31 files changed, 823 insertions, 344 deletions

diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index 6cc60fd58..c17290b9b 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -29,6 +29,7 @@ set(SRCS
             hle/kernel/address_arbiter.cpp
             hle/kernel/event.cpp
             hle/kernel/kernel.cpp
+            hle/kernel/memory.cpp
             hle/kernel/mutex.cpp
             hle/kernel/process.cpp
             hle/kernel/resource_limit.cpp
@@ -115,7 +116,6 @@ set(SRCS
             loader/loader.cpp
             loader/ncch.cpp
             tracer/recorder.cpp
-            mem_map.cpp
             memory.cpp
             settings.cpp
             system.cpp
@@ -157,6 +157,7 @@ set(HEADERS
             hle/kernel/address_arbiter.h
             hle/kernel/event.h
             hle/kernel/kernel.h
+            hle/kernel/memory.h
             hle/kernel/mutex.h
             hle/kernel/process.h
             hle/kernel/resource_limit.h
@@ -245,7 +246,6 @@ set(HEADERS
             loader/ncch.h
             tracer/recorder.h
             tracer/citrace.h
-            mem_map.h
             memory.h
             memory_setup.h
             settings.h
diff --git a/src/core/arm/skyeye_common/armstate.cpp b/src/core/arm/skyeye_common/armstate.cpp
index ccb2eb0eb..0491717dc 100644
--- a/src/core/arm/skyeye_common/armstate.cpp
+++ b/src/core/arm/skyeye_common/armstate.cpp
@@ -4,7 +4,6 @@
 
 #include "common/swap.h"
 #include "common/logging/log.h"
-#include "core/mem_map.h"
 #include "core/memory.h"
 #include "core/arm/skyeye_common/armstate.h"
 #include "core/arm/skyeye_common/vfp/vfp.h"
diff --git a/src/core/arm/skyeye_common/armsupp.cpp b/src/core/arm/skyeye_common/armsupp.cpp
index d31fb9449..883713e86 100644
--- a/src/core/arm/skyeye_common/armsupp.cpp
+++ b/src/core/arm/skyeye_common/armsupp.cpp
@@ -17,7 +17,6 @@
 
 #include "common/logging/log.h"
 
-#include "core/mem_map.h"
 #include "core/arm/skyeye_common/arm_regformat.h"
 #include "core/arm/skyeye_common/armstate.h"
 #include "core/arm/skyeye_common/armsupp.h"
diff --git a/src/core/hle/config_mem.cpp b/src/core/hle/config_mem.cpp
index aea936d2d..b1a72dc0c 100644
--- a/src/core/hle/config_mem.cpp
+++ b/src/core/hle/config_mem.cpp
@@ -25,10 +25,6 @@ void Init() {
     config_mem.sys_core_ver = 0x2;
     config_mem.unit_info = 0x1; // Bit 0 set for Retail
     config_mem.prev_firm = 0;
-    config_mem.app_mem_type = 0x2; // Default app mem type is 0
-    config_mem.app_mem_alloc = 0x06000000; // Set to 96MB, since some games use more than the default (64MB)
-    config_mem.base_mem_alloc = 0x01400000; // Default base memory is 20MB
-    config_mem.sys_mem_alloc = Memory::FCRAM_SIZE - (config_mem.app_mem_alloc + config_mem.base_mem_alloc);
     config_mem.firm_unk = 0;
     config_mem.firm_version_rev = 0;
     config_mem.firm_version_min = 0x40;
@@ -36,7 +32,4 @@ void Init() {
     config_mem.firm_sys_core_ver = 0x2;
 }
 
-void Shutdown() {
-}
-
 } // namespace
diff --git a/src/core/hle/config_mem.h b/src/core/hle/config_mem.h
index 9825a09e8..24a1254f2 100644
--- a/src/core/hle/config_mem.h
+++ b/src/core/hle/config_mem.h
@@ -52,6 +52,5 @@ static_assert(sizeof(ConfigMemDef) == Memory::CONFIG_MEMORY_SIZE, "Config Memory
 extern ConfigMemDef config_mem;
 
 void Init();
-void Shutdown();
 
 } // namespace
diff --git a/src/core/hle/function_wrappers.h b/src/core/hle/function_wrappers.h
index 1a0518926..5846a161b 100644
--- a/src/core/hle/function_wrappers.h
+++ b/src/core/hle/function_wrappers.h
@@ -172,6 +172,14 @@ template<ResultCode func(u32, s64, s64)> void Wrap() {
     FuncReturn(func(PARAM(0), param1, param2).raw);
 }
 
+template<ResultCode func(s64*, Handle, u32)> void Wrap() {
+    s64 param_1 = 0;
+    u32 retval = func(&param_1, PARAM(1), PARAM(2)).raw;
+    Core::g_app_core->SetReg(1, (u32)param_1);
+    Core::g_app_core->SetReg(2, (u32)(param_1 >> 32));
+    FuncReturn(retval);
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Function wrappers that return type u32
 
diff --git a/src/core/hle/hle.cpp b/src/core/hle/hle.cpp
index cd0a400dc..331b1b22a 100644
--- a/src/core/hle/hle.cpp
+++ b/src/core/hle/hle.cpp
@@ -34,8 +34,6 @@ void Reschedule(const char *reason) {
 
 void Init() {
     Service::Init();
-    ConfigMem::Init();
-    SharedPage::Init();
 
     g_reschedule = false;
 
@@ -43,8 +41,6 @@ void Init() {
 }
 
 void Shutdown() {
-    ConfigMem::Shutdown();
-    SharedPage::Shutdown();
     Service::Shutdown();
 
     LOG_DEBUG(Kernel, "shutdown OK");
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 5711c0405..7a401a965 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -7,11 +7,14 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 
+#include "core/hle/config_mem.h"
 #include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/memory.h"
 #include "core/hle/kernel/process.h"
+#include "core/hle/kernel/resource_limit.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/timer.h"
+#include "core/hle/shared_page.h"
 
 namespace Kernel {
 
@@ -119,6 +122,13 @@ void HandleTable::Clear() {
 
 /// Initialize the kernel
 void Init() {
+    ConfigMem::Init();
+    SharedPage::Init();
+
+    // TODO(yuriks): The memory type parameter needs to be determined by the ExHeader field instead
+    // For now it defaults to the one with a largest allocation to the app
+    Kernel::MemoryInit(2); // Allocates 96MB to the application
+
     Kernel::ResourceLimitsInit();
     Kernel::ThreadingInit();
     Kernel::TimersInit();
@@ -131,11 +141,14 @@ void Init() {
 
 /// Shutdown the kernel
 void Shutdown() {
+    g_handle_table.Clear(); // Free all kernel objects
+
     Kernel::ThreadingShutdown();
+    g_current_process = nullptr;
+
     Kernel::TimersShutdown();
     Kernel::ResourceLimitsShutdown();
-    g_handle_table.Clear(); // Free all kernel objects
-    g_current_process = nullptr;
+    Kernel::MemoryShutdown();
 }
 
 } // namespace
diff --git a/src/core/hle/kernel/memory.cpp b/src/core/hle/kernel/memory.cpp
new file mode 100644
index 000000000..e4fc5f3c4
--- /dev/null
+++ b/src/core/hle/kernel/memory.cpp
@@ -0,0 +1,136 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <map>
+#include <memory>
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+#include "common/logging/log.h"
+
+#include "core/hle/config_mem.h"
+#include "core/hle/kernel/memory.h"
+#include "core/hle/kernel/vm_manager.h"
+#include "core/hle/result.h"
+#include "core/hle/shared_page.h"
+#include "core/memory.h"
+#include "core/memory_setup.h"
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+namespace Kernel {
+
+static MemoryRegionInfo memory_regions[3];
+
+/// Size of the APPLICATION, SYSTEM and BASE memory regions (respectively) for each sytem
+/// memory configuration type.
+static const u32 memory_region_sizes[8][3] = {
+    // Old 3DS layouts
+    {0x04000000, 0x02C00000, 0x01400000}, // 0
+    { /* This appears to be unused. */ }, // 1
+    {0x06000000, 0x00C00000, 0x01400000}, // 2
+    {0x05000000, 0x01C00000, 0x01400000}, // 3
+    {0x04800000, 0x02400000, 0x01400000}, // 4
+    {0x02000000, 0x04C00000, 0x01400000}, // 5
+
+    // New 3DS layouts
+    {0x07C00000, 0x06400000, 0x02000000}, // 6
+    {0x0B200000, 0x02E00000, 0x02000000}, // 7
+};
+
+void MemoryInit(u32 mem_type) {
+    // TODO(yuriks): On the n3DS, all o3DS configurations (<=5) are forced to 6 instead.
+    ASSERT_MSG(mem_type <= 5, "New 3DS memory configuration aren't supported yet!");
+    ASSERT(mem_type != 1);
+
+    // The kernel allocation regions (APPLICATION, SYSTEM and BASE) are laid out in sequence, with
+    // the sizes specified in the memory_region_sizes table.
+    VAddr base = 0;
+    for (int i = 0; i < 3; ++i) {
+        memory_regions[i].base = base;
+        memory_regions[i].size = memory_region_sizes[mem_type][i];
+        memory_regions[i].linear_heap_memory = std::make_shared<std::vector<u8>>();
+
+        base += memory_regions[i].size;
+    }
+
+    // We must've allocated the entire FCRAM by the end
+    ASSERT(base == Memory::FCRAM_SIZE);
+
+    using ConfigMem::config_mem;
+    config_mem.app_mem_type = mem_type;
+    // app_mem_malloc does not always match the configured size for memory_region[0]: in case the
+    // n3DS type override is in effect it reports the size the game expects, not the real one.
+    config_mem.app_mem_alloc = memory_region_sizes[mem_type][0];
+    config_mem.sys_mem_alloc = memory_regions[1].size;
+    config_mem.base_mem_alloc = memory_regions[2].size;
+}
+
+void MemoryShutdown() {
+    for (auto& region : memory_regions) {
+        region.base = 0;
+        region.size = 0;
+        region.linear_heap_memory = nullptr;
+    }
+}
+
+MemoryRegionInfo* GetMemoryRegion(MemoryRegion region) {
+    switch (region) {
+    case MemoryRegion::APPLICATION:
+        return &memory_regions[0];
+    case MemoryRegion::SYSTEM:
+        return &memory_regions[1];
+    case MemoryRegion::BASE:
+        return &memory_regions[2];
+    default:
+        UNREACHABLE();
+    }
+}
+
+}
+
+namespace Memory {
+
+namespace {
+
+struct MemoryArea {
+    u32 base;
+    u32 size;
+    const char* name;
+};
+
+// We don't declare the IO regions in here since its handled by other means.
+static MemoryArea memory_areas[] = {
+    {SHARED_MEMORY_VADDR, SHARED_MEMORY_SIZE,     "Shared Memory"}, // Shared memory
+    {VRAM_VADDR,          VRAM_SIZE,              "VRAM"},          // Video memory (VRAM)
+    {DSP_RAM_VADDR,       DSP_RAM_SIZE,           "DSP RAM"},       // DSP memory
+    {TLS_AREA_VADDR,      TLS_AREA_SIZE,          "TLS Area"},      // TLS memory
+};
+
+}
+
+void Init() {
+    InitMemoryMap();
+    LOG_DEBUG(HW_Memory, "initialized OK");
+}
+
+void InitLegacyAddressSpace(Kernel::VMManager& address_space) {
+    using namespace Kernel;
+
+    for (MemoryArea& area : memory_areas) {
+        auto block = std::make_shared<std::vector<u8>>(area.size);
+        address_space.MapMemoryBlock(area.base, std::move(block), 0, area.size, MemoryState::Private).Unwrap();
+    }
+
+    auto cfg_mem_vma = address_space.MapBackingMemory(CONFIG_MEMORY_VADDR,
+            (u8*)&ConfigMem::config_mem, CONFIG_MEMORY_SIZE, MemoryState::Shared).MoveFrom();
+    address_space.Reprotect(cfg_mem_vma, VMAPermission::Read);
+
+    auto shared_page_vma = address_space.MapBackingMemory(SHARED_PAGE_VADDR,
+            (u8*)&SharedPage::shared_page, SHARED_PAGE_SIZE, MemoryState::Shared).MoveFrom();
+    address_space.Reprotect(shared_page_vma, VMAPermission::Read);
+}
+
+} // namespace
diff --git a/src/core/hle/kernel/memory.h b/src/core/hle/kernel/memory.h
new file mode 100644
index 000000000..36690b091
--- /dev/null
+++ b/src/core/hle/kernel/memory.h
@@ -0,0 +1,35 @@
+// Copyright 2014 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+
+#include "common/common_types.h"
+
+#include "core/hle/kernel/process.h"
+
+namespace Kernel {
+
+class VMManager;
+
+struct MemoryRegionInfo {
+    u32 base; // Not an address, but offset from start of FCRAM
+    u32 size;
+
+    std::shared_ptr<std::vector<u8>> linear_heap_memory;
+};
+
+void MemoryInit(u32 mem_type);
+void MemoryShutdown();
+MemoryRegionInfo* GetMemoryRegion(MemoryRegion region);
+
+}
+
+namespace Memory {
+
+void Init();
+void InitLegacyAddressSpace(Kernel::VMManager& address_space);
+
+} // namespace
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index a7892c652..124047a53 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -7,11 +7,11 @@
 #include "common/logging/log.h"
 #include "common/make_unique.h"
 
+#include "core/hle/kernel/memory.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/kernel/vm_manager.h"
-#include "core/mem_map.h"
 #include "core/memory.h"
 
 namespace Kernel {
@@ -36,8 +36,7 @@ SharedPtr<Process> Process::Create(SharedPtr<CodeSet> code_set) {
     process->codeset = std::move(code_set);
     process->flags.raw = 0;
     process->flags.memory_region = MemoryRegion::APPLICATION;
-    process->address_space = Common::make_unique<VMManager>();
-    Memory::InitLegacyAddressSpace(*process->address_space);
+    Memory::InitLegacyAddressSpace(process->vm_manager);
 
     return process;
 }
@@ -93,9 +92,11 @@ void Process::ParseKernelCaps(const u32* kernel_caps, size_t len) {
             mapping.unk_flag = false;
         } else if ((type & 0xFE0) == 0xFC0) { // 0x01FF
             // Kernel version
-            int minor = descriptor & 0xFF;
-            int major = (descriptor >> 8) & 0xFF;
-            LOG_INFO(Loader, "ExHeader kernel version ignored: %d.%d", major, minor);
+            kernel_version = descriptor & 0xFFFF;
+
+            int minor = kernel_version & 0xFF;
+            int major = (kernel_version >> 8) & 0xFF;
+            LOG_INFO(Loader, "ExHeader kernel version: %d.%d", major, minor);
         } else {
             LOG_ERROR(Loader, "Unhandled kernel caps descriptor: 0x%08X", descriptor);
         }
@@ -103,20 +104,153 @@ void Process::ParseKernelCaps(const u32* kernel_caps, size_t len) {
 }
 
 void Process::Run(s32 main_thread_priority, u32 stack_size) {
+    memory_region = GetMemoryRegion(flags.memory_region);
+
     auto MapSegment = [&](CodeSet::Segment& segment, VMAPermission permissions, MemoryState memory_state) {
-        auto vma = address_space->MapMemoryBlock(segment.addr, codeset->memory,
+        auto vma = vm_manager.MapMemoryBlock(segment.addr, codeset->memory,
                 segment.offset, segment.size, memory_state).Unwrap();
-        address_space->Reprotect(vma, permissions);
+        vm_manager.Reprotect(vma, permissions);
+        misc_memory_used += segment.size;
     };
 
+    // Map CodeSet segments
     MapSegment(codeset->code,   VMAPermission::ReadExecute, MemoryState::Code);
     MapSegment(codeset->rodata, VMAPermission::Read,        MemoryState::Code);
     MapSegment(codeset->data,   VMAPermission::ReadWrite,   MemoryState::Private);
 
-    address_space->LogLayout();
+    // Allocate and map stack
+    vm_manager.MapMemoryBlock(Memory::HEAP_VADDR_END - stack_size,
+            std::make_shared<std::vector<u8>>(stack_size, 0), 0, stack_size, MemoryState::Locked
+            ).Unwrap();
+    misc_memory_used += stack_size;
+
+    vm_manager.LogLayout(Log::Level::Debug);
     Kernel::SetupMainThread(codeset->entrypoint, main_thread_priority);
 }
 
+VAddr Process::GetLinearHeapBase() const {
+    return (kernel_version < 0x22C ? Memory::LINEAR_HEAP_VADDR : Memory::NEW_LINEAR_HEAP_SIZE)
+            + memory_region->base;
+}
+
+VAddr Process::GetLinearHeapLimit() const {
+    return GetLinearHeapBase() + memory_region->size;
+}
+
+ResultVal<VAddr> Process::HeapAllocate(VAddr target, u32 size, VMAPermission perms) {
+    if (target < Memory::HEAP_VADDR || target + size > Memory::HEAP_VADDR_END || target + size < target) {
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (heap_memory == nullptr) {
+        // Initialize heap
+        heap_memory = std::make_shared<std::vector<u8>>();
+        heap_start = heap_end = target;
+    }
+
+    // If necessary, expand backing vector to cover new heap extents.
+    if (target < heap_start) {
+        heap_memory->insert(begin(*heap_memory), heap_start - target, 0);
+        heap_start = target;
+        vm_manager.RefreshMemoryBlockMappings(heap_memory.get());
+    }
+    if (target + size > heap_end) {
+        heap_memory->insert(end(*heap_memory), (target + size) - heap_end, 0);
+        heap_end = target + size;
+        vm_manager.RefreshMemoryBlockMappings(heap_memory.get());
+    }
+    ASSERT(heap_end - heap_start == heap_memory->size());
+
+    CASCADE_RESULT(auto vma, vm_manager.MapMemoryBlock(target, heap_memory, target - heap_start, size, MemoryState::Private));
+    vm_manager.Reprotect(vma, perms);
+
+    heap_used += size;
+
+    return MakeResult<VAddr>(heap_end - size);
+}
+
+ResultCode Process::HeapFree(VAddr target, u32 size) {
+    if (target < Memory::HEAP_VADDR || target + size > Memory::HEAP_VADDR_END || target + size < target) {
+        return ERR_INVALID_ADDRESS;
+    }
+
+    ResultCode result = vm_manager.UnmapRange(target, size);
+    if (result.IsError()) return result;
+
+    heap_used -= size;
+
+    return RESULT_SUCCESS;
+}
+
+ResultVal<VAddr> Process::LinearAllocate(VAddr target, u32 size, VMAPermission perms) {
+    auto& linheap_memory = memory_region->linear_heap_memory;
+
+    VAddr heap_end = GetLinearHeapBase() + (u32)linheap_memory->size();
+    // Games and homebrew only ever seem to pass 0 here (which lets the kernel decide the address),
+    // but explicit addresses are also accepted and respected.
+    if (target == 0) {
+        target = heap_end;
+    }
+
+    if (target < GetLinearHeapBase() || target + size > GetLinearHeapLimit() ||
+        target > heap_end || target + size < target) {
+
+        return ERR_INVALID_ADDRESS;
+    }
+
+    // Expansion of the linear heap is only allowed if you do an allocation immediatelly at its
+    // end. It's possible to free gaps in the middle of the heap and then reallocate them later,
+    // but expansions are only allowed at the end.
+    if (target == heap_end) {
+        linheap_memory->insert(linheap_memory->end(), size, 0);
+        vm_manager.RefreshMemoryBlockMappings(linheap_memory.get());
+    }
+
+    // TODO(yuriks): As is, this lets processes map memory allocated by other processes from the
+    // same region. It is unknown if or how the 3DS kernel checks against this.
+    size_t offset = target - GetLinearHeapBase();
+    CASCADE_RESULT(auto vma, vm_manager.MapMemoryBlock(target, linheap_memory, offset, size, MemoryState::Continuous));
+    vm_manager.Reprotect(vma, perms);
+
+    linear_heap_used += size;
+
+    return MakeResult<VAddr>(target);
+}
+
+ResultCode Process::LinearFree(VAddr target, u32 size) {
+    auto& linheap_memory = memory_region->linear_heap_memory;
+
+    if (target < GetLinearHeapBase() || target + size > GetLinearHeapLimit() ||
+        target + size < target) {
+
+        return ERR_INVALID_ADDRESS;
+    }
+
+    VAddr heap_end = GetLinearHeapBase() + (u32)linheap_memory->size();
+    if (target + size > heap_end) {
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    ResultCode result = vm_manager.UnmapRange(target, size);
+    if (result.IsError()) return result;
+
+    linear_heap_used -= size;
+
+    if (target + size == heap_end) {
+        // End of linear heap has been freed, so check what's the last allocated block in it and
+        // reduce the size.
+        auto vma = vm_manager.FindVMA(target);
+        ASSERT(vma != vm_manager.vma_map.end());
+        ASSERT(vma->second.type == VMAType::Free);
+        VAddr new_end = vma->second.base;
+        if (new_end >= GetLinearHeapBase()) {
+            linheap_memory->resize(new_end - GetLinearHeapBase());
+        }
+    }
+
+    return RESULT_SUCCESS;
+}
+
 Kernel::Process::Process() {}
 Kernel::Process::~Process() {}
 
diff --git a/src/core/hle/kernel/process.h b/src/core/hle/kernel/process.h
index 83d3aceae..60e17f251 100644
--- a/src/core/hle/kernel/process.h
+++ b/src/core/hle/kernel/process.h
@@ -15,6 +15,7 @@
 #include "common/common_types.h"
 
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/vm_manager.h"
 
 namespace Kernel {
 
@@ -48,7 +49,7 @@ union ProcessFlags {
 };
 
 class ResourceLimit;
-class VMManager;
+struct MemoryRegionInfo;
 
 struct CodeSet final : public Object {
     static SharedPtr<CodeSet> Create(std::string name, u64 program_id);
@@ -104,14 +105,12 @@ public:
     /// processes access to specific I/O regions and device memory.
     boost::container::static_vector<AddressMapping, 8> address_mappings;
     ProcessFlags flags;
+    /// Kernel compatibility version for this process
+    u16 kernel_version = 0;
 
     /// The id of this process
     u32 process_id = next_process_id++;
 
-    /// Bitmask of the used TLS slots
-    std::bitset<300> used_tls_slots;
-    std::unique_ptr<VMManager> address_space;
-
     /**
      * Parses a list of kernel capability descriptors (as found in the ExHeader) and applies them
      * to this process.
@@ -123,6 +122,36 @@ public:
      */
     void Run(s32 main_thread_priority, u32 stack_size);
 
+
+    ///////////////////////////////////////////////////////////////////////////////////////////////
+    // Memory Management
+
+    VMManager vm_manager;
+
+    // Memory used to back the allocations in the regular heap. A single vector is used to cover
+    // the entire virtual address space extents that bound the allocations, including any holes.
+    // This makes deallocation and reallocation of holes fast and keeps process memory contiguous
+    // in the emulator address space, allowing Memory::GetPointer to be reasonably safe.
+    std::shared_ptr<std::vector<u8>> heap_memory;
+    // The left/right bounds of the address space covered by heap_memory.
+    VAddr heap_start = 0, heap_end = 0;
+
+    u32 heap_used = 0, linear_heap_used = 0, misc_memory_used = 0;
+
+    MemoryRegionInfo* memory_region = nullptr;
+
+    /// Bitmask of the used TLS slots
+    std::bitset<300> used_tls_slots;
+
+    VAddr GetLinearHeapBase() const;
+    VAddr GetLinearHeapLimit() const;
+
+    ResultVal<VAddr> HeapAllocate(VAddr target, u32 size, VMAPermission perms);
+    ResultCode HeapFree(VAddr target, u32 size);
+
+    ResultVal<VAddr> LinearAllocate(VAddr target, u32 size, VMAPermission perms);
+    ResultCode LinearFree(VAddr target, u32 size);
+
 private:
     Process();
     ~Process() override;
diff --git a/src/core/hle/kernel/resource_limit.cpp b/src/core/hle/kernel/resource_limit.cpp
index 94b3e3298..67dde08c2 100644
--- a/src/core/hle/kernel/resource_limit.cpp
+++ b/src/core/hle/kernel/resource_limit.cpp
@@ -6,7 +6,6 @@
 
 #include "common/logging/log.h"
 
-#include "core/mem_map.h"
 #include "core/hle/kernel/resource_limit.h"
 
 namespace Kernel {
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index 29ea6d531..c10126513 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -117,6 +117,7 @@ void Thread::Stop() {
     wait_objects.clear();
 
     Kernel::g_current_process->used_tls_slots[tls_index] = false;
+    g_current_process->misc_memory_used -= Memory::TLS_ENTRY_SIZE;
 
     HLE::Reschedule(__func__);
 }
@@ -414,6 +415,7 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point,
     }
 
     ASSERT_MSG(thread->tls_index != -1, "Out of TLS space");
+    g_current_process->misc_memory_used += Memory::TLS_ENTRY_SIZE;
 
     // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
     // to initialize the context
@@ -504,7 +506,7 @@ void Thread::SetWaitSynchronizationOutput(s32 output) {
 }
 
 VAddr Thread::GetTLSAddress() const {
-    return Memory::TLS_AREA_VADDR + tls_index * 0x200;
+    return Memory::TLS_AREA_VADDR + tls_index * Memory::TLS_ENTRY_SIZE;
 }
 
 ////////////////////////////////////////////////////////////////////////////////////////////////////
diff --git a/src/core/hle/kernel/vm_manager.cpp b/src/core/hle/kernel/vm_manager.cpp
index 205cc7b53..2610acf76 100644
--- a/src/core/hle/kernel/vm_manager.cpp
+++ b/src/core/hle/kernel/vm_manager.cpp
@@ -11,6 +11,15 @@
 
 namespace Kernel {
 
+static const char* GetMemoryStateName(MemoryState state) {
+    static const char* names[] = {
+        "Free", "Reserved", "IO", "Static", "Code", "Private", "Shared", "Continuous", "Aliased",
+        "Alias", "AliasCode", "Locked",
+    };
+
+    return names[(int)state];
+}
+
 bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const {
     ASSERT(base + size == next.base);
     if (permissions != next.permissions ||
@@ -51,11 +60,15 @@ void VMManager::Reset() {
 }
 
 VMManager::VMAHandle VMManager::FindVMA(VAddr target) const {
-    return std::prev(vma_map.upper_bound(target));
+    if (target >= MAX_ADDRESS) {
+        return vma_map.end();
+    } else {
+        return std::prev(vma_map.upper_bound(target));
+    }
 }
 
 ResultVal<VMManager::VMAHandle> VMManager::MapMemoryBlock(VAddr target,
-        std::shared_ptr<std::vector<u8>> block, u32 offset, u32 size, MemoryState state) {
+        std::shared_ptr<std::vector<u8>> block, size_t offset, u32 size, MemoryState state) {
     ASSERT(block != nullptr);
     ASSERT(offset + size <= block->size());
 
@@ -106,10 +119,8 @@ ResultVal<VMManager::VMAHandle> VMManager::MapMMIO(VAddr target, PAddr paddr, u3
     return MakeResult<VMAHandle>(MergeAdjacent(vma_handle));
 }
 
-void VMManager::Unmap(VMAHandle vma_handle) {
-    VMAIter iter = StripIterConstness(vma_handle);
-
-    VirtualMemoryArea& vma = iter->second;
+VMManager::VMAIter VMManager::Unmap(VMAIter vma_handle) {
+    VirtualMemoryArea& vma = vma_handle->second;
     vma.type = VMAType::Free;
     vma.permissions = VMAPermission::None;
     vma.meminfo_state = MemoryState::Free;
@@ -121,26 +132,67 @@ void VMManager::Unmap(VMAHandle vma_handle) {
 
     UpdatePageTableForVMA(vma);
 
-    MergeAdjacent(iter);
+    return MergeAdjacent(vma_handle);
+}
+
+ResultCode VMManager::UnmapRange(VAddr target, u32 size) {
+    CASCADE_RESULT(VMAIter vma, CarveVMARange(target, size));
+    VAddr target_end = target + size;
+
+    VMAIter end = vma_map.end();
+    // The comparison against the end of the range must be done using addresses since VMAs can be
+    // merged during this process, causing invalidation of the iterators.
+    while (vma != end && vma->second.base < target_end) {
+        vma = std::next(Unmap(vma));
+    }
+
+    ASSERT(FindVMA(target)->second.size >= size);
+    return RESULT_SUCCESS;
 }
 
-void VMManager::Reprotect(VMAHandle vma_handle, VMAPermission new_perms) {
+VMManager::VMAHandle VMManager::Reprotect(VMAHandle vma_handle, VMAPermission new_perms) {
     VMAIter iter = StripIterConstness(vma_handle);
 
     VirtualMemoryArea& vma = iter->second;
     vma.permissions = new_perms;
     UpdatePageTableForVMA(vma);
 
-    MergeAdjacent(iter);
+    return MergeAdjacent(iter);
+}
+
+ResultCode VMManager::ReprotectRange(VAddr target, u32 size, VMAPermission new_perms) {
+    CASCADE_RESULT(VMAIter vma, CarveVMARange(target, size));
+    VAddr target_end = target + size;
+
+    VMAIter end = vma_map.end();
+    // The comparison against the end of the range must be done using addresses since VMAs can be
+    // merged during this process, causing invalidation of the iterators.
+    while (vma != end && vma->second.base < target_end) {
+        vma = std::next(StripIterConstness(Reprotect(vma, new_perms)));
+    }
+
+    return RESULT_SUCCESS;
 }
 
-void VMManager::LogLayout() const {
+void VMManager::RefreshMemoryBlockMappings(const std::vector<u8>* block) {
+    // If this ever proves to have a noticeable performance impact, allow users of the function to
+    // specify a specific range of addresses to limit the scan to.
     for (const auto& p : vma_map) {
         const VirtualMemoryArea& vma = p.second;
-        LOG_DEBUG(Kernel, "%08X - %08X  size: %8X %c%c%c", vma.base, vma.base + vma.size, vma.size,
+        if (block == vma.backing_block.get()) {
+            UpdatePageTableForVMA(vma);
+        }
+    }
+}
+
+void VMManager::LogLayout(Log::Level log_level) const {
+    for (const auto& p : vma_map) {
+        const VirtualMemoryArea& vma = p.second;
+        LOG_GENERIC(Log::Class::Kernel, log_level, "%08X - %08X  size: %8X %c%c%c %s",
+            vma.base, vma.base + vma.size, vma.size,
             (u8)vma.permissions & (u8)VMAPermission::Read    ? 'R' : '-',
             (u8)vma.permissions & (u8)VMAPermission::Write   ? 'W' : '-',
-            (u8)vma.permissions & (u8)VMAPermission::Execute ? 'X' : '-');
+            (u8)vma.permissions & (u8)VMAPermission::Execute ? 'X' : '-', GetMemoryStateName(vma.meminfo_state));
     }
 }
 
@@ -151,21 +203,19 @@ VMManager::VMAIter VMManager::StripIterConstness(const VMAHandle & iter) {
 }
 
 ResultVal<VMManager::VMAIter> VMManager::CarveVMA(VAddr base, u32 size) {
-    ASSERT_MSG((size & Memory::PAGE_MASK) == 0, "non-page aligned size: %8X", size);
-    ASSERT_MSG((base & Memory::PAGE_MASK) == 0, "non-page aligned base: %08X", base);
+    ASSERT_MSG((size & Memory::PAGE_MASK) == 0, "non-page aligned size: 0x%8X", size);
+    ASSERT_MSG((base & Memory::PAGE_MASK) == 0, "non-page aligned base: 0x%08X", base);
 
     VMAIter vma_handle = StripIterConstness(FindVMA(base));
     if (vma_handle == vma_map.end()) {
         // Target address is outside the range managed by the kernel
-        return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS,
-                ErrorSummary::InvalidArgument, ErrorLevel::Usage); // 0xE0E01BF5
+        return ERR_INVALID_ADDRESS;
     }
 
     VirtualMemoryArea& vma = vma_handle->second;
     if (vma.type != VMAType::Free) {
         // Region is already allocated
-        return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS,
-                ErrorSummary::InvalidState, ErrorLevel::Usage); // 0xE0A01BF5
+        return ERR_INVALID_ADDRESS_STATE;
     }
 
     u32 start_in_vma = base - vma.base;
@@ -173,8 +223,7 @@ ResultVal<VMManager::VMAIter> VMManager::CarveVMA(VAddr base, u32 size) {
 
     if (end_in_vma > vma.size) {
         // Requested allocation doesn't fit inside VMA
-        return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS,
-                ErrorSummary::InvalidState, ErrorLevel::Usage); // 0xE0A01BF5
+        return ERR_INVALID_ADDRESS_STATE;
     }
 
     if (end_in_vma != vma.size) {
@@ -189,6 +238,35 @@ ResultVal<VMManager::VMAIter> VMManager::CarveVMA(VAddr base, u32 size) {
     return MakeResult<VMAIter>(vma_handle);
 }
 
+ResultVal<VMManager::VMAIter> VMManager::CarveVMARange(VAddr target, u32 size) {
+    ASSERT_MSG((size & Memory::PAGE_MASK) == 0, "non-page aligned size: 0x%8X", size);
+    ASSERT_MSG((target & Memory::PAGE_MASK) == 0, "non-page aligned base: 0x%08X", target);
+
+    VAddr target_end = target + size;
+    ASSERT(target_end >= target);
+    ASSERT(target_end <= MAX_ADDRESS);
+    ASSERT(size > 0);
+
+    VMAIter begin_vma = StripIterConstness(FindVMA(target));
+    VMAIter i_end = vma_map.lower_bound(target_end);
+    for (auto i = begin_vma; i != i_end; ++i) {
+        if (i->second.type == VMAType::Free) {
+            return ERR_INVALID_ADDRESS_STATE;
+        }
+    }
+
+    if (target != begin_vma->second.base) {
+        begin_vma = SplitVMA(begin_vma, target - begin_vma->second.base);
+    }
+
+    VMAIter end_vma = StripIterConstness(FindVMA(target_end));
+    if (end_vma != vma_map.end() && target_end != end_vma->second.base) {
+        end_vma = SplitVMA(end_vma, target_end - end_vma->second.base);
+    }
+
+    return MakeResult<VMAIter>(begin_vma);
+}
+
 VMManager::VMAIter VMManager::SplitVMA(VMAIter vma_handle, u32 offset_in_vma) {
     VirtualMemoryArea& old_vma = vma_handle->second;
     VirtualMemoryArea new_vma = old_vma; // Make a copy of the VMA
diff --git a/src/core/hle/kernel/vm_manager.h b/src/core/hle/kernel/vm_manager.h
index b3795a94a..4e95f1f0c 100644
--- a/src/core/hle/kernel/vm_manager.h
+++ b/src/core/hle/kernel/vm_manager.h
@@ -14,6 +14,14 @@
 
 namespace Kernel {
 
+const ResultCode ERR_INVALID_ADDRESS{ // 0xE0E01BF5
+        ErrorDescription::InvalidAddress, ErrorModule::OS,
+        ErrorSummary::InvalidArgument, ErrorLevel::Usage};
+
+const ResultCode ERR_INVALID_ADDRESS_STATE{ // 0xE0A01BF5
+        ErrorDescription::InvalidAddress, ErrorModule::OS,
+        ErrorSummary::InvalidState, ErrorLevel::Usage};
+
 enum class VMAType : u8 {
     /// VMA represents an unmapped region of the address space.
     Free,
@@ -75,7 +83,7 @@ struct VirtualMemoryArea {
     /// Memory block backing this VMA.
     std::shared_ptr<std::vector<u8>> backing_block = nullptr;
     /// Offset into the backing_memory the mapping starts from.
-    u32 offset = 0;
+    size_t offset = 0;
 
     // Settings for type = BackingMemory
     /// Pointer backing this VMA. It will not be destroyed or freed when the VMA is removed.
@@ -141,7 +149,7 @@ public:
      * @param state MemoryState tag to attach to the VMA.
      */
     ResultVal<VMAHandle> MapMemoryBlock(VAddr target, std::shared_ptr<std::vector<u8>> block,
-            u32 offset, u32 size, MemoryState state);
+            size_t offset, u32 size, MemoryState state);
 
     /**
      * Maps an unmanaged host memory pointer at a given address.
@@ -163,14 +171,23 @@ public:
      */
     ResultVal<VMAHandle> MapMMIO(VAddr target, PAddr paddr, u32 size, MemoryState state);
 
-    /// Unmaps the given VMA.
-    void Unmap(VMAHandle vma);
+    /// Unmaps a range of addresses, splitting VMAs as necessary.
+    ResultCode UnmapRange(VAddr target, u32 size);
 
     /// Changes the permissions of the given VMA.
-    void Reprotect(VMAHandle vma, VMAPermission new_perms);
+    VMAHandle Reprotect(VMAHandle vma, VMAPermission new_perms);
+
+    /// Changes the permissions of a range of addresses, splitting VMAs as necessary.
+    ResultCode ReprotectRange(VAddr target, u32 size, VMAPermission new_perms);
+
+    /**
+     * Scans all VMAs and updates the page table range of any that use the given vector as backing
+     * memory. This should be called after any operation that causes reallocation of the vector.
+     */
+    void RefreshMemoryBlockMappings(const std::vector<u8>* block);
 
     /// Dumps the address space layout to the log, for debugging
-    void LogLayout() const;
+    void LogLayout(Log::Level log_level) const;
 
 private:
     using VMAIter = decltype(vma_map)::iterator;
@@ -178,6 +195,9 @@ private:
     /// Converts a VMAHandle to a mutable VMAIter.
     VMAIter StripIterConstness(const VMAHandle& iter);
 
+    /// Unmaps the given VMA.
+    VMAIter Unmap(VMAIter vma);
+
     /**
      * Carves a VMA of a specific size at the specified address by splitting Free VMAs while doing
      * the appropriate error checking.
@@ -185,6 +205,12 @@ private:
     ResultVal<VMAIter> CarveVMA(VAddr base, u32 size);
 
     /**
+     * Splits the edges of the given range of non-Free VMAs so that there is a VMA split at each
+     * end of the range.
+     */
+    ResultVal<VMAIter> CarveVMARange(VAddr base, u32 size);
+
+    /**
      * Splits a VMA in two, at the specified offset.
      * @returns the right side of the split, with the original iterator becoming the left side.
      */
diff --git a/src/core/hle/service/apt/apt.cpp b/src/core/hle/service/apt/apt.cpp
index 35402341b..6a2fdea2b 100644
--- a/src/core/hle/service/apt/apt.cpp
+++ b/src/core/hle/service/apt/apt.cpp
@@ -16,6 +16,7 @@
 #include "core/hle/hle.h"
 #include "core/hle/kernel/event.h"
 #include "core/hle/kernel/mutex.h"
+#include "core/hle/kernel/process.h"
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/kernel/thread.h"
 
@@ -37,7 +38,7 @@ static Kernel::SharedPtr<Kernel::Mutex> lock;
 static Kernel::SharedPtr<Kernel::Event> notification_event; ///< APT notification event
 static Kernel::SharedPtr<Kernel::Event> parameter_event; ///< APT parameter event
 
-static std::vector<u8> shared_font;
+static std::shared_ptr<std::vector<u8>> shared_font;
 
 static u32 cpu_percent; ///< CPU time available to the running application
 
@@ -74,11 +75,12 @@ void Initialize(Service::Interface* self) {
 void GetSharedFont(Service::Interface* self) {
     u32* cmd_buff = Kernel::GetCommandBuffer();
 
-    if (!shared_font.empty()) {
-        // TODO(bunnei): This function shouldn't copy the shared font every time it's called.
-        // Instead, it should probably map the shared font as RO memory. We don't currently have
-        // an easy way to do this, but the copy should be sufficient for now.
-        memcpy(Memory::GetPointer(SHARED_FONT_VADDR), shared_font.data(), shared_font.size());
+    if (shared_font != nullptr) {
+        // TODO(yuriks): This is a hack to keep this working right now even with our completely
+        // broken shared memory system.
+        shared_font_mem->base_address = SHARED_FONT_VADDR;
+        Kernel::g_current_process->vm_manager.MapMemoryBlock(shared_font_mem->base_address,
+                shared_font, 0, shared_font_mem->size, Kernel::MemoryState::Shared);
 
         cmd_buff[0] = IPC::MakeHeader(0x44, 2, 2);
         cmd_buff[1] = RESULT_SUCCESS.raw; // No error
@@ -391,7 +393,6 @@ void Init() {
     // a homebrew app to do this: https://github.com/citra-emu/3dsutils. Put the resulting file
     // "shared_font.bin" in the Citra "sysdata" directory.
 
-    shared_font.clear();
     std::string filepath = FileUtil::GetUserPath(D_SYSDATA_IDX) + SHARED_FONT;
 
     FileUtil::CreateFullPath(filepath); // Create path if not already created
@@ -399,8 +400,8 @@ void Init() {
 
     if (file.IsOpen()) {
         // Read shared font data
-        shared_font.resize((size_t)file.GetSize());
-        file.ReadBytes(shared_font.data(), (size_t)file.GetSize());
+        shared_font = std::make_shared<std::vector<u8>>((size_t)file.GetSize());
+        file.ReadBytes(shared_font->data(), shared_font->size());
 
         // Create shared font memory object
         using Kernel::MemoryPermission;
@@ -424,7 +425,7 @@ void Init() {
 }
 
 void Shutdown() {
-    shared_font.clear();
+    shared_font = nullptr;
     shared_font_mem = nullptr;
     lock = nullptr;
     notification_event = nullptr;
diff --git a/src/core/hle/service/gsp_gpu.cpp b/src/core/hle/service/gsp_gpu.cpp
index e93c1b436..fde508a13 100644
--- a/src/core/hle/service/gsp_gpu.cpp
+++ b/src/core/hle/service/gsp_gpu.cpp
@@ -4,7 +4,6 @@
 
 #include "common/bit_field.h"
 
-#include "core/mem_map.h"
 #include "core/memory.h"
 #include "core/hle/kernel/event.h"
 #include "core/hle/kernel/shared_memory.h"
@@ -418,7 +417,7 @@ static void ExecuteCommand(const Command& command, u32 thread_id) {
 
     case CommandId::SET_DISPLAY_TRANSFER:
     {
-        auto& params = command.image_copy;
+        auto& params = command.display_transfer;
         WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.input_address)),
                 Memory::VirtualToPhysicalAddress(params.in_buffer_address) >> 3);
         WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.output_address)),
@@ -433,17 +432,22 @@ static void ExecuteCommand(const Command& command, u32 thread_id) {
     // TODO: Check if texture copies are implemented correctly..
     case CommandId::SET_TEXTURE_COPY:
     {
-        auto& params = command.image_copy;
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.input_address)),
+        auto& params = command.texture_copy;
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.input_address),
                 Memory::VirtualToPhysicalAddress(params.in_buffer_address) >> 3);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.output_address)),
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.output_address),
                 Memory::VirtualToPhysicalAddress(params.out_buffer_address) >> 3);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.input_size)), params.in_buffer_size);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.output_size)), params.out_buffer_size);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.flags)), params.flags);
-
-        // TODO: Should this register be set to 1 or should instead its value be OR-ed with 1?
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.trigger)), 1);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.texture_copy.size),
+                params.size);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.texture_copy.input_size),
+                params.in_width_gap);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.texture_copy.output_size),
+                params.out_width_gap);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.flags),
+                params.flags);
+
+        // NOTE: Actual GSP ORs 1 with current register instead of overwriting. Doesn't seem to matter.
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.trigger), 1);
         break;
     }
 
diff --git a/src/core/hle/service/gsp_gpu.h b/src/core/hle/service/gsp_gpu.h
index c89d0a467..8bcb30ad1 100644
--- a/src/core/hle/service/gsp_gpu.h
+++ b/src/core/hle/service/gsp_gpu.h
@@ -127,7 +127,16 @@ struct Command {
             u32 in_buffer_size;
             u32 out_buffer_size;
             u32 flags;
-        } image_copy;
+        } display_transfer;
+
+        struct {
+            u32 in_buffer_address;
+            u32 out_buffer_address;
+            u32 size;
+            u32 in_width_gap;
+            u32 out_width_gap;
+            u32 flags;
+        } texture_copy;
 
         u8 raw_data[0x1C];
     };
diff --git a/src/core/hle/service/y2r_u.cpp b/src/core/hle/service/y2r_u.cpp
index 6e7dafaad..6b1b71fe4 100644
--- a/src/core/hle/service/y2r_u.cpp
+++ b/src/core/hle/service/y2r_u.cpp
@@ -10,7 +10,6 @@
 #include "core/hle/kernel/event.h"
 #include "core/hle/service/y2r_u.h"
 #include "core/hw/y2r.h"
-#include "core/mem_map.h"
 
 #include "video_core/renderer_base.h"
 #include "video_core/utils.h"
diff --git a/src/core/hle/shared_page.cpp b/src/core/hle/shared_page.cpp
index 26d87c7e2..50c5bc01b 100644
--- a/src/core/hle/shared_page.cpp
+++ b/src/core/hle/shared_page.cpp
@@ -18,7 +18,4 @@ void Init() {
     shared_page.running_hw = 0x1; // product
 }
 
-void Shutdown() {
-}
-
 } // namespace
diff --git a/src/core/hle/shared_page.h b/src/core/hle/shared_page.h
index db6a5340b..379bb7b63 100644
--- a/src/core/hle/shared_page.h
+++ b/src/core/hle/shared_page.h
@@ -54,6 +54,5 @@ static_assert(sizeof(SharedPageDef) == Memory::SHARED_PAGE_SIZE, "Shared page st
 extern SharedPageDef shared_page;
 
 void Init();
-void Shutdown();
 
 } // namespace
diff --git a/src/core/hle/svc.cpp b/src/core/hle/svc.cpp
index bb64fdfb7..89ac45a6f 100644
--- a/src/core/hle/svc.cpp
+++ b/src/core/hle/svc.cpp
@@ -10,11 +10,11 @@
 #include "common/symbols.h"
 
 #include "core/core_timing.h"
-#include "core/mem_map.h"
 #include "core/arm/arm_interface.h"
 
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/event.h"
+#include "core/hle/kernel/memory.h"
 #include "core/hle/kernel/mutex.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
@@ -41,32 +41,114 @@ const ResultCode ERR_NOT_FOUND(ErrorDescription::NotFound, ErrorModule::Kernel,
 const ResultCode ERR_PORT_NAME_TOO_LONG(ErrorDescription(30), ErrorModule::OS,
         ErrorSummary::InvalidArgument, ErrorLevel::Usage); // 0xE0E0181E
 
+const ResultCode ERR_MISALIGNED_ADDRESS{ // 0xE0E01BF1
+        ErrorDescription::MisalignedAddress, ErrorModule::OS,
+        ErrorSummary::InvalidArgument, ErrorLevel::Usage};
+const ResultCode ERR_MISALIGNED_SIZE{ // 0xE0E01BF2
+        ErrorDescription::MisalignedSize, ErrorModule::OS,
+        ErrorSummary::InvalidArgument, ErrorLevel::Usage};
+const ResultCode ERR_INVALID_COMBINATION{ // 0xE0E01BEE
+        ErrorDescription::InvalidCombination, ErrorModule::OS,
+        ErrorSummary::InvalidArgument, ErrorLevel::Usage};
+
 enum ControlMemoryOperation {
-    MEMORY_OPERATION_HEAP       = 0x00000003,
-    MEMORY_OPERATION_GSP_HEAP   = 0x00010003,
+    MEMOP_FREE    = 1,
+    MEMOP_RESERVE = 2, // This operation seems to be unsupported in the kernel
+    MEMOP_COMMIT  = 3,
+    MEMOP_MAP     = 4,
+    MEMOP_UNMAP   = 5,
+    MEMOP_PROTECT = 6,
+    MEMOP_OPERATION_MASK = 0xFF,
+
+    MEMOP_REGION_APP    = 0x100,
+    MEMOP_REGION_SYSTEM = 0x200,
+    MEMOP_REGION_BASE   = 0x300,
+    MEMOP_REGION_MASK   = 0xF00,
+
+    MEMOP_LINEAR = 0x10000,
 };
 
 /// Map application or GSP heap memory
 static ResultCode ControlMemory(u32* out_addr, u32 operation, u32 addr0, u32 addr1, u32 size, u32 permissions) {
-    LOG_TRACE(Kernel_SVC,"called operation=0x%08X, addr0=0x%08X, addr1=0x%08X, size=%08X, permissions=0x%08X",
+    using namespace Kernel;
+
+    LOG_DEBUG(Kernel_SVC,"called operation=0x%08X, addr0=0x%08X, addr1=0x%08X, size=0x%X, permissions=0x%08X",
         operation, addr0, addr1, size, permissions);
 
-    switch (operation) {
+    if ((addr0 & Memory::PAGE_MASK) != 0 || (addr1 & Memory::PAGE_MASK) != 0) {
+        return ERR_MISALIGNED_ADDRESS;
+    }
+    if ((size & Memory::PAGE_MASK) != 0) {
+        return ERR_MISALIGNED_SIZE;
+    }
+
+    u32 region = operation & MEMOP_REGION_MASK;
+    operation &= ~MEMOP_REGION_MASK;
+
+    if (region != 0) {
+        LOG_WARNING(Kernel_SVC, "ControlMemory with specified region not supported, region=%X", region);
+    }
+
+    if ((permissions & (u32)MemoryPermission::ReadWrite) != permissions) {
+        return ERR_INVALID_COMBINATION;
+    }
+    VMAPermission vma_permissions = (VMAPermission)permissions;
+
+    auto& process = *g_current_process;
+
+    switch (operation & MEMOP_OPERATION_MASK) {
+    case MEMOP_FREE:
+    {
+        if (addr0 >= Memory::HEAP_VADDR && addr0 < Memory::HEAP_VADDR_END) {
+            ResultCode result = process.HeapFree(addr0, size);
+            if (result.IsError()) return result;
+        } else if (addr0 >= process.GetLinearHeapBase() && addr0 < process.GetLinearHeapLimit()) {
+            ResultCode result = process.LinearFree(addr0, size);
+            if (result.IsError()) return result;
+        } else {
+            return ERR_INVALID_ADDRESS;
+        }
+        *out_addr = addr0;
+        break;
+    }
+
+    case MEMOP_COMMIT:
+    {
+        if (operation & MEMOP_LINEAR) {
+            CASCADE_RESULT(*out_addr, process.LinearAllocate(addr0, size, vma_permissions));
+        } else {
+            CASCADE_RESULT(*out_addr, process.HeapAllocate(addr0, size, vma_permissions));
+        }
+        break;
+    }
 
-    // Map normal heap memory
-    case MEMORY_OPERATION_HEAP:
-        *out_addr = Memory::MapBlock_Heap(size, operation, permissions);
+    case MEMOP_MAP: // TODO: This is just a hack to avoid regressions until memory aliasing is implemented
+    {
+        CASCADE_RESULT(*out_addr, process.HeapAllocate(addr0, size, vma_permissions));
         break;
+    }
+
+    case MEMOP_UNMAP: // TODO: This is just a hack to avoid regressions until memory aliasing is implemented
+    {
+        ResultCode result = process.HeapFree(addr0, size);
+        if (result.IsError()) return result;
+        break;
+    }
 
-    // Map GSP heap memory
-    case MEMORY_OPERATION_GSP_HEAP:
-        *out_addr = Memory::MapBlock_HeapLinear(size, operation, permissions);
+    case MEMOP_PROTECT:
+    {
+        ResultCode result = process.vm_manager.ReprotectRange(addr0, size, vma_permissions);
+        if (result.IsError()) return result;
         break;
+    }
 
-    // Unknown ControlMemory operation
     default:
         LOG_ERROR(Kernel_SVC, "unknown operation=0x%08X", operation);
+        return ERR_INVALID_COMBINATION;
     }
+
+    process.vm_manager.LogLayout(Log::Level::Trace);
+
     return RESULT_SUCCESS;
 }
 
@@ -537,9 +619,9 @@ static ResultCode QueryProcessMemory(MemoryInfo* memory_info, PageInfo* page_inf
     if (process == nullptr)
         return ERR_INVALID_HANDLE;
 
-    auto vma = process->address_space->FindVMA(addr);
+    auto vma = process->vm_manager.FindVMA(addr);
 
-    if (vma == process->address_space->vma_map.end())
+    if (vma == Kernel::g_current_process->vm_manager.vma_map.end())
         return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS, ErrorSummary::InvalidArgument, ErrorLevel::Usage);
 
     memory_info->base_address = vma->second.base;
@@ -692,6 +774,52 @@ static ResultCode CreateMemoryBlock(Handle* out_handle, u32 addr, u32 size, u32
     return RESULT_SUCCESS;
 }
 
+static ResultCode GetProcessInfo(s64* out, Handle process_handle, u32 type) {
+    LOG_TRACE(Kernel_SVC, "called process=0x%08X type=%u", process_handle, type);
+
+    using Kernel::Process;
+    Kernel::SharedPtr<Process> process = Kernel::g_handle_table.Get<Process>(process_handle);
+    if (process == nullptr)
+        return ERR_INVALID_HANDLE;
+
+    switch (type) {
+    case 0:
+    case 2:
+        // TODO(yuriks): Type 0 returns a slightly higher number than type 2, but I'm not sure
+        // what's the difference between them.
+        *out = process->heap_used + process->linear_heap_used + process->misc_memory_used;
+        break;
+    case 1:
+    case 3:
+    case 4:
+    case 5:
+    case 6:
+    case 7:
+    case 8:
+        // These are valid, but not implemented yet
+        LOG_ERROR(Kernel_SVC, "unimplemented GetProcessInfo type=%u", type);
+        break;
+    case 20:
+        *out = Memory::FCRAM_PADDR - process->GetLinearHeapBase();
+        break;
+    default:
+        LOG_ERROR(Kernel_SVC, "unknown GetProcessInfo type=%u", type);
+
+        if (type >= 21 && type <= 23) {
+            return ResultCode( // 0xE0E01BF4
+                    ErrorDescription::NotImplemented, ErrorModule::OS,
+                    ErrorSummary::InvalidArgument, ErrorLevel::Usage);
+        } else {
+            return ResultCode( // 0xD8E007ED
+                    ErrorDescription::InvalidEnumValue, ErrorModule::Kernel,
+                    ErrorSummary::InvalidArgument, ErrorLevel::Permanent);
+        }
+        break;
+    }
+
+    return RESULT_SUCCESS;
+}
+
 namespace {
     struct FunctionDef {
         using Func = void();
@@ -746,7 +874,7 @@ static const FunctionDef SVC_Table[] = {
     {0x28, HLE::Wrap<GetSystemTick>,        "GetSystemTick"},
     {0x29, nullptr,                         "GetHandleInfo"},
     {0x2A, nullptr,                         "GetSystemInfo"},
-    {0x2B, nullptr,                         "GetProcessInfo"},
+    {0x2B, HLE::Wrap<GetProcessInfo>,       "GetProcessInfo"},
     {0x2C, nullptr,                         "GetThreadInfo"},
     {0x2D, HLE::Wrap<ConnectToPort>,        "ConnectToPort"},
     {0x2E, nullptr,                         "SendSyncRequest1"},
diff --git a/src/core/hw/gpu.cpp b/src/core/hw/gpu.cpp
index 3ccbc03b2..68ae38289 100644
--- a/src/core/hw/gpu.cpp
+++ b/src/core/hw/gpu.cpp
@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <cstring>
+#include <numeric>
 #include <type_traits>
 
 #include "common/color.h"
@@ -158,14 +159,59 @@ inline void Write(u32 addr, const T data) {
             u8* src_pointer = Memory::GetPhysicalPointer(config.GetPhysicalInputAddress());
             u8* dst_pointer = Memory::GetPhysicalPointer(config.GetPhysicalOutputAddress());
 
+            if (config.is_texture_copy) {
+                u32 input_width = config.texture_copy.input_width * 16;
+                u32 input_gap = config.texture_copy.input_gap * 16;
+                u32 output_width = config.texture_copy.output_width * 16;
+                u32 output_gap = config.texture_copy.output_gap * 16;
+
+                size_t contiguous_input_size = config.texture_copy.size / input_width * (input_width + input_gap);
+                VideoCore::g_renderer->hw_rasterizer->NotifyPreRead(config.GetPhysicalInputAddress(), contiguous_input_size);
+
+                u32 remaining_size = config.texture_copy.size;
+                u32 remaining_input = input_width;
+                u32 remaining_output = output_width;
+                while (remaining_size > 0) {
+                    u32 copy_size = std::min({ remaining_input, remaining_output, remaining_size });
+
+                    std::memcpy(dst_pointer, src_pointer, copy_size);
+                    src_pointer += copy_size;
+                    dst_pointer += copy_size;
+
+                    remaining_input -= copy_size;
+                    remaining_output -= copy_size;
+                    remaining_size -= copy_size;
+
+                    if (remaining_input == 0) {
+                        remaining_input = input_width;
+                        src_pointer += input_gap;
+                    }
+                    if (remaining_output == 0) {
+                        remaining_output = output_width;
+                        dst_pointer += output_gap;
+                    }
+                }
+
+                LOG_TRACE(HW_GPU, "TextureCopy: 0x%X bytes from 0x%08X(%u+%u)-> 0x%08X(%u+%u), flags 0x%08X",
+                    config.texture_copy.size,
+                    config.GetPhysicalInputAddress(), input_width, input_gap,
+                    config.GetPhysicalOutputAddress(), output_width, output_gap,
+                    config.flags);
+
+                size_t contiguous_output_size = config.texture_copy.size / output_width * (output_width + output_gap);
+                VideoCore::g_renderer->hw_rasterizer->NotifyFlush(config.GetPhysicalOutputAddress(), contiguous_output_size);
+
+                GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PPF);
+                break;
+            }
+
             if (config.scaling > config.ScaleXY) {
                 LOG_CRITICAL(HW_GPU, "Unimplemented display transfer scaling mode %u", config.scaling.Value());
                 UNIMPLEMENTED();
                 break;
             }
 
-            if (config.output_tiled &&
-                    (config.scaling == config.ScaleXY || config.scaling == config.ScaleX)) {
+            if (config.input_linear && config.scaling != config.NoScale) {
                 LOG_CRITICAL(HW_GPU, "Scaling is only implemented on tiled input");
                 UNIMPLEMENTED();
                 break;
@@ -182,23 +228,6 @@ inline void Write(u32 addr, const T data) {
 
             VideoCore::g_renderer->hw_rasterizer->NotifyPreRead(config.GetPhysicalInputAddress(), input_size);
 
-            if (config.raw_copy) {
-                // Raw copies do not perform color conversion nor tiled->linear / linear->tiled conversions
-                // TODO(Subv): Verify if raw copies perform scaling
-                memcpy(dst_pointer, src_pointer, output_size);
-
-                LOG_TRACE(HW_GPU, "DisplayTriggerTransfer: 0x%08x bytes from 0x%08x(%ux%u)-> 0x%08x(%ux%u), output format: %x, flags 0x%08X, Raw copy",
-                    output_size,
-                    config.GetPhysicalInputAddress(), config.input_width.Value(), config.input_height.Value(),
-                    config.GetPhysicalOutputAddress(), config.output_width.Value(), config.output_height.Value(),
-                    config.output_format.Value(), config.flags);
-
-                GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PPF);
-
-                VideoCore::g_renderer->hw_rasterizer->NotifyFlush(config.GetPhysicalOutputAddress(), output_size);
-                break;
-            }
-
             for (u32 y = 0; y < output_height; ++y) {
                 for (u32 x = 0; x < output_width; ++x) {
                     Math::Vec4<u8> src_color;
@@ -220,7 +249,7 @@ inline void Write(u32 addr, const T data) {
                     u32 src_offset;
                     u32 dst_offset;
 
-                    if (config.output_tiled) {
+                    if (config.input_linear) {
                         if (!config.dont_swizzle) {
                             // Interpret the input as linear and the output as tiled
                             u32 coarse_y = y & ~7;
diff --git a/src/core/hw/gpu.h b/src/core/hw/gpu.h
index daad506fe..2e3a9f779 100644
--- a/src/core/hw/gpu.h
+++ b/src/core/hw/gpu.h
@@ -201,12 +201,14 @@ struct Regs {
             u32 flags;
 
             BitField< 0, 1, u32> flip_vertically;  // flips input data vertically
-            BitField< 1, 1, u32> output_tiled;     // Converts from linear to tiled format
-            BitField< 3, 1, u32> raw_copy;         // Copies the data without performing any processing
+            BitField< 1, 1, u32> input_linear;     // Converts from linear to tiled format
+            BitField< 2, 1, u32> crop_input_lines;
+            BitField< 3, 1, u32> is_texture_copy;  // Copies the data without performing any processing and respecting texture copy fields
             BitField< 5, 1, u32> dont_swizzle;
             BitField< 8, 3, PixelFormat> input_format;
             BitField<12, 3, PixelFormat> output_format;
-
+            /// Uses some kind of 32x32 block swizzling mode, instead of the usual 8x8 one.
+            BitField<16, 1, u32> block_32; // TODO(yuriks): unimplemented
             BitField<24, 2, ScalingMode> scaling; // Determines the scaling mode of the transfer
         };
 
@@ -214,10 +216,30 @@ struct Regs {
 
         // it seems that writing to this field triggers the display transfer
         u32 trigger;
+
+        INSERT_PADDING_WORDS(0x1);
+
+        struct {
+            u32 size;
+
+            union {
+                u32 input_size;
+
+                BitField< 0, 16, u32> input_width;
+                BitField<16, 16, u32> input_gap;
+            };
+
+            union {
+                u32 output_size;
+
+                BitField< 0, 16, u32> output_width;
+                BitField<16, 16, u32> output_gap;
+            };
+        } texture_copy;
     } display_transfer_config;
-    ASSERT_MEMBER_SIZE(display_transfer_config, 0x1c);
+    ASSERT_MEMBER_SIZE(display_transfer_config, 0x2c);
 
-    INSERT_PADDING_WORDS(0x331);
+    INSERT_PADDING_WORDS(0x32D);
 
     struct {
         // command list size (in bytes)
diff --git a/src/core/mem_map.cpp b/src/core/mem_map.cpp
deleted file mode 100644
index cbe993fbe..000000000
--- a/src/core/mem_map.cpp
+++ /dev/null
@@ -1,163 +0,0 @@
-// Copyright 2014 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <map>
-#include <memory>
-#include <utility>
-#include <vector>
-
-#include "common/common_types.h"
-#include "common/logging/log.h"
-
-#include "core/hle/config_mem.h"
-#include "core/hle/kernel/vm_manager.h"
-#include "core/hle/result.h"
-#include "core/hle/shared_page.h"
-#include "core/mem_map.h"
-#include "core/memory.h"
-#include "core/memory_setup.h"
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-
-namespace Memory {
-
-namespace {
-
-struct MemoryArea {
-    u32 base;
-    u32 size;
-    const char* name;
-};
-
-// We don't declare the IO regions in here since its handled by other means.
-static MemoryArea memory_areas[] = {
-    {HEAP_VADDR,          HEAP_SIZE,              "Heap"},          // Application heap (main memory)
-    {SHARED_MEMORY_VADDR, SHARED_MEMORY_SIZE,     "Shared Memory"}, // Shared memory
-    {LINEAR_HEAP_VADDR,   LINEAR_HEAP_SIZE,       "Linear Heap"},   // Linear heap (main memory)
-    {VRAM_VADDR,          VRAM_SIZE,              "VRAM"},          // Video memory (VRAM)
-    {DSP_RAM_VADDR,       DSP_RAM_SIZE,           "DSP RAM"},       // DSP memory
-    {TLS_AREA_VADDR,      TLS_AREA_SIZE,          "TLS Area"},      // TLS memory
-};
-
-/// Represents a block of memory mapped by ControlMemory/MapMemoryBlock
-struct MemoryBlock {
-    MemoryBlock() : handle(0), base_address(0), address(0), size(0), operation(0), permissions(0) {
-    }
-    u32 handle;
-    u32 base_address;
-    u32 address;
-    u32 size;
-    u32 operation;
-    u32 permissions;
-
-    const u32 GetVirtualAddress() const{
-        return base_address + address;
-    }
-};
-
-static std::map<u32, MemoryBlock> heap_map;
-static std::map<u32, MemoryBlock> heap_linear_map;
-
-}
-
-u32 MapBlock_Heap(u32 size, u32 operation, u32 permissions) {
-    MemoryBlock block;
-
-    block.base_address  = HEAP_VADDR;
-    block.size          = size;
-    block.operation     = operation;
-    block.permissions   = permissions;
-
-    if (heap_map.size() > 0) {
-        const MemoryBlock last_block = heap_map.rbegin()->second;
-        block.address = last_block.address + last_block.size;
-    }
-    heap_map[block.GetVirtualAddress()] = block;
-
-    return block.GetVirtualAddress();
-}
-
-u32 MapBlock_HeapLinear(u32 size, u32 operation, u32 permissions) {
-    MemoryBlock block;
-
-    block.base_address  = LINEAR_HEAP_VADDR;
-    block.size          = size;
-    block.operation     = operation;
-    block.permissions   = permissions;
-
-    if (heap_linear_map.size() > 0) {
-        const MemoryBlock last_block = heap_linear_map.rbegin()->second;
-        block.address = last_block.address + last_block.size;
-    }
-    heap_linear_map[block.GetVirtualAddress()] = block;
-
-    return block.GetVirtualAddress();
-}
-
-PAddr VirtualToPhysicalAddress(const VAddr addr) {
-    if (addr == 0) {
-        return 0;
-    } else if (addr >= VRAM_VADDR && addr < VRAM_VADDR_END) {
-        return addr - VRAM_VADDR + VRAM_PADDR;
-    } else if (addr >= LINEAR_HEAP_VADDR && addr < LINEAR_HEAP_VADDR_END) {
-        return addr - LINEAR_HEAP_VADDR + FCRAM_PADDR;
-    } else if (addr >= DSP_RAM_VADDR && addr < DSP_RAM_VADDR_END) {
-        return addr - DSP_RAM_VADDR + DSP_RAM_PADDR;
-    } else if (addr >= IO_AREA_VADDR && addr < IO_AREA_VADDR_END) {
-        return addr - IO_AREA_VADDR + IO_AREA_PADDR;
-    }
-
-    LOG_ERROR(HW_Memory, "Unknown virtual address @ 0x%08x", addr);
-    // To help with debugging, set bit on address so that it's obviously invalid.
-    return addr | 0x80000000;
-}
-
-VAddr PhysicalToVirtualAddress(const PAddr addr) {
-    if (addr == 0) {
-        return 0;
-    } else if (addr >= VRAM_PADDR && addr < VRAM_PADDR_END) {
-        return addr - VRAM_PADDR + VRAM_VADDR;
-    } else if (addr >= FCRAM_PADDR && addr < FCRAM_PADDR_END) {
-        return addr - FCRAM_PADDR + LINEAR_HEAP_VADDR;
-    } else if (addr >= DSP_RAM_PADDR && addr < DSP_RAM_PADDR_END) {
-        return addr - DSP_RAM_PADDR + DSP_RAM_VADDR;
-    } else if (addr >= IO_AREA_PADDR && addr < IO_AREA_PADDR_END) {
-        return addr - IO_AREA_PADDR + IO_AREA_VADDR;
-    }
-
-    LOG_ERROR(HW_Memory, "Unknown physical address @ 0x%08x", addr);
-    // To help with debugging, set bit on address so that it's obviously invalid.
-    return addr | 0x80000000;
-}
-
-void Init() {
-    InitMemoryMap();
-    LOG_DEBUG(HW_Memory, "initialized OK");
-}
-
-void InitLegacyAddressSpace(Kernel::VMManager& address_space) {
-    using namespace Kernel;
-
-    for (MemoryArea& area : memory_areas) {
-        auto block = std::make_shared<std::vector<u8>>(area.size);
-        address_space.MapMemoryBlock(area.base, std::move(block), 0, area.size, MemoryState::Private).Unwrap();
-    }
-
-    auto cfg_mem_vma = address_space.MapBackingMemory(CONFIG_MEMORY_VADDR,
-            (u8*)&ConfigMem::config_mem, CONFIG_MEMORY_SIZE, MemoryState::Shared).MoveFrom();
-    address_space.Reprotect(cfg_mem_vma, VMAPermission::Read);
-
-    auto shared_page_vma = address_space.MapBackingMemory(SHARED_PAGE_VADDR,
-            (u8*)&SharedPage::shared_page, SHARED_PAGE_SIZE, MemoryState::Shared).MoveFrom();
-    address_space.Reprotect(shared_page_vma, VMAPermission::Read);
-}
-
-void Shutdown() {
-    heap_map.clear();
-    heap_linear_map.clear();
-
-    LOG_DEBUG(HW_Memory, "shutdown OK");
-}
-
-} // namespace
diff --git a/src/core/mem_map.h b/src/core/mem_map.h
deleted file mode 100644
index 229ef82c5..000000000
--- a/src/core/mem_map.h
+++ /dev/null
@@ -1,46 +0,0 @@
-// Copyright 2014 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 Kernel {
-class VMManager;
-}
-
-namespace Memory {
-
-void Init();
-void InitLegacyAddressSpace(Kernel::VMManager& address_space);
-void Shutdown();
-
-/**
- * Maps a block of memory on the heap
- * @param size Size of block in bytes
- * @param operation Memory map operation type
- * @param permissions Memory allocation permissions
- */
-u32 MapBlock_Heap(u32 size, u32 operation, u32 permissions);
-
-/**
- * Maps a block of memory on the GSP heap
- * @param size Size of block in bytes
- * @param operation Memory map operation type
- * @param permissions Control memory permissions
- */
-u32 MapBlock_HeapLinear(u32 size, u32 operation, u32 permissions);
-
-/**
- * Converts a virtual address inside a region with 1:1 mapping to physical memory to a physical
- * address. This should be used by services to translate addresses for use by the hardware.
- */
-PAddr VirtualToPhysicalAddress(VAddr addr);
-
-/**
- * Undoes a mapping performed by VirtualToPhysicalAddress().
- */
-VAddr PhysicalToVirtualAddress(PAddr addr);
-
-} // namespace
diff --git a/src/core/memory.cpp b/src/core/memory.cpp
index 1f66bb27d..cde390b8a 100644
--- a/src/core/memory.cpp
+++ b/src/core/memory.cpp
@@ -9,7 +9,7 @@
 #include "common/logging/log.h"
 #include "common/swap.h"
 
-#include "core/mem_map.h"
+#include "core/hle/kernel/process.h"
 #include "core/memory.h"
 #include "core/memory_setup.h"
 
@@ -198,4 +198,42 @@ void WriteBlock(const VAddr addr, const u8* data, const size_t size) {
         Write8(addr + offset, data[offset]);
 }
 
+PAddr VirtualToPhysicalAddress(const VAddr addr) {
+    if (addr == 0) {
+        return 0;
+    } else if (addr >= VRAM_VADDR && addr < VRAM_VADDR_END) {
+        return addr - VRAM_VADDR + VRAM_PADDR;
+    } else if (addr >= LINEAR_HEAP_VADDR && addr < LINEAR_HEAP_VADDR_END) {
+        return addr - LINEAR_HEAP_VADDR + FCRAM_PADDR;
+    } else if (addr >= DSP_RAM_VADDR && addr < DSP_RAM_VADDR_END) {
+        return addr - DSP_RAM_VADDR + DSP_RAM_PADDR;
+    } else if (addr >= IO_AREA_VADDR && addr < IO_AREA_VADDR_END) {
+        return addr - IO_AREA_VADDR + IO_AREA_PADDR;
+    } else if (addr >= NEW_LINEAR_HEAP_VADDR && addr < NEW_LINEAR_HEAP_VADDR_END) {
+        return addr - NEW_LINEAR_HEAP_VADDR + FCRAM_PADDR;
+    }
+
+    LOG_ERROR(HW_Memory, "Unknown virtual address @ 0x%08X", addr);
+    // To help with debugging, set bit on address so that it's obviously invalid.
+    return addr | 0x80000000;
+}
+
+VAddr PhysicalToVirtualAddress(const PAddr addr) {
+    if (addr == 0) {
+        return 0;
+    } else if (addr >= VRAM_PADDR && addr < VRAM_PADDR_END) {
+        return addr - VRAM_PADDR + VRAM_VADDR;
+    } else if (addr >= FCRAM_PADDR && addr < FCRAM_PADDR_END) {
+        return addr - FCRAM_PADDR + Kernel::g_current_process->GetLinearHeapBase();
+    } else if (addr >= DSP_RAM_PADDR && addr < DSP_RAM_PADDR_END) {
+        return addr - DSP_RAM_PADDR + DSP_RAM_VADDR;
+    } else if (addr >= IO_AREA_PADDR && addr < IO_AREA_PADDR_END) {
+        return addr - IO_AREA_PADDR + IO_AREA_VADDR;
+    }
+
+    LOG_ERROR(HW_Memory, "Unknown physical address @ 0x%08X", addr);
+    // To help with debugging, set bit on address so that it's obviously invalid.
+    return addr | 0x80000000;
+}
+
 } // namespace
diff --git a/src/core/memory.h b/src/core/memory.h
index 418609de0..5af72b7a7 100644
--- a/src/core/memory.h
+++ b/src/core/memory.h
@@ -15,6 +15,8 @@ namespace Memory {
  * be mapped.
  */
 const u32 PAGE_SIZE = 0x1000;
+const u32 PAGE_MASK = PAGE_SIZE - 1;
+const int PAGE_BITS = 12;
 
 /// Physical memory regions as seen from the ARM11
 enum : PAddr {
@@ -103,8 +105,15 @@ enum : VAddr {
     // hardcoded value.
     /// Area where TLS (Thread-Local Storage) buffers are allocated.
     TLS_AREA_VADDR     = 0x1FF82000,
-    TLS_AREA_SIZE      = 0x00030000, // Each TLS buffer is 0x200 bytes, allows for 300 threads
+    TLS_ENTRY_SIZE     = 0x200,
+    TLS_AREA_SIZE      = 300 * TLS_ENTRY_SIZE + 0x800, // Space for up to 300 threads + round to page size
     TLS_AREA_VADDR_END = TLS_AREA_VADDR + TLS_AREA_SIZE,
+
+
+    /// Equivalent to LINEAR_HEAP_VADDR, but expanded to cover the extra memory in the New 3DS.
+    NEW_LINEAR_HEAP_VADDR     = 0x30000000,
+    NEW_LINEAR_HEAP_SIZE      = 0x10000000,
+    NEW_LINEAR_HEAP_VADDR_END = NEW_LINEAR_HEAP_VADDR + NEW_LINEAR_HEAP_SIZE,
 };
 
 u8 Read8(VAddr addr);
@@ -122,6 +131,17 @@ void WriteBlock(VAddr addr, const u8* data, size_t size);
 u8* GetPointer(VAddr virtual_address);
 
 /**
+* Converts a virtual address inside a region with 1:1 mapping to physical memory to a physical
+* address. This should be used by services to translate addresses for use by the hardware.
+*/
+PAddr VirtualToPhysicalAddress(VAddr addr);
+
+/**
+* Undoes a mapping performed by VirtualToPhysicalAddress().
+*/
+VAddr PhysicalToVirtualAddress(PAddr addr);
+
+/**
  * Gets a pointer to the memory region beginning at the specified physical address.
  *
  * @note This is currently implemented using PhysicalToVirtualAddress().
diff --git a/src/core/memory_setup.h b/src/core/memory_setup.h
index 361bfc816..84ff30120 100644
--- a/src/core/memory_setup.h
+++ b/src/core/memory_setup.h
@@ -10,9 +10,6 @@
 
 namespace Memory {
 
-const u32 PAGE_MASK = PAGE_SIZE - 1;
-const int PAGE_BITS = 12;
-
 void InitMemoryMap();
 
 /**
diff --git a/src/core/system.cpp b/src/core/system.cpp
index 561ff82f0..3cd84bf5e 100644
--- a/src/core/system.cpp
+++ b/src/core/system.cpp
@@ -4,11 +4,11 @@
 
 #include "core/core.h"
 #include "core/core_timing.h"
-#include "core/mem_map.h"
 #include "core/system.h"
 #include "core/hw/hw.h"
 #include "core/hle/hle.h"
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/memory.h"
 
 #include "video_core/video_core.h"
 
@@ -29,7 +29,6 @@ void Shutdown() {
     HLE::Shutdown();
     Kernel::Shutdown();
     HW::Shutdown();
-    Memory::Shutdown();
     CoreTiming::Shutdown();
     Core::Shutdown();
 }