Remove I386_ONLY define. It was only used to prefer a

faster-on-genuine-80386-but-slower-on-80486-onwards innstruction sequence in
the SHA512 code, and had not been enabled in years, if at all.

ok tom@ bcook@

18 files changed, 169 insertions, 57 deletions

diff --git a/lib/libcrypto/arch/alpha/opensslconf.h b/lib/libcrypto/arch/alpha/opensslconf.h
index a74c6df644c..a1331118f40 100644
--- a/lib/libcrypto/arch/alpha/opensslconf.h
+++ b/lib/libcrypto/arch/alpha/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/amd64/opensslconf.h b/lib/libcrypto/arch/amd64/opensslconf.h
index cbd5d53ca4b..f8cbb5d652f 100644
--- a/lib/libcrypto/arch/amd64/opensslconf.h
+++ b/lib/libcrypto/arch/amd64/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/arm/opensslconf.h b/lib/libcrypto/arch/arm/opensslconf.h
index e3795ce73ae..78cd6d856fb 100644
--- a/lib/libcrypto/arch/arm/opensslconf.h
+++ b/lib/libcrypto/arch/arm/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/hppa/opensslconf.h b/lib/libcrypto/arch/hppa/opensslconf.h
index e3795ce73ae..78cd6d856fb 100644
--- a/lib/libcrypto/arch/hppa/opensslconf.h
+++ b/lib/libcrypto/arch/hppa/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/i386/opensslconf.h b/lib/libcrypto/arch/i386/opensslconf.h
index 37137ff1022..bf439696d89 100644
--- a/lib/libcrypto/arch/i386/opensslconf.h
+++ b/lib/libcrypto/arch/i386/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/m88k/opensslconf.h b/lib/libcrypto/arch/m88k/opensslconf.h
index e3795ce73ae..78cd6d856fb 100644
--- a/lib/libcrypto/arch/m88k/opensslconf.h
+++ b/lib/libcrypto/arch/m88k/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/mips64/opensslconf.h b/lib/libcrypto/arch/mips64/opensslconf.h
index ef7a679d94f..226951eded1 100644
--- a/lib/libcrypto/arch/mips64/opensslconf.h
+++ b/lib/libcrypto/arch/mips64/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/powerpc/opensslconf.h b/lib/libcrypto/arch/powerpc/opensslconf.h
index e3795ce73ae..78cd6d856fb 100644
--- a/lib/libcrypto/arch/powerpc/opensslconf.h
+++ b/lib/libcrypto/arch/powerpc/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/sh/opensslconf.h b/lib/libcrypto/arch/sh/opensslconf.h
index e3795ce73ae..78cd6d856fb 100644
--- a/lib/libcrypto/arch/sh/opensslconf.h
+++ b/lib/libcrypto/arch/sh/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/arch/sparc/opensslconf.h b/lib/libcrypto/arch/sparc/opensslconf.h
new file mode 100644
index 00000000000..78cd6d856fb
--- /dev/null
+++ b/lib/libcrypto/arch/sparc/opensslconf.h
@@ -0,0 +1,154 @@
+#include <openssl/opensslfeatures.h>
+/* crypto/opensslconf.h.in */
+
+#if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
+#define OPENSSLDIR "/etc/ssl"
+#endif
+
+#undef OPENSSL_UNISTD
+#define OPENSSL_UNISTD <unistd.h>
+
+#undef OPENSSL_EXPORT_VAR_AS_FUNCTION
+
+#if defined(HEADER_IDEA_H) && !defined(IDEA_INT)
+#define IDEA_INT unsigned int
+#endif
+
+#if defined(HEADER_MD2_H) && !defined(MD2_INT)
+#define MD2_INT unsigned int
+#endif
+
+#if defined(HEADER_RC2_H) && !defined(RC2_INT)
+/* I need to put in a mod for the alpha - eay */
+#define RC2_INT unsigned int
+#endif
+
+#if defined(HEADER_RC4_H)
+#if !defined(RC4_INT)
+/* using int types make the structure larger but make the code faster
+ * on most boxes I have tested - up to %20 faster. */
+/*
+ * I don't know what does "most" mean, but declaring "int" is a must on:
+ * - Intel P6 because partial register stalls are very expensive;
+ * - elder Alpha because it lacks byte load/store instructions;
+ */
+#define RC4_INT unsigned int
+#endif
+#if !defined(RC4_CHUNK)
+/*
+ * This enables code handling data aligned at natural CPU word
+ * boundary. See crypto/rc4/rc4_enc.c for further details.
+ */
+#undef RC4_CHUNK
+#endif
+#endif
+
+#if (defined(HEADER_NEW_DES_H) || defined(HEADER_DES_H)) && !defined(DES_LONG)
+/* If this is set to 'unsigned int' on a DEC Alpha, this gives about a
+ * %20 speed up (longs are 8 bytes, int's are 4). */
+#ifndef DES_LONG
+#define DES_LONG unsigned int
+#endif
+#endif
+
+#if defined(HEADER_BN_H) && !defined(CONFIG_HEADER_BN_H)
+#define CONFIG_HEADER_BN_H
+#define BN_LLONG
+
+/* Should we define BN_DIV2W here? */
+
+/* Only one for the following should be defined */
+/* The prime number generation stuff may not work when
+ * EIGHT_BIT but I don't care since I've only used this mode
+ * for debuging the bignum libraries */
+#undef SIXTY_FOUR_BIT_LONG
+#undef SIXTY_FOUR_BIT
+#define THIRTY_TWO_BIT
+#undef SIXTEEN_BIT
+#undef EIGHT_BIT
+#endif
+
+#if defined(HEADER_RC4_LOCL_H) && !defined(CONFIG_HEADER_RC4_LOCL_H)
+#define CONFIG_HEADER_RC4_LOCL_H
+/* if this is defined data[i] is used instead of *data, this is a %20
+ * speedup on x86 */
+#define RC4_INDEX
+#endif
+
+#if defined(HEADER_BF_LOCL_H) && !defined(CONFIG_HEADER_BF_LOCL_H)
+#define CONFIG_HEADER_BF_LOCL_H
+#undef BF_PTR
+#endif /* HEADER_BF_LOCL_H */
+
+#if defined(HEADER_DES_LOCL_H) && !defined(CONFIG_HEADER_DES_LOCL_H)
+#define CONFIG_HEADER_DES_LOCL_H
+#ifndef DES_DEFAULT_OPTIONS
+/* the following is tweaked from a config script, that is why it is a
+ * protected undef/define */
+#ifndef DES_PTR
+#undef DES_PTR
+#endif
+
+/* This helps C compiler generate the correct code for multiple functional
+ * units.  It reduces register dependancies at the expense of 2 more
+ * registers */
+#ifndef DES_RISC1
+#undef DES_RISC1
+#endif
+
+#ifndef DES_RISC2
+#undef DES_RISC2
+#endif
+
+#if defined(DES_RISC1) && defined(DES_RISC2)
+YOU SHOULD NOT HAVE BOTH DES_RISC1 AND DES_RISC2 DEFINED!!!!!
+#endif
+
+/* Unroll the inner loop, this sometimes helps, sometimes hinders.
+ * Very mucy CPU dependant */
+#ifndef DES_UNROLL
+#define DES_UNROLL
+#endif
+
+/* These default values were supplied by
+ * Peter Gutman <pgut001@cs.auckland.ac.nz>
+ * They are only used if nothing else has been defined */
+#if !defined(DES_PTR) && !defined(DES_RISC1) && !defined(DES_RISC2) && !defined(DES_UNROLL)
+/* Special defines which change the way the code is built depending on the
+   CPU and OS.  For SGI machines you can use _MIPS_SZLONG (32 or 64) to find
+   even newer MIPS CPU's, but at the moment one size fits all for
+   optimization options.  Older Sparc's work better with only UNROLL, but
+   there's no way to tell at compile time what it is you're running on */
+ 
+#if defined( sun )		/* Newer Sparc's */
+#  define DES_PTR
+#  define DES_RISC1
+#  define DES_UNROLL
+#elif defined( __ultrix )	/* Older MIPS */
+#  define DES_PTR
+#  define DES_RISC2
+#  define DES_UNROLL
+#elif defined( __osf1__ )	/* Alpha */
+#  define DES_PTR
+#  define DES_RISC2
+#elif defined ( _AIX )		/* RS6000 */
+  /* Unknown */
+#elif defined( __hpux )		/* HP-PA */
+  /* Unknown */
+#elif defined( __aux )		/* 68K */
+  /* Unknown */
+#elif defined( __dgux )		/* 88K (but P6 in latest boxes) */
+#  define DES_UNROLL
+#elif defined( __sgi )		/* Newer MIPS */
+#  define DES_PTR
+#  define DES_RISC2
+#  define DES_UNROLL
+#elif defined(i386) || defined(__i386__)	/* x86 boxes, should be gcc */
+#  define DES_PTR
+#  define DES_RISC1
+#  define DES_UNROLL
+#endif /* Systems-specific speed defines */
+#endif
+
+#endif /* DES_DEFAULT_OPTIONS */
+#endif /* HEADER_DES_LOCL_H */
diff --git a/lib/libcrypto/arch/sparc64/opensslconf.h b/lib/libcrypto/arch/sparc64/opensslconf.h
index ef7a679d94f..226951eded1 100644
--- a/lib/libcrypto/arch/sparc64/opensslconf.h
+++ b/lib/libcrypto/arch/sparc64/opensslconf.h
@@ -1,9 +1,6 @@
 #include <openssl/opensslfeatures.h>
 /* crypto/opensslconf.h.in */
 
-/* Generate 80386 code? */
-#undef I386_ONLY
-
 #if defined(HEADER_CRYPTLIB_H) && !defined(OPENSSLDIR)
 #define OPENSSLDIR "/etc/ssl"
 #endif
diff --git a/lib/libcrypto/cryptlib.c b/lib/libcrypto/cryptlib.c
index 0f2c5970bc9..fa091fbaeac 100644
--- a/lib/libcrypto/cryptlib.c
+++ b/lib/libcrypto/cryptlib.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: cryptlib.c,v 1.37 2015/09/13 16:56:11 miod Exp $ */
+/* $OpenBSD: cryptlib.c,v 1.38 2016/11/04 13:56:04 miod Exp $ */
 /* ====================================================================
  * Copyright (c) 1998-2006 The OpenSSL Project.  All rights reserved.
  *
@@ -635,7 +635,7 @@ OPENSSL_cpu_caps(void)
 	return *(uint64_t *)OPENSSL_ia32cap_P;
 }
 
-#if defined(OPENSSL_CPUID_OBJ) && !defined(OPENSSL_NO_ASM) && !defined(I386_ONLY)
+#if defined(OPENSSL_CPUID_OBJ) && !defined(OPENSSL_NO_ASM)
 #define OPENSSL_CPUID_SETUP
 typedef unsigned long long IA32CAP;
 void
diff --git a/lib/libcrypto/engine/eng_padlock.c b/lib/libcrypto/engine/eng_padlock.c
index 1c86a343dfe..4f2d426a068 100644
--- a/lib/libcrypto/engine/eng_padlock.c
+++ b/lib/libcrypto/engine/eng_padlock.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: eng_padlock.c,v 1.14 2015/02/07 13:19:15 doug Exp $ */
+/* $OpenBSD: eng_padlock.c,v 1.15 2016/11/04 13:56:05 miod Exp $ */
 /*
  * Support for VIA PadLock Advanced Cryptography Engine (ACE)
  * Written by Michal Ludvig <michal@logix.cz>
@@ -100,7 +100,7 @@
    In addition, because of the heavy use of inline assembler,
    compiler choice is limited to GCC and Microsoft C. */
 #undef COMPILE_HW_PADLOCK
-#if !defined(I386_ONLY) && !defined(OPENSSL_NO_INLINE_ASM)
+#if !defined(OPENSSL_NO_INLINE_ASM)
 # if (defined(__GNUC__) && (defined(__i386__) || defined(__i386)))
 #  define COMPILE_HW_PADLOCK
 # endif
diff --git a/lib/libcrypto/evp/e_aes.c b/lib/libcrypto/evp/e_aes.c
index a6d48085c3c..25199dca366 100644
--- a/lib/libcrypto/evp/e_aes.c
+++ b/lib/libcrypto/evp/e_aes.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: e_aes.c,v 1.29 2015/09/10 15:56:25 jsing Exp $ */
+/* $OpenBSD: e_aes.c,v 1.30 2016/11/04 13:56:05 miod Exp $ */
 /* ====================================================================
  * Copyright (c) 2001-2011 The OpenSSL Project.  All rights reserved.
  *
@@ -143,14 +143,14 @@ void AES_xts_decrypt(const char *inp, char *out, size_t len,
     const AES_KEY *key1, const AES_KEY *key2, const unsigned char iv[16]);
 #endif
 
-#if	defined(AES_ASM) && !defined(I386_ONLY) &&	(  \
+#if	defined(AES_ASM) &&				(  \
 	((defined(__i386)	|| defined(__i386__)	|| \
 	  defined(_M_IX86)) && defined(OPENSSL_IA32_SSE2))|| \
 	defined(__x86_64)	|| defined(__x86_64__)	|| \
 	defined(_M_AMD64)	|| defined(_M_X64)	|| \
 	defined(__INTEL__)				)
 
-extern unsigned int OPENSSL_ia32cap_P[2];
+extern unsigned int OPENSSL_ia32cap_P[];
 
 #ifdef VPAES_ASM
 #define VPAES_CAPABLE	(OPENSSL_ia32cap_P[1]&(1<<(41-32)))
diff --git a/lib/libcrypto/md32_common.h b/lib/libcrypto/md32_common.h
index bad34d22d6a..0dca61710fc 100644
--- a/lib/libcrypto/md32_common.h
+++ b/lib/libcrypto/md32_common.h
@@ -1,4 +1,4 @@
-/* $OpenBSD: md32_common.h,v 1.21 2016/09/04 14:31:29 jsing Exp $ */
+/* $OpenBSD: md32_common.h,v 1.22 2016/11/04 13:56:04 miod Exp $ */
 /* ====================================================================
  * Copyright (c) 1999-2007 The OpenSSL Project.  All rights reserved.
  *
@@ -152,8 +152,8 @@ static inline uint32_t ROTATE(uint32_t a, uint32_t n)
 #if defined(DATA_ORDER_IS_BIG_ENDIAN)
 
 #if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
-# if ((defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)) || \
-      (defined(__x86_64) || defined(__x86_64__))
+# if (defined(__i386) || defined(__i386__) || \
+      defined(__x86_64) || defined(__x86_64__))
     /*
      * This gives ~30-40% performance improvement in SHA-256 compiled
      * with gcc [on P4]. Well, first macro to be frank. We can pull
diff --git a/lib/libcrypto/modes/gcm128.c b/lib/libcrypto/modes/gcm128.c
index dd6d91e8807..6f8a8dd7f4f 100644
--- a/lib/libcrypto/modes/gcm128.c
+++ b/lib/libcrypto/modes/gcm128.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: gcm128.c,v 1.13 2015/09/10 15:56:25 jsing Exp $ */
+/* $OpenBSD: gcm128.c,v 1.14 2016/11/04 13:56:05 miod Exp $ */
 /* ====================================================================
  * Copyright (c) 2010 The OpenSSL Project.  All rights reserved.
  *
@@ -638,8 +638,7 @@ static void gcm_gmult_1bit(u64 Xi[2],const u64 H[2])
 #endif
 
 #if	TABLE_BITS==4 && defined(GHASH_ASM)
-# if	!defined(I386_ONLY) && \
-	(defined(__i386)	|| defined(__i386__)	|| \
+# if	(defined(__i386)	|| defined(__i386__)	|| \
 	 defined(__x86_64)	|| defined(__x86_64__)	|| \
 	 defined(_M_IX86)	|| defined(_M_AMD64)	|| defined(_M_X64))
 #  define GHASH_ASM_X86_OR_64
diff --git a/lib/libcrypto/modes/modes_lcl.h b/lib/libcrypto/modes/modes_lcl.h
index 8e43e480fc1..a90f4a2b37a 100644
--- a/lib/libcrypto/modes/modes_lcl.h
+++ b/lib/libcrypto/modes/modes_lcl.h
@@ -1,4 +1,4 @@
-/* $OpenBSD: modes_lcl.h,v 1.8 2014/07/10 22:45:57 jsing Exp $ */
+/* $OpenBSD: modes_lcl.h,v 1.9 2016/11/04 13:56:05 miod Exp $ */
 /* ====================================================================
  * Copyright (c) 2010 The OpenSSL Project.  All rights reserved.
  *
@@ -34,7 +34,7 @@ typedef unsigned char u8;
 #  define BSWAP4(x) ({	u32 ret=(x);			\
 			asm ("bswapl %0"		\
 			: "+r"(ret));	ret;		})
-# elif (defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)
+# elif (defined(__i386) || defined(__i386__))
 #  define BSWAP8(x) ({	u32 lo=(u64)(x)>>32,hi=(x);	\
 			asm ("bswapl %0; bswapl %1"	\
 			: "+r"(hi),"+r"(lo));		\
diff --git a/lib/libcrypto/sha/sha512.c b/lib/libcrypto/sha/sha512.c
index 7a55c0acc9d..6b95cfa72e1 100644
--- a/lib/libcrypto/sha/sha512.c
+++ b/lib/libcrypto/sha/sha512.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: sha512.c,v 1.14 2015/09/10 15:56:26 jsing Exp $ */
+/* $OpenBSD: sha512.c,v 1.15 2016/11/04 13:56:05 miod Exp $ */
 /* ====================================================================
  * Copyright (c) 2004 The OpenSSL Project.  All rights reserved
  * according to the OpenSSL license [found in ../../LICENSE].
@@ -320,23 +320,12 @@ static const SHA_LONG64 K512[80] = {
 				: "=r"(ret)			\
 				: "0"(ret)); ret;		})
 # elif (defined(__i386) || defined(__i386__))
-#  if defined(I386_ONLY)
-#   define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\
-			 unsigned int hi=p[0],lo=p[1];		\
-				asm("xchgb %%ah,%%al;xchgb %%dh,%%dl;"\
-				    "roll $16,%%eax; roll $16,%%edx; "\
-				    "xchgb %%ah,%%al;xchgb %%dh,%%dl;" \
-				: "=a"(lo),"=d"(hi)		\
-				: "0"(lo),"1"(hi) : "cc");	\
-				((SHA_LONG64)hi)<<32|lo;	})
-#  else
 #   define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\
 			 unsigned int hi=p[0],lo=p[1];		\
 				asm ("bswapl %0; bswapl %1;"	\
 				: "=r"(lo),"=r"(hi)		\
 				: "0"(lo),"1"(hi));		\
 				((SHA_LONG64)hi)<<32|lo;	})
-#  endif
 # elif (defined(_ARCH_PPC) && defined(__64BIT__)) || defined(_ARCH_PPC64)
 #  define ROTR(a,n)	({ SHA_LONG64 ret;		\
 				asm ("rotrdi %0,%1,%2"	\