4 files changed, 40 insertions, 148 deletions

diff --git a/src/lib/libcrypto/arch/i386/Makefile.inc b/src/lib/libcrypto/arch/i386/Makefile.inc
index 6989b35686..bfc701687e 100644
--- a/src/lib/libcrypto/arch/i386/Makefile.inc
+++ b/src/lib/libcrypto/arch/i386/Makefile.inc
@@ -1,4 +1,4 @@
-# $OpenBSD: Makefile.inc,v 1.27 2025/02/14 12:01:58 jsing Exp $
+# $OpenBSD: Makefile.inc,v 1.31 2025/06/28 12:39:10 jsing Exp $
 
 # i386-specific libcrypto build rules
 
@@ -10,23 +10,27 @@ SRCS += crypto_cpu_caps.c
 # aes
 CFLAGS+= -DAES_ASM
 SSLASM+= aes aes-586
-CFLAGS+= -DVPAES_ASM
-SSLASM+= aes vpaes-x86
 SSLASM+= aes aesni-x86
+SRCS += aes_i386.c
+
 # bn
-CFLAGS+= -DOPENSSL_IA32_SSE2
 SSLASM+= bn bn-586
 SSLASM+= bn co-586
 CFLAGS+= -DOPENSSL_BN_ASM_MONT
 SSLASM+= bn x86-mont
+
 # md5
 CFLAGS+= -DMD5_ASM
 SSLASM+= md5 md5-586
+
 # modes
 CFLAGS+= -DGHASH_ASM
 SSLASM+= modes ghash-x86
+SRCS += gcm128_i386.c
+
 # rc4
 SSLASM+= rc4 rc4-586
+
 # sha
 SSLASM+= sha sha1-586
 SSLASM+= sha sha256-586
diff --git a/src/lib/libcrypto/arch/i386/crypto_arch.h b/src/lib/libcrypto/arch/i386/crypto_arch.h
index 3df3963d0b..d2faa36e2e 100644
--- a/src/lib/libcrypto/arch/i386/crypto_arch.h
+++ b/src/lib/libcrypto/arch/i386/crypto_arch.h
@@ -1,4 +1,4 @@
-/*      $OpenBSD: crypto_arch.h,v 1.4 2025/02/14 12:01:58 jsing Exp $ */
+/*      $OpenBSD: crypto_arch.h,v 1.12 2025/07/22 09:18:02 jsing Exp $ */
 /*
  * Copyright (c) 2024 Joel Sing <jsing@openbsd.org>
  *
@@ -15,19 +15,34 @@
  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  */
 
+#include <stdint.h>
+
 #ifndef HEADER_CRYPTO_ARCH_H
 #define HEADER_CRYPTO_ARCH_H
 
 #define HAVE_CRYPTO_CPU_CAPS_INIT
-#define HAVE_CRYPTO_CPU_CAPS_IA32
+
+#ifndef __ASSEMBLER__
+extern uint64_t crypto_cpu_caps_i386;
+#endif
+
+#define CRYPTO_CPU_CAPS_I386_AES        (1ULL << 0)
+#define CRYPTO_CPU_CAPS_I386_CLMUL      (1ULL << 1)
+#define CRYPTO_CPU_CAPS_I386_MMX        (1ULL << 2)
 
 #ifndef OPENSSL_NO_ASM
 
-#define HAVE_AES_CBC_ENCRYPT_INTERNAL
 #define HAVE_AES_SET_ENCRYPT_KEY_INTERNAL
 #define HAVE_AES_SET_DECRYPT_KEY_INTERNAL
 #define HAVE_AES_ENCRYPT_INTERNAL
 #define HAVE_AES_DECRYPT_INTERNAL
+#define HAVE_AES_CBC_ENCRYPT_INTERNAL
+#define HAVE_AES_CCM64_ENCRYPT_INTERNAL
+#define HAVE_AES_CTR32_ENCRYPT_INTERNAL
+#define HAVE_AES_ECB_ENCRYPT_INTERNAL
+#define HAVE_AES_XTS_ENCRYPT_INTERNAL
+
+#define HAVE_GCM128_INIT
 
 #define HAVE_RC4_INTERNAL
 #define HAVE_RC4_SET_KEY_INTERNAL
diff --git a/src/lib/libcrypto/arch/i386/crypto_cpu_caps.c b/src/lib/libcrypto/arch/i386/crypto_cpu_caps.c
index 6bb77411af..07d60f9a3f 100644
--- a/src/lib/libcrypto/arch/i386/crypto_cpu_caps.c
+++ b/src/lib/libcrypto/arch/i386/crypto_cpu_caps.c
@@ -1,4 +1,4 @@
-/*      $OpenBSD: crypto_cpu_caps.c,v 1.3 2024/11/12 13:14:57 jsing Exp $ */
+/*      $OpenBSD: crypto_cpu_caps.c,v 1.6 2025/07/22 09:18:02 jsing Exp $ */
 /*
  * Copyright (c) 2024 Joel Sing <jsing@openbsd.org>
  *
@@ -19,11 +19,15 @@
 
 #include <openssl/crypto.h>
 
+#include "crypto_arch.h"
 #include "x86_arch.h"
 
 /* Legacy architecture specific capabilities, used by perlasm. */
 uint64_t OPENSSL_ia32cap_P;
 
+/* Machine dependent CPU capabilities. */
+uint64_t crypto_cpu_caps_i386;
+
 /* Machine independent CPU capabilities. */
 extern uint64_t crypto_cpu_caps;
 
@@ -85,17 +89,23 @@ crypto_cpu_caps_init(void)
                 caps |= CPUCAP_MASK_FXSR;
         if ((edx & IA32CAP_MASK0_HT) != 0)
                 caps |= CPUCAP_MASK_HT;
-        if ((edx & IA32CAP_MASK0_MMX) != 0)
+        if ((edx & IA32CAP_MASK0_MMX) != 0) {
                 caps |= CPUCAP_MASK_MMX;
+                crypto_cpu_caps_i386 |= CRYPTO_CPU_CAPS_I386_MMX;
+        }
         if ((edx & IA32CAP_MASK0_SSE) != 0)
                 caps |= CPUCAP_MASK_SSE;
         if ((edx & IA32CAP_MASK0_SSE2) != 0)
                 caps |= CPUCAP_MASK_SSE2;
 
-        if ((ecx & IA32CAP_MASK1_AESNI) != 0)
+        if ((ecx & IA32CAP_MASK1_AESNI) != 0) {
                 caps |= CPUCAP_MASK_AESNI;
-        if ((ecx & IA32CAP_MASK1_PCLMUL) != 0)
+                crypto_cpu_caps_i386 |= CRYPTO_CPU_CAPS_I386_AES;
+        }
+        if ((ecx & IA32CAP_MASK1_PCLMUL) != 0) {
                 caps |= CPUCAP_MASK_PCLMUL;
+                crypto_cpu_caps_i386 |= CRYPTO_CPU_CAPS_I386_CLMUL;
+        }
         if ((ecx & IA32CAP_MASK1_SSSE3) != 0)
                 caps |= CPUCAP_MASK_SSSE3;
 
@@ -112,9 +122,3 @@ crypto_cpu_caps_init(void)
 
         OPENSSL_ia32cap_P = caps;
 }
-
-uint64_t
-crypto_cpu_caps_ia32(void)
-{
-        return OPENSSL_ia32cap_P;
-}
diff --git a/src/lib/libcrypto/arch/i386/opensslconf.h b/src/lib/libcrypto/arch/i386/opensslconf.h
index 03cf31b940..dcbe113864 100644
--- a/src/lib/libcrypto/arch/i386/opensslconf.h
+++ b/src/lib/libcrypto/arch/i386/opensslconf.h
@@ -1,9 +1,4 @@
 #include <openssl/opensslfeatures.h>
-/* crypto/opensslconf.h.in */
-
-#if defined(HEADER_CRYPTO_LOCAL_H) && !defined(OPENSSLDIR)
-#define OPENSSLDIR "/etc/ssl"
-#endif
 
 #undef OPENSSL_EXPORT_VAR_AS_FUNCTION
 
@@ -17,30 +12,7 @@
 #endif
 #endif
 
-#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
@@ -49,106 +21,3 @@
 #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 long
-#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 debugging 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_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
-#define DES_PTR
-#endif
-
-/* This helps C compiler generate the correct code for multiple functional
- * units.  It reduces register dependencies at the expense of 2 more
- * registers */
-#ifndef DES_RISC1
-#define 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 much CPU dependent */
-#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 */