From patchwork Fri Jan 12 02:28:59 2024 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Dongsoo Lee X-Patchwork-Id: 13517916 Received: from mail.nsr.re.kr (unknown [210.104.33.65]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 7377C171C8; Fri, 12 Jan 2024 02:30:50 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=nsr.re.kr Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=nsr.re.kr Received: from 210.104.33.70 (nsr.re.kr) (using TLSv1.3 with cipher TLS_AES_128_GCM_SHA256 (128 bits)) by mail.nsr.re.kr with SMTP; Fri, 12 Jan 2024 11:30:34 +0900 X-Sender: letrhee@nsr.re.kr Received: from 192.168.155.188 ([192.168.155.188]) by mail.nsr.re.kr (Crinity Message Backbone-7.0.1) with SMTP ID 155; Fri, 12 Jan 2024 11:30:30 +0900 (KST) From: Dongsoo Lee To: Herbert Xu , "David S. Miller" , Jens Axboe , Eric Biggers , "Theodore Y. Ts'o" , Jaegeuk Kim , Thomas Gleixner , Ingo Molnar , Borislav Petkov , Dave Hansen , x86@kernel.org, "H. Peter Anvin" Cc: linux-crypto@vger.kernel.org, linux-block@vger.kernel.org, linux-fscrypt@vger.kernel.org, linux-kernel@vger.kernel.org, Dongsoo Lee Subject: [PATCH v6 RESEND 5/5] crypto: LEA block cipher x86_64 optimization Date: Fri, 12 Jan 2024 02:28:59 +0000 Message-Id: <20240112022859.2384-6-letrhee@nsr.re.kr> X-Mailer: git-send-email 2.40.1 In-Reply-To: <20240112022859.2384-1-letrhee@nsr.re.kr> References: <20240112022859.2384-1-letrhee@nsr.re.kr> Precedence: bulk X-Mailing-List: linux-fscrypt@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 For the x86_64 environment, we use AVX-512F/AVX2/SSE2 instructions. Since LEA uses 128-bit blocks of four 32-bit integers, for optimization, SSE2 encrypts 4 blocks, AVX2 encrypts 4/8 blocks, and AVX-512F encrypts 4/8/16 blocks at a time. Our submission provides a optimized implementation of ECB, CBC decryption, CTR, and XTS cipher operation modes on x86_64 CPUs supporting. Signed-off-by: Dongsoo Lee --- arch/x86/crypto/Kconfig | 29 + arch/x86/crypto/Makefile | 3 + arch/x86/crypto/lea-x86_64-asm.S | 2272 +++++++++++++++++++++++++++++ arch/x86/crypto/lea-x86_64-glue.c | 820 +++++++++++ 4 files changed, 3124 insertions(+) create mode 100644 arch/x86/crypto/lea-x86_64-asm.S create mode 100644 arch/x86/crypto/lea-x86_64-glue.c diff --git a/arch/x86/crypto/Kconfig b/arch/x86/crypto/Kconfig index 9bbfd01cfa2f..5acd8794cab4 100644 --- a/arch/x86/crypto/Kconfig +++ b/arch/x86/crypto/Kconfig @@ -342,6 +342,35 @@ config CRYPTO_ARIA_GFNI_AVX512_X86_64 Processes 64 blocks in parallel. +config CRYPTO_LEA_X86_64 + tristate "Ciphers: LEA with modes: ECB, CBC, CTR, XTS (SSE2/AVX2/AVX-512F)" + depends on X86 && 64BIT + select CRYPTO_LEA + imply CRYPTO_XTS + imply CRYPTO_CTR + help + LEA is a 128-bit lightweight block cipher developed by South Korea. + + LEA is the a Korean standard (KS X 3246) and is included in the + ISO/IEC 29192-2:2019 standard (Information security - Lightweight + cryptography - Part 2: Block ciphers). + + It consists of 32-bit integer addition, rotation, and XOR, which can + be performed effectively on CPUs that support 32-bit operations. + + It supports 128-bit, 192-bit, and 256-bit keys. + + See: + https://seed.kisa.or.kr/kisa/algorithm/EgovLeaInfo.do + + Architecture: x86_64 using: + - SSE2 (Streaming SIMD Extensions 2) + - AVX2 (Advanced Vector Extensions) + - AVX-512F (Advanced Vector Extensions-512F) + + Processes 4(SSE2, AVX2, AVX-512F), 8(AVX2, AVX-512F), 16(AVX-512F) + blocks in parallel. + config CRYPTO_CHACHA20_X86_64 tristate "Ciphers: ChaCha20, XChaCha20, XChaCha12 (SSSE3/AVX2/AVX-512VL)" depends on X86 && 64BIT diff --git a/arch/x86/crypto/Makefile b/arch/x86/crypto/Makefile index 9aa46093c91b..3c44d43d155d 100644 --- a/arch/x86/crypto/Makefile +++ b/arch/x86/crypto/Makefile @@ -109,6 +109,9 @@ aria-aesni-avx2-x86_64-y := aria-aesni-avx2-asm_64.o aria_aesni_avx2_glue.o obj-$(CONFIG_CRYPTO_ARIA_GFNI_AVX512_X86_64) += aria-gfni-avx512-x86_64.o aria-gfni-avx512-x86_64-y := aria-gfni-avx512-asm_64.o aria_gfni_avx512_glue.o +obj-$(CONFIG_CRYPTO_LEA_X86_64) += lea-x86_64.o +lea-x86_64-y := lea-x86_64-asm.o lea-x86_64-glue.o + quiet_cmd_perlasm = PERLASM $@ cmd_perlasm = $(PERL) $< > $@ $(obj)/%.S: $(src)/%.pl FORCE diff --git a/arch/x86/crypto/lea-x86_64-asm.S b/arch/x86/crypto/lea-x86_64-asm.S new file mode 100644 index 000000000000..97e29e1158b8 --- /dev/null +++ b/arch/x86/crypto/lea-x86_64-asm.S @@ -0,0 +1,2272 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * LEA Cipher SSE2/AVX2/AVX-512F parallel algorithm. + * + * - SSE2: 4 parallel blocks + * - AVX2: 8/4 parallel blocks + * - AVX-512F: 16/8/4 parallel blocks + * + * Copyright (c) 2023 National Security Research. + * Author: Dongsoo Lee + */ + +#include +#include + +.file "lea_x86_64-asm.S" + + +.section .text + +#define PASTE1(a,b) a##b + +#define LEA_MAX_KEYLENGTH (32 * 6 * 4) + +#define LEA_CTX_RK_ENC (0) +#define LEA_CTX_RK_DEC (LEA_MAX_KEYLENGTH) +#define LEA_CTX_ROUND (LEA_MAX_KEYLENGTH * 2) + +#define LOAD_CTR(addr, high, low) \ + mov (addr), high; \ + mov 8(addr), low; \ + bswap high; \ + bswap low + +#define ADD_CTR_AND_STORE(high, low, addr, val) \ + add val, low; \ + adc $0, high; \ + bswap high; \ + bswap low; \ + mov high, (addr); \ + mov low, 8(addr) + +#define XTS_TW_X0 %xmm12 +#define XTS_TW_X1 %xmm13 +#define XTS_TW_I2 %xmm0 +#define XTS_TW_O2 %xmm14 +#define XTS_TW_X3 %xmm15 +#define XTS_TW_X4 %xmm8 +#define XTS_TW_X5 %xmm9 +#define XTS_TW_I6 %xmm1 +#define XTS_TW_O6 %xmm10 +#define XTS_TW_X7 %xmm11 +#define XTS_TW_X8 %xmm2 +#define XTS_MASK %xmm7 + +#define XTS_TW_Y0 %ymm12 +#define XTS_TW_Y1 %ymm13 +#define XTS_TW_Y2 %ymm14 +#define XTS_TW_Y3 %ymm15 + + +#define BSWAPMASK_XMM %xmm7 +#define CTRCONST_XMM %xmm8 + +#define BSWAPMASK_YMM %ymm7 +#define CTRCONST_YMM %ymm8 + +#define CTR_64_low_src %rax +#define CTR_64_high_src %r9 +#define CTR_64_low %r10 +#define CTR_64_high %r11 + +#define XMM(n) PASTE1(%xmm, n) +#define YMM(n) PASTE1(%ymm, n) + +#ifdef CONFIG_AS_AVX512 + +#define ZMM(n) PASTE1(%zmm, n) + +#define XTS_TW_Z0 %zmm12 +#define XTS_TW_Z1 %zmm13 +#define XTS_TW_Z2 %zmm14 +#define XTS_TW_Z3 %zmm15 + +#define BSWAPMASK_ZMM %zmm7 +#define CTRCONST_ZMM %zmm8 + +#endif + +/* XOR, ADD, ROTR */ +#define XAR_SSE2(v0, v1, cur, pre, tmp1, tmp2, rk1, rk2)\ + movd rk1, tmp1; \ + movd rk2, tmp2; \ + pshufd $0, tmp1, tmp1; \ + pshufd $0, tmp2, tmp2; \ + pxor pre, tmp1; \ + pxor tmp2, cur; \ + paddd cur, tmp1; \ + movdqa tmp1, cur; \ + psrld v0, tmp1; \ + pslld v1, cur; \ + pxor tmp1, cur; + +#define XAR_AVX2(v0, v1, cur, pre, tmp1, tmp2, rk1, rk2)\ + vpbroadcastd rk1, tmp1; \ + vpbroadcastd rk2, tmp2; \ + vpxor pre, tmp1, tmp1; \ + vpxor tmp2, cur, cur; \ + vpaddd cur, tmp1, tmp1; \ + vpsrld v0, tmp1, cur; \ + vpslld v1, tmp1, tmp1; \ + vpxor tmp1, cur, cur; + +#ifdef CONFIG_AS_AVX512 +#define XAR_AVX512(v0, _v1, cur, pre, tmp1, tmp2, rk1, rk2) \ + vpbroadcastd rk1, tmp1; \ + vpbroadcastd rk2, tmp2; \ + vpxord pre, tmp1, tmp1; \ + vpxord tmp2, cur, cur; \ + vpaddd cur, tmp1, tmp1; \ + vprord v0, tmp1, cur; +#endif + +/* XOR, SUB, ROTR */ +#define XSR9_SSE2(v0, v3, tnext, tmp1, tmp2, rk1, rk2) \ + movd rk1, tmp1; \ + movdqa v0, tnext; \ + psrld $9, v0; \ + pslld $23, tnext; \ + pshufd $0, tmp1, tmp1; \ + pxor v0, tnext; \ + movd rk2, v0; \ + pxor v3, tmp1; \ + pshufd $0, v0, v0; \ + psubd tmp1, tnext; \ + pxor tnext, v0; + +#define XSR5_SSE2(v1, tnext, tmp1, tmp2, rk1, rk2) \ + movd rk1, tmp1; \ + pshufd $0, tmp1, tmp1; \ + pxor tnext, tmp1; \ + movdqa v1, tnext; \ + psrld $27, v1; \ + pslld $5, tnext; \ + pxor v1, tnext; \ + movd rk2, v1; \ + pshufd $0, v1, v1; \ + psubd tmp1, tnext; \ + pxor tnext, v1; + +#define XSR3_SSE2(v2, tnext, tmp1, tmp2, rk1, rk2) \ + movd rk1, tmp1; \ + movdqa v2, tmp2; \ + psrld $29, v2; \ + pslld $3, tmp2; \ + pshufd $0, tmp1, tmp1; \ + pxor tmp2, v2; \ + movd rk2, tmp2; \ + pxor tnext, tmp1; \ + pshufd $0, tmp2, tmp2; \ + psubd tmp1, v2; \ + pxor tmp2, v2; + +#define XSR9_AVX2(v0, v3, tnext, tmp1, tmp2, rk1, rk2) \ + vpbroadcastd rk1, tmp1; \ + vpsrld $9, v0, tmp2; \ + vpslld $23, v0, v0; \ + vpxor v3, tmp1, tmp1; \ + vpxor tmp2, v0, v0; \ + vpbroadcastd rk2, tmp2; \ + vpsubd tmp1, v0, tnext; \ + vpxor tmp2, tnext, v0; + +#define XSR5_AVX2(v1, tnext, tmp1, tmp2, rk1, rk2) \ + vpbroadcastd rk1, tmp1; \ + vpsrld $27, v1, tmp2; \ + vpslld $5, v1, v1; \ + vpxor tnext, tmp1, tmp1; \ + vpxor tmp2, v1, v1; \ + vpbroadcastd rk2, tmp2; \ + vpsubd tmp1, v1, tnext; \ + vpxor tmp2, tnext, v1; + +#define XSR3_AVX2(v2, tnext, tmp1, tmp2, rk1, rk2) \ + vpbroadcastd rk1, tmp1; \ + vpsrld $29, v2, tmp2; \ + vpslld $3, v2, v2; \ + vpxor tnext, tmp1, tmp1; \ + vpxor tmp2, v2, v2; \ + vpbroadcastd rk2, tmp2; \ + vpsubd tmp1, v2, v2; \ + vpxor tmp2, v2, v2; + +#ifdef CONFIG_AS_AVX512 +#define XSR9_AVX512(v0, v3, tnext, tmp1, tmp2, rk1, rk2)\ + vpbroadcastd rk1, tmp1; \ + vpbroadcastd rk2, tmp2; \ + vprord $9, v0, v0; \ + vpxord v3, tmp1, tmp1; \ + vpsubd tmp1, v0, tnext; \ + vpxord tmp2, tnext, v0 + +#define XSR5_AVX512(v1, tnext, tmp1, tmp2, rk1, rk2) \ + vpbroadcastd rk1, tmp1; \ + vpbroadcastd rk2, tmp2; \ + vprold $5, v1, v1; \ + vpxord tnext, tmp1, tmp1; \ + vpsubd tmp1, v1, tnext; \ + vpxord tmp2, tnext, v1 + +#define XSR3_AVX512(v2, tnext, tmp1, tmp2, rk1, rk2) \ + vpbroadcastd rk1, tmp1; \ + vpbroadcastd rk2, tmp2; \ + vprold $3, v2, v2; \ + vpxord tnext, tmp1, tmp1; \ + vpsubd tmp1, v2, tnext; \ + vpxord tmp2, tnext, v2 +#endif + +#define XAR3(SIMD, cur, pre, tmp1, tmp2, rk1, rk2) \ + PASTE1(XAR_,SIMD)($3, $29, cur, pre, tmp1, tmp2, rk1, rk2) + +#define XAR5(SIMD, cur, pre, tmp1, tmp2, rk1, rk2) \ + PASTE1(XAR_,SIMD)($5, $27, cur, pre, tmp1, tmp2, rk1, rk2) + +#define XAR9(SIMD, cur, pre, tmp1, tmp2, rk1, rk2) \ + PASTE1(XAR_,SIMD)($23, $9, cur, pre, tmp1, tmp2, rk1, rk2) + +#define XSR9(SIMD, v0, v3, tnext, tmp1, tmp2, rk1, rk2) \ + PASTE1(XSR9_,SIMD)(v0, v3, tnext, tmp1, tmp2, rk1, rk2) + +#define XSR5(SIMD, v1, tnext, tmp1, tmp2, rk1, rk2) \ + PASTE1(XSR5_,SIMD)(v1, tnext, tmp1, tmp2, rk1, rk2) + +#define XSR3(SIMD, v2, tnext, tmp1, tmp2, rk1, rk2) \ + PASTE1(XSR3_,SIMD)(v2, tnext, tmp1, tmp2, rk1, rk2) + +/* Could be replaced by the VPGATHERDD command, + * but was not used because it is known to have a non-optimized architecture. +*/ +#define GATHER4_AVX2(i, j, mem) \ + vmovd (j + 0 * 16)(mem), XMM(i); \ + vpinsrd $0x1, (j + 1 * 16)(mem), XMM(i), XMM(i);\ + vpinsrd $0x2, (j + 2 * 16)(mem), XMM(i), XMM(i);\ + vpinsrd $0x3, (j + 3 * 16)(mem), XMM(i), XMM(i); + +#define GATHER8_AVX2(i, ti, j, mem) \ + vmovd (j + 0 * 16)(mem), XMM(i); \ + vpinsrd $0x1, (j + 1 * 16)(mem), XMM(i), XMM(i); \ + vpinsrd $0x2, (j + 2 * 16)(mem), XMM(i), XMM(i); \ + vpinsrd $0x3, (j + 3 * 16)(mem), XMM(i), XMM(i); \ + vmovd (j + 4 * 16)(mem), XMM(ti); \ + vpinsrd $0x1, (j + 5 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x2, (j + 6 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x3, (j + 7 * 16)(mem), XMM(ti), XMM(ti); \ + vinserti128 $0x1, XMM(ti), YMM(i), YMM(i); + +#ifdef CONFIG_AS_AVX512 +#define GATHER16_AVX512(i, ti, j, mem) \ + vmovd (j + 0 * 16)(mem), XMM(i); \ + vpinsrd $0x1, (j + 1 * 16)(mem), XMM(i), XMM(i); \ + vpinsrd $0x2, (j + 2 * 16)(mem), XMM(i), XMM(i); \ + vpinsrd $0x3, (j + 3 * 16)(mem), XMM(i), XMM(i); \ + vmovd (j + 4 * 16)(mem), XMM(ti); \ + vpinsrd $0x1, (j + 5 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x2, (j + 6 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x3, (j + 7 * 16)(mem), XMM(ti), XMM(ti); \ + vinserti32x4 $0x1, XMM(ti), ZMM(i), ZMM(i); \ + vmovd (j + 8 * 16)(mem), XMM(ti); \ + vpinsrd $0x1, (j + 9 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x2, (j + 10 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x3, (j + 11 * 16)(mem), XMM(ti), XMM(ti); \ + vinserti32x4 $0x2, XMM(ti), ZMM(i), ZMM(i); \ + vmovd (j + 12 * 16)(mem), XMM(ti); \ + vpinsrd $0x1, (j + 13 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x2, (j + 14 * 16)(mem), XMM(ti), XMM(ti); \ + vpinsrd $0x3, (j + 15 * 16)(mem), XMM(ti), XMM(ti); \ + vinserti32x4 $0x3, XMM(ti), ZMM(i), ZMM(i); +#endif + +#define GATHER_BLOCK4_AVX2(i0, i1, i2, i3, mem) \ + GATHER4_AVX2(i0, 0, mem); \ + GATHER4_AVX2(i1, 4, mem); \ + GATHER4_AVX2(i2, 8, mem); \ + GATHER4_AVX2(i3, 12, mem); + +#define GATHER_BLOCK8_AVX2(i0, i1, i2, i3, ti0, mem) \ + GATHER8_AVX2(i0, ti0, 0, mem); \ + GATHER8_AVX2(i1, ti0, 4, mem); \ + GATHER8_AVX2(i2, ti0, 8, mem); \ + GATHER8_AVX2(i3, ti0, 12, mem); + +#ifdef CONFIG_AS_AVX512 +#define GATHER_BLOCK8_AVX512(i0, i1, i2, i3, mask, mem) \ + GATHER8_AVX512(i0, mask, 0, mem); \ + GATHER8_AVX512(i1, mask, 4, mem); \ + GATHER8_AVX512(i2, mask, 8, mem); \ + GATHER8_AVX512(i3, mask, 12, mem); + +#define GATHER_BLOCK16_AVX512(i0, i1, i2, i3, ti1, mem) \ + GATHER16_AVX512(i0, ti1, 0, mem); \ + GATHER16_AVX512(i1, ti1, 4, mem); \ + GATHER16_AVX512(i2, ti1, 8, mem); \ + GATHER16_AVX512(i3, ti1, 12, mem); +#endif + +#define SCATTER4_AVX2(i, j, mem) \ + vpextrd $0x0, XMM(i), (j + 0 * 16)(mem);\ + vpextrd $0x1, XMM(i), (j + 1 * 16)(mem);\ + vpextrd $0x2, XMM(i), (j + 2 * 16)(mem);\ + vpextrd $0x3, XMM(i), (j + 3 * 16)(mem); + +#define SCATTER8_AVX2(i, j, mem) \ + vpextrd $0x0, XMM(i), (j + 0 * 16)(mem);\ + vpextrd $0x1, XMM(i), (j + 1 * 16)(mem);\ + vpextrd $0x2, XMM(i), (j + 2 * 16)(mem);\ + vpextrd $0x3, XMM(i), (j + 3 * 16)(mem);\ + vextracti128 $0x1, YMM(i), XMM(i); \ + vpextrd $0x0, XMM(i), (j + 4 * 16)(mem);\ + vpextrd $0x1, XMM(i), (j + 5 * 16)(mem);\ + vpextrd $0x2, XMM(i), (j + 6 * 16)(mem);\ + vpextrd $0x3, XMM(i), (j + 7 * 16)(mem); + +#ifdef CONFIG_AS_AVX512 +/* Could be replaced by the VPSCATTERDD command */ +#define SCATTER16_AVX512(i, ti, j, mem) \ + vpextrd $0x0, XMM(i), (j + 0 * 16)(mem); \ + vpextrd $0x1, XMM(i), (j + 1 * 16)(mem); \ + vpextrd $0x2, XMM(i), (j + 2 * 16)(mem); \ + vpextrd $0x3, XMM(i), (j + 3 * 16)(mem); \ + vextracti32x4 $0x1, ZMM(i), XMM(ti); \ + vpextrd $0x0, XMM(ti), (j + 4 * 16)(mem); \ + vpextrd $0x1, XMM(ti), (j + 5 * 16)(mem); \ + vpextrd $0x2, XMM(ti), (j + 6 * 16)(mem); \ + vpextrd $0x3, XMM(ti), (j + 7 * 16)(mem); \ + vextracti32x4 $0x2, ZMM(i), XMM(ti); \ + vpextrd $0x0, XMM(ti), (j + 8 * 16)(mem); \ + vpextrd $0x1, XMM(ti), (j + 9 * 16)(mem); \ + vpextrd $0x2, XMM(ti), (j + 10 * 16)(mem); \ + vpextrd $0x3, XMM(ti), (j + 11 * 16)(mem); \ + vextracti32x4 $0x3, ZMM(i), XMM(ti); \ + vpextrd $0x0, XMM(ti), (j + 12 * 16)(mem); \ + vpextrd $0x1, XMM(ti), (j + 13 * 16)(mem); \ + vpextrd $0x2, XMM(ti), (j + 14 * 16)(mem); \ + vpextrd $0x3, XMM(ti), (j + 15 * 16)(mem); +#endif + +#define SCATTER_BLOCK4_AVX2(i0, i1, i2, i3, mem)\ + SCATTER4_AVX2(i0, 0, mem); \ + SCATTER4_AVX2(i1, 4, mem); \ + SCATTER4_AVX2(i2, 8, mem); \ + SCATTER4_AVX2(i3, 12, mem); + +#define SCATTER_BLOCK8_AVX2(i0, i1, i2, i3, mem)\ + SCATTER8_AVX2(i0, 0, mem); \ + SCATTER8_AVX2(i1, 4, mem); \ + SCATTER8_AVX2(i2, 8, mem); \ + SCATTER8_AVX2(i3, 12, mem); + +#ifdef CONFIG_AS_AVX512 +#define SCATTER_BLOCK16_AVX512(i0, i1, i2, i3, ti, mem) \ + SCATTER16_AVX512(i0, ti, 0, mem);\ + SCATTER16_AVX512(i1, ti, 4, mem);\ + SCATTER16_AVX512(i2, ti, 8, mem);\ + SCATTER16_AVX512(i3, ti, 12, mem); +#endif + +#define LOAD_BLOCK4_SSE2(x0, x1, x2, x3, mem) \ + movdqu 0 * 16(mem), x0; \ + movdqu 1 * 16(mem), x1; \ + movdqu 2 * 16(mem), x2; \ + movdqu 3 * 16(mem), x3; + +#define LOAD_BLOCK4_AVX2(x0, x1, x2, x3, mem) \ + vmovdqu 0 * 16(mem), x0; \ + vmovdqu 1 * 16(mem), x1; \ + vmovdqu 2 * 16(mem), x2; \ + vmovdqu 3 * 16(mem), x3; + +#define STORE_BLOCK4_SSE2(x0, x1, x2, x3, mem) \ + movdqu x0, 0 * 16(mem); \ + movdqu x1, 1 * 16(mem); \ + movdqu x2, 2 * 16(mem); \ + movdqu x3, 3 * 16(mem); + +#define STORE_BLOCK4_AVX2(x0, x1, x2, x3, mem) \ + vmovdqu x0, 0 * 16(mem); \ + vmovdqu x1, 1 * 16(mem); \ + vmovdqu x2, 2 * 16(mem); \ + vmovdqu x3, 3 * 16(mem); + +#define TRANSPOSE_BLOCK4_SSE2(x0, x1, out_x2, x3, tmp, in_x2) \ + movdqa x0, out_x2; \ + movdqa in_x2, tmp; \ + punpckldq x1, x0; \ + punpckhdq x1, out_x2; \ + punpckldq x3, tmp; \ + punpckhdq x3, in_x2; \ + \ + movdqa x0, x1; \ + movdqa out_x2, x3; \ + punpcklqdq tmp, x0; \ + punpckhqdq tmp, x1; \ + punpcklqdq in_x2, out_x2; \ + punpckhqdq in_x2, x3; + +#define TRANSPOSE_BLOCK4_AVX2(x0, x1, out_x2, x3, tmp, in_x2) \ + vpunpckhdq x1, x0, out_x2; \ + vpunpckldq x1, x0, x0; \ + vpunpckldq x3, in_x2, tmp; \ + vpunpckhdq x3, in_x2, in_x2; \ + \ + vpunpckhqdq tmp, x0, x1; \ + vpunpcklqdq tmp, x0, x0; \ + vpunpckhqdq in_x2, out_x2, x3; \ + vpunpcklqdq in_x2, out_x2, out_x2; + +#define XOR_BLOCK3_SSE2(x0, x1, x2, mem)\ + pxor 0 * 16(mem), x0; \ + pxor 1 * 16(mem), x1; \ + pxor 2 * 16(mem), x2; + +#define XOR_BLOCK3_AVX2(x0, x1, x2, mem)\ + vpxor 0 * 16(mem), x0, x0; \ + vpxor 1 * 16(mem), x1, x1; \ + vpxor 2 * 16(mem), x2, x2; + +#define GEN_CTR_BLOCK_AVX2(x0, x1, x2, x3, xtmp, xcarry, bswapmask, ctrconst, ctr) \ + vpbroadcastd (3 * 4)(ctr), x3; \ + vpbroadcastd (2 * 4)(ctr), x2; \ + vpbroadcastd (1 * 4)(ctr), x1; \ + vpbroadcastd (0 * 4)(ctr), x0; \ + vpshufb bswapmask, x3, x3; \ + vpshufb bswapmask, x2, x2; \ + vpshufb bswapmask, x1, x1; \ + vpshufb bswapmask, x0, x0; \ + \ + vpaddd ctrconst, x3, xtmp; \ + vpandn x3, xtmp, xcarry; \ + vpsrld $31, xcarry, xcarry; \ + vpshufb bswapmask, xtmp, x3; \ + \ + vpaddd xcarry, x2, xtmp; \ + vpandn x2, xtmp, xcarry; \ + vpsrld $31, xcarry, xcarry; \ + vpshufb bswapmask, xtmp, x2; \ + \ + vpaddd xcarry, x1, xtmp; \ + vpandn x1, xtmp, xcarry; \ + vpsrld $31, xcarry, xcarry; \ + vpshufb bswapmask, xtmp, x1; \ + \ + vpaddd xcarry, x0, x0; \ + vpshufb bswapmask, x0, x0; + +#ifdef CONFIG_AS_AVX512 +#define GEN_CTR_BLOCK_AVX512(x0, x1, x2, x3, xtmp, xcarry, bswapmask, ctrconst, ctr) \ + vpbroadcastd (3 * 4)(ctr), x3; \ + vpbroadcastd (2 * 4)(ctr), x2; \ + vpbroadcastd (1 * 4)(ctr), x1; \ + vpbroadcastd (0 * 4)(ctr), x0; \ + vpshufb bswapmask, x3, x3; \ + vpshufb bswapmask, x2, x2; \ + vpshufb bswapmask, x1, x1; \ + vpshufb bswapmask, x0, x0; \ + \ + vpaddd ctrconst, x3, xtmp; \ + vpandnd x3, xtmp, xcarry; \ + vpsrld $0x1f, xcarry, xcarry; \ + vpshufb bswapmask, xtmp, x3; \ + \ + vpaddd xcarry, x2, xtmp; \ + vpandnd x2, xtmp, xcarry; \ + vpsrld $0x1f, xcarry, xcarry; \ + vpshufb bswapmask, xtmp, x2; \ + \ + vpaddd xcarry, x1, xtmp; \ + vpandnd x1, xtmp, xcarry; \ + vpsrld $0x1f, xcarry, xcarry; \ + vpshufb bswapmask, xtmp, x1; \ + \ + vpaddd xcarry, x0, x0; \ + vpshufb bswapmask, x0, x0; +#endif + +#define LEA_1ROUND_ENC(SIMD, i0, i1, i2, i3, tmp1, tmp2, rk, rnd_num) \ + XAR3(SIMD, i3, i2, tmp1, tmp2, \ + (LEA_CTX_RK_ENC + ((rnd_num) * 6 + 4) * 4)(rk), \ + (LEA_CTX_RK_ENC + ((rnd_num) * 6 + 5) * 4)(rk)); \ + XAR5(SIMD, i2, i1, tmp1, tmp2, \ + (LEA_CTX_RK_ENC + ((rnd_num) * 6 + 2) * 4)(rk), \ + (LEA_CTX_RK_ENC + ((rnd_num) * 6 + 3) * 4)(rk)); \ + XAR9(SIMD, i1, i0, tmp1, tmp2, \ + (LEA_CTX_RK_ENC + ((rnd_num) * 6 + 0) * 4)(rk), \ + (LEA_CTX_RK_ENC + ((rnd_num) * 6 + 1) * 4)(rk)) + +#define LEA_4ROUND_ENC(SIMD, i0, i1, i2, i3, tmp1, tmp2, rk, rnd_num) \ + LEA_1ROUND_ENC(SIMD, i0, i1, i2, i3, tmp1, tmp2, rk, rnd_num + 0); \ + LEA_1ROUND_ENC(SIMD, i1, i2, i3, i0, tmp1, tmp2, rk, rnd_num + 1); \ + LEA_1ROUND_ENC(SIMD, i2, i3, i0, i1, tmp1, tmp2, rk, rnd_num + 2); \ + LEA_1ROUND_ENC(SIMD, i3, i0, i1, i2, tmp1, tmp2, rk, rnd_num + 3) + +#define LEA_1ROUND_DEC(SIMD, i0, i1, i2, i3, tnext, tmp1, tmp2, rk, rnd_num) \ + XSR9(SIMD, i0, i3, tnext, tmp1, tmp2, \ + (LEA_CTX_RK_DEC + ((rnd_num) * 6 + 0) * 4)(rk), \ + (LEA_CTX_RK_DEC + ((rnd_num) * 6 + 1) * 4)(rk)); \ + XSR5(SIMD, i1, tnext, tmp1, tmp2, \ + (LEA_CTX_RK_DEC + ((rnd_num) * 6 + 2) * 4)(rk), \ + (LEA_CTX_RK_DEC + ((rnd_num) * 6 + 3) * 4)(rk)); \ + XSR3(SIMD, i2, tnext, tmp1, tmp2, \ + (LEA_CTX_RK_DEC + ((rnd_num) * 6 + 4) * 4)(rk), \ + (LEA_CTX_RK_DEC + ((rnd_num) * 6 + 5) * 4)(rk)); + +#define LEA_4ROUND_DEC(SIMD, i0, i1, i2, i3, tnext, tmp1, tmp2, rk, rnd_num) \ + LEA_1ROUND_DEC(SIMD, i0, i1, i2, i3, tnext, tmp1, tmp2, rk, rnd_num + 0);\ + LEA_1ROUND_DEC(SIMD, i3, i0, i1, i2, tnext, tmp1, tmp2, rk, rnd_num + 1);\ + LEA_1ROUND_DEC(SIMD, i2, i3, i0, i1, tnext, tmp1, tmp2, rk, rnd_num + 2);\ + LEA_1ROUND_DEC(SIMD, i1, i2, i3, i0, tnext, tmp1, tmp2, rk, rnd_num + 3); + + +#define XTS_LOAD_TWEAK_MASK_SSE2(mask) \ + movdqa .Lxts_tweak_mask(%rip), mask + +#define XTS_LOAD_TWEAK_MASK_AVX2(mask) \ + vmovdqa .Lxts_tweak_mask(%rip), mask + +#define CBC_LOAD_SHUFFLE_MASK_AVX2(mask) \ + vmovdqa .Lcbc_shuffle_mask(%rip), mask + +#define LOAD_BSWAP_MASK_AVX2(mask) \ + vmovdqa .Lbswap_mask(%rip), mask + +#define LOAD_CTR_CONST_AVX2(val) \ + vmovdqa .Lctr_add_const(%rip), val + +#ifdef CONFIG_AS_AVX512 + +#define CBC_LOAD_SHUFFLE_MASK_AVX512(mask) \ + vmovdqa64 .Lcbc_shuffle_mask16(%rip), mask + +#define LOAD_BSWAP_MASK_AVX512(mask) \ + vmovdqa64 .Lbswap_mask(%rip), mask + +#define LOAD_CTR_CONST_AVX512(val) \ + vmovdqa64 .Lctr_add_const(%rip), val + +#endif + +#define XTS_NEXT_TWEAK_1BLOCK_SSE2(out0, in0, tmp0, mask) \ + pshufd $0x13, in0, tmp0; \ + movdqa in0, out0; \ + psrad $31, tmp0; \ + pand mask, tmp0; \ + psllq $1, out0; \ + pxor tmp0, out0; + +#define XTS_NEXT_TWEAK_1BLOCK_AVX2(out0, in0, tmp0, mask) \ + vpshufd $0x13, in0, tmp0; \ + vpsrad $31, tmp0, tmp0; \ + vpand mask, tmp0, tmp0; \ + vpsllq $1, in0, out0; \ + vpxor tmp0, out0, out0; + +.align 8 +SYM_FUNC_START_LOCAL(__lea_x86_64_enc_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %xmm0..%xmm3: 4 plaintext blocks + * output: + * %xmm0..%xmm3: 4 encrypted blocks + * temporary: + * $xmm4..%xmm5 + */ + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 0); + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 4); + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 8); + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 12); + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 16); + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Lenc4_sse2_done; + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Lenc4_sse2_done; + LEA_4ROUND_ENC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 28); + +.Lenc4_sse2_done: + RET; +SYM_FUNC_END(__lea_x86_64_enc_4way_sse2) + +.align 8 +SYM_FUNC_START_LOCAL(__lea_x86_64_dec_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %xmm0..%xmm3: 4 encrypted blocks + * output: + * %xmm0..%xmm3: 4 plaintext blocks + * temporary: + * $xmm4..%xmm6 + */ + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 0); + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 4); + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 8); + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 12); + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 16); + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Ldec4_sse2_done; + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Ldec4_sse2_done; + LEA_4ROUND_DEC(SSE2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 28); + +.Ldec4_sse2_done: + RET; +SYM_FUNC_END(__lea_x86_64_dec_4way_sse2) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_enc_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %xmm0..%xmm3: 4 plaintext blocks + * output: + * %xmm0..%xmm3: 4 encrypted blocks + * temporary: + * $xmm4..%xmm5 + */ + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 0); + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 4); + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 8); + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 12); + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 16); + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Lenc4_avx2_done; + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Lenc4_avx2_done; + LEA_4ROUND_ENC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 28); + +.Lenc4_avx2_done: + RET; +SYM_FUNC_END(__lea_x86_64_enc_4way_avx2) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_dec_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %xmm0..%xmm3: 4 encrypted blocks + * output: + * %xmm0..%xmm3: 4 plaintext blocks + * temporary: + * $xmm4..%xmm6 + */ + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 0); + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 4); + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 8); + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 12); + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 16); + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Ldec4_avx2_done; + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Ldec4_avx2_done; + LEA_4ROUND_DEC(AVX2, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 28); +.Ldec4_avx2_done: + RET; +SYM_FUNC_END(__lea_x86_64_dec_4way_avx2) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_enc_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %ymm0..%ymm3: 8 plaintext blocks + * output: + * %ymm0..%ymm3: 8 encrypted blocks + * temporary: + * $ymm4..%ymm5 + */ + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 0); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 4); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 8); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 12); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 16); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Lenc8_avx2_done; + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Lenc8_avx2_done; + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 28); + +.Lenc8_avx2_done: + RET; +SYM_FUNC_END(__lea_x86_64_enc_8way_avx2) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_dec_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %ymm0..%ymm3: 8 encrypted blocks + * output: + * %ymm0..%ymm3: 8 plaintext blocks + * temporary: + * $ymm4..%ymm6 + */ + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 0); + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 4); + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 8); + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 12); + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 16); + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Ldec8_avx2_done; + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Ldec8_avx2_done; + LEA_4ROUND_DEC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 28); +.Ldec8_avx2_done: + RET; +SYM_FUNC_END(__lea_x86_64_dec_8way_avx2) + +#ifdef CONFIG_AS_AVX512 +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_enc_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %xmm0..%xmm3: 4 plaintext blocks + * output: + * %xmm0..%xmm3: 4 encrypted blocks + * temporary: + * $xmm4..%xmm5 + */ + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 0); + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 4); + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 8); + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 12); + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 16); + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Lenc4_avx512_done; + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Lenc4_avx512_done; + LEA_4ROUND_ENC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %rdi, 28); + +.Lenc4_avx512_done: + RET; +SYM_FUNC_END(__lea_x86_64_enc_4way_avx512) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_dec_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %xmm0..%xmm3: 4 encrypted blocks + * output: + * %xmm0..%xmm3: 4 plaintext blocks + * temporary: + * $xmm4..%xmm6 + */ + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 0); + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 4); + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 8); + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 12); + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 16); + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Ldec4_avx512_done; + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Ldec4_avx512_done; + LEA_4ROUND_DEC(AVX512, %xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, %xmm6, %rdi, 28); + +.Ldec4_avx512_done: + RET; +SYM_FUNC_END(__lea_x86_64_dec_4way_avx512) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_enc_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %ymm0..%ymm3: 8 plaintext blocks + * output: + * %ymm0..%ymm3: 8 encrypted blocks + * temporary: + * $ymm4..%ymm5 + */ + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 0); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 4); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 8); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 12); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 16); + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Lenc8_avx512_done; + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Lenc8_avx512_done; + LEA_4ROUND_ENC(AVX2, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %rdi, 28); + +.Lenc8_avx512_done: + RET; +SYM_FUNC_END(__lea_x86_64_enc_8way_avx512) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_dec_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %ymm0..%ymm3: 8 encrypted blocks + * output: + * %ymm0..%ymm3: 8 plaintext blocks + * temporary: + * $ymm4..%ymm6 + */ + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 0); + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 4); + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 8); + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 12); + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 16); + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Ldec8_avx512_done; + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Ldec8_avx512_done; + LEA_4ROUND_DEC(AVX512, %ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, %ymm6, %rdi, 28); +.Ldec8_avx512_done: + RET; +SYM_FUNC_END(__lea_x86_64_dec_8way_avx512) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_enc_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %zmm0..%zmm3: 16 plaintext blocks + * output: + * %zmm0..%zmm3: 16 encrypted blocks + * temporary: + * $zmm4..%zmm5 + */ + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 0); + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 4); + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 8); + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 12); + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 16); + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Lenc16_avx512_done; + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Lenc16_avx512_done; + LEA_4ROUND_ENC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %rdi, 28); + +.Lenc16_avx512_done: + RET; +SYM_FUNC_END(__lea_x86_64_enc_16way_avx512) + +.align 16 +SYM_FUNC_START_LOCAL(__lea_x86_64_dec_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %zmm0..%zmm3: 16 encrypted blocks + * output: + * %zmm0..%zmm3: 16 plaintext blocks + * temporary: + * $zmm4..%zmm6 + */ + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 0); + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 4); + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 8); + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 12); + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 16); + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 20); + + cmpl $24, LEA_CTX_ROUND(%rdi); + je .Ldec16_avx512_done; + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 24); + + cmpl $28, LEA_CTX_ROUND(%rdi); + je .Ldec16_avx512_done; + LEA_4ROUND_DEC(AVX512, %zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, %zmm6, %rdi, 28); +.Ldec16_avx512_done: + RET; +SYM_FUNC_END(__lea_x86_64_dec_16way_avx512) +#endif + +SYM_FUNC_START(lea_x86_64_ecb_enc_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + LOAD_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_enc_4way_sse2 + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + STORE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_enc_4way_sse2) + +SYM_FUNC_START(lea_x86_64_ecb_enc_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + GATHER_BLOCK4_AVX2(0, 1, 2, 3, %rdx); + + call __lea_x86_64_enc_4way_avx2 + + SCATTER_BLOCK4_AVX2(0, 1, 2, 3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_enc_4way_avx2) + +SYM_FUNC_START(lea_x86_64_ecb_enc_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + call __lea_x86_64_enc_8way_avx2; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_enc_8way_avx2) + +#ifdef CONFIG_AS_AVX512 +SYM_FUNC_START(lea_x86_64_ecb_enc_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + GATHER_BLOCK4_AVX2(0, 1, 2, 3, %rdx); + + call __lea_x86_64_enc_4way_avx512 + + SCATTER_BLOCK4_AVX2(0, 1, 2, 3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_enc_4way_avx512) + +SYM_FUNC_START(lea_x86_64_ecb_enc_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + call __lea_x86_64_enc_8way_avx512; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_enc_8way_avx512) + +SYM_FUNC_START(lea_x86_64_ecb_enc_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (16 blocks) + * %rdx: src (16 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rdx); + + call __lea_x86_64_enc_16way_avx512; + + SCATTER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_enc_16way_avx512) +#endif + +SYM_FUNC_START(lea_x86_64_ecb_dec_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + LOAD_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_dec_4way_sse2 + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + STORE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_dec_4way_sse2) + +SYM_FUNC_START(lea_x86_64_ecb_dec_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + GATHER_BLOCK4_AVX2(0, 1, 2, 3, %rdx); + + call __lea_x86_64_dec_4way_avx2 + + SCATTER_BLOCK4_AVX2(0, 1, 2, 3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_dec_4way_avx2) + +SYM_FUNC_START(lea_x86_64_ecb_dec_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + call __lea_x86_64_dec_8way_avx2 + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_dec_8way_avx2) + +#ifdef CONFIG_AS_AVX512 +SYM_FUNC_START(lea_x86_64_ecb_dec_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + GATHER_BLOCK4_AVX2(0, 1, 2, 3, %rdx); + + call __lea_x86_64_dec_4way_avx512 + + SCATTER_BLOCK4_AVX2(0, 1, 2, 3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_dec_4way_avx512) + +SYM_FUNC_START(lea_x86_64_ecb_dec_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + call __lea_x86_64_dec_8way_avx512 + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_dec_8way_avx512) + +SYM_FUNC_START(lea_x86_64_ecb_dec_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (16 blocks) + * %rdx: src (16 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rdx); + + call __lea_x86_64_dec_16way_avx512 + + SCATTER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ecb_dec_16way_avx512) +#endif + +SYM_FUNC_START(lea_x86_64_cbc_dec_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + LOAD_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_dec_4way_sse2 + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + XOR_BLOCK3_SSE2(%xmm1, %xmm5, %xmm3, %rdx); + STORE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_cbc_dec_4way_sse2) + +SYM_FUNC_START(lea_x86_64_cbc_dec_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + GATHER_BLOCK4_AVX2(0, 1, 2, 3, %rdx); + + call __lea_x86_64_dec_4way_avx2 + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + XOR_BLOCK3_AVX2(%xmm1, %xmm5, %xmm3, %rdx); + STORE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_cbc_dec_4way_avx2) + +SYM_FUNC_START(lea_x86_64_cbc_dec_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + CBC_LOAD_SHUFFLE_MASK_AVX2(%ymm7); + vpxor %ymm4, %ymm4, %ymm4; + + vpermd %ymm0, %ymm7, %ymm8; + vpermd %ymm1, %ymm7, %ymm9; + vpermd %ymm2, %ymm7, %ymm10; + vpermd %ymm3, %ymm7, %ymm11; + + vpblendd $0x1, %ymm4, %ymm8, %ymm8; + vpblendd $0x1, %ymm4, %ymm9, %ymm9; + vpblendd $0x1, %ymm4, %ymm10, %ymm10; + vpblendd $0x1, %ymm4, %ymm11, %ymm11; + + call __lea_x86_64_dec_8way_avx2 + + vpxor %ymm8, %ymm0, %ymm0; + vpxor %ymm9, %ymm1, %ymm1; + vpxor %ymm10, %ymm2, %ymm2; + vpxor %ymm11, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_cbc_dec_8way_avx2) + +#ifdef CONFIG_AS_AVX512 +SYM_FUNC_START(lea_x86_64_cbc_dec_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + */ + FRAME_BEGIN + + GATHER_BLOCK4_AVX2(0, 1, 2, 3, %rdx); + + call __lea_x86_64_dec_4way_avx512 + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + XOR_BLOCK3_AVX2(%xmm1, %xmm5, %xmm3, %rdx); + STORE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_cbc_dec_4way_avx512) + +SYM_FUNC_START(lea_x86_64_cbc_dec_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + CBC_LOAD_SHUFFLE_MASK_AVX2(%ymm7); + vpxor %ymm4, %ymm4, %ymm4; + + vpermd %ymm0, %ymm7, %ymm8; + vpermd %ymm1, %ymm7, %ymm9; + vpermd %ymm2, %ymm7, %ymm10; + vpermd %ymm3, %ymm7, %ymm11; + + vpblendd $0x1, %ymm4, %ymm8, %ymm8; + vpblendd $0x1, %ymm4, %ymm9, %ymm9; + vpblendd $0x1, %ymm4, %ymm10, %ymm10; + vpblendd $0x1, %ymm4, %ymm11, %ymm11; + + call __lea_x86_64_dec_8way_avx512 + + vpxor %ymm8, %ymm0, %ymm0; + vpxor %ymm9, %ymm1, %ymm1; + vpxor %ymm10, %ymm2, %ymm2; + vpxor %ymm11, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_cbc_dec_8way_avx512) + +SYM_FUNC_START(lea_x86_64_cbc_dec_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (16 blocks) + * %rdx: src (16 blocks) + */ + FRAME_BEGIN + + vzeroupper; + + GATHER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rdx); + + CBC_LOAD_SHUFFLE_MASK_AVX512(%zmm7); + vpxord %zmm4, %zmm4, %zmm4; + + mov $0x1, %eax; + kmovw %eax, %k1; + + vpermd %zmm0, %zmm7, %zmm8; + vpermd %zmm1, %zmm7, %zmm9; + vpermd %zmm2, %zmm7, %zmm10; + vpermd %zmm3, %zmm7, %zmm11; + + vpblendmd %zmm4, %zmm8, %zmm8{%k1}; + vpblendmd %zmm4, %zmm9, %zmm9{%k1}; + vpblendmd %zmm4, %zmm10, %zmm10{%k1}; + vpblendmd %zmm4, %zmm11, %zmm11{%k1}; + + call __lea_x86_64_dec_16way_avx512 + + vpxord %zmm8, %zmm0, %zmm0; + vpxord %zmm9, %zmm1, %zmm1; + vpxord %zmm10, %zmm2, %zmm2; + vpxord %zmm11, %zmm3, %zmm3; + + SCATTER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_cbc_dec_16way_avx512) +#endif + +SYM_FUNC_START(lea_x86_64_ctr_enc_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: ctr + * temporary: + * CTR_64_high_src(%r9) + * CTR_64_low_src(%rax) + * CTR_64_high(%r11) + * CTR_64_low(%r10) + */ + FRAME_BEGIN + + mov (%rcx), CTR_64_high_src; + mov 8(%rcx), CTR_64_low_src; + bswap CTR_64_high_src; + bswap CTR_64_low_src; + + movdqu (%rcx), %xmm0; + mov CTR_64_low_src, CTR_64_low; + mov CTR_64_high_src, CTR_64_high; + ADD_CTR_AND_STORE(CTR_64_high, CTR_64_low, %rcx, $1); + movdqu (%rcx), %xmm1; + mov CTR_64_low_src, CTR_64_low; + mov CTR_64_high_src, CTR_64_high; + ADD_CTR_AND_STORE(CTR_64_high, CTR_64_low, %rcx, $2); + movdqu (%rcx), %xmm5; + mov CTR_64_low_src, CTR_64_low; + mov CTR_64_high_src, CTR_64_high; + ADD_CTR_AND_STORE(CTR_64_high, CTR_64_low, %rcx, $3); + movdqu (%rcx), %xmm3; + mov CTR_64_low_src, CTR_64_low; + mov CTR_64_high_src, CTR_64_high; + ADD_CTR_AND_STORE(CTR_64_high, CTR_64_low, %rcx, $4); + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + LOAD_BLOCK4_SSE2(%xmm7, %xmm8, %xmm9, %xmm10, %rdx); + + call __lea_x86_64_enc_4way_sse2; + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + + pxor %xmm7, %xmm0; + pxor %xmm8, %xmm1; + pxor %xmm9, %xmm5; + pxor %xmm10, %xmm3; + + STORE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ctr_enc_4way_sse2) + +SYM_FUNC_START(lea_x86_64_ctr_enc_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: ctr + * temporary: + * %r9 + * %rax + */ + FRAME_BEGIN + + LOAD_BSWAP_MASK_AVX2(BSWAPMASK_XMM); + LOAD_CTR_CONST_AVX2(CTRCONST_XMM); + + GEN_CTR_BLOCK_AVX2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, + BSWAPMASK_XMM, CTRCONST_XMM, %rcx); + + LOAD_CTR(%rcx, %r9, %rax); + ADD_CTR_AND_STORE(%r9, %rax, %rcx, $4); + + GATHER_BLOCK4_AVX2(7, 8, 9, 10, %rdx); + + call __lea_x86_64_enc_4way_avx2; + + vpxor %ymm7, %ymm0, %ymm0; + vpxor %ymm8, %ymm1, %ymm1; + vpxor %ymm9, %ymm2, %ymm2; + vpxor %ymm10, %ymm3, %ymm3; + + SCATTER_BLOCK4_AVX2(0, 1, 2, 3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ctr_enc_4way_avx2) + +SYM_FUNC_START(lea_x86_64_ctr_enc_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + * %rcx: ctr + * temporary: + * %r9 + * %rax + */ + FRAME_BEGIN + + vzeroupper; + + + LOAD_BSWAP_MASK_AVX2(BSWAPMASK_YMM); + LOAD_CTR_CONST_AVX2(CTRCONST_YMM); + + GEN_CTR_BLOCK_AVX2(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, + BSWAPMASK_YMM, CTRCONST_YMM, %rcx); + LOAD_CTR(%rcx, %r9, %rax); + ADD_CTR_AND_STORE(%r9, %rax, %rcx, $8); + + GATHER_BLOCK8_AVX2(7, 8, 9, 10, 4, %rdx); + + call __lea_x86_64_enc_8way_avx2; + + vpxor %ymm7, %ymm0, %ymm0; + vpxor %ymm8, %ymm1, %ymm1; + vpxor %ymm9, %ymm2, %ymm2; + vpxor %ymm10, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ctr_enc_8way_avx2) + +#ifdef CONFIG_AS_AVX512 +SYM_FUNC_START(lea_x86_64_ctr_enc_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: ctr + * temporary: + * %r9 + * %rax + */ + FRAME_BEGIN + + LOAD_BSWAP_MASK_AVX2(BSWAPMASK_XMM); + LOAD_CTR_CONST_AVX2(CTRCONST_XMM); + + GEN_CTR_BLOCK_AVX2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5, + BSWAPMASK_XMM, CTRCONST_XMM, %rcx); + + LOAD_CTR(%rcx, %r9, %rax); + ADD_CTR_AND_STORE(%r9, %rax, %rcx, $4); + + GATHER_BLOCK4_AVX2(7, 8, 9, 10, %rdx); + + call __lea_x86_64_enc_4way_avx512; + + vpxor %ymm7, %ymm0, %ymm0; + vpxor %ymm8, %ymm1, %ymm1; + vpxor %ymm9, %ymm2, %ymm2; + vpxor %ymm10, %ymm3, %ymm3; + + SCATTER_BLOCK4_AVX2(0, 1, 2, 3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ctr_enc_4way_avx512) + +SYM_FUNC_START(lea_x86_64_ctr_enc_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + * %rcx: ctr + * temporary: + * %r9 + * %rax + */ + FRAME_BEGIN + + vzeroupper; + + + LOAD_BSWAP_MASK_AVX2(BSWAPMASK_YMM); + LOAD_CTR_CONST_AVX2(CTRCONST_YMM); + + GEN_CTR_BLOCK_AVX2(%ymm0, %ymm1, %ymm2, %ymm3, %ymm4, %ymm5, + BSWAPMASK_YMM, CTRCONST_YMM, %rcx); + LOAD_CTR(%rcx, %r9, %rax); + ADD_CTR_AND_STORE(%r9, %rax, %rcx, $8); + + GATHER_BLOCK8_AVX2(7, 8, 9, 10, 4, %rdx); + + call __lea_x86_64_enc_8way_avx512; + + vpxor %ymm7, %ymm0, %ymm0; + vpxor %ymm8, %ymm1, %ymm1; + vpxor %ymm9, %ymm2, %ymm2; + vpxor %ymm10, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ctr_enc_8way_avx512) + +SYM_FUNC_START(lea_x86_64_ctr_enc_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (16 blocks) + * %rdx: src (16 blocks) + * %rcx: ctr + * temporary: + * %r9 + * %rax + */ + FRAME_BEGIN + + vzeroupper; + + + LOAD_BSWAP_MASK_AVX512(BSWAPMASK_ZMM); + LOAD_CTR_CONST_AVX512(CTRCONST_ZMM); + + GEN_CTR_BLOCK_AVX512(%zmm0, %zmm1, %zmm2, %zmm3, %zmm4, %zmm5, + BSWAPMASK_ZMM, CTRCONST_ZMM, %rcx); + LOAD_CTR(%rcx, %r9, %rax); + ADD_CTR_AND_STORE(%r9, %rax, %rcx, $16); + + GATHER_BLOCK16_AVX512(8, 9, 10, 11, 4, %rdx); + + call __lea_x86_64_enc_16way_avx512; + + vpxord %zmm8, %zmm0, %zmm0; + vpxord %zmm9, %zmm1, %zmm1; + vpxord %zmm10, %zmm2, %zmm2; + vpxord %zmm11, %zmm3, %zmm3; + + SCATTER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_ctr_enc_16way_avx512) +#endif + +SYM_FUNC_START(lea_x86_64_xts_enc_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + LOAD_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + movdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_SSE2(XTS_MASK); + pxor XTS_TW_X0, %xmm0; + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_X1, XTS_TW_X0, %xmm4, XTS_MASK); + pxor XTS_TW_X1, %xmm1; + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_O2, XTS_TW_X1, %xmm4, XTS_MASK); + pxor XTS_TW_O2, %xmm5; + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_X3, XTS_TW_O2, %xmm4, XTS_MASK); + pxor XTS_TW_X3, %xmm3; + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_enc_4way_sse2; + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + + pxor XTS_TW_X0, %xmm0; + pxor XTS_TW_X1, %xmm1; + pxor XTS_TW_O2, %xmm5; + pxor XTS_TW_X3, %xmm3; + + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_X0, XTS_TW_X3, %xmm4, XTS_MASK); + movdqu XTS_TW_X0, (%rcx); + STORE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_enc_4way_sse2) + +SYM_FUNC_START(lea_x86_64_xts_enc_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + LOAD_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + vpxor XTS_TW_X0, %xmm0, %xmm0; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm4, XTS_MASK); + vpxor XTS_TW_X1, %xmm1, %xmm1; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_O2, XTS_TW_X1, %xmm4, XTS_MASK); + vpxor XTS_TW_O2, %xmm5, %xmm5; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_O2, %xmm4, XTS_MASK); + vpxor XTS_TW_X3, %xmm3, %xmm3; + + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_enc_4way_avx2 + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + + vpxor XTS_TW_X0, %xmm0, %xmm0; + vpxor XTS_TW_X1, %xmm1, %xmm1; + vpxor XTS_TW_O2, %xmm5, %xmm5; + vpxor XTS_TW_X3, %xmm3, %xmm3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X0, XTS_TW_X3, %xmm4, XTS_MASK); + vmovdqu XTS_TW_X0, (%rcx); + STORE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_enc_4way_avx2) + +SYM_FUNC_START(lea_x86_64_xts_enc_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + vzeroupper; + + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm6, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I2, XTS_TW_X1, %xmm6, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_I2, %xmm6, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X3, %xmm6, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm6, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm6, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm6, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm6, XTS_MASK); + vmovdqu XTS_TW_X8, (%rcx); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X0, XTS_TW_X1, XTS_TW_O2, XTS_TW_X3, %xmm5, XTS_TW_I2); + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti128 $0x1, XTS_TW_X4, XTS_TW_Y0, XTS_TW_Y0; + vinserti128 $0x1, XTS_TW_X5, XTS_TW_Y1, XTS_TW_Y1; + vinserti128 $0x1, XTS_TW_O6, XTS_TW_Y2, XTS_TW_Y2; + vinserti128 $0x1, XTS_TW_X7, XTS_TW_Y3, XTS_TW_Y3; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + call __lea_x86_64_enc_8way_avx2 + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_enc_8way_avx2) + +#ifdef CONFIG_AS_AVX512 +SYM_FUNC_START(lea_x86_64_xts_enc_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + LOAD_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + vpxor XTS_TW_X0, %xmm0, %xmm0; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm4, XTS_MASK); + vpxor XTS_TW_X1, %xmm1, %xmm1; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_O2, XTS_TW_X1, %xmm4, XTS_MASK); + vpxor XTS_TW_O2, %xmm5, %xmm5; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_O2, %xmm4, XTS_MASK); + vpxor XTS_TW_X3, %xmm3, %xmm3; + + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_enc_4way_avx512 + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + + vpxor XTS_TW_X0, %xmm0, %xmm0; + vpxor XTS_TW_X1, %xmm1, %xmm1; + vpxor XTS_TW_O2, %xmm5, %xmm5; + vpxor XTS_TW_X3, %xmm3, %xmm3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X0, XTS_TW_X3, %xmm4, XTS_MASK); + vmovdqu XTS_TW_X0, (%rcx); + STORE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_enc_4way_avx512) + +SYM_FUNC_START(lea_x86_64_xts_enc_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + vzeroupper; + + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I2, XTS_TW_X1, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_I2, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X3, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm5, XTS_MASK); + vmovdqu XTS_TW_X8, (%rcx); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X0, XTS_TW_X1, XTS_TW_O2, XTS_TW_X3, %xmm5, XTS_TW_I2); + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti128 $0x1, XTS_TW_X4, XTS_TW_Y0, XTS_TW_Y0; + vinserti128 $0x1, XTS_TW_X5, XTS_TW_Y1, XTS_TW_Y1; + vinserti128 $0x1, XTS_TW_O6, XTS_TW_Y2, XTS_TW_Y2; + vinserti128 $0x1, XTS_TW_X7, XTS_TW_Y3, XTS_TW_Y3; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + call __lea_x86_64_enc_8way_avx512 + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_enc_8way_avx512) + +SYM_FUNC_START(lea_x86_64_xts_enc_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (16 blocks) + * %rdx: src (16 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + vzeroupper; + + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I2, XTS_TW_X1, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_I2, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X3, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X0, XTS_TW_X1, XTS_TW_O2, XTS_TW_X3, %xmm5, XTS_TW_I2); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti32x4 $0x1, XTS_TW_X4, XTS_TW_Z0, XTS_TW_Z0; + vinserti32x4 $0x1, XTS_TW_X5, XTS_TW_Z1, XTS_TW_Z1; + vinserti32x4 $0x1, XTS_TW_O6, XTS_TW_Z2, XTS_TW_Z2; + vinserti32x4 $0x1, XTS_TW_X7, XTS_TW_Z3, XTS_TW_Z3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X8, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X7, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X8, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti32x4 $0x2, XTS_TW_X8, XTS_TW_Z0, XTS_TW_Z0; + vinserti32x4 $0x2, XTS_TW_X5, XTS_TW_Z1, XTS_TW_Z1; + vinserti32x4 $0x2, XTS_TW_O6, XTS_TW_Z2, XTS_TW_Z2; + vinserti32x4 $0x2, XTS_TW_X7, XTS_TW_Z3, XTS_TW_Z3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti32x4 $0x3, XTS_TW_X4, XTS_TW_Z0, XTS_TW_Z0; + vinserti32x4 $0x3, XTS_TW_X5, XTS_TW_Z1, XTS_TW_Z1; + vinserti32x4 $0x3, XTS_TW_O6, XTS_TW_Z2, XTS_TW_Z2; + vinserti32x4 $0x3, XTS_TW_X7, XTS_TW_Z3, XTS_TW_Z3; + + vmovdqu XTS_TW_X8, (%rcx); + + GATHER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rdx); + + vpxord XTS_TW_Z0, %zmm0, %zmm0; + vpxord XTS_TW_Z1, %zmm1, %zmm1; + vpxord XTS_TW_Z2, %zmm2, %zmm2; + vpxord XTS_TW_Z3, %zmm3, %zmm3; + + call __lea_x86_64_enc_16way_avx512 + + vpxord XTS_TW_Z0, %zmm0, %zmm0; + vpxord XTS_TW_Z1, %zmm1, %zmm1; + vpxord XTS_TW_Z2, %zmm2, %zmm2; + vpxord XTS_TW_Z3, %zmm3, %zmm3; + + SCATTER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_enc_16way_avx512) +#endif + +SYM_FUNC_START(lea_x86_64_xts_dec_4way_sse2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + LOAD_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + movdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_SSE2(XTS_MASK); + pxor XTS_TW_X0, %xmm0; + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_X1, XTS_TW_X0, %xmm4, XTS_MASK); + pxor XTS_TW_X1, %xmm1; + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_O2, XTS_TW_X1, %xmm4, XTS_MASK); + pxor XTS_TW_O2, %xmm5; + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_X3, XTS_TW_O2, %xmm4, XTS_MASK); + pxor XTS_TW_X3, %xmm3; + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_dec_4way_sse2 + + TRANSPOSE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + + pxor XTS_TW_X0, %xmm0; + pxor XTS_TW_X1, %xmm1; + pxor XTS_TW_O2, %xmm5; + pxor XTS_TW_X3, %xmm3; + + XTS_NEXT_TWEAK_1BLOCK_SSE2(XTS_TW_X0, XTS_TW_X3, %xmm4, XTS_MASK); + movdqu XTS_TW_X0, (%rcx); + STORE_BLOCK4_SSE2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_dec_4way_sse2) + +SYM_FUNC_START(lea_x86_64_xts_dec_4way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + LOAD_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + vpxor XTS_TW_X0, %xmm0, %xmm0; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm4, XTS_MASK); + vpxor XTS_TW_X1, %xmm1, %xmm1; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_O2, XTS_TW_X1, %xmm4, XTS_MASK); + vpxor XTS_TW_O2, %xmm5, %xmm5; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_O2, %xmm4, XTS_MASK); + vpxor XTS_TW_X3, %xmm3, %xmm3; + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_dec_4way_avx2 + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + + vpxor XTS_TW_X0, %xmm0, %xmm0; + vpxor XTS_TW_X1, %xmm1, %xmm1; + vpxor XTS_TW_O2, %xmm5, %xmm5; + vpxor XTS_TW_X3, %xmm3, %xmm3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X0, XTS_TW_X3, %xmm4, XTS_MASK); + vmovdqu XTS_TW_X0, (%rcx); + STORE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_dec_4way_avx2) + +SYM_FUNC_START(lea_x86_64_xts_dec_8way_avx2) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + vzeroupper; + + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I2, XTS_TW_X1, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_I2, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X3, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm5, XTS_MASK); + vmovdqu XTS_TW_X8, (%rcx); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X0, XTS_TW_X1, XTS_TW_O2, XTS_TW_X3, %xmm5, XTS_TW_I2); + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti128 $0x1, XTS_TW_X4, XTS_TW_Y0, XTS_TW_Y0; + vinserti128 $0x1, XTS_TW_X5, XTS_TW_Y1, XTS_TW_Y1; + vinserti128 $0x1, XTS_TW_O6, XTS_TW_Y2, XTS_TW_Y2; + vinserti128 $0x1, XTS_TW_X7, XTS_TW_Y3, XTS_TW_Y3; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + call __lea_x86_64_dec_8way_avx2 + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_dec_8way_avx2) + +#ifdef CONFIG_AS_AVX512 +SYM_FUNC_START(lea_x86_64_xts_dec_4way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (4 blocks) + * %rdx: src (4 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + LOAD_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rdx); + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + vpxor XTS_TW_X0, %xmm0, %xmm0; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm4, XTS_MASK); + vpxor XTS_TW_X1, %xmm1, %xmm1; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_O2, XTS_TW_X1, %xmm4, XTS_MASK); + vpxor XTS_TW_O2, %xmm5, %xmm5; + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_O2, %xmm4, XTS_MASK); + vpxor XTS_TW_X3, %xmm3, %xmm3; + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm2, %xmm3, %xmm4, %xmm5); + + call __lea_x86_64_dec_4way_avx512 + + TRANSPOSE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %xmm4, %xmm2); + + vpxor XTS_TW_X0, %xmm0, %xmm0; + vpxor XTS_TW_X1, %xmm1, %xmm1; + vpxor XTS_TW_O2, %xmm5, %xmm5; + vpxor XTS_TW_X3, %xmm3, %xmm3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X0, XTS_TW_X3, %xmm4, XTS_MASK); + vmovdqu XTS_TW_X0, (%rcx); + STORE_BLOCK4_AVX2(%xmm0, %xmm1, %xmm5, %xmm3, %rsi); + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_dec_4way_avx512) + +SYM_FUNC_START(lea_x86_64_xts_dec_8way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (8 blocks) + * %rdx: src (8 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + vzeroupper; + + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I2, XTS_TW_X1, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_I2, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X3, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm5, XTS_MASK); + vmovdqu XTS_TW_X8, (%rcx); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X0, XTS_TW_X1, XTS_TW_O2, XTS_TW_X3, %xmm5, XTS_TW_I2); + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti128 $0x1, XTS_TW_X4, XTS_TW_Y0, XTS_TW_Y0; + vinserti128 $0x1, XTS_TW_X5, XTS_TW_Y1, XTS_TW_Y1; + vinserti128 $0x1, XTS_TW_O6, XTS_TW_Y2, XTS_TW_Y2; + vinserti128 $0x1, XTS_TW_X7, XTS_TW_Y3, XTS_TW_Y3; + + GATHER_BLOCK8_AVX2(0, 1, 2, 3, 4, %rdx); + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + call __lea_x86_64_dec_8way_avx512 + + vpxor XTS_TW_Y0, %ymm0, %ymm0; + vpxor XTS_TW_Y1, %ymm1, %ymm1; + vpxor XTS_TW_Y2, %ymm2, %ymm2; + vpxor XTS_TW_Y3, %ymm3, %ymm3; + + SCATTER_BLOCK8_AVX2(0, 1, 2, 3, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_dec_8way_avx512) + +SYM_FUNC_START(lea_x86_64_xts_dec_16way_avx512) + /* input: + * %rdi: ctx, CTX + * %rsi: dst (16 blocks) + * %rdx: src (16 blocks) + * %rcx: tweak + */ + FRAME_BEGIN + + vzeroupper; + + vmovdqu (%rcx), XTS_TW_X0; + XTS_LOAD_TWEAK_MASK_AVX2(XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X1, XTS_TW_X0, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I2, XTS_TW_X1, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X3, XTS_TW_I2, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X3, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X0, XTS_TW_X1, XTS_TW_O2, XTS_TW_X3, %xmm5, XTS_TW_I2); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti32x4 $0x1, XTS_TW_X4, XTS_TW_Z0, XTS_TW_Z0; + vinserti32x4 $0x1, XTS_TW_X5, XTS_TW_Z1, XTS_TW_Z1; + vinserti32x4 $0x1, XTS_TW_O6, XTS_TW_Z2, XTS_TW_Z2; + vinserti32x4 $0x1, XTS_TW_X7, XTS_TW_Z3, XTS_TW_Z3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X8, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X4, XTS_TW_X7, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X8, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti32x4 $0x2, XTS_TW_X8, XTS_TW_Z0, XTS_TW_Z0; + vinserti32x4 $0x2, XTS_TW_X5, XTS_TW_Z1, XTS_TW_Z1; + vinserti32x4 $0x2, XTS_TW_O6, XTS_TW_Z2, XTS_TW_Z2; + vinserti32x4 $0x2, XTS_TW_X7, XTS_TW_Z3, XTS_TW_Z3; + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X5, XTS_TW_X4, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_I6, XTS_TW_X5, %xmm5, XTS_MASK); + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X7, XTS_TW_I6, %xmm5, XTS_MASK); + + XTS_NEXT_TWEAK_1BLOCK_AVX2(XTS_TW_X8, XTS_TW_X7, %xmm5, XTS_MASK); + + TRANSPOSE_BLOCK4_AVX2(XTS_TW_X4, XTS_TW_X5, XTS_TW_O6, XTS_TW_X7, %xmm5, XTS_TW_I6); + + vinserti32x4 $0x3, XTS_TW_X4, XTS_TW_Z0, XTS_TW_Z0; + vinserti32x4 $0x3, XTS_TW_X5, XTS_TW_Z1, XTS_TW_Z1; + vinserti32x4 $0x3, XTS_TW_O6, XTS_TW_Z2, XTS_TW_Z2; + vinserti32x4 $0x3, XTS_TW_X7, XTS_TW_Z3, XTS_TW_Z3; + + vmovdqu XTS_TW_X8, (%rcx); + + GATHER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rdx); + + vpxord XTS_TW_Z0, %zmm0, %zmm0; + vpxord XTS_TW_Z1, %zmm1, %zmm1; + vpxord XTS_TW_Z2, %zmm2, %zmm2; + vpxord XTS_TW_Z3, %zmm3, %zmm3; + + call __lea_x86_64_dec_16way_avx512 + + vpxord XTS_TW_Z0, %zmm0, %zmm0; + vpxord XTS_TW_Z1, %zmm1, %zmm1; + vpxord XTS_TW_Z2, %zmm2, %zmm2; + vpxord XTS_TW_Z3, %zmm3, %zmm3; + + SCATTER_BLOCK16_AVX512(0, 1, 2, 3, 4, %rsi); + + vzeroupper; + + FRAME_END + RET; +SYM_FUNC_END(lea_x86_64_xts_dec_16way_avx512) +#endif + +#ifdef CONFIG_AS_AVX512 + +.section .rodata.cst64.bswap_shuffle_mask, "aM", @progbits, 64 +.align 64 +.Lbswap_mask: + .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 + .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 + .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 + .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 + + +.section .rodata.cst64.ctr_add_const, "aM", @progbits, 64 +.align 64 +.Lctr_add_const: + .octa 0x00000003000000020000000100000000 + .octa 0x00000007000000060000000500000004 + .octa 0x0000000b0000000a0000000900000008 + .octa 0x0000000f0000000e0000000d0000000c + +.section .rodata.cst64.cbc_shuffle_mask16, "aM", @progbits, 64 +.align 64 +.Lcbc_shuffle_mask16: + .octa 0x0000000200000001000000000000000f + .octa 0x00000006000000050000000400000003 + .octa 0x0000000a000000090000000800000007 + .octa 0x0000000e0000000d0000000c0000000b + +#else + +.section .rodata.cst32.bswap_shuffle_mask, "aM", @progbits, 32 +.align 32 +.Lbswap_mask: + .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 + .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 + +.section .rodata.cst32.ctr_add_const, "aM", @progbits, 32 +.align 32 +.Lctr_add_const: + .octa 0x00000003000000020000000100000000 + .octa 0x00000007000000060000000500000004 + +#endif + +.section .rodata.cst32.cbc_shuffle_mask, "aM", @progbits, 32 +.align 32 +.Lcbc_shuffle_mask: + .octa 0x00000002000000010000000000000007 + .octa 0x00000006000000050000000400000003 + +.section .rodata.cst16.xts_tweak_mask, "aM", @progbits, 16 +.align 16 +.Lxts_tweak_mask: + .octa 0x00000000000000010000000000000087 diff --git a/arch/x86/crypto/lea-x86_64-glue.c b/arch/x86/crypto/lea-x86_64-glue.c new file mode 100644 index 000000000000..a28d89960090 --- /dev/null +++ b/arch/x86/crypto/lea-x86_64-glue.c @@ -0,0 +1,820 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Glue Code for the SSE2/AVX2/AVX-512F assembler instructions + * for the LEA Cipher + * + * - SSE2: 4 parallel blocks + * - AVX2: 8/4 parallel blocks + * - AVX-512F: 16/8/4 parallel blocks + * + * Copyright (c) 2023 National Security Research. + * Author: Dongsoo Lee + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include "ecb_cbc_helpers.h" + +struct crypto_lea_xts_ctx { + struct crypto_lea_ctx crypt_ctx; + struct crypto_lea_ctx __aligned(LEA_BLOCK_SIZE) tweak_ctx; +}; + +#define LEA_AVX512_MAX_PARALLEL_BLOCKS 16 +#define LEA_AVX2_MAX_PARALLEL_BLOCKS 8 +#define LEA_SSE2_MAX_PARALLEL_BLOCKS 4 +#define LEA_MIN_PARALLEL_BLOCKS LEA_SSE2_MAX_PARALLEL_BLOCKS + +typedef void (*lea_ecb_enc_nway)(const void *ctx, u8 *dst, const u8 *src); +typedef void (*lea_ecb_dec_nway)(const void *ctx, u8 *dst, const u8 *src); + +typedef void (*lea_cbc_dec_nway)(const void *ctx, u8 *dst, const u8 *src); + +typedef void (*lea_ctr_enc_nway)(const void *ctx, u8 *dst, const u8 *src, + u8 *ctr); + +typedef void (*lea_xts_enc_nway)(const void *ctx, u8 *dst, const u8 *src, + u8 *tweak); +typedef void (*lea_xts_dec_nway)(const void *ctx, u8 *dst, const u8 *src, + u8 *tweak); + +struct lea_func_map_t { + lea_ecb_enc_nway ecb_enc_4way; + lea_ecb_dec_nway ecb_dec_4way; + lea_cbc_dec_nway cbc_dec_4way; + lea_ctr_enc_nway ctr_enc_4way; + lea_xts_enc_nway xts_enc_4way; + lea_xts_dec_nway xts_dec_4way; + + lea_ecb_enc_nway ecb_enc_8way; + lea_ecb_dec_nway ecb_dec_8way; + lea_cbc_dec_nway cbc_dec_8way; + lea_ctr_enc_nway ctr_enc_8way; + lea_xts_enc_nway xts_enc_8way; + lea_xts_dec_nway xts_dec_8way; + +#ifdef CONFIG_AS_AVX512 + lea_ecb_enc_nway ecb_enc_16way; + lea_ecb_dec_nway ecb_dec_16way; + lea_cbc_dec_nway cbc_dec_16way; + lea_ctr_enc_nway ctr_enc_16way; + lea_xts_enc_nway xts_enc_16way; + lea_xts_dec_nway xts_dec_16way; +#endif + int proc_blocks; +}; + +#ifdef CONFIG_AS_AVX512 +asmlinkage void lea_x86_64_ecb_enc_16way_avx512(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_dec_16way_avx512(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_enc_8way_avx512(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_dec_8way_avx512(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_enc_4way_avx512(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_dec_4way_avx512(const void *ctx, u8 *dst, + const u8 *src); +#endif + +asmlinkage void lea_x86_64_ecb_enc_8way_avx2(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_dec_8way_avx2(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_enc_4way_avx2(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_dec_4way_avx2(const void *ctx, u8 *dst, + const u8 *src); + +asmlinkage void lea_x86_64_ecb_enc_4way_sse2(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_ecb_dec_4way_sse2(const void *ctx, u8 *dst, + const u8 *src); + +#ifdef CONFIG_AS_AVX512 +asmlinkage void lea_x86_64_cbc_dec_16way_avx512(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_cbc_dec_8way_avx512(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_cbc_dec_4way_avx512(const void *ctx, u8 *dst, + const u8 *src); +#endif + +asmlinkage void lea_x86_64_cbc_dec_8way_avx2(const void *ctx, u8 *dst, + const u8 *src); +asmlinkage void lea_x86_64_cbc_dec_4way_avx2(const void *ctx, u8 *dst, + const u8 *src); + +asmlinkage void lea_x86_64_cbc_dec_4way_sse2(const void *ctx, u8 *dst, + const u8 *src); + +#ifdef CONFIG_AS_AVX512 +asmlinkage void lea_x86_64_ctr_enc_16way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *ctr); +asmlinkage void lea_x86_64_ctr_enc_8way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *ctr); +asmlinkage void lea_x86_64_ctr_enc_4way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *ctr); +#endif + +asmlinkage void lea_x86_64_ctr_enc_8way_avx2(const void *ctx, u8 *dst, + const u8 *src, u8 *ctr); +asmlinkage void lea_x86_64_ctr_enc_4way_avx2(const void *ctx, u8 *dst, + const u8 *src, u8 *ctr); + +asmlinkage void lea_x86_64_ctr_enc_4way_sse2(const void *ctx, u8 *dst, + const u8 *src, u8 *ctr); + +#ifdef CONFIG_AS_AVX512 +asmlinkage void lea_x86_64_xts_enc_16way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_dec_16way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_enc_8way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_dec_8way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_enc_4way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_dec_4way_avx512(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +#endif + +asmlinkage void lea_x86_64_xts_enc_8way_avx2(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_dec_8way_avx2(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_enc_4way_avx2(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_dec_4way_avx2(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); + +asmlinkage void lea_x86_64_xts_enc_4way_sse2(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); +asmlinkage void lea_x86_64_xts_dec_4way_sse2(const void *ctx, u8 *dst, + const u8 *src, u8 *tweak); + +#ifdef CONFIG_AS_AVX512 +static const struct lea_func_map_t lea_avx512_func_map = { + .ecb_enc_4way = lea_x86_64_ecb_enc_4way_avx512, + .ecb_dec_4way = lea_x86_64_ecb_dec_4way_avx512, + .cbc_dec_4way = lea_x86_64_cbc_dec_4way_avx512, + .ctr_enc_4way = lea_x86_64_ctr_enc_4way_avx512, + .xts_enc_4way = lea_x86_64_xts_enc_4way_avx512, + .xts_dec_4way = lea_x86_64_xts_dec_4way_avx512, + + .ecb_enc_8way = lea_x86_64_ecb_enc_8way_avx512, + .ecb_dec_8way = lea_x86_64_ecb_dec_8way_avx512, + .cbc_dec_8way = lea_x86_64_cbc_dec_8way_avx512, + .ctr_enc_8way = lea_x86_64_ctr_enc_8way_avx512, + .xts_enc_8way = lea_x86_64_xts_enc_8way_avx512, + .xts_dec_8way = lea_x86_64_xts_dec_8way_avx512, + + .ecb_enc_16way = lea_x86_64_ecb_enc_16way_avx512, + .ecb_dec_16way = lea_x86_64_ecb_dec_16way_avx512, + .cbc_dec_16way = lea_x86_64_cbc_dec_16way_avx512, + .ctr_enc_16way = lea_x86_64_ctr_enc_16way_avx512, + .xts_enc_16way = lea_x86_64_xts_enc_16way_avx512, + .xts_dec_16way = lea_x86_64_xts_dec_16way_avx512, + + .proc_blocks = LEA_AVX512_MAX_PARALLEL_BLOCKS, +}; +#endif + +static const struct lea_func_map_t lea_avx2_func_map = { + .ecb_enc_4way = lea_x86_64_ecb_enc_4way_avx2, + .ecb_dec_4way = lea_x86_64_ecb_dec_4way_avx2, + .cbc_dec_4way = lea_x86_64_cbc_dec_4way_avx2, + .ctr_enc_4way = lea_x86_64_ctr_enc_4way_avx2, + .xts_enc_4way = lea_x86_64_xts_enc_4way_avx2, + .xts_dec_4way = lea_x86_64_xts_dec_4way_avx2, + + .ecb_enc_8way = lea_x86_64_ecb_enc_8way_avx2, + .ecb_dec_8way = lea_x86_64_ecb_dec_8way_avx2, + .cbc_dec_8way = lea_x86_64_cbc_dec_8way_avx2, + .ctr_enc_8way = lea_x86_64_ctr_enc_8way_avx2, + .xts_enc_8way = lea_x86_64_xts_enc_8way_avx2, + .xts_dec_8way = lea_x86_64_xts_dec_8way_avx2, + + .proc_blocks = LEA_AVX2_MAX_PARALLEL_BLOCKS, +}; + +/* default SSE2 */ +static struct lea_func_map_t lea_func_map = { + .ecb_enc_4way = lea_x86_64_ecb_enc_4way_sse2, + .ecb_dec_4way = lea_x86_64_ecb_dec_4way_sse2, + .cbc_dec_4way = lea_x86_64_cbc_dec_4way_sse2, + .ctr_enc_4way = lea_x86_64_ctr_enc_4way_sse2, + .xts_enc_4way = lea_x86_64_xts_enc_4way_sse2, + .xts_dec_4way = lea_x86_64_xts_dec_4way_sse2, + + .proc_blocks = LEA_SSE2_MAX_PARALLEL_BLOCKS, +}; + +static int ecb_encrypt_nway(struct skcipher_request *req) +{ + ECB_WALK_START(req, LEA_BLOCK_SIZE, LEA_MIN_PARALLEL_BLOCKS); + if (lea_func_map.ecb_enc_16way) + ECB_BLOCK(16, lea_func_map.ecb_enc_16way); + if (lea_func_map.ecb_enc_8way) + ECB_BLOCK(8, lea_func_map.ecb_enc_8way); + ECB_BLOCK(4, lea_func_map.ecb_enc_4way); + ECB_BLOCK(1, lea_encrypt); + ECB_WALK_END(); +} + +static int ecb_decrypt_nway(struct skcipher_request *req) +{ + ECB_WALK_START(req, LEA_BLOCK_SIZE, LEA_MIN_PARALLEL_BLOCKS); + if (lea_func_map.ecb_dec_16way) + ECB_BLOCK(16, lea_func_map.ecb_dec_16way); + if (lea_func_map.ecb_dec_8way) + ECB_BLOCK(8, lea_func_map.ecb_dec_8way); + ECB_BLOCK(4, lea_x86_64_ecb_dec_4way_avx2); + ECB_BLOCK(1, lea_decrypt); + ECB_WALK_END(); +} + +static int cbc_encrypt(struct skcipher_request *req) +{ + CBC_WALK_START(req, LEA_BLOCK_SIZE, -1); + CBC_ENC_BLOCK(lea_encrypt); + CBC_WALK_END(); +} + +static int cbc_decrypt_nway(struct skcipher_request *req) +{ + CBC_WALK_START(req, LEA_BLOCK_SIZE, LEA_MIN_PARALLEL_BLOCKS); + if (lea_func_map.cbc_dec_16way) + CBC_DEC_BLOCK(16, lea_func_map.cbc_dec_16way); + if (lea_func_map.cbc_dec_8way) + CBC_DEC_BLOCK(8, lea_func_map.cbc_dec_8way); + CBC_DEC_BLOCK(4, lea_func_map.cbc_dec_4way); + CBC_DEC_BLOCK(1, lea_decrypt); + CBC_WALK_END(); +} + +static inline void xor_1blk(u8 *out, const u8 *in1, const u8 *in2) +{ + const u64 *_in1 = (const u64 *)in1; + const u64 *_in2 = (const u64 *)in2; + u64 *_out = (u64 *)out; + + _out[0] = _in1[0] ^ _in2[0]; + _out[1] = _in1[1] ^ _in2[1]; +} + +static inline void xts_next_tweak(u8 *out, const u8 *in) +{ + const u64 *_in = (const u64 *)in; + u64 *_out = (u64 *)out; + u64 v0 = _in[0]; + u64 v1 = _in[1]; + u64 carry = (u64)(((s64)v1) >> 63); + + v1 = (v1 << 1) ^ (v0 >> 63); + v0 = (v0 << 1) ^ ((u64)carry & 0x87); + + _out[0] = v0; + _out[1] = v1; +} + +#define CTR_NWAY(NWAY, FUNC) \ + while (nbytes >= NWAY * LEA_BLOCK_SIZE) { \ + FUNC(ctx, dst, src, walk.iv); \ + src += NWAY * LEA_BLOCK_SIZE; \ + dst += NWAY * LEA_BLOCK_SIZE; \ + nbytes -= NWAY * LEA_BLOCK_SIZE; \ + } + +#define XTS_NWAY(NWAY, FUNC) \ + for (; nblocks >= NWAY; nblocks -= NWAY) { \ + FUNC(&ctx->crypt_ctx, dst, src, walk.iv); \ + src += NWAY * LEA_BLOCK_SIZE; \ + dst += NWAY * LEA_BLOCK_SIZE; \ + nbytes -= NWAY * LEA_BLOCK_SIZE; \ + } + +#define XTS_SINGLE(FUNC) \ + for (; nblocks > 0; nblocks -= 1) { \ + u8 __aligned(16) buffer[LEA_BLOCK_SIZE]; \ + xor_1blk(buffer, walk.iv, src); \ + FUNC(&ctx->crypt_ctx, buffer, buffer); \ + xor_1blk(dst, walk.iv, buffer); \ + xts_next_tweak(walk.iv, walk.iv); \ + src += LEA_BLOCK_SIZE; \ + dst += LEA_BLOCK_SIZE; \ + nbytes -= LEA_BLOCK_SIZE; \ + } + +#define XTS_TAIL_ENCRYPT(FUNC) \ + do { \ + u8 __aligned(16) buffer[LEA_BLOCK_SIZE]; \ + xor_1blk(buffer, walk.iv, src); \ + FUNC(&ctx->crypt_ctx, buffer, buffer); \ + xor_1blk(buffer, walk.iv, buffer); \ + memcpy(dst, buffer, LEA_BLOCK_SIZE); \ + memcpy(buffer, src + LEA_BLOCK_SIZE, tail); \ + memcpy(dst + LEA_BLOCK_SIZE, dst, tail); \ + xts_next_tweak(walk.iv, walk.iv); \ + xor_1blk(buffer, walk.iv, buffer); \ + FUNC(&ctx->crypt_ctx, buffer, buffer); \ + xor_1blk(dst, walk.iv, buffer); \ + nbytes -= LEA_BLOCK_SIZE + tail; \ + } while (0) + +#define XTS_TAIL_DECRYPT(FUNC) \ + do { \ + u8 __aligned(16) ntweak[LEA_BLOCK_SIZE]; \ + u8 __aligned(16) buffer[LEA_BLOCK_SIZE]; \ + memcpy(ntweak, walk.iv, LEA_BLOCK_SIZE); \ + xts_next_tweak(walk.iv, ntweak); \ + xor_1blk(buffer, walk.iv, src); \ + FUNC(&ctx->crypt_ctx, buffer, buffer); \ + xor_1blk(buffer, walk.iv, buffer); \ + memcpy(dst, buffer, LEA_BLOCK_SIZE); \ + memcpy(buffer, src + 16, tail); \ + memcpy(dst + 16, dst, tail); \ + xor_1blk(buffer, ntweak, buffer); \ + FUNC(&ctx->crypt_ctx, buffer, buffer); \ + xor_1blk(dst, ntweak, buffer); \ + nbytes -= LEA_BLOCK_SIZE + tail; \ + } while (0) + +static inline int xts_edge_tail_encrypt(struct crypto_lea_xts_ctx *ctx, + struct skcipher_request *req, + struct skcipher_walk *walk, + int edge_tail) +{ + u8 __aligned(16) buffer[LEA_BLOCK_SIZE]; + struct scatterlist sg_src[2]; + struct scatterlist sg_dst[2]; + struct scatterlist *scatter_src; + struct scatterlist *scatter_dst; + const u8 *src; + u8 *dst; + + scatter_src = scatterwalk_ffwd(sg_src, req->src, req->cryptlen); + if (req->src == req->dst) + scatter_dst = scatter_src; + else + scatter_dst = scatterwalk_ffwd(sg_dst, req->dst, req->cryptlen); + + skcipher_request_set_crypt(req, scatter_src, scatter_dst, + LEA_BLOCK_SIZE + edge_tail, req->iv); + + skcipher_walk_virt(walk, req, false); + + src = walk->src.virt.addr; + dst = walk->dst.virt.addr; + + kernel_fpu_begin(); + + xor_1blk(buffer, walk->iv, src); + lea_encrypt(&ctx->crypt_ctx, buffer, buffer); + xor_1blk(buffer, walk->iv, buffer); + + memcpy(dst, buffer, LEA_BLOCK_SIZE); + memcpy(buffer, src + LEA_BLOCK_SIZE, edge_tail); + memcpy(dst + LEA_BLOCK_SIZE, dst, edge_tail); + + xts_next_tweak(walk->iv, walk->iv); + + xor_1blk(buffer, walk->iv, buffer); + lea_encrypt(&ctx->crypt_ctx, buffer, buffer); + xor_1blk(dst, walk->iv, buffer); + + kernel_fpu_end(); + return skcipher_walk_done(walk, 0); +} + +static inline int xts_edge_tail_decrypt(struct crypto_lea_xts_ctx *ctx, + struct skcipher_request *req, + struct skcipher_walk *walk, + int edge_tail) +{ + u8 __aligned(16) buffer[LEA_BLOCK_SIZE]; + u8 __aligned(16) ntweak[LEA_BLOCK_SIZE]; + struct scatterlist sg_src[2]; + struct scatterlist sg_dst[2]; + struct scatterlist *scatter_src; + struct scatterlist *scatter_dst; + + const u8 *src; + u8 *dst; + int ret; + + scatter_src = scatterwalk_ffwd(sg_src, req->src, req->cryptlen); + if (req->src == req->dst) + scatter_dst = scatter_src; + else + scatter_dst = scatterwalk_ffwd(sg_dst, req->dst, req->cryptlen); + + skcipher_request_set_crypt(req, scatter_src, scatter_dst, + LEA_BLOCK_SIZE + edge_tail, req->iv); + + ret = skcipher_walk_virt(walk, req, false); + + src = walk->src.virt.addr; + dst = walk->dst.virt.addr; + + kernel_fpu_begin(); + + memcpy(ntweak, walk->iv, LEA_BLOCK_SIZE); + xts_next_tweak(walk->iv, ntweak); + + xor_1blk(buffer, walk->iv, src); + lea_decrypt(&ctx->crypt_ctx, buffer, buffer); + xor_1blk(buffer, walk->iv, buffer); + + memcpy(dst, buffer, LEA_BLOCK_SIZE); + + memcpy(buffer, src + 16, edge_tail); + memcpy(dst + 16, dst, edge_tail); + + xor_1blk(buffer, ntweak, buffer); + lea_decrypt(&ctx->crypt_ctx, buffer, buffer); + xor_1blk(dst, ntweak, buffer); + + kernel_fpu_end(); + + return skcipher_walk_done(walk, 0); +} + +static int xts_encrypt_nway(struct skcipher_request *req) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + struct crypto_lea_xts_ctx *ctx = crypto_skcipher_ctx(tfm); + struct skcipher_request subreq; + struct skcipher_walk walk; + + int ret; + u32 nblocks; + u32 tail = req->cryptlen % LEA_BLOCK_SIZE; + u32 edge_tail = 0; + + if (req->cryptlen < LEA_BLOCK_SIZE) + return -EINVAL; + + ret = skcipher_walk_virt(&walk, req, false); + if (ret) + return ret; + + if (unlikely(tail != 0 && walk.nbytes < walk.total)) { + u32 req_len = req->cryptlen - LEA_BLOCK_SIZE - tail; + + skcipher_walk_abort(&walk); + + skcipher_request_set_tfm(&subreq, tfm); + skcipher_request_set_callback( + &subreq, skcipher_request_flags(req), NULL, NULL); + skcipher_request_set_crypt(&subreq, req->src, req->dst, req_len, + req->iv); + req = &subreq; + ret = skcipher_walk_virt(&walk, req, false); + if (ret) + return ret; + edge_tail = tail; + tail = 0; + } + + lea_encrypt(&ctx->tweak_ctx, walk.iv, walk.iv); + + while ((nblocks = walk.nbytes / LEA_BLOCK_SIZE) > 0) { + u32 nbytes = walk.nbytes; + const u8 *src = walk.src.virt.addr; + u8 *dst = walk.dst.virt.addr; + bool is_tail = tail != 0 && + (nblocks + 1) * LEA_BLOCK_SIZE > walk.total; + + if (unlikely(is_tail)) + nblocks -= 1; + + kernel_fpu_begin(); + + if (lea_func_map.xts_enc_16way) + XTS_NWAY(16, lea_func_map.xts_enc_16way); + if (lea_func_map.xts_enc_8way) + XTS_NWAY(8, lea_func_map.xts_enc_8way); + XTS_NWAY(4, lea_func_map.xts_enc_4way); + XTS_SINGLE(lea_encrypt); + + if (unlikely(is_tail)) { + XTS_TAIL_ENCRYPT(lea_encrypt); + kernel_fpu_end(); + return skcipher_walk_done(&walk, nbytes); + } + + kernel_fpu_end(); + ret = skcipher_walk_done(&walk, nbytes); + if (ret) + return ret; + } + + if (unlikely(edge_tail != 0)) + ret = xts_edge_tail_encrypt(ctx, req, &walk, edge_tail); + + return ret; +} + +static int xts_decrypt_nway(struct skcipher_request *req) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + struct crypto_lea_xts_ctx *ctx = crypto_skcipher_ctx(tfm); + struct skcipher_request subreq; + struct skcipher_walk walk; + + int ret; + u32 nblocks; + u32 tail = req->cryptlen % LEA_BLOCK_SIZE; + u32 edge_tail = 0; + + if (req->cryptlen < LEA_BLOCK_SIZE) + return -EINVAL; + + ret = skcipher_walk_virt(&walk, req, false); + + if (ret) + return ret; + + if (unlikely(tail != 0 && walk.nbytes < walk.total)) { + u32 req_len = req->cryptlen - LEA_BLOCK_SIZE - tail; + + skcipher_walk_abort(&walk); + + skcipher_request_set_tfm(&subreq, tfm); + skcipher_request_set_callback( + &subreq, skcipher_request_flags(req), NULL, NULL); + skcipher_request_set_crypt(&subreq, req->src, req->dst, req_len, + req->iv); + req = &subreq; + ret = skcipher_walk_virt(&walk, req, false); + if (ret) + return ret; + + edge_tail = tail; + tail = 0; + } + + lea_encrypt(&ctx->tweak_ctx, walk.iv, walk.iv); + + while ((nblocks = walk.nbytes / LEA_BLOCK_SIZE) > 0) { + u32 nbytes = walk.nbytes; + const u8 *src = walk.src.virt.addr; + u8 *dst = walk.dst.virt.addr; + bool is_tail = tail != 0 && + (nblocks + 1) * LEA_BLOCK_SIZE > walk.total; + + if (unlikely(is_tail)) + nblocks -= 1; + + kernel_fpu_begin(); + + if (lea_func_map.xts_dec_16way) + XTS_NWAY(16, lea_func_map.xts_dec_16way); + if (lea_func_map.xts_dec_8way) + XTS_NWAY(8, lea_func_map.xts_dec_8way); + XTS_NWAY(4, lea_func_map.xts_dec_4way); + XTS_SINGLE(lea_decrypt); + + if (unlikely(is_tail)) { + XTS_TAIL_DECRYPT(lea_decrypt); + kernel_fpu_end(); + return skcipher_walk_done(&walk, nbytes); + } + + kernel_fpu_end(); + + ret = skcipher_walk_done(&walk, nbytes); + if (ret) + return ret; + } + + if (unlikely(edge_tail != 0)) + xts_edge_tail_decrypt(ctx, req, &walk, edge_tail); + + return ret; +} + +static int ctr_encrypt_nway(struct skcipher_request *req) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + struct crypto_lea_ctx *ctx = crypto_skcipher_ctx(tfm); + struct skcipher_walk walk; + + u8 __aligned(16) buffer[LEA_BLOCK_SIZE]; + + int ret; + + ret = skcipher_walk_virt(&walk, req, false); + if (ret) + return ret; + + while (walk.nbytes > 0) { + u32 nbytes = walk.nbytes; + const u8 *src = walk.src.virt.addr; + u8 *dst = walk.dst.virt.addr; + + kernel_fpu_begin(); + + if (lea_func_map.ctr_enc_16way) + CTR_NWAY(16, lea_func_map.ctr_enc_16way); + if (lea_func_map.ctr_enc_8way) + CTR_NWAY(8, lea_func_map.ctr_enc_8way); + CTR_NWAY(4, lea_func_map.ctr_enc_4way); + + while (nbytes >= LEA_BLOCK_SIZE) { + lea_encrypt(ctx, buffer, walk.iv); + xor_1blk(dst, buffer, src); + crypto_inc(walk.iv, LEA_BLOCK_SIZE); + + src += LEA_BLOCK_SIZE; + dst += LEA_BLOCK_SIZE; + nbytes -= LEA_BLOCK_SIZE; + } + + if (unlikely(walk.nbytes == walk.total && nbytes != 0)) { + lea_encrypt(ctx, buffer, walk.iv); + crypto_xor_cpy(dst, src, buffer, nbytes); + crypto_inc(walk.iv, LEA_BLOCK_SIZE); + + nbytes = 0; + } + + kernel_fpu_end(); + ret = skcipher_walk_done(&walk, nbytes); + if (ret) + return ret; + } + + return ret; +} + +static int xts_lea_set_key(struct crypto_skcipher *tfm, const u8 *key, + u32 keylen) +{ + struct crypto_lea_xts_ctx *ctx = crypto_skcipher_ctx(tfm); + + struct crypto_lea_ctx *crypt_key = &ctx->crypt_ctx; + struct crypto_lea_ctx *tweak_key = &ctx->tweak_ctx; + + int result; + + result = xts_verify_key(tfm, key, keylen); + if (result) + return result; + + result = lea_set_key(crypt_key, key, keylen / 2); + + if (result) + return result; + + return lea_set_key(tweak_key, key + (keylen / 2), keylen / 2); +} + +static int _lea_set_key(struct crypto_skcipher *tfm, const u8 *key, u32 keylen) +{ + return lea_set_key(crypto_skcipher_ctx(tfm), key, keylen); +} + +static struct skcipher_alg lea_x86_64_algs[] = { + { + .base.cra_name = "__ecb(lea)", + .base.cra_driver_name = "__ecb-lea-x86_64", + .base.cra_priority = 300, + .base.cra_flags = CRYPTO_ALG_INTERNAL, + .base.cra_blocksize = LEA_BLOCK_SIZE, + .base.cra_ctxsize = sizeof(struct crypto_lea_ctx), + .base.cra_module = THIS_MODULE, + .min_keysize = LEA_MIN_KEY_SIZE, + .max_keysize = LEA_MAX_KEY_SIZE, + .walksize = LEA_SSE2_MAX_PARALLEL_BLOCKS * LEA_BLOCK_SIZE, + .setkey = _lea_set_key, + .encrypt = ecb_encrypt_nway, + .decrypt = ecb_decrypt_nway, + }, + { + .base.cra_name = "__ctr(lea)", + .base.cra_driver_name = "__ctr-lea-x86_64", + .base.cra_priority = 300, + .base.cra_flags = CRYPTO_ALG_INTERNAL, + .base.cra_blocksize = 1, + .base.cra_ctxsize = sizeof(struct crypto_lea_ctx), + .base.cra_module = THIS_MODULE, + .min_keysize = LEA_MIN_KEY_SIZE, + .max_keysize = LEA_MAX_KEY_SIZE, + .chunksize = LEA_BLOCK_SIZE, + .walksize = LEA_SSE2_MAX_PARALLEL_BLOCKS * LEA_BLOCK_SIZE, + .ivsize = LEA_BLOCK_SIZE, + .setkey = _lea_set_key, + .encrypt = ctr_encrypt_nway, + .decrypt = ctr_encrypt_nway, + }, + { + .base.cra_name = "__cbc(lea)", + .base.cra_driver_name = "__cbc-lea-x86_64", + .base.cra_priority = 300, + .base.cra_flags = CRYPTO_ALG_INTERNAL, + .base.cra_blocksize = LEA_BLOCK_SIZE, + .base.cra_ctxsize = sizeof(struct crypto_lea_ctx), + .base.cra_module = THIS_MODULE, + .min_keysize = LEA_MIN_KEY_SIZE, + .max_keysize = LEA_MAX_KEY_SIZE, + .walksize = LEA_SSE2_MAX_PARALLEL_BLOCKS * LEA_BLOCK_SIZE, + .ivsize = LEA_BLOCK_SIZE, + .setkey = _lea_set_key, + .encrypt = cbc_encrypt, + .decrypt = cbc_decrypt_nway, + }, + { + .base.cra_name = "__xts(lea)", + .base.cra_driver_name = "__xts-lea-x86_64", + .base.cra_priority = 300, + .base.cra_flags = CRYPTO_ALG_INTERNAL, + .base.cra_blocksize = LEA_BLOCK_SIZE, + .base.cra_ctxsize = sizeof(struct crypto_lea_xts_ctx), + .base.cra_module = THIS_MODULE, + .min_keysize = LEA_MIN_KEY_SIZE * 2, + .max_keysize = LEA_MAX_KEY_SIZE * 2, + .walksize = LEA_SSE2_MAX_PARALLEL_BLOCKS * LEA_BLOCK_SIZE, + .ivsize = LEA_BLOCK_SIZE, + .setkey = xts_lea_set_key, + .encrypt = xts_encrypt_nway, + .decrypt = xts_decrypt_nway, + }, +}; + +static struct simd_skcipher_alg *lea_x86_64_sk_algs[ARRAY_SIZE(lea_x86_64_algs)]; + +static int __init crypto_lea_x86_64_init(void) +{ + const char *feature_name; + int i; + const struct lea_func_map_t *target_call_map; + + target_call_map = &lea_avx512_func_map; + + if (!boot_cpu_has(X86_FEATURE_AVX512F)) + target_call_map = &lea_avx2_func_map; + else if (!cpu_has_xfeatures(XFEATURE_MASK_ZMM_Hi256 | + XFEATURE_MASK_OPMASK, + &feature_name)) + target_call_map = &lea_avx2_func_map; + + if (!boot_cpu_has(X86_FEATURE_AVX2) || !boot_cpu_has(X86_FEATURE_AVX)) + target_call_map = NULL; + else if (!cpu_has_xfeatures(XFEATURE_MASK_YMM, &feature_name)) + target_call_map = NULL; + + if (!boot_cpu_has(X86_FEATURE_XMM2)) { + pr_info("SSE2 instructions are not detected.\n"); + return -ENODEV; + } + + if (!cpu_has_xfeatures(XFEATURE_MASK_SSE, &feature_name)) { + pr_info("CPU feature '%s' is not supported.\n", feature_name); + return -ENODEV; + } + + if (target_call_map != NULL) { + lea_func_map = *target_call_map; + for (i = 0; i < ARRAY_SIZE(lea_x86_64_algs); i++) { + lea_x86_64_algs[i].walksize = + lea_func_map.proc_blocks * LEA_BLOCK_SIZE; + } + } + + return simd_register_skciphers_compat(lea_x86_64_algs, + ARRAY_SIZE(lea_x86_64_sk_algs), + lea_x86_64_sk_algs); +} + +static void __exit crypto_lea_x86_64_exit(void) +{ + simd_unregister_skciphers(lea_x86_64_algs, + ARRAY_SIZE(lea_x86_64_sk_algs), + lea_x86_64_sk_algs); +} + +module_init(crypto_lea_x86_64_init); +module_exit(crypto_lea_x86_64_exit); + +MODULE_DESCRIPTION("LEA Cipher Algorithm, AVX-512F, AVX2, SSE2 SIMD"); +MODULE_AUTHOR("Dongsoo Lee "); +MODULE_LICENSE("GPL"); +MODULE_ALIAS_CRYPTO("lea"); +MODULE_ALIAS_CRYPTO("lea-x86_64");