mirror of
https://github.com/dolphin-emu/dolphin.git
synced 2024-12-28 01:24:43 +01:00
335cf4f2db
A deep-copy method CopyReader has been added to BlobReader (virtual) and all of its subclasses (override). This should create a second BlobReader to open the same set of data but with an independent read pointer so that it doesn't interfere with any reads done on the original Reader. As part of this, IOFile has added code to create a deep copy IOFile pointer onto the same file, with code based on the platform in question to find the file ID from the file pointer and open a new one. There has also been a small piece added to FileInfo to enable a deep copy, but its only subclass at this time already had a copy constructor so this was relatively minor.
316 lines
9.1 KiB
C++
316 lines
9.1 KiB
C++
// Copyright 2022 Dolphin Emulator Project
|
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
|
|
|
#include "DiscIO/NFSBlob.h"
|
|
|
|
#include <algorithm>
|
|
#include <array>
|
|
#include <cstring>
|
|
#include <memory>
|
|
#include <string>
|
|
#include <string_view>
|
|
#include <utility>
|
|
#include <vector>
|
|
|
|
#include <fmt/format.h>
|
|
|
|
#include "Common/Align.h"
|
|
#include "Common/CommonTypes.h"
|
|
#include "Common/Crypto/AES.h"
|
|
#include "Common/IOFile.h"
|
|
#include "Common/Logging/Log.h"
|
|
#include "Common/StringUtil.h"
|
|
#include "Common/Swap.h"
|
|
|
|
namespace DiscIO
|
|
{
|
|
bool NFSFileReader::ReadKey(const std::string& path, const std::string& directory, Key* key_out)
|
|
{
|
|
const std::string_view directory_without_trailing_slash =
|
|
std::string_view(directory).substr(0, directory.size() - 1);
|
|
|
|
std::string parent, parent_name, parent_extension;
|
|
SplitPath(directory_without_trailing_slash, &parent, &parent_name, &parent_extension);
|
|
|
|
if (parent_name + parent_extension != "content")
|
|
{
|
|
ERROR_LOG_FMT(DISCIO, "hif_000000.nfs is not in a directory named 'content': {}", path);
|
|
return false;
|
|
}
|
|
|
|
const std::string key_path = parent + "code/htk.bin";
|
|
File::IOFile key_file(key_path, "rb");
|
|
if (!key_file.ReadBytes(key_out->data(), key_out->size()))
|
|
{
|
|
ERROR_LOG_FMT(DISCIO, "Failed to read from {}", key_path);
|
|
return false;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
std::vector<NFSLBARange> NFSFileReader::GetLBARanges(const NFSHeader& header)
|
|
{
|
|
const size_t lba_range_count =
|
|
std::min<size_t>(Common::swap32(header.lba_range_count), header.lba_ranges.size());
|
|
|
|
std::vector<NFSLBARange> lba_ranges;
|
|
lba_ranges.reserve(lba_range_count);
|
|
|
|
for (size_t i = 0; i < lba_range_count; ++i)
|
|
{
|
|
const NFSLBARange& unswapped_lba_range = header.lba_ranges[i];
|
|
lba_ranges.push_back(NFSLBARange{Common::swap32(unswapped_lba_range.start_block),
|
|
Common::swap32(unswapped_lba_range.num_blocks)});
|
|
}
|
|
|
|
return lba_ranges;
|
|
}
|
|
|
|
std::vector<File::IOFile> NFSFileReader::OpenFiles(const std::string& directory,
|
|
File::IOFile first_file, u64 expected_raw_size,
|
|
u64* raw_size_out)
|
|
{
|
|
const u64 file_count = Common::AlignUp(expected_raw_size, MAX_FILE_SIZE) / MAX_FILE_SIZE;
|
|
|
|
std::vector<File::IOFile> files;
|
|
files.reserve(file_count);
|
|
|
|
*raw_size_out = first_file.GetSize();
|
|
files.emplace_back(std::move(first_file));
|
|
|
|
for (u64 i = 1; i < file_count; ++i)
|
|
{
|
|
const std::string child_path = fmt::format("{}hif_{:06}.nfs", directory, i);
|
|
File::IOFile child(child_path, "rb");
|
|
if (!child)
|
|
{
|
|
ERROR_LOG_FMT(DISCIO, "Failed to open {}", child_path);
|
|
return {};
|
|
}
|
|
|
|
*raw_size_out += child.GetSize();
|
|
files.emplace_back(std::move(child));
|
|
}
|
|
|
|
if (*raw_size_out < expected_raw_size)
|
|
{
|
|
ERROR_LOG_FMT(
|
|
DISCIO,
|
|
"Expected sum of NFS file sizes for {} to be at least {} bytes, but it was {} bytes",
|
|
directory, expected_raw_size, *raw_size_out);
|
|
return {};
|
|
}
|
|
|
|
return files;
|
|
}
|
|
|
|
u64 NFSFileReader::CalculateExpectedRawSize(const std::vector<NFSLBARange>& lba_ranges)
|
|
{
|
|
u64 total_blocks = 0;
|
|
for (const NFSLBARange& range : lba_ranges)
|
|
total_blocks += range.num_blocks;
|
|
|
|
return sizeof(NFSHeader) + total_blocks * BLOCK_SIZE;
|
|
}
|
|
|
|
u64 NFSFileReader::CalculateExpectedDataSize(const std::vector<NFSLBARange>& lba_ranges)
|
|
{
|
|
u32 greatest_block_index = 0;
|
|
for (const NFSLBARange& range : lba_ranges)
|
|
greatest_block_index = std::max(greatest_block_index, range.start_block + range.num_blocks);
|
|
|
|
return u64(greatest_block_index) * BLOCK_SIZE;
|
|
}
|
|
|
|
std::unique_ptr<NFSFileReader> NFSFileReader::Create(File::IOFile first_file,
|
|
const std::string& path)
|
|
{
|
|
std::string directory, filename, extension;
|
|
SplitPath(path, &directory, &filename, &extension);
|
|
if (filename + extension != "hif_000000.nfs")
|
|
return nullptr;
|
|
|
|
std::array<u8, 16> key;
|
|
if (!ReadKey(path, directory, &key))
|
|
return nullptr;
|
|
|
|
NFSHeader header;
|
|
if (!first_file.Seek(0, File::SeekOrigin::Begin) || !first_file.ReadArray(&header, 1) ||
|
|
header.magic != NFS_MAGIC)
|
|
{
|
|
return nullptr;
|
|
}
|
|
|
|
std::vector<NFSLBARange> lba_ranges = GetLBARanges(header);
|
|
|
|
const u64 expected_raw_size = CalculateExpectedRawSize(lba_ranges);
|
|
|
|
u64 raw_size;
|
|
std::vector<File::IOFile> files =
|
|
OpenFiles(directory, std::move(first_file), expected_raw_size, &raw_size);
|
|
|
|
if (files.empty())
|
|
return nullptr;
|
|
|
|
return std::unique_ptr<NFSFileReader>(
|
|
new NFSFileReader(std::move(lba_ranges), std::move(files), key, raw_size));
|
|
}
|
|
|
|
NFSFileReader::NFSFileReader(std::vector<NFSLBARange> lba_ranges, std::vector<File::IOFile> files,
|
|
Key key, u64 raw_size)
|
|
: m_lba_ranges(std::move(lba_ranges)), m_files(std::move(files)),
|
|
m_aes_context(Common::AES::CreateContextDecrypt(key.data())), m_raw_size(raw_size), m_key(key)
|
|
{
|
|
m_data_size = CalculateExpectedDataSize(m_lba_ranges);
|
|
}
|
|
|
|
std::unique_ptr<BlobReader> NFSFileReader::CopyReader() const
|
|
{
|
|
std::vector<File::IOFile> new_files{};
|
|
for (const File::IOFile& file : m_files)
|
|
new_files.push_back(file.Duplicate("rb"));
|
|
return std::unique_ptr<NFSFileReader>(
|
|
new NFSFileReader(m_lba_ranges, std::move(new_files), m_key, m_raw_size));
|
|
}
|
|
|
|
u64 NFSFileReader::GetDataSize() const
|
|
{
|
|
return m_data_size;
|
|
}
|
|
|
|
u64 NFSFileReader::GetRawSize() const
|
|
{
|
|
return m_raw_size;
|
|
}
|
|
|
|
u64 NFSFileReader::ToPhysicalBlockIndex(u64 logical_block_index)
|
|
{
|
|
u64 physical_blocks_so_far = 0;
|
|
|
|
for (const NFSLBARange& range : m_lba_ranges)
|
|
{
|
|
if (logical_block_index >= range.start_block &&
|
|
logical_block_index < range.start_block + range.num_blocks)
|
|
{
|
|
return physical_blocks_so_far + (logical_block_index - range.start_block);
|
|
}
|
|
|
|
physical_blocks_so_far += range.num_blocks;
|
|
}
|
|
|
|
return std::numeric_limits<u64>::max();
|
|
}
|
|
|
|
bool NFSFileReader::ReadEncryptedBlock(u64 physical_block_index)
|
|
{
|
|
constexpr u64 BLOCKS_PER_FILE = MAX_FILE_SIZE / BLOCK_SIZE;
|
|
|
|
const u64 file_index = physical_block_index / BLOCKS_PER_FILE;
|
|
const u64 block_in_file = physical_block_index % BLOCKS_PER_FILE;
|
|
|
|
if (block_in_file == BLOCKS_PER_FILE - 1)
|
|
{
|
|
// Special case. Because of the 0x200 byte header at the very beginning,
|
|
// the last block of each file has its last 0x200 bytes stored in the next file.
|
|
|
|
constexpr size_t PART_1_SIZE = BLOCK_SIZE - sizeof(NFSHeader);
|
|
constexpr size_t PART_2_SIZE = sizeof(NFSHeader);
|
|
|
|
File::IOFile& file_1 = m_files[file_index];
|
|
File::IOFile& file_2 = m_files[file_index + 1];
|
|
|
|
if (!file_1.Seek(sizeof(NFSHeader) + block_in_file * BLOCK_SIZE, File::SeekOrigin::Begin) ||
|
|
!file_1.ReadBytes(m_current_block_encrypted.data(), PART_1_SIZE))
|
|
{
|
|
file_1.ClearError();
|
|
return false;
|
|
}
|
|
|
|
if (!file_2.Seek(0, File::SeekOrigin::Begin) ||
|
|
!file_2.ReadBytes(m_current_block_encrypted.data() + PART_1_SIZE, PART_2_SIZE))
|
|
{
|
|
file_2.ClearError();
|
|
return false;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
// Normal case. The read is offset by 0x200 bytes, but it's all within one file.
|
|
|
|
File::IOFile& file = m_files[file_index];
|
|
|
|
if (!file.Seek(sizeof(NFSHeader) + block_in_file * BLOCK_SIZE, File::SeekOrigin::Begin) ||
|
|
!file.ReadBytes(m_current_block_encrypted.data(), BLOCK_SIZE))
|
|
{
|
|
file.ClearError();
|
|
return false;
|
|
}
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
void NFSFileReader::DecryptBlock(u64 logical_block_index)
|
|
{
|
|
std::array<u8, 16> iv{};
|
|
const u64 swapped_block_index = Common::swap64(logical_block_index);
|
|
std::memcpy(iv.data() + iv.size() - sizeof(swapped_block_index), &swapped_block_index,
|
|
sizeof(swapped_block_index));
|
|
|
|
m_aes_context->Crypt(iv.data(), m_current_block_encrypted.data(),
|
|
m_current_block_decrypted.data(), BLOCK_SIZE);
|
|
}
|
|
|
|
bool NFSFileReader::ReadAndDecryptBlock(u64 logical_block_index)
|
|
{
|
|
const u64 physical_block_index = ToPhysicalBlockIndex(logical_block_index);
|
|
|
|
if (physical_block_index == std::numeric_limits<u64>::max())
|
|
{
|
|
// The block isn't physically present. Treat its contents as all zeroes.
|
|
m_current_block_decrypted.fill(0);
|
|
}
|
|
else
|
|
{
|
|
if (!ReadEncryptedBlock(physical_block_index))
|
|
return false;
|
|
|
|
DecryptBlock(logical_block_index);
|
|
}
|
|
|
|
// Small hack: Set 0x61 of the header to 1 so that VolumeWii realizes that the disc is unencrypted
|
|
if (logical_block_index == 0)
|
|
m_current_block_decrypted[0x61] = 1;
|
|
|
|
return true;
|
|
}
|
|
|
|
bool NFSFileReader::Read(u64 offset, u64 nbytes, u8* out_ptr)
|
|
{
|
|
while (nbytes != 0)
|
|
{
|
|
const u64 logical_block_index = offset / BLOCK_SIZE;
|
|
const u64 offset_in_block = offset % BLOCK_SIZE;
|
|
|
|
if (logical_block_index != m_current_logical_block_index)
|
|
{
|
|
if (!ReadAndDecryptBlock(logical_block_index))
|
|
return false;
|
|
|
|
m_current_logical_block_index = logical_block_index;
|
|
}
|
|
|
|
const u64 bytes_to_copy = std::min(nbytes, BLOCK_SIZE - offset_in_block);
|
|
std::memcpy(out_ptr, m_current_block_decrypted.data() + offset_in_block, bytes_to_copy);
|
|
|
|
offset += bytes_to_copy;
|
|
nbytes -= bytes_to_copy;
|
|
out_ptr += bytes_to_copy;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
} // namespace DiscIO
|