfni_avx2 = 0;
ctx->use_gfni_avx512 = 0;
ctx->use_avx2 = ctx->use_aesni_avx2;
#endif
#ifdef USE_VAES_AVX2
ctx->use_vaes_avx2 = (hwf & HWF_INTEL_VAES_VPCLMUL) && (hwf & HWF_INTEL_AVX2);
ctx->use_avx2 |= ctx->use_vaes_avx2;
#endif
#ifdef USE_GFNI_AVX2
ctx->use_gfni_avx2 = (hwf & HWF_INTEL_GFNI) && (hwf & HWF_INTEL_AVX2);
ctx->use_avx2 |= ctx->use_gfni_avx2;
#endif
#ifdef USE_GFNI_AVX512
ctx->use_gfni_avx512 = (hwf & HWF_INTEL_GFNI) && (hwf & HWF_INTEL_AVX512);
#endif
#ifdef USE_PPC_CRYPTO
ctx->use_ppc8 = (hwf & HWF_PPC_VCRYPTO) != 0;
ctx->use_ppc9 = (hwf & HWF_PPC_VCRYPTO) && (hwf & HWF_PPC_ARCH_3_00);
ctx->use_ppc = ctx->use_ppc8 || ctx->use_ppc9;
#endif
#ifdef USE_AARCH64_CE
ctx->use_aarch64ce = (hwf & HWF_ARM_AES) != 0;
#endif
ctx->keybitlength=keylen*8;
/* Setup bulk encryption routines. */
memset (bulk_ops, 0, sizeof(*bulk_ops));
bulk_ops->cbc_dec = _gcry_camellia_cbc_dec;
bulk_ops->cfb_dec = _gcry_camellia_cfb_dec;
bulk_ops->ctr_enc = _gcry_camellia_ctr_enc;
bulk_ops->ocb_crypt = _gcry_camellia_ocb_crypt;
bulk_ops->ocb_auth = _gcry_camellia_ocb_auth;
bulk_ops->xts_crypt = _gcry_camellia_xts_crypt;
bulk_ops->ecb_crypt = _gcry_camellia_ecb_crypt;
bulk_ops->ctr32le_enc = _gcry_camellia_ctr32le_enc;
if (0)
{ }
#ifdef USE_AESNI_AVX
else if (ctx->use_aesni_avx)
_gcry_camellia_aesni_avx_keygen(ctx, key, keylen);
#endif
#ifdef USE_PPC_CRYPTO
else if (ctx->use_ppc9)
_gcry_camellia_ppc9_keygen(ctx->keytable, key, keylen);
else if (ctx->use_ppc8)
_gcry_camellia_ppc8_keygen(ctx->keytable, key, keylen);
#endif
#ifdef USE_AARCH64_CE
else if (ctx->use_aarch64ce)
_gcry_camellia_aarch64ce_keygen(ctx->keytable, key, keylen);
#endif
else
{
Camellia_Ekeygen(ctx->keybitlength,key,ctx->keytable);
_gcry_burn_stack
((19+34+34)*sizeof(u32)+2*sizeof(void*) /* camellia_setup256 */
+(4+32)*sizeof(u32)+2*sizeof(void*) /* camellia_setup192 */
+0+sizeof(int)+2*sizeof(void*) /* Camellia_Ekeygen */
+3*2*sizeof(void*) /* Function calls. */
);
}
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2)
{
/* Disable AESNI & VAES implementations when GFNI implementation is
* enabled. */
#ifdef USE_AESNI_AVX
ctx->use_aesni_avx = 0;
#endif
#ifdef USE_AESNI_AVX2
ctx->use_aesni_avx2 = 0;
#endif
#ifdef USE_VAES_AVX2
ctx->use_vaes_avx2 = 0;
#endif
}
#endif
return 0;
}
#ifdef USE_ARM_ASM
/* Assembly implementations of Camellia. */
extern void _gcry_camellia_arm_encrypt_block(const KEY_TABLE_TYPE keyTable,
byte *outbuf, const byte *inbuf,
const int keybits);
extern void _gcry_camellia_arm_decrypt_block(const KEY_TABLE_TYPE keyTable,
byte *outbuf, const byte *inbuf,
const int keybits);
static void Camellia_EncryptBlock(const int keyBitLength,
const unsigned char *plaintext,
const KEY_TABLE_TYPE keyTable,
unsigned char *cipherText)
{
_gcry_camellia_arm_encrypt_block(keyTable, cipherText, plaintext,
keyBitLength);
}
static void Camellia_DecryptBlock(const int keyBitLength,
const unsigned char *cipherText,
const KEY_TABLE_TYPE keyTable,
unsigned char *plaintext)
{
_gcry_camellia_arm_decrypt_block(keyTable, plaintext, cipherText,
keyBitLength);
}
#ifdef __aarch64__
# define CAMELLIA_encrypt_stack_burn_size (0)
# define CAMELLIA_decrypt_stack_burn_size (0)
#else
# define CAMELLIA_encrypt_stack_burn_size (15*4)
# define CAMELLIA_decrypt_stack_burn_size (15*4)
#endif
static unsigned int
camellia_encrypt(void *c, byte *outbuf, const byte *inbuf)
{
CAMELLIA_context *ctx = c;
Camellia_EncryptBlock(ctx->keybitlength,inbuf,ctx->keytable,outbuf);
return /*burn_stack*/ (CAMELLIA_encrypt_stack_burn_size);
}
static unsigned int
camellia_decrypt(void *c, byte *outbuf, const byte *inbuf)
{
CAMELLIA_context *ctx=c;
Camellia_DecryptBlock(ctx->keybitlength,inbuf,ctx->keytable,outbuf);
return /*burn_stack*/ (CAMELLIA_decrypt_stack_burn_size);
}
#else /*USE_ARM_ASM*/
static unsigned int
camellia_encrypt(void *c, byte *outbuf, const byte *inbuf)
{
CAMELLIA_context *ctx=c;
+#ifdef USE_GFNI_AVX512
+ if (ctx->use_gfni_avx512)
+ {
+ _gcry_camellia_gfni_avx512_enc_blk1(ctx, outbuf, inbuf);
+ return 0;
+ }
+#endif
+
Camellia_EncryptBlock(ctx->keybitlength,inbuf,ctx->keytable,outbuf);
#define CAMELLIA_encrypt_stack_burn_size \
(sizeof(int)+2*sizeof(unsigned char *)+sizeof(void*/*KEY_TABLE_TYPE*/) \
+4*sizeof(u32)+4*sizeof(u32) \
+2*sizeof(u32*)+4*sizeof(u32) \
+2*2*sizeof(void*) /* Function calls. */ \
)
return /*burn_stack*/ (CAMELLIA_encrypt_stack_burn_size);
}
static unsigned int
camellia_decrypt(void *c, byte *outbuf, const byte *inbuf)
{
CAMELLIA_context *ctx=c;
+#ifdef USE_GFNI_AVX512
+ if (ctx->use_gfni_avx512)
+ {
+ _gcry_camellia_gfni_avx512_dec_blk1(ctx, outbuf, inbuf);
+ return 0;
+ }
+#endif
+
Camellia_DecryptBlock(ctx->keybitlength,inbuf,ctx->keytable,outbuf);
#define CAMELLIA_decrypt_stack_burn_size \
(sizeof(int)+2*sizeof(unsigned char *)+sizeof(void*/*KEY_TABLE_TYPE*/) \
+4*sizeof(u32)+4*sizeof(u32) \
+2*sizeof(u32*)+4*sizeof(u32) \
+2*2*sizeof(void*) /* Function calls. */ \
)
return /*burn_stack*/ (CAMELLIA_decrypt_stack_burn_size);
}
#endif /*!USE_ARM_ASM*/
static unsigned int
camellia_encrypt_blk1_32 (void *priv, byte *outbuf, const byte *inbuf,
size_t num_blks)
{
const CAMELLIA_context *ctx = priv;
unsigned int stack_burn_size = 0;
gcry_assert (num_blks <= 32);
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2 && num_blks >= 2)
{
/* 2 or more parallel block GFNI processing is faster than
* generic C implementation. */
_gcry_camellia_gfni_avx2_enc_blk1_32 (ctx, outbuf, inbuf, num_blks);
return avx2_burn_stack_depth;
}
#endif
#ifdef USE_VAES_AVX2
if (ctx->use_vaes_avx2 && num_blks >= 4)
{
/* 4 or more parallel block VAES processing is faster than
* generic C implementation. */
_gcry_camellia_vaes_avx2_enc_blk1_32 (ctx, outbuf, inbuf, num_blks);
return avx2_burn_stack_depth;
}
#endif
#ifdef USE_AESNI_AVX2
if (ctx->use_aesni_avx2 && num_blks >= 5)
{
/* 5 or more parallel block AESNI processing is faster than
* generic C implementation. */
_gcry_camellia_aesni_avx2_enc_blk1_32 (ctx, outbuf, inbuf, num_blks);
return avx2_burn_stack_depth;
}
#endif
#ifdef USE_AESNI_AVX
while (ctx->use_aesni_avx && num_blks >= 16)
{
_gcry_camellia_aesni_avx_ecb_enc (ctx, outbuf, inbuf);
stack_burn_size = avx_burn_stack_depth;
outbuf += CAMELLIA_BLOCK_SIZE * 16;
inbuf += CAMELLIA_BLOCK_SIZE * 16;
num_blks -= 16;
}
#endif
#ifdef USE_PPC_CRYPTO
while (ctx->use_ppc && num_blks >= 16)
{
camellia_ppc_enc_blk16 (ctx, outbuf, inbuf);
stack_burn_size = ppc_burn_stack_depth;
outbuf += CAMELLIA_BLOCK_SIZE * 16;
inbuf += CAMELLIA_BLOCK_SIZE * 16;
num_blks -= 16;
}
#endif
#ifdef USE_AARCH64_CE
while (ctx->use_aarch64ce && num_blks >= 16)
{
camellia_aarch64ce_enc_blk16 (ctx, outbuf, inbuf);
stack_burn_size = aarch64ce_burn_stack_depth;
outbuf += CAMELLIA_BLOCK_SIZE * 16;
inbuf += CAMELLIA_BLOCK_SIZE * 16;
num_blks -= 16;
}
#endif
while (num_blks)
{
unsigned int nburn = camellia_encrypt((void *)ctx, outbuf, inbuf);
stack_burn_size = nburn > stack_burn_size ? nburn : stack_burn_size;
outbuf += CAMELLIA_BLOCK_SIZE;
inbuf += CAMELLIA_BLOCK_SIZE;
num_blks--;
}
return stack_burn_size;
}
static unsigned int
camellia_encrypt_blk1_64 (void *priv, byte *outbuf, const byte *inbuf,
size_t num_blks)
{
CAMELLIA_context *ctx = priv;
unsigned int stack_burn_size = 0;
unsigned int nburn;
gcry_assert (num_blks <= 64);
#ifdef USE_GFNI_AVX512
if (num_blks == 64 && ctx->use_gfni_avx512)
{
_gcry_camellia_gfni_avx512_enc_blk64 (ctx, outbuf, inbuf);
return avx512_burn_stack_depth;
}
#endif
do
{
unsigned int curr_blks = num_blks > 32 ? 32 : num_blks;
nburn = camellia_encrypt_blk1_32 (ctx, outbuf, inbuf, curr_blks);
stack_burn_size = nburn > stack_burn_size ? nburn : stack_burn_size;
outbuf += curr_blks * 16;
inbuf += curr_blks * 16;
num_blks -= curr_blks;
}
while (num_blks > 0);
return stack_burn_size;
}
static unsigned int
camellia_decrypt_blk1_32 (void *priv, byte *outbuf, const byte *inbuf,
size_t num_blks)
{
const CAMELLIA_context *ctx = priv;
unsigned int stack_burn_size = 0;
gcry_assert (num_blks <= 32);
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2 && num_blks >= 2)
{
/* 2 or more parallel block GFNI processing is faster than
* generic C implementation. */
_gcry_camellia_gfni_avx2_dec_blk1_32 (ctx, outbuf, inbuf, num_blks);
return avx2_burn_stack_depth;
}
#endif
#ifdef USE_VAES_AVX2
if (ctx->use_vaes_avx2 && num_blks >= 4)
{
/* 4 or more parallel block VAES processing is faster than
* generic C implementation. */
_gcry_camellia_vaes_avx2_dec_blk1_32 (ctx, outbuf, inbuf, num_blks);
return avx2_burn_stack_depth;
}
#endif
#ifdef USE_AESNI_AVX2
if (ctx->use_aesni_avx2 && num_blks >= 5)
{
/* 5 or more parallel block AESNI processing is faster than
* generic C implementation. */
_gcry_camellia_aesni_avx2_dec_blk1_32 (ctx, outbuf, inbuf, num_blks);
return avx2_burn_stack_depth;
}
#endif
#ifdef USE_AESNI_AVX
while (ctx->use_aesni_avx && num_blks >= 16)
{
_gcry_camellia_aesni_avx_ecb_dec (ctx, outbuf, inbuf);
stack_burn_size = avx_burn_stack_depth;
outbuf += CAMELLIA_BLOCK_SIZE * 16;
inbuf += CAMELLIA_BLOCK_SIZE * 16;
num_blks -= 16;
}
#endif
#ifdef USE_PPC_CRYPTO
while (ctx->use_ppc && num_blks >= 16)
{
camellia_ppc_dec_blk16 (ctx, outbuf, inbuf);
stack_burn_size = ppc_burn_stack_depth;
outbuf += CAMELLIA_BLOCK_SIZE * 16;
inbuf += CAMELLIA_BLOCK_SIZE * 16;
num_blks -= 16;
}
#endif
#ifdef USE_AARCH64_CE
while (ctx->use_aarch64ce && num_blks >= 16)
{
camellia_aarch64ce_dec_blk16 (ctx, outbuf, inbuf);
stack_burn_size = aarch64ce_burn_stack_depth;
outbuf += CAMELLIA_BLOCK_SIZE * 16;
inbuf += CAMELLIA_BLOCK_SIZE * 16;
num_blks -= 16;
}
#endif
while (num_blks)
{
unsigned int nburn = camellia_decrypt((void *)ctx, outbuf, inbuf);
stack_burn_size = nburn > stack_burn_size ? nburn : stack_burn_size;
outbuf += CAMELLIA_BLOCK_SIZE;
inbuf += CAMELLIA_BLOCK_SIZE;
num_blks--;
}
return stack_burn_size;
}
static unsigned int
camellia_decrypt_blk1_64 (void *priv, byte *outbuf, const byte *inbuf,
size_t num_blks)
{
CAMELLIA_context *ctx = priv;
unsigned int stack_burn_size = 0;
unsigned int nburn;
gcry_assert (num_blks <= 64);
#ifdef USE_GFNI_AVX512
if (num_blks == 64 && ctx->use_gfni_avx512)
{
_gcry_camellia_gfni_avx512_dec_blk64 (ctx, outbuf, inbuf);
return avx512_burn_stack_depth;
}
#endif
do
{
unsigned int curr_blks = num_blks > 32 ? 32 : num_blks;
nburn = camellia_decrypt_blk1_32 (ctx, outbuf, inbuf, curr_blks);
stack_burn_size = nburn > stack_burn_size ? nburn : stack_burn_size;
outbuf += curr_blks * 16;
inbuf += curr_blks * 16;
num_blks -= curr_blks;
}
while (num_blks > 0);
return stack_burn_size;
}
/* Bulk encryption of complete blocks in CTR mode. This function is only
intended for the bulk encryption feature of cipher.c. CTR is expected to be
of size CAMELLIA_BLOCK_SIZE. */
static void
_gcry_camellia_ctr_enc(void *context, unsigned char *ctr,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks)
{
CAMELLIA_context *ctx = context;
unsigned char *outbuf = outbuf_arg;
const unsigned char *inbuf = inbuf_arg;
int burn_stack_depth = 0;
#ifdef USE_GFNI_AVX512
if (ctx->use_gfni_avx512)
{
int did_use_gfni_avx512 = 0;
/* Process data in 64 block chunks. */
while (nblocks >= 64)
{
_gcry_camellia_gfni_avx512_ctr_enc (ctx, outbuf, inbuf, ctr);
nblocks -= 64;
outbuf += 64 * CAMELLIA_BLOCK_SIZE;
inbuf += 64 * CAMELLIA_BLOCK_SIZE;
did_use_gfni_avx512 = 1;
}
if (did_use_gfni_avx512)
{
if (burn_stack_depth < avx512_burn_stack_depth)
burn_stack_depth = avx512_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX2
if (ctx->use_avx2)
{
int did_use_aesni_avx2 = 0;
typeof (&_gcry_camellia_aesni_avx2_ctr_enc) bulk_ctr_fn =
_gcry_camellia_aesni_avx2_ctr_enc;
#ifdef USE_VAES_AVX2
if (ctx->use_vaes_avx2)
bulk_ctr_fn =_gcry_camellia_vaes_avx2_ctr_enc;
#endif
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2)
bulk_ctr_fn =_gcry_camellia_gfni_avx2_ctr_enc;
#endif
/* Process data in 32 block chunks. */
while (nblocks >= 32)
{
bulk_ctr_fn (ctx, outbuf, inbuf, ctr);
nblocks -= 32;
outbuf += 32 * CAMELLIA_BLOCK_SIZE;
inbuf += 32 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx2 = 1;
}
if (did_use_aesni_avx2)
{
if (burn_stack_depth < avx2_burn_stack_depth)
burn_stack_depth = avx2_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX
if (ctx->use_aesni_avx)
{
int did_use_aesni_avx = 0;
/* Process data in 16 block chunks. */
while (nblocks >= 16)
{
_gcry_camellia_aesni_avx_ctr_enc(ctx, outbuf, inbuf, ctr);
nblocks -= 16;
outbuf += 16 * CAMELLIA_BLOCK_SIZE;
inbuf += 16 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx = 1;
}
if (did_use_aesni_avx)
{
if (burn_stack_depth < avx_burn_stack_depth)
burn_stack_depth = avx_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
/* Process remaining blocks. */
if (nblocks)
{
byte tmpbuf[CAMELLIA_BLOCK_SIZE * 32];
unsigned int tmp_used = CAMELLIA_BLOCK_SIZE;
size_t nburn;
nburn = bulk_ctr_enc_128(ctx, camellia_encrypt_blk1_32, outbuf, inbuf,
nblocks, ctr, tmpbuf,
sizeof(tmpbuf) / CAMELLIA_BLOCK_SIZE, &tmp_used);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
wipememory(tmpbuf, tmp_used);
}
if (burn_stack_depth)
_gcry_burn_stack(burn_stack_depth);
}
/* Bulk decryption of complete blocks in CBC mode. This function is only
intended for the bulk encryption feature of cipher.c. */
static void
_gcry_camellia_cbc_dec(void *context, unsigned char *iv,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks)
{
CAMELLIA_context *ctx = context;
unsigned char *outbuf = outbuf_arg;
const unsigned char *inbuf = inbuf_arg;
int burn_stack_depth = 0;
#ifdef USE_GFNI_AVX512
if (ctx->use_gfni_avx512)
{
int did_use_gfni_avx512 = 0;
/* Process data in 64 block chunks. */
while (nblocks >= 64)
{
_gcry_camellia_gfni_avx512_cbc_dec (ctx, outbuf, inbuf, iv);
nblocks -= 64;
outbuf += 64 * CAMELLIA_BLOCK_SIZE;
inbuf += 64 * CAMELLIA_BLOCK_SIZE;
did_use_gfni_avx512 = 1;
}
if (did_use_gfni_avx512)
{
if (burn_stack_depth < avx512_burn_stack_depth)
burn_stack_depth = avx512_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX2
if (ctx->use_avx2)
{
int did_use_aesni_avx2 = 0;
typeof (&_gcry_camellia_aesni_avx2_cbc_dec) bulk_cbc_fn =
_gcry_camellia_aesni_avx2_cbc_dec;
#ifdef USE_VAES_AVX2
if (ctx->use_vaes_avx2)
bulk_cbc_fn =_gcry_camellia_vaes_avx2_cbc_dec;
#endif
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2)
bulk_cbc_fn =_gcry_camellia_gfni_avx2_cbc_dec;
#endif
/* Process data in 32 block chunks. */
while (nblocks >= 32)
{
bulk_cbc_fn (ctx, outbuf, inbuf, iv);
nblocks -= 32;
outbuf += 32 * CAMELLIA_BLOCK_SIZE;
inbuf += 32 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx2 = 1;
}
if (did_use_aesni_avx2)
{
if (burn_stack_depth < avx2_burn_stack_depth)
burn_stack_depth = avx2_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX
if (ctx->use_aesni_avx)
{
int did_use_aesni_avx = 0;
/* Process data in 16 block chunks. */
while (nblocks >= 16)
{
_gcry_camellia_aesni_avx_cbc_dec(ctx, outbuf, inbuf, iv);
nblocks -= 16;
outbuf += 16 * CAMELLIA_BLOCK_SIZE;
inbuf += 16 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx = 1;
}
if (did_use_aesni_avx)
{
if (burn_stack_depth < avx_burn_stack_depth)
burn_stack_depth = avx_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
/* Process remaining blocks. */
if (nblocks)
{
byte tmpbuf[CAMELLIA_BLOCK_SIZE * 32];
unsigned int tmp_used = CAMELLIA_BLOCK_SIZE;
size_t nburn;
nburn = bulk_cbc_dec_128(ctx, camellia_decrypt_blk1_32, outbuf, inbuf,
nblocks, iv, tmpbuf,
sizeof(tmpbuf) / CAMELLIA_BLOCK_SIZE, &tmp_used);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
wipememory(tmpbuf, tmp_used);
}
if (burn_stack_depth)
_gcry_burn_stack(burn_stack_depth);
}
/* Bulk decryption of complete blocks in CFB mode. This function is only
intended for the bulk encryption feature of cipher.c. */
static void
_gcry_camellia_cfb_dec(void *context, unsigned char *iv,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks)
{
CAMELLIA_context *ctx = context;
unsigned char *outbuf = outbuf_arg;
const unsigned char *inbuf = inbuf_arg;
int burn_stack_depth = 0;
#ifdef USE_GFNI_AVX512
if (ctx->use_gfni_avx512)
{
int did_use_gfni_avx512 = 0;
/* Process data in 64 block chunks. */
while (nblocks >= 64)
{
_gcry_camellia_gfni_avx512_cfb_dec (ctx, outbuf, inbuf, iv);
nblocks -= 64;
outbuf += 64 * CAMELLIA_BLOCK_SIZE;
inbuf += 64 * CAMELLIA_BLOCK_SIZE;
did_use_gfni_avx512 = 1;
}
if (did_use_gfni_avx512)
{
if (burn_stack_depth < avx512_burn_stack_depth)
burn_stack_depth = avx512_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX2
if (ctx->use_avx2)
{
int did_use_aesni_avx2 = 0;
typeof (&_gcry_camellia_aesni_avx2_cfb_dec) bulk_cfb_fn =
_gcry_camellia_aesni_avx2_cfb_dec;
#ifdef USE_VAES_AVX2
if (ctx->use_vaes_avx2)
bulk_cfb_fn =_gcry_camellia_vaes_avx2_cfb_dec;
#endif
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2)
bulk_cfb_fn =_gcry_camellia_gfni_avx2_cfb_dec;
#endif
/* Process data in 32 block chunks. */
while (nblocks >= 32)
{
bulk_cfb_fn (ctx, outbuf, inbuf, iv);
nblocks -= 32;
outbuf += 32 * CAMELLIA_BLOCK_SIZE;
inbuf += 32 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx2 = 1;
}
if (did_use_aesni_avx2)
{
if (burn_stack_depth < avx2_burn_stack_depth)
burn_stack_depth = avx2_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX
if (ctx->use_aesni_avx)
{
int did_use_aesni_avx = 0;
/* Process data in 16 block chunks. */
while (nblocks >= 16)
{
_gcry_camellia_aesni_avx_cfb_dec(ctx, outbuf, inbuf, iv);
nblocks -= 16;
outbuf += 16 * CAMELLIA_BLOCK_SIZE;
inbuf += 16 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx = 1;
}
if (did_use_aesni_avx)
{
if (burn_stack_depth < avx_burn_stack_depth)
burn_stack_depth = avx_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
/* Process remaining blocks. */
if (nblocks)
{
byte tmpbuf[CAMELLIA_BLOCK_SIZE * 32];
unsigned int tmp_used = CAMELLIA_BLOCK_SIZE;
size_t nburn;
nburn = bulk_cfb_dec_128(ctx, camellia_encrypt_blk1_32, outbuf, inbuf,
nblocks, iv, tmpbuf,
sizeof(tmpbuf) / CAMELLIA_BLOCK_SIZE, &tmp_used);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
wipememory(tmpbuf, tmp_used);
}
if (burn_stack_depth)
_gcry_burn_stack(burn_stack_depth);
}
/* Bulk encryption/decryption in ECB mode. */
static void
_gcry_camellia_ecb_crypt (void *context, void *outbuf_arg,
const void *inbuf_arg, size_t nblocks, int encrypt)
{
CAMELLIA_context *ctx = context;
unsigned char *outbuf = outbuf_arg;
const unsigned char *inbuf = inbuf_arg;
int burn_stack_depth = 0;
/* Process remaining blocks. */
if (nblocks)
{
size_t nburn;
nburn = bulk_ecb_crypt_128(ctx, encrypt ? camellia_encrypt_blk1_64
: camellia_decrypt_blk1_64,
outbuf, inbuf, nblocks, 64);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
}
if (burn_stack_depth)
_gcry_burn_stack(burn_stack_depth);
}
/* Bulk encryption/decryption of complete blocks in XTS mode. */
static void
_gcry_camellia_xts_crypt (void *context, unsigned char *tweak,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks, int encrypt)
{
CAMELLIA_context *ctx = context;
unsigned char *outbuf = outbuf_arg;
const unsigned char *inbuf = inbuf_arg;
int burn_stack_depth = 0;
/* Process remaining blocks. */
if (nblocks)
{
byte tmpbuf[CAMELLIA_BLOCK_SIZE * 64];
unsigned int tmp_used = CAMELLIA_BLOCK_SIZE;
size_t nburn;
nburn = bulk_xts_crypt_128(ctx, encrypt ? camellia_encrypt_blk1_64
: camellia_decrypt_blk1_64,
outbuf, inbuf, nblocks, tweak, tmpbuf,
sizeof(tmpbuf) / CAMELLIA_BLOCK_SIZE,
&tmp_used);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
wipememory(tmpbuf, tmp_used);
}
if (burn_stack_depth)
_gcry_burn_stack(burn_stack_depth);
}
/* Bulk encryption of complete blocks in CTR32LE mode (for GCM-SIV). */
static void
_gcry_camellia_ctr32le_enc(void *context, unsigned char *ctr,
void *outbuf_arg, const void *inbuf_arg,
size_t nblocks)
{
CAMELLIA_context *ctx = context;
byte *outbuf = outbuf_arg;
const byte *inbuf = inbuf_arg;
int burn_stack_depth = 0;
/* Process remaining blocks. */
if (nblocks)
{
byte tmpbuf[64 * CAMELLIA_BLOCK_SIZE];
unsigned int tmp_used = CAMELLIA_BLOCK_SIZE;
size_t nburn;
nburn = bulk_ctr32le_enc_128 (ctx, camellia_encrypt_blk1_64, outbuf,
inbuf, nblocks, ctr, tmpbuf,
sizeof(tmpbuf) / CAMELLIA_BLOCK_SIZE,
&tmp_used);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
wipememory (tmpbuf, tmp_used);
}
if (burn_stack_depth)
_gcry_burn_stack (burn_stack_depth);
}
/* Bulk encryption/decryption of complete blocks in OCB mode. */
static size_t
_gcry_camellia_ocb_crypt (gcry_cipher_hd_t c, void *outbuf_arg,
const void *inbuf_arg, size_t nblocks, int encrypt)
{
#if defined(USE_PPC_CRYPTO) || defined(USE_AESNI_AVX) || defined(USE_AESNI_AVX2)
CAMELLIA_context *ctx = (void *)&c->context.c;
unsigned char *outbuf = outbuf_arg;
const unsigned char *inbuf = inbuf_arg;
int burn_stack_depth = 0;
u64 blkn = c->u_mode.ocb.data_nblocks;
#else
(void)c;
(void)outbuf_arg;
(void)inbuf_arg;
(void)encrypt;
#endif
#ifdef USE_GFNI_AVX512
if (ctx->use_gfni_avx512)
{
int did_use_gfni_avx512 = 0;
u64 Ls[64];
u64 *l;
if (nblocks >= 64)
{
typeof (&_gcry_camellia_gfni_avx512_ocb_dec) bulk_ocb_fn =
encrypt ? _gcry_camellia_gfni_avx512_ocb_enc
: _gcry_camellia_gfni_avx512_ocb_dec;
l = bulk_ocb_prepare_L_pointers_array_blk64 (c, Ls, blkn);
/* Process data in 64 block chunks. */
while (nblocks >= 64)
{
blkn += 64;
*l = (uintptr_t)(void *)ocb_get_l(c, blkn - blkn % 64);
bulk_ocb_fn (ctx, outbuf, inbuf, c->u_iv.iv, c->u_ctr.ctr, Ls);
nblocks -= 64;
outbuf += 64 * CAMELLIA_BLOCK_SIZE;
inbuf += 64 * CAMELLIA_BLOCK_SIZE;
did_use_gfni_avx512 = 1;
}
}
if (did_use_gfni_avx512)
{
if (burn_stack_depth < avx2_burn_stack_depth)
burn_stack_depth = avx2_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX2
if (ctx->use_avx2)
{
int did_use_aesni_avx2 = 0;
u64 Ls[32];
u64 *l;
if (nblocks >= 32)
{
typeof (&_gcry_camellia_aesni_avx2_ocb_dec) bulk_ocb_fn =
encrypt ? _gcry_camellia_aesni_avx2_ocb_enc
: _gcry_camellia_aesni_avx2_ocb_dec;
#ifdef USE_VAES_AVX2
if (ctx->use_vaes_avx2)
bulk_ocb_fn = encrypt ? _gcry_camellia_vaes_avx2_ocb_enc
: _gcry_camellia_vaes_avx2_ocb_dec;
#endif
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2)
bulk_ocb_fn = encrypt ? _gcry_camellia_gfni_avx2_ocb_enc
: _gcry_camellia_gfni_avx2_ocb_dec;
#endif
l = bulk_ocb_prepare_L_pointers_array_blk32 (c, Ls, blkn);
/* Process data in 32 block chunks. */
while (nblocks >= 32)
{
blkn += 32;
*l = (uintptr_t)(void *)ocb_get_l(c, blkn - blkn % 32);
bulk_ocb_fn (ctx, outbuf, inbuf, c->u_iv.iv, c->u_ctr.ctr, Ls);
nblocks -= 32;
outbuf += 32 * CAMELLIA_BLOCK_SIZE;
inbuf += 32 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx2 = 1;
}
}
if (did_use_aesni_avx2)
{
if (burn_stack_depth < avx2_burn_stack_depth)
burn_stack_depth = avx2_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX
if (ctx->use_aesni_avx)
{
int did_use_aesni_avx = 0;
u64 Ls[16];
u64 *l;
if (nblocks >= 16)
{
l = bulk_ocb_prepare_L_pointers_array_blk16 (c, Ls, blkn);
/* Process data in 16 block chunks. */
while (nblocks >= 16)
{
blkn += 16;
*l = (uintptr_t)(void *)ocb_get_l(c, blkn - blkn % 16);
if (encrypt)
_gcry_camellia_aesni_avx_ocb_enc(ctx, outbuf, inbuf, c->u_iv.iv,
c->u_ctr.ctr, Ls);
else
_gcry_camellia_aesni_avx_ocb_dec(ctx, outbuf, inbuf, c->u_iv.iv,
c->u_ctr.ctr, Ls);
nblocks -= 16;
outbuf += 16 * CAMELLIA_BLOCK_SIZE;
inbuf += 16 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx = 1;
}
}
if (did_use_aesni_avx)
{
if (burn_stack_depth < avx_burn_stack_depth)
burn_stack_depth = avx_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#if defined(USE_PPC_CRYPTO) || defined(USE_AESNI_AVX) || defined(USE_AESNI_AVX2)
/* Process remaining blocks. */
if (nblocks)
{
byte tmpbuf[CAMELLIA_BLOCK_SIZE * 32];
unsigned int tmp_used = CAMELLIA_BLOCK_SIZE;
size_t nburn;
nburn = bulk_ocb_crypt_128 (c, ctx, encrypt ? camellia_encrypt_blk1_32
: camellia_decrypt_blk1_32,
outbuf, inbuf, nblocks, &blkn, encrypt,
tmpbuf, sizeof(tmpbuf) / CAMELLIA_BLOCK_SIZE,
&tmp_used);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
wipememory(tmpbuf, tmp_used);
nblocks = 0;
}
c->u_mode.ocb.data_nblocks = blkn;
if (burn_stack_depth)
_gcry_burn_stack (burn_stack_depth + 4 * sizeof(void *));
#endif
return nblocks;
}
/* Bulk authentication of complete blocks in OCB mode. */
static size_t
_gcry_camellia_ocb_auth (gcry_cipher_hd_t c, const void *abuf_arg,
size_t nblocks)
{
#if defined(USE_PPC_CRYPTO) || defined(USE_AESNI_AVX) || defined(USE_AESNI_AVX2)
CAMELLIA_context *ctx = (void *)&c->context.c;
const unsigned char *abuf = abuf_arg;
int burn_stack_depth = 0;
u64 blkn = c->u_mode.ocb.aad_nblocks;
#else
(void)c;
(void)abuf_arg;
#endif
#ifdef USE_AESNI_AVX2
if (ctx->use_avx2)
{
int did_use_aesni_avx2 = 0;
u64 Ls[32];
u64 *l;
if (nblocks >= 32)
{
typeof (&_gcry_camellia_aesni_avx2_ocb_auth) bulk_auth_fn =
_gcry_camellia_aesni_avx2_ocb_auth;
#ifdef USE_VAES_AVX2
if (ctx->use_vaes_avx2)
bulk_auth_fn = _gcry_camellia_vaes_avx2_ocb_auth;
#endif
#ifdef USE_GFNI_AVX2
if (ctx->use_gfni_avx2)
bulk_auth_fn = _gcry_camellia_gfni_avx2_ocb_auth;
#endif
l = bulk_ocb_prepare_L_pointers_array_blk32 (c, Ls, blkn);
/* Process data in 32 block chunks. */
while (nblocks >= 32)
{
blkn += 32;
*l = (uintptr_t)(void *)ocb_get_l(c, blkn - blkn % 32);
bulk_auth_fn (ctx, abuf, c->u_mode.ocb.aad_offset,
c->u_mode.ocb.aad_sum, Ls);
nblocks -= 32;
abuf += 32 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx2 = 1;
}
}
if (did_use_aesni_avx2)
{
if (burn_stack_depth < avx2_burn_stack_depth)
burn_stack_depth = avx2_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#ifdef USE_AESNI_AVX
if (ctx->use_aesni_avx)
{
int did_use_aesni_avx = 0;
u64 Ls[16];
u64 *l;
if (nblocks >= 16)
{
l = bulk_ocb_prepare_L_pointers_array_blk16 (c, Ls, blkn);
/* Process data in 16 block chunks. */
while (nblocks >= 16)
{
blkn += 16;
*l = (uintptr_t)(void *)ocb_get_l(c, blkn - blkn % 16);
_gcry_camellia_aesni_avx_ocb_auth(ctx, abuf,
c->u_mode.ocb.aad_offset,
c->u_mode.ocb.aad_sum, Ls);
nblocks -= 16;
abuf += 16 * CAMELLIA_BLOCK_SIZE;
did_use_aesni_avx = 1;
}
}
if (did_use_aesni_avx)
{
if (burn_stack_depth < avx_burn_stack_depth)
burn_stack_depth = avx_burn_stack_depth;
}
/* Use generic code to handle smaller chunks... */
}
#endif
#if defined(USE_PPC_CRYPTO) || defined(USE_AESNI_AVX) || defined(USE_AESNI_AVX2)
/* Process remaining blocks. */
if (nblocks)
{
byte tmpbuf[CAMELLIA_BLOCK_SIZE * 32];
unsigned int tmp_used = CAMELLIA_BLOCK_SIZE;
size_t nburn;
nburn = bulk_ocb_auth_128 (c, ctx, camellia_encrypt_blk1_32,
abuf, nblocks, &blkn, tmpbuf,
sizeof(tmpbuf) / CAMELLIA_BLOCK_SIZE,
&tmp_used);
burn_stack_depth = nburn > burn_stack_depth ? nburn : burn_stack_depth;
wipememory(tmpbuf, tmp_used);
nblocks = 0;
}
c->u_mode.ocb.aad_nblocks = blkn;
if (burn_stack_depth)
_gcry_burn_stack (burn_stack_depth + 4 * sizeof(void *));
#endif
return nblocks;
}
static const char *
selftest(void)
{
CAMELLIA_context ctx;
byte scratch[16];
cipher_bulk_ops_t bulk_ops;
/* These test vectors are from RFC-3713 */
static const byte plaintext[]=
{
0x01,0x23,0x45,0x67,0x89,0xab,0xcd,0xef,
0xfe,0xdc,0xba,0x98,0x76,0x54,0x32,0x10
};
static const byte key_128[]=
{
0x01,0x23,0x45,0x67,0x89,0xab,0xcd,0xef,
0xfe,0xdc,0xba,0x98,0x76,0x54,0x32,0x10
};
static const byte ciphertext_128[]=
{
0x67,0x67,0x31,0x38,0x54,0x96,0x69,0x73,
0x08,0x57,0x06,0x56,0x48,0xea,0xbe,0x43
};
static const byte key_192[]=
{
0x01,0x23,0x45,0x67,0x89,0xab,0xcd,0xef,0xfe,0xdc,0xba,0x98,
0x76,0x54,0x32,0x10,0x00,0x11,0x22,0x33,0x44,0x55,0x66,0x77
};
static const byte ciphertext_192[]=
{
0xb4,0x99,0x34,0x01,0xb3,0xe9,0x96,0xf8,
0x4e,0xe5,0xce,0xe7,0xd7,0x9b,0x09,0xb9
};
static const byte key_256[]=
{
0x01,0x23,0x45,0x67,0x89,0xab,0xcd,0xef,0xfe,0xdc,0xba,
0x98,0x76,0x54,0x32,0x10,0x00,0x11,0x22,0x33,0x44,0x55,
0x66,0x77,0x88,0x99,0xaa,0xbb,0xcc,0xdd,0xee,0xff
};
static const byte ciphertext_256[]=
{
0x9a,0xcc,0x23,0x7d,0xff,0x16,0xd7,0x6c,
0x20,0xef,0x7c,0x91,0x9e,0x3a,0x75,0x09
};
camellia_setkey(&ctx,key_128,sizeof(key_128),&bulk_ops);
camellia_encrypt(&ctx,scratch,plaintext);
if(memcmp(scratch,ciphertext_128,sizeof(ciphertext_128))!=0)
return "CAMELLIA-128 test encryption failed.";
camellia_decrypt(&ctx,scratch,scratch);
if(memcmp(scratch,plaintext,sizeof(plaintext))!=0)
return "CAMELLIA-128 test decryption failed.";
camellia_setkey(&ctx,key_192,sizeof(key_192),&bulk_ops);
camellia_encrypt(&ctx,scratch,plaintext);
if(memcmp(scratch,ciphertext_192,sizeof(ciphertext_192))!=0)
return "CAMELLIA-192 test encryption failed.";
camellia_decrypt(&ctx,scratch,scratch);
if(memcmp(scratch,plaintext,sizeof(plaintext))!=0)
return "CAMELLIA-192 test decryption failed.";
camellia_setkey(&ctx,key_256,sizeof(key_256),&bulk_ops);
camellia_encrypt(&ctx,scratch,plaintext);
if(memcmp(scratch,ciphertext_256,sizeof(ciphertext_256))!=0)
return "CAMELLIA-256 test encryption failed.";
camellia_decrypt(&ctx,scratch,scratch);
if(memcmp(scratch,plaintext,sizeof(plaintext))!=0)
return "CAMELLIA-256 test decryption failed.";
return NULL;
}
/* These oids are from
<http://info.isl.ntt.co.jp/crypt/eng/camellia/specifications_oid.html>,
retrieved May 1, 2007. */
static const gcry_cipher_oid_spec_t camellia128_oids[] =
{
{"1.2.392.200011.61.1.1.1.2", GCRY_CIPHER_MODE_CBC},
{"0.3.4401.5.3.1.9.1", GCRY_CIPHER_MODE_ECB},
{"0.3.4401.5.3.1.9.3", GCRY_CIPHER_MODE_OFB},
{"0.3.4401.5.3.1.9.4", GCRY_CIPHER_MODE_CFB},
{ NULL }
};
static const gcry_cipher_oid_spec_t camellia192_oids[] =
{
{"1.2.392.200011.61.1.1.1.3", GCRY_CIPHER_MODE_CBC},
{"0.3.4401.5.3.1.9.21", GCRY_CIPHER_MODE_ECB},
{"0.3.4401.5.3.1.9.23", GCRY_CIPHER_MODE_OFB},
{"0.3.4401.5.3.1.9.24", GCRY_CIPHER_MODE_CFB},
{ NULL }
};
static const gcry_cipher_oid_spec_t camellia256_oids[] =
{
{"1.2.392.200011.61.1.1.1.4", GCRY_CIPHER_MODE_CBC},
{"0.3.4401.5.3.1.9.41", GCRY_CIPHER_MODE_ECB},
{"0.3.4401.5.3.1.9.43", GCRY_CIPHER_MODE_OFB},
{"0.3.4401.5.3.1.9.44", GCRY_CIPHER_MODE_CFB},
{ NULL }
};
gcry_cipher_spec_t _gcry_cipher_spec_camellia128 =
{
GCRY_CIPHER_CAMELLIA128, {0, 0},
"CAMELLIA128",NULL,camellia128_oids,CAMELLIA_BLOCK_SIZE,128,
sizeof(CAMELLIA_context),camellia_setkey,camellia_encrypt,camellia_decrypt
};
gcry_cipher_spec_t _gcry_cipher_spec_camellia192 =
{
GCRY_CIPHER_CAMELLIA192, {0, 0},
"CAMELLIA192",NULL,camellia192_oids,CAMELLIA_BLOCK_SIZE,192,
sizeof(CAMELLIA_context),camellia_setkey,camellia_encrypt,camellia_decrypt
};
gcry_cipher_spec_t _gcry_cipher_spec_camellia256 =
{
GCRY_CIPHER_CAMELLIA256, {0, 0},
"CAMELLIA256",NULL,camellia256_oids,CAMELLIA_BLOCK_SIZE,256,
sizeof(CAMELLIA_context),camellia_setkey,camellia_encrypt,camellia_decrypt
};

File Metadata

Mime Type
text/x-diff
Expires
Thu, Feb 5, 9:40 PM (21 h, 7 m)
Storage Engine
local-disk
Storage Format
Raw Data
Storage Handle
43/46/a77e973a2df77fc8625942e77ea2

Event Timeline