eden-emu
/
eden
mirror of https://git.eden-emu.dev/eden-emu/eden.git
1
0
Fork 0
Code Releases Wiki Activity
Browse Source

use stack pooled vectors 

Signed-off-by: lizzie <lizzie@eden-emu.dev>
pull/2750/head
lizzie 5 months ago
committed by crueter
parent
9b279722ec
commit
e374baffa8
5 changed files with 109 additions and 160 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 13
      src/core/crypto/aes_util.cpp
  2. 24
      src/core/crypto/ctr_encryption_layer.cpp
  3. 33
      src/core/crypto/xts_encryption_layer.cpp
  4. 31
      src/core/file_sys/fssystem/fssystem_aes_ctr_storage.cpp
  5. 56
      src/core/file_sys/fssystem/fssystem_aes_xts_storage.cpp

13
src/core/crypto/aes_util.cpp
View File

@ -73,15 +73,11 @@ void AESCipher<Key, KeySize>::Transcode(const u8* src, std::size_t size, u8* des
mbedtls_cipher_reset(context);
// Only ECB strictly requires block sized chunks.
const auto mode = mbedtls_cipher_get_cipher_mode(context);
std::size_t written = 0;
if (mode != MBEDTLS_MODE_ECB) {
if (mbedtls_cipher_get_cipher_mode(context) != MBEDTLS_MODE_ECB) {
mbedtls_cipher_update(context, src, size, dest, &written);
if (written != size) {
LOG_WARNING(Crypto, "Not all data was processed requested={:016X}, actual={:016X}.",
size, written);
}
if (written != size)
LOG_WARNING(Crypto, "Not all data was processed requested={:016X}, actual={:016X}.", size, written);
return;
}
@ -106,8 +102,7 @@ void AESCipher<Key, KeySize>::Transcode(const u8* src, std::size_t size, u8* des
std::memcpy(dest + offset, block.data(), length);
return;
}
LOG_WARNING(Crypto, "Not all data was processed requested={:016X}, actual={:016X}.",
length, written);
LOG_WARNING(Crypto, "Not all data was processed requested={:016X}, actual={:016X}.", length, written);
}
}
}

24
src/core/crypto/ctr_encryption_layer.cpp
View File

@ -12,26 +12,18 @@ CTREncryptionLayer::CTREncryptionLayer(FileSys::VirtualFile base_, Key128 key_,
: EncryptionLayer(std::move(base_)), base_offset(base_offset_), cipher(key_, Mode::CTR) {}
std::size_t CTREncryptionLayer::Read(u8* data, std::size_t length, std::size_t offset) const {
if (length == 0) {
if (length == 0)
return 0;
}
std::size_t total_read = 0;
// Handle an initial misaligned portion if needed.
const auto sector_offset = offset & 0xF;
if (sector_offset != 0) {
if (auto const sector_offset = offset & 0xF; sector_offset != 0) {
const std::size_t aligned_off = offset - sector_offset;
std::array<u8, 0x10> block{};
const std::size_t got = base->Read(block.data(), block.size(), aligned_off);
if (got == 0) {
return 0;
}
if (auto const got = base->Read(block.data(), block.size(), aligned_off); got != 0) {
UpdateIV(base_offset + aligned_off);
cipher.Transcode(block.data(), got, block.data(), Op::Decrypt);
const std::size_t to_copy = std::min<std::size_t>(length, got > sector_offset ? got - sector_offset : 0);
auto const to_copy = std::min<std::size_t>(length, got > sector_offset ? got - sector_offset : 0);
if (to_copy > 0) {
std::memcpy(data, block.data() + sector_offset, to_copy);
data += to_copy;
@ -39,12 +31,11 @@ std::size_t CTREncryptionLayer::Read(u8* data, std::size_t length, std::size_t o
length -= to_copy;
total_read += to_copy;
}
} else {
return 0;
}
if (length == 0) {
return total_read;
}
if (length > 0) {
// Now aligned to 0x10
UpdateIV(base_offset + offset);
const std::size_t got = base->Read(data, length, offset);
@ -52,6 +43,7 @@ std::size_t CTREncryptionLayer::Read(u8* data, std::size_t length, std::size_t o
cipher.Transcode(data, got, data, Op::Decrypt);
total_read += got;
}
}
return total_read;
}

33
src/core/crypto/xts_encryption_layer.cpp
View File

@ -17,29 +17,21 @@ XTSEncryptionLayer::XTSEncryptionLayer(FileSys::VirtualFile base_, Key256 key_)
: EncryptionLayer(std::move(base_)), cipher(key_, Mode::XTS) {}
std::size_t XTSEncryptionLayer::Read(u8* data, std::size_t length, std::size_t offset) const {
if (length == 0) {
if (length == 0)
return 0;
}
std::size_t total_read = 0;
const std::size_t sector_size = XTS_SECTOR_SIZE;
const std::size_t sector_offset = offset % sector_size;
// Handle initial unaligned part within a sector.
if (sector_offset != 0) {
if (auto const sector_offset = offset % XTS_SECTOR_SIZE; sector_offset != 0) {
const std::size_t aligned_off = offset - sector_offset;
std::array<u8, XTS_SECTOR_SIZE> block{};
std::size_t got = base->Read(block.data(), sector_size, aligned_off);
if (got == 0) {
return 0;
}
if (got < sector_size) {
std::memset(block.data() + got, 0, sector_size - got);
}
cipher.XTSTranscode(block.data(), sector_size, block.data(), aligned_off / sector_size,
sector_size, Op::Decrypt);
if (auto const got = base->Read(block.data(), XTS_SECTOR_SIZE, aligned_off); got > 0) {
if (got < XTS_SECTOR_SIZE)
std::memset(block.data() + got, 0, XTS_SECTOR_SIZE - got);
cipher.XTSTranscode(block.data(), XTS_SECTOR_SIZE, block.data(), aligned_off / XTS_SECTOR_SIZE,
XTS_SECTOR_SIZE, Op::Decrypt);
const std::size_t to_copy = std::min<std::size_t>(length, got > sector_offset ? got - sector_offset : 0);
auto const to_copy = std::min<std::size_t>(length, got > sector_offset ? got - sector_offset : 0);
if (to_copy > 0) {
std::memcpy(data, block.data() + sector_offset, to_copy);
data += to_copy;
@ -47,12 +39,12 @@ std::size_t XTSEncryptionLayer::Read(u8* data, std::size_t length, std::size_t o
length -= to_copy;
total_read += to_copy;
}
} else {
return 0;
}
if (length == 0) {
return total_read;
}
if (length > 0) {
// Process aligned middle inplace, in sector sized multiples.
while (length >= sector_size) {
const std::size_t req = (length / sector_size) * sector_size;
@ -74,7 +66,6 @@ std::size_t XTSEncryptionLayer::Read(u8* data, std::size_t length, std::size_t o
break;
}
}
// Handle tail within a sector, if any.
if (length > 0) {
std::array<u8, XTS_SECTOR_SIZE> block{};
@ -90,7 +81,7 @@ std::size_t XTSEncryptionLayer::Read(u8* data, std::size_t length, std::size_t o
total_read += to_copy;
}
}
}
return total_read;
}
} // namespace Core::Crypto

31
src/core/file_sys/fssystem/fssystem_aes_ctr_storage.cpp
View File

@ -4,8 +4,8 @@
// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
#include <boost/container/static_vector.hpp>
#include "common/alignment.h"
#include <vector>
#include "common/swap.h"
#include "core/file_sys/fssystem/fssystem_aes_ctr_storage.h"
#include "core/file_sys/fssystem/fssystem_utility.h"
@ -84,35 +84,24 @@ size_t AesCtrStorage::Write(const u8* buffer, size_t size, size_t offset) {
std::memcpy(ctr.data(), m_iv.data(), IvSize);
AddCounter(ctr.data(), IvSize, offset / BlockSize);
// Loop until all data is written.
size_t remaining = size;
s64 cur_offset = 0;
// Get a pooled buffer.
thread_local std::vector<u8> pooled_buffer;
if (pooled_buffer.size() < BlockSize) {
pooled_buffer.resize(BlockSize);
}
while (remaining > 0) {
// Loop until all data is written using a pooled buffer residing on the stack (blocksize = 0x10)
boost::container::static_vector<u8, BlockSize> pooled_buffer;
for (size_t remaining = size; remaining > 0; ) {
// Determine data we're writing and where.
const size_t write_size = std::min(pooled_buffer.size(), remaining);
auto const write_size = (std::min)(pooled_buffer.size(), remaining);
u8* write_buf = pooled_buffer.data();
// Encrypt the data.
// Encrypt the data and then write it.
m_cipher->SetIV(ctr);
m_cipher->Transcode(buffer, write_size, write_buf, Core::Crypto::Op::Encrypt);
m_base_storage->Write(write_buf, write_size, offset);
// Write the encrypted data.
m_base_storage->Write(write_buf, write_size, offset + cur_offset);
// Advance.
cur_offset += write_size;
// Advance next write chunk
offset += write_size;
remaining -= write_size;
if (remaining > 0) {
if (remaining > 0)
AddCounter(ctr.data(), IvSize, write_size / BlockSize);
}
}
return size;
}

56
src/core/file_sys/fssystem/fssystem_aes_xts_storage.cpp
View File

@ -6,11 +6,11 @@
#include <algorithm>
#include <array>
#include <vector>
#include <boost/container/static_vector.hpp>
#include "common/alignment.h"
#include "common/swap.h"
#include "core/file_sys/fssystem/fssystem_aes_xts_storage.h"
#include "core/file_sys/fssystem/fssystem_nca_header.h"
#include "core/file_sys/fssystem/fssystem_utility.h"
namespace FileSys {
@ -45,18 +45,12 @@ AesXtsStorage::AesXtsStorage(VirtualFile base, const void* key1, const void* key
size_t AesXtsStorage::Read(u8* buffer, size_t size, size_t offset) const {
// Allow zero-size reads.
if (size == 0) {
if (size == 0)
return size;
}
// Ensure buffer is valid.
// Ensure buffer is valid and we can only read at block aligned offsets.
ASSERT(buffer != nullptr);
// We can only read at block aligned offsets.
ASSERT(Common::IsAligned(offset, AesBlockSize));
ASSERT(Common::IsAligned(size, AesBlockSize));
// Read the data.
ASSERT(Common::IsAligned(offset, AesBlockSize) && Common::IsAligned(size, AesBlockSize));
m_base_storage->Read(buffer, size, offset);
// Setup the counter.
@ -64,29 +58,21 @@ size_t AesXtsStorage::Read(u8* buffer, size_t size, size_t offset) const {
std::memcpy(ctr.data(), m_iv.data(), IvSize);
AddCounter(ctr.data(), IvSize, offset / m_block_size);
// Handle any unaligned data before the start.
// Handle any unaligned data before the start; then read said data into a local pooled
// buffer that resides on the stack, do not use the global memory allocator this is a
// very tiny (512 bytes) buffer so should be fine to keep on the stack (Nca::XtsBlockSize wide buffer)
size_t processed_size = 0;
if ((offset % m_block_size) != 0) {
// Decrypt into our pooled stack buffer (max bound = NCA::XtsBlockSize)
boost::container::static_vector<u8, NcaHeader::XtsBlockSize> tmp_buf;
// Determine the size of the pre-data read.
const size_t skip_size =
static_cast<size_t>(offset - Common::AlignDown(offset, m_block_size));
const size_t data_size = (std::min)(size, m_block_size - skip_size);
// Decrypt into a thread-local pooled buffer to avoid per-call allocations.
{
thread_local std::vector<u8> tmp_buf;
if (tmp_buf.size() < m_block_size) {
tmp_buf.resize(m_block_size);
}
std::fill_n(tmp_buf.begin(), m_block_size, u8{0});
auto const skip_size = size_t(offset - Common::AlignDown(offset, m_block_size));
auto const data_size = (std::min)(size, m_block_size - skip_size);
std::fill_n(tmp_buf.begin(), skip_size, u8{0});
std::memcpy(tmp_buf.data() + skip_size, buffer, data_size);
m_cipher->SetIV(ctr);
m_cipher->Transcode(tmp_buf.data(), m_block_size, tmp_buf.data(),
Core::Crypto::Op::Decrypt);
m_cipher->Transcode(tmp_buf.data(), m_block_size, tmp_buf.data(), Core::Crypto::Op::Decrypt);
std::memcpy(buffer, tmp_buf.data() + skip_size, data_size);
}
AddCounter(ctr.data(), IvSize, 1);
processed_size += data_size;
@ -94,20 +80,16 @@ size_t AesXtsStorage::Read(u8* buffer, size_t size, size_t offset) const {
}
// Decrypt aligned chunks.
char* cur = reinterpret_cast<char*>(buffer) + processed_size;
size_t remaining = size - processed_size;
while (remaining > 0) {
const size_t cur_size = (std::min)(m_block_size, remaining);
auto* cur = buffer + processed_size;
for (size_t remaining = size - processed_size; remaining > 0; ) {
auto const cur_size = (std::min)(m_block_size, remaining);
m_cipher->SetIV(ctr);
m_cipher->Transcode(cur, cur_size, cur, Core::Crypto::Op::Decrypt);
auto* char_cur = reinterpret_cast<char*>(cur); //same repr cur - diff signedness
m_cipher->Transcode(char_cur, cur_size, char_cur, Core::Crypto::Op::Decrypt);
remaining -= cur_size;
cur += cur_size;
AddCounter(ctr.data(), IvSize, 1);
}
return size;
}

Write Preview
Loading…
Cancel
Save