Skip to content

Commit

Permalink
crypto: added runtime checks for SHA hardware
Browse files Browse the repository at this point in the history 
added checks for Apple Clang compilation
  • Loading branch information
@edtubbs
edtubbs committed Feb 3, 2023
1 parent 315a84d commit d6950a3
Show file tree
Hide file tree
Showing 7 changed files with 198 additions and 28 deletions.
67 changes: 58 additions & 9 deletions src/crypto/sha1.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,10 @@
# if defined(__ARM_NEON)|| defined(_MSC_VER) || defined(__GNUC__)
# include <arm_neon.h>
# endif
/** Apple Clang **/
# if defined(__APPLE__) && defined(__apple_build_version__)
# include <sys/sysctl.h>
# endif
/** GCC and LLVM Clang, but not Apple Clang */
# if defined(__GNUC__) && !defined(__apple_build_version__)
# if defined(__ARM_ACLE) || defined(__ARM_FEATURE_CRYPTO)
Expand All @@ -38,7 +42,6 @@ namespace
namespace sha1
{

#ifndef USE_AVX2
/** One round of SHA-1. */
void inline Round(uint32_t a, uint32_t& b, uint32_t c, uint32_t d, uint32_t& e, uint32_t f, uint32_t k, uint32_t w)
{
Expand All @@ -51,7 +54,6 @@ uint32_t inline f2(uint32_t b, uint32_t c, uint32_t d) { return b ^ c ^ d; }
uint32_t inline f3(uint32_t b, uint32_t c, uint32_t d) { return (b & c) | (d & (b | c)); }

uint32_t inline left(uint32_t x) { return (x << 1) | (x >> 31); }
#endif

/** Initialize SHA-1 state. */
void inline Initialize(uint32_t* s)
Expand All @@ -68,8 +70,8 @@ const uint32_t k2 = 0x6ED9EBA1ul;
const uint32_t k3 = 0x8F1BBCDCul;
const uint32_t k4 = 0xCA62C1D6ul;

/** Perform a SHA-1 transformation, processing a 64-byte chunk. */
void Transform(uint32_t* s, const unsigned char* chunk)
/** Perform a SHA-1 transformation, processing a 64-byte chunk. (ARMv8) */
void Transform_ARMV8(uint32_t* s, const unsigned char* chunk)
{
#if defined(USE_ARMV8) || defined(USE_ARMV82)
// this entire block is experimental
Expand Down Expand Up @@ -242,12 +244,23 @@ void Transform(uint32_t* s, const unsigned char* chunk)
/** Save state */
vst1q_u32(&s[0], ABCD);
s[4] = E0;
#endif
}

#elif USE_AVX2
/** Perform a SHA-1 transformation, processing a 64-byte chunk. (AVX2) */
void Transform_AVX2(uint32_t* s, const unsigned char* chunk)
{
#if USE_AVX2
// Perform SHA1 one block (Intel AVX2)
EXPERIMENTAL_FEATURE

sha1_one_block_avx2(chunk, s);
#else
#endif
}

/** Perform a SHA-1 transformation, processing a 64-byte chunk. */
void Transform(uint32_t* s, const unsigned char* chunk)
{
// Perform SHA one block (legacy)

uint32_t a = s[0], b = s[1], c = s[2], d = s[3], e = s[4];
Expand Down Expand Up @@ -343,13 +356,44 @@ void Transform(uint32_t* s, const unsigned char* chunk)
s[2] += c;
s[3] += d;
s[4] += e;
}

/** Define SHA1 hardware */
#if defined(__linux__)
#define HWCAP_SHA1 (1<<5)
#include <sys/auxv.h>
#elif defined(__WIN64__)
#include <intrin.h>
bool isAVX (void) {
int cpuinfo[4];
__cpuid(cpuinfo, 1);
return ((cpuinfo[2] & (1 << 28)) != 0);
}
#endif

/** Define a function pointer for Transform */
void (*transform_ptr) (uint32_t*, const unsigned char*) = &Transform;

/** Initialize the function pointer */
void inline Initialize_transform_ptr(void)
{
// Override the function pointer for ARMV8/AVX2
#if ((defined(USE_ARMV8) || defined(USE_ARMV82)) && defined(__APPLE__))
if (sysctlbyname("hw.optional.arm.FEAT_SHA1", NULL, NULL, NULL, 0) == 0)
transform_ptr = &Transform_ARMV8;
#elif (defined(USE_ARMV8) || defined(USE_ARMV82))
if (getauxval(AT_HWCAP) & HWCAP_SHA1)
transform_ptr = &Transform_ARMV8;
#elif USE_AVX2 && defined(__linux__)
if (__builtin_cpu_supports("avx2"))
transform_ptr = &Transform_AVX2;
#elif USE_AVX2 && defined(__WIN64__)
if (isAVX)
transform_ptr = &Transform_AVX2;
#endif
}

} // namespace sha1

} // namespace

////// SHA1
Expand All @@ -368,12 +412,12 @@ CSHA1& CSHA1::Write(const unsigned char* data, size_t len)
memcpy(buf + bufsize, data, 64 - bufsize);
bytes += 64 - bufsize;
data += 64 - bufsize;
sha1::Transform(s, buf);
sha1::transform_ptr(s, buf);
bufsize = 0;
}
while (end >= data + 64) {
// Process full chunks directly from the source.
sha1::Transform(s, data);
sha1::transform_ptr(s, data);
bytes += 64;
data += 64;
}
Expand Down Expand Up @@ -405,3 +449,8 @@ CSHA1& CSHA1::Reset()
sha1::Initialize(s);
return *this;
}

void detect_sha1_hardware()
{
sha1::Initialize_transform_ptr();
}
2 changes: 2 additions & 0 deletions src/crypto/sha1.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -25,4 +25,6 @@ class CSHA1
CSHA1& Reset();
};

void detect_sha1_hardware(void);

#endif // BITCOIN_CRYPTO_SHA1_H
69 changes: 60 additions & 9 deletions src/crypto/sha256.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -23,13 +23,16 @@
# if defined(__ARM_NEON) || defined(_MSC_VER) || defined(__GNUC__)
# include <arm_neon.h>
# endif
/** Apple Clang **/
# if defined(__APPLE__) && defined(__apple_build_version__)
# include <sys/sysctl.h>
# endif
/** GCC and LLVM Clang, but not Apple Clang */
# if defined(__GNUC__) && !defined(__apple_build_version__)
# if defined(__ARM_ACLE) || defined(__ARM_FEATURE_CRYPTO)
# include "compat/arm_acle_selector.h"
# endif
# endif
#endif /** ARM Headers */

static const uint32_t K[] =
{
Expand All @@ -50,14 +53,14 @@ static const uint32_t K[] =
0x748F82EE, 0x78A5636F, 0x84C87814, 0x8CC70208,
0x90BEFFFA, 0xA4506CEB, 0xBEF9A3F7, 0xC67178F2,
};
#endif /** ARM Headers */

// Internal implementation code.
namespace
{
/// Internal SHA-256 implementation.
namespace sha256
{
#ifndef USE_AVX2
uint32_t inline Ch(uint32_t x, uint32_t y, uint32_t z) { return z ^ (x & (y ^ z)); }
uint32_t inline Maj(uint32_t x, uint32_t y, uint32_t z) { return (x & y) | (z & (x | y)); }
uint32_t inline Sigma0(uint32_t x) { return (x >> 2 | x << 30) ^ (x >> 13 | x << 19) ^ (x >> 22 | x << 10); }
Expand All @@ -73,7 +76,6 @@ void inline Round(uint32_t a, uint32_t b, uint32_t c, uint32_t& d, uint32_t e, u
d += t1;
h = t1 + t2;
}
#endif

/** Initialize SHA-256 state. */
void inline Initialize(uint32_t* s)
Expand All @@ -88,8 +90,8 @@ void inline Initialize(uint32_t* s)
s[7] = 0x5be0cd19ul;
}

/** Perform one SHA-256 transformation, processing a 64-byte chunk. */
void Transform(uint32_t* s, const unsigned char* chunk)
/** Perform one SHA-256 transformation, processing a 64-byte chunk. (ARMv8) */
void Transform_ARMV8(uint32_t* s, const unsigned char* chunk)
{
#if defined(USE_ARMV8) || defined(USE_ARMV82)
// entire block is experimental
Expand Down Expand Up @@ -248,12 +250,22 @@ void Transform(uint32_t* s, const unsigned char* chunk)
/** Save state */
vst1q_u32(&s[0], STATE0);
vst1q_u32(&s[4], STATE1);
#endif
}

#elif USE_AVX2
/** Perform one SHA-256 transformation, processing a 64-byte chunk. (AVX2) */
void Transform_AVX2(uint32_t* s, const unsigned char* chunk)
{
#if USE_AVX2
// Perform SHA256 one block (Intel AVX2)
EXPERIMENTAL_FEATURE
sha256_one_block_avx2(chunk, s);
#else
#endif
}

/** Perform one SHA-256 transformation, processing a 64-byte chunk. */
void Transform(uint32_t* s, const unsigned char* chunk)
{
// Perform SHA256 one block (legacy)
uint32_t a = s[0], b = s[1], c = s[2], d = s[3], e = s[4], f = s[5], g = s[6], h = s[7];
uint32_t w0, w1, w2, w3, w4, w5, w6, w7, w8, w9, w10, w11, w12, w13, w14, w15;
Expand Down Expand Up @@ -334,6 +346,40 @@ void Transform(uint32_t* s, const unsigned char* chunk)
s[5] += f;
s[6] += g;
s[7] += h;
}

/** Define SHA256 hardware */
#if defined(__linux__)
#define HWCAP_SHA2 (1<<6)
#include <sys/auxv.h>
#elif defined(__WIN64__)
#include <intrin.h>
bool isAVX (void) {
int cpuinfo[4];
__cpuid(cpuinfo, 1);
return ((cpuinfo[2] & (1 << 28)) != 0);
}
#endif

/** Define a function pointer for Transform */
void (*transform_ptr) (uint32_t*, const unsigned char*) = &Transform;

/** Initialize the function pointer */
void inline Initialize_transform_ptr(void)
{
// Override the function pointer for ARMV8/AVX2
#if ((defined(USE_ARMV8) || defined(USE_ARMV82)) && defined(__APPLE__))
if (sysctlbyname("hw.optional.arm.FEAT_SHA256", NULL, NULL, NULL, 0) == 0)
transform_ptr = &Transform_ARMV8;
#elif (defined(USE_ARMV8) || defined(USE_ARMV82))
if (getauxval(AT_HWCAP) & HWCAP_SHA2)
transform_ptr = &Transform_ARMV8;
#elif USE_AVX2 && defined(__linux__)
if (__builtin_cpu_supports("avx2"))
transform_ptr = &Transform_AVX2;
#elif USE_AVX2 && defined(__WIN64__)
if (isAVX)
transform_ptr = &Transform_AVX2;
#endif
}

Expand All @@ -357,12 +403,12 @@ CSHA256& CSHA256::Write(const unsigned char* data, size_t len)
memcpy(buf + bufsize, data, 64 - bufsize);
bytes += 64 - bufsize;
data += 64 - bufsize;
sha256::Transform(s, buf);
sha256::transform_ptr(s, buf);
bufsize = 0;
}
while (end >= data + 64) {
// Process full chunks directly from the source.
sha256::Transform(s, data);
sha256::transform_ptr(s, data);
bytes += 64;
data += 64;
}
Expand Down Expand Up @@ -397,3 +443,8 @@ CSHA256& CSHA256::Reset()
sha256::Initialize(s);
return *this;
}

void detect_sha256_hardware()
{
sha256::Initialize_transform_ptr();
}
2 changes: 2 additions & 0 deletions src/crypto/sha256.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -25,4 +25,6 @@ class CSHA256
CSHA256& Reset();
};

void detect_sha256_hardware();

#endif // BITCOIN_CRYPTO_SHA256_H
Loading

0 comments on commit d6950a3

Please sign in to comment.