Import OpenSSL 1.1.1h.vendor/openssl/1.1.1h

86 files changed, 1210 insertions, 3246 deletions

diff --git a/crypto/aes/aes_core.c b/crypto/aes/aes_core.c
index 687dd5829baa..ad00c729e700 100644
--- a/crypto/aes/aes_core.c
+++ b/crypto/aes/aes_core.c
@@ -673,357 +673,6 @@ void AES_decrypt(const unsigned char *in, unsigned char *out,
 
     InvCipher(in, out, rk, key->rounds);
 }
-
-# ifndef OPENSSL_SMALL_FOOTPRINT
-void AES_ctr32_encrypt(const unsigned char *in, unsigned char *out,
-                       size_t blocks, const AES_KEY *key,
-                       const unsigned char *ivec);
-
-static void RawToBits(const u8 raw[64], u64 bits[8])
-{
-    int i, j;
-    u64 in, out;
-
-    memset(bits, 0, 64);
-    for (i = 0; i < 8; i++) {
-        in = 0;
-        for (j = 0; j < 8; j++)
-            in |= ((u64)raw[i * 8 + j]) << (8 * j);
-        out = in & 0xF0F0F0F00F0F0F0FuLL;
-        out |= (in & 0x0F0F0F0F00000000uLL) >> 28;
-        out |= (in & 0x00000000F0F0F0F0uLL) << 28;
-        in = out & 0xCCCC3333CCCC3333uLL;
-        in |= (out & 0x3333000033330000uLL) >> 14;
-        in |= (out & 0x0000CCCC0000CCCCuLL) << 14;
-        out = in & 0xAA55AA55AA55AA55uLL;
-        out |= (in & 0x5500550055005500uLL) >> 7;
-        out |= (in & 0x00AA00AA00AA00AAuLL) << 7;
-        for (j = 0; j < 8; j++) {
-            bits[j] |= (out & 0xFFuLL) << (8 * i);
-            out = out >> 8;
-        }
-    }
-}
-
-static void BitsToRaw(const u64 bits[8], u8 raw[64])
-{
-    int i, j;
-    u64 in, out;
-
-    for (i = 0; i < 8; i++) {
-        in = 0;
-        for (j = 0; j < 8; j++)
-            in |= ((bits[j] >> (8 * i)) & 0xFFuLL) << (8 * j);
-        out = in & 0xF0F0F0F00F0F0F0FuLL;
-        out |= (in & 0x0F0F0F0F00000000uLL) >> 28;
-        out |= (in & 0x00000000F0F0F0F0uLL) << 28;
-        in = out & 0xCCCC3333CCCC3333uLL;
-        in |= (out & 0x3333000033330000uLL) >> 14;
-        in |= (out & 0x0000CCCC0000CCCCuLL) << 14;
-        out = in & 0xAA55AA55AA55AA55uLL;
-        out |= (in & 0x5500550055005500uLL) >> 7;
-        out |= (in & 0x00AA00AA00AA00AAuLL) << 7;
-        for (j = 0; j < 8; j++) {
-            raw[i * 8 + j] = (u8)out;
-            out = out >> 8;
-        }
-    }
-}
-
-static void BitsXtime(u64 state[8])
-{
-    u64 b;
-
-    b = state[7];
-    state[7] = state[6];
-    state[6] = state[5];
-    state[5] = state[4];
-    state[4] = state[3] ^ b;
-    state[3] = state[2] ^ b;
-    state[2] = state[1];
-    state[1] = state[0] ^ b;
-    state[0] = b;
-}
-
-/*
- * This S-box implementation follows a circuit described in
- * Boyar and Peralta: "A new combinational logic minimization
- * technique with applications to cryptology."
- * https://eprint.iacr.org/2009/191.pdf
- *
- * The math is similar to above, in that it uses
- * a tower field of GF(2^2^2^2) but with a different
- * basis representation, that is better suited to
- * logic designs.
- */
-static void BitsSub(u64 state[8])
-{
-    u64 x0, x1, x2, x3, x4, x5, x6, x7;
-    u64 y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11;
-    u64 y12, y13, y14, y15, y16, y17, y18, y19, y20, y21;
-    u64 t0, t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11;
-    u64 t12, t13, t14, t15, t16, t17, t18, t19, t20, t21;
-    u64 t22, t23, t24, t25, t26, t27, t28, t29, t30, t31;
-    u64 t32, t33, t34, t35, t36, t37, t38, t39, t40, t41;
-    u64 t42, t43, t44, t45, t46, t47, t48, t49, t50, t51;
-    u64 t52, t53, t54, t55, t56, t57, t58, t59, t60, t61;
-    u64 t62, t63, t64, t65, t66, t67;
-    u64 z0, z1, z2, z3, z4, z5, z6, z7, z8, z9, z10, z11;
-    u64 z12, z13, z14, z15, z16, z17;
-    u64 s0, s1, s2, s3, s4, s5, s6, s7;
-
-    x7 = state[0];
-    x6 = state[1];
-    x5 = state[2];
-    x4 = state[3];
-    x3 = state[4];
-    x2 = state[5];
-    x1 = state[6];
-    x0 = state[7];
-    y14 = x3 ^ x5;
-    y13 = x0 ^ x6;
-    y9 = x0 ^ x3;
-    y8 = x0 ^ x5;
-    t0 = x1 ^ x2;
-    y1 = t0 ^ x7;
-    y4 = y1 ^ x3;
-    y12 = y13 ^ y14;
-    y2 = y1 ^ x0;
-    y5 = y1 ^ x6;
-    y3 = y5 ^ y8;
-    t1 = x4 ^ y12;
-    y15 = t1 ^ x5;
-    y20 = t1 ^ x1;
-    y6 = y15 ^ x7;
-    y10 = y15 ^ t0;
-    y11 = y20 ^ y9;
-    y7 = x7 ^ y11;
-    y17 = y10 ^ y11;
-    y19 = y10 ^ y8;
-    y16 = t0 ^ y11;
-    y21 = y13 ^ y16;
-    y18 = x0 ^ y16;
-    t2 = y12 & y15;
-    t3 = y3 & y6;
-    t4 = t3 ^ t2;
-    t5 = y4 & x7;
-    t6 = t5 ^ t2;
-    t7 = y13 & y16;
-    t8 = y5 & y1;
-    t9 = t8 ^ t7;
-    t10 = y2 & y7;
-    t11 = t10 ^ t7;
-    t12 = y9 & y11;
-    t13 = y14 & y17;
-    t14 = t13 ^ t12;
-    t15 = y8 & y10;
-    t16 = t15 ^ t12;
-    t17 = t4 ^ t14;
-    t18 = t6 ^ t16;
-    t19 = t9 ^ t14;
-    t20 = t11 ^ t16;
-    t21 = t17 ^ y20;
-    t22 = t18 ^ y19;
-    t23 = t19 ^ y21;
-    t24 = t20 ^ y18;
-    t25 = t21 ^ t22;
-    t26 = t21 & t23;
-    t27 = t24 ^ t26;
-    t28 = t25 & t27;
-    t29 = t28 ^ t22;
-    t30 = t23 ^ t24;
-    t31 = t22 ^ t26;
-    t32 = t31 & t30;
-    t33 = t32 ^ t24;
-    t34 = t23 ^ t33;
-    t35 = t27 ^ t33;
-    t36 = t24 & t35;
-    t37 = t36 ^ t34;
-    t38 = t27 ^ t36;
-    t39 = t29 & t38;
-    t40 = t25 ^ t39;
-    t41 = t40 ^ t37;
-    t42 = t29 ^ t33;
-    t43 = t29 ^ t40;
-    t44 = t33 ^ t37;
-    t45 = t42 ^ t41;
-    z0 = t44 & y15;
-    z1 = t37 & y6;
-    z2 = t33 & x7;
-    z3 = t43 & y16;
-    z4 = t40 & y1;
-    z5 = t29 & y7;
-    z6 = t42 & y11;
-    z7 = t45 & y17;
-    z8 = t41 & y10;
-    z9 = t44 & y12;
-    z10 = t37 & y3;
-    z11 = t33 & y4;
-    z12 = t43 & y13;
-    z13 = t40 & y5;
-    z14 = t29 & y2;
-    z15 = t42 & y9;
-    z16 = t45 & y14;
-    z17 = t41 & y8;
-    t46 = z15 ^ z16;
-    t47 = z10 ^ z11;
-    t48 = z5 ^ z13;
-    t49 = z9 ^ z10;
-    t50 = z2 ^ z12;
-    t51 = z2 ^ z5;
-    t52 = z7 ^ z8;
-    t53 = z0 ^ z3;
-    t54 = z6 ^ z7;
-    t55 = z16 ^ z17;
-    t56 = z12 ^ t48;
-    t57 = t50 ^ t53;
-    t58 = z4 ^ t46;
-    t59 = z3 ^ t54;
-    t60 = t46 ^ t57;
-    t61 = z14 ^ t57;
-    t62 = t52 ^ t58;
-    t63 = t49 ^ t58;
-    t64 = z4 ^ t59;
-    t65 = t61 ^ t62;
-    t66 = z1 ^ t63;
-    s0 = t59 ^ t63;
-    s6 = ~(t56 ^ t62);
-    s7 = ~(t48 ^ t60);
-    t67 = t64 ^ t65;
-    s3 = t53 ^ t66;
-    s4 = t51 ^ t66;
-    s5 = t47 ^ t65;
-    s1 = ~(t64 ^ s3);
-    s2 = ~(t55 ^ t67);
-    state[0] = s7;
-    state[1] = s6;
-    state[2] = s5;
-    state[3] = s4;
-    state[4] = s3;
-    state[5] = s2;
-    state[6] = s1;
-    state[7] = s0;
-}
-
-static void BitsShiftRows(u64 state[8])
-{
-    u64 s, s0;
-    int i;
-
-    for (i = 0; i < 8; i++) {
-        s = state[i];
-        s0 = s & 0x1111111111111111uLL;
-        s0 |= ((s & 0x2220222022202220uLL) >> 4) | ((s & 0x0002000200020002uLL) << 12);
-        s0 |= ((s & 0x4400440044004400uLL) >> 8) | ((s & 0x0044004400440044uLL) << 8);
-        s0 |= ((s & 0x8000800080008000uLL) >> 12) | ((s & 0x0888088808880888uLL) << 4);
-        state[i] = s0;
-    }
-}
-
-static void BitsMixColumns(u64 state[8])
-{
-    u64 s1, s;
-    u64 s0[8];
-    int i;
-
-    for (i = 0; i < 8; i++) {
-        s1 = state[i];
-        s = s1;
-        s ^= ((s & 0xCCCCCCCCCCCCCCCCuLL) >> 2) | ((s & 0x3333333333333333uLL) << 2);
-        s ^= ((s & 0xAAAAAAAAAAAAAAAAuLL) >> 1) | ((s & 0x5555555555555555uLL) << 1);
-        s ^= s1;
-        s0[i] = s;
-    }
-    BitsXtime(state);
-    for (i = 0; i < 8; i++) {
-        s1 = state[i];
-        s = s0[i];
-        s ^= s1;
-        s ^= ((s1 & 0xEEEEEEEEEEEEEEEEuLL) >> 1) | ((s1 & 0x1111111111111111uLL) << 3);
-        state[i] = s;
-    }
-}
-
-static void BitsAddRoundKey(u64 state[8], const u64 key[8])
-{
-    int i;
-
-    for (i = 0; i < 8; i++)
-        state[i] ^= key[i];
-}
-
-void AES_ctr32_encrypt(const unsigned char *in, unsigned char *out,
-                       size_t blocks, const AES_KEY *key,
-                       const unsigned char *ivec)
-{
-    struct {
-        u8 cipher[64];
-        u64 state[8];
-        u64 rd_key[AES_MAXNR + 1][8];
-    } *bs;
-    u32 ctr32;
-    int i;
-
-    ctr32 = GETU32(ivec + 12);
-    if (blocks >= 4
-            && (bs = OPENSSL_malloc(sizeof(*bs)))) {
-        for (i = 0; i < key->rounds + 1; i++) {
-            memcpy(bs->cipher + 0, &key->rd_key[4 * i], 16);
-            memcpy(bs->cipher + 16, bs->cipher, 16);
-            memcpy(bs->cipher + 32, bs->cipher, 32);
-            RawToBits(bs->cipher, bs->rd_key[i]);
-        }
-        while (blocks) {
-            memcpy(bs->cipher, ivec, 12);
-            PUTU32(bs->cipher + 12, ctr32);
-            ctr32++;
-            memcpy(bs->cipher + 16, ivec, 12);
-            PUTU32(bs->cipher + 28, ctr32);
-            ctr32++;
-            memcpy(bs->cipher + 32, ivec, 12);
-            PUTU32(bs->cipher + 44, ctr32);
-            ctr32++;
-            memcpy(bs->cipher + 48, ivec, 12);
-            PUTU32(bs->cipher + 60, ctr32);
-            ctr32++;
-            RawToBits(bs->cipher, bs->state);
-            BitsAddRoundKey(bs->state, bs->rd_key[0]);
-            for (i = 1; i < key->rounds; i++) {
-                BitsSub(bs->state);
-                BitsShiftRows(bs->state);
-                BitsMixColumns(bs->state);
-                BitsAddRoundKey(bs->state, bs->rd_key[i]);
-            }
-            BitsSub(bs->state);
-            BitsShiftRows(bs->state);
-            BitsAddRoundKey(bs->state, bs->rd_key[key->rounds]);
-            BitsToRaw(bs->state, bs->cipher);
-            for (i = 0; i < 64 && blocks; i++) {
-                out[i] = in[i] ^ bs->cipher[i];
-                if ((i & 15) == 15)
-                    blocks--;
-            }
-            in += i;
-            out += i;
-        }
-        OPENSSL_clear_free(bs, sizeof(*bs));
-    } else {
-        unsigned char cipher[16];
-
-        while (blocks) {
-            memcpy(cipher, ivec, 12);
-            PUTU32(cipher + 12, ctr32);
-            AES_encrypt(cipher, cipher, key);
-            for (i = 0; i < 16; i++)
-                out[i] = in[i] ^ cipher[i];
-            in += 16;
-            out += 16;
-            ctr32++;
-            blocks--;
-        }
-    }
-}
-# endif
 #elif !defined(AES_ASM)
 /*-
 Te0[x] = S [x].[02, 01, 01, 03];
diff --git a/crypto/aes/aes_ige.c b/crypto/aes/aes_ige.c
index dce4ef11be4f..804b3a723d1f 100644
--- a/crypto/aes/aes_ige.c
+++ b/crypto/aes/aes_ige.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2006-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2006-2020 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
@@ -12,11 +12,6 @@
 #include <openssl/aes.h>
 #include "aes_local.h"
 
-#define N_WORDS (AES_BLOCK_SIZE / sizeof(unsigned long))
-typedef struct {
-    unsigned long data[N_WORDS];
-} aes_block_t;
-
 /* XXX: probably some better way to do this */
 #if defined(__i386__) || defined(__x86_64__)
 # define UNALIGNED_MEMOPS_ARE_FAST 1
@@ -24,6 +19,15 @@ typedef struct {
 # define UNALIGNED_MEMOPS_ARE_FAST 0
 #endif
 
+#define N_WORDS (AES_BLOCK_SIZE / sizeof(unsigned long))
+typedef struct {
+    unsigned long data[N_WORDS];
+#if defined(__GNUC__) && UNALIGNED_MEMOPS_ARE_FAST
+} aes_block_t __attribute((__aligned__(1)));
+#else
+} aes_block_t;
+#endif
+
 #if UNALIGNED_MEMOPS_ARE_FAST
 # define load_block(d, s)        (d) = *(const aes_block_t *)(s)
 # define store_block(d, s)       *(aes_block_t *)(d) = (s)
diff --git a/crypto/aes/asm/aesni-mb-x86_64.pl b/crypto/aes/asm/aesni-mb-x86_64.pl
index be2434f12041..a80cfdc13948 100755
--- a/crypto/aes/asm/aesni-mb-x86_64.pl
+++ b/crypto/aes/asm/aesni-mb-x86_64.pl
@@ -70,7 +70,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/aes/asm/aesni-sha1-x86_64.pl b/crypto/aes/asm/aesni-sha1-x86_64.pl
index 42fe5d469ea6..04fd13be5e09 100755
--- a/crypto/aes/asm/aesni-sha1-x86_64.pl
+++ b/crypto/aes/asm/aesni-sha1-x86_64.pl
@@ -108,7 +108,7 @@ $avx=1 if (!$avx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
 $avx=1 if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	   `ml64 2>&1` =~ /Version ([0-9]+)\./ &&
 	   $1>=10);
-$avx=1 if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/ && $2>=3.0);
+$avx=1 if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/ && $2>=3.0);
 
 $shaext=1;	### set to zero if compiling for 1.0.1
 
diff --git a/crypto/aes/asm/aesni-sha256-x86_64.pl b/crypto/aes/asm/aesni-sha256-x86_64.pl
index f1b144eccbdd..ff9b18507da0 100755
--- a/crypto/aes/asm/aesni-sha256-x86_64.pl
+++ b/crypto/aes/asm/aesni-sha256-x86_64.pl
@@ -70,7 +70,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=12);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/asn1/d2i_pr.c b/crypto/asn1/d2i_pr.c
index 6ec010738049..7b127d2092fa 100644
--- a/crypto/asn1/d2i_pr.c
+++ b/crypto/asn1/d2i_pr.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -56,6 +56,8 @@ EVP_PKEY *d2i_PrivateKey(int type, EVP_PKEY **a, const unsigned char **pp,
                 goto err;
             EVP_PKEY_free(ret);
             ret = tmp;
+            if (EVP_PKEY_type(type) != EVP_PKEY_base_id(ret))
+                goto err;
         } else {
             ASN1err(ASN1_F_D2I_PRIVATEKEY, ERR_R_ASN1_LIB);
             goto err;
diff --git a/crypto/asn1/x_algor.c b/crypto/asn1/x_algor.c
index 4c4a718850ee..c9a8f1e9d1d4 100644
--- a/crypto/asn1/x_algor.c
+++ b/crypto/asn1/x_algor.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1998-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1998-2020 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
@@ -92,3 +92,35 @@ int X509_ALGOR_cmp(const X509_ALGOR *a, const X509_ALGOR *b)
         return 0;
     return ASN1_TYPE_cmp(a->parameter, b->parameter);
 }
+
+int X509_ALGOR_copy(X509_ALGOR *dest, const X509_ALGOR *src)
+{
+    if (src == NULL || dest == NULL)
+	return 0;
+
+    if (dest->algorithm)
+         ASN1_OBJECT_free(dest->algorithm);
+    dest->algorithm = NULL;
+
+    if (dest->parameter)
+        ASN1_TYPE_free(dest->parameter);
+    dest->parameter = NULL;
+
+    if (src->algorithm)
+        if ((dest->algorithm = OBJ_dup(src->algorithm)) == NULL)
+	    return 0;
+
+    if (src->parameter) {
+        dest->parameter = ASN1_TYPE_new();
+        if (dest->parameter == NULL)
+            return 0;
+
+        /* Assuming this is also correct for a BOOL.
+         * set does copy as a side effect.
+         */
+        if (ASN1_TYPE_set1(dest->parameter, 
+                src->parameter->type, src->parameter->value.ptr) == 0)
+            return 0;
+    }
+    return 1;
+}
diff --git a/crypto/bio/b_print.c b/crypto/bio/b_print.c
index 8ef90ac1d4f8..41b7f5e2f61d 100644
--- a/crypto/bio/b_print.c
+++ b/crypto/bio/b_print.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -635,7 +635,11 @@ fmtfp(char **sbuffer,
             fvalue = tmpvalue;
     }
     ufvalue = abs_val(fvalue);
-    if (ufvalue > ULONG_MAX) {
+    /*
+     * By subtracting 65535 (2^16-1) we cancel the low order 15 bits
+     * of ULONG_MAX to avoid using imprecise floating point values.
+     */
+    if (ufvalue >= (double)(ULONG_MAX - 65535) + 65536.0) {
         /* Number too big */
         return 0;
     }
diff --git a/crypto/bio/bss_acpt.c b/crypto/bio/bss_acpt.c
index 5a2cb50dfc39..4461eae2333d 100644
--- a/crypto/bio/bss_acpt.c
+++ b/crypto/bio/bss_acpt.c
@@ -434,8 +434,10 @@ static long acpt_ctrl(BIO *b, int cmd, long num, void *ptr)
                 b->init = 1;
             } else if (num == 1) {
                 OPENSSL_free(data->param_serv);
-                data->param_serv = BUF_strdup(ptr);
-                b->init = 1;
+                if ((data->param_serv = OPENSSL_strdup(ptr)) == NULL)
+                    ret = 0;
+                else
+                    b->init = 1;
             } else if (num == 2) {
                 data->bind_mode |= BIO_SOCK_NONBLOCK;
             } else if (num == 3) {
diff --git a/crypto/bio/bss_conn.c b/crypto/bio/bss_conn.c
index dd43a406018c..807a82b23ba2 100644
--- a/crypto/bio/bss_conn.c
+++ b/crypto/bio/bss_conn.c
@@ -186,8 +186,17 @@ static int conn_state(BIO *b, BIO_CONNECT *c)
 
         case BIO_CONN_S_BLOCKED_CONNECT:
             i = BIO_sock_error(b->num);
-            if (i) {
+            if (i != 0) {
                 BIO_clear_retry_flags(b);
+                if ((c->addr_iter = BIO_ADDRINFO_next(c->addr_iter)) != NULL) {
+                    /*
+                     * if there are more addresses to try, do that first
+                     */
+                    BIO_closesocket(b->num);
+                    c->state = BIO_CONN_S_CREATE_SOCKET;
+                    ERR_clear_error();
+                    break;
+                }
                 SYSerr(SYS_F_CONNECT, i);
                 ERR_add_error_data(4,
                                    "hostname=", c->param_hostname,
@@ -407,12 +416,13 @@ static long conn_ctrl(BIO *b, int cmd, long num, void *ptr)
     case BIO_C_SET_CONNECT:
         if (ptr != NULL) {
             b->init = 1;
-            if (num == 0) {
+            if (num == 0) { /* BIO_set_conn_hostname */
                 char *hold_service = data->param_service;
                 /* We affect the hostname regardless.  However, the input
                  * string might contain a host:service spec, so we must
                  * parse it, which might or might not affect the service
                  */
+
                 OPENSSL_free(data->param_hostname);
                 data->param_hostname = NULL;
                 ret = BIO_parse_hostserv(ptr,
@@ -421,19 +431,29 @@ static long conn_ctrl(BIO *b, int cmd, long num, void *ptr)
                                          BIO_PARSE_PRIO_HOST);
                 if (hold_service != data->param_service)
                     OPENSSL_free(hold_service);
-            } else if (num == 1) {
+            } else if (num == 1) { /* BIO_set_conn_port */
                 OPENSSL_free(data->param_service);
-                data->param_service = BUF_strdup(ptr);
-            } else if (num == 2) {
+                if ((data->param_service = OPENSSL_strdup(ptr)) == NULL)
+                    ret = 0;
+            } else if (num == 2) { /* BIO_set_conn_address */
                 const BIO_ADDR *addr = (const BIO_ADDR *)ptr;
+                char *host = BIO_ADDR_hostname_string(addr, 1);
+                char *service = BIO_ADDR_service_string(addr, 1);
+
+                ret = host != NULL && service != NULL;
                 if (ret) {
-                    data->param_hostname = BIO_ADDR_hostname_string(addr, 1);
-                    data->param_service = BIO_ADDR_service_string(addr, 1);
+                    OPENSSL_free(data->param_hostname);
+                    data->param_hostname = host;
+                    OPENSSL_free(data->param_service);
+                    data->param_service = service;
                     BIO_ADDRINFO_free(data->addr_first);
                     data->addr_first = NULL;
                     data->addr_iter = NULL;
+                } else {
+                    OPENSSL_free(host);
+                    OPENSSL_free(service);
                 }
-            } else if (num == 3) {
+            } else if (num == 3) { /* BIO_set_conn_ip_family */
                 data->connect_family = *(int *)ptr;
             } else {
                 ret = 0;
diff --git a/crypto/bn/asm/rsaz-avx2.pl b/crypto/bn/asm/rsaz-avx2.pl
index 9b54f9d9ed3c..0be771febc16 100755
--- a/crypto/bn/asm/rsaz-avx2.pl
+++ b/crypto/bn/asm/rsaz-avx2.pl
@@ -66,7 +66,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$addx = ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|based on LLVM) ([0-9]+)\.([0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|based on LLVM) ([0-9]+)\.([0-9]+)/) {
 	my $ver = $2 + $3/100.0;	# 3.1->3.01, 3.10->3.10
 	$avx = ($ver>=3.0) + ($ver>=3.01);
 	$addx = ($ver>=3.03);
diff --git a/crypto/bn/asm/rsaz-x86_64.pl b/crypto/bn/asm/rsaz-x86_64.pl
index 8172dace74ee..cf08ce9b8356 100755
--- a/crypto/bn/asm/rsaz-x86_64.pl
+++ b/crypto/bn/asm/rsaz-x86_64.pl
@@ -81,7 +81,7 @@ if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$addx = ($1>=12);
 }
 
-if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
 	my $ver = $2 + $3/100.0;	# 3.1->3.01, 3.10->3.10
 	$addx = ($ver>=3.03);
 }
diff --git a/crypto/bn/asm/x86_64-mont.pl b/crypto/bn/asm/x86_64-mont.pl
index e00cac448b89..f14d4e63b975 100755
--- a/crypto/bn/asm/x86_64-mont.pl
+++ b/crypto/bn/asm/x86_64-mont.pl
@@ -75,7 +75,7 @@ if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$addx = ($1>=12);
 }
 
-if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
 	my $ver = $2 + $3/100.0;	# 3.1->3.01, 3.10->3.10
 	$addx = ($ver>=3.03);
 }
diff --git a/crypto/bn/asm/x86_64-mont5.pl b/crypto/bn/asm/x86_64-mont5.pl
index 887770b30db1..8c37d132e476 100755
--- a/crypto/bn/asm/x86_64-mont5.pl
+++ b/crypto/bn/asm/x86_64-mont5.pl
@@ -60,7 +60,7 @@ if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$addx = ($1>=12);
 }
 
-if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
 	my $ver = $2 + $3/100.0;	# 3.1->3.01, 3.10->3.10
 	$addx = ($ver>=3.03);
 }
diff --git a/crypto/bn/bn_gcd.c b/crypto/bn/bn_gcd.c
index ef81acb77ba6..0941f7b97f3f 100644
--- a/crypto/bn/bn_gcd.c
+++ b/crypto/bn/bn_gcd.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -10,22 +10,189 @@
 #include "internal/cryptlib.h"
 #include "bn_local.h"
 
-/* solves ax == 1 (mod n) */
-static BIGNUM *BN_mod_inverse_no_branch(BIGNUM *in,
-                                        const BIGNUM *a, const BIGNUM *n,
-                                        BN_CTX *ctx);
-
-BIGNUM *BN_mod_inverse(BIGNUM *in,
-                       const BIGNUM *a, const BIGNUM *n, BN_CTX *ctx)
+/*
+ * bn_mod_inverse_no_branch is a special version of BN_mod_inverse. It does
+ * not contain branches that may leak sensitive information.
+ *
+ * This is a static function, we ensure all callers in this file pass valid
+ * arguments: all passed pointers here are non-NULL.
+ */
+static ossl_inline
+BIGNUM *bn_mod_inverse_no_branch(BIGNUM *in,
+                                 const BIGNUM *a, const BIGNUM *n,
+                                 BN_CTX *ctx, int *pnoinv)
 {
-    BIGNUM *rv;
-    int noinv;
-    rv = int_bn_mod_inverse(in, a, n, ctx, &noinv);
-    if (noinv)
-        BNerr(BN_F_BN_MOD_INVERSE, BN_R_NO_INVERSE);
-    return rv;
+    BIGNUM *A, *B, *X, *Y, *M, *D, *T, *R = NULL;
+    BIGNUM *ret = NULL;
+    int sign;
+
+    bn_check_top(a);
+    bn_check_top(n);
+
+    BN_CTX_start(ctx);
+    A = BN_CTX_get(ctx);
+    B = BN_CTX_get(ctx);
+    X = BN_CTX_get(ctx);
+    D = BN_CTX_get(ctx);
+    M = BN_CTX_get(ctx);
+    Y = BN_CTX_get(ctx);
+    T = BN_CTX_get(ctx);
+    if (T == NULL)
+        goto err;
+
+    if (in == NULL)
+        R = BN_new();
+    else
+        R = in;
+    if (R == NULL)
+        goto err;
+
+    BN_one(X);
+    BN_zero(Y);
+    if (BN_copy(B, a) == NULL)
+        goto err;
+    if (BN_copy(A, n) == NULL)
+        goto err;
+    A->neg = 0;
+
+    if (B->neg || (BN_ucmp(B, A) >= 0)) {
+        /*
+         * Turn BN_FLG_CONSTTIME flag on, so that when BN_div is invoked,
+         * BN_div_no_branch will be called eventually.
+         */
+         {
+            BIGNUM local_B;
+            bn_init(&local_B);
+            BN_with_flags(&local_B, B, BN_FLG_CONSTTIME);
+            if (!BN_nnmod(B, &local_B, A, ctx))
+                goto err;
+            /* Ensure local_B goes out of scope before any further use of B */
+        }
+    }
+    sign = -1;
+    /*-
+     * From  B = a mod |n|,  A = |n|  it follows that
+     *
+     *      0 <= B < A,
+     *     -sign*X*a  ==  B   (mod |n|),
+     *      sign*Y*a  ==  A   (mod |n|).
+     */
+
+    while (!BN_is_zero(B)) {
+        BIGNUM *tmp;
+
+        /*-
+         *      0 < B < A,
+         * (*) -sign*X*a  ==  B   (mod |n|),
+         *      sign*Y*a  ==  A   (mod |n|)
+         */
+
+        /*
+         * Turn BN_FLG_CONSTTIME flag on, so that when BN_div is invoked,
+         * BN_div_no_branch will be called eventually.
+         */
+        {
+            BIGNUM local_A;
+            bn_init(&local_A);
+            BN_with_flags(&local_A, A, BN_FLG_CONSTTIME);
+
+            /* (D, M) := (A/B, A%B) ... */
+            if (!BN_div(D, M, &local_A, B, ctx))
+                goto err;
+            /* Ensure local_A goes out of scope before any further use of A */
+        }
+
+        /*-
+         * Now
+         *      A = D*B + M;
+         * thus we have
+         * (**)  sign*Y*a  ==  D*B + M   (mod |n|).
+         */
+
+        tmp = A;                /* keep the BIGNUM object, the value does not
+                                 * matter */
+
+        /* (A, B) := (B, A mod B) ... */
+        A = B;
+        B = M;
+        /* ... so we have  0 <= B < A  again */
+
+        /*-
+         * Since the former  M  is now  B  and the former  B  is now  A,
+         * (**) translates into
+         *       sign*Y*a  ==  D*A + B    (mod |n|),
+         * i.e.
+         *       sign*Y*a - D*A  ==  B    (mod |n|).
+         * Similarly, (*) translates into
+         *      -sign*X*a  ==  A          (mod |n|).
+         *
+         * Thus,
+         *   sign*Y*a + D*sign*X*a  ==  B  (mod |n|),
+         * i.e.
+         *        sign*(Y + D*X)*a  ==  B  (mod |n|).
+         *
+         * So if we set  (X, Y, sign) := (Y + D*X, X, -sign), we arrive back at
+         *      -sign*X*a  ==  B   (mod |n|),
+         *       sign*Y*a  ==  A   (mod |n|).
+         * Note that  X  and  Y  stay non-negative all the time.
+         */
+
+        if (!BN_mul(tmp, D, X, ctx))
+            goto err;
+        if (!BN_add(tmp, tmp, Y))
+            goto err;
+
+        M = Y;                  /* keep the BIGNUM object, the value does not
+                                 * matter */
+        Y = X;
+        X = tmp;
+        sign = -sign;
+    }
+
+    /*-
+     * The while loop (Euclid's algorithm) ends when
+     *      A == gcd(a,n);
+     * we have
+     *       sign*Y*a  ==  A  (mod |n|),
+     * where  Y  is non-negative.
+     */
+
+    if (sign < 0) {
+        if (!BN_sub(Y, n, Y))
+            goto err;
+    }
+    /* Now  Y*a  ==  A  (mod |n|).  */
+
+    if (BN_is_one(A)) {
+        /* Y*a == 1  (mod |n|) */
+        if (!Y->neg && BN_ucmp(Y, n) < 0) {
+            if (!BN_copy(R, Y))
+                goto err;
+        } else {
+            if (!BN_nnmod(R, Y, n, ctx))
+                goto err;
+        }
+    } else {
+        *pnoinv = 1;
+        /* caller sets the BN_R_NO_INVERSE error */
+        goto err;
+    }
+
+    ret = R;
+    *pnoinv = 0;
+
+ err:
+    if ((ret == NULL) && (in == NULL))
+        BN_free(R);
+    BN_CTX_end(ctx);
+    bn_check_top(ret);
+    return ret;
 }
 
+/*
+ * This is an internal function, we assume all callers pass valid arguments:
+ * all pointers passed here are assumed non-NULL.
+ */
 BIGNUM *int_bn_mod_inverse(BIGNUM *in,
                            const BIGNUM *a, const BIGNUM *n, BN_CTX *ctx,
                            int *pnoinv)
@@ -36,17 +203,15 @@ BIGNUM *int_bn_mod_inverse(BIGNUM *in,
 
     /* This is invalid input so we don't worry about constant time here */
     if (BN_abs_is_word(n, 1) || BN_is_zero(n)) {
-        if (pnoinv != NULL)
-            *pnoinv = 1;
+        *pnoinv = 1;
         return NULL;
     }
 
-    if (pnoinv != NULL)
-        *pnoinv = 0;
+    *pnoinv = 0;
 
     if ((BN_get_flags(a, BN_FLG_CONSTTIME) != 0)
         || (BN_get_flags(n, BN_FLG_CONSTTIME) != 0)) {
-        return BN_mod_inverse_no_branch(in, a, n, ctx);
+        return bn_mod_inverse_no_branch(in, a, n, ctx, pnoinv);
     }
 
     bn_check_top(a);
@@ -332,8 +497,7 @@ BIGNUM *int_bn_mod_inverse(BIGNUM *in,
                 goto err;
         }
     } else {
-        if (pnoinv)
-            *pnoinv = 1;
+        *pnoinv = 1;
         goto err;
     }
     ret = R;
@@ -345,175 +509,27 @@ BIGNUM *int_bn_mod_inverse(BIGNUM *in,
     return ret;
 }
 
-/*
- * BN_mod_inverse_no_branch is a special version of BN_mod_inverse. It does
- * not contain branches that may leak sensitive information.
- */
-static BIGNUM *BN_mod_inverse_no_branch(BIGNUM *in,
-                                        const BIGNUM *a, const BIGNUM *n,
-                                        BN_CTX *ctx)
+/* solves ax == 1 (mod n) */
+BIGNUM *BN_mod_inverse(BIGNUM *in,
+                       const BIGNUM *a, const BIGNUM *n, BN_CTX *ctx)
 {
-    BIGNUM *A, *B, *X, *Y, *M, *D, *T, *R = NULL;
-    BIGNUM *ret = NULL;
-    int sign;
-
-    bn_check_top(a);
-    bn_check_top(n);
-
-    BN_CTX_start(ctx);
-    A = BN_CTX_get(ctx);
-    B = BN_CTX_get(ctx);
-    X = BN_CTX_get(ctx);
-    D = BN_CTX_get(ctx);
-    M = BN_CTX_get(ctx);
-    Y = BN_CTX_get(ctx);
-    T = BN_CTX_get(ctx);
-    if (T == NULL)
-        goto err;
-
-    if (in == NULL)
-        R = BN_new();
-    else
-        R = in;
-    if (R == NULL)
-        goto err;
-
-    BN_one(X);
-    BN_zero(Y);
-    if (BN_copy(B, a) == NULL)
-        goto err;
-    if (BN_copy(A, n) == NULL)
-        goto err;
-    A->neg = 0;
-
-    if (B->neg || (BN_ucmp(B, A) >= 0)) {
-        /*
-         * Turn BN_FLG_CONSTTIME flag on, so that when BN_div is invoked,
-         * BN_div_no_branch will be called eventually.
-         */
-         {
-            BIGNUM local_B;
-            bn_init(&local_B);
-            BN_with_flags(&local_B, B, BN_FLG_CONSTTIME);
-            if (!BN_nnmod(B, &local_B, A, ctx))
-                goto err;
-            /* Ensure local_B goes out of scope before any further use of B */
-        }
-    }
-    sign = -1;
-    /*-
-     * From  B = a mod |n|,  A = |n|  it follows that
-     *
-     *      0 <= B < A,
-     *     -sign*X*a  ==  B   (mod |n|),
-     *      sign*Y*a  ==  A   (mod |n|).
-     */
-
-    while (!BN_is_zero(B)) {
-        BIGNUM *tmp;
-
-        /*-
-         *      0 < B < A,
-         * (*) -sign*X*a  ==  B   (mod |n|),
-         *      sign*Y*a  ==  A   (mod |n|)
-         */
-
-        /*
-         * Turn BN_FLG_CONSTTIME flag on, so that when BN_div is invoked,
-         * BN_div_no_branch will be called eventually.
-         */
-        {
-            BIGNUM local_A;
-            bn_init(&local_A);
-            BN_with_flags(&local_A, A, BN_FLG_CONSTTIME);
+    BN_CTX *new_ctx = NULL;
+    BIGNUM *rv;
+    int noinv = 0;
 
-            /* (D, M) := (A/B, A%B) ... */
-            if (!BN_div(D, M, &local_A, B, ctx))
-                goto err;
-            /* Ensure local_A goes out of scope before any further use of A */
+    if (ctx == NULL) {
+        ctx = new_ctx = BN_CTX_new();
+        if (ctx == NULL) {
+            BNerr(BN_F_BN_MOD_INVERSE, ERR_R_MALLOC_FAILURE);
+            return NULL;
         }
-
-        /*-
-         * Now
-         *      A = D*B + M;
-         * thus we have
-         * (**)  sign*Y*a  ==  D*B + M   (mod |n|).
-         */
-
-        tmp = A;                /* keep the BIGNUM object, the value does not
-                                 * matter */
-
-        /* (A, B) := (B, A mod B) ... */
-        A = B;
-        B = M;
-        /* ... so we have  0 <= B < A  again */
-
-        /*-
-         * Since the former  M  is now  B  and the former  B  is now  A,
-         * (**) translates into
-         *       sign*Y*a  ==  D*A + B    (mod |n|),
-         * i.e.
-         *       sign*Y*a - D*A  ==  B    (mod |n|).
-         * Similarly, (*) translates into
-         *      -sign*X*a  ==  A          (mod |n|).
-         *
-         * Thus,
-         *   sign*Y*a + D*sign*X*a  ==  B  (mod |n|),
-         * i.e.
-         *        sign*(Y + D*X)*a  ==  B  (mod |n|).
-         *
-         * So if we set  (X, Y, sign) := (Y + D*X, X, -sign), we arrive back at
-         *      -sign*X*a  ==  B   (mod |n|),
-         *       sign*Y*a  ==  A   (mod |n|).
-         * Note that  X  and  Y  stay non-negative all the time.
-         */
-
-        if (!BN_mul(tmp, D, X, ctx))
-            goto err;
-        if (!BN_add(tmp, tmp, Y))
-            goto err;
-
-        M = Y;                  /* keep the BIGNUM object, the value does not
-                                 * matter */
-        Y = X;
-        X = tmp;
-        sign = -sign;
-    }
-
-    /*-
-     * The while loop (Euclid's algorithm) ends when
-     *      A == gcd(a,n);
-     * we have
-     *       sign*Y*a  ==  A  (mod |n|),
-     * where  Y  is non-negative.
-     */
-
-    if (sign < 0) {
-        if (!BN_sub(Y, n, Y))
-            goto err;
     }
-    /* Now  Y*a  ==  A  (mod |n|).  */
 
-    if (BN_is_one(A)) {
-        /* Y*a == 1  (mod |n|) */
-        if (!Y->neg && BN_ucmp(Y, n) < 0) {
-            if (!BN_copy(R, Y))
-                goto err;
-        } else {
-            if (!BN_nnmod(R, Y, n, ctx))
-                goto err;
-        }
-    } else {
-        BNerr(BN_F_BN_MOD_INVERSE_NO_BRANCH, BN_R_NO_INVERSE);
-        goto err;
-    }
-    ret = R;
- err:
-    if ((ret == NULL) && (in == NULL))
-        BN_free(R);
-    BN_CTX_end(ctx);
-    bn_check_top(ret);
-    return ret;
+    rv = int_bn_mod_inverse(in, a, n, ctx, &noinv);
+    if (noinv)
+        BNerr(BN_F_BN_MOD_INVERSE, BN_R_NO_INVERSE);
+    BN_CTX_free(new_ctx);
+    return rv;
 }
 
 /*-
diff --git a/crypto/bn/bn_lib.c b/crypto/bn/bn_lib.c
index 86d4956c8a8c..eb4a31849bef 100644
--- a/crypto/bn/bn_lib.c
+++ b/crypto/bn/bn_lib.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -87,6 +87,15 @@ const BIGNUM *BN_value_one(void)
     return &const_one;
 }
 
+/*
+ * Old Visual Studio ARM compiler miscompiles BN_num_bits_word()
+ * https://mta.openssl.org/pipermail/openssl-users/2018-August/008465.html
+ */
+#if defined(_MSC_VER) && defined(_ARM_) && defined(_WIN32_WCE) \
+    && _MSC_VER>=1400 && _MSC_VER<1501
+# define MS_BROKEN_BN_num_bits_word
+# pragma optimize("", off)
+#endif
 int BN_num_bits_word(BN_ULONG l)
 {
     BN_ULONG x, mask;
@@ -131,6 +140,9 @@ int BN_num_bits_word(BN_ULONG l)
 
     return bits;
 }
+#ifdef MS_BROKEN_BN_num_bits_word
+# pragma optimize("", on)
+#endif
 
 /*
  * This function still leaks `a->dmax`: it's caller's responsibility to
@@ -322,15 +334,19 @@ BIGNUM *BN_dup(const BIGNUM *a)
 
 BIGNUM *BN_copy(BIGNUM *a, const BIGNUM *b)
 {
+    int bn_words;
+
     bn_check_top(b);
 
+    bn_words = BN_get_flags(b, BN_FLG_CONSTTIME) ? b->dmax : b->top;
+
     if (a == b)
         return a;
-    if (bn_wexpand(a, b->top) == NULL)
+    if (bn_wexpand(a, bn_words) == NULL)
         return NULL;
 
     if (b->top > 0)
-        memcpy(a->d, b->d, sizeof(b->d[0]) * b->top);
+        memcpy(a->d, b->d, sizeof(b->d[0]) * bn_words);
 
     a->neg = b->neg;
     a->top = b->top;
diff --git a/crypto/bn/bn_mpi.c b/crypto/bn/bn_mpi.c
index bdbe822415c7..0902da5d076e 100644
--- a/crypto/bn/bn_mpi.c
+++ b/crypto/bn/bn_mpi.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -45,7 +45,7 @@ BIGNUM *BN_mpi2bn(const unsigned char *d, int n, BIGNUM *ain)
     int neg = 0;
     BIGNUM *a = NULL;
 
-    if (n < 4) {
+    if (n < 4 || (d[0] & 0x80) != 0) {
         BNerr(BN_F_BN_MPI2BN, BN_R_INVALID_LENGTH);
         return NULL;
     }
diff --git a/crypto/chacha/asm/chacha-x86.pl b/crypto/chacha/asm/chacha-x86.pl
index a1f5694b691e..492fda5f114c 100755
--- a/crypto/chacha/asm/chacha-x86.pl
+++ b/crypto/chacha/asm/chacha-x86.pl
@@ -62,7 +62,7 @@ $ymm=1 if ($xmm && !$ymm && $ARGV[0] eq "win32" &&
 		$1>=10);	# first version supporting AVX
 
 $ymm=1 if ($xmm && !$ymm &&
-		`$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/ &&
+		`$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/ &&
 		$2>=3.0);	# first version supporting AVX
 
 $a="eax";
diff --git a/crypto/chacha/asm/chacha-x86_64.pl b/crypto/chacha/asm/chacha-x86_64.pl
index 647d2537ceae..227ee59ff2ba 100755
--- a/crypto/chacha/asm/chacha-x86_64.pl
+++ b/crypto/chacha/asm/chacha-x86_64.pl
@@ -85,7 +85,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/cmac/cmac.c b/crypto/cmac/cmac.c
index 6989c32d0660..1fac53101687 100644
--- a/crypto/cmac/cmac.c
+++ b/crypto/cmac/cmac.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2010-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2010-2020 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
@@ -116,11 +116,18 @@ int CMAC_Init(CMAC_CTX *ctx, const void *key, size_t keylen,
         return 1;
     }
     /* Initialise context */
-    if (cipher && !EVP_EncryptInit_ex(ctx->cctx, cipher, impl, NULL, NULL))
-        return 0;
+    if (cipher != NULL) {
+        /* Ensure we can't use this ctx until we also have a key */
+        ctx->nlast_block = -1;
+        if (!EVP_EncryptInit_ex(ctx->cctx, cipher, impl, NULL, NULL))
+            return 0;
+    }
     /* Non-NULL key means initialisation complete */
-    if (key) {
+    if (key != NULL) {
         int bl;
+
+        /* If anything fails then ensure we can't use this ctx */
+        ctx->nlast_block = -1;
         if (!EVP_CIPHER_CTX_cipher(ctx->cctx))
             return 0;
         if (!EVP_CIPHER_CTX_set_key_length(ctx->cctx, keylen))
@@ -128,7 +135,7 @@ int CMAC_Init(CMAC_CTX *ctx, const void *key, size_t keylen,
         if (!EVP_EncryptInit_ex(ctx->cctx, NULL, NULL, key, zero_iv))
             return 0;
         bl = EVP_CIPHER_CTX_block_size(ctx->cctx);
-        if (!EVP_Cipher(ctx->cctx, ctx->tbl, zero_iv, bl))
+        if (EVP_Cipher(ctx->cctx, ctx->tbl, zero_iv, bl) <= 0)
             return 0;
         make_kn(ctx->k1, ctx->tbl, bl);
         make_kn(ctx->k2, ctx->k1, bl);
@@ -166,12 +173,12 @@ int CMAC_Update(CMAC_CTX *ctx, const void *in, size_t dlen)
             return 1;
         data += nleft;
         /* Else not final block so encrypt it */
-        if (!EVP_Cipher(ctx->cctx, ctx->tbl, ctx->last_block, bl))
+        if (EVP_Cipher(ctx->cctx, ctx->tbl, ctx->last_block, bl) <= 0)
             return 0;
     }
     /* Encrypt all but one of the complete blocks left */
     while (dlen > bl) {
-        if (!EVP_Cipher(ctx->cctx, ctx->tbl, data, bl))
+        if (EVP_Cipher(ctx->cctx, ctx->tbl, data, bl) <= 0)
             return 0;
         dlen -= bl;
         data += bl;
diff --git a/crypto/cms/cms_lib.c b/crypto/cms/cms_lib.c
index 57afba436115..be4c2c703f1a 100644
--- a/crypto/cms/cms_lib.c
+++ b/crypto/cms/cms_lib.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2008-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2008-2020 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
@@ -92,12 +92,13 @@ BIO *CMS_dataInit(CMS_ContentInfo *cms, BIO *icont)
 
     default:
         CMSerr(CMS_F_CMS_DATAINIT, CMS_R_UNSUPPORTED_TYPE);
-        return NULL;
+        goto err;
     }
 
     if (cmsbio)
         return BIO_push(cmsbio, cont);
 
+err:
     if (!icont)
         BIO_free(cont);
     return NULL;
diff --git a/crypto/cms/cms_sd.c b/crypto/cms/cms_sd.c
index 29ba4c1b1334..3f2a782565a8 100644
--- a/crypto/cms/cms_sd.c
+++ b/crypto/cms/cms_sd.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2008-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2008-2020 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
@@ -897,8 +897,10 @@ int CMS_add_simple_smimecap(STACK_OF(X509_ALGOR) **algs,
     ASN1_INTEGER *key = NULL;
     if (keysize > 0) {
         key = ASN1_INTEGER_new();
-        if (key == NULL || !ASN1_INTEGER_set(key, keysize))
+        if (key == NULL || !ASN1_INTEGER_set(key, keysize)) {
+            ASN1_INTEGER_free(key);
             return 0;
+        }
     }
     alg = X509_ALGOR_new();
     if (alg == NULL) {
diff --git a/crypto/conf/conf_def.c b/crypto/conf/conf_def.c
index ca76fa3679b8..3d710f12ae07 100644
--- a/crypto/conf/conf_def.c
+++ b/crypto/conf/conf_def.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -376,11 +376,13 @@ static int def_load_bio(CONF *conf, BIO *in, long *line)
                     if (biosk == NULL) {
                         if ((biosk = sk_BIO_new_null()) == NULL) {
                             CONFerr(CONF_F_DEF_LOAD_BIO, ERR_R_MALLOC_FAILURE);
+                            BIO_free(next);
                             goto err;
                         }
                     }
                     if (!sk_BIO_push(biosk, in)) {
                         CONFerr(CONF_F_DEF_LOAD_BIO, ERR_R_MALLOC_FAILURE);
+                        BIO_free(next);
                         goto err;
                     }
                     /* continue with reading from the included BIO */
diff --git a/crypto/ec/asm/ecp_nistz256-armv4.pl b/crypto/ec/asm/ecp_nistz256-armv4.pl
index ea538c0698d5..fa833ce6aaf3 100755
--- a/crypto/ec/asm/ecp_nistz256-armv4.pl
+++ b/crypto/ec/asm/ecp_nistz256-armv4.pl
@@ -1517,9 +1517,9 @@ ecp_nistz256_point_add:
 	ldr	$t2,[sp,#32*18+12]	@ ~is_equal(S1,S2)
 	mvn	$t0,$t0			@ -1/0 -> 0/-1
 	mvn	$t1,$t1			@ -1/0 -> 0/-1
-	orr	$a0,$t0
-	orr	$a0,$t1
-	orrs	$a0,$t2			@ set flags
+	orr	$a0,$a0,$t0
+	orr	$a0,$a0,$t1
+	orrs	$a0,$a0,$t2		@ set flags
 
 	@ if(~is_equal(U1,U2) | in1infty | in2infty | ~is_equal(S1,S2))
 	bne	.Ladd_proceed
diff --git a/crypto/ec/asm/ecp_nistz256-avx2.pl b/crypto/ec/asm/ecp_nistz256-avx2.pl
deleted file mode 100755
index 5071d09ac2ec..000000000000
--- a/crypto/ec/asm/ecp_nistz256-avx2.pl
+++ /dev/null
@@ -1,2080 +0,0 @@
-#! /usr/bin/env perl
-# Copyright 2014-2020 The OpenSSL Project Authors. All Rights Reserved.
-# Copyright (c) 2014, Intel Corporation. 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
-#
-# Originally written by Shay Gueron (1, 2), and Vlad Krasnov (1)
-# (1) Intel Corporation, Israel Development Center, Haifa, Israel
-# (2) University of Haifa, Israel
-#
-# Reference:
-# S.Gueron and V.Krasnov, "Fast Prime Field Elliptic Curve Cryptography with
-#                          256 Bit Primes"
-
-$flavour = shift;
-$output  = shift;
-if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
-
-$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
-
-$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
-( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
-( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
-die "can't locate x86_64-xlate.pl";
-
-open OUT,"| \"$^X\" $xlate $flavour $output";
-*STDOUT=*OUT;
-
-if (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1`
-		=~ /GNU assembler version ([2-9]\.[0-9]+)/) {
-	$avx = ($1>=2.19) + ($1>=2.22);
-	$addx = ($1>=2.23);
-}
-
-if (!$addx && $win64 && ($flavour =~ /nasm/ || $ENV{ASM} =~ /nasm/) &&
-	    `nasm -v 2>&1` =~ /NASM version ([2-9]\.[0-9]+)/) {
-	$avx = ($1>=2.09) + ($1>=2.10);
-	$addx = ($1>=2.10);
-}
-
-if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
-	    `ml64 2>&1` =~ /Version ([0-9]+)\./) {
-	$avx = ($1>=10) + ($1>=11);
-	$addx = ($1>=12);
-}
-
-if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|based on LLVM) ([0-9]+)\.([0-9]+)/) {
-	my $ver = $2 + $3/100.0;	# 3.1->3.01, 3.10->3.10
-	$avx = ($ver>=3.0) + ($ver>=3.01);
-	$addx = ($ver>=3.03);
-}
-
-if ($avx>=2) {{
-$digit_size = "\$29";
-$n_digits = "\$9";
-
-$code.=<<___;
-.text
-
-.align 64
-.LAVX2_AND_MASK:
-.LAVX2_POLY:
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x000001ff, 0x000001ff, 0x000001ff, 0x000001ff
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x00040000, 0x00040000, 0x00040000, 0x00040000
-.quad 0x1fe00000, 0x1fe00000, 0x1fe00000, 0x1fe00000
-.quad 0x00ffffff, 0x00ffffff, 0x00ffffff, 0x00ffffff
-
-.LAVX2_POLY_x2:
-.quad 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC
-.quad 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC
-.quad 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC
-.quad 0x400007FC, 0x400007FC, 0x400007FC, 0x400007FC
-.quad 0x3FFFFFFE, 0x3FFFFFFE, 0x3FFFFFFE, 0x3FFFFFFE
-.quad 0x3FFFFFFE, 0x3FFFFFFE, 0x3FFFFFFE, 0x3FFFFFFE
-.quad 0x400FFFFE, 0x400FFFFE, 0x400FFFFE, 0x400FFFFE
-.quad 0x7F7FFFFE, 0x7F7FFFFE, 0x7F7FFFFE, 0x7F7FFFFE
-.quad 0x03FFFFFC, 0x03FFFFFC, 0x03FFFFFC, 0x03FFFFFC
-
-.LAVX2_POLY_x8:
-.quad 0xFFFFFFF8, 0xFFFFFFF8, 0xFFFFFFF8, 0xFFFFFFF8
-.quad 0xFFFFFFF8, 0xFFFFFFF8, 0xFFFFFFF8, 0xFFFFFFF8
-.quad 0xFFFFFFF8, 0xFFFFFFF8, 0xFFFFFFF8, 0xFFFFFFF8
-.quad 0x80000FF8, 0x80000FF8, 0x80000FF8, 0x80000FF8
-.quad 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC
-.quad 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC, 0x7FFFFFFC
-.quad 0x801FFFFC, 0x801FFFFC, 0x801FFFFC, 0x801FFFFC
-.quad 0xFEFFFFFC, 0xFEFFFFFC, 0xFEFFFFFC, 0xFEFFFFFC
-.quad 0x07FFFFF8, 0x07FFFFF8, 0x07FFFFF8, 0x07FFFFF8
-
-.LONE:
-.quad 0x00000020, 0x00000020, 0x00000020, 0x00000020
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x1fffc000, 0x1fffc000, 0x1fffc000, 0x1fffc000
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x1f7fffff, 0x1f7fffff, 0x1f7fffff, 0x1f7fffff
-.quad 0x03ffffff, 0x03ffffff, 0x03ffffff, 0x03ffffff
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-
-# RR = 2^266 mod p in AVX2 format, to transform from the native OpenSSL
-# Montgomery form (*2^256) to our format (*2^261)
-
-.LTO_MONT_AVX2:
-.quad 0x00000400, 0x00000400, 0x00000400, 0x00000400
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x1ff80000, 0x1ff80000, 0x1ff80000, 0x1ff80000
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x0fffffff, 0x0fffffff, 0x0fffffff, 0x0fffffff
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x00000003, 0x00000003, 0x00000003, 0x00000003
-
-.LFROM_MONT_AVX2:
-.quad 0x00000001, 0x00000001, 0x00000001, 0x00000001
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-.quad 0x1ffffe00, 0x1ffffe00, 0x1ffffe00, 0x1ffffe00
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x1fffffff, 0x1fffffff, 0x1fffffff, 0x1fffffff
-.quad 0x1ffbffff, 0x1ffbffff, 0x1ffbffff, 0x1ffbffff
-.quad 0x001fffff, 0x001fffff, 0x001fffff, 0x001fffff
-.quad 0x00000000, 0x00000000, 0x00000000, 0x00000000
-
-.LIntOne:
-.long 1,1,1,1,1,1,1,1
-___
-
-{
-# This function receives a pointer to an array of four affine points
-# (X, Y, <1>) and rearranges the data for AVX2 execution, while
-# converting it to 2^29 radix redundant form
-
-my ($X0,$X1,$X2,$X3, $Y0,$Y1,$Y2,$Y3,
-    $T0,$T1,$T2,$T3, $T4,$T5,$T6,$T7)=map("%ymm$_",(0..15));
-
-$code.=<<___;
-.globl	ecp_nistz256_avx2_transpose_convert
-.type	ecp_nistz256_avx2_transpose_convert,\@function,2
-.align 64
-ecp_nistz256_avx2_transpose_convert:
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	lea	-8-16*10(%rsp), %rsp
-	vmovaps	%xmm6, -8-16*10(%rax)
-	vmovaps	%xmm7, -8-16*9(%rax)
-	vmovaps	%xmm8, -8-16*8(%rax)
-	vmovaps	%xmm9, -8-16*7(%rax)
-	vmovaps	%xmm10, -8-16*6(%rax)
-	vmovaps	%xmm11, -8-16*5(%rax)
-	vmovaps	%xmm12, -8-16*4(%rax)
-	vmovaps	%xmm13, -8-16*3(%rax)
-	vmovaps	%xmm14, -8-16*2(%rax)
-	vmovaps	%xmm15, -8-16*1(%rax)
-___
-$code.=<<___;
-	# Load the data
-	vmovdqa		32*0(%rsi), $X0
-	lea		112(%rsi), %rax		# size optimization
-	vmovdqa		32*1(%rsi), $Y0
-	lea		.LAVX2_AND_MASK(%rip), %rdx
-	vmovdqa		32*2(%rsi), $X1
-	vmovdqa		32*3(%rsi), $Y1
-	vmovdqa		32*4-112(%rax), $X2
-	vmovdqa		32*5-112(%rax), $Y2
-	vmovdqa		32*6-112(%rax), $X3
-	vmovdqa		32*7-112(%rax), $Y3
-
-	# Transpose X and Y independently
-	vpunpcklqdq	$X1, $X0, $T0		# T0 = [B2 A2 B0 A0]
-	vpunpcklqdq	$X3, $X2, $T1		# T1 = [D2 C2 D0 C0]
-	vpunpckhqdq	$X1, $X0, $T2		# T2 = [B3 A3 B1 A1]
-	vpunpckhqdq	$X3, $X2, $T3		# T3 = [D3 C3 D1 C1]
-
-	vpunpcklqdq	$Y1, $Y0, $T4
-	vpunpcklqdq	$Y3, $Y2, $T5
-	vpunpckhqdq	$Y1, $Y0, $T6
-	vpunpckhqdq	$Y3, $Y2, $T7
-
-	vperm2i128	\$0x20, $T1, $T0, $X0	# X0 = [D0 C0 B0 A0]
-	vperm2i128	\$0x20, $T3, $T2, $X1	# X1 = [D1 C1 B1 A1]
-	vperm2i128	\$0x31, $T1, $T0, $X2	# X2 = [D2 C2 B2 A2]
-	vperm2i128	\$0x31, $T3, $T2, $X3	# X3 = [D3 C3 B3 A3]
-
-	vperm2i128	\$0x20, $T5, $T4, $Y0
-	vperm2i128	\$0x20, $T7, $T6, $Y1
-	vperm2i128	\$0x31, $T5, $T4, $Y2
-	vperm2i128	\$0x31, $T7, $T6, $Y3
-	vmovdqa		(%rdx), $T7
-
-	vpand		(%rdx), $X0, $T0	# out[0] = in[0] & mask;
-	vpsrlq		\$29, $X0, $X0
-	vpand		$T7, $X0, $T1		# out[1] = (in[0] >> shift) & mask;
-	vpsrlq		\$29, $X0, $X0
-	vpsllq		\$6, $X1, $T2
-	vpxor		$X0, $T2, $T2
-	vpand		$T7, $T2, $T2		# out[2] = ((in[0] >> (shift*2)) ^ (in[1] << (64-shift*2))) & mask;
-	vpsrlq		\$23, $X1, $X1
-	vpand		$T7, $X1, $T3		# out[3] = (in[1] >> ((shift*3)%64)) & mask;
-	vpsrlq		\$29, $X1, $X1
-	vpsllq		\$12, $X2, $T4
-	vpxor		$X1, $T4, $T4
-	vpand		$T7, $T4, $T4		# out[4] = ((in[1] >> ((shift*4)%64)) ^ (in[2] << (64*2-shift*4))) & mask;
-	vpsrlq		\$17, $X2, $X2
-	vpand		$T7, $X2, $T5		# out[5] = (in[2] >> ((shift*5)%64)) & mask;
-	vpsrlq		\$29, $X2, $X2
-	vpsllq		\$18, $X3, $T6
-	vpxor		$X2, $T6, $T6
-	vpand		$T7, $T6, $T6		# out[6] = ((in[2] >> ((shift*6)%64)) ^ (in[3] << (64*3-shift*6))) & mask;
-	vpsrlq		\$11, $X3, $X3
-	 vmovdqa	$T0, 32*0(%rdi)
-	 lea		112(%rdi), %rax		# size optimization
-	vpand		$T7, $X3, $T0		# out[7] = (in[3] >> ((shift*7)%64)) & mask;
-	vpsrlq		\$29, $X3, $X3		# out[8] = (in[3] >> ((shift*8)%64)) & mask;
-
-	vmovdqa		$T1, 32*1(%rdi)
-	vmovdqa		$T2, 32*2(%rdi)
-	vmovdqa		$T3, 32*3(%rdi)
-	vmovdqa		$T4, 32*4-112(%rax)
-	vmovdqa		$T5, 32*5-112(%rax)
-	vmovdqa		$T6, 32*6-112(%rax)
-	vmovdqa		$T0, 32*7-112(%rax)
-	vmovdqa		$X3, 32*8-112(%rax)
-	lea		448(%rdi), %rax		# size optimization
-
-	vpand		$T7, $Y0, $T0		# out[0] = in[0] & mask;
-	vpsrlq		\$29, $Y0, $Y0
-	vpand		$T7, $Y0, $T1		# out[1] = (in[0] >> shift) & mask;
-	vpsrlq		\$29, $Y0, $Y0
-	vpsllq		\$6, $Y1, $T2
-	vpxor		$Y0, $T2, $T2
-	vpand		$T7, $T2, $T2		# out[2] = ((in[0] >> (shift*2)) ^ (in[1] << (64-shift*2))) & mask;
-	vpsrlq		\$23, $Y1, $Y1
-	vpand		$T7, $Y1, $T3		# out[3] = (in[1] >> ((shift*3)%64)) & mask;
-	vpsrlq		\$29, $Y1, $Y1
-	vpsllq		\$12, $Y2, $T4
-	vpxor		$Y1, $T4, $T4
-	vpand		$T7, $T4, $T4		# out[4] = ((in[1] >> ((shift*4)%64)) ^ (in[2] << (64*2-shift*4))) & mask;
-	vpsrlq		\$17, $Y2, $Y2
-	vpand		$T7, $Y2, $T5		# out[5] = (in[2] >> ((shift*5)%64)) & mask;
-	vpsrlq		\$29, $Y2, $Y2
-	vpsllq		\$18, $Y3, $T6
-	vpxor		$Y2, $T6, $T6
-	vpand		$T7, $T6, $T6		# out[6] = ((in[2] >> ((shift*6)%64)) ^ (in[3] << (64*3-shift*6))) & mask;
-	vpsrlq		\$11, $Y3, $Y3
-	 vmovdqa	$T0, 32*9-448(%rax)
-	vpand		$T7, $Y3, $T0		# out[7] = (in[3] >> ((shift*7)%64)) & mask;
-	vpsrlq		\$29, $Y3, $Y3		# out[8] = (in[3] >> ((shift*8)%64)) & mask;
-
-	vmovdqa		$T1, 32*10-448(%rax)
-	vmovdqa		$T2, 32*11-448(%rax)
-	vmovdqa		$T3, 32*12-448(%rax)
-	vmovdqa		$T4, 32*13-448(%rax)
-	vmovdqa		$T5, 32*14-448(%rax)
-	vmovdqa		$T6, 32*15-448(%rax)
-	vmovdqa		$T0, 32*16-448(%rax)
-	vmovdqa		$Y3, 32*17-448(%rax)
-
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	movaps	16*0(%rsp), %xmm6
-	movaps	16*1(%rsp), %xmm7
-	movaps	16*2(%rsp), %xmm8
-	movaps	16*3(%rsp), %xmm9
-	movaps	16*4(%rsp), %xmm10
-	movaps	16*5(%rsp), %xmm11
-	movaps	16*6(%rsp), %xmm12
-	movaps	16*7(%rsp), %xmm13
-	movaps	16*8(%rsp), %xmm14
-	movaps	16*9(%rsp), %xmm15
-	lea	8+16*10(%rsp), %rsp
-___
-$code.=<<___;
-	ret
-.size	ecp_nistz256_avx2_transpose_convert,.-ecp_nistz256_avx2_transpose_convert
-___
-}
-{
-################################################################################
-# This function receives a pointer to an array of four AVX2 formatted points
-# (X, Y, Z) convert the data to normal representation, and rearranges the data
-
-my ($D0,$D1,$D2,$D3, $D4,$D5,$D6,$D7, $D8)=map("%ymm$_",(0..8));
-my ($T0,$T1,$T2,$T3, $T4,$T5,$T6)=map("%ymm$_",(9..15));
-
-$code.=<<___;
-
-.globl	ecp_nistz256_avx2_convert_transpose_back
-.type	ecp_nistz256_avx2_convert_transpose_back,\@function,2
-.align	32
-ecp_nistz256_avx2_convert_transpose_back:
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	lea	-8-16*10(%rsp), %rsp
-	vmovaps	%xmm6, -8-16*10(%rax)
-	vmovaps	%xmm7, -8-16*9(%rax)
-	vmovaps	%xmm8, -8-16*8(%rax)
-	vmovaps	%xmm9, -8-16*7(%rax)
-	vmovaps	%xmm10, -8-16*6(%rax)
-	vmovaps	%xmm11, -8-16*5(%rax)
-	vmovaps	%xmm12, -8-16*4(%rax)
-	vmovaps	%xmm13, -8-16*3(%rax)
-	vmovaps	%xmm14, -8-16*2(%rax)
-	vmovaps	%xmm15, -8-16*1(%rax)
-___
-$code.=<<___;
-	mov	\$3, %ecx
-
-.Lconv_loop:
-	vmovdqa		32*0(%rsi), $D0
-	lea		160(%rsi), %rax		# size optimization
-	vmovdqa		32*1(%rsi), $D1
-	vmovdqa		32*2(%rsi), $D2
-	vmovdqa		32*3(%rsi), $D3
-	vmovdqa		32*4-160(%rax), $D4
-	vmovdqa		32*5-160(%rax), $D5
-	vmovdqa		32*6-160(%rax), $D6
-	vmovdqa		32*7-160(%rax), $D7
-	vmovdqa		32*8-160(%rax), $D8
-
-	vpsllq		\$29, $D1, $D1
-	vpsllq		\$58, $D2, $T0
-	vpaddq		$D1, $D0, $D0
-	vpaddq		$T0, $D0, $D0		# out[0] = (in[0]) ^ (in[1] << shift*1) ^ (in[2] << shift*2);
-
-	vpsrlq		\$6, $D2, $D2
-	vpsllq		\$23, $D3, $D3
-	vpsllq		\$52, $D4, $T1
-	vpaddq		$D2, $D3, $D3
-	vpaddq		$D3, $T1, $D1		# out[1] = (in[2] >> (64*1-shift*2)) ^ (in[3] << shift*3%64) ^ (in[4] << shift*4%64);
-
-	vpsrlq		\$12, $D4, $D4
-	vpsllq		\$17, $D5, $D5
-	vpsllq		\$46, $D6, $T2
-	vpaddq		$D4, $D5, $D5
-	vpaddq		$D5, $T2, $D2		# out[2] = (in[4] >> (64*2-shift*4)) ^ (in[5] << shift*5%64) ^ (in[6] << shift*6%64);
-
-	vpsrlq		\$18, $D6, $D6
-	vpsllq		\$11, $D7, $D7
-	vpsllq		\$40, $D8, $T3
-	vpaddq		$D6, $D7, $D7
-	vpaddq		$D7, $T3, $D3		# out[3] = (in[6] >> (64*3-shift*6)) ^ (in[7] << shift*7%64) ^ (in[8] << shift*8%64);
-
-	vpunpcklqdq	$D1, $D0, $T0		# T0 = [B2 A2 B0 A0]
-	vpunpcklqdq	$D3, $D2, $T1		# T1 = [D2 C2 D0 C0]
-	vpunpckhqdq	$D1, $D0, $T2		# T2 = [B3 A3 B1 A1]
-	vpunpckhqdq	$D3, $D2, $T3		# T3 = [D3 C3 D1 C1]
-
-	vperm2i128	\$0x20, $T1, $T0, $D0	# X0 = [D0 C0 B0 A0]
-	vperm2i128	\$0x20, $T3, $T2, $D1	# X1 = [D1 C1 B1 A1]
-	vperm2i128	\$0x31, $T1, $T0, $D2	# X2 = [D2 C2 B2 A2]
-	vperm2i128	\$0x31, $T3, $T2, $D3	# X3 = [D3 C3 B3 A3]
-
-	vmovdqa		$D0, 32*0(%rdi)
-	vmovdqa		$D1, 32*3(%rdi)
-	vmovdqa		$D2, 32*6(%rdi)
-	vmovdqa		$D3, 32*9(%rdi)
-
-	lea		32*9(%rsi), %rsi
-	lea		32*1(%rdi), %rdi
-
-	dec	%ecx
-	jnz	.Lconv_loop
-
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	movaps	16*0(%rsp), %xmm6
-	movaps	16*1(%rsp), %xmm7
-	movaps	16*2(%rsp), %xmm8
-	movaps	16*3(%rsp), %xmm9
-	movaps	16*4(%rsp), %xmm10
-	movaps	16*5(%rsp), %xmm11
-	movaps	16*6(%rsp), %xmm12
-	movaps	16*7(%rsp), %xmm13
-	movaps	16*8(%rsp), %xmm14
-	movaps	16*9(%rsp), %xmm15
-	lea	8+16*10(%rsp), %rsp
-___
-$code.=<<___;
-	ret
-.size	ecp_nistz256_avx2_convert_transpose_back,.-ecp_nistz256_avx2_convert_transpose_back
-___
-}
-{
-my ($r_ptr,$a_ptr,$b_ptr,$itr)=("%rdi","%rsi","%rdx","%ecx");
-my ($ACC0,$ACC1,$ACC2,$ACC3,$ACC4,$ACC5,$ACC6,$ACC7,$ACC8)=map("%ymm$_",(0..8));
-my ($B,$Y,$T0,$AND_MASK,$OVERFLOW)=map("%ymm$_",(9..13));
-
-sub NORMALIZE {
-my $ret=<<___;
-	vpsrlq		$digit_size, $ACC0, $T0
-	vpand		$AND_MASK, $ACC0, $ACC0
-	vpaddq		$T0, $ACC1, $ACC1
-
-	vpsrlq		$digit_size, $ACC1, $T0
-	vpand		$AND_MASK, $ACC1, $ACC1
-	vpaddq		$T0, $ACC2, $ACC2
-
-	vpsrlq		$digit_size, $ACC2, $T0
-	vpand		$AND_MASK, $ACC2, $ACC2
-	vpaddq		$T0, $ACC3, $ACC3
-
-	vpsrlq		$digit_size, $ACC3, $T0
-	vpand		$AND_MASK, $ACC3, $ACC3
-	vpaddq		$T0, $ACC4, $ACC4
-
-	vpsrlq		$digit_size, $ACC4, $T0
-	vpand		$AND_MASK, $ACC4, $ACC4
-	vpaddq		$T0, $ACC5, $ACC5
-
-	vpsrlq		$digit_size, $ACC5, $T0
-	vpand		$AND_MASK, $ACC5, $ACC5
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpsrlq		$digit_size, $ACC6, $T0
-	vpand		$AND_MASK, $ACC6, $ACC6
-	vpaddq		$T0, $ACC7, $ACC7
-
-	vpsrlq		$digit_size, $ACC7, $T0
-	vpand		$AND_MASK, $ACC7, $ACC7
-	vpaddq		$T0, $ACC8, $ACC8
-	#vpand		$AND_MASK, $ACC8, $ACC8
-___
-    $ret;
-}
-
-sub STORE {
-my $ret=<<___;
-	vmovdqa		$ACC0, 32*0(%rdi)
-	lea		160(%rdi), %rax		# size optimization
-	vmovdqa		$ACC1, 32*1(%rdi)
-	vmovdqa		$ACC2, 32*2(%rdi)
-	vmovdqa		$ACC3, 32*3(%rdi)
-	vmovdqa		$ACC4, 32*4-160(%rax)
-	vmovdqa		$ACC5, 32*5-160(%rax)
-	vmovdqa		$ACC6, 32*6-160(%rax)
-	vmovdqa		$ACC7, 32*7-160(%rax)
-	vmovdqa		$ACC8, 32*8-160(%rax)
-___
-    $ret;
-}
-
-$code.=<<___;
-.type	avx2_normalize,\@abi-omnipotent
-.align	32
-avx2_normalize:
-	vpsrlq		$digit_size, $ACC0, $T0
-	vpand		$AND_MASK, $ACC0, $ACC0
-	vpaddq		$T0, $ACC1, $ACC1
-
-	vpsrlq		$digit_size, $ACC1, $T0
-	vpand		$AND_MASK, $ACC1, $ACC1
-	vpaddq		$T0, $ACC2, $ACC2
-
-	vpsrlq		$digit_size, $ACC2, $T0
-	vpand		$AND_MASK, $ACC2, $ACC2
-	vpaddq		$T0, $ACC3, $ACC3
-
-	vpsrlq		$digit_size, $ACC3, $T0
-	vpand		$AND_MASK, $ACC3, $ACC3
-	vpaddq		$T0, $ACC4, $ACC4
-
-	vpsrlq		$digit_size, $ACC4, $T0
-	vpand		$AND_MASK, $ACC4, $ACC4
-	vpaddq		$T0, $ACC5, $ACC5
-
-	vpsrlq		$digit_size, $ACC5, $T0
-	vpand		$AND_MASK, $ACC5, $ACC5
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpsrlq		$digit_size, $ACC6, $T0
-	vpand		$AND_MASK, $ACC6, $ACC6
-	vpaddq		$T0, $ACC7, $ACC7
-
-	vpsrlq		$digit_size, $ACC7, $T0
-	vpand		$AND_MASK, $ACC7, $ACC7
-	vpaddq		$T0, $ACC8, $ACC8
-	#vpand		$AND_MASK, $ACC8, $ACC8
-
-	ret
-.size	avx2_normalize,.-avx2_normalize
-
-.type	avx2_normalize_n_store,\@abi-omnipotent
-.align	32
-avx2_normalize_n_store:
-	vpsrlq		$digit_size, $ACC0, $T0
-	vpand		$AND_MASK, $ACC0, $ACC0
-	vpaddq		$T0, $ACC1, $ACC1
-
-	vpsrlq		$digit_size, $ACC1, $T0
-	vpand		$AND_MASK, $ACC1, $ACC1
-	 vmovdqa	$ACC0, 32*0(%rdi)
-	 lea		160(%rdi), %rax		# size optimization
-	vpaddq		$T0, $ACC2, $ACC2
-
-	vpsrlq		$digit_size, $ACC2, $T0
-	vpand		$AND_MASK, $ACC2, $ACC2
-	 vmovdqa	$ACC1, 32*1(%rdi)
-	vpaddq		$T0, $ACC3, $ACC3
-
-	vpsrlq		$digit_size, $ACC3, $T0
-	vpand		$AND_MASK, $ACC3, $ACC3
-	 vmovdqa	$ACC2, 32*2(%rdi)
-	vpaddq		$T0, $ACC4, $ACC4
-
-	vpsrlq		$digit_size, $ACC4, $T0
-	vpand		$AND_MASK, $ACC4, $ACC4
-	 vmovdqa	$ACC3, 32*3(%rdi)
-	vpaddq		$T0, $ACC5, $ACC5
-
-	vpsrlq		$digit_size, $ACC5, $T0
-	vpand		$AND_MASK, $ACC5, $ACC5
-	 vmovdqa	$ACC4, 32*4-160(%rax)
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpsrlq		$digit_size, $ACC6, $T0
-	vpand		$AND_MASK, $ACC6, $ACC6
-	 vmovdqa	$ACC5, 32*5-160(%rax)
-	vpaddq		$T0, $ACC7, $ACC7
-
-	vpsrlq		$digit_size, $ACC7, $T0
-	vpand		$AND_MASK, $ACC7, $ACC7
-	 vmovdqa	$ACC6, 32*6-160(%rax)
-	vpaddq		$T0, $ACC8, $ACC8
-	#vpand		$AND_MASK, $ACC8, $ACC8
-	 vmovdqa	$ACC7, 32*7-160(%rax)
-	 vmovdqa	$ACC8, 32*8-160(%rax)
-
-	ret
-.size	avx2_normalize_n_store,.-avx2_normalize_n_store
-
-################################################################################
-# void avx2_mul_x4(void* RESULTx4, void *Ax4, void *Bx4);
-.type	avx2_mul_x4,\@abi-omnipotent
-.align	32
-avx2_mul_x4:
-	lea	.LAVX2_POLY(%rip), %rax
-
-	vpxor	$ACC0, $ACC0, $ACC0
-	vpxor	$ACC1, $ACC1, $ACC1
-	vpxor	$ACC2, $ACC2, $ACC2
-	vpxor	$ACC3, $ACC3, $ACC3
-	vpxor	$ACC4, $ACC4, $ACC4
-	vpxor	$ACC5, $ACC5, $ACC5
-	vpxor	$ACC6, $ACC6, $ACC6
-	vpxor	$ACC7, $ACC7, $ACC7
-
-	vmovdqa	32*7(%rax), %ymm14
-	vmovdqa	32*8(%rax), %ymm15
-
-	mov	$n_digits, $itr
-	lea	-512($a_ptr), $a_ptr	# strategic bias to control u-op density
-	jmp	.Lavx2_mul_x4_loop
-
-.align	32
-.Lavx2_mul_x4_loop:
-	vmovdqa		32*0($b_ptr), $B
-	lea		32*1($b_ptr), $b_ptr
-
-	vpmuludq	32*0+512($a_ptr), $B, $T0
-	vpmuludq	32*1+512($a_ptr), $B, $OVERFLOW	# borrow $OVERFLOW
-	vpaddq		$T0, $ACC0, $ACC0
-	vpmuludq	32*2+512($a_ptr), $B, $T0
-	vpaddq		$OVERFLOW, $ACC1, $ACC1
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpmuludq	32*3+512($a_ptr), $B, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC2
-	vpmuludq	32*4+512($a_ptr), $B, $T0
-	vpaddq		$OVERFLOW, $ACC3, $ACC3
-	vpmuludq	32*5+512($a_ptr), $B, $OVERFLOW
-	vpaddq		$T0, $ACC4, $ACC4
-	vpmuludq	32*6+512($a_ptr), $B, $T0
-	vpaddq		$OVERFLOW, $ACC5, $ACC5
-	vpmuludq	32*7+512($a_ptr), $B, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC6
-
-	# Skip some multiplications, optimizing for the constant poly
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*8+512($a_ptr), $B, $ACC8
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	.byte		0x67
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $OVERFLOW
-	.byte		0x67
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $T0
-	vpaddq		$OVERFLOW, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC7, $ACC6
-	vpaddq		$OVERFLOW, $ACC8, $ACC7
-
-	dec	$itr
-	jnz	.Lavx2_mul_x4_loop
-
-	vpxor	$ACC8, $ACC8, $ACC8
-
-	ret
-.size	avx2_mul_x4,.-avx2_mul_x4
-
-# Function optimized for the constant 1
-################################################################################
-# void avx2_mul_by1_x4(void* RESULTx4, void *Ax4);
-.type	avx2_mul_by1_x4,\@abi-omnipotent
-.align	32
-avx2_mul_by1_x4:
-	lea	.LAVX2_POLY(%rip), %rax
-
-	vpxor	$ACC0, $ACC0, $ACC0
-	vpxor	$ACC1, $ACC1, $ACC1
-	vpxor	$ACC2, $ACC2, $ACC2
-	vpxor	$ACC3, $ACC3, $ACC3
-	vpxor	$ACC4, $ACC4, $ACC4
-	vpxor	$ACC5, $ACC5, $ACC5
-	vpxor	$ACC6, $ACC6, $ACC6
-	vpxor	$ACC7, $ACC7, $ACC7
-	vpxor	$ACC8, $ACC8, $ACC8
-
-	vmovdqa	32*3+.LONE(%rip), %ymm14
-	vmovdqa	32*7+.LONE(%rip), %ymm15
-
-	mov	$n_digits, $itr
-	jmp	.Lavx2_mul_by1_x4_loop
-
-.align	32
-.Lavx2_mul_by1_x4_loop:
-	vmovdqa		32*0($a_ptr), $B
-	.byte		0x48,0x8d,0xb6,0x20,0,0,0	# lea	32*1($a_ptr), $a_ptr
-
-	vpsllq		\$5, $B, $OVERFLOW
-	vpmuludq	%ymm14, $B, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC3
-	.byte		0x67
-	vpmuludq	$AND_MASK, $B, $T0
-	vpand		$AND_MASK, $ACC0, $Y
-	vpaddq		$T0, $ACC4, $ACC4
-	vpaddq		$T0, $ACC5, $ACC5
-	vpaddq		$T0, $ACC6, $ACC6
-	vpsllq		\$23, $B, $T0
-
-	.byte		0x67,0x67
-	vpmuludq	%ymm15, $B, $OVERFLOW
-	vpsubq		$T0, $ACC6, $ACC6
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	vpaddq		$OVERFLOW, $ACC7, $ACC7
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	.byte		0x67,0x67
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $OVERFLOW
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	32*7(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC6, $ACC5
-	vpaddq		$T0, $ACC7, $ACC6
-	vpmuludq	32*8(%rax), $Y, $ACC7
-
-	dec	$itr
-	jnz	.Lavx2_mul_by1_x4_loop
-
-	ret
-.size	avx2_mul_by1_x4,.-avx2_mul_by1_x4
-
-################################################################################
-# void avx2_sqr_x4(void* RESULTx4, void *Ax4, void *Bx4);
-.type	avx2_sqr_x4,\@abi-omnipotent
-.align	32
-avx2_sqr_x4:
-	lea		.LAVX2_POLY(%rip), %rax
-
-	vmovdqa		32*7(%rax), %ymm14
-	vmovdqa		32*8(%rax), %ymm15
-
-	vmovdqa		32*0($a_ptr), $B
-	vmovdqa		32*1($a_ptr), $ACC1
-	vmovdqa		32*2($a_ptr), $ACC2
-	vmovdqa		32*3($a_ptr), $ACC3
-	vmovdqa		32*4($a_ptr), $ACC4
-	vmovdqa		32*5($a_ptr), $ACC5
-	vmovdqa		32*6($a_ptr), $ACC6
-	vmovdqa		32*7($a_ptr), $ACC7
-	vpaddq		$ACC1, $ACC1, $ACC1	# 2*$ACC0..7
-	vmovdqa		32*8($a_ptr), $ACC8
-	vpaddq		$ACC2, $ACC2, $ACC2
-	vmovdqa		$ACC1, 32*0(%rcx)
-	vpaddq		$ACC3, $ACC3, $ACC3
-	vmovdqa		$ACC2, 32*1(%rcx)
-	vpaddq		$ACC4, $ACC4, $ACC4
-	vmovdqa		$ACC3, 32*2(%rcx)
-	vpaddq		$ACC5, $ACC5, $ACC5
-	vmovdqa		$ACC4, 32*3(%rcx)
-	vpaddq		$ACC6, $ACC6, $ACC6
-	vmovdqa		$ACC5, 32*4(%rcx)
-	vpaddq		$ACC7, $ACC7, $ACC7
-	vmovdqa		$ACC6, 32*5(%rcx)
-	vpaddq		$ACC8, $ACC8, $ACC8
-	vmovdqa		$ACC7, 32*6(%rcx)
-	vmovdqa		$ACC8, 32*7(%rcx)
-
-	#itr		1
-	vpmuludq	$B, $B, $ACC0
-	vpmuludq	$B, $ACC1, $ACC1
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpmuludq	$B, $ACC2, $ACC2
-	vpmuludq	$B, $ACC3, $ACC3
-	vpmuludq	$B, $ACC4, $ACC4
-	vpmuludq	$B, $ACC5, $ACC5
-	vpmuludq	$B, $ACC6, $ACC6
-	 vpmuludq	$AND_MASK, $Y, $T0
-	vpmuludq	$B, $ACC7, $ACC7
-	vpmuludq	$B, $ACC8, $ACC8
-	 vmovdqa	32*1($a_ptr), $B
-
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		2
-	vpmuludq	$B, $B, $OVERFLOW
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpmuludq	32*1(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC1, $ACC1
-	vpmuludq	32*2(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC2
-	vpmuludq	32*3(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC3, $ACC3
-	vpmuludq	32*4(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC4, $ACC4
-	vpmuludq	32*5(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC5, $ACC5
-	vpmuludq	32*6(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*7(%rcx), $B, $ACC8
-	 vmovdqa	32*2($a_ptr), $B
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		3
-	vpmuludq	$B, $B, $T0
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpmuludq	32*2(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC2
-	vpmuludq	32*3(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC3, $ACC3
-	vpmuludq	32*4(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC4, $ACC4
-	vpmuludq	32*5(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC5, $ACC5
-	vpmuludq	32*6(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*7(%rcx), $B, $ACC8
-	 vmovdqa	32*3($a_ptr), $B
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		4
-	vpmuludq	$B, $B, $OVERFLOW
-	vpmuludq	32*3(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC3, $ACC3
-	vpmuludq	32*4(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC4, $ACC4
-	vpmuludq	32*5(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC5, $ACC5
-	vpmuludq	32*6(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*7(%rcx), $B, $ACC8
-	 vmovdqa	32*4($a_ptr), $B
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		5
-	vpmuludq	$B, $B, $T0
-	vpmuludq	32*4(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC4, $ACC4
-	vpmuludq	32*5(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC5, $ACC5
-	vpmuludq	32*6(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*7(%rcx), $B, $ACC8
-	 vmovdqa	32*5($a_ptr), $B
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3+.LAVX2_POLY(%rip), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		6
-	vpmuludq	$B, $B, $OVERFLOW
-	vpmuludq	32*5(%rcx), $B, $T0
-	vpaddq		$OVERFLOW, $ACC5, $ACC5
-	vpmuludq	32*6(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*7(%rcx), $B, $ACC8
-	 vmovdqa	32*6($a_ptr), $B
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		7
-	vpmuludq	$B, $B, $T0
-	vpmuludq	32*6(%rcx), $B, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC6
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*7(%rcx), $B, $ACC8
-	 vmovdqa	32*7($a_ptr), $B
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		8
-	vpmuludq	$B, $B, $OVERFLOW
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	 vpaddq		$OVERFLOW, $ACC7, $ACC7
-	 vpmuludq	32*7(%rcx), $B, $ACC8
-	 vmovdqa	32*8($a_ptr), $B
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	 vpand		$AND_MASK, $ACC0, $Y
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	#itr		9
-	vpmuludq	$B, $B, $ACC8
-
-	vpmuludq	$AND_MASK, $Y, $T0
-	vpaddq		$T0, $ACC0, $OVERFLOW
-	vpsrlq		$digit_size, $OVERFLOW, $OVERFLOW
-	vpaddq		$T0, $ACC1, $ACC0
-	vpaddq		$T0, $ACC2, $ACC1
-	vpmuludq	32*3(%rax), $Y, $T0
-	vpaddq		$OVERFLOW, $ACC0, $ACC0
-	vpaddq		$T0, $ACC3, $ACC2
-	vmovdqa		$ACC4, $ACC3
-	vpsllq		\$18, $Y, $T0
-	vmovdqa		$ACC5, $ACC4
-	vpmuludq	%ymm14, $Y, $OVERFLOW
-	vpaddq		$T0, $ACC6, $ACC5
-	vpmuludq	%ymm15, $Y, $T0
-	vpaddq		$OVERFLOW, $ACC7, $ACC6
-	vpaddq		$T0, $ACC8, $ACC7
-
-	vpxor		$ACC8, $ACC8, $ACC8
-
-	ret
-.size	avx2_sqr_x4,.-avx2_sqr_x4
-
-################################################################################
-# void avx2_sub_x4(void* RESULTx4, void *Ax4, void *Bx4);
-.type	avx2_sub_x4,\@abi-omnipotent
-.align	32
-avx2_sub_x4:
-	vmovdqa	32*0($a_ptr), $ACC0
-	lea	160($a_ptr), $a_ptr
-	lea	.LAVX2_POLY_x8+128(%rip), %rax
-	lea	128($b_ptr), $b_ptr
-	vmovdqa	32*1-160($a_ptr), $ACC1
-	vmovdqa	32*2-160($a_ptr), $ACC2
-	vmovdqa	32*3-160($a_ptr), $ACC3
-	vmovdqa	32*4-160($a_ptr), $ACC4
-	vmovdqa	32*5-160($a_ptr), $ACC5
-	vmovdqa	32*6-160($a_ptr), $ACC6
-	vmovdqa	32*7-160($a_ptr), $ACC7
-	vmovdqa	32*8-160($a_ptr), $ACC8
-
-	vpaddq	32*0-128(%rax), $ACC0, $ACC0
-	vpaddq	32*1-128(%rax), $ACC1, $ACC1
-	vpaddq	32*2-128(%rax), $ACC2, $ACC2
-	vpaddq	32*3-128(%rax), $ACC3, $ACC3
-	vpaddq	32*4-128(%rax), $ACC4, $ACC4
-	vpaddq	32*5-128(%rax), $ACC5, $ACC5
-	vpaddq	32*6-128(%rax), $ACC6, $ACC6
-	vpaddq	32*7-128(%rax), $ACC7, $ACC7
-	vpaddq	32*8-128(%rax), $ACC8, $ACC8
-
-	vpsubq	32*0-128($b_ptr), $ACC0, $ACC0
-	vpsubq	32*1-128($b_ptr), $ACC1, $ACC1
-	vpsubq	32*2-128($b_ptr), $ACC2, $ACC2
-	vpsubq	32*3-128($b_ptr), $ACC3, $ACC3
-	vpsubq	32*4-128($b_ptr), $ACC4, $ACC4
-	vpsubq	32*5-128($b_ptr), $ACC5, $ACC5
-	vpsubq	32*6-128($b_ptr), $ACC6, $ACC6
-	vpsubq	32*7-128($b_ptr), $ACC7, $ACC7
-	vpsubq	32*8-128($b_ptr), $ACC8, $ACC8
-
-	ret
-.size	avx2_sub_x4,.-avx2_sub_x4
-
-.type	avx2_select_n_store,\@abi-omnipotent
-.align	32
-avx2_select_n_store:
-	vmovdqa	`8+32*9*8`(%rsp), $Y
-	vpor	`8+32*9*8+32`(%rsp), $Y, $Y
-
-	vpandn	$ACC0, $Y, $ACC0
-	vpandn	$ACC1, $Y, $ACC1
-	vpandn	$ACC2, $Y, $ACC2
-	vpandn	$ACC3, $Y, $ACC3
-	vpandn	$ACC4, $Y, $ACC4
-	vpandn	$ACC5, $Y, $ACC5
-	vpandn	$ACC6, $Y, $ACC6
-	vmovdqa	`8+32*9*8+32`(%rsp), $B
-	vpandn	$ACC7, $Y, $ACC7
-	vpandn	`8+32*9*8`(%rsp), $B, $B
-	vpandn	$ACC8, $Y, $ACC8
-
-	vpand	32*0(%rsi), $B, $T0
-	lea	160(%rsi), %rax
-	vpand	32*1(%rsi), $B, $Y
-	vpxor	$T0, $ACC0, $ACC0
-	vpand	32*2(%rsi), $B, $T0
-	vpxor	$Y, $ACC1, $ACC1
-	vpand	32*3(%rsi), $B, $Y
-	vpxor	$T0, $ACC2, $ACC2
-	vpand	32*4-160(%rax), $B, $T0
-	vpxor	$Y, $ACC3, $ACC3
-	vpand	32*5-160(%rax), $B, $Y
-	vpxor	$T0, $ACC4, $ACC4
-	vpand	32*6-160(%rax), $B, $T0
-	vpxor	$Y, $ACC5, $ACC5
-	vpand	32*7-160(%rax), $B, $Y
-	vpxor	$T0, $ACC6, $ACC6
-	vpand	32*8-160(%rax), $B, $T0
-	vmovdqa	`8+32*9*8+32`(%rsp), $B
-	vpxor	$Y, $ACC7, $ACC7
-
-	vpand	32*0(%rdx), $B, $Y
-	lea	160(%rdx), %rax
-	vpxor	$T0, $ACC8, $ACC8
-	vpand	32*1(%rdx), $B, $T0
-	vpxor	$Y, $ACC0, $ACC0
-	vpand	32*2(%rdx), $B, $Y
-	vpxor	$T0, $ACC1, $ACC1
-	vpand	32*3(%rdx), $B, $T0
-	vpxor	$Y, $ACC2, $ACC2
-	vpand	32*4-160(%rax), $B, $Y
-	vpxor	$T0, $ACC3, $ACC3
-	vpand	32*5-160(%rax), $B, $T0
-	vpxor	$Y, $ACC4, $ACC4
-	vpand	32*6-160(%rax), $B, $Y
-	vpxor	$T0, $ACC5, $ACC5
-	vpand	32*7-160(%rax), $B, $T0
-	vpxor	$Y, $ACC6, $ACC6
-	vpand	32*8-160(%rax), $B, $Y
-	vpxor	$T0, $ACC7, $ACC7
-	vpxor	$Y, $ACC8, $ACC8
-	`&STORE`
-
-	ret
-.size	avx2_select_n_store,.-avx2_select_n_store
-___
-$code.=<<___	if (0);				# inlined
-################################################################################
-# void avx2_mul_by2_x4(void* RESULTx4, void *Ax4);
-.type	avx2_mul_by2_x4,\@abi-omnipotent
-.align	32
-avx2_mul_by2_x4:
-	vmovdqa	32*0($a_ptr), $ACC0
-	lea	160($a_ptr), %rax
-	vmovdqa	32*1($a_ptr), $ACC1
-	vmovdqa	32*2($a_ptr), $ACC2
-	vmovdqa	32*3($a_ptr), $ACC3
-	vmovdqa	32*4-160(%rax), $ACC4
-	vmovdqa	32*5-160(%rax), $ACC5
-	vmovdqa	32*6-160(%rax), $ACC6
-	vmovdqa	32*7-160(%rax), $ACC7
-	vmovdqa	32*8-160(%rax), $ACC8
-
-	vpaddq	$ACC0, $ACC0, $ACC0
-	vpaddq	$ACC1, $ACC1, $ACC1
-	vpaddq	$ACC2, $ACC2, $ACC2
-	vpaddq	$ACC3, $ACC3, $ACC3
-	vpaddq	$ACC4, $ACC4, $ACC4
-	vpaddq	$ACC5, $ACC5, $ACC5
-	vpaddq	$ACC6, $ACC6, $ACC6
-	vpaddq	$ACC7, $ACC7, $ACC7
-	vpaddq	$ACC8, $ACC8, $ACC8
-
-	ret
-.size	avx2_mul_by2_x4,.-avx2_mul_by2_x4
-___
-my ($r_ptr_in,$a_ptr_in,$b_ptr_in)=("%rdi","%rsi","%rdx");
-my ($r_ptr,$a_ptr,$b_ptr)=("%r8","%r9","%r10");
-
-$code.=<<___;
-################################################################################
-# void ecp_nistz256_avx2_point_add_affine_x4(void* RESULTx4, void *Ax4, void *Bx4);
-.globl	ecp_nistz256_avx2_point_add_affine_x4
-.type	ecp_nistz256_avx2_point_add_affine_x4,\@function,3
-.align	32
-ecp_nistz256_avx2_point_add_affine_x4:
-	mov	%rsp, %rax
-	push    %rbp
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	lea	-16*10(%rsp), %rsp
-	vmovaps	%xmm6, -8-16*10(%rax)
-	vmovaps	%xmm7, -8-16*9(%rax)
-	vmovaps	%xmm8, -8-16*8(%rax)
-	vmovaps	%xmm9, -8-16*7(%rax)
-	vmovaps	%xmm10, -8-16*6(%rax)
-	vmovaps	%xmm11, -8-16*5(%rax)
-	vmovaps	%xmm12, -8-16*4(%rax)
-	vmovaps	%xmm13, -8-16*3(%rax)
-	vmovaps	%xmm14, -8-16*2(%rax)
-	vmovaps	%xmm15, -8-16*1(%rax)
-___
-$code.=<<___;
-	lea	-8(%rax), %rbp
-
-# Result + 32*0 = Result.X
-# Result + 32*9 = Result.Y
-# Result + 32*18 = Result.Z
-
-# A + 32*0 = A.X
-# A + 32*9 = A.Y
-# A + 32*18 = A.Z
-
-# B + 32*0 = B.X
-# B + 32*9 = B.Y
-
-	sub	\$`32*9*8+32*2+32*8`, %rsp
-	and	\$-64, %rsp
-
-	mov	$r_ptr_in, $r_ptr
-	mov	$a_ptr_in, $a_ptr
-	mov	$b_ptr_in, $b_ptr
-
-	vmovdqa	32*0($a_ptr_in), %ymm0
-	vmovdqa	.LAVX2_AND_MASK(%rip), $AND_MASK
-	vpxor	%ymm1, %ymm1, %ymm1
-	lea	256($a_ptr_in), %rax		# size optimization
-	vpor	32*1($a_ptr_in), %ymm0, %ymm0
-	vpor	32*2($a_ptr_in), %ymm0, %ymm0
-	vpor	32*3($a_ptr_in), %ymm0, %ymm0
-	vpor	32*4-256(%rax), %ymm0, %ymm0
-	lea	256(%rax), %rcx			# size optimization
-	vpor	32*5-256(%rax), %ymm0, %ymm0
-	vpor	32*6-256(%rax), %ymm0, %ymm0
-	vpor	32*7-256(%rax), %ymm0, %ymm0
-	vpor	32*8-256(%rax), %ymm0, %ymm0
-	vpor	32*9-256(%rax), %ymm0, %ymm0
-	vpor	32*10-256(%rax), %ymm0, %ymm0
-	vpor	32*11-256(%rax), %ymm0, %ymm0
-	vpor	32*12-512(%rcx), %ymm0, %ymm0
-	vpor	32*13-512(%rcx), %ymm0, %ymm0
-	vpor	32*14-512(%rcx), %ymm0, %ymm0
-	vpor	32*15-512(%rcx), %ymm0, %ymm0
-	vpor	32*16-512(%rcx), %ymm0, %ymm0
-	vpor	32*17-512(%rcx), %ymm0, %ymm0
-	vpcmpeqq %ymm1, %ymm0, %ymm0
-	vmovdqa	%ymm0, `32*9*8`(%rsp)
-
-	vpxor	%ymm1, %ymm1, %ymm1
-	vmovdqa	32*0($b_ptr), %ymm0
-	lea	256($b_ptr), %rax		# size optimization
-	vpor	32*1($b_ptr), %ymm0, %ymm0
-	vpor	32*2($b_ptr), %ymm0, %ymm0
-	vpor	32*3($b_ptr), %ymm0, %ymm0
-	vpor	32*4-256(%rax), %ymm0, %ymm0
-	lea	256(%rax), %rcx			# size optimization
-	vpor	32*5-256(%rax), %ymm0, %ymm0
-	vpor	32*6-256(%rax), %ymm0, %ymm0
-	vpor	32*7-256(%rax), %ymm0, %ymm0
-	vpor	32*8-256(%rax), %ymm0, %ymm0
-	vpor	32*9-256(%rax), %ymm0, %ymm0
-	vpor	32*10-256(%rax), %ymm0, %ymm0
-	vpor	32*11-256(%rax), %ymm0, %ymm0
-	vpor	32*12-512(%rcx), %ymm0, %ymm0
-	vpor	32*13-512(%rcx), %ymm0, %ymm0
-	vpor	32*14-512(%rcx), %ymm0, %ymm0
-	vpor	32*15-512(%rcx), %ymm0, %ymm0
-	vpor	32*16-512(%rcx), %ymm0, %ymm0
-	vpor	32*17-512(%rcx), %ymm0, %ymm0
-	vpcmpeqq %ymm1, %ymm0, %ymm0
-	vmovdqa	%ymm0, `32*9*8+32`(%rsp)
-
-	#	Z1^2 = Z1*Z1
-	lea	`32*9*2`($a_ptr), %rsi
-	lea	`32*9*2`(%rsp), %rdi
-	lea	`32*9*8+32*2`(%rsp), %rcx	# temporary vector
-	call	avx2_sqr_x4
-	call	avx2_normalize_n_store
-
-	#	U2 = X2*Z1^2
-	lea	`32*9*0`($b_ptr), %rsi
-	lea	`32*9*2`(%rsp), %rdx
-	lea	`32*9*0`(%rsp), %rdi
-	call	avx2_mul_x4
-	#call	avx2_normalize
-	`&STORE`
-
-	#	S2 = Z1*Z1^2 = Z1^3
-	lea	`32*9*2`($a_ptr), %rsi
-	lea	`32*9*2`(%rsp), %rdx
-	lea	`32*9*1`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#	S2 = S2*Y2 = Y2*Z1^3
-	lea	`32*9*1`($b_ptr), %rsi
-	lea	`32*9*1`(%rsp), %rdx
-	lea	`32*9*1`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#	H = U2 - U1 = U2 - X1
-	lea	`32*9*0`(%rsp), %rsi
-	lea	`32*9*0`($a_ptr), %rdx
-	lea	`32*9*3`(%rsp), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize_n_store
-
-	#	R = S2 - S1 = S2 - Y1
-	lea	`32*9*1`(%rsp), %rsi
-	lea	`32*9*1`($a_ptr), %rdx
-	lea	`32*9*4`(%rsp), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize_n_store
-
-	#	Z3 = H*Z1*Z2
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*2`($a_ptr), %rdx
-	lea	`32*9*2`($r_ptr), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize
-
-	lea	.LONE(%rip), %rsi
-	lea	`32*9*2`($a_ptr), %rdx
-	call	avx2_select_n_store
-
-	#	R^2 = R^2
-	lea	`32*9*4`(%rsp), %rsi
-	lea	`32*9*6`(%rsp), %rdi
-	lea	`32*9*8+32*2`(%rsp), %rcx	# temporary vector
-	call	avx2_sqr_x4
-	call	avx2_normalize_n_store
-
-	#	H^2 = H^2
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*5`(%rsp), %rdi
-	call	avx2_sqr_x4
-	call	avx2_normalize_n_store
-
-	#	H^3 = H^2*H
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*5`(%rsp), %rdx
-	lea	`32*9*7`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#	U2 = U1*H^2
-	lea	`32*9*0`($a_ptr), %rsi
-	lea	`32*9*5`(%rsp), %rdx
-	lea	`32*9*0`(%rsp), %rdi
-	call	avx2_mul_x4
-	#call	avx2_normalize
-	`&STORE`
-
-	#	Hsqr = U2*2
-	#lea	32*9*0(%rsp), %rsi
-	#lea	32*9*5(%rsp), %rdi
-	#call	avx2_mul_by2_x4
-
-	vpaddq	$ACC0, $ACC0, $ACC0	# inlined avx2_mul_by2_x4
-	lea	`32*9*5`(%rsp), %rdi
-	vpaddq	$ACC1, $ACC1, $ACC1
-	vpaddq	$ACC2, $ACC2, $ACC2
-	vpaddq	$ACC3, $ACC3, $ACC3
-	vpaddq	$ACC4, $ACC4, $ACC4
-	vpaddq	$ACC5, $ACC5, $ACC5
-	vpaddq	$ACC6, $ACC6, $ACC6
-	vpaddq	$ACC7, $ACC7, $ACC7
-	vpaddq	$ACC8, $ACC8, $ACC8
-	call	avx2_normalize_n_store
-
-	#	X3 = R^2 - H^3
-	#lea	32*9*6(%rsp), %rsi
-	#lea	32*9*7(%rsp), %rdx
-	#lea	32*9*5(%rsp), %rcx
-	#lea	32*9*0($r_ptr), %rdi
-	#call	avx2_sub_x4
-	#NORMALIZE
-	#STORE
-
-	#	X3 = X3 - U2*2
-	#lea	32*9*0($r_ptr), %rsi
-	#lea	32*9*0($r_ptr), %rdi
-	#call	avx2_sub_x4
-	#NORMALIZE
-	#STORE
-
-	lea	`32*9*6+128`(%rsp), %rsi
-	lea	.LAVX2_POLY_x2+128(%rip), %rax
-	lea	`32*9*7+128`(%rsp), %rdx
-	lea	`32*9*5+128`(%rsp), %rcx
-	lea	`32*9*0`($r_ptr), %rdi
-
-	vmovdqa	32*0-128(%rsi), $ACC0
-	vmovdqa	32*1-128(%rsi), $ACC1
-	vmovdqa	32*2-128(%rsi), $ACC2
-	vmovdqa	32*3-128(%rsi), $ACC3
-	vmovdqa	32*4-128(%rsi), $ACC4
-	vmovdqa	32*5-128(%rsi), $ACC5
-	vmovdqa	32*6-128(%rsi), $ACC6
-	vmovdqa	32*7-128(%rsi), $ACC7
-	vmovdqa	32*8-128(%rsi), $ACC8
-
-	vpaddq	32*0-128(%rax), $ACC0, $ACC0
-	vpaddq	32*1-128(%rax), $ACC1, $ACC1
-	vpaddq	32*2-128(%rax), $ACC2, $ACC2
-	vpaddq	32*3-128(%rax), $ACC3, $ACC3
-	vpaddq	32*4-128(%rax), $ACC4, $ACC4
-	vpaddq	32*5-128(%rax), $ACC5, $ACC5
-	vpaddq	32*6-128(%rax), $ACC6, $ACC6
-	vpaddq	32*7-128(%rax), $ACC7, $ACC7
-	vpaddq	32*8-128(%rax), $ACC8, $ACC8
-
-	vpsubq	32*0-128(%rdx), $ACC0, $ACC0
-	vpsubq	32*1-128(%rdx), $ACC1, $ACC1
-	vpsubq	32*2-128(%rdx), $ACC2, $ACC2
-	vpsubq	32*3-128(%rdx), $ACC3, $ACC3
-	vpsubq	32*4-128(%rdx), $ACC4, $ACC4
-	vpsubq	32*5-128(%rdx), $ACC5, $ACC5
-	vpsubq	32*6-128(%rdx), $ACC6, $ACC6
-	vpsubq	32*7-128(%rdx), $ACC7, $ACC7
-	vpsubq	32*8-128(%rdx), $ACC8, $ACC8
-
-	vpsubq	32*0-128(%rcx), $ACC0, $ACC0
-	vpsubq	32*1-128(%rcx), $ACC1, $ACC1
-	vpsubq	32*2-128(%rcx), $ACC2, $ACC2
-	vpsubq	32*3-128(%rcx), $ACC3, $ACC3
-	vpsubq	32*4-128(%rcx), $ACC4, $ACC4
-	vpsubq	32*5-128(%rcx), $ACC5, $ACC5
-	vpsubq	32*6-128(%rcx), $ACC6, $ACC6
-	vpsubq	32*7-128(%rcx), $ACC7, $ACC7
-	vpsubq	32*8-128(%rcx), $ACC8, $ACC8
-	call	avx2_normalize
-
-	lea	32*0($b_ptr), %rsi
-	lea	32*0($a_ptr), %rdx
-	call	avx2_select_n_store
-
-	#	H = U2 - X3
-	lea	`32*9*0`(%rsp), %rsi
-	lea	`32*9*0`($r_ptr), %rdx
-	lea	`32*9*3`(%rsp), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize_n_store
-
-	#
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*4`(%rsp), %rdx
-	lea	`32*9*3`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#
-	lea	`32*9*7`(%rsp), %rsi
-	lea	`32*9*1`($a_ptr), %rdx
-	lea	`32*9*1`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*1`(%rsp), %rdx
-	lea	`32*9*1`($r_ptr), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize
-
-	lea	32*9($b_ptr), %rsi
-	lea	32*9($a_ptr), %rdx
-	call	avx2_select_n_store
-
-	#lea	32*9*0($r_ptr), %rsi
-	#lea	32*9*0($r_ptr), %rdi
-	#call	avx2_mul_by1_x4
-	#NORMALIZE
-	#STORE
-
-	lea	`32*9*1`($r_ptr), %rsi
-	lea	`32*9*1`($r_ptr), %rdi
-	call	avx2_mul_by1_x4
-	call	avx2_normalize_n_store
-
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	movaps	%xmm6, -16*10(%rbp)
-	movaps	%xmm7, -16*9(%rbp)
-	movaps	%xmm8, -16*8(%rbp)
-	movaps	%xmm9, -16*7(%rbp)
-	movaps	%xmm10, -16*6(%rbp)
-	movaps	%xmm11, -16*5(%rbp)
-	movaps	%xmm12, -16*4(%rbp)
-	movaps	%xmm13, -16*3(%rbp)
-	movaps	%xmm14, -16*2(%rbp)
-	movaps	%xmm15, -16*1(%rbp)
-___
-$code.=<<___;
-	mov	%rbp, %rsp
-	pop	%rbp
-	ret
-.size	ecp_nistz256_avx2_point_add_affine_x4,.-ecp_nistz256_avx2_point_add_affine_x4
-
-################################################################################
-# void ecp_nistz256_avx2_point_add_affines_x4(void* RESULTx4, void *Ax4, void *Bx4);
-.globl	ecp_nistz256_avx2_point_add_affines_x4
-.type	ecp_nistz256_avx2_point_add_affines_x4,\@function,3
-.align	32
-ecp_nistz256_avx2_point_add_affines_x4:
-	mov	%rsp, %rax
-	push    %rbp
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	lea	-16*10(%rsp), %rsp
-	vmovaps	%xmm6, -8-16*10(%rax)
-	vmovaps	%xmm7, -8-16*9(%rax)
-	vmovaps	%xmm8, -8-16*8(%rax)
-	vmovaps	%xmm9, -8-16*7(%rax)
-	vmovaps	%xmm10, -8-16*6(%rax)
-	vmovaps	%xmm11, -8-16*5(%rax)
-	vmovaps	%xmm12, -8-16*4(%rax)
-	vmovaps	%xmm13, -8-16*3(%rax)
-	vmovaps	%xmm14, -8-16*2(%rax)
-	vmovaps	%xmm15, -8-16*1(%rax)
-___
-$code.=<<___;
-	lea	-8(%rax), %rbp
-
-# Result + 32*0 = Result.X
-# Result + 32*9 = Result.Y
-# Result + 32*18 = Result.Z
-
-# A + 32*0 = A.X
-# A + 32*9 = A.Y
-
-# B + 32*0 = B.X
-# B + 32*9 = B.Y
-
-	sub	\$`32*9*8+32*2+32*8`, %rsp
-	and	\$-64, %rsp
-
-	mov	$r_ptr_in, $r_ptr
-	mov	$a_ptr_in, $a_ptr
-	mov	$b_ptr_in, $b_ptr
-
-	vmovdqa	32*0($a_ptr_in), %ymm0
-	vmovdqa	.LAVX2_AND_MASK(%rip), $AND_MASK
-	vpxor	%ymm1, %ymm1, %ymm1
-	lea	256($a_ptr_in), %rax		# size optimization
-	vpor	32*1($a_ptr_in), %ymm0, %ymm0
-	vpor	32*2($a_ptr_in), %ymm0, %ymm0
-	vpor	32*3($a_ptr_in), %ymm0, %ymm0
-	vpor	32*4-256(%rax), %ymm0, %ymm0
-	lea	256(%rax), %rcx			# size optimization
-	vpor	32*5-256(%rax), %ymm0, %ymm0
-	vpor	32*6-256(%rax), %ymm0, %ymm0
-	vpor	32*7-256(%rax), %ymm0, %ymm0
-	vpor	32*8-256(%rax), %ymm0, %ymm0
-	vpor	32*9-256(%rax), %ymm0, %ymm0
-	vpor	32*10-256(%rax), %ymm0, %ymm0
-	vpor	32*11-256(%rax), %ymm0, %ymm0
-	vpor	32*12-512(%rcx), %ymm0, %ymm0
-	vpor	32*13-512(%rcx), %ymm0, %ymm0
-	vpor	32*14-512(%rcx), %ymm0, %ymm0
-	vpor	32*15-512(%rcx), %ymm0, %ymm0
-	vpor	32*16-512(%rcx), %ymm0, %ymm0
-	vpor	32*17-512(%rcx), %ymm0, %ymm0
-	vpcmpeqq %ymm1, %ymm0, %ymm0
-	vmovdqa	%ymm0, `32*9*8`(%rsp)
-
-	vpxor	%ymm1, %ymm1, %ymm1
-	vmovdqa	32*0($b_ptr), %ymm0
-	lea	256($b_ptr), %rax		# size optimization
-	vpor	32*1($b_ptr), %ymm0, %ymm0
-	vpor	32*2($b_ptr), %ymm0, %ymm0
-	vpor	32*3($b_ptr), %ymm0, %ymm0
-	vpor	32*4-256(%rax), %ymm0, %ymm0
-	lea	256(%rax), %rcx			# size optimization
-	vpor	32*5-256(%rax), %ymm0, %ymm0
-	vpor	32*6-256(%rax), %ymm0, %ymm0
-	vpor	32*7-256(%rax), %ymm0, %ymm0
-	vpor	32*8-256(%rax), %ymm0, %ymm0
-	vpor	32*9-256(%rax), %ymm0, %ymm0
-	vpor	32*10-256(%rax), %ymm0, %ymm0
-	vpor	32*11-256(%rax), %ymm0, %ymm0
-	vpor	32*12-512(%rcx), %ymm0, %ymm0
-	vpor	32*13-512(%rcx), %ymm0, %ymm0
-	vpor	32*14-512(%rcx), %ymm0, %ymm0
-	vpor	32*15-512(%rcx), %ymm0, %ymm0
-	vpor	32*16-512(%rcx), %ymm0, %ymm0
-	vpor	32*17-512(%rcx), %ymm0, %ymm0
-	vpcmpeqq %ymm1, %ymm0, %ymm0
-	vmovdqa	%ymm0, `32*9*8+32`(%rsp)
-
-	#	H = U2 - U1 = X2 - X1
-	lea	`32*9*0`($b_ptr), %rsi
-	lea	`32*9*0`($a_ptr), %rdx
-	lea	`32*9*3`(%rsp), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize_n_store
-
-	#	R = S2 - S1 = Y2 - Y1
-	lea	`32*9*1`($b_ptr), %rsi
-	lea	`32*9*1`($a_ptr), %rdx
-	lea	`32*9*4`(%rsp), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize_n_store
-
-	#	Z3 = H*Z1*Z2 = H
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*2`($r_ptr), %rdi
-	call	avx2_mul_by1_x4
-	call	avx2_normalize
-
-	vmovdqa	`32*9*8`(%rsp), $B
-	vpor	`32*9*8+32`(%rsp), $B, $B
-
-	vpandn	$ACC0, $B, $ACC0
-	lea	.LONE+128(%rip), %rax
-	vpandn	$ACC1, $B, $ACC1
-	vpandn	$ACC2, $B, $ACC2
-	vpandn	$ACC3, $B, $ACC3
-	vpandn	$ACC4, $B, $ACC4
-	vpandn	$ACC5, $B, $ACC5
-	vpandn	$ACC6, $B, $ACC6
-	vpandn	$ACC7, $B, $ACC7
-
-	vpand	32*0-128(%rax), $B, $T0
-	 vpandn	$ACC8, $B, $ACC8
-	vpand	32*1-128(%rax), $B, $Y
-	vpxor	$T0, $ACC0, $ACC0
-	vpand	32*2-128(%rax), $B, $T0
-	vpxor	$Y, $ACC1, $ACC1
-	vpand	32*3-128(%rax), $B, $Y
-	vpxor	$T0, $ACC2, $ACC2
-	vpand	32*4-128(%rax), $B, $T0
-	vpxor	$Y, $ACC3, $ACC3
-	vpand	32*5-128(%rax), $B, $Y
-	vpxor	$T0, $ACC4, $ACC4
-	vpand	32*6-128(%rax), $B, $T0
-	vpxor	$Y, $ACC5, $ACC5
-	vpand	32*7-128(%rax), $B, $Y
-	vpxor	$T0, $ACC6, $ACC6
-	vpand	32*8-128(%rax), $B, $T0
-	vpxor	$Y, $ACC7, $ACC7
-	vpxor	$T0, $ACC8, $ACC8
-	`&STORE`
-
-	#	R^2 = R^2
-	lea	`32*9*4`(%rsp), %rsi
-	lea	`32*9*6`(%rsp), %rdi
-	lea	`32*9*8+32*2`(%rsp), %rcx	# temporary vector
-	call	avx2_sqr_x4
-	call	avx2_normalize_n_store
-
-	#	H^2 = H^2
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*5`(%rsp), %rdi
-	call	avx2_sqr_x4
-	call	avx2_normalize_n_store
-
-	#	H^3 = H^2*H
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*5`(%rsp), %rdx
-	lea	`32*9*7`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#	U2 = U1*H^2
-	lea	`32*9*0`($a_ptr), %rsi
-	lea	`32*9*5`(%rsp), %rdx
-	lea	`32*9*0`(%rsp), %rdi
-	call	avx2_mul_x4
-	#call	avx2_normalize
-	`&STORE`
-
-	#	Hsqr = U2*2
-	#lea	32*9*0(%rsp), %rsi
-	#lea	32*9*5(%rsp), %rdi
-	#call	avx2_mul_by2_x4
-
-	vpaddq	$ACC0, $ACC0, $ACC0	# inlined avx2_mul_by2_x4
-	lea	`32*9*5`(%rsp), %rdi
-	vpaddq	$ACC1, $ACC1, $ACC1
-	vpaddq	$ACC2, $ACC2, $ACC2
-	vpaddq	$ACC3, $ACC3, $ACC3
-	vpaddq	$ACC4, $ACC4, $ACC4
-	vpaddq	$ACC5, $ACC5, $ACC5
-	vpaddq	$ACC6, $ACC6, $ACC6
-	vpaddq	$ACC7, $ACC7, $ACC7
-	vpaddq	$ACC8, $ACC8, $ACC8
-	call	avx2_normalize_n_store
-
-	#	X3 = R^2 - H^3
-	#lea	32*9*6(%rsp), %rsi
-	#lea	32*9*7(%rsp), %rdx
-	#lea	32*9*5(%rsp), %rcx
-	#lea	32*9*0($r_ptr), %rdi
-	#call	avx2_sub_x4
-	#NORMALIZE
-	#STORE
-
-	#	X3 = X3 - U2*2
-	#lea	32*9*0($r_ptr), %rsi
-	#lea	32*9*0($r_ptr), %rdi
-	#call	avx2_sub_x4
-	#NORMALIZE
-	#STORE
-
-	lea	`32*9*6+128`(%rsp), %rsi
-	lea	.LAVX2_POLY_x2+128(%rip), %rax
-	lea	`32*9*7+128`(%rsp), %rdx
-	lea	`32*9*5+128`(%rsp), %rcx
-	lea	`32*9*0`($r_ptr), %rdi
-
-	vmovdqa	32*0-128(%rsi), $ACC0
-	vmovdqa	32*1-128(%rsi), $ACC1
-	vmovdqa	32*2-128(%rsi), $ACC2
-	vmovdqa	32*3-128(%rsi), $ACC3
-	vmovdqa	32*4-128(%rsi), $ACC4
-	vmovdqa	32*5-128(%rsi), $ACC5
-	vmovdqa	32*6-128(%rsi), $ACC6
-	vmovdqa	32*7-128(%rsi), $ACC7
-	vmovdqa	32*8-128(%rsi), $ACC8
-
-	vpaddq	32*0-128(%rax), $ACC0, $ACC0
-	vpaddq	32*1-128(%rax), $ACC1, $ACC1
-	vpaddq	32*2-128(%rax), $ACC2, $ACC2
-	vpaddq	32*3-128(%rax), $ACC3, $ACC3
-	vpaddq	32*4-128(%rax), $ACC4, $ACC4
-	vpaddq	32*5-128(%rax), $ACC5, $ACC5
-	vpaddq	32*6-128(%rax), $ACC6, $ACC6
-	vpaddq	32*7-128(%rax), $ACC7, $ACC7
-	vpaddq	32*8-128(%rax), $ACC8, $ACC8
-
-	vpsubq	32*0-128(%rdx), $ACC0, $ACC0
-	vpsubq	32*1-128(%rdx), $ACC1, $ACC1
-	vpsubq	32*2-128(%rdx), $ACC2, $ACC2
-	vpsubq	32*3-128(%rdx), $ACC3, $ACC3
-	vpsubq	32*4-128(%rdx), $ACC4, $ACC4
-	vpsubq	32*5-128(%rdx), $ACC5, $ACC5
-	vpsubq	32*6-128(%rdx), $ACC6, $ACC6
-	vpsubq	32*7-128(%rdx), $ACC7, $ACC7
-	vpsubq	32*8-128(%rdx), $ACC8, $ACC8
-
-	vpsubq	32*0-128(%rcx), $ACC0, $ACC0
-	vpsubq	32*1-128(%rcx), $ACC1, $ACC1
-	vpsubq	32*2-128(%rcx), $ACC2, $ACC2
-	vpsubq	32*3-128(%rcx), $ACC3, $ACC3
-	vpsubq	32*4-128(%rcx), $ACC4, $ACC4
-	vpsubq	32*5-128(%rcx), $ACC5, $ACC5
-	vpsubq	32*6-128(%rcx), $ACC6, $ACC6
-	vpsubq	32*7-128(%rcx), $ACC7, $ACC7
-	vpsubq	32*8-128(%rcx), $ACC8, $ACC8
-	call	avx2_normalize
-
-	lea	32*0($b_ptr), %rsi
-	lea	32*0($a_ptr), %rdx
-	call	avx2_select_n_store
-
-	#	H = U2 - X3
-	lea	`32*9*0`(%rsp), %rsi
-	lea	`32*9*0`($r_ptr), %rdx
-	lea	`32*9*3`(%rsp), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize_n_store
-
-	#	H = H*R
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*4`(%rsp), %rdx
-	lea	`32*9*3`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#	S2 = S1 * H^3
-	lea	`32*9*7`(%rsp), %rsi
-	lea	`32*9*1`($a_ptr), %rdx
-	lea	`32*9*1`(%rsp), %rdi
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	#
-	lea	`32*9*3`(%rsp), %rsi
-	lea	`32*9*1`(%rsp), %rdx
-	lea	`32*9*1`($r_ptr), %rdi
-	call	avx2_sub_x4
-	call	avx2_normalize
-
-	lea	32*9($b_ptr), %rsi
-	lea	32*9($a_ptr), %rdx
-	call	avx2_select_n_store
-
-	#lea	32*9*0($r_ptr), %rsi
-	#lea	32*9*0($r_ptr), %rdi
-	#call	avx2_mul_by1_x4
-	#NORMALIZE
-	#STORE
-
-	lea	`32*9*1`($r_ptr), %rsi
-	lea	`32*9*1`($r_ptr), %rdi
-	call	avx2_mul_by1_x4
-	call	avx2_normalize_n_store
-
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	movaps	%xmm6, -16*10(%rbp)
-	movaps	%xmm7, -16*9(%rbp)
-	movaps	%xmm8, -16*8(%rbp)
-	movaps	%xmm9, -16*7(%rbp)
-	movaps	%xmm10, -16*6(%rbp)
-	movaps	%xmm11, -16*5(%rbp)
-	movaps	%xmm12, -16*4(%rbp)
-	movaps	%xmm13, -16*3(%rbp)
-	movaps	%xmm14, -16*2(%rbp)
-	movaps	%xmm15, -16*1(%rbp)
-___
-$code.=<<___;
-	mov	%rbp, %rsp
-	pop	%rbp
-	ret
-.size	ecp_nistz256_avx2_point_add_affines_x4,.-ecp_nistz256_avx2_point_add_affines_x4
-
-################################################################################
-# void ecp_nistz256_avx2_to_mont(void* RESULTx4, void *Ax4);
-.globl	ecp_nistz256_avx2_to_mont
-.type	ecp_nistz256_avx2_to_mont,\@function,2
-.align	32
-ecp_nistz256_avx2_to_mont:
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	lea	-8-16*10(%rsp), %rsp
-	vmovaps	%xmm6, -8-16*10(%rax)
-	vmovaps	%xmm7, -8-16*9(%rax)
-	vmovaps	%xmm8, -8-16*8(%rax)
-	vmovaps	%xmm9, -8-16*7(%rax)
-	vmovaps	%xmm10, -8-16*6(%rax)
-	vmovaps	%xmm11, -8-16*5(%rax)
-	vmovaps	%xmm12, -8-16*4(%rax)
-	vmovaps	%xmm13, -8-16*3(%rax)
-	vmovaps	%xmm14, -8-16*2(%rax)
-	vmovaps	%xmm15, -8-16*1(%rax)
-___
-$code.=<<___;
-	vmovdqa	.LAVX2_AND_MASK(%rip), $AND_MASK
-	lea	.LTO_MONT_AVX2(%rip), %rdx
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	movaps	16*0(%rsp), %xmm6
-	movaps	16*1(%rsp), %xmm7
-	movaps	16*2(%rsp), %xmm8
-	movaps	16*3(%rsp), %xmm9
-	movaps	16*4(%rsp), %xmm10
-	movaps	16*5(%rsp), %xmm11
-	movaps	16*6(%rsp), %xmm12
-	movaps	16*7(%rsp), %xmm13
-	movaps	16*8(%rsp), %xmm14
-	movaps	16*9(%rsp), %xmm15
-	lea	8+16*10(%rsp), %rsp
-___
-$code.=<<___;
-	ret
-.size	ecp_nistz256_avx2_to_mont,.-ecp_nistz256_avx2_to_mont
-
-################################################################################
-# void ecp_nistz256_avx2_from_mont(void* RESULTx4, void *Ax4);
-.globl	ecp_nistz256_avx2_from_mont
-.type	ecp_nistz256_avx2_from_mont,\@function,2
-.align	32
-ecp_nistz256_avx2_from_mont:
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	lea	-8-16*10(%rsp), %rsp
-	vmovaps	%xmm6, -8-16*10(%rax)
-	vmovaps	%xmm7, -8-16*9(%rax)
-	vmovaps	%xmm8, -8-16*8(%rax)
-	vmovaps	%xmm9, -8-16*7(%rax)
-	vmovaps	%xmm10, -8-16*6(%rax)
-	vmovaps	%xmm11, -8-16*5(%rax)
-	vmovaps	%xmm12, -8-16*4(%rax)
-	vmovaps	%xmm13, -8-16*3(%rax)
-	vmovaps	%xmm14, -8-16*2(%rax)
-	vmovaps	%xmm15, -8-16*1(%rax)
-___
-$code.=<<___;
-	vmovdqa	.LAVX2_AND_MASK(%rip), $AND_MASK
-	lea	.LFROM_MONT_AVX2(%rip), %rdx
-	call	avx2_mul_x4
-	call	avx2_normalize_n_store
-
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	movaps	16*0(%rsp), %xmm6
-	movaps	16*1(%rsp), %xmm7
-	movaps	16*2(%rsp), %xmm8
-	movaps	16*3(%rsp), %xmm9
-	movaps	16*4(%rsp), %xmm10
-	movaps	16*5(%rsp), %xmm11
-	movaps	16*6(%rsp), %xmm12
-	movaps	16*7(%rsp), %xmm13
-	movaps	16*8(%rsp), %xmm14
-	movaps	16*9(%rsp), %xmm15
-	lea	8+16*10(%rsp), %rsp
-___
-$code.=<<___;
-	ret
-.size	ecp_nistz256_avx2_from_mont,.-ecp_nistz256_avx2_from_mont
-
-################################################################################
-# void ecp_nistz256_avx2_set1(void* RESULTx4);
-.globl	ecp_nistz256_avx2_set1
-.type	ecp_nistz256_avx2_set1,\@function,1
-.align	32
-ecp_nistz256_avx2_set1:
-	lea	.LONE+128(%rip), %rax
-	lea	128(%rdi), %rdi
-	vzeroupper
-	vmovdqa	32*0-128(%rax), %ymm0
-	vmovdqa	32*1-128(%rax), %ymm1
-	vmovdqa	32*2-128(%rax), %ymm2
-	vmovdqa	32*3-128(%rax), %ymm3
-	vmovdqa	32*4-128(%rax), %ymm4
-	vmovdqa	32*5-128(%rax), %ymm5
-	vmovdqa	%ymm0, 32*0-128(%rdi)
-	vmovdqa	32*6-128(%rax), %ymm0
-	vmovdqa	%ymm1, 32*1-128(%rdi)
-	vmovdqa	32*7-128(%rax), %ymm1
-	vmovdqa	%ymm2, 32*2-128(%rdi)
-	vmovdqa	32*8-128(%rax), %ymm2
-	vmovdqa	%ymm3, 32*3-128(%rdi)
-	vmovdqa	%ymm4, 32*4-128(%rdi)
-	vmovdqa	%ymm5, 32*5-128(%rdi)
-	vmovdqa	%ymm0, 32*6-128(%rdi)
-	vmovdqa	%ymm1, 32*7-128(%rdi)
-	vmovdqa	%ymm2, 32*8-128(%rdi)
-
-	vzeroupper
-	ret
-.size	ecp_nistz256_avx2_set1,.-ecp_nistz256_avx2_set1
-___
-}
-{
-################################################################################
-# void ecp_nistz256_avx2_multi_gather_w7(void* RESULT, void *in,
-#			    int index0, int index1, int index2, int index3);
-################################################################################
-
-my ($val,$in_t,$index0,$index1,$index2,$index3)=("%rdi","%rsi","%edx","%ecx","%r8d","%r9d");
-my ($INDEX0,$INDEX1,$INDEX2,$INDEX3)=map("%ymm$_",(0..3));
-my ($R0a,$R0b,$R1a,$R1b,$R2a,$R2b,$R3a,$R3b)=map("%ymm$_",(4..11));
-my ($M0,$T0,$T1,$TMP0)=map("%ymm$_",(12..15));
-
-$code.=<<___;
-.globl	ecp_nistz256_avx2_multi_gather_w7
-.type	ecp_nistz256_avx2_multi_gather_w7,\@function,6
-.align	32
-ecp_nistz256_avx2_multi_gather_w7:
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	lea	-8-16*10(%rsp), %rsp
-	vmovaps	%xmm6, -8-16*10(%rax)
-	vmovaps	%xmm7, -8-16*9(%rax)
-	vmovaps	%xmm8, -8-16*8(%rax)
-	vmovaps	%xmm9, -8-16*7(%rax)
-	vmovaps	%xmm10, -8-16*6(%rax)
-	vmovaps	%xmm11, -8-16*5(%rax)
-	vmovaps	%xmm12, -8-16*4(%rax)
-	vmovaps	%xmm13, -8-16*3(%rax)
-	vmovaps	%xmm14, -8-16*2(%rax)
-	vmovaps	%xmm15, -8-16*1(%rax)
-___
-$code.=<<___;
-	lea	.LIntOne(%rip), %rax
-
-	vmovd	$index0, %xmm0
-	vmovd	$index1, %xmm1
-	vmovd	$index2, %xmm2
-	vmovd	$index3, %xmm3
-
-	vpxor	$R0a, $R0a, $R0a
-	vpxor	$R0b, $R0b, $R0b
-	vpxor	$R1a, $R1a, $R1a
-	vpxor	$R1b, $R1b, $R1b
-	vpxor	$R2a, $R2a, $R2a
-	vpxor	$R2b, $R2b, $R2b
-	vpxor	$R3a, $R3a, $R3a
-	vpxor	$R3b, $R3b, $R3b
-	vmovdqa	(%rax), $M0
-
-	vpermd	$INDEX0, $R0a, $INDEX0
-	vpermd	$INDEX1, $R0a, $INDEX1
-	vpermd	$INDEX2, $R0a, $INDEX2
-	vpermd	$INDEX3, $R0a, $INDEX3
-
-	mov	\$64, %ecx
-	lea	112($val), $val		# size optimization
-	jmp	.Lmulti_select_loop_avx2
-
-# INDEX=0, corresponds to the point at infty (0,0)
-.align	32
-.Lmulti_select_loop_avx2:
-	vpcmpeqd	$INDEX0, $M0, $TMP0
-
-	vmovdqa		`32*0+32*64*2*0`($in_t), $T0
-	vmovdqa		`32*1+32*64*2*0`($in_t), $T1
-	vpand		$TMP0, $T0, $T0
-	vpand		$TMP0, $T1, $T1
-	vpxor		$T0, $R0a, $R0a
-	vpxor		$T1, $R0b, $R0b
-
-	vpcmpeqd	$INDEX1, $M0, $TMP0
-
-	vmovdqa		`32*0+32*64*2*1`($in_t), $T0
-	vmovdqa		`32*1+32*64*2*1`($in_t), $T1
-	vpand		$TMP0, $T0, $T0
-	vpand		$TMP0, $T1, $T1
-	vpxor		$T0, $R1a, $R1a
-	vpxor		$T1, $R1b, $R1b
-
-	vpcmpeqd	$INDEX2, $M0, $TMP0
-
-	vmovdqa		`32*0+32*64*2*2`($in_t), $T0
-	vmovdqa		`32*1+32*64*2*2`($in_t), $T1
-	vpand		$TMP0, $T0, $T0
-	vpand		$TMP0, $T1, $T1
-	vpxor		$T0, $R2a, $R2a
-	vpxor		$T1, $R2b, $R2b
-
-	vpcmpeqd	$INDEX3, $M0, $TMP0
-
-	vmovdqa		`32*0+32*64*2*3`($in_t), $T0
-	vmovdqa		`32*1+32*64*2*3`($in_t), $T1
-	vpand		$TMP0, $T0, $T0
-	vpand		$TMP0, $T1, $T1
-	vpxor		$T0, $R3a, $R3a
-	vpxor		$T1, $R3b, $R3b
-
-	vpaddd		(%rax), $M0, $M0	# increment
-	lea		32*2($in_t), $in_t
-
-        dec	%ecx
-	jnz	.Lmulti_select_loop_avx2
-
-	vmovdqu	$R0a, 32*0-112($val)
-	vmovdqu	$R0b, 32*1-112($val)
-	vmovdqu	$R1a, 32*2-112($val)
-	vmovdqu	$R1b, 32*3-112($val)
-	vmovdqu	$R2a, 32*4-112($val)
-	vmovdqu	$R2b, 32*5-112($val)
-	vmovdqu	$R3a, 32*6-112($val)
-	vmovdqu	$R3b, 32*7-112($val)
-
-	vzeroupper
-___
-$code.=<<___	if ($win64);
-	movaps	16*0(%rsp), %xmm6
-	movaps	16*1(%rsp), %xmm7
-	movaps	16*2(%rsp), %xmm8
-	movaps	16*3(%rsp), %xmm9
-	movaps	16*4(%rsp), %xmm10
-	movaps	16*5(%rsp), %xmm11
-	movaps	16*6(%rsp), %xmm12
-	movaps	16*7(%rsp), %xmm13
-	movaps	16*8(%rsp), %xmm14
-	movaps	16*9(%rsp), %xmm15
-	lea	8+16*10(%rsp), %rsp
-___
-$code.=<<___;
-	ret
-.size	ecp_nistz256_avx2_multi_gather_w7,.-ecp_nistz256_avx2_multi_gather_w7
-
-.extern	OPENSSL_ia32cap_P
-.globl	ecp_nistz_avx2_eligible
-.type	ecp_nistz_avx2_eligible,\@abi-omnipotent
-.align	32
-ecp_nistz_avx2_eligible:
-	mov	OPENSSL_ia32cap_P+8(%rip),%eax
-	shr	\$5,%eax
-	and	\$1,%eax
-	ret
-.size	ecp_nistz_avx2_eligible,.-ecp_nistz_avx2_eligible
-___
-}
-}} else {{	# assembler is too old
-$code.=<<___;
-.text
-
-.globl	ecp_nistz256_avx2_transpose_convert
-.globl	ecp_nistz256_avx2_convert_transpose_back
-.globl	ecp_nistz256_avx2_point_add_affine_x4
-.globl	ecp_nistz256_avx2_point_add_affines_x4
-.globl	ecp_nistz256_avx2_to_mont
-.globl	ecp_nistz256_avx2_from_mont
-.globl	ecp_nistz256_avx2_set1
-.globl	ecp_nistz256_avx2_multi_gather_w7
-.type	ecp_nistz256_avx2_multi_gather_w7,\@abi-omnipotent
-ecp_nistz256_avx2_transpose_convert:
-ecp_nistz256_avx2_convert_transpose_back:
-ecp_nistz256_avx2_point_add_affine_x4:
-ecp_nistz256_avx2_point_add_affines_x4:
-ecp_nistz256_avx2_to_mont:
-ecp_nistz256_avx2_from_mont:
-ecp_nistz256_avx2_set1:
-ecp_nistz256_avx2_multi_gather_w7:
-	.byte	0x0f,0x0b	# ud2
-	ret
-.size	ecp_nistz256_avx2_multi_gather_w7,.-ecp_nistz256_avx2_multi_gather_w7
-
-.globl	ecp_nistz_avx2_eligible
-.type	ecp_nistz_avx2_eligible,\@abi-omnipotent
-ecp_nistz_avx2_eligible:
-	xor	%eax,%eax
-	ret
-.size	ecp_nistz_avx2_eligible,.-ecp_nistz_avx2_eligible
-___
-}}
-
-foreach (split("\n",$code)) {
-	s/\`([^\`]*)\`/eval($1)/geo;
-
-	print $_,"\n";
-}
-
-close STDOUT or die "error closing STDOUT: $!";
diff --git a/crypto/ec/asm/ecp_nistz256-x86_64.pl b/crypto/ec/asm/ecp_nistz256-x86_64.pl
index de9b194510bf..b50ee70191b2 100755
--- a/crypto/ec/asm/ecp_nistz256-x86_64.pl
+++ b/crypto/ec/asm/ecp_nistz256-x86_64.pl
@@ -72,7 +72,7 @@ if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$addx = ($1>=12);
 }
 
-if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
 	my $ver = $2 + $3/100.0;	# 3.1->3.01, 3.10->3.10
 	$avx = ($ver>=3.0) + ($ver>=3.01);
 	$addx = ($ver>=3.03);
diff --git a/crypto/ec/asm/x25519-x86_64.pl b/crypto/ec/asm/x25519-x86_64.pl
index 3d9d1dc1ad0c..62599dacaccd 100755
--- a/crypto/ec/asm/x25519-x86_64.pl
+++ b/crypto/ec/asm/x25519-x86_64.pl
@@ -90,7 +90,7 @@ if (!$addx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$addx = ($1>=12);
 }
 
-if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
+if (!$addx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+)\.([0-9]+)/) {
 	my $ver = $2 + $3/100.0;	# 3.1->3.01, 3.10->3.10
 	$addx = ($ver>=3.03);
 }
diff --git a/crypto/ec/ec_ameth.c b/crypto/ec/ec_ameth.c
index 221038373921..5098bd7a6602 100644
--- a/crypto/ec/ec_ameth.c
+++ b/crypto/ec/ec_ameth.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2006-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2006-2020 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
@@ -23,7 +23,7 @@ static int ecdh_cms_decrypt(CMS_RecipientInfo *ri);
 static int ecdh_cms_encrypt(CMS_RecipientInfo *ri);
 #endif
 
-static int eckey_param2type(int *pptype, void **ppval, EC_KEY *ec_key)
+static int eckey_param2type(int *pptype, void **ppval, const EC_KEY *ec_key)
 {
     const EC_GROUP *group;
     int nid;
@@ -35,7 +35,14 @@ static int eckey_param2type(int *pptype, void **ppval, EC_KEY *ec_key)
         && (nid = EC_GROUP_get_curve_name(group)))
         /* we have a 'named curve' => just set the OID */
     {
-        *ppval = OBJ_nid2obj(nid);
+        ASN1_OBJECT *asn1obj = OBJ_nid2obj(nid);
+
+        if (asn1obj == NULL || OBJ_length(asn1obj) == 0) {
+            ASN1_OBJECT_free(asn1obj);
+            ECerr(EC_F_ECKEY_PARAM2TYPE, EC_R_MISSING_OID);
+            return 0;
+        }
+        *ppval = asn1obj;
         *pptype = V_ASN1_OBJECT;
     } else {                    /* explicit parameters */
 
@@ -43,7 +50,17 @@ static int eckey_param2type(int *pptype, void **ppval, EC_KEY *ec_key)
         pstr = ASN1_STRING_new();
         if (pstr == NULL)
             return 0;
-        pstr->length = i2d_ECParameters(ec_key, &pstr->data);
+
+        /*
+         * The cast in the following line is intentional as the
+         * `i2d_ECParameters` signature can't be constified (see discussion at
+         * https://github.com/openssl/openssl/pull/9347 where related and
+         * required constification backports were rejected).
+         *
+         * This cast should be safe anyway, because we can expect
+         * `i2d_ECParameters()` to treat the first argument as if it was const.
+         */
+        pstr->length = i2d_ECParameters((EC_KEY *)ec_key, &pstr->data);
         if (pstr->length <= 0) {
             ASN1_STRING_free(pstr);
             ECerr(EC_F_ECKEY_PARAM2TYPE, ERR_R_EC_LIB);
@@ -57,7 +74,7 @@ static int eckey_param2type(int *pptype, void **ppval, EC_KEY *ec_key)
 
 static int eckey_pub_encode(X509_PUBKEY *pk, const EVP_PKEY *pkey)
 {
-    EC_KEY *ec_key = pkey->pkey.ec;
+    const EC_KEY *ec_key = pkey->pkey.ec;
     void *pval = NULL;
     int ptype;
     unsigned char *penc = NULL, *p;
diff --git a/crypto/ec/ec_asn1.c b/crypto/ec/ec_asn1.c
index 006f9a5dea17..7b7c75ce8443 100644
--- a/crypto/ec/ec_asn1.c
+++ b/crypto/ec/ec_asn1.c
@@ -137,6 +137,12 @@ struct ec_parameters_st {
     ASN1_INTEGER *cofactor;
 } /* ECPARAMETERS */ ;
 
+typedef enum {
+    ECPKPARAMETERS_TYPE_NAMED = 0,
+    ECPKPARAMETERS_TYPE_EXPLICIT,
+    ECPKPARAMETERS_TYPE_IMPLICIT
+} ecpk_parameters_type_t;
+
 struct ecpk_parameters_st {
     int type;
     union {
@@ -535,9 +541,10 @@ ECPKPARAMETERS *EC_GROUP_get_ecpkparameters(const EC_GROUP *group,
             return NULL;
         }
     } else {
-        if (ret->type == 0)
+        if (ret->type == ECPKPARAMETERS_TYPE_NAMED)
             ASN1_OBJECT_free(ret->value.named_curve);
-        else if (ret->type == 1 && ret->value.parameters)
+        else if (ret->type == ECPKPARAMETERS_TYPE_EXPLICIT
+                 && ret->value.parameters != NULL)
             ECPARAMETERS_free(ret->value.parameters);
     }
 
@@ -547,15 +554,22 @@ ECPKPARAMETERS *EC_GROUP_get_ecpkparameters(const EC_GROUP *group,
          */
         tmp = EC_GROUP_get_curve_name(group);
         if (tmp) {
-            ret->type = 0;
-            if ((ret->value.named_curve = OBJ_nid2obj(tmp)) == NULL)
+            ASN1_OBJECT *asn1obj = OBJ_nid2obj(tmp);
+
+            if (asn1obj == NULL || OBJ_length(asn1obj) == 0) {
+                ASN1_OBJECT_free(asn1obj);
+                ECerr(EC_F_EC_GROUP_GET_ECPKPARAMETERS, EC_R_MISSING_OID);
                 ok = 0;
+            } else {
+                ret->type = ECPKPARAMETERS_TYPE_NAMED;
+                ret->value.named_curve = asn1obj;
+            }
         } else
             /* we don't know the nid => ERROR */
             ok = 0;
     } else {
         /* use the ECPARAMETERS structure */
-        ret->type = 1;
+        ret->type = ECPKPARAMETERS_TYPE_EXPLICIT;
         if ((ret->value.parameters =
              EC_GROUP_get_ecparameters(group, NULL)) == NULL)
             ok = 0;
@@ -894,7 +908,8 @@ EC_GROUP *EC_GROUP_new_from_ecpkparameters(const ECPKPARAMETERS *params)
         return NULL;
     }
 
-    if (params->type == 0) {    /* the curve is given by an OID */
+    if (params->type == ECPKPARAMETERS_TYPE_NAMED) {
+        /* the curve is given by an OID */
         tmp = OBJ_obj2nid(params->value.named_curve);
         if ((ret = EC_GROUP_new_by_curve_name(tmp)) == NULL) {
             ECerr(EC_F_EC_GROUP_NEW_FROM_ECPKPARAMETERS,
@@ -902,15 +917,16 @@ EC_GROUP *EC_GROUP_new_from_ecpkparameters(const ECPKPARAMETERS *params)
             return NULL;
         }
         EC_GROUP_set_asn1_flag(ret, OPENSSL_EC_NAMED_CURVE);
-    } else if (params->type == 1) { /* the parameters are given by a
-                                     * ECPARAMETERS structure */
+    } else if (params->type == ECPKPARAMETERS_TYPE_EXPLICIT) {
+        /* the parameters are given by an ECPARAMETERS structure */
         ret = EC_GROUP_new_from_ecparameters(params->value.parameters);
         if (!ret) {
             ECerr(EC_F_EC_GROUP_NEW_FROM_ECPKPARAMETERS, ERR_R_EC_LIB);
             return NULL;
         }
         EC_GROUP_set_asn1_flag(ret, OPENSSL_EC_EXPLICIT_CURVE);
-    } else if (params->type == 2) { /* implicitlyCA */
+    } else if (params->type == ECPKPARAMETERS_TYPE_IMPLICIT) {
+        /* implicit parameters inherited from CA - unsupported */
         return NULL;
     } else {
         ECerr(EC_F_EC_GROUP_NEW_FROM_ECPKPARAMETERS, EC_R_ASN1_ERROR);
@@ -940,6 +956,9 @@ EC_GROUP *d2i_ECPKParameters(EC_GROUP **a, const unsigned char **in, long len)
         return NULL;
     }
 
+    if (params->type == ECPKPARAMETERS_TYPE_EXPLICIT)
+        group->decoded_from_explicit_params = 1;
+
     if (a) {
         EC_GROUP_free(*a);
         *a = group;
@@ -991,6 +1010,9 @@ EC_KEY *d2i_ECPrivateKey(EC_KEY **a, const unsigned char **in, long len)
     if (priv_key->parameters) {
         EC_GROUP_free(ret->group);
         ret->group = EC_GROUP_new_from_ecpkparameters(priv_key->parameters);
+        if (ret->group != NULL
+            && priv_key->parameters->type == ECPKPARAMETERS_TYPE_EXPLICIT)
+            ret->group->decoded_from_explicit_params = 1;
     }
 
     if (ret->group == NULL) {
diff --git a/crypto/ec/ec_err.c b/crypto/ec/ec_err.c
index ce3493823218..bfe74226503e 100644
--- a/crypto/ec/ec_err.c
+++ b/crypto/ec/ec_err.c
@@ -1,6 +1,6 @@
 /*
  * Generated by util/mkerr.pl DO NOT EDIT
- * Copyright 1995-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -341,6 +341,7 @@ static const ERR_STRING_DATA EC_str_reasons[] = {
     {ERR_PACK(ERR_LIB_EC, 0, EC_R_LADDER_POST_FAILURE), "ladder post failure"},
     {ERR_PACK(ERR_LIB_EC, 0, EC_R_LADDER_PRE_FAILURE), "ladder pre failure"},
     {ERR_PACK(ERR_LIB_EC, 0, EC_R_LADDER_STEP_FAILURE), "ladder step failure"},
+    {ERR_PACK(ERR_LIB_EC, 0, EC_R_MISSING_OID), "missing OID"},
     {ERR_PACK(ERR_LIB_EC, 0, EC_R_MISSING_PARAMETERS), "missing parameters"},
     {ERR_PACK(ERR_LIB_EC, 0, EC_R_MISSING_PRIVATE_KEY), "missing private key"},
     {ERR_PACK(ERR_LIB_EC, 0, EC_R_NEED_NEW_SETUP_VALUES),
diff --git a/crypto/ec/ec_key.c b/crypto/ec/ec_key.c
index 08aaac5d8a6f..23efbd015ca4 100644
--- a/crypto/ec/ec_key.c
+++ b/crypto/ec/ec_key.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2002-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2002-2020 The OpenSSL Project Authors. All Rights Reserved.
  * Copyright (c) 2002, Oracle and/or its affiliates. All rights reserved
  *
  * Licensed under the OpenSSL license (the "License").  You may not use
@@ -14,6 +14,7 @@
 #include "internal/refcount.h"
 #include <openssl/err.h>
 #include <openssl/engine.h>
+#include "crypto/bn.h"
 
 EC_KEY *EC_KEY_new(void)
 {
@@ -416,17 +417,86 @@ const BIGNUM *EC_KEY_get0_private_key(const EC_KEY *key)
 
 int EC_KEY_set_private_key(EC_KEY *key, const BIGNUM *priv_key)
 {
+    int fixed_top;
+    const BIGNUM *order = NULL;
+    BIGNUM *tmp_key = NULL;
+
     if (key->group == NULL || key->group->meth == NULL)
         return 0;
+
+    /*
+     * Not only should key->group be set, but it should also be in a valid
+     * fully initialized state.
+     *
+     * Specifically, to operate in constant time, we need that the group order
+     * is set, as we use its length as the fixed public size of any scalar used
+     * as an EC private key.
+     */
+    order = EC_GROUP_get0_order(key->group);
+    if (order == NULL || BN_is_zero(order))
+        return 0; /* This should never happen */
+
     if (key->group->meth->set_private != NULL
         && key->group->meth->set_private(key, priv_key) == 0)
         return 0;
     if (key->meth->set_private != NULL
         && key->meth->set_private(key, priv_key) == 0)
         return 0;
+
+    /*
+     * We should never leak the bit length of the secret scalar in the key,
+     * so we always set the `BN_FLG_CONSTTIME` flag on the internal `BIGNUM`
+     * holding the secret scalar.
+     *
+     * This is important also because `BN_dup()` (and `BN_copy()`) do not
+     * propagate the `BN_FLG_CONSTTIME` flag from the source `BIGNUM`, and
+     * this brings an extra risk of inadvertently losing the flag, even when
+     * the caller specifically set it.
+     *
+     * The propagation has been turned on and off a few times in the past
+     * years because in some conditions has shown unintended consequences in
+     * some code paths, so at the moment we can't fix this in the BN layer.
+     *
+     * In `EC_KEY_set_private_key()` we can work around the propagation by
+     * manually setting the flag after `BN_dup()` as we know for sure that
+     * inside the EC module the `BN_FLG_CONSTTIME` is always treated
+     * correctly and should not generate unintended consequences.
+     *
+     * Setting the BN_FLG_CONSTTIME flag alone is never enough, we also have
+     * to preallocate the BIGNUM internal buffer to a fixed public size big
+     * enough that operations performed during the processing never trigger
+     * a realloc which would leak the size of the scalar through memory
+     * accesses.
+     *
+     * Fixed Length
+     * ------------
+     *
+     * The order of the large prime subgroup of the curve is our choice for
+     * a fixed public size, as that is generally the upper bound for
+     * generating a private key in EC cryptosystems and should fit all valid
+     * secret scalars.
+     *
+     * For preallocating the BIGNUM storage we look at the number of "words"
+     * required for the internal representation of the order, and we
+     * preallocate 2 extra "words" in case any of the subsequent processing
+     * might temporarily overflow the order length.
+     */
+    tmp_key = BN_dup(priv_key);
+    if (tmp_key == NULL)
+        return 0;
+
+    BN_set_flags(tmp_key, BN_FLG_CONSTTIME);
+
+    fixed_top = bn_get_top(order) + 2;
+    if (bn_wexpand(tmp_key, fixed_top) == NULL) {
+        BN_clear_free(tmp_key);
+        return 0;
+    }
+
     BN_clear_free(key->priv_key);
-    key->priv_key = BN_dup(priv_key);
-    return (key->priv_key == NULL) ? 0 : 1;
+    key->priv_key = tmp_key;
+
+    return 1;
 }
 
 const EC_POINT *EC_KEY_get0_public_key(const EC_KEY *key)
@@ -494,6 +564,13 @@ void EC_KEY_clear_flags(EC_KEY *key, int flags)
     key->flags &= ~flags;
 }
 
+int EC_KEY_decoded_from_explicit_params(const EC_KEY *key)
+{
+    if (key == NULL || key->group == NULL)
+        return -1;
+    return key->group->decoded_from_explicit_params;
+}
+
 size_t EC_KEY_key2buf(const EC_KEY *key, point_conversion_form_t form,
                         unsigned char **pbuf, BN_CTX *ctx)
 {
diff --git a/crypto/ec/ec_lib.c b/crypto/ec/ec_lib.c
index 6832383cad51..08db89fceeb5 100644
--- a/crypto/ec/ec_lib.c
+++ b/crypto/ec/ec_lib.c
@@ -211,6 +211,7 @@ int EC_GROUP_copy(EC_GROUP *dest, const EC_GROUP *src)
 
     dest->asn1_flag = src->asn1_flag;
     dest->asn1_form = src->asn1_form;
+    dest->decoded_from_explicit_params = src->decoded_from_explicit_params;
 
     if (src->seed) {
         OPENSSL_free(dest->seed);
diff --git a/crypto/ec/ec_local.h b/crypto/ec/ec_local.h
index e656fbd5e775..64725a9c92f4 100644
--- a/crypto/ec/ec_local.h
+++ b/crypto/ec/ec_local.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2001-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2001-2020 The OpenSSL Project Authors. All Rights Reserved.
  * Copyright (c) 2002, Oracle and/or its affiliates. All rights reserved
  *
  * Licensed under the OpenSSL license (the "License").  You may not use
@@ -209,6 +209,8 @@ struct ec_group_st {
     BIGNUM *order, *cofactor;
     int curve_name;             /* optional NID for named curve */
     int asn1_flag;              /* flag to control the asn1 encoding */
+    int decoded_from_explicit_params; /* set if decoded from explicit
+                                       * curve parameters encoding */
     point_conversion_form_t asn1_form;
     unsigned char *seed;        /* optional seed for parameters (appears in
                                  * ASN1) */
diff --git a/crypto/ec/ecp_nistp224.c b/crypto/ec/ecp_nistp224.c
index 9a9ced8f1343..6f7d66c8bea4 100644
--- a/crypto/ec/ecp_nistp224.c
+++ b/crypto/ec/ecp_nistp224.c
@@ -72,6 +72,7 @@ typedef uint64_t u64;
  */
 
 typedef uint64_t limb;
+typedef uint64_t limb_aX __attribute((__aligned__(1)));
 typedef uint128_t widelimb;
 
 typedef limb felem[4];
@@ -307,10 +308,10 @@ const EC_METHOD *EC_GFp_nistp224_method(void)
  */
 static void bin28_to_felem(felem out, const u8 in[28])
 {
-    out[0] = *((const uint64_t *)(in)) & 0x00ffffffffffffff;
-    out[1] = (*((const uint64_t *)(in + 7))) & 0x00ffffffffffffff;
-    out[2] = (*((const uint64_t *)(in + 14))) & 0x00ffffffffffffff;
-    out[3] = (*((const uint64_t *)(in+20))) >> 8;
+    out[0] = *((const limb *)(in)) & 0x00ffffffffffffff;
+    out[1] = (*((const limb_aX *)(in + 7))) & 0x00ffffffffffffff;
+    out[2] = (*((const limb_aX *)(in + 14))) & 0x00ffffffffffffff;
+    out[3] = (*((const limb_aX *)(in + 20))) >> 8;
 }
 
 static void felem_to_bin28(u8 out[28], const felem in)
diff --git a/crypto/ec/ecp_nistp521.c b/crypto/ec/ecp_nistp521.c
index 75eeba853679..08b32787293b 100644
--- a/crypto/ec/ecp_nistp521.c
+++ b/crypto/ec/ecp_nistp521.c
@@ -128,6 +128,7 @@ static const felem_bytearray nistp521_curve_params[5] = {
 # define NLIMBS 9
 
 typedef uint64_t limb;
+typedef limb limb_aX __attribute((__aligned__(1)));
 typedef limb felem[NLIMBS];
 typedef uint128_t largefelem[NLIMBS];
 
@@ -141,14 +142,14 @@ static const limb bottom58bits = 0x3ffffffffffffff;
 static void bin66_to_felem(felem out, const u8 in[66])
 {
     out[0] = (*((limb *) & in[0])) & bottom58bits;
-    out[1] = (*((limb *) & in[7]) >> 2) & bottom58bits;
-    out[2] = (*((limb *) & in[14]) >> 4) & bottom58bits;
-    out[3] = (*((limb *) & in[21]) >> 6) & bottom58bits;
-    out[4] = (*((limb *) & in[29])) & bottom58bits;
-    out[5] = (*((limb *) & in[36]) >> 2) & bottom58bits;
-    out[6] = (*((limb *) & in[43]) >> 4) & bottom58bits;
-    out[7] = (*((limb *) & in[50]) >> 6) & bottom58bits;
-    out[8] = (*((limb *) & in[58])) & bottom57bits;
+    out[1] = (*((limb_aX *) & in[7]) >> 2) & bottom58bits;
+    out[2] = (*((limb_aX *) & in[14]) >> 4) & bottom58bits;
+    out[3] = (*((limb_aX *) & in[21]) >> 6) & bottom58bits;
+    out[4] = (*((limb_aX *) & in[29])) & bottom58bits;
+    out[5] = (*((limb_aX *) & in[36]) >> 2) & bottom58bits;
+    out[6] = (*((limb_aX *) & in[43]) >> 4) & bottom58bits;
+    out[7] = (*((limb_aX *) & in[50]) >> 6) & bottom58bits;
+    out[8] = (*((limb_aX *) & in[58])) & bottom57bits;
 }
 
 /*
@@ -159,14 +160,14 @@ static void felem_to_bin66(u8 out[66], const felem in)
 {
     memset(out, 0, 66);
     (*((limb *) & out[0])) = in[0];
-    (*((limb *) & out[7])) |= in[1] << 2;
-    (*((limb *) & out[14])) |= in[2] << 4;
-    (*((limb *) & out[21])) |= in[3] << 6;
-    (*((limb *) & out[29])) = in[4];
-    (*((limb *) & out[36])) |= in[5] << 2;
-    (*((limb *) & out[43])) |= in[6] << 4;
-    (*((limb *) & out[50])) |= in[7] << 6;
-    (*((limb *) & out[58])) = in[8];
+    (*((limb_aX *) & out[7])) |= in[1] << 2;
+    (*((limb_aX *) & out[14])) |= in[2] << 4;
+    (*((limb_aX *) & out[21])) |= in[3] << 6;
+    (*((limb_aX *) & out[29])) = in[4];
+    (*((limb_aX *) & out[36])) |= in[5] << 2;
+    (*((limb_aX *) & out[43])) |= in[6] << 4;
+    (*((limb_aX *) & out[50])) |= in[7] << 6;
+    (*((limb_aX *) & out[58])) = in[8];
 }
 
 /* BN_to_felem converts an OpenSSL BIGNUM into an felem */
diff --git a/crypto/ec/ecp_nistz256.c b/crypto/ec/ecp_nistz256.c
index ba9268138862..5005249b05ea 100644
--- a/crypto/ec/ecp_nistz256.c
+++ b/crypto/ec/ecp_nistz256.c
@@ -929,207 +929,6 @@ __owur static int ecp_nistz256_mult_precompute(EC_GROUP *group, BN_CTX *ctx)
     return ret;
 }
 
-/*
- * Note that by default ECP_NISTZ256_AVX2 is undefined. While it's great
- * code processing 4 points in parallel, corresponding serial operation
- * is several times slower, because it uses 29x29=58-bit multiplication
- * as opposite to 64x64=128-bit in integer-only scalar case. As result
- * it doesn't provide *significant* performance improvement. Note that
- * just defining ECP_NISTZ256_AVX2 is not sufficient to make it work,
- * you'd need to compile even asm/ecp_nistz256-avx.pl module.
- */
-#if defined(ECP_NISTZ256_AVX2)
-# if !(defined(__x86_64) || defined(__x86_64__) || \
-       defined(_M_AMD64) || defined(_M_X64)) || \
-     !(defined(__GNUC__) || defined(_MSC_VER)) /* this is for ALIGN32 */
-#  undef ECP_NISTZ256_AVX2
-# else
-/* Constant time access, loading four values, from four consecutive tables */
-void ecp_nistz256_avx2_multi_gather_w7(void *result, const void *in,
-                                       int index0, int index1, int index2,
-                                       int index3);
-void ecp_nistz256_avx2_transpose_convert(void *RESULTx4, const void *in);
-void ecp_nistz256_avx2_convert_transpose_back(void *result, const void *Ax4);
-void ecp_nistz256_avx2_point_add_affine_x4(void *RESULTx4, const void *Ax4,
-                                           const void *Bx4);
-void ecp_nistz256_avx2_point_add_affines_x4(void *RESULTx4, const void *Ax4,
-                                            const void *Bx4);
-void ecp_nistz256_avx2_to_mont(void *RESULTx4, const void *Ax4);
-void ecp_nistz256_avx2_from_mont(void *RESULTx4, const void *Ax4);
-void ecp_nistz256_avx2_set1(void *RESULTx4);
-int ecp_nistz_avx2_eligible(void);
-
-static void booth_recode_w7(unsigned char *sign,
-                            unsigned char *digit, unsigned char in)
-{
-    unsigned char s, d;
-
-    s = ~((in >> 7) - 1);
-    d = (1 << 8) - in - 1;
-    d = (d & s) | (in & ~s);
-    d = (d >> 1) + (d & 1);
-
-    *sign = s & 1;
-    *digit = d;
-}
-
-/*
- * ecp_nistz256_avx2_mul_g performs multiplication by G, using only the
- * precomputed table. It does 4 affine point additions in parallel,
- * significantly speeding up point multiplication for a fixed value.
- */
-static void ecp_nistz256_avx2_mul_g(P256_POINT *r,
-                                    unsigned char p_str[33],
-                                    const P256_POINT_AFFINE(*preComputedTable)[64])
-{
-    const unsigned int window_size = 7;
-    const unsigned int mask = (1 << (window_size + 1)) - 1;
-    unsigned int wvalue;
-    /* Using 4 windows at a time */
-    unsigned char sign0, digit0;
-    unsigned char sign1, digit1;
-    unsigned char sign2, digit2;
-    unsigned char sign3, digit3;
-    unsigned int idx = 0;
-    BN_ULONG tmp[P256_LIMBS];
-    int i;
-
-    ALIGN32 BN_ULONG aX4[4 * 9 * 3] = { 0 };
-    ALIGN32 BN_ULONG bX4[4 * 9 * 2] = { 0 };
-    ALIGN32 P256_POINT_AFFINE point_arr[4];
-    ALIGN32 P256_POINT res_point_arr[4];
-
-    /* Initial four windows */
-    wvalue = *((u16 *) & p_str[0]);
-    wvalue = (wvalue << 1) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign0, &digit0, wvalue);
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign1, &digit1, wvalue);
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign2, &digit2, wvalue);
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign3, &digit3, wvalue);
-
-    ecp_nistz256_avx2_multi_gather_w7(point_arr, preComputedTable[0],
-                                      digit0, digit1, digit2, digit3);
-
-    ecp_nistz256_neg(tmp, point_arr[0].Y);
-    copy_conditional(point_arr[0].Y, tmp, sign0);
-    ecp_nistz256_neg(tmp, point_arr[1].Y);
-    copy_conditional(point_arr[1].Y, tmp, sign1);
-    ecp_nistz256_neg(tmp, point_arr[2].Y);
-    copy_conditional(point_arr[2].Y, tmp, sign2);
-    ecp_nistz256_neg(tmp, point_arr[3].Y);
-    copy_conditional(point_arr[3].Y, tmp, sign3);
-
-    ecp_nistz256_avx2_transpose_convert(aX4, point_arr);
-    ecp_nistz256_avx2_to_mont(aX4, aX4);
-    ecp_nistz256_avx2_to_mont(&aX4[4 * 9], &aX4[4 * 9]);
-    ecp_nistz256_avx2_set1(&aX4[4 * 9 * 2]);
-
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign0, &digit0, wvalue);
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign1, &digit1, wvalue);
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign2, &digit2, wvalue);
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    idx += window_size;
-    booth_recode_w7(&sign3, &digit3, wvalue);
-
-    ecp_nistz256_avx2_multi_gather_w7(point_arr, preComputedTable[4 * 1],
-                                      digit0, digit1, digit2, digit3);
-
-    ecp_nistz256_neg(tmp, point_arr[0].Y);
-    copy_conditional(point_arr[0].Y, tmp, sign0);
-    ecp_nistz256_neg(tmp, point_arr[1].Y);
-    copy_conditional(point_arr[1].Y, tmp, sign1);
-    ecp_nistz256_neg(tmp, point_arr[2].Y);
-    copy_conditional(point_arr[2].Y, tmp, sign2);
-    ecp_nistz256_neg(tmp, point_arr[3].Y);
-    copy_conditional(point_arr[3].Y, tmp, sign3);
-
-    ecp_nistz256_avx2_transpose_convert(bX4, point_arr);
-    ecp_nistz256_avx2_to_mont(bX4, bX4);
-    ecp_nistz256_avx2_to_mont(&bX4[4 * 9], &bX4[4 * 9]);
-    /* Optimized when both inputs are affine */
-    ecp_nistz256_avx2_point_add_affines_x4(aX4, aX4, bX4);
-
-    for (i = 2; i < 9; i++) {
-        wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-        wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-        idx += window_size;
-        booth_recode_w7(&sign0, &digit0, wvalue);
-        wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-        wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-        idx += window_size;
-        booth_recode_w7(&sign1, &digit1, wvalue);
-        wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-        wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-        idx += window_size;
-        booth_recode_w7(&sign2, &digit2, wvalue);
-        wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-        wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-        idx += window_size;
-        booth_recode_w7(&sign3, &digit3, wvalue);
-
-        ecp_nistz256_avx2_multi_gather_w7(point_arr,
-                                          preComputedTable[4 * i],
-                                          digit0, digit1, digit2, digit3);
-
-        ecp_nistz256_neg(tmp, point_arr[0].Y);
-        copy_conditional(point_arr[0].Y, tmp, sign0);
-        ecp_nistz256_neg(tmp, point_arr[1].Y);
-        copy_conditional(point_arr[1].Y, tmp, sign1);
-        ecp_nistz256_neg(tmp, point_arr[2].Y);
-        copy_conditional(point_arr[2].Y, tmp, sign2);
-        ecp_nistz256_neg(tmp, point_arr[3].Y);
-        copy_conditional(point_arr[3].Y, tmp, sign3);
-
-        ecp_nistz256_avx2_transpose_convert(bX4, point_arr);
-        ecp_nistz256_avx2_to_mont(bX4, bX4);
-        ecp_nistz256_avx2_to_mont(&bX4[4 * 9], &bX4[4 * 9]);
-
-        ecp_nistz256_avx2_point_add_affine_x4(aX4, aX4, bX4);
-    }
-
-    ecp_nistz256_avx2_from_mont(&aX4[4 * 9 * 0], &aX4[4 * 9 * 0]);
-    ecp_nistz256_avx2_from_mont(&aX4[4 * 9 * 1], &aX4[4 * 9 * 1]);
-    ecp_nistz256_avx2_from_mont(&aX4[4 * 9 * 2], &aX4[4 * 9 * 2]);
-
-    ecp_nistz256_avx2_convert_transpose_back(res_point_arr, aX4);
-    /* Last window is performed serially */
-    wvalue = *((u16 *) & p_str[(idx - 1) / 8]);
-    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-    booth_recode_w7(&sign0, &digit0, wvalue);
-    ecp_nistz256_gather_w7((P256_POINT_AFFINE *)r,
-                           preComputedTable[36], digit0);
-    ecp_nistz256_neg(tmp, r->Y);
-    copy_conditional(r->Y, tmp, sign0);
-    memcpy(r->Z, ONE, sizeof(ONE));
-    /* Sum the four windows */
-    ecp_nistz256_point_add(r, r, &res_point_arr[0]);
-    ecp_nistz256_point_add(r, r, &res_point_arr[1]);
-    ecp_nistz256_point_add(r, r, &res_point_arr[2]);
-    ecp_nistz256_point_add(r, r, &res_point_arr[3]);
-}
-# endif
-#endif
-
 __owur static int ecp_nistz256_set_from_affine(EC_POINT *out, const EC_GROUP *group,
                                                const P256_POINT_AFFINE *in,
                                                BN_CTX *ctx)
@@ -1219,6 +1018,8 @@ __owur static int ecp_nistz256_points_mul(const EC_GROUP *group,
         }
 
         if (preComputedTable) {
+            BN_ULONG infty;
+
             if ((BN_num_bits(scalar) > 256)
                 || BN_is_negative(scalar)) {
                 if ((tmp_scalar = BN_CTX_get(ctx)) == NULL)
@@ -1250,67 +1051,58 @@ __owur static int ecp_nistz256_points_mul(const EC_GROUP *group,
             for (; i < 33; i++)
                 p_str[i] = 0;
 
-#if defined(ECP_NISTZ256_AVX2)
-            if (ecp_nistz_avx2_eligible()) {
-                ecp_nistz256_avx2_mul_g(&p.p, p_str, preComputedTable);
-            } else
-#endif
-            {
-                BN_ULONG infty;
+            /* First window */
+            wvalue = (p_str[0] << 1) & mask;
+            idx += window_size;
 
-                /* First window */
-                wvalue = (p_str[0] << 1) & mask;
-                idx += window_size;
+            wvalue = _booth_recode_w7(wvalue);
 
-                wvalue = _booth_recode_w7(wvalue);
+            ecp_nistz256_gather_w7(&p.a, preComputedTable[0],
+                                   wvalue >> 1);
 
-                ecp_nistz256_gather_w7(&p.a, preComputedTable[0],
-                                       wvalue >> 1);
-
-                ecp_nistz256_neg(p.p.Z, p.p.Y);
-                copy_conditional(p.p.Y, p.p.Z, wvalue & 1);
-
-                /*
-                 * Since affine infinity is encoded as (0,0) and
-                 * Jacobian ias (,,0), we need to harmonize them
-                 * by assigning "one" or zero to Z.
-                 */
-                infty = (p.p.X[0] | p.p.X[1] | p.p.X[2] | p.p.X[3] |
-                         p.p.Y[0] | p.p.Y[1] | p.p.Y[2] | p.p.Y[3]);
-                if (P256_LIMBS == 8)
-                    infty |= (p.p.X[4] | p.p.X[5] | p.p.X[6] | p.p.X[7] |
-                              p.p.Y[4] | p.p.Y[5] | p.p.Y[6] | p.p.Y[7]);
-
-                infty = 0 - is_zero(infty);
-                infty = ~infty;
-
-                p.p.Z[0] = ONE[0] & infty;
-                p.p.Z[1] = ONE[1] & infty;
-                p.p.Z[2] = ONE[2] & infty;
-                p.p.Z[3] = ONE[3] & infty;
-                if (P256_LIMBS == 8) {
-                    p.p.Z[4] = ONE[4] & infty;
-                    p.p.Z[5] = ONE[5] & infty;
-                    p.p.Z[6] = ONE[6] & infty;
-                    p.p.Z[7] = ONE[7] & infty;
-                }
+            ecp_nistz256_neg(p.p.Z, p.p.Y);
+            copy_conditional(p.p.Y, p.p.Z, wvalue & 1);
 
-                for (i = 1; i < 37; i++) {
-                    unsigned int off = (idx - 1) / 8;
-                    wvalue = p_str[off] | p_str[off + 1] << 8;
-                    wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
-                    idx += window_size;
+            /*
+             * Since affine infinity is encoded as (0,0) and
+             * Jacobian is (,,0), we need to harmonize them
+             * by assigning "one" or zero to Z.
+             */
+            infty = (p.p.X[0] | p.p.X[1] | p.p.X[2] | p.p.X[3] |
+                     p.p.Y[0] | p.p.Y[1] | p.p.Y[2] | p.p.Y[3]);
+            if (P256_LIMBS == 8)
+                infty |= (p.p.X[4] | p.p.X[5] | p.p.X[6] | p.p.X[7] |
+                          p.p.Y[4] | p.p.Y[5] | p.p.Y[6] | p.p.Y[7]);
+
+            infty = 0 - is_zero(infty);
+            infty = ~infty;
+
+            p.p.Z[0] = ONE[0] & infty;
+            p.p.Z[1] = ONE[1] & infty;
+            p.p.Z[2] = ONE[2] & infty;
+            p.p.Z[3] = ONE[3] & infty;
+            if (P256_LIMBS == 8) {
+                p.p.Z[4] = ONE[4] & infty;
+                p.p.Z[5] = ONE[5] & infty;
+                p.p.Z[6] = ONE[6] & infty;
+                p.p.Z[7] = ONE[7] & infty;
+            }
 
-                    wvalue = _booth_recode_w7(wvalue);
+            for (i = 1; i < 37; i++) {
+                unsigned int off = (idx - 1) / 8;
+                wvalue = p_str[off] | p_str[off + 1] << 8;
+                wvalue = (wvalue >> ((idx - 1) % 8)) & mask;
+                idx += window_size;
 
-                    ecp_nistz256_gather_w7(&t.a,
-                                           preComputedTable[i], wvalue >> 1);
+                wvalue = _booth_recode_w7(wvalue);
 
-                    ecp_nistz256_neg(t.p.Z, t.a.Y);
-                    copy_conditional(t.a.Y, t.p.Z, wvalue & 1);
+                ecp_nistz256_gather_w7(&t.a,
+                                       preComputedTable[i], wvalue >> 1);
 
-                    ecp_nistz256_point_add_affine(&p.p, &p.p, &t.a);
-                }
+                ecp_nistz256_neg(t.p.Z, t.a.Y);
+                copy_conditional(t.a.Y, t.p.Z, wvalue & 1);
+
+                ecp_nistz256_point_add_affine(&p.p, &p.p, &t.a);
             }
         } else {
             p_is_infinity = 1;
diff --git a/crypto/engine/eng_lib.c b/crypto/engine/eng_lib.c
index b851ff695756..5bd584c5999a 100644
--- a/crypto/engine/eng_lib.c
+++ b/crypto/engine/eng_lib.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2001-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2001-2020 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
@@ -171,6 +171,7 @@ void engine_cleanup_int(void)
         cleanup_stack = NULL;
     }
     CRYPTO_THREAD_lock_free(global_engine_lock);
+    global_engine_lock = NULL;
 }
 
 /* Now the "ex_data" support */
diff --git a/crypto/err/openssl.txt b/crypto/err/openssl.txt
index 35512f9caf96..0b5873ebbcb7 100644
--- a/crypto/err/openssl.txt
+++ b/crypto/err/openssl.txt
@@ -934,6 +934,8 @@ PEM_F_PEM_READ_PRIVATEKEY:124:PEM_read_PrivateKey
 PEM_F_PEM_SIGNFINAL:112:PEM_SignFinal
 PEM_F_PEM_WRITE:113:PEM_write
 PEM_F_PEM_WRITE_BIO:114:PEM_write_bio
+PEM_F_PEM_WRITE_BIO_PRIVATEKEY_TRADITIONAL:147:\
+	PEM_write_bio_PrivateKey_traditional
 PEM_F_PEM_WRITE_PRIVATEKEY:139:PEM_write_PrivateKey
 PEM_F_PEM_X509_INFO_READ:115:PEM_X509_INFO_read
 PEM_F_PEM_X509_INFO_READ_BIO:116:PEM_X509_INFO_read_bio
@@ -1742,6 +1744,7 @@ X509_F_X509_NAME_PRINT:117:X509_NAME_print
 X509_F_X509_OBJECT_NEW:150:X509_OBJECT_new
 X509_F_X509_PRINT_EX_FP:118:X509_print_ex_fp
 X509_F_X509_PUBKEY_DECODE:148:x509_pubkey_decode
+X509_F_X509_PUBKEY_GET:161:X509_PUBKEY_get
 X509_F_X509_PUBKEY_GET0:119:X509_PUBKEY_get0
 X509_F_X509_PUBKEY_SET:120:X509_PUBKEY_set
 X509_F_X509_REQ_CHECK_PRIVATE_KEY:144:X509_REQ_check_private_key
@@ -2164,6 +2167,7 @@ EC_R_KEYS_NOT_SET:140:keys not set
 EC_R_LADDER_POST_FAILURE:136:ladder post failure
 EC_R_LADDER_PRE_FAILURE:153:ladder pre failure
 EC_R_LADDER_STEP_FAILURE:162:ladder step failure
+EC_R_MISSING_OID:167:missing OID
 EC_R_MISSING_PARAMETERS:124:missing parameters
 EC_R_MISSING_PRIVATE_KEY:125:missing private key
 EC_R_NEED_NEW_SETUP_VALUES:157:need new setup values
@@ -2398,6 +2402,7 @@ PEM_R_UNEXPECTED_DEK_IV:130:unexpected dek iv
 PEM_R_UNSUPPORTED_CIPHER:113:unsupported cipher
 PEM_R_UNSUPPORTED_ENCRYPTION:114:unsupported encryption
 PEM_R_UNSUPPORTED_KEY_COMPONENTS:126:unsupported key components
+PEM_R_UNSUPPORTED_PUBLIC_KEY_TYPE:110:unsupported public key type
 PKCS12_R_CANT_PACK_STRUCTURE:100:cant pack structure
 PKCS12_R_CONTENT_TYPE_NOT_DATA:121:content type not data
 PKCS12_R_DECODE_ERROR:101:decode error
diff --git a/crypto/evp/e_aes.c b/crypto/evp/e_aes.c
index a1b7d50bbff8..405ddbf9bf09 100644
--- a/crypto/evp/e_aes.c
+++ b/crypto/evp/e_aes.c
@@ -130,11 +130,6 @@ void bsaes_xts_decrypt(const unsigned char *inp, unsigned char *out,
                        size_t len, const AES_KEY *key1,
                        const AES_KEY *key2, const unsigned char iv[16]);
 #endif
-#if !defined(AES_ASM) && !defined(AES_CTR_ASM)			\
-	&& defined(OPENSSL_AES_CONST_TIME)			\
-	&& !defined(OPENSSL_SMALL_FOOTPRINT)
-# define AES_CTR_ASM
-#endif
 #ifdef AES_CTR_ASM
 void AES_ctr32_encrypt(const unsigned char *in, unsigned char *out,
                        size_t blocks, const AES_KEY *key,
diff --git a/crypto/evp/encode.c b/crypto/evp/encode.c
index 9307ff046424..85926434c300 100644
--- a/crypto/evp/encode.c
+++ b/crypto/evp/encode.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -423,7 +423,7 @@ static int evp_decodeblock_int(EVP_ENCODE_CTX *ctx, unsigned char *t,
         table = data_ascii2bin;
 
     /* trim white space from the start of the line. */
-    while ((conv_ascii2bin(*f, table) == B64_WS) && (n > 0)) {
+    while ((n > 0) && (conv_ascii2bin(*f, table) == B64_WS)) {
         f++;
         n--;
     }
diff --git a/crypto/mem_sec.c b/crypto/mem_sec.c
index 9e0f6702f406..b5f959ba15d5 100644
--- a/crypto/mem_sec.c
+++ b/crypto/mem_sec.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2015-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2015-2020 The OpenSSL Project Authors. All Rights Reserved.
  * Copyright 2004-2014, Akamai Technologies. All Rights Reserved.
  *
  * Licensed under the OpenSSL license (the "License").  You may not use
@@ -502,7 +502,7 @@ static void sh_done(void)
     OPENSSL_free(sh.freelist);
     OPENSSL_free(sh.bittable);
     OPENSSL_free(sh.bitmalloc);
-    if (sh.map_result != NULL && sh.map_size)
+    if (sh.map_result != MAP_FAILED && sh.map_size)
         munmap(sh.map_result, sh.map_size);
     memset(&sh, 0, sizeof(sh));
 }
diff --git a/crypto/modes/asm/aesni-gcm-x86_64.pl b/crypto/modes/asm/aesni-gcm-x86_64.pl
index 959efedb0de7..60f03e4fe25b 100755
--- a/crypto/modes/asm/aesni-gcm-x86_64.pl
+++ b/crypto/modes/asm/aesni-gcm-x86_64.pl
@@ -66,7 +66,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/modes/asm/ghash-x86_64.pl b/crypto/modes/asm/ghash-x86_64.pl
index 0a0bfd575cee..9bdba41d1de5 100755
--- a/crypto/modes/asm/ghash-x86_64.pl
+++ b/crypto/modes/asm/ghash-x86_64.pl
@@ -116,7 +116,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/modes/cbc128.c b/crypto/modes/cbc128.c
index fc7e0b60510b..c85e37c6a546 100644
--- a/crypto/modes/cbc128.c
+++ b/crypto/modes/cbc128.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2008-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2008-2020 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
@@ -15,6 +15,12 @@
 # define STRICT_ALIGNMENT 0
 #endif
 
+#if defined(__GNUC__) && !STRICT_ALIGNMENT
+typedef size_t size_t_aX __attribute((__aligned__(1)));
+#else
+typedef size_t size_t_aX;
+#endif
+
 void CRYPTO_cbc128_encrypt(const unsigned char *in, unsigned char *out,
                            size_t len, const void *key,
                            unsigned char ivec[16], block128_f block)
@@ -40,8 +46,8 @@ void CRYPTO_cbc128_encrypt(const unsigned char *in, unsigned char *out,
     } else {
         while (len >= 16) {
             for (n = 0; n < 16; n += sizeof(size_t))
-                *(size_t *)(out + n) =
-                    *(size_t *)(in + n) ^ *(size_t *)(iv + n);
+                *(size_t_aX *)(out + n) =
+                    *(size_t_aX *)(in + n) ^ *(size_t_aX *)(iv + n);
             (*block) (out, out, key);
             iv = out;
             len -= 16;
@@ -96,7 +102,8 @@ void CRYPTO_cbc128_decrypt(const unsigned char *in, unsigned char *out,
             }
         } else if (16 % sizeof(size_t) == 0) { /* always true */
             while (len >= 16) {
-                size_t *out_t = (size_t *)out, *iv_t = (size_t *)iv;
+                size_t_aX *out_t = (size_t_aX *)out;
+                size_t_aX *iv_t = (size_t_aX *)iv;
 
                 (*block) (in, out, key);
                 for (n = 0; n < 16 / sizeof(size_t); n++)
@@ -125,8 +132,10 @@ void CRYPTO_cbc128_decrypt(const unsigned char *in, unsigned char *out,
             }
         } else if (16 % sizeof(size_t) == 0) { /* always true */
             while (len >= 16) {
-                size_t c, *out_t = (size_t *)out, *ivec_t = (size_t *)ivec;
-                const size_t *in_t = (const size_t *)in;
+                size_t c;
+                size_t_aX *out_t = (size_t_aX *)out;
+                size_t_aX *ivec_t = (size_t_aX *)ivec;
+                const size_t_aX *in_t = (const size_t_aX *)in;
 
                 (*block) (in, tmp.c, key);
                 for (n = 0; n < 16 / sizeof(size_t); n++) {
diff --git a/crypto/modes/ccm128.c b/crypto/modes/ccm128.c
index 424722811c16..655b10350201 100644
--- a/crypto/modes/ccm128.c
+++ b/crypto/modes/ccm128.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2011-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2011-2020 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
@@ -11,6 +11,14 @@
 #include "modes_local.h"
 #include <string.h>
 
+#ifndef STRICT_ALIGNMENT
+# ifdef __GNUC__
+typedef u64 u64_a1 __attribute((__aligned__(1)));
+# else
+typedef u64 u64_a1;
+# endif
+#endif
+
 /*
  * First you setup M and L parameters and pass the key schedule. This is
  * called once per session setup...
@@ -170,8 +178,8 @@ int CRYPTO_ccm128_encrypt(CCM128_CONTEXT *ctx,
         ctx->cmac.u[0] ^= temp.u[0];
         ctx->cmac.u[1] ^= temp.u[1];
 #else
-        ctx->cmac.u[0] ^= ((u64 *)inp)[0];
-        ctx->cmac.u[1] ^= ((u64 *)inp)[1];
+        ctx->cmac.u[0] ^= ((u64_a1 *)inp)[0];
+        ctx->cmac.u[1] ^= ((u64_a1 *)inp)[1];
 #endif
         (*block) (ctx->cmac.c, ctx->cmac.c, key);
         (*block) (ctx->nonce.c, scratch.c, key);
@@ -181,8 +189,8 @@ int CRYPTO_ccm128_encrypt(CCM128_CONTEXT *ctx,
         temp.u[1] ^= scratch.u[1];
         memcpy(out, temp.c, 16);
 #else
-        ((u64 *)out)[0] = scratch.u[0] ^ ((u64 *)inp)[0];
-        ((u64 *)out)[1] = scratch.u[1] ^ ((u64 *)inp)[1];
+        ((u64_a1 *)out)[0] = scratch.u[0] ^ ((u64_a1 *)inp)[0];
+        ((u64_a1 *)out)[1] = scratch.u[1] ^ ((u64_a1 *)inp)[1];
 #endif
         inp += 16;
         out += 16;
@@ -254,8 +262,10 @@ int CRYPTO_ccm128_decrypt(CCM128_CONTEXT *ctx,
         ctx->cmac.u[1] ^= (scratch.u[1] ^= temp.u[1]);
         memcpy(out, scratch.c, 16);
 #else
-        ctx->cmac.u[0] ^= (((u64 *)out)[0] = scratch.u[0] ^ ((u64 *)inp)[0]);
-        ctx->cmac.u[1] ^= (((u64 *)out)[1] = scratch.u[1] ^ ((u64 *)inp)[1]);
+        ctx->cmac.u[0] ^= (((u64_a1 *)out)[0]
+                            = scratch.u[0] ^ ((u64_a1 *)inp)[0]);
+        ctx->cmac.u[1] ^= (((u64_a1 *)out)[1]
+                            = scratch.u[1] ^ ((u64_a1 *)inp)[1]);
 #endif
         (*block) (ctx->cmac.c, ctx->cmac.c, key);
 
diff --git a/crypto/modes/cfb128.c b/crypto/modes/cfb128.c
index b6bec414a966..b2530007b6e4 100644
--- a/crypto/modes/cfb128.c
+++ b/crypto/modes/cfb128.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2008-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2008-2020 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
@@ -11,6 +11,12 @@
 #include "modes_local.h"
 #include <string.h>
 
+#if defined(__GNUC__) && !defined(STRICT_ALIGNMENT)
+typedef size_t size_t_aX __attribute((__aligned__(1)));
+#else
+typedef size_t size_t_aX;
+#endif
+
 /*
  * The input and output encrypted as though 128bit cfb mode is being used.
  * The extra state information to record how much of the 128bit block we have
@@ -43,8 +49,9 @@ void CRYPTO_cfb128_encrypt(const unsigned char *in, unsigned char *out,
                 while (len >= 16) {
                     (*block) (ivec, ivec, key);
                     for (; n < 16; n += sizeof(size_t)) {
-                        *(size_t *)(out + n) =
-                            *(size_t *)(ivec + n) ^= *(size_t *)(in + n);
+                        *(size_t_aX *)(out + n) =
+                            *(size_t_aX *)(ivec + n)
+                                ^= *(size_t_aX *)(in + n);
                     }
                     len -= 16;
                     out += 16;
@@ -92,9 +99,10 @@ void CRYPTO_cfb128_encrypt(const unsigned char *in, unsigned char *out,
                 while (len >= 16) {
                     (*block) (ivec, ivec, key);
                     for (; n < 16; n += sizeof(size_t)) {
-                        size_t t = *(size_t *)(in + n);
-                        *(size_t *)(out + n) = *(size_t *)(ivec + n) ^ t;
-                        *(size_t *)(ivec + n) = t;
+                        size_t t = *(size_t_aX *)(in + n);
+                        *(size_t_aX *)(out + n)
+                            = *(size_t_aX *)(ivec + n) ^ t;
+                        *(size_t_aX *)(ivec + n) = t;
                     }
                     len -= 16;
                     out += 16;
diff --git a/crypto/modes/ctr128.c b/crypto/modes/ctr128.c
index ae35116e9524..1ed7decedfd3 100644
--- a/crypto/modes/ctr128.c
+++ b/crypto/modes/ctr128.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2008-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2008-2020 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
@@ -11,6 +11,12 @@
 #include "modes_local.h"
 #include <string.h>
 
+#if defined(__GNUC__) && !defined(STRICT_ALIGNMENT)
+typedef size_t size_t_aX __attribute((__aligned__(1)));
+#else
+typedef size_t size_t_aX;
+#endif
+
 /*
  * NOTE: the IV/counter CTR mode is big-endian.  The code itself is
  * endian-neutral.
@@ -97,8 +103,9 @@ void CRYPTO_ctr128_encrypt(const unsigned char *in, unsigned char *out,
                 (*block) (ivec, ecount_buf, key);
                 ctr128_inc_aligned(ivec);
                 for (n = 0; n < 16; n += sizeof(size_t))
-                    *(size_t *)(out + n) =
-                        *(size_t *)(in + n) ^ *(size_t *)(ecount_buf + n);
+                    *(size_t_aX *)(out + n) =
+                        *(size_t_aX *)(in + n)
+                        ^ *(size_t_aX *)(ecount_buf + n);
                 len -= 16;
                 out += 16;
                 in += 16;
diff --git a/crypto/modes/gcm128.c b/crypto/modes/gcm128.c
index 48775e6d05ff..0c0bf3cda5b5 100644
--- a/crypto/modes/gcm128.c
+++ b/crypto/modes/gcm128.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2010-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2010-2020 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
@@ -11,6 +11,12 @@
 #include "modes_local.h"
 #include <string.h>
 
+#if defined(__GNUC__) && !defined(STRICT_ALIGNMENT)
+typedef size_t size_t_aX __attribute((__aligned__(1)));
+#else
+typedef size_t size_t_aX;
+#endif
+
 #if defined(BSWAP4) && defined(STRICT_ALIGNMENT)
 /* redefine, because alignment is ensured */
 # undef  GETU32
@@ -1080,8 +1086,8 @@ int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx,
                 size_t j = GHASH_CHUNK;
 
                 while (j) {
-                    size_t *out_t = (size_t *)out;
-                    const size_t *in_t = (const size_t *)in;
+                    size_t_aX *out_t = (size_t_aX *)out;
+                    const size_t_aX *in_t = (const size_t_aX *)in;
 
                     (*block) (ctx->Yi.c, ctx->EKi.c, key);
                     ++ctr;
@@ -1107,8 +1113,8 @@ int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx,
                 size_t j = i;
 
                 while (len >= 16) {
-                    size_t *out_t = (size_t *)out;
-                    const size_t *in_t = (const size_t *)in;
+                    size_t_aX *out_t = (size_t_aX *)out;
+                    const size_t_aX *in_t = (const size_t_aX *)in;
 
                     (*block) (ctx->Yi.c, ctx->EKi.c, key);
                     ++ctr;
@@ -1318,8 +1324,8 @@ int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx,
 
                 GHASH(ctx, in, GHASH_CHUNK);
                 while (j) {
-                    size_t *out_t = (size_t *)out;
-                    const size_t *in_t = (const size_t *)in;
+                    size_t_aX *out_t = (size_t_aX *)out;
+                    const size_t_aX *in_t = (const size_t_aX *)in;
 
                     (*block) (ctx->Yi.c, ctx->EKi.c, key);
                     ++ctr;
@@ -1343,8 +1349,8 @@ int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx,
             if ((i = (len & (size_t)-16))) {
                 GHASH(ctx, in, i);
                 while (len >= 16) {
-                    size_t *out_t = (size_t *)out;
-                    const size_t *in_t = (const size_t *)in;
+                    size_t_aX *out_t = (size_t_aX *)out;
+                    const size_t_aX *in_t = (const size_t_aX *)in;
 
                     (*block) (ctx->Yi.c, ctx->EKi.c, key);
                     ++ctr;
diff --git a/crypto/modes/modes_local.h b/crypto/modes/modes_local.h
index f2ae01d11afd..28c32c0643f4 100644
--- a/crypto/modes/modes_local.h
+++ b/crypto/modes/modes_local.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2010-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2010-2020 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
@@ -37,6 +37,14 @@ typedef unsigned char u8;
 # endif
 #endif
 
+#ifndef STRICT_ALIGNMENT
+# ifdef __GNUC__
+typedef u32 u32_a1 __attribute((__aligned__(1)));
+# else
+typedef u32 u32_a1;
+# endif
+#endif
+
 #if !defined(PEDANTIC) && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
 # if defined(__GNUC__) && __GNUC__>=2
 #  if defined(__x86_64) || defined(__x86_64__)
@@ -86,8 +94,8 @@ _asm mov eax, val _asm bswap eax}
 # endif
 #endif
 #if defined(BSWAP4) && !defined(STRICT_ALIGNMENT)
-# define GETU32(p)       BSWAP4(*(const u32 *)(p))
-# define PUTU32(p,v)     *(u32 *)(p) = BSWAP4(v)
+# define GETU32(p)       BSWAP4(*(const u32_a1 *)(p))
+# define PUTU32(p,v)     *(u32_a1 *)(p) = BSWAP4(v)
 #else
 # define GETU32(p)       ((u32)(p)[0]<<24|(u32)(p)[1]<<16|(u32)(p)[2]<<8|(u32)(p)[3])
 # define PUTU32(p,v)     ((p)[0]=(u8)((v)>>24),(p)[1]=(u8)((v)>>16),(p)[2]=(u8)((v)>>8),(p)[3]=(u8)(v))
diff --git a/crypto/modes/ofb128.c b/crypto/modes/ofb128.c
index 44bdf888db1a..a3469712b2de 100644
--- a/crypto/modes/ofb128.c
+++ b/crypto/modes/ofb128.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2008-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2008-2020 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
@@ -11,6 +11,12 @@
 #include "modes_local.h"
 #include <string.h>
 
+#if defined(__GNUC__) && !defined(STRICT_ALIGNMENT)
+typedef size_t size_t_aX __attribute((__aligned__(1)));
+#else
+typedef size_t size_t_aX;
+#endif
+
 /*
  * The input and output encrypted as though 128bit ofb mode is being used.
  * The extra state information to record how much of the 128bit block we have
@@ -41,8 +47,9 @@ void CRYPTO_ofb128_encrypt(const unsigned char *in, unsigned char *out,
             while (len >= 16) {
                 (*block) (ivec, ivec, key);
                 for (; n < 16; n += sizeof(size_t))
-                    *(size_t *)(out + n) =
-                        *(size_t *)(in + n) ^ *(size_t *)(ivec + n);
+                    *(size_t_aX *)(out + n) =
+                        *(size_t_aX *)(in + n)
+                        ^ *(size_t_aX *)(ivec + n);
                 len -= 16;
                 out += 16;
                 in += 16;
diff --git a/crypto/modes/xts128.c b/crypto/modes/xts128.c
index b5bda5e6402d..fe1626c62e10 100644
--- a/crypto/modes/xts128.c
+++ b/crypto/modes/xts128.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2011-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2011-2020 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
@@ -11,6 +11,14 @@
 #include "modes_local.h"
 #include <string.h>
 
+#ifndef STRICT_ALIGNMENT
+# ifdef __GNUC__
+typedef u64 u64_a1 __attribute((__aligned__(1)));
+# else
+typedef u64 u64_a1;
+# endif
+#endif
+
 int CRYPTO_xts128_encrypt(const XTS128_CONTEXT *ctx,
                           const unsigned char iv[16],
                           const unsigned char *inp, unsigned char *out,
@@ -45,8 +53,8 @@ int CRYPTO_xts128_encrypt(const XTS128_CONTEXT *ctx,
         scratch.u[0] ^= tweak.u[0];
         scratch.u[1] ^= tweak.u[1];
 #else
-        scratch.u[0] = ((u64 *)inp)[0] ^ tweak.u[0];
-        scratch.u[1] = ((u64 *)inp)[1] ^ tweak.u[1];
+        scratch.u[0] = ((u64_a1 *)inp)[0] ^ tweak.u[0];
+        scratch.u[1] = ((u64_a1 *)inp)[1] ^ tweak.u[1];
 #endif
         (*ctx->block1) (scratch.c, scratch.c, ctx->key1);
 #if defined(STRICT_ALIGNMENT)
@@ -54,8 +62,8 @@ int CRYPTO_xts128_encrypt(const XTS128_CONTEXT *ctx,
         scratch.u[1] ^= tweak.u[1];
         memcpy(out, scratch.c, 16);
 #else
-        ((u64 *)out)[0] = scratch.u[0] ^= tweak.u[0];
-        ((u64 *)out)[1] = scratch.u[1] ^= tweak.u[1];
+        ((u64_a1 *)out)[0] = scratch.u[0] ^= tweak.u[0];
+        ((u64_a1 *)out)[1] = scratch.u[1] ^= tweak.u[1];
 #endif
         inp += 16;
         out += 16;
@@ -128,8 +136,8 @@ int CRYPTO_xts128_encrypt(const XTS128_CONTEXT *ctx,
         scratch.u[0] ^= tweak1.u[0];
         scratch.u[1] ^= tweak1.u[1];
 #else
-        scratch.u[0] = ((u64 *)inp)[0] ^ tweak1.u[0];
-        scratch.u[1] = ((u64 *)inp)[1] ^ tweak1.u[1];
+        scratch.u[0] = ((u64_a1 *)inp)[0] ^ tweak1.u[0];
+        scratch.u[1] = ((u64_a1 *)inp)[1] ^ tweak1.u[1];
 #endif
         (*ctx->block1) (scratch.c, scratch.c, ctx->key1);
         scratch.u[0] ^= tweak1.u[0];
@@ -148,8 +156,8 @@ int CRYPTO_xts128_encrypt(const XTS128_CONTEXT *ctx,
         scratch.u[1] ^= tweak.u[1];
         memcpy(out, scratch.c, 16);
 #else
-        ((u64 *)out)[0] = scratch.u[0] ^ tweak.u[0];
-        ((u64 *)out)[1] = scratch.u[1] ^ tweak.u[1];
+        ((u64_a1 *)out)[0] = scratch.u[0] ^ tweak.u[0];
+        ((u64_a1 *)out)[1] = scratch.u[1] ^ tweak.u[1];
 #endif
     }
 
diff --git a/crypto/o_str.c b/crypto/o_str.c
index 9ad7a89dcadf..eb9f21cc0c45 100644
--- a/crypto/o_str.c
+++ b/crypto/o_str.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2003-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2003-2020 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
@@ -220,7 +220,7 @@ char *OPENSSL_buf2hexstr(const unsigned char *buffer, long len)
 
 int openssl_strerror_r(int errnum, char *buf, size_t buflen)
 {
-#if defined(_MSC_VER) && _MSC_VER>=1400
+#if defined(_MSC_VER) && _MSC_VER>=1400 && !defined(_WIN32_WCE)
     return !strerror_s(buf, buflen, errnum);
 #elif defined(_GNU_SOURCE)
     char *err;
diff --git a/crypto/o_time.c b/crypto/o_time.c
index 6d764f55e2e8..3502edda6238 100644
--- a/crypto/o_time.c
+++ b/crypto/o_time.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2001-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2001-2020 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
@@ -41,7 +41,7 @@ struct tm *OPENSSL_gmtime(const time_t *timer, struct tm *result)
     if (gmtime_r(timer, result) == NULL)
         return NULL;
     ts = result;
-#elif defined (OPENSSL_SYS_WINDOWS) && defined(_MSC_VER) && _MSC_VER >= 1400
+#elif defined (OPENSSL_SYS_WINDOWS) && defined(_MSC_VER) && _MSC_VER >= 1400 && !defined(_WIN32_WCE)
     if (gmtime_s(result, timer))
         return NULL;
     ts = result;
diff --git a/crypto/pem/pem_err.c b/crypto/pem/pem_err.c
index f642030aa539..0f3cb02407e6 100644
--- a/crypto/pem/pem_err.c
+++ b/crypto/pem/pem_err.c
@@ -1,6 +1,6 @@
 /*
  * Generated by util/mkerr.pl DO NOT EDIT
- * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -60,6 +60,8 @@ static const ERR_STRING_DATA PEM_str_functs[] = {
     {ERR_PACK(ERR_LIB_PEM, PEM_F_PEM_SIGNFINAL, 0), "PEM_SignFinal"},
     {ERR_PACK(ERR_LIB_PEM, PEM_F_PEM_WRITE, 0), "PEM_write"},
     {ERR_PACK(ERR_LIB_PEM, PEM_F_PEM_WRITE_BIO, 0), "PEM_write_bio"},
+    {ERR_PACK(ERR_LIB_PEM, PEM_F_PEM_WRITE_BIO_PRIVATEKEY_TRADITIONAL, 0),
+     "PEM_write_bio_PrivateKey_traditional"},
     {ERR_PACK(ERR_LIB_PEM, PEM_F_PEM_WRITE_PRIVATEKEY, 0),
      "PEM_write_PrivateKey"},
     {ERR_PACK(ERR_LIB_PEM, PEM_F_PEM_X509_INFO_READ, 0), "PEM_X509_INFO_read"},
@@ -109,6 +111,8 @@ static const ERR_STRING_DATA PEM_str_reasons[] = {
     "unsupported encryption"},
     {ERR_PACK(ERR_LIB_PEM, 0, PEM_R_UNSUPPORTED_KEY_COMPONENTS),
     "unsupported key components"},
+    {ERR_PACK(ERR_LIB_PEM, 0, PEM_R_UNSUPPORTED_PUBLIC_KEY_TYPE),
+    "unsupported public key type"},
     {0, NULL}
 };
 
diff --git a/crypto/pem/pem_lib.c b/crypto/pem/pem_lib.c
index 64baf7108ea4..a26322119aa7 100644
--- a/crypto/pem/pem_lib.c
+++ b/crypto/pem/pem_lib.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -332,7 +332,7 @@ int PEM_ASN1_write_bio(i2d_of_void *i2d, const char *name, BIO *bp,
         }
     }
 
-    if ((dsize = i2d(x, NULL)) < 0) {
+    if ((dsize = i2d(x, NULL)) <= 0) {
         PEMerr(PEM_F_PEM_ASN1_WRITE_BIO, ERR_R_ASN1_LIB);
         dsize = 0;
         goto err;
@@ -791,7 +791,7 @@ static int get_header_and_data(BIO *bp, BIO **header, BIO **data, char *name,
 {
     BIO *tmp = *header;
     char *linebuf, *p;
-    int len, line, ret = 0, end = 0;
+    int len, line, ret = 0, end = 0, prev_partial_line_read = 0, partial_line_read = 0;
     /* 0 if not seen (yet), 1 if reading header, 2 if finished header */
     enum header_status got_header = MAYBE_HEADER;
     unsigned int flags_mask;
@@ -809,10 +809,18 @@ static int get_header_and_data(BIO *bp, BIO **header, BIO **data, char *name,
         flags_mask = ~0u;
         len = BIO_gets(bp, linebuf, LINESIZE);
         if (len <= 0) {
-            PEMerr(PEM_F_GET_HEADER_AND_DATA, PEM_R_SHORT_HEADER);
+            PEMerr(PEM_F_GET_HEADER_AND_DATA, PEM_R_BAD_END_LINE);
             goto err;
         }
 
+        /*
+         * Check if line has been read completely or if only part of the line
+         * has been read. Keep the previous value to ignore newlines that
+         * appear due to reading a line up until the char before the newline.
+         */
+        prev_partial_line_read = partial_line_read;
+        partial_line_read = len == LINESIZE-1 && linebuf[LINESIZE-2] != '\n';
+
         if (got_header == MAYBE_HEADER) {
             if (memchr(linebuf, ':', len) != NULL)
                 got_header = IN_HEADER;
@@ -823,13 +831,19 @@ static int get_header_and_data(BIO *bp, BIO **header, BIO **data, char *name,
 
         /* Check for end of header. */
         if (linebuf[0] == '\n') {
-            if (got_header == POST_HEADER) {
-                /* Another blank line is an error. */
-                PEMerr(PEM_F_GET_HEADER_AND_DATA, PEM_R_BAD_END_LINE);
-                goto err;
+            /*
+             * If previous line has been read only partially this newline is a
+             * regular newline at the end of a line and not an empty line.
+             */
+            if (!prev_partial_line_read) {
+                if (got_header == POST_HEADER) {
+                    /* Another blank line is an error. */
+                    PEMerr(PEM_F_GET_HEADER_AND_DATA, PEM_R_BAD_END_LINE);
+                    goto err;
+                }
+                got_header = POST_HEADER;
+                tmp = *data;
             }
-            got_header = POST_HEADER;
-            tmp = *data;
             continue;
         }
 
diff --git a/crypto/pem/pem_pkey.c b/crypto/pem/pem_pkey.c
index e58cdf4a3e0b..4a9492724487 100644
--- a/crypto/pem/pem_pkey.c
+++ b/crypto/pem/pem_pkey.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -108,6 +108,12 @@ int PEM_write_bio_PrivateKey_traditional(BIO *bp, EVP_PKEY *x,
                                          pem_password_cb *cb, void *u)
 {
     char pem_str[80];
+
+    if (x->ameth == NULL || x->ameth->old_priv_encode == NULL) {
+        PEMerr(PEM_F_PEM_WRITE_BIO_PRIVATEKEY_TRADITIONAL,
+               PEM_R_UNSUPPORTED_PUBLIC_KEY_TYPE);
+        return 0;
+    }
     BIO_snprintf(pem_str, 80, "%s PRIVATE KEY", x->ameth->pem_str);
     return PEM_ASN1_write_bio((i2d_of_void *)i2d_PrivateKey,
                               pem_str, bp, x, enc, kstr, klen, cb, u);
diff --git a/crypto/pem/pvkfmt.c b/crypto/pem/pvkfmt.c
index 1fc19c17f913..a933b7c1813c 100644
--- a/crypto/pem/pvkfmt.c
+++ b/crypto/pem/pvkfmt.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2005-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2005-2020 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
@@ -29,10 +29,10 @@ static unsigned int read_ledword(const unsigned char **in)
 {
     const unsigned char *p = *in;
     unsigned int ret;
-    ret = *p++;
-    ret |= (*p++ << 8);
-    ret |= (*p++ << 16);
-    ret |= (*p++ << 24);
+    ret = (unsigned int)*p++;
+    ret |= (unsigned int)*p++ << 8;
+    ret |= (unsigned int)*p++ << 16;
+    ret |= (unsigned int)*p++ << 24;
     *in = p;
     return ret;
 }
@@ -875,9 +875,9 @@ int i2b_PVK_bio(BIO *out, EVP_PKEY *pk, int enclevel,
     wrlen = BIO_write(out, tmp, outlen);
     OPENSSL_free(tmp);
     if (wrlen == outlen) {
-        PEMerr(PEM_F_I2B_PVK_BIO, PEM_R_BIO_WRITE_FAILURE);
         return outlen;
     }
+    PEMerr(PEM_F_I2B_PVK_BIO, PEM_R_BIO_WRITE_FAILURE);
     return -1;
 }
 
diff --git a/crypto/poly1305/asm/poly1305-x86.pl b/crypto/poly1305/asm/poly1305-x86.pl
index 4aaf63a0a1dd..2ae16a230b66 100755
--- a/crypto/poly1305/asm/poly1305-x86.pl
+++ b/crypto/poly1305/asm/poly1305-x86.pl
@@ -71,7 +71,7 @@ if ($sse2) {
 	$avx = ($1>=2.09) + ($1>=2.10);
 	}
 
-	if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/) {
+	if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/) {
 		$avx = ($2>=3.0) + ($2>3.0);
 	}
 }
diff --git a/crypto/poly1305/asm/poly1305-x86_64.pl b/crypto/poly1305/asm/poly1305-x86_64.pl
index c014be1ca9fa..5f834d8faf2a 100755
--- a/crypto/poly1305/asm/poly1305-x86_64.pl
+++ b/crypto/poly1305/asm/poly1305-x86_64.pl
@@ -90,7 +90,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=12);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/rand/drbg_ctr.c b/crypto/rand/drbg_ctr.c
index 0f0ad1b37be4..a757d0a258ab 100644
--- a/crypto/rand/drbg_ctr.c
+++ b/crypto/rand/drbg_ctr.c
@@ -63,15 +63,15 @@ static void ctr_XOR(RAND_DRBG_CTR *ctr, const unsigned char *in, size_t inlen)
  * Process a complete block using BCC algorithm of SP 800-90A 10.3.3
  */
 __owur static int ctr_BCC_block(RAND_DRBG_CTR *ctr, unsigned char *out,
-                                const unsigned char *in)
+                                const unsigned char *in, int len)
 {
     int i, outlen = AES_BLOCK_SIZE;
 
-    for (i = 0; i < 16; i++)
+    for (i = 0; i < len; i++)
         out[i] ^= in[i];
 
-    if (!EVP_CipherUpdate(ctr->ctx_df, out, &outlen, out, AES_BLOCK_SIZE)
-        || outlen != AES_BLOCK_SIZE)
+    if (!EVP_CipherUpdate(ctr->ctx_df, out, &outlen, out, len)
+        || outlen != len)
         return 0;
     return 1;
 }
@@ -82,12 +82,16 @@ __owur static int ctr_BCC_block(RAND_DRBG_CTR *ctr, unsigned char *out,
  */
 __owur static int ctr_BCC_blocks(RAND_DRBG_CTR *ctr, const unsigned char *in)
 {
-    if (!ctr_BCC_block(ctr, ctr->KX, in)
-        || !ctr_BCC_block(ctr, ctr->KX + 16, in))
-        return 0;
-    if (ctr->keylen != 16 && !ctr_BCC_block(ctr, ctr->KX + 32, in))
-        return 0;
-    return 1;
+    unsigned char in_tmp[48];
+    unsigned char num_of_blk = 2;
+
+    memcpy(in_tmp, in, 16);
+    memcpy(in_tmp + 16, in, 16);
+    if (ctr->keylen != 16) {
+        memcpy(in_tmp + 32, in, 16);
+        num_of_blk = 3;
+    }
+    return ctr_BCC_block(ctr, ctr->KX, in_tmp, AES_BLOCK_SIZE * num_of_blk);
 }
 
 /*
@@ -96,19 +100,14 @@ __owur static int ctr_BCC_blocks(RAND_DRBG_CTR *ctr, const unsigned char *in)
  */
 __owur static int ctr_BCC_init(RAND_DRBG_CTR *ctr)
 {
+    unsigned char bltmp[48] = {0};
+    unsigned char num_of_blk;
+
     memset(ctr->KX, 0, 48);
-    memset(ctr->bltmp, 0, 16);
-    if (!ctr_BCC_block(ctr, ctr->KX, ctr->bltmp))
-        return 0;
-    ctr->bltmp[3] = 1;
-    if (!ctr_BCC_block(ctr, ctr->KX + 16, ctr->bltmp))
-        return 0;
-    if (ctr->keylen != 16) {
-        ctr->bltmp[3] = 2;
-        if (!ctr_BCC_block(ctr, ctr->KX + 32, ctr->bltmp))
-            return 0;
-    }
-    return 1;
+    num_of_blk = ctr->keylen == 16 ? 2 : 3;
+    bltmp[(AES_BLOCK_SIZE * 1) + 3] = 1;
+    bltmp[(AES_BLOCK_SIZE * 2) + 3] = 2;
+    return ctr_BCC_block(ctr, ctr->KX, bltmp, num_of_blk * AES_BLOCK_SIZE);
 }
 
 /*
@@ -197,20 +196,20 @@ __owur static int ctr_df(RAND_DRBG_CTR *ctr,
         || !ctr_BCC_final(ctr))
         return 0;
     /* Set up key K */
-    if (!EVP_CipherInit_ex(ctr->ctx, ctr->cipher, NULL, ctr->KX, NULL, 1))
+    if (!EVP_CipherInit_ex(ctr->ctx_ecb, NULL, NULL, ctr->KX, NULL, -1))
         return 0;
     /* X follows key K */
-    if (!EVP_CipherUpdate(ctr->ctx, ctr->KX, &outlen, ctr->KX + ctr->keylen,
+    if (!EVP_CipherUpdate(ctr->ctx_ecb, ctr->KX, &outlen, ctr->KX + ctr->keylen,
                           AES_BLOCK_SIZE)
         || outlen != AES_BLOCK_SIZE)
         return 0;
-    if (!EVP_CipherUpdate(ctr->ctx, ctr->KX + 16, &outlen, ctr->KX,
+    if (!EVP_CipherUpdate(ctr->ctx_ecb, ctr->KX + 16, &outlen, ctr->KX,
                           AES_BLOCK_SIZE)
         || outlen != AES_BLOCK_SIZE)
         return 0;
     if (ctr->keylen != 16)
-        if (!EVP_CipherUpdate(ctr->ctx, ctr->KX + 32, &outlen, ctr->KX + 16,
-                              AES_BLOCK_SIZE)
+        if (!EVP_CipherUpdate(ctr->ctx_ecb, ctr->KX + 32, &outlen,
+                              ctr->KX + 16, AES_BLOCK_SIZE)
             || outlen != AES_BLOCK_SIZE)
             return 0;
     return 1;
@@ -229,31 +228,25 @@ __owur static int ctr_update(RAND_DRBG *drbg,
 {
     RAND_DRBG_CTR *ctr = &drbg->data.ctr;
     int outlen = AES_BLOCK_SIZE;
+    unsigned char V_tmp[48], out[48];
+    unsigned char len;
 
     /* correct key is already set up. */
+    memcpy(V_tmp, ctr->V, 16);
     inc_128(ctr);
-    if (!EVP_CipherUpdate(ctr->ctx, ctr->K, &outlen, ctr->V, AES_BLOCK_SIZE)
-        || outlen != AES_BLOCK_SIZE)
-        return 0;
-
-    /* If keylen longer than 128 bits need extra encrypt */
-    if (ctr->keylen != 16) {
+    memcpy(V_tmp + 16, ctr->V, 16);
+    if (ctr->keylen == 16) {
+        len = 32;
+    } else {
         inc_128(ctr);
-        if (!EVP_CipherUpdate(ctr->ctx, ctr->K+16, &outlen, ctr->V,
-                              AES_BLOCK_SIZE)
-            || outlen != AES_BLOCK_SIZE)
-            return 0;
+        memcpy(V_tmp + 32, ctr->V, 16);
+        len = 48;
     }
-    inc_128(ctr);
-    if (!EVP_CipherUpdate(ctr->ctx, ctr->V, &outlen, ctr->V, AES_BLOCK_SIZE)
-        || outlen != AES_BLOCK_SIZE)
+    if (!EVP_CipherUpdate(ctr->ctx_ecb, out, &outlen, V_tmp, len)
+            || outlen != len)
         return 0;
-
-    /* If 192 bit key part of V is on end of K */
-    if (ctr->keylen == 24) {
-        memcpy(ctr->V + 8, ctr->V, 8);
-        memcpy(ctr->V, ctr->K + 24, 8);
-    }
+    memcpy(ctr->K, out, ctr->keylen);
+    memcpy(ctr->V, out + ctr->keylen, 16);
 
     if ((drbg->flags & RAND_DRBG_FLAG_CTR_NO_DF) == 0) {
         /* If no input reuse existing derived value */
@@ -268,7 +261,8 @@ __owur static int ctr_update(RAND_DRBG *drbg,
         ctr_XOR(ctr, in2, in2len);
     }
 
-    if (!EVP_CipherInit_ex(ctr->ctx, ctr->cipher, NULL, ctr->K, NULL, 1))
+    if (!EVP_CipherInit_ex(ctr->ctx_ecb, NULL, NULL, ctr->K, NULL, -1)
+        || !EVP_CipherInit_ex(ctr->ctx_ctr, NULL, NULL, ctr->K, NULL, -1))
         return 0;
     return 1;
 }
@@ -285,8 +279,10 @@ __owur static int drbg_ctr_instantiate(RAND_DRBG *drbg,
 
     memset(ctr->K, 0, sizeof(ctr->K));
     memset(ctr->V, 0, sizeof(ctr->V));
-    if (!EVP_CipherInit_ex(ctr->ctx, ctr->cipher, NULL, ctr->K, NULL, 1))
+    if (!EVP_CipherInit_ex(ctr->ctx_ecb, NULL, NULL, ctr->K, NULL, -1))
         return 0;
+
+    inc_128(ctr);
     if (!ctr_update(drbg, entropy, entropylen, pers, perslen, nonce, noncelen))
         return 0;
     return 1;
@@ -296,20 +292,40 @@ __owur static int drbg_ctr_reseed(RAND_DRBG *drbg,
                                   const unsigned char *entropy, size_t entropylen,
                                   const unsigned char *adin, size_t adinlen)
 {
+    RAND_DRBG_CTR *ctr = &drbg->data.ctr;
+
     if (entropy == NULL)
         return 0;
+
+    inc_128(ctr);
     if (!ctr_update(drbg, entropy, entropylen, adin, adinlen, NULL, 0))
         return 0;
     return 1;
 }
 
+static void ctr96_inc(unsigned char *counter)
+{
+    u32 n = 12, c = 1;
+
+    do {
+        --n;
+        c += counter[n];
+        counter[n] = (u8)c;
+        c >>= 8;
+    } while (n);
+}
+
 __owur static int drbg_ctr_generate(RAND_DRBG *drbg,
                                     unsigned char *out, size_t outlen,
                                     const unsigned char *adin, size_t adinlen)
 {
     RAND_DRBG_CTR *ctr = &drbg->data.ctr;
+    unsigned int ctr32, blocks;
+    int outl, buflen;
 
     if (adin != NULL && adinlen != 0) {
+        inc_128(ctr);
+
         if (!ctr_update(drbg, adin, adinlen, NULL, 0, NULL, 0))
             return 0;
         /* This means we reuse derived value */
@@ -321,28 +337,53 @@ __owur static int drbg_ctr_generate(RAND_DRBG *drbg,
         adinlen = 0;
     }
 
-    for ( ; ; ) {
-        int outl = AES_BLOCK_SIZE;
+    inc_128(ctr);
 
+    if (outlen == 0) {
         inc_128(ctr);
-        if (outlen < 16) {
-            /* Use K as temp space as it will be updated */
-            if (!EVP_CipherUpdate(ctr->ctx, ctr->K, &outl, ctr->V,
-                                  AES_BLOCK_SIZE)
-                || outl != AES_BLOCK_SIZE)
-                return 0;
-            memcpy(out, ctr->K, outlen);
-            break;
-        }
-        if (!EVP_CipherUpdate(ctr->ctx, out, &outl, ctr->V, AES_BLOCK_SIZE)
-            || outl != AES_BLOCK_SIZE)
+
+        if (!ctr_update(drbg, adin, adinlen, NULL, 0, NULL, 0))
             return 0;
-        out += 16;
-        outlen -= 16;
-        if (outlen == 0)
-            break;
+        return 1;
     }
 
+    memset(out, 0, outlen);
+
+    do {
+        if (!EVP_CipherInit_ex(ctr->ctx_ctr,
+                               NULL, NULL, NULL, ctr->V, -1))
+            return 0;
+
+        /*-
+         * outlen has type size_t while EVP_CipherUpdate takes an
+         * int argument and thus cannot be guaranteed to process more
+         * than 2^31-1 bytes at a time. We process such huge generate
+         * requests in 2^30 byte chunks, which is the greatest multiple
+         * of AES block size lower than or equal to 2^31-1.
+         */
+        buflen = outlen > (1U << 30) ? (1U << 30) : outlen;
+        blocks = (buflen + 15) / 16;
+
+        ctr32 = GETU32(ctr->V + 12) + blocks;
+        if (ctr32 < blocks) {
+            /* 32-bit counter overflow into V. */
+            if (ctr32 != 0) {
+                blocks -= ctr32;
+                buflen = blocks * 16;
+                ctr32 = 0;
+            }
+            ctr96_inc(ctr->V);
+        }
+        PUTU32(ctr->V + 12, ctr32);
+
+        if (!EVP_CipherUpdate(ctr->ctx_ctr, out, &outl, out, buflen)
+            || outl != buflen)
+            return 0;
+
+        out += buflen;
+        outlen -= buflen;
+    } while (outlen);
+
     if (!ctr_update(drbg, adin, adinlen, NULL, 0, NULL, 0))
         return 0;
     return 1;
@@ -350,7 +391,8 @@ __owur static int drbg_ctr_generate(RAND_DRBG *drbg,
 
 static int drbg_ctr_uninstantiate(RAND_DRBG *drbg)
 {
-    EVP_CIPHER_CTX_free(drbg->data.ctr.ctx);
+    EVP_CIPHER_CTX_free(drbg->data.ctr.ctx_ecb);
+    EVP_CIPHER_CTX_free(drbg->data.ctr.ctx_ctr);
     EVP_CIPHER_CTX_free(drbg->data.ctr.ctx_df);
     OPENSSL_cleanse(&drbg->data.ctr, sizeof(drbg->data.ctr));
     return 1;
@@ -374,25 +416,36 @@ int drbg_ctr_init(RAND_DRBG *drbg)
         return 0;
     case NID_aes_128_ctr:
         keylen = 16;
-        ctr->cipher = EVP_aes_128_ecb();
+        ctr->cipher_ecb = EVP_aes_128_ecb();
+        ctr->cipher_ctr = EVP_aes_128_ctr();
         break;
     case NID_aes_192_ctr:
         keylen = 24;
-        ctr->cipher = EVP_aes_192_ecb();
+        ctr->cipher_ecb = EVP_aes_192_ecb();
+        ctr->cipher_ctr = EVP_aes_192_ctr();
         break;
     case NID_aes_256_ctr:
         keylen = 32;
-        ctr->cipher = EVP_aes_256_ecb();
+        ctr->cipher_ecb = EVP_aes_256_ecb();
+        ctr->cipher_ctr = EVP_aes_256_ctr();
         break;
     }
 
     drbg->meth = &drbg_ctr_meth;
 
     ctr->keylen = keylen;
-    if (ctr->ctx == NULL)
-        ctr->ctx = EVP_CIPHER_CTX_new();
-    if (ctr->ctx == NULL)
+    if (ctr->ctx_ecb == NULL)
+        ctr->ctx_ecb = EVP_CIPHER_CTX_new();
+    if (ctr->ctx_ctr == NULL)
+        ctr->ctx_ctr = EVP_CIPHER_CTX_new();
+    if (ctr->ctx_ecb == NULL || ctr->ctx_ctr == NULL
+        || !EVP_CipherInit_ex(ctr->ctx_ecb,
+                              ctr->cipher_ecb, NULL, NULL, NULL, 1)
+        || !EVP_CipherInit_ex(ctr->ctx_ctr,
+                              ctr->cipher_ctr, NULL, NULL, NULL, 1))
         return 0;
+
+    drbg->meth = &drbg_ctr_meth;
     drbg->strength = keylen * 8;
     drbg->seedlen = keylen + 16;
 
@@ -410,7 +463,8 @@ int drbg_ctr_init(RAND_DRBG *drbg)
         if (ctr->ctx_df == NULL)
             return 0;
         /* Set key schedule for df_key */
-        if (!EVP_CipherInit_ex(ctr->ctx_df, ctr->cipher, NULL, df_key, NULL, 1))
+        if (!EVP_CipherInit_ex(ctr->ctx_df,
+                               ctr->cipher_ecb, NULL, df_key, NULL, 1))
             return 0;
 
         drbg->min_entropylen = ctr->keylen;
diff --git a/crypto/rand/drbg_lib.c b/crypto/rand/drbg_lib.c
index faf0590c6c28..8c7c28c9703a 100644
--- a/crypto/rand/drbg_lib.c
+++ b/crypto/rand/drbg_lib.c
@@ -327,13 +327,6 @@ int RAND_DRBG_instantiate(RAND_DRBG *drbg,
         max_entropylen += drbg->max_noncelen;
     }
 
-    drbg->reseed_next_counter = tsan_load(&drbg->reseed_prop_counter);
-    if (drbg->reseed_next_counter) {
-        drbg->reseed_next_counter++;
-        if(!drbg->reseed_next_counter)
-            drbg->reseed_next_counter = 1;
-    }
-
     if (drbg->get_entropy != NULL)
         entropylen = drbg->get_entropy(drbg, &entropy, min_entropy,
                                        min_entropylen, max_entropylen, 0);
@@ -359,9 +352,15 @@ int RAND_DRBG_instantiate(RAND_DRBG *drbg,
     }
 
     drbg->state = DRBG_READY;
-    drbg->reseed_gen_counter = 1;
+    drbg->generate_counter = 1;
     drbg->reseed_time = time(NULL);
-    tsan_store(&drbg->reseed_prop_counter, drbg->reseed_next_counter);
+    if (drbg->enable_reseed_propagation) {
+        if (drbg->parent == NULL)
+            tsan_counter(&drbg->reseed_counter);
+        else
+            tsan_store(&drbg->reseed_counter,
+                       tsan_load(&drbg->parent->reseed_counter));
+    }
 
  end:
     if (entropy != NULL && drbg->cleanup_entropy != NULL)
@@ -428,14 +427,6 @@ int RAND_DRBG_reseed(RAND_DRBG *drbg,
     }
 
     drbg->state = DRBG_ERROR;
-
-    drbg->reseed_next_counter = tsan_load(&drbg->reseed_prop_counter);
-    if (drbg->reseed_next_counter) {
-        drbg->reseed_next_counter++;
-        if(!drbg->reseed_next_counter)
-            drbg->reseed_next_counter = 1;
-    }
-
     if (drbg->get_entropy != NULL)
         entropylen = drbg->get_entropy(drbg, &entropy, drbg->strength,
                                        drbg->min_entropylen,
@@ -451,9 +442,15 @@ int RAND_DRBG_reseed(RAND_DRBG *drbg,
         goto end;
 
     drbg->state = DRBG_READY;
-    drbg->reseed_gen_counter = 1;
+    drbg->generate_counter = 1;
     drbg->reseed_time = time(NULL);
-    tsan_store(&drbg->reseed_prop_counter, drbg->reseed_next_counter);
+    if (drbg->enable_reseed_propagation) {
+        if (drbg->parent == NULL)
+            tsan_counter(&drbg->reseed_counter);
+        else
+            tsan_store(&drbg->reseed_counter,
+                       tsan_load(&drbg->parent->reseed_counter));
+    }
 
  end:
     if (entropy != NULL && drbg->cleanup_entropy != NULL)
@@ -554,7 +551,9 @@ int rand_drbg_restart(RAND_DRBG *drbg,
             drbg->meth->reseed(drbg, adin, adinlen, NULL, 0);
         } else if (reseeded == 0) {
             /* do a full reseeding if it has not been done yet above */
-            RAND_DRBG_reseed(drbg, NULL, 0, 0);
+            if (!RAND_DRBG_reseed(drbg, NULL, 0, 0)) {
+                RANDerr(RAND_F_RAND_DRBG_RESTART, RAND_R_RESEED_ERROR);
+            }
         }
     }
 
@@ -612,7 +611,7 @@ int RAND_DRBG_generate(RAND_DRBG *drbg, unsigned char *out, size_t outlen,
     }
 
     if (drbg->reseed_interval > 0) {
-        if (drbg->reseed_gen_counter >= drbg->reseed_interval)
+        if (drbg->generate_counter >= drbg->reseed_interval)
             reseed_required = 1;
     }
     if (drbg->reseed_time_interval > 0) {
@@ -621,11 +620,8 @@ int RAND_DRBG_generate(RAND_DRBG *drbg, unsigned char *out, size_t outlen,
             || now - drbg->reseed_time >= drbg->reseed_time_interval)
             reseed_required = 1;
     }
-    if (drbg->parent != NULL) {
-        unsigned int reseed_counter = tsan_load(&drbg->reseed_prop_counter);
-        if (reseed_counter > 0
-                && tsan_load(&drbg->parent->reseed_prop_counter)
-                   != reseed_counter)
+    if (drbg->enable_reseed_propagation && drbg->parent != NULL) {
+        if (drbg->reseed_counter != tsan_load(&drbg->parent->reseed_counter))
             reseed_required = 1;
     }
 
@@ -644,7 +640,7 @@ int RAND_DRBG_generate(RAND_DRBG *drbg, unsigned char *out, size_t outlen,
         return 0;
     }
 
-    drbg->reseed_gen_counter++;
+    drbg->generate_counter++;
 
     return 1;
 }
@@ -706,8 +702,7 @@ int RAND_DRBG_set_callbacks(RAND_DRBG *drbg,
                             RAND_DRBG_get_nonce_fn get_nonce,
                             RAND_DRBG_cleanup_nonce_fn cleanup_nonce)
 {
-    if (drbg->state != DRBG_UNINITIALISED
-            || drbg->parent != NULL)
+    if (drbg->state != DRBG_UNINITIALISED)
         return 0;
     drbg->get_entropy = get_entropy;
     drbg->cleanup_entropy = cleanup_entropy;
@@ -883,8 +878,9 @@ static RAND_DRBG *drbg_setup(RAND_DRBG *parent)
     if (parent == NULL && rand_drbg_enable_locking(drbg) == 0)
         goto err;
 
-    /* enable seed propagation */
-    tsan_store(&drbg->reseed_prop_counter, 1);
+    /* enable reseed propagation */
+    drbg->enable_reseed_propagation = 1;
+    drbg->reseed_counter = 1;
 
     /*
      * Ignore instantiation error to support just-in-time instantiation.
diff --git a/crypto/rand/rand_lib.c b/crypto/rand/rand_lib.c
index ab4e9b5486cb..ba3a29e58468 100644
--- a/crypto/rand/rand_lib.c
+++ b/crypto/rand/rand_lib.c
@@ -174,8 +174,6 @@ size_t rand_drbg_get_entropy(RAND_DRBG *drbg,
                                    prediction_resistance,
                                    (unsigned char *)&drbg, sizeof(drbg)) != 0)
                 bytes = bytes_needed;
-            drbg->reseed_next_counter
-                = tsan_load(&drbg->parent->reseed_prop_counter);
             rand_drbg_unlock(drbg->parent);
 
             rand_pool_add_end(pool, bytes, 8 * bytes);
diff --git a/crypto/rand/rand_local.h b/crypto/rand/rand_local.h
index 1bc9bf7d266d..a5de5252dcdc 100644
--- a/crypto/rand/rand_local.h
+++ b/crypto/rand/rand_local.h
@@ -138,9 +138,11 @@ typedef struct rand_drbg_method_st {
  * The state of a DRBG AES-CTR.
  */
 typedef struct rand_drbg_ctr_st {
-    EVP_CIPHER_CTX *ctx;
+    EVP_CIPHER_CTX *ctx_ecb;
+    EVP_CIPHER_CTX *ctx_ctr;
     EVP_CIPHER_CTX *ctx_df;
-    const EVP_CIPHER *cipher;
+    const EVP_CIPHER *cipher_ecb;
+    const EVP_CIPHER *cipher_ctr;
     size_t keylen;
     unsigned char K[32];
     unsigned char V[16];
@@ -233,7 +235,7 @@ struct rand_drbg_st {
     size_t max_perslen, max_adinlen;
 
     /* Counts the number of generate requests since the last reseed. */
-    unsigned int reseed_gen_counter;
+    unsigned int generate_counter;
     /*
      * Maximum number of generate requests until a reseed is required.
      * This value is ignored if it is zero.
@@ -246,9 +248,15 @@ struct rand_drbg_st {
      * This value is ignored if it is zero.
      */
     time_t reseed_time_interval;
+
+    /*
+     * Enables reseed propagation (see following comment)
+     */
+    unsigned int enable_reseed_propagation;
+
     /*
      * Counts the number of reseeds since instantiation.
-     * This value is ignored if it is zero.
+     * This value is ignored if enable_reseed_propagation is zero.
      *
      * This counter is used only for seed propagation from the <master> DRBG
      * to its two children, the <public> and <private> DRBG. This feature is
@@ -256,8 +264,7 @@ struct rand_drbg_st {
      * is added by RAND_add() or RAND_seed() will have an immediate effect on
      * the output of RAND_bytes() resp. RAND_priv_bytes().
      */
-    TSAN_QUALIFIER unsigned int reseed_prop_counter;
-    unsigned int reseed_next_counter;
+    TSAN_QUALIFIER unsigned int reseed_counter;
 
     size_t seedlen;
     DRBG_STATUS state;
diff --git a/crypto/rand/rand_unix.c b/crypto/rand/rand_unix.c
index fe457cab4a3b..da66773e4ab9 100644
--- a/crypto/rand/rand_unix.c
+++ b/crypto/rand/rand_unix.c
@@ -26,12 +26,12 @@
 #  include <sys/utsname.h>
 # endif
 #endif
-#if defined(__FreeBSD__) && !defined(OPENSSL_SYS_UEFI)
+#if (defined(__FreeBSD__) || defined(__NetBSD__)) && !defined(OPENSSL_SYS_UEFI)
 # include <sys/types.h>
 # include <sys/sysctl.h>
 # include <sys/param.h>
 #endif
-#if defined(__OpenBSD__) || defined(__NetBSD__)
+#if defined(__OpenBSD__)
 # include <sys/param.h>
 #endif
 
@@ -247,10 +247,12 @@ static ssize_t sysctl_random(char *buf, size_t buflen)
      * when the sysctl returns long and we want to request something not a
      * multiple of longs, which should never be the case.
      */
+#if   defined(__FreeBSD__)
     if (!ossl_assert(buflen % sizeof(long) == 0)) {
         errno = EINVAL;
         return -1;
     }
+#endif
 
     /*
      * On NetBSD before 4.0 KERN_ARND was an alias for KERN_URND, and only
@@ -268,7 +270,7 @@ static ssize_t sysctl_random(char *buf, size_t buflen)
     mib[1] = KERN_ARND;
 
     do {
-        len = buflen;
+        len = buflen > 256 ? 256 : buflen;
         if (sysctl(mib, 2, buf, &len, NULL, 0) == -1)
             return done > 0 ? done : -1;
         done += len;
@@ -409,7 +411,8 @@ static struct random_device {
 } random_devices[OSSL_NELEM(random_device_paths)];
 static int keep_random_devices_open = 1;
 
-#   if defined(__linux) && defined(DEVRANDOM_WAIT)
+#   if defined(__linux) && defined(DEVRANDOM_WAIT) \
+       && defined(OPENSSL_RAND_SEED_GETRANDOM)
 static void *shm_addr;
 
 static void cleanup_shm(void)
@@ -487,7 +490,7 @@ static int wait_random_seeded(void)
     }
     return seeded;
 }
-#   else /* defined __linux */
+#   else /* defined __linux && DEVRANDOM_WAIT && OPENSSL_RAND_SEED_GETRANDOM */
 static int wait_random_seeded(void)
 {
     return 1;
diff --git a/crypto/rand/randfile.c b/crypto/rand/randfile.c
index ba121eefbf09..229ce864a312 100644
--- a/crypto/rand/randfile.c
+++ b/crypto/rand/randfile.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -26,7 +26,7 @@
 #ifndef OPENSSL_NO_POSIX_IO
 # include <sys/stat.h>
 # include <fcntl.h>
-# ifdef _WIN32
+# if defined(_WIN32) && !defined(_WIN32_WCE)
 #  include <windows.h>
 #  include <io.h>
 #  define stat    _stat
diff --git a/crypto/rsa/rsa_ameth.c b/crypto/rsa/rsa_ameth.c
index 6692a51ed8fe..fb045544a832 100644
--- a/crypto/rsa/rsa_ameth.c
+++ b/crypto/rsa/rsa_ameth.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2006-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2006-2020 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
@@ -118,6 +118,15 @@ static int rsa_pub_decode(EVP_PKEY *pkey, X509_PUBKEY *pubkey)
 
 static int rsa_pub_cmp(const EVP_PKEY *a, const EVP_PKEY *b)
 {
+    /*
+     * Don't check the public/private key, this is mostly for smart
+     * cards.
+     */
+    if (((RSA_flags(a->pkey.rsa) & RSA_METHOD_FLAG_NO_CHECK))
+            || (RSA_flags(b->pkey.rsa) & RSA_METHOD_FLAG_NO_CHECK)) {
+        return 1;
+    }
+
     if (BN_cmp(b->pkey.rsa->n, a->pkey.rsa->n) != 0
         || BN_cmp(b->pkey.rsa->e, a->pkey.rsa->e) != 0)
         return 0;
diff --git a/crypto/sha/asm/sha1-586.pl b/crypto/sha/asm/sha1-586.pl
index f7b85e6c6d7e..b72869b86d4a 100644
--- a/crypto/sha/asm/sha1-586.pl
+++ b/crypto/sha/asm/sha1-586.pl
@@ -144,7 +144,7 @@ $ymm=1 if ($xmm && !$ymm && $ARGV[0] eq "win32" &&
 		`ml 2>&1` =~ /Version ([0-9]+)\./ &&
 		$1>=10);	# first version supporting AVX
 
-$ymm=1 if ($xmm && !$ymm && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/ &&
+$ymm=1 if ($xmm && !$ymm && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/ &&
 		$2>=3.0);	# first version supporting AVX
 
 $shaext=$xmm;	### set to zero if compiling for 1.0.1
diff --git a/crypto/sha/asm/sha1-mb-x86_64.pl b/crypto/sha/asm/sha1-mb-x86_64.pl
index a78266d62fc3..47c588715163 100755
--- a/crypto/sha/asm/sha1-mb-x86_64.pl
+++ b/crypto/sha/asm/sha1-mb-x86_64.pl
@@ -66,7 +66,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/sha/asm/sha1-x86_64.pl b/crypto/sha/asm/sha1-x86_64.pl
index 0680d6d0a2b2..89146d137b9b 100755
--- a/crypto/sha/asm/sha1-x86_64.pl
+++ b/crypto/sha/asm/sha1-x86_64.pl
@@ -119,7 +119,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/sha/asm/sha256-586.pl b/crypto/sha/asm/sha256-586.pl
index 6989d59bd735..7523dbc165fa 100755
--- a/crypto/sha/asm/sha256-586.pl
+++ b/crypto/sha/asm/sha256-586.pl
@@ -96,7 +96,7 @@ if ($xmm && !$avx && $ARGV[0] eq "win32" &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if ($xmm && !$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/) {
+if ($xmm && !$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/sha/asm/sha256-mb-x86_64.pl b/crypto/sha/asm/sha256-mb-x86_64.pl
index 81ac814effc4..c8cc6efac95c 100755
--- a/crypto/sha/asm/sha256-mb-x86_64.pl
+++ b/crypto/sha/asm/sha256-mb-x86_64.pl
@@ -67,7 +67,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/sha/asm/sha512-x86_64.pl b/crypto/sha/asm/sha512-x86_64.pl
index 64ae641c1407..ad37850e6130 100755
--- a/crypto/sha/asm/sha512-x86_64.pl
+++ b/crypto/sha/asm/sha512-x86_64.pl
@@ -135,7 +135,7 @@ if (!$avx && $win64 && ($flavour =~ /masm/ || $ENV{ASM} =~ /ml64/) &&
 	$avx = ($1>=10) + ($1>=11);
 }
 
-if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:^clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
+if (!$avx && `$ENV{CC} -v 2>&1` =~ /((?:clang|LLVM) version|.*based on LLVM) ([0-9]+\.[0-9]+)/) {
 	$avx = ($2>=3.0) + ($2>3.0);
 }
 
diff --git a/crypto/store/loader_file.c b/crypto/store/loader_file.c
index 8f1d20e74aa4..9c9e3bd08506 100644
--- a/crypto/store/loader_file.c
+++ b/crypto/store/loader_file.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2016-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2016-2020 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
@@ -429,6 +429,42 @@ static OSSL_STORE_INFO *try_decode_PrivateKey(const char *pem_name,
         }
     } else {
         int i;
+#ifndef OPENSSL_NO_ENGINE
+        ENGINE *curengine = ENGINE_get_first();
+
+        while (curengine != NULL) {
+            ENGINE_PKEY_ASN1_METHS_PTR asn1meths =
+                ENGINE_get_pkey_asn1_meths(curengine);
+
+            if (asn1meths != NULL) {
+                const int *nids = NULL;
+                int nids_n = asn1meths(curengine, NULL, &nids, 0);
+
+                for (i = 0; i < nids_n; i++) {
+                    EVP_PKEY_ASN1_METHOD *ameth2 = NULL;
+                    EVP_PKEY *tmp_pkey = NULL;
+                    const unsigned char *tmp_blob = blob;
+
+                    if (!asn1meths(curengine, &ameth2, NULL, nids[i]))
+                        continue;
+                    if (ameth2 == NULL
+                        || ameth2->pkey_flags & ASN1_PKEY_ALIAS)
+                        continue;
+
+                    tmp_pkey = d2i_PrivateKey(ameth2->pkey_id, NULL,
+                                              &tmp_blob, len);
+                    if (tmp_pkey != NULL) {
+                        if (pkey != NULL)
+                            EVP_PKEY_free(tmp_pkey);
+                        else
+                            pkey = tmp_pkey;
+                        (*matchcount)++;
+                    }
+                }
+            }
+            curengine = ENGINE_get_next(curengine);
+        }
+#endif
 
         for (i = 0; i < EVP_PKEY_asn1_get_count(); i++) {
             EVP_PKEY *tmp_pkey = NULL;
diff --git a/crypto/store/store_lib.c b/crypto/store/store_lib.c
index fb8184d2d9b5..fb71f84725b1 100644
--- a/crypto/store/store_lib.c
+++ b/crypto/store/store_lib.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2016-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2016-2020 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
@@ -218,7 +218,11 @@ int OSSL_STORE_eof(OSSL_STORE_CTX *ctx)
 
 int OSSL_STORE_close(OSSL_STORE_CTX *ctx)
 {
-    int loader_ret = ctx->loader->close(ctx->loader_ctx);
+    int loader_ret;
+
+    if (ctx == NULL)
+        return 1;
+    loader_ret = ctx->loader->close(ctx->loader_ctx);
 
     OPENSSL_free(ctx);
     return loader_ret;
diff --git a/crypto/ts/ts_rsp_sign.c b/crypto/ts/ts_rsp_sign.c
index 041a187da68c..342582f024b2 100644
--- a/crypto/ts/ts_rsp_sign.c
+++ b/crypto/ts/ts_rsp_sign.c
@@ -57,12 +57,14 @@ static ASN1_INTEGER *def_serial_cb(struct TS_resp_ctx *ctx, void *data)
         goto err;
     if (!ASN1_INTEGER_set(serial, 1))
         goto err;
+
     return serial;
 
  err:
     TSerr(TS_F_DEF_SERIAL_CB, ERR_R_MALLOC_FAILURE);
     TS_RESP_CTX_set_status_info(ctx, TS_STATUS_REJECTION,
                                 "Error during serial number generation.");
+    ASN1_INTEGER_free(serial);
     return NULL;
 }
 
diff --git a/crypto/ui/ui_openssl.c b/crypto/ui/ui_openssl.c
index 168de4630dcc..9526c16536cb 100644
--- a/crypto/ui/ui_openssl.c
+++ b/crypto/ui/ui_openssl.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2001-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2001-2020 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
@@ -439,6 +439,16 @@ static int open_console(UI *ui)
             is_a_tty = 0;
         else
 #  endif
+#  ifdef EPERM
+            /*
+             * Linux can return EPERM (Operation not permitted),
+             * e.g. if a daemon executes openssl via fork()+execve()
+             * This should be ok
+             */
+        if (errno == EPERM)
+            is_a_tty = 0;
+        else
+#  endif
 #  ifdef ENODEV
             /*
              * MacOS X returns ENODEV (Operation not supported by device),
diff --git a/crypto/whrlpool/wp_block.c b/crypto/whrlpool/wp_block.c
index c21c04dbc1bb..39ad009c01bf 100644
--- a/crypto/whrlpool/wp_block.c
+++ b/crypto/whrlpool/wp_block.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2005-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2005-2020 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
@@ -63,6 +63,20 @@ typedef unsigned long long u64;
 # undef STRICT_ALIGNMENT
 #endif
 
+#ifndef STRICT_ALIGNMENT
+# ifdef __GNUC__
+typedef u64 u64_a1 __attribute((__aligned__(1)));
+# else
+typedef u64 u64_a1;
+# endif
+#endif
+
+#if defined(__GNUC__) && !defined(STRICT_ALIGNMENT)
+typedef u64 u64_aX __attribute((__aligned__(1)));
+#else
+typedef u64 u64_aX;
+#endif
+
 #undef SMALL_REGISTER_BANK
 #if defined(__i386) || defined(__i386__) || defined(_M_IX86)
 # define SMALL_REGISTER_BANK
@@ -191,13 +205,13 @@ typedef unsigned long long u64;
 # define LL(c0,c1,c2,c3,c4,c5,c6,c7)   c0,c1,c2,c3,c4,c5,c6,c7, \
                                         c0,c1,c2,c3,c4,c5,c6,c7
 # define C0(K,i)       (((u64*)(Cx.c+0))[2*K.c[(i)*8+0]])
-# define C1(K,i)       (((u64*)(Cx.c+7))[2*K.c[(i)*8+1]])
-# define C2(K,i)       (((u64*)(Cx.c+6))[2*K.c[(i)*8+2]])
-# define C3(K,i)       (((u64*)(Cx.c+5))[2*K.c[(i)*8+3]])
-# define C4(K,i)       (((u64*)(Cx.c+4))[2*K.c[(i)*8+4]])
-# define C5(K,i)       (((u64*)(Cx.c+3))[2*K.c[(i)*8+5]])
-# define C6(K,i)       (((u64*)(Cx.c+2))[2*K.c[(i)*8+6]])
-# define C7(K,i)       (((u64*)(Cx.c+1))[2*K.c[(i)*8+7]])
+# define C1(K,i)       (((u64_a1*)(Cx.c+7))[2*K.c[(i)*8+1]])
+# define C2(K,i)       (((u64_a1*)(Cx.c+6))[2*K.c[(i)*8+2]])
+# define C3(K,i)       (((u64_a1*)(Cx.c+5))[2*K.c[(i)*8+3]])
+# define C4(K,i)       (((u64_a1*)(Cx.c+4))[2*K.c[(i)*8+4]])
+# define C5(K,i)       (((u64_a1*)(Cx.c+3))[2*K.c[(i)*8+5]])
+# define C6(K,i)       (((u64_a1*)(Cx.c+2))[2*K.c[(i)*8+6]])
+# define C7(K,i)       (((u64_a1*)(Cx.c+1))[2*K.c[(i)*8+7]])
 #endif
 
 static const
@@ -531,7 +545,7 @@ void whirlpool_block(WHIRLPOOL_CTX *ctx, const void *inp, size_t n)
         } else
 # endif
         {
-            const u64 *pa = (const u64 *)p;
+            const u64_aX *pa = (const u64_aX *)p;
             S.q[0] = (K.q[0] = H->q[0]) ^ pa[0];
             S.q[1] = (K.q[1] = H->q[1]) ^ pa[1];
             S.q[2] = (K.q[2] = H->q[2]) ^ pa[2];
@@ -769,7 +783,7 @@ void whirlpool_block(WHIRLPOOL_CTX *ctx, const void *inp, size_t n)
         } else
 # endif
         {
-            const u64 *pa = (const u64 *)p;
+            const u64_aX *pa = (const u64_aX *)p;
             H->q[0] ^= S.q[0] ^ pa[0];
             H->q[1] ^= S.q[1] ^ pa[1];
             H->q[2] ^= S.q[2] ^ pa[2];
diff --git a/crypto/x509/x509_err.c b/crypto/x509/x509_err.c
index c110d908090e..bdd1e67cd3fd 100644
--- a/crypto/x509/x509_err.c
+++ b/crypto/x509/x509_err.c
@@ -1,6 +1,6 @@
 /*
  * Generated by util/mkerr.pl DO NOT EDIT
- * Copyright 1995-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -79,6 +79,7 @@ static const ERR_STRING_DATA X509_str_functs[] = {
     {ERR_PACK(ERR_LIB_X509, X509_F_X509_PRINT_EX_FP, 0), "X509_print_ex_fp"},
     {ERR_PACK(ERR_LIB_X509, X509_F_X509_PUBKEY_DECODE, 0),
      "x509_pubkey_decode"},
+    {ERR_PACK(ERR_LIB_X509, X509_F_X509_PUBKEY_GET, 0), "X509_PUBKEY_get"},
     {ERR_PACK(ERR_LIB_X509, X509_F_X509_PUBKEY_GET0, 0), "X509_PUBKEY_get0"},
     {ERR_PACK(ERR_LIB_X509, X509_F_X509_PUBKEY_SET, 0), "X509_PUBKEY_set"},
     {ERR_PACK(ERR_LIB_X509, X509_F_X509_REQ_CHECK_PRIVATE_KEY, 0),
diff --git a/crypto/x509/x509_local.h b/crypto/x509/x509_local.h
index c517a7745637..10807e1def04 100644
--- a/crypto/x509/x509_local.h
+++ b/crypto/x509/x509_local.h
@@ -1,5 +1,5 @@
 /*
- * Copyright 2014-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2014-2020 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
@@ -145,3 +145,5 @@ DEFINE_STACK_OF(STACK_OF_X509_NAME_ENTRY)
 
 void x509_set_signature_info(X509_SIG_INFO *siginf, const X509_ALGOR *alg,
                              const ASN1_STRING *sig);
+int x509_likely_issued(X509 *issuer, X509 *subject);
+int x509_signing_allowed(const X509 *issuer, const X509 *subject);
diff --git a/crypto/x509/x509_req.c b/crypto/x509/x509_req.c
index 7ba0f26495f9..dd674926ddb5 100644
--- a/crypto/x509/x509_req.c
+++ b/crypto/x509/x509_req.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -286,6 +286,18 @@ void X509_REQ_get0_signature(const X509_REQ *req, const ASN1_BIT_STRING **psig,
         *palg = &req->sig_alg;
 }
 
+void X509_REQ_set0_signature(X509_REQ *req, ASN1_BIT_STRING *psig)
+{
+    if (req->signature)
+           ASN1_BIT_STRING_free(req->signature);
+    req->signature = psig;
+}
+
+int X509_REQ_set1_signature_algo(X509_REQ *req, X509_ALGOR *palg)
+{
+    return X509_ALGOR_copy(&req->sig_alg, palg);
+}
+
 int X509_REQ_get_signature_nid(const X509_REQ *req)
 {
     return OBJ_obj2nid(req->sig_alg.algorithm);
diff --git a/crypto/x509/x509_txt.c b/crypto/x509/x509_txt.c
index 4755b39eb4eb..02bde640d8e8 100644
--- a/crypto/x509/x509_txt.c
+++ b/crypto/x509/x509_txt.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2017 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -174,6 +174,8 @@ const char *X509_verify_cert_error_string(long n)
         return "OCSP verification failed";
     case X509_V_ERR_OCSP_CERT_UNKNOWN:
         return "OCSP unknown cert";
+    case X509_V_ERR_EC_KEY_EXPLICIT_PARAMS:
+        return "Certificate public key has explicit ECC parameters";
 
     default:
         /* Printing an error number into a static buffer is not thread-safe */
diff --git a/crypto/x509/x509_vfy.c b/crypto/x509/x509_vfy.c
index 41625e75ad6a..801055f5a087 100644
--- a/crypto/x509/x509_vfy.c
+++ b/crypto/x509/x509_vfy.c
@@ -80,6 +80,7 @@ static int get_issuer_sk(X509 **issuer, X509_STORE_CTX *ctx, X509 *x);
 static int check_dane_issuer(X509_STORE_CTX *ctx, int depth);
 static int check_key_level(X509_STORE_CTX *ctx, X509 *cert);
 static int check_sig_level(X509_STORE_CTX *ctx, X509 *cert);
+static int check_curve(X509 *cert);
 
 static int get_crl_score(X509_STORE_CTX *ctx, X509 **pissuer,
                          unsigned int *preasons, X509_CRL *crl, X509 *x);
@@ -104,7 +105,12 @@ static int null_callback(int ok, X509_STORE_CTX *e)
     return ok;
 }
 
-/* Return 1 is a certificate is self signed */
+/*
+ * Return 1 if given cert is considered self-signed, 0 if not or on error.
+ * This does not verify self-signedness but relies on x509v3_cache_extensions()
+ * matching issuer and subject names (i.e., the cert being self-issued) and any
+ * present authority key identifier matching the subject key identifier, etc.
+ */
 static int cert_self_signed(X509 *x)
 {
     if (X509_check_purpose(x, -1, 0) != 1)
@@ -131,10 +137,9 @@ static X509 *lookup_cert_match(X509_STORE_CTX *ctx, X509 *x)
         xtmp = sk_X509_value(certs, i);
         if (!X509_cmp(xtmp, x))
             break;
+        xtmp = NULL;
     }
-    if (i < sk_X509_num(certs))
-        X509_up_ref(xtmp);
-    else
+    if (xtmp != NULL && !X509_up_ref(xtmp))
         xtmp = NULL;
     sk_X509_pop_free(certs, X509_free);
     return xtmp;
@@ -267,17 +272,24 @@ int X509_verify_cert(X509_STORE_CTX *ctx)
         return -1;
     }
 
+    if (!X509_up_ref(ctx->cert)) {
+        X509err(X509_F_X509_VERIFY_CERT, ERR_R_INTERNAL_ERROR);
+        ctx->error = X509_V_ERR_UNSPECIFIED;
+        return -1;
+    }
+
     /*
      * first we make sure the chain we are going to build is present and that
      * the first entry is in place
      */
-    if (((ctx->chain = sk_X509_new_null()) == NULL) ||
-        (!sk_X509_push(ctx->chain, ctx->cert))) {
+    if ((ctx->chain = sk_X509_new_null()) == NULL
+            || !sk_X509_push(ctx->chain, ctx->cert)) {
+        X509_free(ctx->cert);
         X509err(X509_F_X509_VERIFY_CERT, ERR_R_MALLOC_FAILURE);
         ctx->error = X509_V_ERR_OUT_OF_MEM;
         return -1;
     }
-    X509_up_ref(ctx->cert);
+
     ctx->num_untrusted = 1;
 
     /* If the peer's public key is too weak, we can stop early. */
@@ -319,30 +331,26 @@ static X509 *find_issuer(X509_STORE_CTX *ctx, STACK_OF(X509) *sk, X509 *x)
     return rv;
 }
 
-/* Given a possible certificate and issuer check them */
-
+/*
+ * Check that the given certificate 'x' is issued by the certificate 'issuer'
+ * and the issuer is not yet in ctx->chain, where the exceptional case
+ * that 'x' is self-issued and ctx->chain has just one element is allowed.
+ */
 static int check_issued(X509_STORE_CTX *ctx, X509 *x, X509 *issuer)
 {
-    int ret;
-    if (x == issuer)
-        return cert_self_signed(x);
-    ret = X509_check_issued(issuer, x);
-    if (ret == X509_V_OK) {
+    if (x509_likely_issued(issuer, x) != X509_V_OK)
+        return 0;
+    if ((x->ex_flags & EXFLAG_SI) == 0 || sk_X509_num(ctx->chain) != 1) {
         int i;
         X509 *ch;
-        /* Special case: single self signed certificate */
-        if (cert_self_signed(x) && sk_X509_num(ctx->chain) == 1)
-            return 1;
+
         for (i = 0; i < sk_X509_num(ctx->chain); i++) {
             ch = sk_X509_value(ctx->chain, i);
-            if (ch == issuer || !X509_cmp(ch, issuer)) {
-                ret = X509_V_ERR_PATH_LOOP;
-                break;
-            }
+            if (ch == issuer || X509_cmp(ch, issuer) == 0)
+                return 0;
         }
     }
-
-    return (ret == X509_V_OK);
+    return 1;
 }
 
 /* Alternative lookup method: look from a STACK stored in other_ctx */
@@ -350,11 +358,15 @@ static int check_issued(X509_STORE_CTX *ctx, X509 *x, X509 *issuer)
 static int get_issuer_sk(X509 **issuer, X509_STORE_CTX *ctx, X509 *x)
 {
     *issuer = find_issuer(ctx, ctx->other_ctx, x);
-    if (*issuer) {
-        X509_up_ref(*issuer);
-        return 1;
-    } else
-        return 0;
+
+    if (*issuer == NULL || !X509_up_ref(*issuer))
+        goto err;
+
+    return 1;
+
+ err:
+    *issuer = NULL;
+    return 0;
 }
 
 static STACK_OF(X509) *lookup_certs_sk(X509_STORE_CTX *ctx, X509_NAME *nm)
@@ -366,15 +378,21 @@ static STACK_OF(X509) *lookup_certs_sk(X509_STORE_CTX *ctx, X509_NAME *nm)
     for (i = 0; i < sk_X509_num(ctx->other_ctx); i++) {
         x = sk_X509_value(ctx->other_ctx, i);
         if (X509_NAME_cmp(nm, X509_get_subject_name(x)) == 0) {
+            if (!X509_up_ref(x)) {
+                sk_X509_pop_free(sk, X509_free);
+                X509err(X509_F_LOOKUP_CERTS_SK, ERR_R_INTERNAL_ERROR);
+                ctx->error = X509_V_ERR_UNSPECIFIED;
+                return NULL;
+            }
             if (sk == NULL)
                 sk = sk_X509_new_null();
-            if (sk == NULL || sk_X509_push(sk, x) == 0) {
+            if (sk == NULL || !sk_X509_push(sk, x)) {
+                X509_free(x);
                 sk_X509_pop_free(sk, X509_free);
                 X509err(X509_F_LOOKUP_CERTS_SK, ERR_R_MALLOC_FAILURE);
                 ctx->error = X509_V_ERR_OUT_OF_MEM;
                 return NULL;
             }
-            X509_up_ref(x);
         }
     }
     return sk;
@@ -508,6 +526,14 @@ static int check_chain_extensions(X509_STORE_CTX *ctx)
                 ret = 1;
             break;
         }
+        if ((ctx->param->flags & X509_V_FLAG_X509_STRICT) && num > 1) {
+            /* Check for presence of explicit elliptic curve parameters */
+            ret = check_curve(x);
+            if (ret < 0)
+                ctx->error = X509_V_ERR_UNSPECIFIED;
+            else if (ret == 0)
+                ctx->error = X509_V_ERR_EC_KEY_EXPLICIT_PARAMS;
+        }
         if ((x->ex_flags & EXFLAG_CA) == 0
             && x->ex_pathlen != -1
             && (ctx->param->flags & X509_V_FLAG_X509_STRICT)) {
@@ -1699,6 +1725,7 @@ int x509_check_cert_time(X509_STORE_CTX *ctx, X509 *x, int depth)
     return 1;
 }
 
+/* verify the issuer signatures and cert times of ctx->chain */
 static int internal_verify(X509_STORE_CTX *ctx)
 {
     int n = sk_X509_num(ctx->chain) - 1;
@@ -1717,7 +1744,7 @@ static int internal_verify(X509_STORE_CTX *ctx)
     }
 
     if (ctx->check_issued(ctx, xi, xi))
-        xs = xi;
+        xs = xi; /* the typical case: last cert in the chain is self-issued */
     else {
         if (ctx->param->flags & X509_V_FLAG_PARTIAL_CHAIN) {
             xs = xi;
@@ -1736,22 +1763,50 @@ static int internal_verify(X509_STORE_CTX *ctx)
      * is allowed to reset errors (at its own peril).
      */
     while (n >= 0) {
-        EVP_PKEY *pkey;
-
         /*
-         * Skip signature check for self signed certificates unless explicitly
-         * asked for.  It doesn't add any security and just wastes time.  If
-         * the issuer's public key is unusable, report the issuer certificate
-         * and its depth (rather than the depth of the subject).
+         * For each iteration of this loop:
+         * n is the subject depth
+         * xs is the subject cert, for which the signature is to be checked
+         * xi is the supposed issuer cert containing the public key to use
+         * Initially xs == xi if the last cert in the chain is self-issued.
+         *
+         * Skip signature check for self-signed certificates unless explicitly
+         * asked for because it does not add any security and just wastes time.
          */
-        if (xs != xi || (ctx->param->flags & X509_V_FLAG_CHECK_SS_SIGNATURE)) {
+        if (xs != xi || ((ctx->param->flags & X509_V_FLAG_CHECK_SS_SIGNATURE)
+                         && (xi->ex_flags & EXFLAG_SS) != 0)) {
+            EVP_PKEY *pkey;
+            /*
+             * If the issuer's public key is not available or its key usage
+             * does not support issuing the subject cert, report the issuer
+             * cert and its depth (rather than n, the depth of the subject).
+             */
+            int issuer_depth = n + (xs == xi ? 0 : 1);
+            /*
+             * According to https://tools.ietf.org/html/rfc5280#section-6.1.4
+             * step (n) we must check any given key usage extension in a CA cert
+             * when preparing the verification of a certificate issued by it.
+             * According to https://tools.ietf.org/html/rfc5280#section-4.2.1.3
+             * we must not verify a certifiate signature if the key usage of the
+             * CA certificate that issued the certificate prohibits signing.
+             * In case the 'issuing' certificate is the last in the chain and is
+             * not a CA certificate but a 'self-issued' end-entity cert (i.e.,
+             * xs == xi && !(xi->ex_flags & EXFLAG_CA)) RFC 5280 does not apply
+             * (see https://tools.ietf.org/html/rfc6818#section-2) and thus
+             * we are free to ignore any key usage restrictions on such certs.
+             */
+            int ret = xs == xi && (xi->ex_flags & EXFLAG_CA) == 0
+                ? X509_V_OK : x509_signing_allowed(xi, xs);
+
+            if (ret != X509_V_OK && !verify_cb_cert(ctx, xi, issuer_depth, ret))
+                return 0;
             if ((pkey = X509_get0_pubkey(xi)) == NULL) {
-                if (!verify_cb_cert(ctx, xi, xi != xs ? n+1 : n,
-                        X509_V_ERR_UNABLE_TO_DECODE_ISSUER_PUBLIC_KEY))
+                ret = X509_V_ERR_UNABLE_TO_DECODE_ISSUER_PUBLIC_KEY;
+                if (!verify_cb_cert(ctx, xi, issuer_depth, ret))
                     return 0;
             } else if (X509_verify(xs, pkey) <= 0) {
-                if (!verify_cb_cert(ctx, xs, n,
-                                    X509_V_ERR_CERT_SIGNATURE_FAILURE))
+                ret = X509_V_ERR_CERT_SIGNATURE_FAILURE;
+                if (!verify_cb_cert(ctx, xs, n, ret))
                     return 0;
             }
         }
@@ -3158,7 +3213,16 @@ static int build_chain(X509_STORE_CTX *ctx)
             /* Drop this issuer from future consideration */
             (void) sk_X509_delete_ptr(sktmp, xtmp);
 
+            if (!X509_up_ref(xtmp)) {
+                X509err(X509_F_BUILD_CHAIN, ERR_R_INTERNAL_ERROR);
+                trust = X509_TRUST_REJECTED;
+                ctx->error = X509_V_ERR_UNSPECIFIED;
+                search = 0;
+                continue;
+            }
+
             if (!sk_X509_push(ctx->chain, xtmp)) {
+                X509_free(xtmp);
                 X509err(X509_F_BUILD_CHAIN, ERR_R_MALLOC_FAILURE);
                 trust = X509_TRUST_REJECTED;
                 ctx->error = X509_V_ERR_OUT_OF_MEM;
@@ -3166,7 +3230,7 @@ static int build_chain(X509_STORE_CTX *ctx)
                 continue;
             }
 
-            X509_up_ref(x = xtmp);
+            x = xtmp;
             ++ctx->num_untrusted;
             ss = cert_self_signed(xtmp);
 
@@ -3258,6 +3322,32 @@ static int check_key_level(X509_STORE_CTX *ctx, X509 *cert)
 }
 
 /*
+ * Check whether the public key of ``cert`` does not use explicit params
+ * for an elliptic curve.
+ *
+ * Returns 1 on success, 0 if check fails, -1 for other errors.
+ */
+static int check_curve(X509 *cert)
+{
+#ifndef OPENSSL_NO_EC
+    EVP_PKEY *pkey = X509_get0_pubkey(cert);
+
+    /* Unsupported or malformed key */
+    if (pkey == NULL)
+        return -1;
+
+    if (EVP_PKEY_id(pkey) == EVP_PKEY_EC) {
+        int ret;
+
+        ret = EC_KEY_decoded_from_explicit_params(EVP_PKEY_get0_EC_KEY(pkey));
+        return ret < 0 ? ret : !ret;
+    }
+#endif
+
+    return 1;
+}
+
+/*
  * Check whether the signature digest algorithm of ``cert`` meets the security
  * level of ``ctx``.  Should not be checked for trust anchors (whether
  * self-signed or otherwise).
diff --git a/crypto/x509/x_pubkey.c b/crypto/x509/x_pubkey.c
index 4f694b93fb00..9be7e9286571 100644
--- a/crypto/x509/x_pubkey.c
+++ b/crypto/x509/x_pubkey.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1995-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1995-2020 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
@@ -169,8 +169,11 @@ EVP_PKEY *X509_PUBKEY_get0(X509_PUBKEY *key)
 EVP_PKEY *X509_PUBKEY_get(X509_PUBKEY *key)
 {
     EVP_PKEY *ret = X509_PUBKEY_get0(key);
-    if (ret != NULL)
-        EVP_PKEY_up_ref(ret);
+
+    if (ret != NULL && !EVP_PKEY_up_ref(ret)) {
+        X509err(X509_F_X509_PUBKEY_GET, ERR_R_INTERNAL_ERROR);
+        ret = NULL;
+    }
     return ret;
 }
 
diff --git a/crypto/x509v3/pcy_data.c b/crypto/x509v3/pcy_data.c
index 073505951322..8c7bc69576a4 100644
--- a/crypto/x509v3/pcy_data.c
+++ b/crypto/x509v3/pcy_data.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 2004-2018 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2004-2020 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
@@ -52,6 +52,7 @@ X509_POLICY_DATA *policy_data_new(POLICYINFO *policy,
     ret = OPENSSL_zalloc(sizeof(*ret));
     if (ret == NULL) {
         X509V3err(X509V3_F_POLICY_DATA_NEW, ERR_R_MALLOC_FAILURE);
+        ASN1_OBJECT_free(id);
         return NULL;
     }
     ret->expected_policy_set = sk_ASN1_OBJECT_new_null();
diff --git a/crypto/x509v3/v3_alt.c b/crypto/x509v3/v3_alt.c
index 7ac2911b91af..4dce0041012e 100644
--- a/crypto/x509v3/v3_alt.c
+++ b/crypto/x509v3/v3_alt.c
@@ -1,5 +1,5 @@
 /*
- * Copyright 1999-2019 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 1999-2020 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
@@ -275,6 +275,7 @@ static int copy_issuer(X509V3_CTX *ctx, GENERAL_NAMES *gens)
     num = sk_GENERAL_NAME_num(ialt);
     if (!sk_GENERAL_NAME_reserve(gens, num)) {
         X509V3err(X509V3_F_COPY_ISSUER, ERR_R_MALLOC_FAILURE);
+        sk_GENERAL_NAME_free(ialt);
         goto err;
     }
 
diff --git a/crypto/x509v3/v3_purp.c b/crypto/x509v3/v3_purp.c
index f023c6489548..2b06dba05398 100644
--- a/crypto/x509v3/v3_purp.c
+++ b/crypto/x509v3/v3_purp.c
@@ -13,6 +13,7 @@
 #include <openssl/x509v3.h>
 #include <openssl/x509_vfy.h>
 #include "crypto/x509.h"
+#include "../x509/x509_local.h" /* for x509_signing_allowed() */
 #include "internal/tsan_assist.h"
 
 static void x509v3_cache_extensions(X509 *x);
@@ -344,6 +345,21 @@ static int setup_crldp(X509 *x)
     return 1;
 }
 
+/* Check that issuer public key algorithm matches subject signature algorithm */
+static int check_sig_alg_match(const EVP_PKEY *pkey, const X509 *subject)
+{
+    int pkey_nid;
+
+    if (pkey == NULL)
+        return X509_V_ERR_NO_ISSUER_PUBLIC_KEY;
+    if (OBJ_find_sigid_algs(OBJ_obj2nid(subject->cert_info.signature.algorithm),
+                            NULL, &pkey_nid) == 0)
+        return X509_V_ERR_UNSUPPORTED_SIGNATURE_ALGORITHM;
+    if (EVP_PKEY_type(pkey_nid) != EVP_PKEY_base_id(pkey))
+        return X509_V_ERR_SIGNATURE_ALGORITHM_MISMATCH;
+    return X509_V_OK;
+}
+
 #define V1_ROOT (EXFLAG_V1|EXFLAG_SS)
 #define ku_reject(x, usage) \
         (((x)->ex_flags & EXFLAG_KUSAGE) && !((x)->ex_kusage & (usage)))
@@ -496,11 +512,11 @@ static void x509v3_cache_extensions(X509 *x)
         x->ex_flags |= EXFLAG_INVALID;
     /* Does subject name match issuer ? */
     if (!X509_NAME_cmp(X509_get_subject_name(x), X509_get_issuer_name(x))) {
-        x->ex_flags |= EXFLAG_SI;
-        /* If SKID matches AKID also indicate self signed */
-        if (X509_check_akid(x, x->akid) == X509_V_OK &&
-            !ku_reject(x, KU_KEY_CERT_SIGN))
-            x->ex_flags |= EXFLAG_SS;
+        x->ex_flags |= EXFLAG_SI; /* cert is self-issued */
+        if (X509_check_akid(x, x->akid) == X509_V_OK /* SKID matches AKID */
+                /* .. and the signature alg matches the PUBKEY alg: */
+                && check_sig_alg_match(X509_get0_pubkey(x), x) == X509_V_OK)
+            x->ex_flags |= EXFLAG_SS; /* indicate self-signed */
     }
     x->altname = X509_get_ext_d2i(x, NID_subject_alt_name, &i, NULL);
     if (x->altname == NULL && i != -1)
@@ -793,6 +809,23 @@ static int no_check(const X509_PURPOSE *xp, const X509 *x, int ca)
 }
 
 /*-
+ * Check if certificate I<issuer> is allowed to issue certificate I<subject>
+ * according to the B<keyUsage> field of I<issuer> if present
+ * depending on any proxyCertInfo extension of I<subject>.
+ * Returns 0 for OK, or positive for reason for rejection
+ * where reason codes match those for X509_verify_cert().
+ */
+int x509_signing_allowed(const X509 *issuer, const X509 *subject)
+{
+    if (subject->ex_flags & EXFLAG_PROXY) {
+        if (ku_reject(issuer, KU_DIGITAL_SIGNATURE))
+            return X509_V_ERR_KEYUSAGE_NO_DIGITAL_SIGNATURE;
+    } else if (ku_reject(issuer, KU_KEY_CERT_SIGN))
+        return X509_V_ERR_KEYUSAGE_NO_CERTSIGN;
+    return X509_V_OK;
+}
+
+/*-
  * Various checks to see if one certificate issued the second.
  * This can be used to prune a set of possible issuer certificates
  * which have been looked up using some simple method such as by
@@ -800,13 +833,24 @@ static int no_check(const X509_PURPOSE *xp, const X509 *x, int ca)
  * These are:
  * 1. Check issuer_name(subject) == subject_name(issuer)
  * 2. If akid(subject) exists check it matches issuer
- * 3. If key_usage(issuer) exists check it supports certificate signing
+ * 3. Check that issuer public key algorithm matches subject signature algorithm
+ * 4. If key_usage(issuer) exists check it supports certificate signing
  * returns 0 for OK, positive for reason for mismatch, reasons match
  * codes for X509_verify_cert()
  */
 
 int X509_check_issued(X509 *issuer, X509 *subject)
 {
+    int ret;
+
+    if ((ret = x509_likely_issued(issuer, subject)) != X509_V_OK)
+        return ret;
+    return x509_signing_allowed(issuer, subject);
+}
+
+/* do the checks 1., 2., and 3. as described above for X509_check_issued() */
+int x509_likely_issued(X509 *issuer, X509 *subject)
+{
     if (X509_NAME_cmp(X509_get_subject_name(issuer),
                       X509_get_issuer_name(subject)))
         return X509_V_ERR_SUBJECT_ISSUER_MISMATCH;
@@ -824,12 +868,8 @@ int X509_check_issued(X509 *issuer, X509 *subject)
             return ret;
     }
 
-    if (subject->ex_flags & EXFLAG_PROXY) {
-        if (ku_reject(issuer, KU_DIGITAL_SIGNATURE))
-            return X509_V_ERR_KEYUSAGE_NO_DIGITAL_SIGNATURE;
-    } else if (ku_reject(issuer, KU_KEY_CERT_SIGN))
-        return X509_V_ERR_KEYUSAGE_NO_CERTSIGN;
-    return X509_V_OK;
+    /* check if the subject signature alg matches the issuer's PUBKEY alg */
+    return check_sig_alg_match(X509_get0_pubkey(issuer), subject);
 }
 
 int X509_check_akid(X509 *issuer, AUTHORITY_KEYID *akid)