Memory: Use a table based lookup scheme to read from memory regions
This commit is contained in:
parent
52158c1b8d
commit
dd4430609a
5 changed files with 167 additions and 121 deletions
|
@ -212,6 +212,7 @@ set(HEADERS
|
|||
loader/ncch.h
|
||||
mem_map.h
|
||||
memory.h
|
||||
memory_setup.h
|
||||
settings.h
|
||||
system.h
|
||||
)
|
||||
|
|
|
@ -7,8 +7,11 @@
|
|||
#include "common/common_types.h"
|
||||
#include "common/logging/log.h"
|
||||
|
||||
#include "core/hle/config_mem.h"
|
||||
#include "core/hle/shared_page.h"
|
||||
#include "core/mem_map.h"
|
||||
#include "core/memory.h"
|
||||
#include "core/memory_setup.h"
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
|
@ -26,18 +29,19 @@ namespace {
|
|||
|
||||
struct MemoryArea {
|
||||
u8** ptr;
|
||||
size_t size;
|
||||
u32 base;
|
||||
u32 size;
|
||||
};
|
||||
|
||||
// We don't declare the IO regions in here since its handled by other means.
|
||||
static MemoryArea memory_areas[] = {
|
||||
{&g_exefs_code, PROCESS_IMAGE_MAX_SIZE},
|
||||
{&g_heap, HEAP_SIZE },
|
||||
{&g_shared_mem, SHARED_MEMORY_SIZE },
|
||||
{&g_heap_linear, LINEAR_HEAP_SIZE },
|
||||
{&g_vram, VRAM_SIZE },
|
||||
{&g_dsp_mem, DSP_RAM_SIZE },
|
||||
{&g_tls_mem, TLS_AREA_SIZE },
|
||||
{&g_exefs_code, PROCESS_IMAGE_VADDR, PROCESS_IMAGE_MAX_SIZE},
|
||||
{&g_heap, HEAP_VADDR, HEAP_SIZE },
|
||||
{&g_shared_mem, SHARED_MEMORY_VADDR, SHARED_MEMORY_SIZE },
|
||||
{&g_heap_linear, LINEAR_HEAP_VADDR, LINEAR_HEAP_SIZE },
|
||||
{&g_vram, VRAM_VADDR, VRAM_SIZE },
|
||||
{&g_dsp_mem, DSP_RAM_VADDR, DSP_RAM_SIZE },
|
||||
{&g_tls_mem, TLS_AREA_VADDR, TLS_AREA_SIZE },
|
||||
};
|
||||
|
||||
/// Represents a block of memory mapped by ControlMemory/MapMemoryBlock
|
||||
|
@ -132,9 +136,14 @@ VAddr PhysicalToVirtualAddress(const PAddr addr) {
|
|||
}
|
||||
|
||||
void Init() {
|
||||
InitMemoryMap();
|
||||
|
||||
for (MemoryArea& area : memory_areas) {
|
||||
*area.ptr = new u8[area.size];
|
||||
MapMemoryRegion(area.base, area.size, *area.ptr);
|
||||
}
|
||||
MapMemoryRegion(CONFIG_MEMORY_VADDR, CONFIG_MEMORY_SIZE, (u8*)&ConfigMem::config_mem);
|
||||
MapMemoryRegion(SHARED_PAGE_VADDR, SHARED_PAGE_SIZE, (u8*)&SharedPage::shared_page);
|
||||
|
||||
LOG_DEBUG(HW_Memory, "initialized OK, RAM at %p", g_heap);
|
||||
}
|
||||
|
|
|
@ -1,7 +1,10 @@
|
|||
// Copyright 2014 Citra Emulator Project
|
||||
// Copyright 2015 Citra Emulator Project
|
||||
// Licensed under GPLv2 or any later version
|
||||
// Refer to the license.txt file included.
|
||||
|
||||
#include <array>
|
||||
|
||||
#include "common/assert.h"
|
||||
#include "common/common_types.h"
|
||||
#include "common/logging/log.h"
|
||||
#include "common/swap.h"
|
||||
|
@ -14,126 +17,134 @@
|
|||
|
||||
namespace Memory {
|
||||
|
||||
const u32 PAGE_MASK = PAGE_SIZE - 1;
|
||||
const int PAGE_BITS = 12;
|
||||
|
||||
enum class PageType {
|
||||
/// Page is unmapped and should cause an access error.
|
||||
Unmapped,
|
||||
/// Page is mapped to regular memory. This is the only type you can get pointers to.
|
||||
Memory,
|
||||
/// Page is mapped to a I/O region. Writing and reading to this page is handled by functions.
|
||||
Special,
|
||||
};
|
||||
|
||||
/**
|
||||
* A (reasonably) fast way of allowing switchable and remmapable process address spaces. It loosely
|
||||
* mimics the way a real CPU page table works, but instead is optimized for minimal decoding and
|
||||
* fetching requirements when acessing. In the usual case of an access to regular memory, it only
|
||||
* requires an indexed fetch and a check for NULL.
|
||||
*/
|
||||
struct PageTable {
|
||||
static const size_t NUM_ENTRIES = 1 << (32 - PAGE_BITS);
|
||||
|
||||
/**
|
||||
* Array of memory pointers backing each page. An entry can only be non-null if the
|
||||
* corresponding entry in the `attributes` array is of type `Memory`.
|
||||
*/
|
||||
std::array<u8*, NUM_ENTRIES> pointers;
|
||||
|
||||
/**
|
||||
* Array of fine grained page attributes. If it is set to any value other than `Memory`, then
|
||||
* the corresponding entry in `pointer` MUST be set to null.
|
||||
*/
|
||||
std::array<PageType, NUM_ENTRIES> attributes;
|
||||
};
|
||||
|
||||
/// Singular page table used for the singleton process
|
||||
static PageTable main_page_table;
|
||||
/// Currently active page table
|
||||
static PageTable* current_page_table = &main_page_table;
|
||||
|
||||
static void MapPages(u32 base, u32 size, u8* memory, PageType type) {
|
||||
LOG_DEBUG(HW_Memory, "Mapping %p onto %08X-%08X", memory, base * PAGE_SIZE, (base + size) * PAGE_SIZE);
|
||||
|
||||
u32 end = base + size;
|
||||
|
||||
while (base != end) {
|
||||
ASSERT_MSG(base < PageTable::NUM_ENTRIES, "out of range mapping at %08X", base);
|
||||
|
||||
if (current_page_table->attributes[base] != PageType::Unmapped) {
|
||||
LOG_ERROR(HW_Memory, "overlapping memory ranges at %08X", base * PAGE_SIZE);
|
||||
}
|
||||
current_page_table->attributes[base] = type;
|
||||
current_page_table->pointers[base] = memory;
|
||||
|
||||
base += 1;
|
||||
memory += PAGE_SIZE;
|
||||
}
|
||||
}
|
||||
|
||||
void InitMemoryMap() {
|
||||
main_page_table.pointers.fill(nullptr);
|
||||
main_page_table.attributes.fill(PageType::Unmapped);
|
||||
}
|
||||
|
||||
void MapMemoryRegion(VAddr base, u32 size, u8* target) {
|
||||
ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: %08X", size);
|
||||
ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: %08X", base);
|
||||
MapPages(base / PAGE_SIZE, size / PAGE_SIZE, target, PageType::Memory);
|
||||
}
|
||||
|
||||
void MapIoRegion(VAddr base, u32 size) {
|
||||
ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: %08X", size);
|
||||
ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: %08X", base);
|
||||
MapPages(base / PAGE_SIZE, size / PAGE_SIZE, nullptr, PageType::Special);
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
inline void Read(T &var, const VAddr vaddr) {
|
||||
// TODO: Figure out the fastest order of tests for both read and write (they are probably different).
|
||||
// TODO: Make sure this represents the mirrors in a correct way.
|
||||
// Could just do a base-relative read, too.... TODO
|
||||
T Read(const VAddr vaddr) {
|
||||
const u8* page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS];
|
||||
if (page_pointer) {
|
||||
return *reinterpret_cast<const T*>(page_pointer + (vaddr & PAGE_MASK));
|
||||
}
|
||||
|
||||
// Kernel memory command buffer
|
||||
if (vaddr >= TLS_AREA_VADDR && vaddr < TLS_AREA_VADDR_END) {
|
||||
var = *((const T*)&g_tls_mem[vaddr - TLS_AREA_VADDR]);
|
||||
|
||||
// ExeFS:/.code is loaded here
|
||||
} else if ((vaddr >= PROCESS_IMAGE_VADDR) && (vaddr < PROCESS_IMAGE_VADDR_END)) {
|
||||
var = *((const T*)&g_exefs_code[vaddr - PROCESS_IMAGE_VADDR]);
|
||||
|
||||
// FCRAM - linear heap
|
||||
} else if ((vaddr >= LINEAR_HEAP_VADDR) && (vaddr < LINEAR_HEAP_VADDR_END)) {
|
||||
var = *((const T*)&g_heap_linear[vaddr - LINEAR_HEAP_VADDR]);
|
||||
|
||||
// FCRAM - application heap
|
||||
} else if ((vaddr >= HEAP_VADDR) && (vaddr < HEAP_VADDR_END)) {
|
||||
var = *((const T*)&g_heap[vaddr - HEAP_VADDR]);
|
||||
|
||||
// Shared memory
|
||||
} else if ((vaddr >= SHARED_MEMORY_VADDR) && (vaddr < SHARED_MEMORY_VADDR_END)) {
|
||||
var = *((const T*)&g_shared_mem[vaddr - SHARED_MEMORY_VADDR]);
|
||||
|
||||
// Config memory
|
||||
} else if ((vaddr >= CONFIG_MEMORY_VADDR) && (vaddr < CONFIG_MEMORY_VADDR_END)) {
|
||||
const u8* raw_memory = (const u8*)&ConfigMem::config_mem;
|
||||
var = *((const T*)&raw_memory[vaddr - CONFIG_MEMORY_VADDR]);
|
||||
|
||||
// Shared page
|
||||
} else if ((vaddr >= SHARED_PAGE_VADDR) && (vaddr < SHARED_PAGE_VADDR_END)) {
|
||||
const u8* raw_memory = (const u8*)&SharedPage::shared_page;
|
||||
var = *((const T*)&raw_memory[vaddr - SHARED_PAGE_VADDR]);
|
||||
|
||||
// DSP memory
|
||||
} else if ((vaddr >= DSP_RAM_VADDR) && (vaddr < DSP_RAM_VADDR_END)) {
|
||||
var = *((const T*)&g_dsp_mem[vaddr - DSP_RAM_VADDR]);
|
||||
|
||||
// VRAM
|
||||
} else if ((vaddr >= VRAM_VADDR) && (vaddr < VRAM_VADDR_END)) {
|
||||
var = *((const T*)&g_vram[vaddr - VRAM_VADDR]);
|
||||
|
||||
} else {
|
||||
LOG_ERROR(HW_Memory, "unknown Read%lu @ 0x%08X", sizeof(var) * 8, vaddr);
|
||||
PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
|
||||
switch (type) {
|
||||
case PageType::Unmapped:
|
||||
LOG_ERROR(HW_Memory, "unmapped Read%lu @ 0x%08X", sizeof(T) * 8, vaddr);
|
||||
return 0;
|
||||
case PageType::Memory:
|
||||
ASSERT_MSG(false, "Mapped memory page without a pointer @ %08X", vaddr);
|
||||
case PageType::Special:
|
||||
LOG_ERROR(HW_Memory, "I/O reads aren't implemented yet @ %08X", vaddr);
|
||||
return 0;
|
||||
default:
|
||||
UNREACHABLE();
|
||||
}
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
inline void Write(const VAddr vaddr, const T data) {
|
||||
void Write(const VAddr vaddr, const T data) {
|
||||
u8* page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS];
|
||||
if (page_pointer) {
|
||||
*reinterpret_cast<T*>(page_pointer + (vaddr & PAGE_MASK)) = data;
|
||||
return;
|
||||
}
|
||||
|
||||
// Kernel memory command buffer
|
||||
if (vaddr >= TLS_AREA_VADDR && vaddr < TLS_AREA_VADDR_END) {
|
||||
*(T*)&g_tls_mem[vaddr - TLS_AREA_VADDR] = data;
|
||||
|
||||
// ExeFS:/.code is loaded here
|
||||
} else if ((vaddr >= PROCESS_IMAGE_VADDR) && (vaddr < PROCESS_IMAGE_VADDR_END)) {
|
||||
*(T*)&g_exefs_code[vaddr - PROCESS_IMAGE_VADDR] = data;
|
||||
|
||||
// FCRAM - linear heap
|
||||
} else if ((vaddr >= LINEAR_HEAP_VADDR) && (vaddr < LINEAR_HEAP_VADDR_END)) {
|
||||
*(T*)&g_heap_linear[vaddr - LINEAR_HEAP_VADDR] = data;
|
||||
|
||||
// FCRAM - application heap
|
||||
} else if ((vaddr >= HEAP_VADDR) && (vaddr < HEAP_VADDR_END)) {
|
||||
*(T*)&g_heap[vaddr - HEAP_VADDR] = data;
|
||||
|
||||
// Shared memory
|
||||
} else if ((vaddr >= SHARED_MEMORY_VADDR) && (vaddr < SHARED_MEMORY_VADDR_END)) {
|
||||
*(T*)&g_shared_mem[vaddr - SHARED_MEMORY_VADDR] = data;
|
||||
|
||||
// VRAM
|
||||
} else if ((vaddr >= VRAM_VADDR) && (vaddr < VRAM_VADDR_END)) {
|
||||
*(T*)&g_vram[vaddr - VRAM_VADDR] = data;
|
||||
|
||||
// DSP memory
|
||||
} else if ((vaddr >= DSP_RAM_VADDR) && (vaddr < DSP_RAM_VADDR_END)) {
|
||||
*(T*)&g_dsp_mem[vaddr - DSP_RAM_VADDR] = data;
|
||||
|
||||
//} else if ((vaddr & 0xFFFF0000) == 0x1FF80000) {
|
||||
// ASSERT_MSG(MEMMAP, false, "umimplemented write to Configuration Memory");
|
||||
//} else if ((vaddr & 0xFFFFF000) == 0x1FF81000) {
|
||||
// ASSERT_MSG(MEMMAP, false, "umimplemented write to shared page");
|
||||
|
||||
// Error out...
|
||||
} else {
|
||||
LOG_ERROR(HW_Memory, "unknown Write%lu 0x%08X @ 0x%08X", sizeof(data) * 8, (u32)data, vaddr);
|
||||
PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
|
||||
switch (type) {
|
||||
case PageType::Unmapped:
|
||||
LOG_ERROR(HW_Memory, "unmapped Write%lu 0x%08X @ 0x%08X", sizeof(data) * 8, (u32) data, vaddr);
|
||||
return;
|
||||
case PageType::Memory:
|
||||
ASSERT_MSG(false, "Mapped memory page without a pointer @ %08X", vaddr);
|
||||
case PageType::Special:
|
||||
LOG_ERROR(HW_Memory, "I/O writes aren't implemented yet @ %08X", vaddr);
|
||||
return;
|
||||
default:
|
||||
UNREACHABLE();
|
||||
}
|
||||
}
|
||||
|
||||
u8* GetPointer(const VAddr vaddr) {
|
||||
// Kernel memory command buffer
|
||||
if (vaddr >= TLS_AREA_VADDR && vaddr < TLS_AREA_VADDR_END) {
|
||||
return g_tls_mem + (vaddr - TLS_AREA_VADDR);
|
||||
|
||||
// ExeFS:/.code is loaded here
|
||||
} else if ((vaddr >= PROCESS_IMAGE_VADDR) && (vaddr < PROCESS_IMAGE_VADDR_END)) {
|
||||
return g_exefs_code + (vaddr - PROCESS_IMAGE_VADDR);
|
||||
|
||||
// FCRAM - linear heap
|
||||
} else if ((vaddr >= LINEAR_HEAP_VADDR) && (vaddr < LINEAR_HEAP_VADDR_END)) {
|
||||
return g_heap_linear + (vaddr - LINEAR_HEAP_VADDR);
|
||||
|
||||
// FCRAM - application heap
|
||||
} else if ((vaddr >= HEAP_VADDR) && (vaddr < HEAP_VADDR_END)) {
|
||||
return g_heap + (vaddr - HEAP_VADDR);
|
||||
|
||||
// Shared memory
|
||||
} else if ((vaddr >= SHARED_MEMORY_VADDR) && (vaddr < SHARED_MEMORY_VADDR_END)) {
|
||||
return g_shared_mem + (vaddr - SHARED_MEMORY_VADDR);
|
||||
|
||||
// VRAM
|
||||
} else if ((vaddr >= VRAM_VADDR) && (vaddr < VRAM_VADDR_END)) {
|
||||
return g_vram + (vaddr - VRAM_VADDR);
|
||||
|
||||
} else {
|
||||
LOG_ERROR(HW_Memory, "unknown GetPointer @ 0x%08x", vaddr);
|
||||
return 0;
|
||||
u8* page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS];
|
||||
if (page_pointer) {
|
||||
return page_pointer + (vaddr & PAGE_MASK);
|
||||
}
|
||||
|
||||
LOG_ERROR(HW_Memory, "unknown GetPointer @ 0x%08x", vaddr);
|
||||
return nullptr;
|
||||
}
|
||||
|
||||
u8* GetPhysicalPointer(PAddr address) {
|
||||
|
@ -141,27 +152,19 @@ u8* GetPhysicalPointer(PAddr address) {
|
|||
}
|
||||
|
||||
u8 Read8(const VAddr addr) {
|
||||
u8 data = 0;
|
||||
Read<u8>(data, addr);
|
||||
return data;
|
||||
return Read<u8>(addr);
|
||||
}
|
||||
|
||||
u16 Read16(const VAddr addr) {
|
||||
u16_le data = 0;
|
||||
Read<u16_le>(data, addr);
|
||||
return data;
|
||||
return Read<u16_le>(addr);
|
||||
}
|
||||
|
||||
u32 Read32(const VAddr addr) {
|
||||
u32_le data = 0;
|
||||
Read<u32_le>(data, addr);
|
||||
return data;
|
||||
return Read<u32_le>(addr);
|
||||
}
|
||||
|
||||
u64 Read64(const VAddr addr) {
|
||||
u64_le data = 0;
|
||||
Read<u64_le>(data, addr);
|
||||
return data;
|
||||
return Read<u64_le>(addr);
|
||||
}
|
||||
|
||||
void Write8(const VAddr addr, const u8 data) {
|
||||
|
|
|
@ -8,6 +8,10 @@
|
|||
|
||||
namespace Memory {
|
||||
|
||||
/**
|
||||
* Page size used by the ARM architecture. This is the smallest granularity with which memory can
|
||||
* be mapped.
|
||||
*/
|
||||
const u32 PAGE_SIZE = 0x1000;
|
||||
|
||||
/// Physical memory regions as seen from the ARM11
|
||||
|
|
29
src/core/memory_setup.h
Normal file
29
src/core/memory_setup.h
Normal file
|
@ -0,0 +1,29 @@
|
|||
// Copyright 2015 Citra Emulator Project
|
||||
// Licensed under GPLv2 or any later version
|
||||
// Refer to the license.txt file included.
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "common/common_types.h"
|
||||
|
||||
namespace Memory {
|
||||
|
||||
void InitMemoryMap();
|
||||
|
||||
/**
|
||||
* Maps an allocated buffer onto a region of the emulated process address space.
|
||||
*
|
||||
* @param base The address to start mapping at. Must be page-aligned.
|
||||
* @param size The amount of bytes to map. Must be page-aligned.
|
||||
* @param target Buffer with the memory backing the mapping. Must be of length at least `size`.
|
||||
*/
|
||||
void MapMemoryRegion(VAddr base, u32 size, u8* target);
|
||||
|
||||
/**
|
||||
* Maps a region of the emulated process address space as a IO region.
|
||||
* @note Currently this can only be used to mark a region as being IO, since actual memory-mapped
|
||||
* IO isn't yet supported.
|
||||
*/
|
||||
void MapIoRegion(VAddr base, u32 size);
|
||||
|
||||
}
|
Loading…
Reference in a new issue