Source-Changes-HG archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
[src-draft/trunk]: src/sys New permutation-based AES implementation using SSSE3.
details: https://anonhg.NetBSD.org/src-all/rev/7b5680ba9f51
branches: trunk
changeset: 935262:7b5680ba9f51
user: Taylor R Campbell <riastradh%NetBSD.org@localhost>
date: Fri Jun 26 23:17:30 2020 +0000
description:
New permutation-based AES implementation using SSSE3.
This covers a lot of CPUs -- particularly lower-end CPUs over the
past decade which lack AES-NI.
Derived from Mike Hamburg's public domain vpaes software; see
<https://crypto.stanford.edu/vpaes/> for details.
diffstat:
sys/arch/x86/conf/files.x86 | 3 +
sys/arch/x86/x86/identcpu.c | 5 +
sys/crypto/aes/arch/x86/aes_ssse3.c | 559 +++++++++++++++++++++++++++
sys/crypto/aes/arch/x86/aes_ssse3.h | 36 +
sys/crypto/aes/arch/x86/aes_ssse3_impl.c | 167 ++++++++
sys/crypto/aes/arch/x86/aes_ssse3_internal.h | 43 ++
sys/crypto/aes/arch/x86/aes_ssse3_subr.c | 214 ++++++++++
sys/crypto/aes/arch/x86/aes_ssse3_subr.h | 60 ++
sys/crypto/aes/arch/x86/files.aesssse3 | 8 +
sys/crypto/aes/arch/x86/immintrin.h | 62 ++
10 files changed, 1157 insertions(+), 0 deletions(-)
diffs (truncated from 1267 to 300 lines):
diff -r 83cd09a5a2f6 -r 7b5680ba9f51 sys/arch/x86/conf/files.x86
--- a/sys/arch/x86/conf/files.x86 Fri Jun 26 21:15:43 2020 +0000
+++ b/sys/arch/x86/conf/files.x86 Fri Jun 26 23:17:30 2020 +0000
@@ -174,3 +174,6 @@
# Bitsliced AES with SSE2
include "crypto/aes/arch/x86/files.aessse2"
+
+# Permutation-based AES with PSHUFB
+include "crypto/aes/arch/x86/files.aesssse3"
diff -r 83cd09a5a2f6 -r 7b5680ba9f51 sys/arch/x86/x86/identcpu.c
--- a/sys/arch/x86/x86/identcpu.c Fri Jun 26 21:15:43 2020 +0000
+++ b/sys/arch/x86/x86/identcpu.c Fri Jun 26 23:17:30 2020 +0000
@@ -41,6 +41,7 @@
#include <crypto/aes/arch/x86/aes_ni.h>
#include <crypto/aes/arch/x86/aes_sse2.h>
+#include <crypto/aes/arch/x86/aes_ssse3.h>
#include <crypto/aes/arch/x86/aes_via.h>
#include <uvm/uvm_extern.h>
@@ -1006,6 +1007,10 @@
#endif
if (cpu_feature[4] & CPUID_VIA_HAS_ACE)
aes_md_init(&aes_via_impl);
+ else if (i386_has_sse && i386_has_sse2 &&
+ (cpu_feature[1] & CPUID2_SSE3) &&
+ (cpu_feature[1] & CPUID2_SSSE3))
+ aes_md_init(&aes_ssse3_impl);
else if (i386_has_sse && i386_has_sse2)
aes_md_init(&aes_sse2_impl);
} else {
diff -r 83cd09a5a2f6 -r 7b5680ba9f51 sys/crypto/aes/arch/x86/aes_ssse3.c
--- /dev/null Thu Jan 01 00:00:00 1970 +0000
+++ b/sys/crypto/aes/arch/x86/aes_ssse3.c Fri Jun 26 23:17:30 2020 +0000
@@ -0,0 +1,559 @@
+/* $NetBSD$ */
+
+/*-
+ * Copyright (c) 2020 The NetBSD Foundation, Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * Permutation-based AES using SSSE3, derived from Mike Hamburg's VPAES
+ * software, at <https://crypto.stanford.edu/vpaes/>, described in
+ *
+ * Mike Hamburg, `Accelerating AES with Vector Permute
+ * Instructions', in Christophe Clavier and Kris Gaj (eds.),
+ * Cryptographic Hardware and Embedded Systems -- CHES 2009,
+ * Springer LNCS 5747, pp. 18-32.
+ *
+ * https://link.springer.com/chapter/10.1007/978-3-642-04138-9_2
+ */
+
+#include <sys/cdefs.h>
+__KERNEL_RCSID(1, "$NetBSD$");
+
+#include <sys/types.h>
+
+#include <sys/systm.h>
+
+#include "aes_ssse3_internal.h"
+
+static const union m128const {
+ uint64_t u64[2];
+ __m128i m;
+}
+mc_forward[4] = {
+ {.u64 = {0x0407060500030201, 0x0C0F0E0D080B0A09}},
+ {.u64 = {0x080B0A0904070605, 0x000302010C0F0E0D}},
+ {.u64 = {0x0C0F0E0D080B0A09, 0x0407060500030201}},
+ {.u64 = {0x000302010C0F0E0D, 0x080B0A0904070605}},
+},
+mc_backward[4] = {
+ {.u64 = {0x0605040702010003, 0x0E0D0C0F0A09080B}},
+ {.u64 = {0x020100030E0D0C0F, 0x0A09080B06050407}},
+ {.u64 = {0x0E0D0C0F0A09080B, 0x0605040702010003}},
+ {.u64 = {0x0A09080B06050407, 0x020100030E0D0C0F}},
+},
+ipt[2] = {
+ {.u64 = {0xC2B2E8985A2A7000, 0xCABAE09052227808}},
+ {.u64 = {0x4C01307D317C4D00, 0xCD80B1FCB0FDCC81}},
+},
+opt[2] = {
+ {.u64 = {0xFF9F4929D6B66000, 0xF7974121DEBE6808}},
+ {.u64 = {0x01EDBD5150BCEC00, 0xE10D5DB1B05C0CE0}},
+},
+dipt[2] = {
+ {.u64 = {0x0F505B040B545F00, 0x154A411E114E451A}},
+ {.u64 = {0x86E383E660056500, 0x12771772F491F194}},
+},
+sb1[2] = {
+ {.u64 = {0xB19BE18FCB503E00, 0xA5DF7A6E142AF544}},
+ {.u64 = {0x3618D415FAE22300, 0x3BF7CCC10D2ED9EF}},
+},
+sb2[2] = {
+ {.u64 = {0xE27A93C60B712400, 0x5EB7E955BC982FCD}},
+ {.u64 = {0x69EB88400AE12900, 0xC2A163C8AB82234A}},
+},
+sbo[2] = {
+ {.u64 = {0xD0D26D176FBDC700, 0x15AABF7AC502A878}},
+ {.u64 = {0xCFE474A55FBB6A00, 0x8E1E90D1412B35FA}},
+},
+dsb9[2] = {
+ {.u64 = {0x851C03539A86D600, 0xCAD51F504F994CC9}},
+ {.u64 = {0xC03B1789ECD74900, 0x725E2C9EB2FBA565}},
+},
+dsbd[2] = {
+ {.u64 = {0x7D57CCDFE6B1A200, 0xF56E9B13882A4439}},
+ {.u64 = {0x3CE2FAF724C6CB00, 0x2931180D15DEEFD3}},
+},
+dsbb[2] = {
+ {.u64 = {0xD022649296B44200, 0x602646F6B0F2D404}},
+ {.u64 = {0xC19498A6CD596700, 0xF3FF0C3E3255AA6B}},
+},
+dsbe[2] = {
+ {.u64 = {0x46F2929626D4D000, 0x2242600464B4F6B0}},
+ {.u64 = {0x0C55A6CDFFAAC100, 0x9467F36B98593E32}},
+},
+dsbo[2] = {
+ {.u64 = {0x1387EA537EF94000, 0xC7AA6DB9D4943E2D}},
+ {.u64 = {0x12D7560F93441D00, 0xCA4B8159D8C58E9C}},
+},
+dks1[2] = {
+ {.u64 = {0xB6116FC87ED9A700, 0x4AED933482255BFC}},
+ {.u64 = {0x4576516227143300, 0x8BB89FACE9DAFDCE}},
+},
+dks2[2] = {
+ {.u64 = {0x27438FEBCCA86400, 0x4622EE8AADC90561}},
+ {.u64 = {0x815C13CE4F92DD00, 0x73AEE13CBD602FF2}},
+},
+dks3[2] = {
+ {.u64 = {0x03C4C50201C6C700, 0xF83F3EF9FA3D3CFB}},
+ {.u64 = {0xEE1921D638CFF700, 0xA5526A9D7384BC4B}},
+},
+dks4[2] = {
+ {.u64 = {0xE3C390B053732000, 0xA080D3F310306343}},
+ {.u64 = {0xA0CA214B036982E8, 0x2F45AEC48CE60D67}},
+},
+deskew[2] = {
+ {.u64 = {0x07E4A34047A4E300, 0x1DFEB95A5DBEF91A}},
+ {.u64 = {0x5F36B5DC83EA6900, 0x2841C2ABF49D1E77}},
+},
+sr[4] = {
+ {.u64 = {0x0706050403020100, 0x0F0E0D0C0B0A0908}},
+ {.u64 = {0x030E09040F0A0500, 0x0B06010C07020D08}},
+ {.u64 = {0x0F060D040B020900, 0x070E050C030A0108}},
+ {.u64 = {0x0B0E0104070A0D00, 0x0306090C0F020508}},
+},
+rcon = {.u64 = {0x1F8391B9AF9DEEB6, 0x702A98084D7C7D81}},
+s63 = {.u64 = {0x5B5B5B5B5B5B5B5B, 0x5B5B5B5B5B5B5B5B}},
+of = {.u64 = {0x0F0F0F0F0F0F0F0F, 0x0F0F0F0F0F0F0F0F}},
+inv = {.u64 = {0x0E05060F0D080180, 0x040703090A0B0C02}},
+inva = {.u64 = {0x01040A060F0B0780, 0x030D0E0C02050809}};
+
+static inline __m128i
+loadroundkey(const uint32_t *rk32)
+{
+ return _mm_load_si128((const void *)rk32);
+}
+
+static inline void
+storeroundkey(uint32_t *rk32, __m128i rk)
+{
+ _mm_store_si128((void *)rk32, rk);
+}
+
+/* Given abcdefgh, set *lo = 0b0d0f0h and *hi = 0a0c0e0g. */
+static inline void
+bytes2nybbles(__m128i *restrict lo, __m128i *restrict hi, __m128i x)
+{
+
+ *lo = x & of.m;
+ *hi = _mm_srli_epi32(x & ~of.m, 4);
+}
+
+/* Given 0p0q0r0s, return 0x0y0z0w where x = a/p, y = a/q, &c. */
+static inline __m128i
+gf16_inva(__m128i x)
+{
+ return _mm_shuffle_epi8(inva.m, x);
+}
+
+/* Given 0p0q0r0s, return 0x0y0z0w where x = 1/p, y = 1/q, &c. */
+static inline __m128i
+gf16_inv(__m128i x)
+{
+ return _mm_shuffle_epi8(inv.m, x);
+}
+
+/*
+ * t is a pair of maps respectively from low and high nybbles to bytes.
+ * Apply t the nybbles, and add the results in GF(2).
+ */
+static __m128i
+aes_schedule_transform(__m128i x, const union m128const t[static 2])
+{
+ __m128i lo, hi;
+
+ bytes2nybbles(&lo, &hi, x);
+ return _mm_shuffle_epi8(t[0].m, lo) ^ _mm_shuffle_epi8(t[1].m, hi);
+}
+
+static inline void
+subbytes(__m128i *io, __m128i *jo, __m128i x)
+{
+ __m128i k, i, ak, j;
+
+ bytes2nybbles(&k, &i, x);
+ ak = gf16_inva(k);
+ j = i ^ k;
+ *io = j ^ gf16_inv(ak ^ gf16_inv(i));
+ *jo = i ^ gf16_inv(ak ^ gf16_inv(j));
+}
+
+static __m128i
+aes_schedule_low_round(__m128i rk, __m128i prk)
+{
+ __m128i io, jo;
+
+ /* smear prk */
+ prk ^= _mm_slli_si128(prk, 4);
+ prk ^= _mm_slli_si128(prk, 8);
+ prk ^= s63.m;
+
+ /* subbytes */
+ subbytes(&io, &jo, rk);
+ rk = _mm_shuffle_epi8(sb1[0].m, io) ^ _mm_shuffle_epi8(sb1[1].m, jo);
+
+ /* add in smeared stuff */
+ return rk ^ prk;
+}
+
+static __m128i
+aes_schedule_round(__m128i rk, __m128i prk, __m128i *rcon_rot)
+{
+
+ /* extract rcon from rcon_rot */
+ prk ^= _mm_alignr_epi8(_mm_setzero_si128(), *rcon_rot, 15);
+ *rcon_rot = _mm_alignr_epi8(*rcon_rot, *rcon_rot, 15);
+
+ /* rotate */
+ rk = _mm_shuffle_epi32(rk, 0xff);
+ rk = _mm_alignr_epi8(rk, rk, 1);
+
+ return aes_schedule_low_round(rk, prk);
+}
+
+static __m128i
+aes_schedule_mangle_enc(__m128i x, __m128i sr_i)
+{
+ __m128i y = _mm_setzero_si128();
+
+ x ^= s63.m;
+
+ x = _mm_shuffle_epi8(x, mc_forward[0].m);
+ y ^= x;
+ x = _mm_shuffle_epi8(x, mc_forward[0].m);
+ y ^= x;
+ x = _mm_shuffle_epi8(x, mc_forward[0].m);
+ y ^= x;
+
+ return _mm_shuffle_epi8(y, sr_i);
+}
+
+static __m128i
+aes_schedule_mangle_last_enc(__m128i x, __m128i sr_i)
+{
+
+ return aes_schedule_transform(_mm_shuffle_epi8(x, sr_i) ^ s63.m, opt);
+}
+
+static __m128i
+aes_schedule_mangle_dec(__m128i x, __m128i sr_i)
+{
+ __m128i y = _mm_setzero_si128();
+
+ x = aes_schedule_transform(x, dks1);
Home |
Main Index |
Thread Index |
Old Index