bn: Move ia32cap_P references from x86_64-mont.pl to C.

Replace |bn_sqr8x_mont|'s unused |bp| parameter with a flag that
indicates whether MULX and ADX are enabled.

Bug: 673
Change-Id: I56632ad51bdc2f7f5ddd4278419d67e467b84d8f
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/65587
Reviewed-by: Bob Beck <bbe@google.com>
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>

crypto/fipsmodule/bn/asm/x86_64-mont.pl

@@ -89,8 +89,6 @@
 $code=<<___;
 .text
 
-.extern	OPENSSL_ia32cap_P
-
 .globl	bn_mul_mont_nohw
 .type	bn_mul_mont_nohw,\@function,6
 .align	16
@@ -789,7 +787,7 @@
 # int bn_sqr8x_mont(
 my $rptr="%rdi";	# const BN_ULONG *rptr,
 my $aptr="%rsi";	# const BN_ULONG *aptr,
-my $bptr="%rdx";	# not used
+my $mulx_adx_capable="%rdx"; # Different than upstream!
 my $nptr="%rcx";	# const BN_ULONG *nptr,
 my $n0  ="%r8";		# const BN_ULONG *n0);
 my $num ="%r9";		# int num, has to be divisible by 8
@@ -893,11 +891,8 @@
 ___
 $code.=<<___ if ($addx);
 #ifndef MY_ASSEMBLER_IS_TOO_OLD_FOR_512AVX
-	leaq	OPENSSL_ia32cap_P(%rip),%rax
-	mov	8(%rax),%eax
-	and	\$0x80100,%eax
-	cmp	\$0x80100,%eax
-	jne	.Lsqr8x_nox
+	test	$mulx_adx_capable,$mulx_adx_capable
+	jz	.Lsqr8x_nox
 
 	call	bn_sqrx8x_internal	# see x86_64-mont5 module
 					# %rax	top-most carry

crypto/fipsmodule/bn/bn_test.cc

@@ -2864,8 +2864,8 @@ TEST_F(BNTest, BNMulMontABI) {
               mont->n0, words);
 #if !defined(MY_ASSEMBLER_IS_TOO_OLD_FOR_512AVX)
     if (bn_sqr8x_mont_capable(words)) {
-      CHECK_ABI(bn_sqr8x_mont, r.data(), a.data(), a.data(), mont->N.d,
-                mont->n0, words);
+      CHECK_ABI(bn_sqr8x_mont, r.data(), a.data(), bn_mulx_adx_capable(),
+                mont->N.d, mont->n0, words);
     }
 #endif // !defined(MY_ASSEMBLER_IS_TOO_OLD_FOR_512AVX)
 #else

crypto/fipsmodule/bn/internal.h

@@ -407,6 +407,10 @@ int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
                 const BN_ULONG *np, const BN_ULONG *n0, size_t num);
 
 #if defined(OPENSSL_X86_64)
+OPENSSL_INLINE int bn_mulx_adx_capable(void) {
+  // MULX is in BMI2.
+  return CRYPTO_is_BMI2_capable() && CRYPTO_is_ADX_capable();
+}
 int bn_mul_mont_nohw(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
                      const BN_ULONG *np, const BN_ULONG *n0, size_t num);
 OPENSSL_INLINE int bn_mul4x_mont_capable(size_t num) {
@@ -416,16 +420,14 @@ int bn_mul4x_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
                   const BN_ULONG *np, const BN_ULONG *n0, size_t num);
 #if !defined(MY_ASSEMBLER_IS_TOO_OLD_FOR_512AVX)
 OPENSSL_INLINE int bn_mulx4x_mont_capable(size_t num) {
-  // MULX is in BMI2.
-  return bn_mul4x_mont_capable(num) && CRYPTO_is_BMI2_capable() &&
-         CRYPTO_is_ADX_capable();
+  return bn_mul4x_mont_capable(num) && bn_mulx_adx_capable();
 }
 int bn_mulx4x_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
                    const BN_ULONG *np, const BN_ULONG *n0, size_t num);
 OPENSSL_INLINE int bn_sqr8x_mont_capable(size_t num) {
   return (num >= 8) && ((num & 7) == 0);
 }
-int bn_sqr8x_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *unused_bp,
+int bn_sqr8x_mont(BN_ULONG *rp, const BN_ULONG *ap, BN_ULONG mulx_adx_capable,
                   const BN_ULONG *np, const BN_ULONG *n0, size_t num);
 #endif // !defined(MY_ASSEMBLER_IS_TOO_OLD_FOR_512AVX)
 #endif // defined(OPENSSL_X86_64)

crypto/fipsmodule/bn/montgomery.c

@@ -639,7 +639,7 @@ int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
 {
 #if !defined(MY_ASSEMBLER_IS_TOO_OLD_FOR_512AVX)
   if (ap == bp && bn_sqr8x_mont_capable(num)) {
-    return bn_sqr8x_mont(rp, ap, bp, np, n0, num);
+    return bn_sqr8x_mont(rp, ap, bn_mulx_adx_capable(), np, n0, num);
   }
   if (bn_mulx4x_mont_capable(num)) {
     return bn_mulx4x_mont(rp, ap, bp, np, n0, num);