Move OPENSSL_armcap changes and aarch64 dispatch tests to 1MU

crypto/fipsmodule/aes/asm/aesv8-armx.pl

@@ -96,6 +96,13 @@
 .Lenc_key:
 ___
 $code.=<<___	if ($flavour =~ /64/);
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern        BORINGSSL_function_hit
+	adrp	x9,:pg_hi21:BORINGSSL_function_hit
+	add     x9, x9, :lo12:BORINGSSL_function_hit
+	mov     w10, #1
+	strb    w10, [x9,#3] // kFlag_aes_hw_set_encrypt_key
+#endif
 	// Armv8.3-A PAuth: even though x30 is pushed to stack it is not popped later.
 	AARCH64_VALID_CALL_TARGET
 	stp	x29,x30,[sp,#-16]!
@@ -340,6 +347,17 @@ ()
 .type	${prefix}_${dir}crypt,%function
 .align	5
 ${prefix}_${dir}crypt:
+___
+$code.=<<___	if ($flavour =~ /64/);
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern        BORINGSSL_function_hit
+	adrp	x9,:pg_hi21:BORINGSSL_function_hit
+	add     x9, x9, :lo12:BORINGSSL_function_hit
+	mov     w10, #1
+	strb    w10, [x9,#1] // kFlag_aes_hw_encrypt
+#endif
+___
+$code.=<<___;
 	AARCH64_VALID_CALL_TARGET
 	ldr	$rounds,[$key,#240]
 	vld1.32	{$rndkey0},[$key],#16
@@ -719,6 +737,13 @@ ()
 ${prefix}_ctr32_encrypt_blocks:
 ___
 $code.=<<___	if ($flavour =~ /64/);
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern        BORINGSSL_function_hit
+	adrp	x9,:pg_hi21:BORINGSSL_function_hit
+	add     x9, x9, :lo12:BORINGSSL_function_hit
+	mov     w10, #1
+	strb    w10, [x9] // kFlag_aes_hw_ctr32_encrypt_blocks
+#endif
 	// Armv8.3-A PAuth: even though x30 is pushed to stack it is not popped later.
 	AARCH64_VALID_CALL_TARGET
 	stp		x29,x30,[sp,#-16]!

crypto/fipsmodule/aes/asm/vpaes-armv8.pl

@@ -261,6 +261,13 @@
 .type	vpaes_encrypt,%function
 .align	4
 vpaes_encrypt:
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern        BORINGSSL_function_hit
+	adrp	x9,:pg_hi21:BORINGSSL_function_hit
+	add     x9, x9, :lo12:BORINGSSL_function_hit
+	mov     w10, #1
+	strb    w10, [x9,#4] // kFlag_vpaes_encrypt
+#endif
 	AARCH64_SIGN_LINK_REGISTER
 	stp	x29,x30,[sp,#-16]!
 	add	x29,sp,#0
@@ -1069,6 +1076,13 @@
 .type	vpaes_set_encrypt_key,%function
 .align	4
 vpaes_set_encrypt_key:
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern        BORINGSSL_function_hit
+	adrp	x9,:pg_hi21:BORINGSSL_function_hit
+	add     x9, x9, :lo12:BORINGSSL_function_hit
+	mov     w10, #1
+	strb    w10, [x9,#5] // kFlag_vpaes_set_encrypt_key
+#endif
 	AARCH64_SIGN_LINK_REGISTER
 	stp	x29,x30,[sp,#-16]!
 	add	x29,sp,#0

crypto/fipsmodule/bcm.c

@@ -41,6 +41,7 @@
 #include "../internal.h"
 
 #include "cpucap/cpucap.c"
+#include "cpucap/cpu_aarch64.c"
 #include "cpucap/cpu_aarch64_apple.c"
 #include "cpucap/cpu_aarch64_fuchsia.c"
 #include "cpucap/cpu_aarch64_linux.c"

crypto/fipsmodule/cpucap/cpu_aarch64.c

@@ -0,0 +1,52 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0 OR ISC
+
+#if defined(OPENSSL_AARCH64) && !defined(OPENSSL_STATIC_ARMCAP)
+
+#include "cpu_aarch64.h"
+
+void handle_cpu_env(uint32_t *out, const char *in) {
+  const int invert = in[0] == '~';
+  const int or = in[0] == '|';
+  const int skip_first_byte = invert || or;
+  const int hex = in[skip_first_byte] == '0' && in[skip_first_byte+1] == 'x';
+  uint32_t armcap = out[0];
+
+  int sscanf_result;
+  uint32_t v;
+  if (hex) {
+    sscanf_result = sscanf(in + skip_first_byte + 2, "%" PRIx32, &v);
+  } else {
+    sscanf_result = sscanf(in + skip_first_byte, "%" PRIu32, &v);
+  }
+
+  if (!sscanf_result) {
+    return;
+  }
+
+  // Detect if the user is trying to use the environment variable to set
+  // a capability that is _not_ available on the CPU:
+  // If the runtime capability check (e.g via getauxval() on Linux)
+  // returned a non-zero hwcap in `armcap` (out)
+  // and a bit set in the requested `v` is not set in `armcap`,
+  // abort instead of crashing later.
+  // The case of invert cannot enable an unexisting capability;
+  // it can only disable an existing one.
+  if (!invert && armcap && (~armcap & v))
+  {
+    fprintf(stderr,
+            "Fatal Error: HW capability found: 0x%02X, but HW capability requested: 0x%02X.\n",
+            armcap, v);
+    exit(1);
+  }
+
+  if (invert) {
+    out[0] &= ~v;
+  } else if (or) {
+    out[0] |= v;
+  } else {
+    out[0] = v;
+  }
+}
+
+#endif // OPENSSL_AARCH64 && !OPENSSL_STATIC_ARMCAP

crypto/fipsmodule/cpucap/cpu_aarch64.h

@@ -0,0 +1,31 @@
+// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
+// SPDX-License-Identifier: Apache-2.0 OR ISC
+
+#ifndef OPENSSL_HEADER_CPUCAP_CPU_AARCH64_H
+#define OPENSSL_HEADER_CPUCAP_CPU_AARCH64_H
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+#include <inttypes.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#if defined(OPENSSL_AARCH64) && !defined(OPENSSL_STATIC_ARMCAP)
+
+// cpu_aarch64 contains common functions used across multiple cpu_aarch64_* files
+
+// handle_cpu_env applies the value from |in| to the CPUID values in |out[0]|.
+// See the comment in |OPENSSL_cpuid_setup| about this.
+void handle_cpu_env(uint32_t *out, const char *in);
+
+#endif // OPENSSL_AARCH64 && !OPENSSL_STATIC_ARMCAP
+
+#if defined(__cplusplus)
+}
+#endif
+
+#endif // OPENSSL_HEADER_CPUCAP_CPU_AARCH64_H

crypto/fipsmodule/cpucap/cpu_aarch64_apple.c

@@ -22,6 +22,7 @@
 
 #include <openssl/arm_arch.h>
 
+#include "cpu_aarch64.h"
 
 extern uint32_t OPENSSL_armcap_P;
 
@@ -67,6 +68,21 @@ void OPENSSL_cpuid_setup(void) {
   if (has_hw_feature("hw.optional.armv8_2_sha512")) {
     OPENSSL_armcap_P |= ARMV8_SHA512;
   }
+
+  // OPENSSL_armcap is a 32-bit, unsigned value which may start with "0x" to
+  // indicate a hex value. Prior to the 32-bit value, a '~' or '|' may be given.
+  //
+  // If the '~' prefix is present:
+  //   the value is inverted and ANDed with the probed CPUID result
+  // If the '|' prefix is present:
+  //   the value is ORed with the probed CPUID result
+  // Otherwise:
+  //   the value is taken as the result of the CPUID
+  const char *env;
+  env = getenv("OPENSSL_armcap");
+  if (env != NULL) {
+    handle_cpu_env(&OPENSSL_armcap_P, env);
+  }
 }
 
 #endif  // OPENSSL_AARCH64 && OPENSSL_APPLE && !OPENSSL_STATIC_ARMCAP

crypto/fipsmodule/cpucap/cpu_aarch64_linux.c

@@ -22,46 +22,13 @@
 #ifndef __STDC_FORMAT_MACROS
 #define __STDC_FORMAT_MACROS
 #endif
-#include <inttypes.h>
-
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
 
 #include <openssl/arm_arch.h>
 
+#include "cpu_aarch64.h"
 
 extern uint32_t OPENSSL_armcap_P;
 
-// handle_cpu_env applies the value from |in| to the CPUID values in |out[0]|
-// and |out[1]|. See the comment in |OPENSSL_cpuid_setup| about this.
-static void handle_cpu_env(uint32_t *out, const char *in) {
-  const int invert = in[0] == '~';
-  const int or = in[0] == '|';
-  const int skip_first_byte = invert || or;
-  const int hex = in[skip_first_byte] == '0' && in[skip_first_byte+1] == 'x';
-
-  int sscanf_result;
-  uint32_t v;
-  if (hex) {
-    sscanf_result = sscanf(in + invert + 2, "%" PRIx32, &v);
-  } else {
-    sscanf_result = sscanf(in + invert, "%" PRIu32, &v);
-  }
-
-  if (!sscanf_result) {
-    return;
-  }
-
-  if (invert) {
-    out[0] &= ~v;
-  } else if (or) {
-    out[0] |= v;
-  } else {
-    out[0] = v;
-  }
-}
-
 void OPENSSL_cpuid_setup(void) {
   unsigned long hwcap = getauxval(AT_HWCAP);
 

crypto/fipsmodule/sha/asm/sha512-armv8.pl

@@ -359,6 +359,13 @@ sub BODY_00_xx {
 .align	6
 sha256_block_armv8:
 .Lv8_entry:
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern        BORINGSSL_function_hit
+	adrp	x9,:pg_hi21:BORINGSSL_function_hit
+	add     x9, x9, :lo12:BORINGSSL_function_hit
+	mov     w10, #1
+	strb    w10, [x9,#6] // kFlag_sha256_hw
+#endif
 	// Armv8.3-A PAuth: even though x30 is pushed to stack it is not popped later.
 	stp		x29,x30,[sp,#-16]!
 	add		x29,sp,#0
@@ -445,6 +452,13 @@ sub BODY_00_xx {
 .align	6
 sha512_block_armv8:
 .Lv8_entry:
+#ifdef BORINGSSL_DISPATCH_TEST
+.extern        BORINGSSL_function_hit
+	adrp	x9,:pg_hi21:BORINGSSL_function_hit
+	add     x9, x9, :lo12:BORINGSSL_function_hit
+	mov     w10, #1
+	strb    w10, [x9,#2] // kFlag_sha512_hw
+#endif
 	stp		x29,x30,[sp,#-16]!
 	add		x29,sp,#0