chacha20poly1305: import x86_64 primitives from OpenSSL

x86_64 only at the moment. SSSE3, AVX, AVX2, AVX512.

Signed-off-by: Samuel Neves <sneves@dei.uc.pt>

1 files changed, 334 insertions, 305 deletions

diff --git a/src/crypto/chacha20poly1305.c b/src/crypto/chacha20poly1305.c
index a9c3bf8..ac033f0 100644
--- a/src/crypto/chacha20poly1305.c
+++ b/src/crypto/chacha20poly1305.c
@@ -1,6 +1,34 @@
-/*
- * Copyright (C) 2015-2017 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
+/* Copyright (C) 2015-2017 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved.
  * Copyright 2015 Martin Willi.
+ * Copyright 2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * - Redistributions of source code must retain copyright notices,
+ * this list of conditions and the following disclaimer.
+ * - 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.
+ * - Neither the name of the CRYPTOGAMS nor the names of its
+ * copyright holder and contributors may be used to endorse or
+ * promote products derived from this software without specific
+ * prior written permission.
+ * ALTERNATIVELY, provided that this notice is retained in full, this
+ * product may be distributed under the terms of the GNU General Public
+ * License (GPL), in which case the provisions of the GPL apply INSTEAD OF
+ * those given above.
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER 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 COPYRIGHT
+ * OWNER 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.
  */
 
 #include "chacha20poly1305.h"
@@ -15,27 +43,39 @@
 #if defined(CONFIG_X86_64)
 #include <asm/cpufeature.h>
 #include <asm/processor.h>
+asmlinkage void poly1305_init_x86_64(void *ctx, const unsigned char key[16]);
+asmlinkage void poly1305_blocks_x86_64(void *ctx, const unsigned char *inp, size_t len, u32 padbit);
+asmlinkage void poly1305_emit_x86_64(void *ctx, unsigned char mac[16], const u32 nonce[4]);
 #ifdef CONFIG_AS_SSSE3
-asmlinkage void hchacha20_asm_ssse3(u8 *derived_key, const u8 *nonce, const u8 *key);
-asmlinkage void chacha20_asm_block_xor_ssse3(u32 *state, u8 *dst, const u8 *src);
-asmlinkage void chacha20_asm_4block_xor_ssse3(u32 *state, u8 *dst, const u8 *src);
+asmlinkage void hchacha20_ssse3(u8 *derived_key, const u8 *nonce, const u8 *key);
+asmlinkage void chacha20_ssse3(unsigned char *out, const unsigned char *in, size_t len, const unsigned int key[8], const unsigned int counter[4]);
 #endif
-#ifdef CONFIG_AS_AVX2
-asmlinkage void chacha20_asm_8block_xor_avx2(u32 *state, u8 *dst, const u8 *src);
+#ifdef CONFIG_AS_AVX
+asmlinkage void poly1305_emit_avx(void *ctx, u8 mac[16], const u32 nonce[4]);
+asmlinkage void poly1305_blocks_avx(void *ctx, const u8 *inp, size_t len, u32 padbit);
 #endif
-asmlinkage void poly1305_asm_block_sse2(u32 *h, const u8 *src, const u32 *r, unsigned int blocks);
-asmlinkage void poly1305_asm_2block_sse2(u32 *h, const u8 *src, const u32 *r, unsigned int blocks, const u32 *u);
 #ifdef CONFIG_AS_AVX2
-asmlinkage void poly1305_asm_4block_avx2(u32 *h, const u8 *src, const u32 *r, unsigned int blocks, const u32 *u);
+asmlinkage void chacha20_avx2(unsigned char *out, const unsigned char *in, size_t len, const unsigned int key[8], const unsigned int counter[4]);
+asmlinkage void poly1305_blocks_avx2(void *ctx, const unsigned char *inp, size_t len, u32 padbit);
 #endif
-static bool chacha20poly1305_use_avx2 __read_mostly;
+#ifdef CONFIG_AS_AVX512
+asmlinkage void chacha20_avx512(unsigned char *out, const unsigned char *in, size_t len, const unsigned int key[8], const unsigned int counter[4]);
+asmlinkage void poly1305_blocks_avx512(void *ctx, const unsigned char *inp, size_t len, u32 padbit);
+#endif
+
 static bool chacha20poly1305_use_ssse3 __read_mostly;
-static bool chacha20poly1305_use_sse2 __read_mostly;
+static bool chacha20poly1305_use_avx __read_mostly;
+static bool chacha20poly1305_use_avx2 __read_mostly;
+static bool chacha20poly1305_use_avx512 __read_mostly;
+
 void chacha20poly1305_fpu_init(void)
 {
-	chacha20poly1305_use_sse2 = boot_cpu_has(X86_FEATURE_XMM2);
 	chacha20poly1305_use_ssse3 = boot_cpu_has(X86_FEATURE_SSSE3);
+	chacha20poly1305_use_avx = boot_cpu_has(X86_FEATURE_AVX) && cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM, NULL);
 	chacha20poly1305_use_avx2 = boot_cpu_has(X86_FEATURE_AVX) && boot_cpu_has(X86_FEATURE_AVX2) && cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM, NULL);
+#ifndef COMPAT_CANNOT_USE_AVX512
+	chacha20poly1305_use_avx512 = boot_cpu_has(X86_FEATURE_AVX) && boot_cpu_has(X86_FEATURE_AVX2) && boot_cpu_has(X86_FEATURE_AVX512F) && cpu_has_xfeatures(XFEATURE_MASK_SSE | XFEATURE_MASK_YMM | XFEATURE_MASK_ZMM_Hi256, NULL);
+#endif
 }
 #elif IS_ENABLED(CONFIG_KERNEL_MODE_NEON)
 #include <asm/hwcap.h>
@@ -77,21 +117,6 @@ static inline u32 rotl32(u32 v, u8 n)
 	return (v << n) | (v >> (sizeof(v) * 8 - n));
 }
 
-static inline u64 mlt(u64 a, u64 b)
-{
-	return a * b;
-}
-
-static inline u32 sr(u64 v, u_char n)
-{
-	return v >> n;
-}
-
-static inline u32 and(u32 v, u32 mask)
-{
-	return v & mask;
-}
-
 struct chacha20_ctx {
 	u32 state[CHACHA20_BLOCK_SIZE / sizeof(u32)];
 } __aligned(32);
@@ -232,17 +257,13 @@ static void hchacha20_generic(u8 derived_key[CHACHA20POLY1305_KEYLEN], const u8
 
 static inline void hchacha20(u8 derived_key[CHACHA20POLY1305_KEYLEN], const u8 nonce[16], const u8 key[CHACHA20POLY1305_KEYLEN], bool have_simd)
 {
-	if (!have_simd)
-		goto no_simd;
-
 #if defined(CONFIG_X86_64) && defined(CONFIG_AS_SSSE3)
-	if (chacha20poly1305_use_ssse3) {
-		hchacha20_asm_ssse3(derived_key, nonce, key);
+	if (have_simd && chacha20poly1305_use_ssse3) {
+		hchacha20_ssse3(derived_key, nonce, key);
 		return;
 	}
 #endif
 
-no_simd:
 	hchacha20_generic(derived_key, nonce, key);
 }
 
@@ -266,7 +287,7 @@ static void chacha20_keysetup(struct chacha20_ctx *ctx, const u8 key[CHACHA20_KE
 	ctx->state[15] = le32_to_cpuvp(nonce + 4);
 }
 
-static void chacha20_crypt(struct chacha20_ctx *ctx, u8 *dst, const u8 *src, unsigned int bytes, bool have_simd)
+static void chacha20_crypt(struct chacha20_ctx *ctx, u8 *dst, const u8 *src, u32 bytes, bool have_simd)
 {
 	u8 buf[CHACHA20_BLOCK_SIZE];
 
@@ -281,37 +302,23 @@ static void chacha20_crypt(struct chacha20_ctx *ctx, u8 *dst, const u8 *src, uns
 		goto no_simd;
 
 #if defined(CONFIG_X86_64)
+#ifdef CONFIG_AS_AVX512
+	if (chacha20poly1305_use_avx512) {
+		chacha20_avx512(dst, src, bytes, &ctx->state[4], &ctx->state[12]);
+		ctx->state[12] += (bytes + 63) / 64;
+		return;
+	}
+#endif
 #ifdef CONFIG_AS_AVX2
 	if (chacha20poly1305_use_avx2) {
-		while (bytes >= CHACHA20_BLOCK_SIZE * 8) {
-			chacha20_asm_8block_xor_avx2(ctx->state, dst, src);
-			bytes -= CHACHA20_BLOCK_SIZE * 8;
-			src += CHACHA20_BLOCK_SIZE * 8;
-			dst += CHACHA20_BLOCK_SIZE * 8;
-			ctx->state[12] += 8;
-		}
+		chacha20_avx2(dst, src, bytes, &ctx->state[4], &ctx->state[12]);
+		ctx->state[12] += (bytes + 63) / 64;
+		return;
 	}
 #endif
 #ifdef CONFIG_AS_SSSE3
-	while (bytes >= CHACHA20_BLOCK_SIZE * 4) {
-		chacha20_asm_4block_xor_ssse3(ctx->state, dst, src);
-		bytes -= CHACHA20_BLOCK_SIZE * 4;
-		src += CHACHA20_BLOCK_SIZE * 4;
-		dst += CHACHA20_BLOCK_SIZE * 4;
-		ctx->state[12] += 4;
-	}
-	while (bytes >= CHACHA20_BLOCK_SIZE) {
-		chacha20_asm_block_xor_ssse3(ctx->state, dst, src);
-		bytes -= CHACHA20_BLOCK_SIZE;
-		src += CHACHA20_BLOCK_SIZE;
-		dst += CHACHA20_BLOCK_SIZE;
-		ctx->state[12]++;
-	}
-	if (bytes) {
-		memcpy(buf, src, bytes);
-		chacha20_asm_block_xor_ssse3(ctx->state, buf, buf);
-		memcpy(dst, buf, bytes);
-	}
+	chacha20_ssse3(dst, src, bytes, &ctx->state[4], &ctx->state[12]);
+	ctx->state[12] += (bytes + 63) / 64;
 	return;
 #endif
 #elif IS_ENABLED(CONFIG_KERNEL_MODE_NEON)
@@ -352,261 +359,283 @@ no_simd:
 		crypto_xor(dst, buf, bytes);
 	}
 }
+typedef void (*poly1305_blocks_f)(void *ctx, const u8 *inp, size_t len, u32 padbit);
+typedef void (*poly1305_emit_f)(void *ctx, u8 mac[16], const u32 nonce[4]);
 
 struct poly1305_ctx {
-	/* key */
-	u32 r[5];
-	/* finalize key */
-	u32 s[4];
-	/* accumulator */
+	u8 opaque[24 * sizeof(u64)];
+	u32 nonce[4];
+	u8 data[POLY1305_BLOCK_SIZE];
+	size_t num;
+	struct {
+		poly1305_blocks_f blocks;
+		poly1305_emit_f emit;
+	} func;
+} __aligned(8);
+
+#ifndef CONFIG_X86_64
+struct poly1305_internal {
 	u32 h[5];
-	/* partial buffer */
-	u8 buf[POLY1305_BLOCK_SIZE];
-	/* bytes used in partial buffer */
-	unsigned int buflen;
-	/* derived key u set? */
-	bool uset;
-	/* derived keys r^3, r^4 set? */
-	bool wset;
-	/* derived Poly1305 key r^2 */
-	u32 u[5];
-	/* derived Poly1305 key r^3 */
-	u32 r3[5];
-	/* derived Poly1305 key r^4 */
-	u32 r4[5];
+	u32 r[4];
 };
 
-static void poly1305_init(struct poly1305_ctx *ctx, const u8 key[POLY1305_KEY_SIZE])
+static void poly1305_init_generic(void *ctx, const u8 key[16])
 {
-	memset(ctx, 0, sizeof(struct poly1305_ctx));
+	struct poly1305_internal *st = (struct poly1305_internal *)ctx;
+
+	/* h = 0 */
+	st->h[0] = 0;
+	st->h[1] = 0;
+	st->h[2] = 0;
+	st->h[3] = 0;
+	st->h[4] = 0;
+
 	/* r &= 0xffffffc0ffffffc0ffffffc0fffffff */
-	ctx->r[0] = (le32_to_cpuvp(key +  0) >> 0) & 0x3ffffff;
-	ctx->r[1] = (get_unaligned_le32(key +  3) >> 2) & 0x3ffff03;
-	ctx->r[2] = (get_unaligned_le32(key +  6) >> 4) & 0x3ffc0ff;
-	ctx->r[3] = (get_unaligned_le32(key +  9) >> 6) & 0x3f03fff;
-	ctx->r[4] = (le32_to_cpuvp(key + 12) >> 8) & 0x00fffff;
-	ctx->s[0] = le32_to_cpuvp(key +  16);
-	ctx->s[1] = le32_to_cpuvp(key +  20);
-	ctx->s[2] = le32_to_cpuvp(key +  24);
-	ctx->s[3] = le32_to_cpuvp(key +  28);
+	st->r[0] = le32_to_cpuvp(&key[ 0]) & 0x0fffffff;
+	st->r[1] = le32_to_cpuvp(&key[ 4]) & 0x0ffffffc;
+	st->r[2] = le32_to_cpuvp(&key[ 8]) & 0x0ffffffc;
+	st->r[3] = le32_to_cpuvp(&key[12]) & 0x0ffffffc;
 }
 
-static unsigned int poly1305_generic_blocks(struct poly1305_ctx *ctx, const u8 *src, unsigned int srclen, u32 hibit)
+static void
+poly1305_blocks_generic(void *ctx, const u8 *inp, size_t len, u32 padbit)
 {
-	u32 r0, r1, r2, r3, r4;
-	u32 s1, s2, s3, s4;
-	u32 h0, h1, h2, h3, h4;
-	u64 d0, d1, d2, d3, d4;
-
-	r0 = ctx->r[0];
-	r1 = ctx->r[1];
-	r2 = ctx->r[2];
-	r3 = ctx->r[3];
-	r4 = ctx->r[4];
-
-	s1 = r1 * 5;
-	s2 = r2 * 5;
-	s3 = r3 * 5;
-	s4 = r4 * 5;
-
-	h0 = ctx->h[0];
-	h1 = ctx->h[1];
-	h2 = ctx->h[2];
-	h3 = ctx->h[3];
-	h4 = ctx->h[4];
-
-	while (likely(srclen >= POLY1305_BLOCK_SIZE)) {
+#define CONSTANT_TIME_CARRY(a,b) ((a ^ ((a ^ b) | ((a - b) ^ b))) >> (sizeof(a) * 8 - 1))
+	struct poly1305_internal *st = (struct poly1305_internal *)ctx;
+	u32 r0, r1, r2, r3;
+	u32 s1, s2, s3;
+	u32 h0, h1, h2, h3, h4, c;
+	u64 d0, d1, d2, d3;
+
+	r0 = st->r[0];
+	r1 = st->r[1];
+	r2 = st->r[2];
+	r3 = st->r[3];
+
+	s1 = r1 + (r1 >> 2);
+	s2 = r2 + (r2 >> 2);
+	s3 = r3 + (r3 >> 2);
+
+	h0 = st->h[0];
+	h1 = st->h[1];
+	h2 = st->h[2];
+	h3 = st->h[3];
+	h4 = st->h[4];
+
+	while (len >= POLY1305_BLOCK_SIZE) {
 		/* h += m[i] */
-		h0 += (le32_to_cpuvp(src +  0) >> 0) & 0x3ffffff;
-		h1 += (get_unaligned_le32(src +  3) >> 2) & 0x3ffffff;
-		h2 += (get_unaligned_le32(src +  6) >> 4) & 0x3ffffff;
-		h3 += (get_unaligned_le32(src +  9) >> 6) & 0x3ffffff;
-		h4 += (le32_to_cpuvp(src + 12) >> 8) | hibit;
-
-		/* h *= r */
-		d0 = mlt(h0, r0) + mlt(h1, s4) + mlt(h2, s3) + mlt(h3, s2) + mlt(h4, s1);
-		d1 = mlt(h0, r1) + mlt(h1, r0) + mlt(h2, s4) + mlt(h3, s3) + mlt(h4, s2);
-		d2 = mlt(h0, r2) + mlt(h1, r1) + mlt(h2, r0) + mlt(h3, s4) + mlt(h4, s3);
-		d3 = mlt(h0, r3) + mlt(h1, r2) + mlt(h2, r1) + mlt(h3, r0) + mlt(h4, s4);
-		d4 = mlt(h0, r4) + mlt(h1, r3) + mlt(h2, r2) + mlt(h3, r1) + mlt(h4, r0);
-
-		/* (partial) h %= p */
-		d1 += sr(d0, 26);     h0 = and(d0, 0x3ffffff);
-		d2 += sr(d1, 26);     h1 = and(d1, 0x3ffffff);
-		d3 += sr(d2, 26);     h2 = and(d2, 0x3ffffff);
-		d4 += sr(d3, 26);     h3 = and(d3, 0x3ffffff);
-		h0 += sr(d4, 26) * 5; h4 = and(d4, 0x3ffffff);
-		h1 += h0 >> 26;       h0 = h0 & 0x3ffffff;
-
-		src += POLY1305_BLOCK_SIZE;
-		srclen -= POLY1305_BLOCK_SIZE;
+		h0 = (u32)(d0 = (u64)h0 + le32_to_cpuvp(inp + 0));
+		h1 = (u32)(d1 = (u64)h1 + (d0 >> 32) + le32_to_cpuvp(inp + 4));
+		h2 = (u32)(d2 = (u64)h2 + (d1 >> 32) + le32_to_cpuvp(inp + 8));
+		h3 = (u32)(d3 = (u64)h3 + (d2 >> 32) + le32_to_cpuvp(inp + 12));
+		h4 += (u32)(d3 >> 32) + padbit;
+
+		/* h *= r "%" p, where "%" stands for "partial remainder" */
+		d0 = ((u64)h0 * r0) +
+		     ((u64)h1 * s3) +
+		     ((u64)h2 * s2) +
+		     ((u64)h3 * s1);
+		d1 = ((u64)h0 * r1) +
+		     ((u64)h1 * r0) +
+		     ((u64)h2 * s3) +
+		     ((u64)h3 * s2) +
+		     (h4 * s1);
+		d2 = ((u64)h0 * r2) +
+		     ((u64)h1 * r1) +
+		     ((u64)h2 * r0) +
+		     ((u64)h3 * s3) +
+		     (h4 * s2);
+		d3 = ((u64)h0 * r3) +
+		     ((u64)h1 * r2) +
+		     ((u64)h2 * r1) +
+		     ((u64)h3 * r0) +
+		     (h4 * s3);
+		h4 = (h4 * r0);
+
+		/* last reduction step: */
+		/* a) h4:h0 = h4<<128 + d3<<96 + d2<<64 + d1<<32 + d0 */
+		h0 = (u32)d0;
+		h1 = (u32)(d1 += d0 >> 32);
+		h2 = (u32)(d2 += d1 >> 32);
+		h3 = (u32)(d3 += d2 >> 32);
+		h4 += (u32)(d3 >> 32);
+		/* b) (h4:h0 += (h4:h0>>130) * 5) %= 2^130 */
+		c = (h4 >> 2) + (h4 & ~3U);
+		h4 &= 3;
+		h0 += c;
+		h1 += (c = CONSTANT_TIME_CARRY(h0,c));
+		h2 += (c = CONSTANT_TIME_CARRY(h1,c));
+		h3 += (c = CONSTANT_TIME_CARRY(h2,c));
+		h4 += CONSTANT_TIME_CARRY(h3,c);
+		/*
+		 * Occasional overflows to 3rd bit of h4 are taken care of
+		 * "naturally". If after this point we end up at the top of
+		 * this loop, then the overflow bit will be accounted for
+		 * in next iteration. If we end up in poly1305_emit, then
+		 * comparison to modulus below will still count as "carry
+		 * into 131st bit", so that properly reduced value will be
+		 * picked in conditional move.
+		 */
+
+		inp += POLY1305_BLOCK_SIZE;
+		len -= POLY1305_BLOCK_SIZE;
 	}
 
-	ctx->h[0] = h0;
-	ctx->h[1] = h1;
-	ctx->h[2] = h2;
-	ctx->h[3] = h3;
-	ctx->h[4] = h4;
-
-	return srclen;
+	st->h[0] = h0;
+	st->h[1] = h1;
+	st->h[2] = h2;
+	st->h[3] = h3;
+	st->h[4] = h4;
+#undef CONSTANT_TIME_CARRY
 }
 
-#ifdef CONFIG_X86_64
-static void poly1305_simd_mult(u32 *a, const u32 *b)
+static void poly1305_emit_generic(void *ctx, u8 mac[16], const u32 nonce[4])
 {
-	u8 m[POLY1305_BLOCK_SIZE];
-
-	memset(m, 0, sizeof(m));
-	/* The poly1305 block function adds a hi-bit to the accumulator which
-	 * we don't need for key multiplication; compensate for it.
-	 */
-	a[4] -= 1U << 24;
-	poly1305_asm_block_sse2(a, m, b, 1);
+	struct poly1305_internal *st = (struct poly1305_internal *)ctx;
+	__le32 *omac = (__force __le32 *)mac;
+	u32 h0, h1, h2, h3, h4;
+	u32 g0, g1, g2, g3, g4;
+	u64 t;
+	u32 mask;
+
+	h0 = st->h[0];
+	h1 = st->h[1];
+	h2 = st->h[2];
+	h3 = st->h[3];
+	h4 = st->h[4];
+
+	/* compare to modulus by computing h + -p */
+	g0 = (u32)(t = (u64)h0 + 5);
+	g1 = (u32)(t = (u64)h1 + (t >> 32));
+	g2 = (u32)(t = (u64)h2 + (t >> 32));
+	g3 = (u32)(t = (u64)h3 + (t >> 32));
+	g4 = h4 + (u32)(t >> 32);
+
+	/* if there was carry into 131st bit, h3:h0 = g3:g0 */
+	mask = 0 - (g4 >> 2);
+	g0 &= mask;
+	g1 &= mask;
+	g2 &= mask;
+	g3 &= mask;
+	mask = ~mask;
+	h0 = (h0 & mask) | g0;
+	h1 = (h1 & mask) | g1;
+	h2 = (h2 & mask) | g2;
+	h3 = (h3 & mask) | g3;
+
+	/* mac = (h + nonce) % (2^128) */
+	h0 = (u32)(t = (u64)h0 + nonce[0]);
+	h1 = (u32)(t = (u64)h1 + (t >> 32) + nonce[1]);
+	h2 = (u32)(t = (u64)h2 + (t >> 32) + nonce[2]);
+	h3 = (u32)(t = (u64)h3 + (t >> 32) + nonce[3]);
+
+	omac[0] = cpu_to_le32(h0);
+	omac[1] = cpu_to_le32(h1);
+	omac[2] = cpu_to_le32(h2);
+	omac[3] = cpu_to_le32(h3);
 }
+#endif /* !CONFIG_X86_64 */
 
-static unsigned int poly1305_simd_blocks(struct poly1305_ctx *ctx, const u8 *src, unsigned int srclen)
+void poly1305_init(struct poly1305_ctx *ctx, const u8 key[POLY1305_KEY_SIZE], bool have_simd)
 {
-	unsigned int blocks;
+	ctx->nonce[0] = le32_to_cpuvp(&key[16]);
+	ctx->nonce[1] = le32_to_cpuvp(&key[20]);
+	ctx->nonce[2] = le32_to_cpuvp(&key[24]);
+	ctx->nonce[3] = le32_to_cpuvp(&key[28]);
 
+#ifdef CONFIG_X86_64
+	poly1305_init_x86_64(ctx->opaque, key);
+	ctx->func.blocks = poly1305_blocks_x86_64;
+	ctx->func.emit = poly1305_emit_x86_64;
+#ifdef CONFIG_AS_AVX512
+	if(chacha20poly1305_use_avx512 && have_simd) {
+		ctx->func.blocks = poly1305_blocks_avx512;
+		ctx->func.emit = poly1305_emit_avx;
+	} else
+#endif
 #ifdef CONFIG_AS_AVX2
-	if (chacha20poly1305_use_avx2 && srclen >= POLY1305_BLOCK_SIZE * 4) {
-		if (unlikely(!ctx->wset)) {
-			if (!ctx->uset) {
-				memcpy(ctx->u, ctx->r, sizeof(ctx->u));
-				poly1305_simd_mult(ctx->u, ctx->r);
-				ctx->uset = true;
-			}
-			memcpy(ctx->r3, ctx->u, sizeof(ctx->u));
-			poly1305_simd_mult(ctx->r3, ctx->r);
-			memcpy(ctx->r4, ctx->r3, sizeof(ctx->u));
-			poly1305_simd_mult(ctx->r4, ctx->r);
-			ctx->wset = true;
-		}
-		blocks = srclen / (POLY1305_BLOCK_SIZE * 4);
-		poly1305_asm_4block_avx2(ctx->h, src, ctx->r, blocks, ctx->u);
-		src += POLY1305_BLOCK_SIZE * 4 * blocks;
-		srclen -= POLY1305_BLOCK_SIZE * 4 * blocks;
-	}
+	if (chacha20poly1305_use_avx2 && have_simd) {
+		ctx->func.blocks = poly1305_blocks_avx2;
+		ctx->func.emit = poly1305_emit_avx;
+	} else
 #endif
-	if (likely(srclen >= POLY1305_BLOCK_SIZE * 2)) {
-		if (unlikely(!ctx->uset)) {
-			memcpy(ctx->u, ctx->r, sizeof(ctx->u));
-			poly1305_simd_mult(ctx->u, ctx->r);
-			ctx->uset = true;
-		}
-		blocks = srclen / (POLY1305_BLOCK_SIZE * 2);
-		poly1305_asm_2block_sse2(ctx->h, src, ctx->r, blocks, ctx->u);
-		src += POLY1305_BLOCK_SIZE * 2 * blocks;
-		srclen -= POLY1305_BLOCK_SIZE * 2 * blocks;
-	}
-	if (srclen >= POLY1305_BLOCK_SIZE) {
-		poly1305_asm_block_sse2(ctx->h, src, ctx->r, 1);
-		srclen -= POLY1305_BLOCK_SIZE;
+#ifdef CONFIG_AS_AVX
+	if (chacha20poly1305_use_avx && have_simd) {
+		ctx->func.blocks = poly1305_blocks_avx;
+		ctx->func.emit = poly1305_emit_avx;
 	}
-	return srclen;
-}
 #endif
+#else
+	poly1305_init_generic(ctx->opaque, key);
+#endif
+	ctx->num = 0;
+}
 
-static void poly1305_update(struct poly1305_ctx *ctx, const u8 *src, unsigned int srclen, bool have_simd)
+void poly1305_update(struct poly1305_ctx *ctx, const u8 *inp, size_t len)
 {
-	unsigned int bytes;
-
-	if (unlikely(ctx->buflen)) {
-		bytes = min(srclen, POLY1305_BLOCK_SIZE - ctx->buflen);
-		memcpy(ctx->buf + ctx->buflen, src, bytes);
-		src += bytes;
-		srclen -= bytes;
-		ctx->buflen += bytes;
-
-		if (ctx->buflen == POLY1305_BLOCK_SIZE) {
 #ifdef CONFIG_X86_64
-			if (have_simd && chacha20poly1305_use_sse2)
-				poly1305_simd_blocks(ctx, ctx->buf, POLY1305_BLOCK_SIZE);
-			else
+	const poly1305_blocks_f blocks = ctx->func.blocks;
+#else
+	const poly1305_blocks_f blocks = poly1305_blocks_generic;
 #endif
-				poly1305_generic_blocks(ctx, ctx->buf, POLY1305_BLOCK_SIZE, 1U << 24);
-			ctx->buflen = 0;
+
+	const size_t num = ctx->num;
+	size_t rem;;
+
+	if (num) {
+		rem = POLY1305_BLOCK_SIZE - num;
+		if (len >= rem) {
+			memcpy(ctx->data + num, inp, rem);
+			blocks(ctx->opaque, ctx->data, POLY1305_BLOCK_SIZE, 1);
+			inp += rem;
+			len -= rem;
+		} else {
+			/* Still not enough data to process a block. */
+			memcpy(ctx->data + num, inp, len);
+			ctx->num = num + len;
+			return;
 		}
 	}
 
-	if (likely(srclen >= POLY1305_BLOCK_SIZE)) {
-#ifdef CONFIG_X86_64
-		if (have_simd && chacha20poly1305_use_sse2)
-			bytes = poly1305_simd_blocks(ctx, src, srclen);
-		else
-#endif
-			bytes = poly1305_generic_blocks(ctx, src, srclen, 1U << 24);
-		src += srclen - bytes;
-		srclen = bytes;
-	}
+	rem = len % POLY1305_BLOCK_SIZE;
+	len -= rem;
 
-	if (unlikely(srclen)) {
-		ctx->buflen = srclen;
-		memcpy(ctx->buf, src, srclen);
+	if (len >= POLY1305_BLOCK_SIZE) {
+		blocks(ctx->opaque, inp, len, 1);
+		inp += len;
 	}
+
+	if (rem)
+		memcpy(ctx->data, inp, rem);
+
+	ctx->num = rem;
 }
 
-static void poly1305_finish(struct poly1305_ctx *ctx, u8 *dst)
+void poly1305_finish(struct poly1305_ctx * ctx, u8 mac[16])
 {
-	__le32 *mac = (__le32 *)dst;
-	u32 h0, h1, h2, h3, h4;
-	u32 g0, g1, g2, g3, g4;
-	u32 mask;
-	u64 f = 0;
+#ifdef CONFIG_X86_64
+	poly1305_blocks_f blocks = ctx->func.blocks;
+	poly1305_emit_f emit = ctx->func.emit;
+#else
+	poly1305_blocks_f blocks = poly1305_blocks_generic;
+	poly1305_emit_f emit = poly1305_emit_generic;
+#endif
+	size_t num = ctx->num;
 
-	if (unlikely(ctx->buflen)) {
-		ctx->buf[ctx->buflen++] = 1;
-		memset(ctx->buf + ctx->buflen, 0, POLY1305_BLOCK_SIZE - ctx->buflen);
-		poly1305_generic_blocks(ctx, ctx->buf, POLY1305_BLOCK_SIZE, 0);
+	if (num) {
+		ctx->data[num++] = 1;   /* pad bit */
+		while (num < POLY1305_BLOCK_SIZE)
+			ctx->data[num++] = 0;
+		blocks(ctx->opaque, ctx->data, POLY1305_BLOCK_SIZE, 0);
 	}
 
-	/* fully carry h */
-	h0 = ctx->h[0];
-	h1 = ctx->h[1];
-	h2 = ctx->h[2];
-	h3 = ctx->h[3];
-	h4 = ctx->h[4];
-
-	h2 += (h1 >> 26);     h1 = h1 & 0x3ffffff;
-	h3 += (h2 >> 26);     h2 = h2 & 0x3ffffff;
-	h4 += (h3 >> 26);     h3 = h3 & 0x3ffffff;
-	h0 += (h4 >> 26) * 5; h4 = h4 & 0x3ffffff;
-	h1 += (h0 >> 26);     h0 = h0 & 0x3ffffff;
-
-	/* compute h + -p */
-	g0 = h0 + 5;
-	g1 = h1 + (g0 >> 26);              g0 &= 0x3ffffff;
-	g2 = h2 + (g1 >> 26);              g1 &= 0x3ffffff;
-	g3 = h3 + (g2 >> 26);              g2 &= 0x3ffffff;
-	g4 = h4 + (g3 >> 26) - (1U << 26); g3 &= 0x3ffffff;
-
-	/* select h if h < p, or h + -p if h >= p */
-	mask = (g4 >> ((sizeof(u32) * 8) - 1)) - 1;
-	g0 &= mask;
-	g1 &= mask;
-	g2 &= mask;
-	g3 &= mask;
-	g4 &= mask;
-	mask = ~mask;
-	h0 = (h0 & mask) | g0;
-	h1 = (h1 & mask) | g1;
-	h2 = (h2 & mask) | g2;
-	h3 = (h3 & mask) | g3;
-	h4 = (h4 & mask) | g4;
-
-	/* h = h % (2^128) */
-	h0 = (h0 >>  0) | (h1 << 26);
-	h1 = (h1 >>  6) | (h2 << 20);
-	h2 = (h2 >> 12) | (h3 << 14);
-	h3 = (h3 >> 18) | (h4 <<  8);
-
-	/* mac = (h + s) % (2^128) */
-	f = (f >> 32) + h0 + ctx->s[0]; mac[0] = cpu_to_le32(f);
-	f = (f >> 32) + h1 + ctx->s[1]; mac[1] = cpu_to_le32(f);
-	f = (f >> 32) + h2 + ctx->s[2]; mac[2] = cpu_to_le32(f);
-	f = (f >> 32) + h3 + ctx->s[3]; mac[3] = cpu_to_le32(f);
+	emit(ctx->opaque, mac, ctx->nonce);
+
+	/* zero out the state */
+	memzero_explicit(ctx, sizeof(*ctx));
 }
 
+
 static const u8 pad0[16] = { 0 };
 
 static struct crypto_alg chacha20_alg = {
@@ -636,22 +665,22 @@ static inline void __chacha20poly1305_encrypt(u8 *dst, const u8 *src, const size
 	chacha20_keysetup(&chacha20_state, key, (u8 *)&le_nonce);
 
 	chacha20_crypt(&chacha20_state, block0, block0, sizeof(block0), have_simd);
-	poly1305_init(&poly1305_state, block0);
+	poly1305_init(&poly1305_state, block0, have_simd);
 	memzero_explicit(block0, sizeof(block0));
 
-	poly1305_update(&poly1305_state, ad, ad_len, have_simd);
-	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, ad, ad_len);
+	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf);
 
 	chacha20_crypt(&chacha20_state, dst, src, src_len, have_simd);
 
-	poly1305_update(&poly1305_state, dst, src_len, have_simd);
-	poly1305_update(&poly1305_state, pad0, (0x10 - src_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, dst, src_len);
+	poly1305_update(&poly1305_state, pad0, (0x10 - src_len) & 0xf);
 
 	len = cpu_to_le64(ad_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	len = cpu_to_le64(src_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	poly1305_finish(&poly1305_state, dst + src_len);
 
@@ -687,11 +716,11 @@ bool chacha20poly1305_encrypt_sg(struct scatterlist *dst, struct scatterlist *sr
 	chacha20_keysetup(&chacha20_state, key, (u8 *)&le_nonce);
 
 	chacha20_crypt(&chacha20_state, block0, block0, sizeof(block0), have_simd);
-	poly1305_init(&poly1305_state, block0);
+	poly1305_init(&poly1305_state, block0, have_simd);
 	memzero_explicit(block0, sizeof(block0));
 
-	poly1305_update(&poly1305_state, ad, ad_len, have_simd);
-	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, ad, ad_len);
+	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf);
 
 	if (likely(src_len)) {
 		blkcipher_walk_init(&walk, dst, src, src_len);
@@ -700,25 +729,25 @@ bool chacha20poly1305_encrypt_sg(struct scatterlist *dst, struct scatterlist *sr
 			size_t chunk_len = rounddown(walk.nbytes, CHACHA20_BLOCK_SIZE);
 
 			chacha20_crypt(&chacha20_state, walk.dst.virt.addr, walk.src.virt.addr, chunk_len, have_simd);
-			poly1305_update(&poly1305_state, walk.dst.virt.addr, chunk_len, have_simd);
+			poly1305_update(&poly1305_state, walk.dst.virt.addr, chunk_len);
 			ret = blkcipher_walk_done(&chacha20_desc, &walk, walk.nbytes % CHACHA20_BLOCK_SIZE);
 		}
 		if (walk.nbytes) {
 			chacha20_crypt(&chacha20_state, walk.dst.virt.addr, walk.src.virt.addr, walk.nbytes, have_simd);
-			poly1305_update(&poly1305_state, walk.dst.virt.addr, walk.nbytes, have_simd);
+			poly1305_update(&poly1305_state, walk.dst.virt.addr, walk.nbytes);
 			ret = blkcipher_walk_done(&chacha20_desc, &walk, 0);
 		}
 	}
 	if (unlikely(ret))
 		goto err;
 
-	poly1305_update(&poly1305_state, pad0, (0x10 - src_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, pad0, (0x10 - src_len) & 0xf);
 
 	len = cpu_to_le64(ad_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	len = cpu_to_le64(src_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	poly1305_finish(&poly1305_state, mac);
 	scatterwalk_map_and_copy(mac, dst, src_len, sizeof(mac), 1);
@@ -749,21 +778,21 @@ static inline bool __chacha20poly1305_decrypt(u8 *dst, const u8 *src, const size
 	chacha20_keysetup(&chacha20_state, key, (u8 *)&le_nonce);
 
 	chacha20_crypt(&chacha20_state, block0, block0, sizeof(block0), have_simd);
-	poly1305_init(&poly1305_state, block0);
+	poly1305_init(&poly1305_state, block0, have_simd);
 	memzero_explicit(block0, sizeof(block0));
 
-	poly1305_update(&poly1305_state, ad, ad_len, have_simd);
-	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, ad, ad_len);
+	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf);
 
 	dst_len = src_len - POLY1305_MAC_SIZE;
-	poly1305_update(&poly1305_state, src, dst_len, have_simd);
-	poly1305_update(&poly1305_state, pad0, (0x10 - dst_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, src, dst_len);
+	poly1305_update(&poly1305_state, pad0, (0x10 - dst_len) & 0xf);
 
 	len = cpu_to_le64(ad_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	len = cpu_to_le64(dst_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	poly1305_finish(&poly1305_state, mac);
 	memzero_explicit(&poly1305_state, sizeof(poly1305_state));
@@ -811,11 +840,11 @@ bool chacha20poly1305_decrypt_sg(struct scatterlist *dst, struct scatterlist *sr
 	chacha20_keysetup(&chacha20_state, key, (u8 *)&le_nonce);
 
 	chacha20_crypt(&chacha20_state, block0, block0, sizeof(block0), have_simd);
-	poly1305_init(&poly1305_state, block0);
+	poly1305_init(&poly1305_state, block0, have_simd);
 	memzero_explicit(block0, sizeof(block0));
 
-	poly1305_update(&poly1305_state, ad, ad_len, have_simd);
-	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, ad, ad_len);
+	poly1305_update(&poly1305_state, pad0, (0x10 - ad_len) & 0xf);
 
 	dst_len = src_len - POLY1305_MAC_SIZE;
 	if (likely(dst_len)) {
@@ -824,12 +853,12 @@ bool chacha20poly1305_decrypt_sg(struct scatterlist *dst, struct scatterlist *sr
 		while (walk.nbytes >= CHACHA20_BLOCK_SIZE) {
 			size_t chunk_len = rounddown(walk.nbytes, CHACHA20_BLOCK_SIZE);
 
-			poly1305_update(&poly1305_state, walk.src.virt.addr, chunk_len, have_simd);
+			poly1305_update(&poly1305_state, walk.src.virt.addr, chunk_len);
 			chacha20_crypt(&chacha20_state, walk.dst.virt.addr, walk.src.virt.addr, chunk_len, have_simd);
 			ret = blkcipher_walk_done(&chacha20_desc, &walk, walk.nbytes % CHACHA20_BLOCK_SIZE);
 		}
 		if (walk.nbytes) {
-			poly1305_update(&poly1305_state, walk.src.virt.addr, walk.nbytes, have_simd);
+			poly1305_update(&poly1305_state, walk.src.virt.addr, walk.nbytes);
 			chacha20_crypt(&chacha20_state, walk.dst.virt.addr, walk.src.virt.addr, walk.nbytes, have_simd);
 			ret = blkcipher_walk_done(&chacha20_desc, &walk, 0);
 		}
@@ -837,13 +866,13 @@ bool chacha20poly1305_decrypt_sg(struct scatterlist *dst, struct scatterlist *sr
 	if (unlikely(ret))
 		goto err;
 
-	poly1305_update(&poly1305_state, pad0, (0x10 - dst_len) & 0xf, have_simd);
+	poly1305_update(&poly1305_state, pad0, (0x10 - dst_len) & 0xf);
 
 	len = cpu_to_le64(ad_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	len = cpu_to_le64(dst_len);
-	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len), have_simd);
+	poly1305_update(&poly1305_state, (u8 *)&len, sizeof(len));
 
 	poly1305_finish(&poly1305_state, computed_mac);
 	memzero_explicit(&poly1305_state, sizeof(poly1305_state));