intermediate changes

ref:cde9a383711a11544ce7e107a78147fb96cc4029

6 files changed, 1533 insertions, 0 deletions

diff --git a/contrib/libs/openssl/crypto/poly1305/poly1305.c b/contrib/libs/openssl/crypto/poly1305/poly1305.c
new file mode 100644
index 0000000000..e7f5b92c8f
--- /dev/null
+++ b/contrib/libs/openssl/crypto/poly1305/poly1305.c
@@ -0,0 +1,531 @@
+/*
+ * Copyright 2015-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License").  You may not use
+ * this file except in compliance with the License.  You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include <openssl/crypto.h>
+
+#include "crypto/poly1305.h"
+#include "poly1305_local.h"
+
+size_t Poly1305_ctx_size(void)
+{
+    return sizeof(struct poly1305_context);
+}
+
+/* pick 32-bit unsigned integer in little endian order */
+static unsigned int U8TOU32(const unsigned char *p)
+{
+    return (((unsigned int)(p[0] & 0xff)) |
+            ((unsigned int)(p[1] & 0xff) << 8) |
+            ((unsigned int)(p[2] & 0xff) << 16) |
+            ((unsigned int)(p[3] & 0xff) << 24));
+}
+
+/*
+ * Implementations can be classified by amount of significant bits in
+ * words making up the multi-precision value, or in other words radix
+ * or base of numerical representation, e.g. base 2^64, base 2^32,
+ * base 2^26. Complementary characteristic is how wide is the result of
+ * multiplication of pair of digits, e.g. it would take 128 bits to
+ * accommodate multiplication result in base 2^64 case. These are used
+ * interchangeably. To describe implementation that is. But interface
+ * is designed to isolate this so that low-level primitives implemented
+ * in assembly can be self-contained/self-coherent.
+ */
+#ifndef POLY1305_ASM
+/*
+ * Even though there is __int128 reference implementation targeting
+ * 64-bit platforms provided below, it's not obvious that it's optimal
+ * choice for every one of them. Depending on instruction set overall
+ * amount of instructions can be comparable to one in __int64
+ * implementation. Amount of multiplication instructions would be lower,
+ * but not necessarily overall. And in out-of-order execution context,
+ * it is the latter that can be crucial...
+ *
+ * On related note. Poly1305 author, D. J. Bernstein, discusses and
+ * provides floating-point implementations of the algorithm in question.
+ * It made a lot of sense by the time of introduction, because most
+ * then-modern processors didn't have pipelined integer multiplier.
+ * [Not to mention that some had non-constant timing for integer
+ * multiplications.] Floating-point instructions on the other hand could
+ * be issued every cycle, which allowed to achieve better performance.
+ * Nowadays, with SIMD and/or out-or-order execution, shared or
+ * even emulated FPU, it's more complicated, and floating-point
+ * implementation is not necessarily optimal choice in every situation,
+ * rather contrary...
+ *
+ *                                              <appro@openssl.org>
+ */
+
+typedef unsigned int u32;
+
+/*
+ * poly1305_blocks processes a multiple of POLY1305_BLOCK_SIZE blocks
+ * of |inp| no longer than |len|. Behaviour for |len| not divisible by
+ * block size is unspecified in general case, even though in reference
+ * implementation the trailing chunk is simply ignored. Per algorithm
+ * specification, every input block, complete or last partial, is to be
+ * padded with a bit past most significant byte. The latter kind is then
+ * padded with zeros till block size. This last partial block padding
+ * is caller(*)'s responsibility, and because of this the last partial
+ * block is always processed with separate call with |len| set to
+ * POLY1305_BLOCK_SIZE and |padbit| to 0. In all other cases |padbit|
+ * should be set to 1 to perform implicit padding with 128th bit.
+ * poly1305_blocks does not actually check for this constraint though,
+ * it's caller(*)'s responsibility to comply.
+ *
+ * (*)  In the context "caller" is not application code, but higher
+ *      level Poly1305_* from this very module, so that quirks are
+ *      handled locally.
+ */
+static void
+poly1305_blocks(void *ctx, const unsigned char *inp, size_t len, u32 padbit);
+
+/*
+ * Type-agnostic "rip-off" from constant_time.h
+ */
+# define CONSTANT_TIME_CARRY(a,b) ( \
+         (a ^ ((a ^ b) | ((a - b) ^ b))) >> (sizeof(a) * 8 - 1) \
+         )
+
+# if (defined(__SIZEOF_INT128__) && __SIZEOF_INT128__==16) && \
+     (defined(__SIZEOF_LONG__) && __SIZEOF_LONG__==8)
+
+typedef unsigned long u64;
+typedef __uint128_t u128;
+
+typedef struct {
+    u64 h[3];
+    u64 r[2];
+} poly1305_internal;
+
+/* pick 32-bit unsigned integer in little endian order */
+static u64 U8TOU64(const unsigned char *p)
+{
+    return (((u64)(p[0] & 0xff)) |
+            ((u64)(p[1] & 0xff) << 8) |
+            ((u64)(p[2] & 0xff) << 16) |
+            ((u64)(p[3] & 0xff) << 24) |
+            ((u64)(p[4] & 0xff) << 32) |
+            ((u64)(p[5] & 0xff) << 40) |
+            ((u64)(p[6] & 0xff) << 48) |
+            ((u64)(p[7] & 0xff) << 56));
+}
+
+/* store a 32-bit unsigned integer in little endian */
+static void U64TO8(unsigned char *p, u64 v)
+{
+    p[0] = (unsigned char)((v) & 0xff);
+    p[1] = (unsigned char)((v >> 8) & 0xff);
+    p[2] = (unsigned char)((v >> 16) & 0xff);
+    p[3] = (unsigned char)((v >> 24) & 0xff);
+    p[4] = (unsigned char)((v >> 32) & 0xff);
+    p[5] = (unsigned char)((v >> 40) & 0xff);
+    p[6] = (unsigned char)((v >> 48) & 0xff);
+    p[7] = (unsigned char)((v >> 56) & 0xff);
+}
+
+static void poly1305_init(void *ctx, const unsigned char key[16])
+{
+    poly1305_internal *st = (poly1305_internal *) ctx;
+
+    /* h = 0 */
+    st->h[0] = 0;
+    st->h[1] = 0;
+    st->h[2] = 0;
+
+    /* r &= 0xffffffc0ffffffc0ffffffc0fffffff */
+    st->r[0] = U8TOU64(&key[0]) & 0x0ffffffc0fffffff;
+    st->r[1] = U8TOU64(&key[8]) & 0x0ffffffc0ffffffc;
+}
+
+static void
+poly1305_blocks(void *ctx, const unsigned char *inp, size_t len, u32 padbit)
+{
+    poly1305_internal *st = (poly1305_internal *)ctx;
+    u64 r0, r1;
+    u64 s1;
+    u64 h0, h1, h2, c;
+    u128 d0, d1;
+
+    r0 = st->r[0];
+    r1 = st->r[1];
+
+    s1 = r1 + (r1 >> 2);
+
+    h0 = st->h[0];
+    h1 = st->h[1];
+    h2 = st->h[2];
+
+    while (len >= POLY1305_BLOCK_SIZE) {
+        /* h += m[i] */
+        h0 = (u64)(d0 = (u128)h0 + U8TOU64(inp + 0));
+        h1 = (u64)(d1 = (u128)h1 + (d0 >> 64) + U8TOU64(inp + 8));
+        /*
+         * padbit can be zero only when original len was
+         * POLY1306_BLOCK_SIZE, but we don't check
+         */
+        h2 += (u64)(d1 >> 64) + padbit;
+
+        /* h *= r "%" p, where "%" stands for "partial remainder" */
+        d0 = ((u128)h0 * r0) +
+             ((u128)h1 * s1);
+        d1 = ((u128)h0 * r1) +
+             ((u128)h1 * r0) +
+             (h2 * s1);
+        h2 = (h2 * r0);
+
+        /* last reduction step: */
+        /* a) h2:h0 = h2<<128 + d1<<64 + d0 */
+        h0 = (u64)d0;
+        h1 = (u64)(d1 += d0 >> 64);
+        h2 += (u64)(d1 >> 64);
+        /* b) (h2:h0 += (h2:h0>>130) * 5) %= 2^130 */
+        c = (h2 >> 2) + (h2 & ~3UL);
+        h2 &= 3;
+        h0 += c;
+        h1 += (c = CONSTANT_TIME_CARRY(h0,c));
+        h2 += CONSTANT_TIME_CARRY(h1,c);
+        /*
+         * Occasional overflows to 3rd bit of h2 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;
+    }
+
+    st->h[0] = h0;
+    st->h[1] = h1;
+    st->h[2] = h2;
+}
+
+static void poly1305_emit(void *ctx, unsigned char mac[16],
+                          const u32 nonce[4])
+{
+    poly1305_internal *st = (poly1305_internal *) ctx;
+    u64 h0, h1, h2;
+    u64 g0, g1, g2;
+    u128 t;
+    u64 mask;
+
+    h0 = st->h[0];
+    h1 = st->h[1];
+    h2 = st->h[2];
+
+    /* compare to modulus by computing h + -p */
+    g0 = (u64)(t = (u128)h0 + 5);
+    g1 = (u64)(t = (u128)h1 + (t >> 64));
+    g2 = h2 + (u64)(t >> 64);
+
+    /* if there was carry into 131st bit, h1:h0 = g1:g0 */
+    mask = 0 - (g2 >> 2);
+    g0 &= mask;
+    g1 &= mask;
+    mask = ~mask;
+    h0 = (h0 & mask) | g0;
+    h1 = (h1 & mask) | g1;
+
+    /* mac = (h + nonce) % (2^128) */
+    h0 = (u64)(t = (u128)h0 + nonce[0] + ((u64)nonce[1]<<32));
+    h1 = (u64)(t = (u128)h1 + nonce[2] + ((u64)nonce[3]<<32) + (t >> 64));
+
+    U64TO8(mac + 0, h0);
+    U64TO8(mac + 8, h1);
+}
+
+# else
+
+#  if defined(_WIN32) && !defined(__MINGW32__)
+typedef unsigned __int64 u64;
+#  elif defined(__arch64__)
+typedef unsigned long u64;
+#  else
+typedef unsigned long long u64;
+#  endif
+
+typedef struct {
+    u32 h[5];
+    u32 r[4];
+} poly1305_internal;
+
+/* store a 32-bit unsigned integer in little endian */
+static void U32TO8(unsigned char *p, unsigned int v)
+{
+    p[0] = (unsigned char)((v) & 0xff);
+    p[1] = (unsigned char)((v >> 8) & 0xff);
+    p[2] = (unsigned char)((v >> 16) & 0xff);
+    p[3] = (unsigned char)((v >> 24) & 0xff);
+}
+
+static void poly1305_init(void *ctx, const unsigned char key[16])
+{
+    poly1305_internal *st = (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 */
+    st->r[0] = U8TOU32(&key[0]) & 0x0fffffff;
+    st->r[1] = U8TOU32(&key[4]) & 0x0ffffffc;
+    st->r[2] = U8TOU32(&key[8]) & 0x0ffffffc;
+    st->r[3] = U8TOU32(&key[12]) & 0x0ffffffc;
+}
+
+static void
+poly1305_blocks(void *ctx, const unsigned char *inp, size_t len, u32 padbit)
+{
+    poly1305_internal *st = (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 = (u32)(d0 = (u64)h0 + U8TOU32(inp + 0));
+        h1 = (u32)(d1 = (u64)h1 + (d0 >> 32) + U8TOU32(inp + 4));
+        h2 = (u32)(d2 = (u64)h2 + (d1 >> 32) + U8TOU32(inp + 8));
+        h3 = (u32)(d3 = (u64)h3 + (d2 >> 32) + U8TOU32(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;
+    }
+
+    st->h[0] = h0;
+    st->h[1] = h1;
+    st->h[2] = h2;
+    st->h[3] = h3;
+    st->h[4] = h4;
+}
+
+static void poly1305_emit(void *ctx, unsigned char mac[16],
+                          const u32 nonce[4])
+{
+    poly1305_internal *st = (poly1305_internal *) ctx;
+    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]);
+
+    U32TO8(mac + 0, h0);
+    U32TO8(mac + 4, h1);
+    U32TO8(mac + 8, h2);
+    U32TO8(mac + 12, h3);
+}
+# endif
+#else
+int poly1305_init(void *ctx, const unsigned char key[16], void *func);
+void poly1305_blocks(void *ctx, const unsigned char *inp, size_t len,
+                     unsigned int padbit);
+void poly1305_emit(void *ctx, unsigned char mac[16],
+                   const unsigned int nonce[4]);
+#endif
+
+void Poly1305_Init(POLY1305 *ctx, const unsigned char key[32])
+{
+    ctx->nonce[0] = U8TOU32(&key[16]);
+    ctx->nonce[1] = U8TOU32(&key[20]);
+    ctx->nonce[2] = U8TOU32(&key[24]);
+    ctx->nonce[3] = U8TOU32(&key[28]);
+
+#ifndef POLY1305_ASM
+    poly1305_init(ctx->opaque, key);
+#else
+    /*
+     * Unlike reference poly1305_init assembly counterpart is expected
+     * to return a value: non-zero if it initializes ctx->func, and zero
+     * otherwise. Latter is to simplify assembly in cases when there no
+     * multiple code paths to switch between.
+     */
+    if (!poly1305_init(ctx->opaque, key, &ctx->func)) {
+        ctx->func.blocks = poly1305_blocks;
+        ctx->func.emit = poly1305_emit;
+    }
+#endif
+
+    ctx->num = 0;
+
+}
+
+#ifdef POLY1305_ASM
+/*
+ * This "eclipses" poly1305_blocks and poly1305_emit, but it's
+ * conscious choice imposed by -Wshadow compiler warnings.
+ */
+# define poly1305_blocks (*poly1305_blocks_p)
+# define poly1305_emit   (*poly1305_emit_p)
+#endif
+
+void Poly1305_Update(POLY1305 *ctx, const unsigned char *inp, size_t len)
+{
+#ifdef POLY1305_ASM
+    /*
+     * As documented, poly1305_blocks is never called with input
+     * longer than single block and padbit argument set to 0. This
+     * property is fluently used in assembly modules to optimize
+     * padbit handling on loop boundary.
+     */
+    poly1305_blocks_f poly1305_blocks_p = ctx->func.blocks;
+#endif
+    size_t rem, num;
+
+    if ((num = ctx->num)) {
+        rem = POLY1305_BLOCK_SIZE - num;
+        if (len >= rem) {
+            memcpy(ctx->data + num, inp, rem);
+            poly1305_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;
+        }
+    }
+
+    rem = len % POLY1305_BLOCK_SIZE;
+    len -= rem;
+
+    if (len >= POLY1305_BLOCK_SIZE) {
+        poly1305_blocks(ctx->opaque, inp, len, 1);
+        inp += len;
+    }
+
+    if (rem)
+        memcpy(ctx->data, inp, rem);
+
+    ctx->num = rem;
+}
+
+void Poly1305_Final(POLY1305 *ctx, unsigned char mac[16])
+{
+#ifdef POLY1305_ASM
+    poly1305_blocks_f poly1305_blocks_p = ctx->func.blocks;
+    poly1305_emit_f poly1305_emit_p = ctx->func.emit;
+#endif
+    size_t num;
+
+    if ((num = ctx->num)) {
+        ctx->data[num++] = 1;   /* pad bit */
+        while (num < POLY1305_BLOCK_SIZE)
+            ctx->data[num++] = 0;
+        poly1305_blocks(ctx->opaque, ctx->data, POLY1305_BLOCK_SIZE, 0);
+    }
+
+    poly1305_emit(ctx->opaque, mac, ctx->nonce);
+
+    /* zero out the state */
+    OPENSSL_cleanse(ctx, sizeof(*ctx));
+}
diff --git a/contrib/libs/openssl/crypto/poly1305/poly1305_ameth.c b/contrib/libs/openssl/crypto/poly1305/poly1305_ameth.c
new file mode 100644
index 0000000000..0dddf79626
--- /dev/null
+++ b/contrib/libs/openssl/crypto/poly1305/poly1305_ameth.c
@@ -0,0 +1,122 @@
+/*
+ * Copyright 2007-2021 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License").  You may not use
+ * this file except in compliance with the License.  You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include "internal/cryptlib.h"
+#include <openssl/evp.h>
+#include "crypto/asn1.h"
+#include "crypto/poly1305.h"
+#include "poly1305_local.h"
+#include "crypto/evp.h"
+
+/*
+ * POLY1305 "ASN1" method. This is just here to indicate the maximum
+ * POLY1305 output length and to free up a POLY1305 key.
+ */
+
+static int poly1305_size(const EVP_PKEY *pkey)
+{
+    return POLY1305_DIGEST_SIZE;
+}
+
+static void poly1305_key_free(EVP_PKEY *pkey)
+{
+    ASN1_OCTET_STRING *os = EVP_PKEY_get0(pkey);
+    if (os != NULL) {
+        if (os->data != NULL)
+            OPENSSL_cleanse(os->data, os->length);
+        ASN1_OCTET_STRING_free(os);
+    }
+}
+
+static int poly1305_pkey_ctrl(EVP_PKEY *pkey, int op, long arg1, void *arg2)
+{
+    /* nothing, (including ASN1_PKEY_CTRL_DEFAULT_MD_NID), is supported */
+    return -2;
+}
+
+static int poly1305_pkey_public_cmp(const EVP_PKEY *a, const EVP_PKEY *b)
+{
+    return ASN1_OCTET_STRING_cmp(EVP_PKEY_get0(a), EVP_PKEY_get0(b)) == 0;
+}
+
+static int poly1305_set_priv_key(EVP_PKEY *pkey, const unsigned char *priv,
+                                 size_t len)
+{
+    ASN1_OCTET_STRING *os;
+
+    if (pkey->pkey.ptr != NULL || len != POLY1305_KEY_SIZE)
+        return 0;
+
+    os = ASN1_OCTET_STRING_new();
+    if (os == NULL)
+        return 0;
+
+    if (!ASN1_OCTET_STRING_set(os, priv, len)) {
+        ASN1_OCTET_STRING_free(os);
+        return 0;
+    }
+
+    pkey->pkey.ptr = os;
+    return 1;
+}
+
+static int poly1305_get_priv_key(const EVP_PKEY *pkey, unsigned char *priv,
+                                 size_t *len)
+{
+    ASN1_OCTET_STRING *os = (ASN1_OCTET_STRING *)pkey->pkey.ptr;
+
+    if (priv == NULL) {
+        *len = POLY1305_KEY_SIZE;
+        return 1;
+    }
+
+    if (os == NULL || *len < POLY1305_KEY_SIZE)
+        return 0;
+
+    memcpy(priv, ASN1_STRING_get0_data(os), ASN1_STRING_length(os));
+    *len = POLY1305_KEY_SIZE;
+
+    return 1;
+}
+
+const EVP_PKEY_ASN1_METHOD poly1305_asn1_meth = {
+    EVP_PKEY_POLY1305,
+    EVP_PKEY_POLY1305,
+    0,
+
+    "POLY1305",
+    "OpenSSL POLY1305 method",
+
+    0, 0, poly1305_pkey_public_cmp, 0,
+
+    0, 0, 0,
+
+    poly1305_size,
+    0, 0,
+    0, 0, 0, 0, 0, 0, 0,
+
+    poly1305_key_free,
+    poly1305_pkey_ctrl,
+    NULL,
+    NULL,
+
+    NULL,
+    NULL,
+    NULL,
+
+    NULL,
+    NULL,
+    NULL,
+
+    poly1305_set_priv_key,
+    NULL,
+    poly1305_get_priv_key,
+    NULL,
+};
diff --git a/contrib/libs/openssl/crypto/poly1305/poly1305_base2_44.c b/contrib/libs/openssl/crypto/poly1305/poly1305_base2_44.c
new file mode 100644
index 0000000000..b6313d01ba
--- /dev/null
+++ b/contrib/libs/openssl/crypto/poly1305/poly1305_base2_44.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright 2016 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License").  You may not use
+ * this file except in compliance with the License.  You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+/*
+ * This module is meant to be used as template for base 2^44 assembly
+ * implementation[s]. On side note compiler-generated code is not
+ * slower than compiler-generated base 2^64 code on [high-end] x86_64,
+ * even though amount of multiplications is 50% higher. Go figure...
+ */
+#include <stdlib.h>
+
+typedef unsigned char u8;
+typedef unsigned int u32;
+typedef unsigned long u64;
+typedef unsigned __int128 u128;
+
+typedef struct {
+    u64 h[3];
+    u64 s[2];
+    u64 r[3];
+} poly1305_internal;
+
+#define POLY1305_BLOCK_SIZE 16
+
+/* pick 64-bit unsigned integer in little endian order */
+static u64 U8TOU64(const unsigned char *p)
+{
+    return (((u64)(p[0] & 0xff)) |
+            ((u64)(p[1] & 0xff) << 8) |
+            ((u64)(p[2] & 0xff) << 16) |
+            ((u64)(p[3] & 0xff) << 24) |
+            ((u64)(p[4] & 0xff) << 32) |
+            ((u64)(p[5] & 0xff) << 40) |
+            ((u64)(p[6] & 0xff) << 48) |
+            ((u64)(p[7] & 0xff) << 56));
+}
+
+/* store a 64-bit unsigned integer in little endian */
+static void U64TO8(unsigned char *p, u64 v)
+{
+    p[0] = (unsigned char)((v) & 0xff);
+    p[1] = (unsigned char)((v >> 8) & 0xff);
+    p[2] = (unsigned char)((v >> 16) & 0xff);
+    p[3] = (unsigned char)((v >> 24) & 0xff);
+    p[4] = (unsigned char)((v >> 32) & 0xff);
+    p[5] = (unsigned char)((v >> 40) & 0xff);
+    p[6] = (unsigned char)((v >> 48) & 0xff);
+    p[7] = (unsigned char)((v >> 56) & 0xff);
+}
+
+int poly1305_init(void *ctx, const unsigned char key[16])
+{
+    poly1305_internal *st = (poly1305_internal *)ctx;
+    u64 r0, r1;
+
+    /* h = 0 */
+    st->h[0] = 0;
+    st->h[1] = 0;
+    st->h[2] = 0;
+
+    r0 = U8TOU64(&key[0]) & 0x0ffffffc0fffffff;
+    r1 = U8TOU64(&key[8]) & 0x0ffffffc0ffffffc;
+
+    /* break r1:r0 to three 44-bit digits, masks are 1<<44-1 */
+    st->r[0] = r0 & 0x0fffffffffff;
+    st->r[1] = ((r0 >> 44) | (r1 << 20))  & 0x0fffffffffff;
+    st->r[2] = (r1 >> 24);
+
+    st->s[0] = (st->r[1] + (st->r[1] << 2)) << 2;
+    st->s[1] = (st->r[2] + (st->r[2] << 2)) << 2;
+
+    return 0;
+}
+
+void poly1305_blocks(void *ctx, const unsigned char *inp, size_t len,
+                     u32 padbit)
+{
+    poly1305_internal *st = (poly1305_internal *)ctx;
+    u64 r0, r1, r2;
+    u64 s1, s2;
+    u64 h0, h1, h2, c;
+    u128 d0, d1, d2;
+    u64 pad = (u64)padbit << 40;
+
+    r0 = st->r[0];
+    r1 = st->r[1];
+    r2 = st->r[2];
+
+    s1 = st->s[0];
+    s2 = st->s[1];
+
+    h0 = st->h[0];
+    h1 = st->h[1];
+    h2 = st->h[2];
+
+    while (len >= POLY1305_BLOCK_SIZE) {
+        u64 m0, m1;
+
+        m0 = U8TOU64(inp + 0);
+        m1 = U8TOU64(inp + 8);
+
+        /* h += m[i], m[i] is broken to 44-bit digits */
+        h0 += m0 & 0x0fffffffffff;
+        h1 += ((m0 >> 44) | (m1 << 20))  & 0x0fffffffffff;
+        h2 +=  (m1 >> 24) + pad;
+
+        /* h *= r "%" p, where "%" stands for "partial remainder" */
+        d0 = ((u128)h0 * r0) + ((u128)h1 * s2) + ((u128)h2 * s1);
+        d1 = ((u128)h0 * r1) + ((u128)h1 * r0) + ((u128)h2 * s2);
+        d2 = ((u128)h0 * r2) + ((u128)h1 * r1) + ((u128)h2 * r0);
+
+        /* "lazy" reduction step */
+        h0 = (u64)d0 & 0x0fffffffffff;
+        h1 = (u64)(d1 += (u64)(d0 >> 44)) & 0x0fffffffffff;
+        h2 = (u64)(d2 += (u64)(d1 >> 44)) & 0x03ffffffffff; /* last 42 bits */
+
+        c = (d2 >> 42);
+        h0 += c + (c << 2);
+
+        inp += POLY1305_BLOCK_SIZE;
+        len -= POLY1305_BLOCK_SIZE;
+    }
+
+    st->h[0] = h0;
+    st->h[1] = h1;
+    st->h[2] = h2;
+}
+
+void poly1305_emit(void *ctx, unsigned char mac[16], const u32 nonce[4])
+{
+    poly1305_internal *st = (poly1305_internal *) ctx;
+    u64 h0, h1, h2;
+    u64 g0, g1, g2;
+    u128 t;
+    u64 mask;
+
+    h0 = st->h[0];
+    h1 = st->h[1];
+    h2 = st->h[2];
+
+    /* after "lazy" reduction, convert 44+bit digits to 64-bit ones */
+    h0 = (u64)(t = (u128)h0 + (h1 << 44));              h1 >>= 20;
+    h1 = (u64)(t = (u128)h1 + (h2 << 24) + (t >> 64));  h2 >>= 40;
+    h2 += (u64)(t >> 64);
+
+    /* compare to modulus by computing h + -p */
+    g0 = (u64)(t = (u128)h0 + 5);
+    g1 = (u64)(t = (u128)h1 + (t >> 64));
+    g2 = h2 + (u64)(t >> 64);
+
+    /* if there was carry into 131st bit, h1:h0 = g1:g0 */
+    mask = 0 - (g2 >> 2);
+    g0 &= mask;
+    g1 &= mask;
+    mask = ~mask;
+    h0 = (h0 & mask) | g0;
+    h1 = (h1 & mask) | g1;
+
+    /* mac = (h + nonce) % (2^128) */
+    h0 = (u64)(t = (u128)h0 + nonce[0] + ((u64)nonce[1]<<32));
+    h1 = (u64)(t = (u128)h1 + nonce[2] + ((u64)nonce[3]<<32) + (t >> 64));
+
+    U64TO8(mac + 0, h0);
+    U64TO8(mac + 8, h1);
+}
diff --git a/contrib/libs/openssl/crypto/poly1305/poly1305_ieee754.c b/contrib/libs/openssl/crypto/poly1305/poly1305_ieee754.c
new file mode 100644
index 0000000000..7cfd968645
--- /dev/null
+++ b/contrib/libs/openssl/crypto/poly1305/poly1305_ieee754.c
@@ -0,0 +1,488 @@
+/*
+ * Copyright 2016-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License").  You may not use
+ * this file except in compliance with the License.  You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+/*
+ * This module is meant to be used as template for non-x87 floating-
+ * point assembly modules. The template itself is x86_64-specific
+ * though, as it was debugged on x86_64. So that implementor would
+ * have to recognize platform-specific parts, UxTOy and inline asm,
+ * and act accordingly.
+ *
+ * Huh? x86_64-specific code as template for non-x87? Note seven, which
+ * is not a typo, but reference to 80-bit precision. This module on the
+ * other hand relies on 64-bit precision operations, which are default
+ * for x86_64 code. And since we are at it, just for sense of it,
+ * large-block performance in cycles per processed byte for *this* code
+ * is:
+ *                      gcc-4.8         icc-15.0        clang-3.4(*)
+ *
+ * Westmere             4.96            5.09            4.37
+ * Sandy Bridge         4.95            4.90            4.17
+ * Haswell              4.92            4.87            3.78
+ * Bulldozer            4.67            4.49            4.68
+ * VIA Nano             7.07            7.05            5.98
+ * Silvermont           10.6            9.61            12.6
+ *
+ * (*)  clang managed to discover parallelism and deployed SIMD;
+ *
+ * And for range of other platforms with unspecified gcc versions:
+ *
+ * Freescale e300       12.5
+ * PPC74x0              10.8
+ * POWER6               4.92
+ * POWER7               4.50
+ * POWER8               4.10
+ *
+ * z10                  11.2
+ * z196+                7.30
+ *
+ * UltraSPARC III       16.0
+ * SPARC T4             16.1
+ */
+
+#if !(defined(__GNUC__) && __GNUC__>=2)
+# error "this is gcc-specific template"
+#endif
+
+#include <stdlib.h>
+
+typedef unsigned char u8;
+typedef unsigned int u32;
+typedef unsigned long long u64;
+typedef union { double d; u64 u; } elem64;
+
+#define TWO(p)          ((double)(1ULL<<(p)))
+#define TWO0            TWO(0)
+#define TWO32           TWO(32)
+#define TWO64           (TWO32*TWO(32))
+#define TWO96           (TWO64*TWO(32))
+#define TWO130          (TWO96*TWO(34))
+
+#define EXP(p)          ((1023ULL+(p))<<52)
+
+#if defined(__x86_64__) || (defined(__PPC__) && defined(__LITTLE_ENDIAN__))
+# define U8TOU32(p)     (*(const u32 *)(p))
+# define U32TO8(p,v)    (*(u32 *)(p) = (v))
+#elif defined(__PPC__)
+# define U8TOU32(p)     ({u32 ret; asm ("lwbrx	%0,0,%1":"=r"(ret):"b"(p)); ret; })
+# define U32TO8(p,v)    asm ("stwbrx %0,0,%1"::"r"(v),"b"(p):"memory")
+#elif defined(__s390x__)
+# define U8TOU32(p)     ({u32 ret; asm ("lrv	%0,%1":"=d"(ret):"m"(*(u32 *)(p))); ret; })
+# define U32TO8(p,v)    asm ("strv	%1,%0":"=m"(*(u32 *)(p)):"d"(v))
+#endif
+
+#ifndef U8TOU32
+# define U8TOU32(p)     ((u32)(p)[0]     | (u32)(p)[1]<<8 |     \
+                         (u32)(p)[2]<<16 | (u32)(p)[3]<<24  )
+#endif
+#ifndef U32TO8
+# define U32TO8(p,v)    ((p)[0] = (u8)(v),       (p)[1] = (u8)((v)>>8), \
+                         (p)[2] = (u8)((v)>>16), (p)[3] = (u8)((v)>>24) )
+#endif
+
+typedef struct {
+    elem64 h[4];
+    double r[8];
+    double s[6];
+} poly1305_internal;
+
+/* "round toward zero (truncate), mask all exceptions" */
+#if defined(__x86_64__)
+static const u32 mxcsr = 0x7f80;
+#elif defined(__PPC__)
+static const u64 one = 1;
+#elif defined(__s390x__)
+static const u32 fpc = 1;
+#elif defined(__sparc__)
+static const u64 fsr = 1ULL<<30;
+#elif defined(__mips__)
+static const u32 fcsr = 1;
+#else
+#error "unrecognized platform"
+#endif
+
+int poly1305_init(void *ctx, const unsigned char key[16])
+{
+    poly1305_internal *st = (poly1305_internal *) ctx;
+    elem64 r0, r1, r2, r3;
+
+    /* h = 0, biased */
+#if 0
+    st->h[0].d = TWO(52)*TWO0;
+    st->h[1].d = TWO(52)*TWO32;
+    st->h[2].d = TWO(52)*TWO64;
+    st->h[3].d = TWO(52)*TWO96;
+#else
+    st->h[0].u = EXP(52+0);
+    st->h[1].u = EXP(52+32);
+    st->h[2].u = EXP(52+64);
+    st->h[3].u = EXP(52+96);
+#endif
+
+    if (key) {
+        /*
+         * set "truncate" rounding mode
+         */
+#if defined(__x86_64__)
+        u32 mxcsr_orig;
+
+        asm volatile ("stmxcsr	%0":"=m"(mxcsr_orig));
+        asm volatile ("ldmxcsr	%0"::"m"(mxcsr));
+#elif defined(__PPC__)
+        double fpscr_orig, fpscr = *(double *)&one;
+
+        asm volatile ("mffs	%0":"=f"(fpscr_orig));
+        asm volatile ("mtfsf	255,%0"::"f"(fpscr));
+#elif defined(__s390x__)
+        u32 fpc_orig;
+
+        asm volatile ("stfpc	%0":"=m"(fpc_orig));
+        asm volatile ("lfpc	%0"::"m"(fpc));
+#elif defined(__sparc__)
+        u64 fsr_orig;
+
+        asm volatile ("stx	%%fsr,%0":"=m"(fsr_orig));
+        asm volatile ("ldx	%0,%%fsr"::"m"(fsr));
+#elif defined(__mips__)
+        u32 fcsr_orig;
+
+        asm volatile ("cfc1	%0,$31":"=r"(fcsr_orig));
+        asm volatile ("ctc1	%0,$31"::"r"(fcsr));
+#endif
+
+        /* r &= 0xffffffc0ffffffc0ffffffc0fffffff */
+        r0.u = EXP(52+0)  | (U8TOU32(&key[0])  & 0x0fffffff);
+        r1.u = EXP(52+32) | (U8TOU32(&key[4])  & 0x0ffffffc);
+        r2.u = EXP(52+64) | (U8TOU32(&key[8])  & 0x0ffffffc);
+        r3.u = EXP(52+96) | (U8TOU32(&key[12]) & 0x0ffffffc);
+
+        st->r[0] = r0.d - TWO(52)*TWO0;
+        st->r[2] = r1.d - TWO(52)*TWO32;
+        st->r[4] = r2.d - TWO(52)*TWO64;
+        st->r[6] = r3.d - TWO(52)*TWO96;
+
+        st->s[0] = st->r[2] * (5.0/TWO130);
+        st->s[2] = st->r[4] * (5.0/TWO130);
+        st->s[4] = st->r[6] * (5.0/TWO130);
+
+        /*
+         * base 2^32 -> base 2^16
+         */
+        st->r[1] = (st->r[0] + TWO(52)*TWO(16)*TWO0) -
+                               TWO(52)*TWO(16)*TWO0;
+        st->r[0] -= st->r[1];
+
+        st->r[3] = (st->r[2] + TWO(52)*TWO(16)*TWO32) -
+                               TWO(52)*TWO(16)*TWO32;
+        st->r[2] -= st->r[3];
+
+        st->r[5] = (st->r[4] + TWO(52)*TWO(16)*TWO64) -
+                               TWO(52)*TWO(16)*TWO64;
+        st->r[4] -= st->r[5];
+
+        st->r[7] = (st->r[6] + TWO(52)*TWO(16)*TWO96) -
+                               TWO(52)*TWO(16)*TWO96;
+        st->r[6] -= st->r[7];
+
+        st->s[1] = (st->s[0] + TWO(52)*TWO(16)*TWO0/TWO96) -
+                               TWO(52)*TWO(16)*TWO0/TWO96;
+        st->s[0] -= st->s[1];
+
+        st->s[3] = (st->s[2] + TWO(52)*TWO(16)*TWO32/TWO96) -
+                               TWO(52)*TWO(16)*TWO32/TWO96;
+        st->s[2] -= st->s[3];
+
+        st->s[5] = (st->s[4] + TWO(52)*TWO(16)*TWO64/TWO96) -
+                               TWO(52)*TWO(16)*TWO64/TWO96;
+        st->s[4] -= st->s[5];
+
+        /*
+         * restore original FPU control register
+         */
+#if defined(__x86_64__)
+        asm volatile ("ldmxcsr	%0"::"m"(mxcsr_orig));
+#elif defined(__PPC__)
+        asm volatile ("mtfsf	255,%0"::"f"(fpscr_orig));
+#elif defined(__s390x__)
+        asm volatile ("lfpc	%0"::"m"(fpc_orig));
+#elif defined(__sparc__)
+        asm volatile ("ldx	%0,%%fsr"::"m"(fsr_orig));
+#elif defined(__mips__)
+        asm volatile ("ctc1	%0,$31"::"r"(fcsr_orig));
+#endif
+    }
+
+    return 0;
+}
+
+void poly1305_blocks(void *ctx, const unsigned char *inp, size_t len,
+                     int padbit)
+{
+    poly1305_internal *st = (poly1305_internal *)ctx;
+    elem64 in0, in1, in2, in3;
+    u64 pad = (u64)padbit<<32;
+
+    double x0, x1, x2, x3;
+    double h0lo, h0hi, h1lo, h1hi, h2lo, h2hi, h3lo, h3hi;
+    double c0lo, c0hi, c1lo, c1hi, c2lo, c2hi, c3lo, c3hi;
+
+    const double r0lo = st->r[0];
+    const double r0hi = st->r[1];
+    const double r1lo = st->r[2];
+    const double r1hi = st->r[3];
+    const double r2lo = st->r[4];
+    const double r2hi = st->r[5];
+    const double r3lo = st->r[6];
+    const double r3hi = st->r[7];
+
+    const double s1lo = st->s[0];
+    const double s1hi = st->s[1];
+    const double s2lo = st->s[2];
+    const double s2hi = st->s[3];
+    const double s3lo = st->s[4];
+    const double s3hi = st->s[5];
+
+    /*
+     * set "truncate" rounding mode
+     */
+#if defined(__x86_64__)
+    u32 mxcsr_orig;
+
+    asm volatile ("stmxcsr	%0":"=m"(mxcsr_orig));
+    asm volatile ("ldmxcsr	%0"::"m"(mxcsr));
+#elif defined(__PPC__)
+    double fpscr_orig, fpscr = *(double *)&one;
+
+    asm volatile ("mffs		%0":"=f"(fpscr_orig));
+    asm volatile ("mtfsf	255,%0"::"f"(fpscr));
+#elif defined(__s390x__)
+    u32 fpc_orig;
+
+    asm volatile ("stfpc	%0":"=m"(fpc_orig));
+    asm volatile ("lfpc		%0"::"m"(fpc));
+#elif defined(__sparc__)
+    u64 fsr_orig;
+
+    asm volatile ("stx		%%fsr,%0":"=m"(fsr_orig));
+    asm volatile ("ldx		%0,%%fsr"::"m"(fsr));
+#elif defined(__mips__)
+    u32 fcsr_orig;
+
+    asm volatile ("cfc1		%0,$31":"=r"(fcsr_orig));
+    asm volatile ("ctc1		%0,$31"::"r"(fcsr));
+#endif
+
+    /*
+     * load base 2^32 and de-bias
+     */
+    h0lo = st->h[0].d - TWO(52)*TWO0;
+    h1lo = st->h[1].d - TWO(52)*TWO32;
+    h2lo = st->h[2].d - TWO(52)*TWO64;
+    h3lo = st->h[3].d - TWO(52)*TWO96;
+
+#ifdef __clang__
+    h0hi = 0;
+    h1hi = 0;
+    h2hi = 0;
+    h3hi = 0;
+#else
+    in0.u = EXP(52+0)  | U8TOU32(&inp[0]);
+    in1.u = EXP(52+32) | U8TOU32(&inp[4]);
+    in2.u = EXP(52+64) | U8TOU32(&inp[8]);
+    in3.u = EXP(52+96) | U8TOU32(&inp[12]) | pad;
+
+    x0 = in0.d - TWO(52)*TWO0;
+    x1 = in1.d - TWO(52)*TWO32;
+    x2 = in2.d - TWO(52)*TWO64;
+    x3 = in3.d - TWO(52)*TWO96;
+
+    x0 += h0lo;
+    x1 += h1lo;
+    x2 += h2lo;
+    x3 += h3lo;
+
+    goto fast_entry;
+#endif
+
+    do {
+        in0.u = EXP(52+0)  | U8TOU32(&inp[0]);
+        in1.u = EXP(52+32) | U8TOU32(&inp[4]);
+        in2.u = EXP(52+64) | U8TOU32(&inp[8]);
+        in3.u = EXP(52+96) | U8TOU32(&inp[12]) | pad;
+
+        x0 = in0.d - TWO(52)*TWO0;
+        x1 = in1.d - TWO(52)*TWO32;
+        x2 = in2.d - TWO(52)*TWO64;
+        x3 = in3.d - TWO(52)*TWO96;
+
+        /*
+         * note that there are multiple ways to accumulate input, e.g.
+         * one can as well accumulate to h0lo-h1lo-h1hi-h2hi...
+         */
+        h0lo += x0;
+        h0hi += x1;
+        h2lo += x2;
+        h2hi += x3;
+
+        /*
+         * carries that cross 32n-bit (and 130-bit) boundaries
+         */
+        c0lo = (h0lo + TWO(52)*TWO32)  - TWO(52)*TWO32;
+        c1lo = (h1lo + TWO(52)*TWO64)  - TWO(52)*TWO64;
+        c2lo = (h2lo + TWO(52)*TWO96)  - TWO(52)*TWO96;
+        c3lo = (h3lo + TWO(52)*TWO130) - TWO(52)*TWO130;
+
+        c0hi = (h0hi + TWO(52)*TWO32)  - TWO(52)*TWO32;
+        c1hi = (h1hi + TWO(52)*TWO64)  - TWO(52)*TWO64;
+        c2hi = (h2hi + TWO(52)*TWO96)  - TWO(52)*TWO96;
+        c3hi = (h3hi + TWO(52)*TWO130) - TWO(52)*TWO130;
+
+        /*
+         * base 2^48 -> base 2^32 with last reduction step
+         */
+        x1 =  (h1lo - c1lo) + c0lo;
+        x2 =  (h2lo - c2lo) + c1lo;
+        x3 =  (h3lo - c3lo) + c2lo;
+        x0 =  (h0lo - c0lo) + c3lo * (5.0/TWO130);
+
+        x1 += (h1hi - c1hi) + c0hi;
+        x2 += (h2hi - c2hi) + c1hi;
+        x3 += (h3hi - c3hi) + c2hi;
+        x0 += (h0hi - c0hi) + c3hi * (5.0/TWO130);
+
+#ifndef __clang__
+    fast_entry:
+#endif
+        /*
+         * base 2^32 * base 2^16 = base 2^48
+         */
+        h0lo = s3lo * x1 + s2lo * x2 + s1lo * x3 + r0lo * x0;
+        h1lo = r0lo * x1 + s3lo * x2 + s2lo * x3 + r1lo * x0;
+        h2lo = r1lo * x1 + r0lo * x2 + s3lo * x3 + r2lo * x0;
+        h3lo = r2lo * x1 + r1lo * x2 + r0lo * x3 + r3lo * x0;
+
+        h0hi = s3hi * x1 + s2hi * x2 + s1hi * x3 + r0hi * x0;
+        h1hi = r0hi * x1 + s3hi * x2 + s2hi * x3 + r1hi * x0;
+        h2hi = r1hi * x1 + r0hi * x2 + s3hi * x3 + r2hi * x0;
+        h3hi = r2hi * x1 + r1hi * x2 + r0hi * x3 + r3hi * x0;
+
+        inp += 16;
+        len -= 16;
+
+    } while (len >= 16);
+
+    /*
+     * carries that cross 32n-bit (and 130-bit) boundaries
+     */
+    c0lo = (h0lo + TWO(52)*TWO32)  - TWO(52)*TWO32;
+    c1lo = (h1lo + TWO(52)*TWO64)  - TWO(52)*TWO64;
+    c2lo = (h2lo + TWO(52)*TWO96)  - TWO(52)*TWO96;
+    c3lo = (h3lo + TWO(52)*TWO130) - TWO(52)*TWO130;
+
+    c0hi = (h0hi + TWO(52)*TWO32)  - TWO(52)*TWO32;
+    c1hi = (h1hi + TWO(52)*TWO64)  - TWO(52)*TWO64;
+    c2hi = (h2hi + TWO(52)*TWO96)  - TWO(52)*TWO96;
+    c3hi = (h3hi + TWO(52)*TWO130) - TWO(52)*TWO130;
+
+    /*
+     * base 2^48 -> base 2^32 with last reduction step
+     */
+    x1 =  (h1lo - c1lo) + c0lo;
+    x2 =  (h2lo - c2lo) + c1lo;
+    x3 =  (h3lo - c3lo) + c2lo;
+    x0 =  (h0lo - c0lo) + c3lo * (5.0/TWO130);
+
+    x1 += (h1hi - c1hi) + c0hi;
+    x2 += (h2hi - c2hi) + c1hi;
+    x3 += (h3hi - c3hi) + c2hi;
+    x0 += (h0hi - c0hi) + c3hi * (5.0/TWO130);
+
+    /*
+     * store base 2^32, with bias
+     */
+    st->h[1].d = x1 + TWO(52)*TWO32;
+    st->h[2].d = x2 + TWO(52)*TWO64;
+    st->h[3].d = x3 + TWO(52)*TWO96;
+    st->h[0].d = x0 + TWO(52)*TWO0;
+
+    /*
+     * restore original FPU control register
+     */
+#if defined(__x86_64__)
+    asm volatile ("ldmxcsr	%0"::"m"(mxcsr_orig));
+#elif defined(__PPC__)
+    asm volatile ("mtfsf	255,%0"::"f"(fpscr_orig));
+#elif defined(__s390x__)
+    asm volatile ("lfpc		%0"::"m"(fpc_orig));
+#elif defined(__sparc__)
+    asm volatile ("ldx		%0,%%fsr"::"m"(fsr_orig));
+#elif defined(__mips__)
+    asm volatile ("ctc1		%0,$31"::"r"(fcsr_orig));
+#endif
+}
+
+void poly1305_emit(void *ctx, unsigned char mac[16], const u32 nonce[4])
+{
+    poly1305_internal *st = (poly1305_internal *) ctx;
+    u64 h0, h1, h2, h3, h4;
+    u32 g0, g1, g2, g3, g4;
+    u64 t;
+    u32 mask;
+
+    /*
+     * thanks to bias masking exponent gives integer result
+     */
+    h0 = st->h[0].u & 0x000fffffffffffffULL;
+    h1 = st->h[1].u & 0x000fffffffffffffULL;
+    h2 = st->h[2].u & 0x000fffffffffffffULL;
+    h3 = st->h[3].u & 0x000fffffffffffffULL;
+
+    /*
+     * can be partially reduced, so reduce...
+     */
+    h4 = h3>>32; h3 &= 0xffffffffU;
+    g4 = h4&-4;
+    h4 &= 3;
+    g4 += g4>>2;
+
+    h0 += g4;
+    h1 += h0>>32; h0 &= 0xffffffffU;
+    h2 += h1>>32; h1 &= 0xffffffffU;
+    h3 += h2>>32; h2 &= 0xffffffffU;
+
+    /* compute h + -p */
+    g0 = (u32)(t = h0 + 5);
+    g1 = (u32)(t = h1 + (t >> 32));
+    g2 = (u32)(t = h2 + (t >> 32));
+    g3 = (u32)(t = h3 + (t >> 32));
+    g4 = h4 + (u32)(t >> 32);
+
+    /* if there was carry, select g0-g3 */
+    mask = 0 - (g4 >> 2);
+    g0 &= mask;
+    g1 &= mask;
+    g2 &= mask;
+    g3 &= mask;
+    mask = ~mask;
+    g0 |= (h0 & mask);
+    g1 |= (h1 & mask);
+    g2 |= (h2 & mask);
+    g3 |= (h3 & mask);
+
+    /* mac = (h + nonce) % (2^128) */
+    g0 = (u32)(t = (u64)g0 + nonce[0]);
+    g1 = (u32)(t = (u64)g1 + (t >> 32) + nonce[1]);
+    g2 = (u32)(t = (u64)g2 + (t >> 32) + nonce[2]);
+    g3 = (u32)(t = (u64)g3 + (t >> 32) + nonce[3]);
+
+    U32TO8(mac + 0, g0);
+    U32TO8(mac + 4, g1);
+    U32TO8(mac + 8, g2);
+    U32TO8(mac + 12, g3);
+}
diff --git a/contrib/libs/openssl/crypto/poly1305/poly1305_local.h b/contrib/libs/openssl/crypto/poly1305/poly1305_local.h
new file mode 100644
index 0000000000..6d4d9dc5b6
--- /dev/null
+++ b/contrib/libs/openssl/crypto/poly1305/poly1305_local.h
@@ -0,0 +1,27 @@
+/*
+ * Copyright 2015-2016 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License").  You may not use
+ * this file except in compliance with the License.  You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+typedef void (*poly1305_blocks_f) (void *ctx, const unsigned char *inp,
+                                   size_t len, unsigned int padbit);
+typedef void (*poly1305_emit_f) (void *ctx, unsigned char mac[16],
+                                 const unsigned int nonce[4]);
+
+struct poly1305_context {
+    double opaque[24];  /* large enough to hold internal state, declared
+                         * 'double' to ensure at least 64-bit invariant
+                         * alignment across all platforms and
+                         * configurations */
+    unsigned int nonce[4];
+    unsigned char data[POLY1305_BLOCK_SIZE];
+    size_t num;
+    struct {
+        poly1305_blocks_f blocks;
+        poly1305_emit_f emit;
+    } func;
+};
diff --git a/contrib/libs/openssl/crypto/poly1305/poly1305_pmeth.c b/contrib/libs/openssl/crypto/poly1305/poly1305_pmeth.c
new file mode 100644
index 0000000000..49a799a12f
--- /dev/null
+++ b/contrib/libs/openssl/crypto/poly1305/poly1305_pmeth.c
@@ -0,0 +1,194 @@
+/*
+ * Copyright 2007-2018 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the OpenSSL license (the "License").  You may not use
+ * this file except in compliance with the License.  You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdio.h>
+#include "internal/cryptlib.h"
+#include <openssl/x509.h>
+#include <openssl/x509v3.h>
+#include <openssl/evp.h>
+#include <openssl/err.h>
+#include "crypto/poly1305.h"
+#include "poly1305_local.h"
+#include "crypto/evp.h"
+
+/* POLY1305 pkey context structure */
+
+typedef struct {
+    ASN1_OCTET_STRING ktmp;     /* Temp storage for key */
+    POLY1305 ctx;
+} POLY1305_PKEY_CTX;
+
+static int pkey_poly1305_init(EVP_PKEY_CTX *ctx)
+{
+    POLY1305_PKEY_CTX *pctx;
+
+    if ((pctx = OPENSSL_zalloc(sizeof(*pctx))) == NULL) {
+        CRYPTOerr(CRYPTO_F_PKEY_POLY1305_INIT, ERR_R_MALLOC_FAILURE);
+        return 0;
+    }
+    pctx->ktmp.type = V_ASN1_OCTET_STRING;
+
+    EVP_PKEY_CTX_set_data(ctx, pctx);
+    EVP_PKEY_CTX_set0_keygen_info(ctx, NULL, 0);
+    return 1;
+}
+
+static void pkey_poly1305_cleanup(EVP_PKEY_CTX *ctx)
+{
+    POLY1305_PKEY_CTX *pctx = EVP_PKEY_CTX_get_data(ctx);
+
+    if (pctx != NULL) {
+        OPENSSL_clear_free(pctx->ktmp.data, pctx->ktmp.length);
+        OPENSSL_clear_free(pctx, sizeof(*pctx));
+        EVP_PKEY_CTX_set_data(ctx, NULL);
+    }
+}
+
+static int pkey_poly1305_copy(EVP_PKEY_CTX *dst, EVP_PKEY_CTX *src)
+{
+    POLY1305_PKEY_CTX *sctx, *dctx;
+
+    /* allocate memory for dst->data and a new POLY1305_CTX in dst->data->ctx */
+    if (!pkey_poly1305_init(dst))
+        return 0;
+    sctx = EVP_PKEY_CTX_get_data(src);
+    dctx = EVP_PKEY_CTX_get_data(dst);
+    if (ASN1_STRING_get0_data(&sctx->ktmp) != NULL &&
+        !ASN1_STRING_copy(&dctx->ktmp, &sctx->ktmp)) {
+        /* cleanup and free the POLY1305_PKEY_CTX in dst->data */
+        pkey_poly1305_cleanup(dst);
+        return 0;
+    }
+    memcpy(&dctx->ctx, &sctx->ctx, sizeof(POLY1305));
+    return 1;
+}
+
+static int pkey_poly1305_keygen(EVP_PKEY_CTX *ctx, EVP_PKEY *pkey)
+{
+    ASN1_OCTET_STRING *key;
+    POLY1305_PKEY_CTX *pctx = EVP_PKEY_CTX_get_data(ctx);
+
+    if (ASN1_STRING_get0_data(&pctx->ktmp) == NULL)
+        return 0;
+    key = ASN1_OCTET_STRING_dup(&pctx->ktmp);
+    if (key == NULL)
+        return 0;
+    return EVP_PKEY_assign_POLY1305(pkey, key);
+}
+
+static int int_update(EVP_MD_CTX *ctx, const void *data, size_t count)
+{
+    POLY1305_PKEY_CTX *pctx = EVP_PKEY_CTX_get_data(EVP_MD_CTX_pkey_ctx(ctx));
+
+    Poly1305_Update(&pctx->ctx, data, count);
+    return 1;
+}
+
+static int poly1305_signctx_init(EVP_PKEY_CTX *ctx, EVP_MD_CTX *mctx)
+{
+    POLY1305_PKEY_CTX *pctx = ctx->data;
+    ASN1_OCTET_STRING *key = (ASN1_OCTET_STRING *)ctx->pkey->pkey.ptr;
+
+    if (key->length != POLY1305_KEY_SIZE)
+        return 0;
+    EVP_MD_CTX_set_flags(mctx, EVP_MD_CTX_FLAG_NO_INIT);
+    EVP_MD_CTX_set_update_fn(mctx, int_update);
+    Poly1305_Init(&pctx->ctx, key->data);
+    return 1;
+}
+static int poly1305_signctx(EVP_PKEY_CTX *ctx, unsigned char *sig, size_t *siglen,
+                            EVP_MD_CTX *mctx)
+{
+    POLY1305_PKEY_CTX *pctx = ctx->data;
+
+    *siglen = POLY1305_DIGEST_SIZE;
+    if (sig != NULL)
+        Poly1305_Final(&pctx->ctx, sig);
+    return 1;
+}
+
+static int pkey_poly1305_ctrl(EVP_PKEY_CTX *ctx, int type, int p1, void *p2)
+{
+    POLY1305_PKEY_CTX *pctx = EVP_PKEY_CTX_get_data(ctx);
+    const unsigned char *key;
+    size_t len;
+
+    switch (type) {
+
+    case EVP_PKEY_CTRL_MD:
+        /* ignore */
+        break;
+
+    case EVP_PKEY_CTRL_SET_MAC_KEY:
+    case EVP_PKEY_CTRL_DIGESTINIT:
+        if (type == EVP_PKEY_CTRL_SET_MAC_KEY) {
+            /* user explicitly setting the key */
+            key = p2;
+            len = p1;
+        } else {
+            /* user indirectly setting the key via EVP_DigestSignInit */
+            key = EVP_PKEY_get0_poly1305(EVP_PKEY_CTX_get0_pkey(ctx), &len);
+        }
+        if (key == NULL || len != POLY1305_KEY_SIZE ||
+            !ASN1_OCTET_STRING_set(&pctx->ktmp, key, len))
+            return 0;
+        Poly1305_Init(&pctx->ctx, ASN1_STRING_get0_data(&pctx->ktmp));
+        break;
+
+    default:
+        return -2;
+
+    }
+    return 1;
+}
+
+static int pkey_poly1305_ctrl_str(EVP_PKEY_CTX *ctx,
+                                  const char *type, const char *value)
+{
+    if (value == NULL)
+        return 0;
+    if (strcmp(type, "key") == 0)
+        return EVP_PKEY_CTX_str2ctrl(ctx, EVP_PKEY_CTRL_SET_MAC_KEY, value);
+    if (strcmp(type, "hexkey") == 0)
+        return EVP_PKEY_CTX_hex2ctrl(ctx, EVP_PKEY_CTRL_SET_MAC_KEY, value);
+    return -2;
+}
+
+const EVP_PKEY_METHOD poly1305_pkey_meth = {
+    EVP_PKEY_POLY1305,
+    EVP_PKEY_FLAG_SIGCTX_CUSTOM, /* we don't deal with a separate MD */
+    pkey_poly1305_init,
+    pkey_poly1305_copy,
+    pkey_poly1305_cleanup,
+
+    0, 0,
+
+    0,
+    pkey_poly1305_keygen,
+
+    0, 0,
+
+    0, 0,
+
+    0, 0,
+
+    poly1305_signctx_init,
+    poly1305_signctx,
+
+    0, 0,
+
+    0, 0,
+
+    0, 0,
+
+    0, 0,
+
+    pkey_poly1305_ctrl,
+    pkey_poly1305_ctrl_str
+};