openssl/crypto/ppccap.c
Richard Levitte 0e9725bcb9 Following the license change, modify the boilerplates in crypto/
[skip ci]

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/7827)
2018-12-06 15:32:17 +01:00

365 lines
11 KiB
C

/*
* Copyright 2009-2018 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 <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <setjmp.h>
#include <signal.h>
#include <unistd.h>
#if defined(__linux) || defined(_AIX)
# include <sys/utsname.h>
#endif
#if defined(_AIX53) /* defined even on post-5.3 */
# include <sys/systemcfg.h>
# if !defined(__power_set)
# define __power_set(a) (_system_configuration.implementation & (a))
# endif
#endif
#if defined(__APPLE__) && defined(__MACH__)
# include <sys/types.h>
# include <sys/sysctl.h>
#endif
#include <openssl/crypto.h>
#include <openssl/bn.h>
#include <internal/cryptlib.h>
#include <internal/chacha.h>
#include "bn/bn_lcl.h"
#include "ppc_arch.h"
unsigned int OPENSSL_ppccap_P = 0;
static sigset_t all_masked;
#ifdef OPENSSL_BN_ASM_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_int(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
const BN_ULONG *np, const BN_ULONG *n0, int num);
int bn_mul4x_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 < 4)
return 0;
if ((num & 3) == 0)
return bn_mul4x_mont_int(rp, ap, bp, np, n0, num);
/*
* There used to be [optional] call to bn_mul_mont_fpu64 here,
* but above subroutine is faster on contemporary processors.
* Formulation means that there might be old processors where
* FPU code path would be faster, POWER6 perhaps, but there was
* no opportunity to figure it out...
*/
return bn_mul_mont_int(rp, ap, bp, np, n0, num);
}
#endif
void sha256_block_p8(void *ctx, const void *inp, size_t len);
void sha256_block_ppc(void *ctx, const void *inp, size_t len);
void sha256_block_data_order(void *ctx, const void *inp, size_t len);
void sha256_block_data_order(void *ctx, const void *inp, size_t len)
{
OPENSSL_ppccap_P & PPC_CRYPTO207 ? sha256_block_p8(ctx, inp, len) :
sha256_block_ppc(ctx, inp, len);
}
void sha512_block_p8(void *ctx, const void *inp, size_t len);
void sha512_block_ppc(void *ctx, const void *inp, size_t len);
void sha512_block_data_order(void *ctx, const void *inp, size_t len);
void sha512_block_data_order(void *ctx, const void *inp, size_t len)
{
OPENSSL_ppccap_P & PPC_CRYPTO207 ? sha512_block_p8(ctx, inp, len) :
sha512_block_ppc(ctx, inp, len);
}
#ifndef OPENSSL_NO_CHACHA
void ChaCha20_ctr32_int(unsigned char *out, const unsigned char *inp,
size_t len, const unsigned int key[8],
const unsigned int counter[4]);
void ChaCha20_ctr32_vmx(unsigned char *out, const unsigned char *inp,
size_t len, const unsigned int key[8],
const unsigned int counter[4]);
void ChaCha20_ctr32_vsx(unsigned char *out, const unsigned char *inp,
size_t len, const unsigned int key[8],
const unsigned int counter[4]);
void ChaCha20_ctr32(unsigned char *out, const unsigned char *inp,
size_t len, const unsigned int key[8],
const unsigned int counter[4])
{
OPENSSL_ppccap_P & PPC_CRYPTO207
? ChaCha20_ctr32_vsx(out, inp, len, key, counter)
: OPENSSL_ppccap_P & PPC_ALTIVEC
? ChaCha20_ctr32_vmx(out, inp, len, key, counter)
: ChaCha20_ctr32_int(out, inp, len, key, counter);
}
#endif
#ifndef OPENSSL_NO_POLY1305
void poly1305_init_int(void *ctx, const unsigned char key[16]);
void poly1305_blocks(void *ctx, const unsigned char *inp, size_t len,
unsigned int padbit);
void poly1305_emit(void *ctx, unsigned char mac[16],
const unsigned int nonce[4]);
void poly1305_init_fpu(void *ctx, const unsigned char key[16]);
void poly1305_blocks_fpu(void *ctx, const unsigned char *inp, size_t len,
unsigned int padbit);
void poly1305_emit_fpu(void *ctx, unsigned char mac[16],
const unsigned int nonce[4]);
int poly1305_init(void *ctx, const unsigned char key[16], void *func[2]);
int poly1305_init(void *ctx, const unsigned char key[16], void *func[2])
{
if (sizeof(size_t) == 4 && (OPENSSL_ppccap_P & PPC_FPU)) {
poly1305_init_fpu(ctx, key);
func[0] = (void*)(uintptr_t)poly1305_blocks_fpu;
func[1] = (void*)(uintptr_t)poly1305_emit_fpu;
} else {
poly1305_init_int(ctx, key);
func[0] = (void*)(uintptr_t)poly1305_blocks;
func[1] = (void*)(uintptr_t)poly1305_emit;
}
return 1;
}
#endif
#ifdef ECP_NISTZ256_ASM
void ecp_nistz256_mul_mont(unsigned long res[4], const unsigned long a[4],
const unsigned long b[4]);
void ecp_nistz256_to_mont(unsigned long res[4], const unsigned long in[4]);
void ecp_nistz256_to_mont(unsigned long res[4], const unsigned long in[4])
{
static const unsigned long RR[] = { 0x0000000000000003U,
0xfffffffbffffffffU,
0xfffffffffffffffeU,
0x00000004fffffffdU };
ecp_nistz256_mul_mont(res, in, RR);
}
void ecp_nistz256_from_mont(unsigned long res[4], const unsigned long in[4]);
void ecp_nistz256_from_mont(unsigned long res[4], const unsigned long in[4])
{
static const unsigned long one[] = { 1, 0, 0, 0 };
ecp_nistz256_mul_mont(res, in, one);
}
#endif
static sigjmp_buf ill_jmp;
static void ill_handler(int sig)
{
siglongjmp(ill_jmp, sig);
}
void OPENSSL_fpu_probe(void);
void OPENSSL_ppc64_probe(void);
void OPENSSL_altivec_probe(void);
void OPENSSL_crypto207_probe(void);
void OPENSSL_madd300_probe(void);
/*
* Use a weak reference to getauxval() so we can use it if it is available
* but don't break the build if it is not. Note that this is *link-time*
* feature detection, not *run-time*. In other words if we link with
* symbol present, it's expected to be present even at run-time.
*/
#if defined(__GNUC__) && __GNUC__>=2 && defined(__ELF__)
extern unsigned long getauxval(unsigned long type) __attribute__ ((weak));
#else
static unsigned long (*getauxval) (unsigned long) = NULL;
#endif
/* I wish <sys/auxv.h> was universally available */
#define HWCAP 16 /* AT_HWCAP */
#define HWCAP_PPC64 (1U << 30)
#define HWCAP_ALTIVEC (1U << 28)
#define HWCAP_FPU (1U << 27)
#define HWCAP_POWER6_EXT (1U << 9)
#define HWCAP_VSX (1U << 7)
#define HWCAP2 26 /* AT_HWCAP2 */
#define HWCAP_VEC_CRYPTO (1U << 25)
#define HWCAP_ARCH_3_00 (1U << 23)
# if defined(__GNUC__) && __GNUC__>=2
__attribute__ ((constructor))
# endif
void OPENSSL_cpuid_setup(void)
{
char *e;
struct sigaction ill_oact, ill_act;
sigset_t oset;
static int trigger = 0;
if (trigger)
return;
trigger = 1;
if ((e = getenv("OPENSSL_ppccap"))) {
OPENSSL_ppccap_P = strtoul(e, NULL, 0);
return;
}
OPENSSL_ppccap_P = 0;
#if defined(_AIX)
OPENSSL_ppccap_P |= PPC_FPU;
if (sizeof(size_t) == 4) {
struct utsname uts;
# if defined(_SC_AIX_KERNEL_BITMODE)
if (sysconf(_SC_AIX_KERNEL_BITMODE) != 64)
return;
# endif
if (uname(&uts) != 0 || atoi(uts.version) < 6)
return;
}
# if defined(__power_set)
/*
* Value used in __power_set is a single-bit 1<<n one denoting
* specific processor class. Incidentally 0xffffffff<<n can be
* used to denote specific processor and its successors.
*/
if (sizeof(size_t) == 4) {
/* In 32-bit case PPC_FPU64 is always fastest [if option] */
if (__power_set(0xffffffffU<<13)) /* POWER5 and later */
OPENSSL_ppccap_P |= PPC_FPU64;
} else {
/* In 64-bit case PPC_FPU64 is fastest only on POWER6 */
if (__power_set(0x1U<<14)) /* POWER6 */
OPENSSL_ppccap_P |= PPC_FPU64;
}
if (__power_set(0xffffffffU<<14)) /* POWER6 and later */
OPENSSL_ppccap_P |= PPC_ALTIVEC;
if (__power_set(0xffffffffU<<16)) /* POWER8 and later */
OPENSSL_ppccap_P |= PPC_CRYPTO207;
if (__power_set(0xffffffffU<<17)) /* POWER9 and later */
OPENSSL_ppccap_P |= PPC_MADD300;
return;
# endif
#endif
#if defined(__APPLE__) && defined(__MACH__)
OPENSSL_ppccap_P |= PPC_FPU;
{
int val;
size_t len = sizeof(val);
if (sysctlbyname("hw.optional.64bitops", &val, &len, NULL, 0) == 0) {
if (val)
OPENSSL_ppccap_P |= PPC_FPU64;
}
len = sizeof(val);
if (sysctlbyname("hw.optional.altivec", &val, &len, NULL, 0) == 0) {
if (val)
OPENSSL_ppccap_P |= PPC_ALTIVEC;
}
return;
}
#endif
if (getauxval != NULL) {
unsigned long hwcap = getauxval(HWCAP);
if (hwcap & HWCAP_FPU) {
OPENSSL_ppccap_P |= PPC_FPU;
if (sizeof(size_t) == 4) {
/* In 32-bit case PPC_FPU64 is always fastest [if option] */
if (hwcap & HWCAP_PPC64)
OPENSSL_ppccap_P |= PPC_FPU64;
} else {
/* In 64-bit case PPC_FPU64 is fastest only on POWER6 */
if (hwcap & HWCAP_POWER6_EXT)
OPENSSL_ppccap_P |= PPC_FPU64;
}
}
if (hwcap & HWCAP_ALTIVEC) {
OPENSSL_ppccap_P |= PPC_ALTIVEC;
if ((hwcap & HWCAP_VSX) && (getauxval(HWCAP2) & HWCAP_VEC_CRYPTO))
OPENSSL_ppccap_P |= PPC_CRYPTO207;
}
if (hwcap & HWCAP_ARCH_3_00) {
OPENSSL_ppccap_P |= PPC_MADD300;
}
return;
}
sigfillset(&all_masked);
sigdelset(&all_masked, SIGILL);
sigdelset(&all_masked, SIGTRAP);
#ifdef SIGEMT
sigdelset(&all_masked, SIGEMT);
#endif
sigdelset(&all_masked, SIGFPE);
sigdelset(&all_masked, SIGBUS);
sigdelset(&all_masked, SIGSEGV);
memset(&ill_act, 0, sizeof(ill_act));
ill_act.sa_handler = ill_handler;
ill_act.sa_mask = all_masked;
sigprocmask(SIG_SETMASK, &ill_act.sa_mask, &oset);
sigaction(SIGILL, &ill_act, &ill_oact);
if (sigsetjmp(ill_jmp,1) == 0) {
OPENSSL_fpu_probe();
OPENSSL_ppccap_P |= PPC_FPU;
if (sizeof(size_t) == 4) {
#ifdef __linux
struct utsname uts;
if (uname(&uts) == 0 && strcmp(uts.machine, "ppc64") == 0)
#endif
if (sigsetjmp(ill_jmp, 1) == 0) {
OPENSSL_ppc64_probe();
OPENSSL_ppccap_P |= PPC_FPU64;
}
} else {
/*
* Wanted code detecting POWER6 CPU and setting PPC_FPU64
*/
}
}
if (sigsetjmp(ill_jmp, 1) == 0) {
OPENSSL_altivec_probe();
OPENSSL_ppccap_P |= PPC_ALTIVEC;
if (sigsetjmp(ill_jmp, 1) == 0) {
OPENSSL_crypto207_probe();
OPENSSL_ppccap_P |= PPC_CRYPTO207;
}
}
if (sigsetjmp(ill_jmp, 1) == 0) {
OPENSSL_madd300_probe();
OPENSSL_ppccap_P |= PPC_MADD300;
}
sigaction(SIGILL, &ill_oact, NULL);
sigprocmask(SIG_SETMASK, &oset, NULL);
}