// used: __readfsdword
#include <intrin.h>
// used: d3d11
#include <d3d11.h>

#include "memory.h"

// used: l_print
#include "log.h"
// used: chartohexint
#include "crt.h"
// used: pe64
#include "pe64.h"
#include "../core/spoofcall/invoker.h"
#include "../core/spoofcall/lazy_importer.hpp"

bool MEM::Setup()
{
	bool bSuccess = true;

	const void* hSDL3 = spoof_call<void*>(_fake_addr, &GetModuleBaseHandle, SDL3_DLL);
	const void* hDbgHelp = spoof_call<void*>(_fake_addr, &GetModuleBaseHandle, DBGHELP_DLL);
	const void* hTier0 = GetModuleBaseHandle(TIER0_DLL);

	if (hSDL3 == nullptr || hDbgHelp == nullptr)
		return false;
	
	fnUnDecorateSymbolName = reinterpret_cast<decltype(fnUnDecorateSymbolName)>(GetExportAddress(hDbgHelp, CS_XOR("UnDecorateSymbolName")));
	bSuccess &= (fnUnDecorateSymbolName != nullptr);

	fnSetRelativeMouseMode = reinterpret_cast<decltype(fnSetRelativeMouseMode)>(GetExportAddress(hSDL3, "SDL_SetRelativeMouseMode"));
	bSuccess &= (fnSetRelativeMouseMode != nullptr);

	fnSetWindowGrab = reinterpret_cast<decltype(fnSetWindowGrab)>(GetExportAddress(hSDL3, "SDL_SetWindowGrab"));
	bSuccess &= (fnSetWindowGrab != nullptr);

	fnWarpMouseInWindow = reinterpret_cast<decltype(fnWarpMouseInWindow)>(GetExportAddress(hSDL3, "SDL_WarpMouseInWindow"));
	bSuccess &= (fnWarpMouseInWindow != nullptr);
	L_PRINT(LOG_INFO) << CS_XOR("[Memory] Loaded fnWarpMouseInWindow");

	fnCreateMaterial = reinterpret_cast<decltype(fnCreateMaterial)>(FindPattern(MATERIAL_SYSTEM2_DLL, CS_XOR("48 89 5C 24 ? 48 89 6C 24 ? 56 57 41 56 48 81 EC ? ? ? ? 48 8D 0D")));
	bSuccess &= (fnCreateMaterial != nullptr);
	L_PRINT(LOG_INFO) << CS_XOR("[Memory] Loaded fnCreateMaterial");

	load_key_value = reinterpret_cast<decltype(load_key_value)>(GetExportAddress(hTier0, CS_XOR("?LoadKV3@@YA_NPEAVKeyValues3@@PEAVCUtlString@@PEBDAEBUKV3ID_t@@2@Z")));
	bSuccess &= (load_key_value != nullptr);
	L_PRINT(LOG_INFO) << CS_XOR("[Memory] Loaded load_key_value");

	return bSuccess;
}

#pragma region memory_allocation

/*
 * overload global new/delete operators with our allocators
 * - @note: ensure that all sdk classes that can be instantiated have an overloaded constructor and/or game allocator, otherwise marked as non-constructible
 */
void* __cdecl operator new(const std::size_t nSize)
{
	return MEM::HeapAlloc(nSize);
}

void* __cdecl operator new[](const std::size_t nSize)
{
	return MEM::HeapAlloc(nSize);
}

void __cdecl operator delete(void* pMemory) noexcept
{
	MEM::HeapFree(pMemory);
}

void __cdecl operator delete[](void* pMemory) noexcept
{
	MEM::HeapFree(pMemory);
}

void* MEM::HeapAlloc(const std::size_t nSize)
{
	const HANDLE hHeap = ::GetProcessHeap();
	return ::HeapAlloc(hHeap, 0UL, nSize);
}

void MEM::HeapFree(void* pMemory)
{
	if (pMemory != nullptr)
	{
		const HANDLE hHeap = ::GetProcessHeap();
		::HeapFree(hHeap, 0UL, pMemory);
	}
}

void* MEM::HeapRealloc(void* pMemory, const std::size_t nNewSize)
{
	if (pMemory == nullptr)
		return HeapAlloc(nNewSize);

	if (nNewSize == 0UL)
	{
		HeapFree(pMemory);
		return nullptr;
	}

	const HANDLE hHeap = ::GetProcessHeap();
	return ::HeapReAlloc(hHeap, 0UL, pMemory, nNewSize);
}

#pragma endregion

// @todo: move to win.cpp (or platform.cpp?) except getsectioninfo
#pragma region memory_get

void* MEM::GetModuleBaseHandle(const wchar_t* wszModuleName)
{
	const _PEB* pPEB = reinterpret_cast<_PEB*>(__readgsqword(0x60));

	if (wszModuleName == nullptr)
		return pPEB->ImageBaseAddress;

	void* pModuleBase = nullptr;
	for (LIST_ENTRY* pListEntry = pPEB->Ldr->InMemoryOrderModuleList.Flink; pListEntry != &pPEB->Ldr->InMemoryOrderModuleList; pListEntry = pListEntry->Flink)
	{
		const _LDR_DATA_TABLE_ENTRY* pEntry = CONTAINING_RECORD(pListEntry, _LDR_DATA_TABLE_ENTRY, InMemoryOrderLinks);

		if (pEntry->FullDllName.Buffer != nullptr && CRT::StringCompare(wszModuleName, pEntry->BaseDllName.Buffer) == 0)
		{
			pModuleBase = pEntry->DllBase;
			break;
		}
	}

	if (pModuleBase == nullptr)
		L_PRINT(LOG_ERROR) << CS_XOR("module base not found: \"") << wszModuleName << CS_XOR("\"");

	return pModuleBase;
}

const wchar_t* MEM::GetModuleBaseFileName(const void* hModuleBase)
{
	const _PEB* pPEB = reinterpret_cast<_PEB*>(__readgsqword(0x60));

	if (hModuleBase == nullptr)
		hModuleBase = pPEB->ImageBaseAddress;

	::EnterCriticalSection(pPEB->LoaderLock);

	const wchar_t* wszModuleName = nullptr;
	for (LIST_ENTRY* pListEntry = pPEB->Ldr->InMemoryOrderModuleList.Flink; pListEntry != &pPEB->Ldr->InMemoryOrderModuleList; pListEntry = pListEntry->Flink)
	{
		const _LDR_DATA_TABLE_ENTRY* pEntry = CONTAINING_RECORD(pListEntry, _LDR_DATA_TABLE_ENTRY, InMemoryOrderLinks);

		if (pEntry->DllBase == hModuleBase)
		{
			wszModuleName = pEntry->BaseDllName.Buffer;
			break;
		}
	}

	::LeaveCriticalSection(pPEB->LoaderLock);

	return wszModuleName;
}

void* MEM::GetExportAddress(const void* hModuleBase, const char* szProcedureName)
{
	const auto pBaseAddress = static_cast<const std::uint8_t*>(hModuleBase);

	const auto pIDH = static_cast<const IMAGE_DOS_HEADER*>(hModuleBase);
	if (pIDH->e_magic != IMAGE_DOS_SIGNATURE)
		return nullptr;

	const auto pINH = reinterpret_cast<const IMAGE_NT_HEADERS64*>(pBaseAddress + pIDH->e_lfanew);
	if (pINH->Signature != IMAGE_NT_SIGNATURE)
		return nullptr;

	const IMAGE_OPTIONAL_HEADER64* pIOH = &pINH->OptionalHeader;
	const std::uintptr_t nExportDirectorySize = pIOH->DataDirectory[IMAGE_DIRECTORY_ENTRY_EXPORT].Size;
	const std::uintptr_t uExportDirectoryAddress = pIOH->DataDirectory[IMAGE_DIRECTORY_ENTRY_EXPORT].VirtualAddress;

	if (nExportDirectorySize == 0U || uExportDirectoryAddress == 0U)
	{
		L_PRINT(LOG_ERROR) << CS_XOR("module has no exports: \"") << GetModuleBaseFileName(hModuleBase) << CS_XOR("\"");
		return nullptr;
	}

	const auto pIED = reinterpret_cast<const IMAGE_EXPORT_DIRECTORY*>(pBaseAddress + uExportDirectoryAddress);
	const auto pNamesRVA = reinterpret_cast<const std::uint32_t*>(pBaseAddress + pIED->AddressOfNames);
	const auto pNameOrdinalsRVA = reinterpret_cast<const std::uint16_t*>(pBaseAddress + pIED->AddressOfNameOrdinals);
	const auto pFunctionsRVA = reinterpret_cast<const std::uint32_t*>(pBaseAddress + pIED->AddressOfFunctions);

	// Perform binary search to find the export by name
	std::size_t nRight = pIED->NumberOfNames, nLeft = 0U;
	while (nRight != nLeft)
	{
		// Avoid INT_MAX/2 overflow
		const std::size_t uMiddle = nLeft + ((nRight - nLeft) >> 1U);
		const int iResult = CRT::StringCompare(szProcedureName, reinterpret_cast<const char*>(pBaseAddress + pNamesRVA[uMiddle]));

		if (iResult == 0)
		{
			const std::uint32_t uFunctionRVA = pFunctionsRVA[pNameOrdinalsRVA[uMiddle]];

#ifdef _DEBUG
			L_PRINT(LOG_INFO) << CS_XOR("export found: \"") << reinterpret_cast<const char*>(pBaseAddress + pNamesRVA[uMiddle]) << CS_XOR("\" in \"") << GetModuleBaseFileName(hModuleBase) << CS_XOR("\" at: ") << L::AddFlags(LOG_MODE_INT_SHOWBASE | LOG_MODE_INT_FORMAT_HEX) << uFunctionRVA;
#else
			L_PRINT(LOG_INFO) << CS_XOR("export found: ") << szProcedureName;
#endif // _DEBUG

			// Check if it's a forwarded export
			if (uFunctionRVA >= uExportDirectoryAddress && uFunctionRVA - uExportDirectoryAddress < nExportDirectorySize)
			{
				// Forwarded exports are not supported
				break;
			}

			return const_cast<std::uint8_t*>(pBaseAddress) + uFunctionRVA;
		}

		if (iResult > 0)
			nLeft = uMiddle + 1;
		else
			nRight = uMiddle;
	}

	L_PRINT(LOG_ERROR) << CS_XOR("export not found: ") << szProcedureName;

	// Export not found
	return nullptr;
}

bool MEM::GetSectionInfo(const void* hModuleBase, const char* szSectionName, std::uint8_t** ppSectionStart, std::size_t* pnSectionSize)
{
	const auto pBaseAddress = static_cast<const std::uint8_t*>(hModuleBase);

	const auto pIDH = static_cast<const IMAGE_DOS_HEADER*>(hModuleBase);
	if (pIDH->e_magic != IMAGE_DOS_SIGNATURE)
		return false;

	const auto pINH = reinterpret_cast<const IMAGE_NT_HEADERS*>(pBaseAddress + pIDH->e_lfanew);
	if (pINH->Signature != IMAGE_NT_SIGNATURE)
		return false;

	const IMAGE_SECTION_HEADER* pISH = IMAGE_FIRST_SECTION(pINH);

	// go through all code sections
	for (WORD i = 0U; i < pINH->FileHeader.NumberOfSections; i++, pISH++)
	{
		// @test: use case insensitive comparison instead?
		if (CRT::StringCompareN(szSectionName, reinterpret_cast<const char*>(pISH->Name), IMAGE_SIZEOF_SHORT_NAME) == 0)
		{
			if (ppSectionStart != nullptr)
				*ppSectionStart = const_cast<std::uint8_t*>(pBaseAddress) + pISH->VirtualAddress;

			if (pnSectionSize != nullptr)
				*pnSectionSize = pISH->SizeOfRawData;

			return true;
		}
	}

	L_PRINT(LOG_ERROR) << CS_XOR("code section not found: \"") << szSectionName << CS_XOR("\"");
	return false;
}

#pragma endregion

#pragma region memory_search
UTILPtr MEM::FindPatterns(const wchar_t* wszModuleName, const char* szPattern)
{
	// convert pattern string to byte array
	const std::size_t nApproximateBufferSize = (CRT::StringLength(szPattern) >> 1U) + 1U;
	std::uint8_t* arrByteBuffer = static_cast<std::uint8_t*>(MEM_STACKALLOC(nApproximateBufferSize));
	char* szMaskBuffer = static_cast<char*>(MEM_STACKALLOC(nApproximateBufferSize));
	PatternToBytes(szPattern, arrByteBuffer, szMaskBuffer);

	// @test: use search with straight in-place conversion? do not think it will be faster, cuz of bunch of new checks that gonna be performed for each iteration
	return FindPattern(wszModuleName, reinterpret_cast<const char*>(arrByteBuffer), szMaskBuffer);
}
std::uint8_t* MEM::FindPattern(const wchar_t* wszModuleName, const char* szPattern)
{
	// convert pattern string to byte array
	const std::size_t nApproximateBufferSize = (CRT::StringLength(szPattern) >> 1U) + 1U;
	std::uint8_t* arrByteBuffer = static_cast<std::uint8_t*>(MEM_STACKALLOC(nApproximateBufferSize));
	char* szMaskBuffer = static_cast<char*>(MEM_STACKALLOC(nApproximateBufferSize));
	PatternToBytes(szPattern, arrByteBuffer, szMaskBuffer);

	// @test: use search with straight in-place conversion? do not think it will be faster, cuz of bunch of new checks that gonna be performed for each iteration
	return FindPattern(wszModuleName, reinterpret_cast<const char*>(arrByteBuffer), szMaskBuffer);
}
std::uint8_t* MEM::FindPattern(const wchar_t* wszModuleName, const char* szBytePattern, const char* szByteMask)
{
	const void* hModuleBase = spoof_call<void*>(_fake_addr, &GetModuleBaseHandle, wszModuleName);

	if (hModuleBase == nullptr)
	{
		L_PRINT(LOG_ERROR) << CS_XOR("failed to get module handle for: \"") << wszModuleName << CS_XOR("\"");
		return nullptr;
	}

	const auto pBaseAddress = static_cast<const std::uint8_t*>(hModuleBase);

	const auto pIDH = static_cast<const IMAGE_DOS_HEADER*>(hModuleBase);
	if (pIDH->e_magic != IMAGE_DOS_SIGNATURE)
	{
		L_PRINT(LOG_ERROR) << CS_XOR("failed to get module size, image is invalid");
		return nullptr;
	}

	const auto pINH = reinterpret_cast<const IMAGE_NT_HEADERS*>(pBaseAddress + pIDH->e_lfanew);
	if (pINH->Signature != IMAGE_NT_SIGNATURE)
	{
		L_PRINT(LOG_ERROR) << CS_XOR("failed to get module size, image is invalid");
		return nullptr;
	}

	const std::uint8_t* arrByteBuffer = reinterpret_cast<const std::uint8_t*>(szBytePattern);
	const std::size_t nByteCount = CRT::StringLength(szByteMask);

	std::uint8_t* pFoundAddress = nullptr;

	// perform little overhead to keep all patterns unique
#ifdef CS_PARANOID_PATTERN_UNIQUENESS 
	const std::vector<std::uint8_t*> vecFoundOccurrences =  FindPatternAllOccurrencesEx(pBaseAddress, pINH->OptionalHeader.SizeOfImage, arrByteBuffer, nByteCount, szByteMask);

	// notify user about non-unique pattern
	if (!vecFoundOccurrences.empty())
	{
		// notify user about non-unique pattern
		if (vecFoundOccurrences.size() > 1U)
		{
			char* szPattern = static_cast<char*>(MEM_STACKALLOC((nByteCount << 1U) + nByteCount));
			[[maybe_unused]] const std::size_t nConvertedPatternLength = BytesToPattern(arrByteBuffer, nByteCount, szPattern);

			L_PRINT(LOG_WARNING) << CS_XOR("found more than one occurrence with \"") << szPattern << CS_XOR("\" pattern, consider updating it!");

			MEM_STACKFREE(szPattern);
		}

		// return first found occurrence
		pFoundAddress = vecFoundOccurrences[0];
	}
#else
	// @todo: we also can go through code sections and skip noexec pages, but will it really improve performance? / or at least for all occurrences search
	// https://docs.microsoft.com/en-us/windows/win32/api/winnt/ns-winnt-image_section_header
#if 0
	IMAGE_SECTION_HEADER* pCurrentSection = IMAGE_FIRST_SECTION(pINH);
	for (WORD i = 0U; i != pINH->FileHeader.NumberOfSections; i++)
	{
		// check does page have executable code
		if (pCurrentSection->Characteristics & IMAGE_SCN_CNT_CODE || pCurrentSection->Characteristics & IMAGE_SCN_MEM_EXECUTE)
		{
			pFoundAddress = FindPatternEx(pBaseAddress + pCurrentSection->VirtualAddress, pCurrentSection->SizeOfRawData, arrByteBuffer, nByteCount, szByteMask);

			if (pFoundAddress != nullptr)
				break;
		}

		++pCurrentSection;
	}
#else
	pFoundAddress = FindPatternEx(pBaseAddress, pINH->OptionalHeader.SizeOfImage, arrByteBuffer, nByteCount, szByteMask);
#endif
#endif

	if (pFoundAddress == nullptr)
	{
		char* szPattern = static_cast<char*>(MEM_STACKALLOC((nByteCount << 1U) + nByteCount));
		[[maybe_unused]] const std::size_t nConvertedPatternLength = BytesToPattern(arrByteBuffer, nByteCount, szPattern);

		L_PRINT(LOG_ERROR) << CS_XOR("pattern not found: \"") << szPattern << CS_XOR("\"");

		MEM_STACKFREE(szPattern);
	}

	return pFoundAddress;
}

// @todo: msvc poorly optimizes this, it looks even better w/o optimization at all
std::uint8_t* MEM::FindPatternEx(const std::uint8_t* pRegionStart, const std::size_t nRegionSize, const std::uint8_t* arrByteBuffer, const std::size_t nByteCount, const char* szByteMask)
{
	std::uint8_t* pCurrentAddress = const_cast<std::uint8_t*>(pRegionStart);
	const std::uint8_t* pRegionEnd = pRegionStart + nRegionSize - nByteCount;
	const bool bIsMaskUsed = (szByteMask != nullptr);

	while (pCurrentAddress < pRegionEnd)
	{
		// check the first byte before entering the loop, otherwise if there two consecutive bytes of first byte in the buffer, we may skip both and fail the search
		if ((bIsMaskUsed && *szByteMask == '?') || *pCurrentAddress == *arrByteBuffer)
		{
			if (nByteCount == 1)
				return pCurrentAddress;

			// compare the least byte sequence and continue on wildcard or skip forward on first mismatched byte
			std::size_t nComparedBytes = 0U;
			while ((bIsMaskUsed && szByteMask[nComparedBytes + 1U] == '?') || pCurrentAddress[nComparedBytes + 1U] == arrByteBuffer[nComparedBytes + 1U])
			{
				// check does byte sequence match
				if (++nComparedBytes == nByteCount - 1U)
					return pCurrentAddress;
			}

			// skip non suitable bytes
			pCurrentAddress += nComparedBytes;
		}

		++pCurrentAddress;
	}

	return nullptr;
}

std::vector<std::uint8_t*> MEM::FindPatternAllOccurrencesEx(const std::uint8_t* pRegionStart, const std::size_t nRegionSize, const std::uint8_t* arrByteBuffer, const std::size_t nByteCount, const char* szByteMask)
{
	const std::uint8_t* pRegionEnd = pRegionStart + nRegionSize - nByteCount;
	const bool bIsMaskUsed = (szByteMask != nullptr);

	// container for addresses of the all found occurrences
	std::vector<std::uint8_t*> vecOccurrences = {};

	for (std::uint8_t* pCurrentByte = const_cast<std::uint8_t*>(pRegionStart); pCurrentByte < pRegionEnd; ++pCurrentByte)
	{
		// do a first byte check before entering the loop, otherwise if there two consecutive bytes of first byte in the buffer, we may skip both and fail the search
		if ((!bIsMaskUsed || *szByteMask != '?') && *pCurrentByte != *arrByteBuffer)
			continue;

		// check for bytes sequence match
		bool bSequenceMatch = true;
		for (std::size_t i = 1U; i < nByteCount; i++)
		{
			// compare sequence and continue on wildcard or skip forward on first mismatched byte
			if ((!bIsMaskUsed || szByteMask[i] != '?') && pCurrentByte[i] != arrByteBuffer[i])
			{
				// skip non suitable bytes
				pCurrentByte += i - 1U;

				bSequenceMatch = false;
				break;
			}
		}

		// check did we found address
		if (bSequenceMatch)
			vecOccurrences.push_back(pCurrentByte);
	}

	return vecOccurrences;
}

#pragma endregion

#pragma region memory_extra

std::size_t MEM::PatternToBytes(const char* szPattern, std::uint8_t* pOutByteBuffer, char* szOutMaskBuffer)
{
	std::uint8_t* pCurrentByte = pOutByteBuffer;

	while (*szPattern != '\0')
	{
		// check is a wildcard
		if (*szPattern == '?')
		{
			++szPattern;
#ifdef CS_PARANOID
			CS_ASSERT(*szPattern == '\0' || *szPattern == ' ' || *szPattern == '?'); // we're expect that next character either terminating null, whitespace or part of double wildcard (note that it's required if your pattern written without whitespaces)
#endif

			// ignore that
			*pCurrentByte++ = 0U;
			*szOutMaskBuffer++ = '?';
		}
		// check is not space
		else if (*szPattern != ' ')
		{
			// convert two consistent numbers in a row to byte value
			std::uint8_t uByte = static_cast<std::uint8_t>(CRT::CharToHexInt(*szPattern) << 4);

			++szPattern;
#ifdef CS_PARANOID
			CS_ASSERT(*szPattern != '\0' && *szPattern != '?' && *szPattern != ' '); // we're expect that byte always represented by two numbers in a row
#endif

			uByte |= static_cast<std::uint8_t>(CRT::CharToHexInt(*szPattern));

			*pCurrentByte++ = uByte;
			*szOutMaskBuffer++ = 'x';
		}

		++szPattern;
	}

	// zero terminate both buffers
	*pCurrentByte = 0U;
	*szOutMaskBuffer = '\0';

	return pCurrentByte - pOutByteBuffer;
}

std::size_t MEM::BytesToPattern(const std::uint8_t* pByteBuffer, const std::size_t nByteCount, char* szOutBuffer)
{
	char* szCurrentPattern = szOutBuffer;

	for (std::size_t i = 0U; i < nByteCount; i++)
	{
		// manually convert byte to chars
		const char* szHexByte = &CRT::_TWO_DIGITS_HEX_LUT[pByteBuffer[i] * 2U];
		*szCurrentPattern++ = szHexByte[0];
		*szCurrentPattern++ = szHexByte[1];
		*szCurrentPattern++ = ' ';
	}
	*--szCurrentPattern = '\0';

	return szCurrentPattern - szOutBuffer;
}

#pragma endregion