Move NEON dispatch in bn_mul_mont to C

This clears the last reference to OPENSSL_armcap_P from assembly!

Bug: 673
Change-Id: Id5d6115535742b2e980ed262d920ae28941841e8
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/65868
Reviewed-by: Bob Beck <bbe@google.com>
Commit-Queue: David Benjamin <davidben@google.com>

4 files changed, 39 insertions, 39 deletions

diff --git a/crypto/fipsmodule/bn/asm/armv4-mont.pl b/crypto/fipsmodule/bn/asm/armv4-mont.pl
index dcbaee5..491cc84 100644
--- a/crypto/fipsmodule/bn/asm/armv4-mont.pl
+++ b/crypto/fipsmodule/bn/asm/armv4-mont.pl
@@ -111,37 +111,13 @@ $code=<<___;
 .code	32
 #endif
 
-#if __ARM_MAX_ARCH__>=7
-.align	5
-.LOPENSSL_armcap:
-.word	OPENSSL_armcap_P-.Lbn_mul_mont
-#endif
-
-.global	bn_mul_mont
-.type	bn_mul_mont,%function
+.global	bn_mul_mont_nohw
+.type	bn_mul_mont_nohw,%function
 
 .align	5
-bn_mul_mont:
-.Lbn_mul_mont:
+bn_mul_mont_nohw:
 	ldr	ip,[sp,#4]		@ load num
 	stmdb	sp!,{r0,r2}		@ sp points at argument block
-#if __ARM_MAX_ARCH__>=7
-	tst	ip,#7
-	bne	.Lialu
-	adr	r0,.Lbn_mul_mont
-	ldr	r2,.LOPENSSL_armcap
-	ldr	r0,[r0,r2]
-#ifdef	__APPLE__
-	ldr	r0,[r0]
-#endif
-	tst	r0,#ARMV7_NEON		@ NEON available?
-	ldmia	sp, {r0,r2}
-	beq	.Lialu
-	add	sp,sp,#8
-	b	bn_mul8x_mont_neon
-.align	4
-.Lialu:
-#endif
 	cmp	ip,#2
 	mov	$num,ip			@ load num
 #ifdef	__thumb2__
@@ -292,7 +268,7 @@ bn_mul_mont:
 	moveq	pc,lr			@ be binary compatible with V4, yet
 	bx	lr			@ interoperable with Thumb ISA:-)
 #endif
-.size	bn_mul_mont,.-bn_mul_mont
+.size	bn_mul_mont_nohw,.-bn_mul_mont_nohw
 ___
 {
 my ($A0,$A1,$A2,$A3)=map("d$_",(0..3));
@@ -311,6 +287,7 @@ $code.=<<___;
 .arch	armv7-a
 .fpu	neon
 
+.global	bn_mul8x_mont_neon
 .type	bn_mul8x_mont_neon,%function
 .align	5
 bn_mul8x_mont_neon:
@@ -744,11 +721,6 @@ ___
 }
 $code.=<<___;
 .asciz	"Montgomery multiplication for ARMv4/NEON, CRYPTOGAMS by <appro\@openssl.org>"
-.align	2
-#if __ARM_MAX_ARCH__>=7
-.comm	OPENSSL_armcap_P,4,4
-.hidden	OPENSSL_armcap_P
-#endif
 ___
 
 foreach (split("\n",$code)) {
diff --git a/crypto/fipsmodule/bn/bn_test.cc b/crypto/fipsmodule/bn/bn_test.cc
index 90e0117..d62f6e4 100644
--- a/crypto/fipsmodule/bn/bn_test.cc
+++ b/crypto/fipsmodule/bn/bn_test.cc
@@ -2902,6 +2902,17 @@ TEST_F(BNTest, BNMulMontABI) {
       CHECK_ABI(bn_sqr8x_mont, r.data(), a.data(), bn_mulx_adx_capable(),
                 mont->N.d, mont->n0, words);
     }
+#elif defined(OPENSSL_ARM)
+    if (bn_mul8x_mont_neon_capable(words)) {
+      CHECK_ABI(bn_mul8x_mont_neon, r.data(), a.data(), b.data(), mont->N.d,
+                mont->n0, words);
+      CHECK_ABI(bn_mul8x_mont_neon, r.data(), a.data(), a.data(), mont->N.d,
+                mont->n0, words);
+    }
+    CHECK_ABI(bn_mul_mont_nohw, r.data(), a.data(), b.data(), mont->N.d,
+              mont->n0, words);
+    CHECK_ABI(bn_mul_mont_nohw, r.data(), a.data(), a.data(), mont->N.d,
+              mont->n0, words);
 #else
     CHECK_ABI(bn_mul_mont, r.data(), a.data(), b.data(), mont->N.d, mont->n0,
               words);
diff --git a/crypto/fipsmodule/bn/internal.h b/crypto/fipsmodule/bn/internal.h
index 363a97e..0271160 100644
--- a/crypto/fipsmodule/bn/internal.h
+++ b/crypto/fipsmodule/bn/internal.h
@@ -409,7 +409,7 @@ OPENSSL_INLINE int bn_mulx_adx_capable(void) {
 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) {
-  return (num >= 8) && ((num & 3) == 0);
+  return num >= 8 && (num & 3) == 0;
 }
 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);
@@ -419,14 +419,22 @@ OPENSSL_INLINE int bn_mulx4x_mont_capable(size_t num) {
 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);
+  return num >= 8 && (num & 7) == 0;
 }
 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(OPENSSL_X86_64)
-
+#elif defined(OPENSSL_ARM)
+OPENSSL_INLINE int bn_mul8x_mont_neon_capable(size_t num) {
+  return (num & 7) == 0 && CRYPTO_is_NEON_capable();
+}
+int bn_mul8x_mont_neon(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+                       const BN_ULONG *np, const BN_ULONG *n0, size_t num);
+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);
 #endif
 
+#endif  // OPENSSL_BN_ASM_MONT
+
 #if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64)
 #define OPENSSL_BN_ASM_MONT5
 
diff --git a/crypto/fipsmodule/bn/montgomery.c b/crypto/fipsmodule/bn/montgomery.c
index 7a4ca2f..cf483b0 100644
--- a/crypto/fipsmodule/bn/montgomery.c
+++ b/crypto/fipsmodule/bn/montgomery.c
@@ -507,8 +507,7 @@ void bn_mod_mul_montgomery_small(BN_ULONG *r, const BN_ULONG *a,
 
 #if defined(OPENSSL_BN_ASM_MONT) && defined(OPENSSL_X86_64)
 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)
-{
+                const BN_ULONG *np, const BN_ULONG *n0, size_t num) {
   if (ap == bp && bn_sqr8x_mont_capable(num)) {
     return bn_sqr8x_mont(rp, ap, bn_mulx_adx_capable(), np, n0, num);
   }
@@ -521,3 +520,13 @@ int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
   return bn_mul_mont_nohw(rp, ap, bp, np, n0, num);
 }
 #endif
+
+#if defined(OPENSSL_BN_ASM_MONT) && defined(OPENSSL_ARM)
+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 (bn_mul8x_mont_neon_capable(num)) {
+    return bn_mul8x_mont_neon(rp, ap, bp, np, n0, num);
+  }
+  return bn_mul_mont_nohw(rp, ap, bp, np, n0, num);
+}
+#endif