Split bignum code out of the sparcv9cap.c

Fixes #15978

Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/16019)

crypto/aes/asm/aes-sparcv9.pl

@@ -82,7 +82,10 @@ sub _data_word()
 }
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef  __arch64__
 .register	%g2,#scratch

crypto/aes/asm/aesfx-sparcv9.pl

@@ -39,7 +39,10 @@ $output = pop and open STDOUT,">$output";
 my ($inp,$out,$key,$rounds,$tmp,$mask) = map("%o$_",(0..5));
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #define LOCALS (STACK_BIAS+STACK_FRAME)
 

crypto/aes/asm/aest4-sparcv9.pl

@@ -91,7 +91,10 @@ $::evp=1;	# if $evp is set to 0, script generates module with
 my ($inp,$out,$key,$rounds,$tmp,$mask)=map("%o$_",(0..5));
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef	__arch64__
 .register	%g2,#scratch

crypto/bn/asm/sparct4-mont.pl

@@ -86,7 +86,10 @@ require "sparcv9_modes.pl";
 $output = pop and open STDOUT,">$output";
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef	__arch64__
 .register	%g2,#scratch

crypto/bn/asm/sparcv9-gf2m.pl

@@ -38,7 +38,10 @@ $tab="%l0";
 ($lo,$hi,$b)=("%g1",$a8,"%o7"); $a=$lo;
 
 $code.=<<___;
-#include <sparc_arch.h>
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef __arch64__
 .register	%g2,#scratch

crypto/bn/asm/sparcv9-mont.pl

@@ -83,7 +83,10 @@ $tpj="%l7";
 $fname="bn_mul_mont_int";
 
 $code=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 .section	".text",#alloc,#execinstr
 

crypto/bn/asm/sparcv9a-mont.pl

@@ -126,7 +126,10 @@ $nhia="%f56"; $nhib="%f58"; $nhic="%f60"; $nhid="%f62";
 $ASI_FL16_P=0xD2;	# magic ASI value to engage 16-bit FP load
 
 $code=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 .section	".text",#alloc,#execinstr
 

crypto/bn/asm/vis3-mont.pl

@@ -31,7 +31,10 @@ $frame = "STACK_FRAME";
 $bias = "STACK_BIAS";
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef	__arch64__
 .register	%g2,#scratch

crypto/bn/bn_exp.c

@@ -29,8 +29,7 @@
 
 #undef SPARC_T4_MONT
 #if defined(OPENSSL_BN_ASM_MONT) && (defined(__sparc__) || defined(__sparc))
-# include "sparc_arch.h"
-extern unsigned int OPENSSL_sparcv9cap_P[];
+# include "crypto/sparc_arch.h"
 # define SPARC_T4_MONT
 #endif
 

crypto/bn/bn_sparc.c

@@ -0,0 +1,77 @@
+/*
+ * Copyright 2005-2021 The OpenSSL Project Authors. All Rights Reserved.
+ *
+ * Licensed under the Apache License 2.0 (the "License").  You may not use
+ * this file except in compliance with the License.  You can obtain a copy
+ * in the file LICENSE in the source distribution or at
+ * https://www.openssl.org/source/license.html
+ */
+
+#include <stdlib.h>
+#include <openssl/bn.h>
+#include "internal/cryptlib.h"
+#include "crypto/sparc_arch.h"
+#include "bn_local.h"    /* for definition of bn_mul_mont */
+
+int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+                const BN_ULONG *np, const BN_ULONG *n0, int num)
+{
+    int bn_mul_mont_vis3(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+                         const BN_ULONG *np, const BN_ULONG *n0, int num);
+    int bn_mul_mont_fpu(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+                        const BN_ULONG *np, const BN_ULONG *n0, int num);
+    int bn_mul_mont_int(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
+                        const BN_ULONG *np, const BN_ULONG *n0, int num);
+
+    if (!(num & 1) && num >= 6) {
+        if ((num & 15) == 0 && num <= 64 &&
+            (OPENSSL_sparcv9cap_P[1] & (CFR_MONTMUL | CFR_MONTSQR)) ==
+            (CFR_MONTMUL | CFR_MONTSQR)) {
+            typedef int (*bn_mul_mont_f) (BN_ULONG *rp, const BN_ULONG *ap,
+                                          const BN_ULONG *bp,
+                                          const BN_ULONG *np,
+                                          const BN_ULONG *n0);
+            int bn_mul_mont_t4_8(BN_ULONG *rp, const BN_ULONG *ap,
+                                 const BN_ULONG *bp, const BN_ULONG *np,
+                                 const BN_ULONG *n0);
+            int bn_mul_mont_t4_16(BN_ULONG *rp, const BN_ULONG *ap,
+                                  const BN_ULONG *bp, const BN_ULONG *np,
+                                  const BN_ULONG *n0);
+            int bn_mul_mont_t4_24(BN_ULONG *rp, const BN_ULONG *ap,
+                                  const BN_ULONG *bp, const BN_ULONG *np,
+                                  const BN_ULONG *n0);
+            int bn_mul_mont_t4_32(BN_ULONG *rp, const BN_ULONG *ap,
+                                  const BN_ULONG *bp, const BN_ULONG *np,
+                                  const BN_ULONG *n0);
+            static const bn_mul_mont_f funcs[4] = {
+                bn_mul_mont_t4_8, bn_mul_mont_t4_16,
+                bn_mul_mont_t4_24, bn_mul_mont_t4_32
+            };
+            bn_mul_mont_f worker = funcs[num / 16 - 1];
+
+            if ((*worker) (rp, ap, bp, np, n0))
+                return 1;
+            /* retry once and fall back */
+            if ((*worker) (rp, ap, bp, np, n0))
+                return 1;
+            return bn_mul_mont_vis3(rp, ap, bp, np, n0, num);
+        }
+        if ((OPENSSL_sparcv9cap_P[0] & SPARCV9_VIS3))
+            return bn_mul_mont_vis3(rp, ap, bp, np, n0, num);
+        else if (num >= 8 &&
+                 /*
+                  * bn_mul_mont_fpu doesn't use FMADD, we just use the
+                  * flag to detect when FPU path is preferable in cases
+                  * when current heuristics is unreliable. [it works
+                  * out because FMADD-capable processors where FPU
+                  * code path is undesirable are also VIS3-capable and
+                  * VIS3 code path takes precedence.]
+                  */
+                 ( (OPENSSL_sparcv9cap_P[0] & SPARCV9_FMADD) ||
+                   (OPENSSL_sparcv9cap_P[0] &
+                    (SPARCV9_PREFER_FPU | SPARCV9_VIS1)) ==
+                   (SPARCV9_PREFER_FPU | SPARCV9_VIS1) ))
+            return bn_mul_mont_fpu(rp, ap, bp, np, n0, num);
+    }
+    return bn_mul_mont_int(rp, ap, bp, np, n0, num);
+}

crypto/bn/build.info

@@ -40,7 +40,7 @@ IF[{- !$disabled{asm} -}]
   ENDIF
 
   $BNASM_sparcv9=asm/sparcv8plus.S sparcv9-mont.S sparcv9a-mont.S vis3-mont.S \
-          sparct4-mont.S
+          sparct4-mont.S bn_sparc.c
   $BNDEF_sparcv9=OPENSSL_BN_ASM_MONT
   $BNASM_sparcv9_ec2m=sparcv9-gf2m.S
   $BNDEF_sparcv9_ec2m=OPENSSL_BN_ASM_GF2m

crypto/camellia/asm/cmllt4-sparcv9.pl

@@ -66,7 +66,10 @@ $::evp=1;	# if $evp is set to 0, script generates module with
 my ($inp,$out,$key,$rounds,$tmp,$mask)=map("%o$_",(0..5));
 
 $code=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 .text
 

crypto/des/asm/dest4-sparcv9.pl

@@ -37,7 +37,10 @@ require "sparcv9_modes.pl";
 $output=pop and open STDOUT,">$output";
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef	__arch64__
 .register       %g2,#scratch

crypto/ec/asm/ecp_nistz256-sparcv9.pl

@@ -34,7 +34,10 @@
 $output = pop and open STDOUT,">$output";
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #define LOCALS	(STACK_BIAS+STACK_FRAME)
 #ifdef	__arch64__

crypto/evp/e_camellia.c

@@ -46,8 +46,7 @@ typedef struct {
 #if defined(AES_ASM) && (defined(__sparc) || defined(__sparc__))
 /* ---------^^^ this is not a typo, just a way to detect that
  * assembler support was in general requested... */
-# include "sparc_arch.h"
-
+# include "crypto/sparc_arch.h"
 
 static int cmll_t4_init_key(EVP_CIPHER_CTX *ctx, const unsigned char *key,
                             const unsigned char *iv, int enc)

crypto/evp/e_des.c

@@ -37,9 +37,7 @@ typedef struct {
 # if defined(AES_ASM) && (defined(__sparc) || defined(__sparc__))
 /* ----------^^^ this is not a typo, just a way to detect that
  * assembler support was in general requested... */
-#  include "sparc_arch.h"
-
-extern unsigned int OPENSSL_sparcv9cap_P[];
+#  include "crypto/sparc_arch.h"
 
 #  define SPARC_DES_CAPABLE       (OPENSSL_sparcv9cap_P[1] & CFR_DES)
 

crypto/evp/e_des3.c

@@ -40,9 +40,7 @@ typedef struct {
 # if defined(AES_ASM) && (defined(__sparc) || defined(__sparc__))
 /* ---------^^^ this is not a typo, just a way to detect that
  * assembler support was in general requested... */
-#  include "sparc_arch.h"
-
-extern unsigned int OPENSSL_sparcv9cap_P[];
+#  include "crypto/sparc_arch.h"
 
 #  define SPARC_DES_CAPABLE       (OPENSSL_sparcv9cap_P[1] & CFR_DES)
 

crypto/md5/asm/md5-sparcv9.pl

@@ -203,7 +203,10 @@ ___
 }
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef __arch64__
 .register	%g2,#scratch

crypto/modes/asm/ghash-sparcv9.pl

@@ -80,7 +80,10 @@ $inp="%i2";
 $len="%i3";
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef  __arch64__
 .register	%g2,#scratch

crypto/modes/gcm128.c

@@ -674,10 +674,9 @@ void gcm_ghash_v8(u64 Xi[2], const u128 Htable[16], const u8 *inp,
                   size_t len);
 #  endif
 # elif defined(__sparc__) || defined(__sparc)
-#  include "sparc_arch.h"
+#  include "crypto/sparc_arch.h"
 #  define GHASH_ASM_SPARC
 #  define GCM_FUNCREF_4BIT
-extern unsigned int OPENSSL_sparcv9cap_P[];
 void gcm_init_vis3(u128 Htable[16], const u64 Xi[2]);
 void gcm_gmult_vis3(u64 Xi[2], const u128 Htable[16]);
 void gcm_ghash_vis3(u64 Xi[2], const u128 Htable[16], const u8 *inp,

crypto/poly1305/asm/poly1305-sparcv9.pl

@@ -52,7 +52,10 @@ my ($h0,$h1,$h2,$h3, $t0,$t1,$t2)	= map("%o$_",(0..5,7));
 my ($d0,$d1,$d2,$d3)			= map("%g$_",(1..4));
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef	__arch64__
 .register	%g2,#scratch

crypto/sha/asm/sha1-sparcv9.pl

@@ -187,7 +187,10 @@ ___
 }
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef __arch64__
 .register	%g2,#scratch

crypto/sha/asm/sha512-sparcv9.pl

@@ -394,7 +394,10 @@ ___
 } if ($SZ==8);
 
 $code.=<<___;
-#include "sparc_arch.h"
+#ifndef __ASSEMBLER__
+# define __ASSEMBLER__ 1
+#endif
+#include "crypto/sparc_arch.h"
 
 #ifdef __arch64__
 .register	%g2,#scratch

crypto/sparcv9cap.c

@@ -16,78 +16,13 @@
 #include <unistd.h>
 #include <openssl/bn.h>
 #include "internal/cryptlib.h"
-#include "bn/bn_local.h"    /* for definition of bn_mul_mont */
-
-#include "sparc_arch.h"
+#include "crypto/sparc_arch.h"
 
 #if defined(__GNUC__) && defined(__linux)
 __attribute__ ((visibility("hidden")))
 #endif
 unsigned int OPENSSL_sparcv9cap_P[2] = { SPARCV9_TICK_PRIVILEGED, 0 };
 
-int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
-                const BN_ULONG *np, const BN_ULONG *n0, int num)
-{
-    int bn_mul_mont_vis3(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
-                         const BN_ULONG *np, const BN_ULONG *n0, int num);
-    int bn_mul_mont_fpu(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
-                        const BN_ULONG *np, const BN_ULONG *n0, int num);
-    int bn_mul_mont_int(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
-                        const BN_ULONG *np, const BN_ULONG *n0, int num);
-
-    if (!(num & 1) && num >= 6) {
-        if ((num & 15) == 0 && num <= 64 &&
-            (OPENSSL_sparcv9cap_P[1] & (CFR_MONTMUL | CFR_MONTSQR)) ==
-            (CFR_MONTMUL | CFR_MONTSQR)) {
-            typedef int (*bn_mul_mont_f) (BN_ULONG *rp, const BN_ULONG *ap,
-                                          const BN_ULONG *bp,
-                                          const BN_ULONG *np,
-                                          const BN_ULONG *n0);
-            int bn_mul_mont_t4_8(BN_ULONG *rp, const BN_ULONG *ap,
-                                 const BN_ULONG *bp, const BN_ULONG *np,
-                                 const BN_ULONG *n0);
-            int bn_mul_mont_t4_16(BN_ULONG *rp, const BN_ULONG *ap,
-                                  const BN_ULONG *bp, const BN_ULONG *np,
-                                  const BN_ULONG *n0);
-            int bn_mul_mont_t4_24(BN_ULONG *rp, const BN_ULONG *ap,
-                                  const BN_ULONG *bp, const BN_ULONG *np,
-                                  const BN_ULONG *n0);
-            int bn_mul_mont_t4_32(BN_ULONG *rp, const BN_ULONG *ap,
-                                  const BN_ULONG *bp, const BN_ULONG *np,
-                                  const BN_ULONG *n0);
-            static const bn_mul_mont_f funcs[4] = {
-                bn_mul_mont_t4_8, bn_mul_mont_t4_16,
-                bn_mul_mont_t4_24, bn_mul_mont_t4_32
-            };
-            bn_mul_mont_f worker = funcs[num / 16 - 1];
-
-            if ((*worker) (rp, ap, bp, np, n0))
-                return 1;
-            /* retry once and fall back */
-            if ((*worker) (rp, ap, bp, np, n0))
-                return 1;
-            return bn_mul_mont_vis3(rp, ap, bp, np, n0, num);
-        }
-        if ((OPENSSL_sparcv9cap_P[0] & SPARCV9_VIS3))
-            return bn_mul_mont_vis3(rp, ap, bp, np, n0, num);
-        else if (num >= 8 &&
-                 /*
-                  * bn_mul_mont_fpu doesn't use FMADD, we just use the
-                  * flag to detect when FPU path is preferable in cases
-                  * when current heuristics is unreliable. [it works
-                  * out because FMADD-capable processors where FPU
-                  * code path is undesirable are also VIS3-capable and
-                  * VIS3 code path takes precedence.]
-                  */
-                 ( (OPENSSL_sparcv9cap_P[0] & SPARCV9_FMADD) ||
-                   (OPENSSL_sparcv9cap_P[0] &
-                    (SPARCV9_PREFER_FPU | SPARCV9_VIS1)) ==
-                   (SPARCV9_PREFER_FPU | SPARCV9_VIS1) ))
-            return bn_mul_mont_fpu(rp, ap, bp, np, n0, num);
-    }
-    return bn_mul_mont_int(rp, ap, bp, np, n0, num);
-}
-
 unsigned long _sparcv9_rdtick(void);
 void _sparcv9_vis1_probe(void);
 unsigned long _sparcv9_vis1_instrument(void);

include/crypto/aes_platform.h

@@ -61,7 +61,7 @@ void AES_xts_decrypt(const unsigned char *inp, unsigned char *out, size_t len,
 
 # if defined(OPENSSL_CPUID_OBJ)
 #  if (defined(__powerpc__) || defined(__ppc__) || defined(_ARCH_PPC))
-#   include "ppc_arch.h"
+#   include "crypto/ppc_arch.h"
 #   ifdef VPAES_ASM
 #    define VPAES_CAPABLE (OPENSSL_ppccap_P & PPC_ALTIVEC)
 #   endif
@@ -233,8 +233,7 @@ void gcm_ghash_avx(u64 Xi[2], const u128 Htable[16], const u8 *in, size_t len);
 # elif defined(AES_ASM) && (defined(__sparc) || defined(__sparc__))
 
 /* Fujitsu SPARC64 X support */
-extern unsigned int OPENSSL_sparcv9cap_P[];
-#  include "sparc_arch.h"
+#  include "crypto/sparc_arch.h"
 
 #  define SPARC_AES_CAPABLE       (OPENSSL_sparcv9cap_P[1] & CFR_AES)
 #  define HWAES_CAPABLE           (OPENSSL_sparcv9cap_P[0] & SPARCV9_FJAESX)

include/crypto/cmll_platform.h

@@ -14,8 +14,7 @@
 # if defined(CMLL_ASM) && (defined(__sparc) || defined(__sparc__))
 
 /* Fujitsu SPARC64 X support */
-extern unsigned int OPENSSL_sparcv9cap_P[];
-#  include "sparc_arch.h"
+#  include "crypto/sparc_arch.h"
 
 #  ifndef OPENSSL_NO_CAMELLIA
 #   define SPARC_CMLL_CAPABLE      (OPENSSL_sparcv9cap_P[1] & CFR_CAMELLIA)

include/crypto/des_platform.h

@@ -14,8 +14,7 @@
 # if defined(DES_ASM) && (defined(__sparc) || defined(__sparc__))
 
 /* Fujitsu SPARC64 X support */
-extern unsigned int OPENSSL_sparcv9cap_P[];
-#  include "sparc_arch.h"
+#  include "crypto/sparc_arch.h"
 
 #  ifndef OPENSSL_NO_DES
 #   define SPARC_DES_CAPABLE      (OPENSSL_sparcv9cap_P[1] & CFR_DES)

crypto/sparc_arch.h → include/crypto/sparc_arch.h

@@ -1,5 +1,5 @@
 /*
- * Copyright 2012-2016 The OpenSSL Project Authors. All Rights Reserved.
+ * Copyright 2012-2021 The OpenSSL Project Authors. All Rights Reserved.
  *
  * Licensed under the Apache License 2.0 (the "License").  You may not use
  * this file except in compliance with the License.  You can obtain a copy
@@ -115,4 +115,8 @@
         mov     tmp, %o7;
 # endif
 
+# ifndef __ASSEMBLER__
+extern unsigned int OPENSSL_sparcv9cap_P[2];
+# endif
+
 #endif                          /* OSSL_CRYPTO_SPARC_ARCH_H */