always use openssl for AES
Some checks are pending
Build Debian packages / ${{ matrix.dist }} (bookworm) (push) Waiting to run
Build Debian packages / ${{ matrix.dist }} (bullseye) (push) Waiting to run
Build Debian packages / ${{ matrix.dist }} (buster) (push) Waiting to run
Build on FreeBSD / with UPnP (push) Waiting to run
Build on OSX / With USE_UPNP=${{ matrix.with_upnp }} (no) (push) Waiting to run
Build on OSX / With USE_UPNP=${{ matrix.with_upnp }} (yes) (push) Waiting to run
Build on Windows / ${{ matrix.arch }} (clang-x86_64, x64-clang, clang, CLANG64) (push) Waiting to run
Build on Windows / ${{ matrix.arch }} (i686, x86, gcc, MINGW32) (push) Waiting to run
Build on Windows / ${{ matrix.arch }} (ucrt-x86_64, x64-ucrt, gcc, UCRT64) (push) Waiting to run
Build on Windows / ${{ matrix.arch }} (x86_64, x64, gcc, MINGW64) (push) Waiting to run
Build on Windows / CMake ${{ matrix.arch }} (clang-x86_64, x64-clang, clang, CLANG64) (push) Waiting to run
Build on Windows / CMake ${{ matrix.arch }} (i686, x86, gcc, MINGW32) (push) Waiting to run
Build on Windows / CMake ${{ matrix.arch }} (ucrt-x86_64, x64-ucrt, gcc, UCRT64) (push) Waiting to run
Build on Windows / CMake ${{ matrix.arch }} (x86_64, x64, gcc, MINGW64) (push) Waiting to run
Build on Windows / XP (push) Waiting to run
Build on Ubuntu / Make with USE_UPNP=${{ matrix.with_upnp }} (no) (push) Waiting to run
Build on Ubuntu / Make with USE_UPNP=${{ matrix.with_upnp }} (yes) (push) Waiting to run
Build on Ubuntu / CMake with -DWITH_UPNP=${{ matrix.with_upnp }} (OFF) (push) Waiting to run
Build on Ubuntu / CMake with -DWITH_UPNP=${{ matrix.with_upnp }} (ON) (push) Waiting to run
Build containers / Building container for ${{ matrix.platform }} (i386, linux/386) (push) Waiting to run
Build containers / Pushing merged manifest (push) Blocked by required conditions
Build containers / Building container for ${{ matrix.platform }} (amd64, linux/amd64) (push) Waiting to run
Build containers / Building container for ${{ matrix.platform }} (arm64, linux/arm64) (push) Waiting to run
Build containers / Building container for ${{ matrix.platform }} (armv7, linux/arm/v7) (push) Waiting to run

This commit is contained in:
orignal 2024-12-06 20:25:22 -05:00
parent 097813a6ca
commit 786da057f2
5 changed files with 111 additions and 586 deletions

View file

@ -19,6 +19,7 @@
#if OPENSSL_HKDF
#include <openssl/kdf.h>
#endif
#include "CPU.h"
#include "Crypto.h"
#include "Ed25519.h"
#include "I2PEndian.h"
@ -515,429 +516,105 @@ namespace crypto
}
// AES
#if SUPPORTS_AES
#define KeyExpansion256(round0,round1) \
"pshufd $0xff, %%xmm2, %%xmm2 \n" \
"movaps %%xmm1, %%xmm4 \n" \
"pslldq $4, %%xmm4 \n" \
"pxor %%xmm4, %%xmm1 \n" \
"pslldq $4, %%xmm4 \n" \
"pxor %%xmm4, %%xmm1 \n" \
"pslldq $4, %%xmm4 \n" \
"pxor %%xmm4, %%xmm1 \n" \
"pxor %%xmm2, %%xmm1 \n" \
"movaps %%xmm1, "#round0"(%[sched]) \n" \
"aeskeygenassist $0, %%xmm1, %%xmm4 \n" \
"pshufd $0xaa, %%xmm4, %%xmm2 \n" \
"movaps %%xmm3, %%xmm4 \n" \
"pslldq $4, %%xmm4 \n" \
"pxor %%xmm4, %%xmm3 \n" \
"pslldq $4, %%xmm4 \n" \
"pxor %%xmm4, %%xmm3 \n" \
"pslldq $4, %%xmm4 \n" \
"pxor %%xmm4, %%xmm3 \n" \
"pxor %%xmm2, %%xmm3 \n" \
"movaps %%xmm3, "#round1"(%[sched]) \n"
#endif
#if SUPPORTS_AES
void ECBCryptoAESNI::ExpandKey (const AESKey& key)
ECBEncryption::ECBEncryption ()
{
__asm__
(
"movups (%[key]), %%xmm1 \n"
"movups 16(%[key]), %%xmm3 \n"
"movaps %%xmm1, (%[sched]) \n"
"movaps %%xmm3, 16(%[sched]) \n"
"aeskeygenassist $1, %%xmm3, %%xmm2 \n"
KeyExpansion256(32,48)
"aeskeygenassist $2, %%xmm3, %%xmm2 \n"
KeyExpansion256(64,80)
"aeskeygenassist $4, %%xmm3, %%xmm2 \n"
KeyExpansion256(96,112)
"aeskeygenassist $8, %%xmm3, %%xmm2 \n"
KeyExpansion256(128,144)
"aeskeygenassist $16, %%xmm3, %%xmm2 \n"
KeyExpansion256(160,176)
"aeskeygenassist $32, %%xmm3, %%xmm2 \n"
KeyExpansion256(192,208)
"aeskeygenassist $64, %%xmm3, %%xmm2 \n"
// key expansion final
"pshufd $0xff, %%xmm2, %%xmm2 \n"
"movaps %%xmm1, %%xmm4 \n"
"pslldq $4, %%xmm4 \n"
"pxor %%xmm4, %%xmm1 \n"
"pslldq $4, %%xmm4 \n"
"pxor %%xmm4, %%xmm1 \n"
"pslldq $4, %%xmm4 \n"
"pxor %%xmm4, %%xmm1 \n"
"pxor %%xmm2, %%xmm1 \n"
"movups %%xmm1, 224(%[sched]) \n"
: // output
: [key]"r"((const uint8_t *)key), [sched]"r"(GetKeySchedule ()) // input
: "%xmm1", "%xmm2", "%xmm3", "%xmm4", "memory" // clogged
);
m_Ctx = EVP_CIPHER_CTX_new ();
}
#endif
#if SUPPORTS_AES
#define EncryptAES256(sched) \
"pxor (%["#sched"]), %%xmm0 \n" \
"aesenc 16(%["#sched"]), %%xmm0 \n" \
"aesenc 32(%["#sched"]), %%xmm0 \n" \
"aesenc 48(%["#sched"]), %%xmm0 \n" \
"aesenc 64(%["#sched"]), %%xmm0 \n" \
"aesenc 80(%["#sched"]), %%xmm0 \n" \
"aesenc 96(%["#sched"]), %%xmm0 \n" \
"aesenc 112(%["#sched"]), %%xmm0 \n" \
"aesenc 128(%["#sched"]), %%xmm0 \n" \
"aesenc 144(%["#sched"]), %%xmm0 \n" \
"aesenc 160(%["#sched"]), %%xmm0 \n" \
"aesenc 176(%["#sched"]), %%xmm0 \n" \
"aesenc 192(%["#sched"]), %%xmm0 \n" \
"aesenc 208(%["#sched"]), %%xmm0 \n" \
"aesenclast 224(%["#sched"]), %%xmm0 \n"
#endif
void ECBEncryption::Encrypt (const ChipherBlock * in, ChipherBlock * out)
ECBEncryption::~ECBEncryption ()
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
"movups (%[in]), %%xmm0 \n"
EncryptAES256(sched)
"movups %%xmm0, (%[out]) \n"
:
: [sched]"r"(GetKeySchedule ()), [in]"r"(in), [out]"r"(out)
: "%xmm0", "memory"
);
}
else
#endif
{
AES_encrypt (in->buf, out->buf, &m_Key);
}
if (m_Ctx)
EVP_CIPHER_CTX_free (m_Ctx);
}
void ECBEncryption::Encrypt (const uint8_t * in, uint8_t * out)
{
EVP_EncryptInit_ex (m_Ctx, EVP_aes_256_ecb(), NULL, m_Key, NULL);
EVP_CIPHER_CTX_set_padding (m_Ctx, 0);
int len;
EVP_EncryptUpdate (m_Ctx, out, &len, in, 16);
EVP_EncryptFinal_ex (m_Ctx, out + len, &len);
}
#if SUPPORTS_AES
#define DecryptAES256(sched) \
"pxor 224(%["#sched"]), %%xmm0 \n" \
"aesdec 208(%["#sched"]), %%xmm0 \n" \
"aesdec 192(%["#sched"]), %%xmm0 \n" \
"aesdec 176(%["#sched"]), %%xmm0 \n" \
"aesdec 160(%["#sched"]), %%xmm0 \n" \
"aesdec 144(%["#sched"]), %%xmm0 \n" \
"aesdec 128(%["#sched"]), %%xmm0 \n" \
"aesdec 112(%["#sched"]), %%xmm0 \n" \
"aesdec 96(%["#sched"]), %%xmm0 \n" \
"aesdec 80(%["#sched"]), %%xmm0 \n" \
"aesdec 64(%["#sched"]), %%xmm0 \n" \
"aesdec 48(%["#sched"]), %%xmm0 \n" \
"aesdec 32(%["#sched"]), %%xmm0 \n" \
"aesdec 16(%["#sched"]), %%xmm0 \n" \
"aesdeclast (%["#sched"]), %%xmm0 \n"
#endif
void ECBDecryption::Decrypt (const ChipherBlock * in, ChipherBlock * out)
ECBDecryption::ECBDecryption ()
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
"movups (%[in]), %%xmm0 \n"
DecryptAES256(sched)
"movups %%xmm0, (%[out]) \n"
:
: [sched]"r"(GetKeySchedule ()), [in]"r"(in), [out]"r"(out)
: "%xmm0", "memory"
);
}
else
#endif
{
AES_decrypt (in->buf, out->buf, &m_Key);
}
m_Ctx = EVP_CIPHER_CTX_new ();
}
ECBDecryption::~ECBDecryption ()
{
if (m_Ctx)
EVP_CIPHER_CTX_free (m_Ctx);
}
void ECBDecryption::Decrypt (const uint8_t * in, uint8_t * out)
{
EVP_DecryptInit_ex (m_Ctx, EVP_aes_256_ecb(), NULL, m_Key, NULL);
EVP_CIPHER_CTX_set_padding (m_Ctx, 0);
int len;
EVP_DecryptUpdate (m_Ctx, out, &len, in, 16);
EVP_DecryptFinal_ex (m_Ctx, out + len, &len);
}
#if SUPPORTS_AES
#define CallAESIMC(offset) \
"movaps "#offset"(%[shed]), %%xmm0 \n" \
"aesimc %%xmm0, %%xmm0 \n" \
"movaps %%xmm0, "#offset"(%[shed]) \n"
#endif
void ECBEncryption::SetKey (const AESKey& key)
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
ExpandKey (key);
}
else
#endif
{
AES_set_encrypt_key (key, 256, &m_Key);
}
CBCEncryption::CBCEncryption ()
{
m_Ctx = EVP_CIPHER_CTX_new ();
//memset ((uint8_t *)m_LastBlock, 0, 16);
}
void ECBDecryption::SetKey (const AESKey& key)
CBCEncryption::~CBCEncryption ()
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
ExpandKey (key); // expand encryption key first
// then invert it using aesimc
__asm__
(
CallAESIMC(16)
CallAESIMC(32)
CallAESIMC(48)
CallAESIMC(64)
CallAESIMC(80)
CallAESIMC(96)
CallAESIMC(112)
CallAESIMC(128)
CallAESIMC(144)
CallAESIMC(160)
CallAESIMC(176)
CallAESIMC(192)
CallAESIMC(208)
:
: [shed]"r"(GetKeySchedule ())
: "%xmm0", "memory"
);
}
else
#endif
{
AES_set_decrypt_key (key, 256, &m_Key);
}
}
void CBCEncryption::Encrypt (int numBlocks, const ChipherBlock * in, ChipherBlock * out)
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
"movups (%[iv]), %%xmm1 \n"
"1: \n"
"movups (%[in]), %%xmm0 \n"
"pxor %%xmm1, %%xmm0 \n"
EncryptAES256(sched)
"movaps %%xmm0, %%xmm1 \n"
"movups %%xmm0, (%[out]) \n"
"add $16, %[in] \n"
"add $16, %[out] \n"
"dec %[num] \n"
"jnz 1b \n"
"movups %%xmm1, (%[iv]) \n"
:
: [iv]"r"((uint8_t *)m_LastBlock), [sched]"r"(m_ECBEncryption.GetKeySchedule ()),
[in]"r"(in), [out]"r"(out), [num]"r"(numBlocks)
: "%xmm0", "%xmm1", "cc", "memory"
);
}
else
#endif
{
for (int i = 0; i < numBlocks; i++)
{
*m_LastBlock.GetChipherBlock () ^= in[i];
m_ECBEncryption.Encrypt (m_LastBlock.GetChipherBlock (), m_LastBlock.GetChipherBlock ());
out[i] = *m_LastBlock.GetChipherBlock ();
}
}
}
if (m_Ctx)
EVP_CIPHER_CTX_free (m_Ctx);
}
void CBCEncryption::Encrypt (const uint8_t * in, std::size_t len, uint8_t * out)
{
// len/16
int numBlocks = len >> 4;
if (numBlocks > 0)
Encrypt (numBlocks, (const ChipherBlock *)in, (ChipherBlock *)out);
EVP_EncryptInit_ex (m_Ctx, EVP_aes_256_cbc(), NULL, m_Key, m_IV);
EVP_CIPHER_CTX_set_padding (m_Ctx, 0);
int l;
EVP_EncryptUpdate (m_Ctx, out, &l, in, len);
EVP_EncryptFinal_ex (m_Ctx, out + l, &l);
}
void CBCEncryption::Encrypt (const uint8_t * in, uint8_t * out)
CBCDecryption::CBCDecryption ()
{
m_Ctx = EVP_CIPHER_CTX_new ();
//memset ((uint8_t *)m_IV, 0, 16);
}
CBCDecryption::~CBCDecryption ()
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
"movups (%[iv]), %%xmm1 \n"
"movups (%[in]), %%xmm0 \n"
"pxor %%xmm1, %%xmm0 \n"
EncryptAES256(sched)
"movups %%xmm0, (%[out]) \n"
"movups %%xmm0, (%[iv]) \n"
:
: [iv]"r"((uint8_t *)m_LastBlock), [sched]"r"(m_ECBEncryption.GetKeySchedule ()),
[in]"r"(in), [out]"r"(out)
: "%xmm0", "%xmm1", "memory"
);
}
else
#endif
Encrypt (1, (const ChipherBlock *)in, (ChipherBlock *)out);
}
void CBCDecryption::Decrypt (int numBlocks, const ChipherBlock * in, ChipherBlock * out)
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
"movups (%[iv]), %%xmm1 \n"
"1: \n"
"movups (%[in]), %%xmm0 \n"
"movaps %%xmm0, %%xmm2 \n"
DecryptAES256(sched)
"pxor %%xmm1, %%xmm0 \n"
"movups %%xmm0, (%[out]) \n"
"movaps %%xmm2, %%xmm1 \n"
"add $16, %[in] \n"
"add $16, %[out] \n"
"dec %[num] \n"
"jnz 1b \n"
"movups %%xmm1, (%[iv]) \n"
:
: [iv]"r"((uint8_t *)m_IV), [sched]"r"(m_ECBDecryption.GetKeySchedule ()),
[in]"r"(in), [out]"r"(out), [num]"r"(numBlocks)
: "%xmm0", "%xmm1", "%xmm2", "cc", "memory"
);
}
else
#endif
{
for (int i = 0; i < numBlocks; i++)
{
ChipherBlock tmp = in[i];
m_ECBDecryption.Decrypt (in + i, out + i);
out[i] ^= *m_IV.GetChipherBlock ();
*m_IV.GetChipherBlock () = tmp;
}
}
}
if (m_Ctx)
EVP_CIPHER_CTX_free (m_Ctx);
}
void CBCDecryption::Decrypt (const uint8_t * in, std::size_t len, uint8_t * out)
{
int numBlocks = len >> 4;
if (numBlocks > 0)
Decrypt (numBlocks, (const ChipherBlock *)in, (ChipherBlock *)out);
}
void CBCDecryption::Decrypt (const uint8_t * in, uint8_t * out)
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
"movups (%[iv]), %%xmm1 \n"
"movups (%[in]), %%xmm0 \n"
"movups %%xmm0, (%[iv]) \n"
DecryptAES256(sched)
"pxor %%xmm1, %%xmm0 \n"
"movups %%xmm0, (%[out]) \n"
:
: [iv]"r"((uint8_t *)m_IV), [sched]"r"(m_ECBDecryption.GetKeySchedule ()),
[in]"r"(in), [out]"r"(out)
: "%xmm0", "%xmm1", "memory"
);
}
else
#endif
Decrypt (1, (const ChipherBlock *)in, (ChipherBlock *)out);
// len/16
EVP_DecryptInit_ex (m_Ctx, EVP_aes_256_cbc(), NULL, m_Key, m_IV);
EVP_CIPHER_CTX_set_padding (m_Ctx, 0);
int l;
EVP_DecryptUpdate (m_Ctx, out, &l, in, len);
EVP_DecryptFinal_ex (m_Ctx, out + l, &l);
}
void TunnelEncryption::Encrypt (const uint8_t * in, uint8_t * out)
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
// encrypt IV
"movups (%[in]), %%xmm0 \n"
EncryptAES256(sched_iv)
"movaps %%xmm0, %%xmm1 \n"
// double IV encryption
EncryptAES256(sched_iv)
"movups %%xmm0, (%[out]) \n"
// encrypt data, IV is xmm1
"1: \n"
"add $16, %[in] \n"
"add $16, %[out] \n"
"movups (%[in]), %%xmm0 \n"
"pxor %%xmm1, %%xmm0 \n"
EncryptAES256(sched_l)
"movaps %%xmm0, %%xmm1 \n"
"movups %%xmm0, (%[out]) \n"
"dec %[num] \n"
"jnz 1b \n"
:
: [sched_iv]"r"(m_IVEncryption.GetKeySchedule ()), [sched_l]"r"(m_LayerEncryption.ECB().GetKeySchedule ()),
[in]"r"(in), [out]"r"(out), [num]"r"(63) // 63 blocks = 1008 bytes
: "%xmm0", "%xmm1", "cc", "memory"
);
}
else
#endif
{
m_IVEncryption.Encrypt ((const ChipherBlock *)in, (ChipherBlock *)out); // iv
m_LayerEncryption.SetIV (out);
m_LayerEncryption.Encrypt (in + 16, i2p::tunnel::TUNNEL_DATA_ENCRYPTED_SIZE, out + 16); // data
m_IVEncryption.Encrypt ((ChipherBlock *)out, (ChipherBlock *)out); // double iv
}
m_IVEncryption.Encrypt (in, out); // iv
m_LayerEncryption.SetIV (out);
m_LayerEncryption.Encrypt (in + 16, i2p::tunnel::TUNNEL_DATA_ENCRYPTED_SIZE, out + 16); // data
m_IVEncryption.Encrypt (out, out); // double iv
}
void TunnelDecryption::Decrypt (const uint8_t * in, uint8_t * out)
{
#if SUPPORTS_AES
if(i2p::cpu::aesni)
{
__asm__
(
// decrypt IV
"movups (%[in]), %%xmm0 \n"
DecryptAES256(sched_iv)
"movaps %%xmm0, %%xmm1 \n"
// double IV encryption
DecryptAES256(sched_iv)
"movups %%xmm0, (%[out]) \n"
// decrypt data, IV is xmm1
"1: \n"
"add $16, %[in] \n"
"add $16, %[out] \n"
"movups (%[in]), %%xmm0 \n"
"movaps %%xmm0, %%xmm2 \n"
DecryptAES256(sched_l)
"pxor %%xmm1, %%xmm0 \n"
"movups %%xmm0, (%[out]) \n"
"movaps %%xmm2, %%xmm1 \n"
"dec %[num] \n"
"jnz 1b \n"
:
: [sched_iv]"r"(m_IVDecryption.GetKeySchedule ()), [sched_l]"r"(m_LayerDecryption.ECB().GetKeySchedule ()),
[in]"r"(in), [out]"r"(out), [num]"r"(63) // 63 blocks = 1008 bytes
: "%xmm0", "%xmm1", "%xmm2", "cc", "memory"
);
}
else
#endif
{
m_IVDecryption.Decrypt ((const ChipherBlock *)in, (ChipherBlock *)out); // iv
m_LayerDecryption.SetIV (out);
m_LayerDecryption.Decrypt (in + 16, i2p::tunnel::TUNNEL_DATA_ENCRYPTED_SIZE, out + 16); // data
m_IVDecryption.Decrypt ((ChipherBlock *)out, (ChipherBlock *)out); // double iv
}
m_IVDecryption.Decrypt (in, out); // iv
m_LayerDecryption.SetIV (out);
m_LayerDecryption.Decrypt (in + 16, i2p::tunnel::TUNNEL_DATA_ENCRYPTED_SIZE, out + 16); // data
m_IVDecryption.Decrypt (out, out); // double iv
}
// AEAD/ChaCha20/Poly1305
@ -1159,7 +836,6 @@ namespace crypto
void InitCrypto (bool precomputation, bool aesni, bool force)
{
i2p::cpu::Detect (aesni, force);
/* auto numLocks = CRYPTO_num_locks();
for (int i = 0; i < numLocks; i++)
m_OpenSSLMutexes.emplace_back (new std::mutex);