kernel: make current thread pointer thread local

KPageTable: Remove extraneous assert

.ci/scripts/windows/docker.sh

@@ -1,12 +1,27 @@
 #!/bin/bash -ex
 
+set -e
+
 cd /yuzu
 
 ccache -s
 
 mkdir build || true && cd build
-cmake .. -G Ninja -DDISPLAY_VERSION=$1 -DCMAKE_TOOLCHAIN_FILE="$(pwd)/../CMakeModules/MinGWCross.cmake" -DUSE_CCACHE=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DCMAKE_BUILD_TYPE=Release -DENABLE_QT_TRANSLATION=ON
-ninja
+LDFLAGS="-fuse-ld=lld"
+# -femulated-tls required due to an incompatibility between GCC and Clang
+# TODO(lat9nq): If this is widespread, we probably need to add this to CMakeLists where appropriate
+cmake .. \
+    -DCMAKE_BUILD_TYPE=Release \
+    -DCMAKE_CXX_FLAGS="-femulated-tls" \
+    -DCMAKE_TOOLCHAIN_FILE="$(pwd)/../CMakeModules/MinGWClangCross.cmake" \
+    -DDISPLAY_VERSION=$1 \
+    -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON \
+    -DENABLE_QT_TRANSLATION=ON \
+    -DUSE_CCACHE=ON \
+    -DYUZU_USE_BUNDLED_SDL2=OFF \
+    -DYUZU_USE_EXTERNAL_SDL2=OFF \
+    -GNinja
+ninja yuzu yuzu-cmd
 
 ccache -s
 

CMakeModules/MinGWClangCross.cmake

@@ -0,0 +1,55 @@
+set(MINGW_PREFIX                /usr/x86_64-w64-mingw32/)
+set(CMAKE_SYSTEM_NAME           Windows)
+set(CMAKE_SYSTEM_PROCESSOR      x86_64)
+
+set(CMAKE_FIND_ROOT_PATH        ${MINGW_PREFIX})
+set(SDL2_PATH                   ${MINGW_PREFIX})
+set(MINGW_TOOL_PREFIX           ${CMAKE_SYSTEM_PROCESSOR}-w64-mingw32-)
+
+# Specify the cross compiler
+set(CMAKE_C_COMPILER            ${MINGW_TOOL_PREFIX}clang)
+set(CMAKE_CXX_COMPILER          ${MINGW_TOOL_PREFIX}clang++)
+set(CMAKE_RC_COMPILER           ${MINGW_TOOL_PREFIX}windres)
+set(CMAKE_C_COMPILER_AR         ${MINGW_TOOL_PREFIX}ar)
+set(CMAKE_CXX_COMPILER_AR       ${MINGW_TOOL_PREFIX}ar)
+set(CMAKE_C_COMPILER_RANLIB     ${MINGW_TOOL_PREFIX}ranlib)
+set(CMAKE_CXX_COMPILER_RANLIB   ${MINGW_TOOL_PREFIX}ranlib)
+
+# Mingw tools
+set(STRIP                       ${MINGW_TOOL_PREFIX}strip)
+set(WINDRES                     ${MINGW_TOOL_PREFIX}windres)
+set(ENV{PKG_CONFIG}             ${MINGW_TOOL_PREFIX}pkg-config)
+
+# ccache wrapper
+option(USE_CCACHE "Use ccache for compilation" OFF)
+if(USE_CCACHE)
+    find_program(CCACHE ccache)
+    if(CCACHE)
+        message(STATUS "Using ccache found in PATH")
+        set_property(GLOBAL PROPERTY RULE_LAUNCH_COMPILE ${CCACHE})
+        set_property(GLOBAL PROPERTY RULE_LAUNCH_LINK ${CCACHE})
+    else(CCACHE)
+        message(WARNING "USE_CCACHE enabled, but no ccache found")
+    endif(CCACHE)
+endif(USE_CCACHE)
+
+# Search for programs in the build host directories
+set(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)
+
+
+# Echo modified cmake vars to screen for debugging purposes
+if(NOT DEFINED ENV{MINGW_DEBUG_INFO})
+        message("")
+        message("Custom cmake vars: (blank = system default)")
+        message("-----------------------------------------")
+        message("* CMAKE_C_COMPILER                     : ${CMAKE_C_COMPILER}")
+        message("* CMAKE_CXX_COMPILER                   : ${CMAKE_CXX_COMPILER}")
+        message("* CMAKE_RC_COMPILER                    : ${CMAKE_RC_COMPILER}")
+        message("* WINDRES                              : ${WINDRES}")
+        message("* ENV{PKG_CONFIG}                      : $ENV{PKG_CONFIG}")
+        message("* STRIP                                : ${STRIP}")
+        message("* USE_CCACHE                           : ${USE_CCACHE}")
+        message("")
+        # So that the debug info only appears once
+        set(ENV{MINGW_DEBUG_INFO} SHOWN)
+endif()

dist/qt_themes/default/style.qss

@@ -58,6 +58,19 @@ QPushButton#GPUStatusBarButton:!checked {
     color: #109010;
 }
 
+QPushButton#DockingStatusBarButton {
+    min-width: 0px;
+    color: #000000;
+    border: 1px solid transparent;
+    background-color: transparent;
+    padding: 0px 3px 0px 3px;
+    text-align: center;
+}
+
+QPushButton#DockingStatusBarButton:hover {
+    border: 1px solid #76797C;
+}
+
 QPushButton#buttonRefreshDevices {
     min-width: 21px;
     min-height: 21px;

dist/qt_themes/qdarkstyle/style.qss

@@ -1304,6 +1304,19 @@ QPushButton#GPUStatusBarButton:!checked {
     color: #40dd40;
 }
 
+QPushButton#DockingStatusBarButton {
+    min-width: 0px;
+    color: #ffffff;
+    border: 1px solid transparent;
+    background-color: transparent;
+    padding: 0px 3px 0px 3px;
+    text-align: center;
+}
+
+QPushButton#DockingStatusBarButton:hover {
+    border: 1px solid #76797C;
+}
+
 QPushButton#buttonRefreshDevices {
     min-width: 23px;
     min-height: 23px;

dist/qt_themes/qdarkstyle_midnight_blue/style.qss

@@ -2207,6 +2207,19 @@ QPushButton#GPUStatusBarButton:!checked {
   color: #40dd40;
 }
 
+QPushButton#DockingStatusBarButton {
+    min-width: 0px;
+    color: #ffffff;
+    border: 1px solid transparent;
+    background-color: transparent;
+    padding: 0px 3px 0px 3px;
+    text-align: center;
+}
+
+QPushButton#DockingStatusBarButton:hover {
+    border: 1px solid #76797C;
+}
+
 QPushButton#buttonRefreshDevices {
   min-width: 19px;
   min-height: 19px;

externals/CMakeLists.txt

@@ -40,6 +40,11 @@ target_include_directories(mbedtls PUBLIC ./mbedtls/include)
 add_library(microprofile INTERFACE)
 target_include_directories(microprofile INTERFACE ./microprofile)
 
+# GCC bugs
+if (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL "12" AND CMAKE_CXX_COMPILER_ID STREQUAL "GNU" AND MINGW)
+    target_compile_options(microprofile INTERFACE "-Wno-array-bounds")
+endif()
+
 # libusb
 if (NOT LIBUSB_FOUND OR YUZU_USE_BUNDLED_LIBUSB)
     add_subdirectory(libusb)

externals/microprofile/microprofile.h

@@ -1246,7 +1246,7 @@ struct MicroProfileScopeLock
 {
     bool bUseLock;
     std::recursive_mutex& m;
-    MicroProfileScopeLock(std::recursive_mutex& m) : bUseLock(g_bUseLock), m(m)
+    MicroProfileScopeLock(std::recursive_mutex& m_) : bUseLock(g_bUseLock), m(m_)
     {
         if(bUseLock)
             m.lock();

externals/microprofile/microprofileui.h

@@ -213,8 +213,8 @@ struct MicroProfileCustom
 
 struct SOptionDesc
 {
-    SOptionDesc(){}
-    SOptionDesc(uint8_t nSubType, uint8_t nIndex, const char* fmt, ...):nSubType(nSubType), nIndex(nIndex)
+    SOptionDesc()=default;
+    SOptionDesc(uint8_t nSubType_, uint8_t nIndex_, const char* fmt, ...):nSubType(nSubType_), nIndex(nIndex_)
     {
         va_list args;
         va_start (args, fmt);
@@ -573,10 +573,10 @@ inline void MicroProfileToolTipMeta(MicroProfileStringArray* pToolTip)
                     }
                     else
                     {
-                        for(int i = 0; i < MICROPROFILE_META_MAX; ++i)
+                        for(int k = 0; k < MICROPROFILE_META_MAX; ++k)
                         {
-                            nMetaSumInclusive[i] += nMetaSum[i];
-                            nMetaSum[i] = 0;
+                            nMetaSumInclusive[k] += nMetaSum[k];
+                            nMetaSum[k] = 0;
                         }
                     }
                     break;
@@ -708,10 +708,10 @@ inline void MicroProfileDrawFloatTooltip(uint32_t nX, uint32_t nY, uint32_t nTok
 
     if(UI.nMouseLeftMod)
     {
-        int nIndex = (g_MicroProfileUI.LockedToolTipFront + MICROPROFILE_TOOLTIP_MAX_LOCKED - 1) % MICROPROFILE_TOOLTIP_MAX_LOCKED;
-        g_MicroProfileUI.nLockedToolTipColor[nIndex] = S.TimerInfo[nTimerId].nColor;
-        MicroProfileStringArrayCopy(&g_MicroProfileUI.LockedToolTips[nIndex], &ToolTip);
-        g_MicroProfileUI.LockedToolTipFront = nIndex;
+        int nToolTipIndex = (g_MicroProfileUI.LockedToolTipFront + MICROPROFILE_TOOLTIP_MAX_LOCKED - 1) % MICROPROFILE_TOOLTIP_MAX_LOCKED;
+        g_MicroProfileUI.nLockedToolTipColor[nToolTipIndex] = S.TimerInfo[nTimerId].nColor;
+        MicroProfileStringArrayCopy(&g_MicroProfileUI.LockedToolTips[nToolTipIndex], &ToolTip);
+        g_MicroProfileUI.LockedToolTipFront = nToolTipIndex;
 
     }
 }
@@ -917,9 +917,8 @@ inline void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int
         float fStart = floor(fMsBase*fRcpStep) * fStep;
         for(float f = fStart; f < fMsEnd; )
         {
-            float fStart = f;
             float fNext = f + fStep;
-            MicroProfileDrawBox(((fStart-fMsBase) * fMsToScreen), nBaseY, (fNext-fMsBase) * fMsToScreen+1, nBaseY + nHeight, UI.nOpacityBackground | g_nMicroProfileBackColors[nColorIndex++ & 1]);
+            MicroProfileDrawBox(((f-fMsBase) * fMsToScreen), nBaseY, (fNext-fMsBase) * fMsToScreen+1, nBaseY + nHeight, UI.nOpacityBackground | g_nMicroProfileBackColors[nColorIndex++ & 1]);
             f = fNext;
         }
     }
@@ -1116,9 +1115,9 @@ inline void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int
 
                         nMaxStackDepth = MicroProfileMax(nMaxStackDepth, nStackPos);
                         float fMsStart = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickStart);
-                        float fMsEnd = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickEnd);
+                        float fMsEnd2 = fToMs * MicroProfileLogTickDifference(nBaseTicks, nTickEnd);
                         float fXStart = fMsStart * fMsToScreen;
-                        float fXEnd = fMsEnd * fMsToScreen;
+                        float fXEnd = fMsEnd2 * fMsToScreen;
                         float fYStart = (float)(nY + nStackPos * nYDelta);
                         float fYEnd = fYStart + (MICROPROFILE_DETAILED_BAR_HEIGHT);
                         float fXDist = MicroProfileMax(fXStart - fMouseX, fMouseX - fXEnd);
@@ -1269,22 +1268,22 @@ inline void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int
         if(UI.nRangeBegin != UI.nRangeEnd)
         {
             float fMsStart = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, UI.nRangeBegin);
-            float fMsEnd = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, UI.nRangeEnd);
+            float fMsEnd3 = fToMsCpu * MicroProfileLogTickDifference(nBaseTicksCpu, UI.nRangeEnd);
             float fXStart = fMsStart * fMsToScreen;
-            float fXEnd = fMsEnd * fMsToScreen;
+            float fXEnd = fMsEnd3 * fMsToScreen;
             MicroProfileDrawBox(fXStart, nBaseY, fXEnd, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT, MicroProfileBoxTypeFlat);
             MicroProfileDrawLineVertical(fXStart, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT | 0x44000000);
             MicroProfileDrawLineVertical(fXEnd, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT | 0x44000000);
 
             fMsStart += fDetailedOffset;
-            fMsEnd += fDetailedOffset;
+            fMsEnd3 += fDetailedOffset;
             char sBuffer[32];
             uint32_t nLenStart = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsStart);
             float fStartTextWidth = (float)((1+MICROPROFILE_TEXT_WIDTH) * nLenStart);
             float fStartTextX = fXStart - fStartTextWidth - 2;
             MicroProfileDrawBox(fStartTextX, nBaseY, fStartTextX + fStartTextWidth + 2, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);
             MicroProfileDrawText(fStartTextX+1, nBaseY, UINT32_MAX, sBuffer, nLenStart);
-            uint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsEnd);
+            uint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsEnd3);
             MicroProfileDrawBox(fXEnd+1, nBaseY, fXEnd+1+(1+MICROPROFILE_TEXT_WIDTH) * nLenEnd + 3, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);
             MicroProfileDrawText(fXEnd+2, nBaseY+1, UINT32_MAX, sBuffer, nLenEnd);
 
@@ -1297,9 +1296,9 @@ inline void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int
         if(UI.nRangeBeginGpu != UI.nRangeEndGpu)
         {
             float fMsStart = fToMsGpu * MicroProfileLogTickDifference(nBaseTicksGpu, UI.nRangeBeginGpu);
-            float fMsEnd = fToMsGpu * MicroProfileLogTickDifference(nBaseTicksGpu, UI.nRangeEndGpu);
+            float fMsEnd4 = fToMsGpu * MicroProfileLogTickDifference(nBaseTicksGpu, UI.nRangeEndGpu);
             float fXStart = fMsStart * fMsToScreen;
-            float fXEnd = fMsEnd * fMsToScreen;
+            float fXEnd = fMsEnd4 * fMsToScreen;
             MicroProfileDrawBox(fXStart, nBaseY, fXEnd, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU, MicroProfileBoxTypeFlat);
             MicroProfileDrawLineVertical(fXStart, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU | 0x44000000);
             MicroProfileDrawLineVertical(fXEnd, nBaseY, nHeight, MICROPROFILE_FRAME_COLOR_HIGHTLIGHT_GPU | 0x44000000);
@@ -1307,14 +1306,14 @@ inline void MicroProfileDrawDetailedBars(uint32_t nWidth, uint32_t nHeight, int
             nBaseY += MICROPROFILE_TEXT_HEIGHT+1;
 
             fMsStart += fDetailedOffset;
-            fMsEnd += fDetailedOffset;
+            fMsEnd4 += fDetailedOffset;
             char sBuffer[32];
             uint32_t nLenStart = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsStart);
             float fStartTextWidth = (float)((1+MICROPROFILE_TEXT_WIDTH) * nLenStart);
             float fStartTextX = fXStart - fStartTextWidth - 2;
             MicroProfileDrawBox(fStartTextX, nBaseY, fStartTextX + fStartTextWidth + 2, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);
             MicroProfileDrawText(fStartTextX+1, nBaseY, UINT32_MAX, sBuffer, nLenStart);
-            uint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsEnd);
+            uint32_t nLenEnd = snprintf(sBuffer, sizeof(sBuffer)-1, "%.2fms", fMsEnd4);
             MicroProfileDrawBox(fXEnd+1, nBaseY, fXEnd+1+(1+MICROPROFILE_TEXT_WIDTH) * nLenEnd + 3, MICROPROFILE_TEXT_HEIGHT + 2 + nBaseY, 0x33000000, MicroProfileBoxTypeFlat);
             MicroProfileDrawText(fXEnd+2, nBaseY+1, UINT32_MAX, sBuffer, nLenEnd);
         }
@@ -1716,8 +1715,8 @@ bool MicroProfileDrawGraph(uint32_t nScreenWidth, uint32_t nScreenHeight)
         uint32_t nTextCount = 0;
         uint32_t nGraphIndex = (S.nGraphPut + MICROPROFILE_GRAPH_HISTORY - int(MICROPROFILE_GRAPH_HISTORY*(1.f - fMouseXPrc))) % MICROPROFILE_GRAPH_HISTORY;
 
-        uint32_t nX = UI.nMouseX;
-        uint32_t nY = UI.nMouseY + 20;
+        uint32_t nMouseX = UI.nMouseX;
+        uint32_t nMouseY = UI.nMouseY + 20;
 
         for(uint32_t i = 0; i < MICROPROFILE_MAX_GRAPHS; ++i)
         {
@@ -1736,7 +1735,7 @@ bool MicroProfileDrawGraph(uint32_t nScreenWidth, uint32_t nScreenHeight)
         }
         if(nTextCount)
         {
-            MicroProfileDrawFloatWindow(nX, nY, Strings.ppStrings, Strings.nNumStrings, 0, pColors);
+            MicroProfileDrawFloatWindow(nMouseX, nMouseY, Strings.ppStrings, Strings.nNumStrings, 0, pColors);
         }
 
         if(UI.nMouseRight)
@@ -2321,8 +2320,8 @@ inline void MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight)
     uint32_t    nMenuX[MICROPROFILE_MENU_MAX] = {0};
     uint32_t nNumMenuItems = 0;
 
-    int nLen = snprintf(buffer, 127, "MicroProfile");
-    MicroProfileDrawText(nX, nY, UINT32_MAX, buffer, nLen);
+    int nMPTextLen = snprintf(buffer, 127, "MicroProfile");
+    MicroProfileDrawText(nX, nY, UINT32_MAX, buffer, nMPTextLen);
     nX += (sizeof("MicroProfile")+2) * (MICROPROFILE_TEXT_WIDTH+1);
     pMenuText[nNumMenuItems++] = "Mode";
     pMenuText[nNumMenuItems++] = "Groups";
@@ -2438,16 +2437,16 @@ inline void MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight)
         int nNumLines = 0;
         bool bSelected = false;
         const char* pString = CB(nNumLines, &bSelected);
-        uint32_t nWidth = 0, nHeight = 0;
+        uint32_t nTextWidth = 0, nTextHeight = 0;
         while(pString)
         {
-            nWidth = MicroProfileMax<int>(nWidth, (int)strlen(pString));
+            nTextWidth = MicroProfileMax<int>(nTextWidth, (int)strlen(pString));
             nNumLines++;
             pString = CB(nNumLines, &bSelected);
         }
-        nWidth = (2+nWidth) * (MICROPROFILE_TEXT_WIDTH+1);
-        nHeight = nNumLines * (MICROPROFILE_TEXT_HEIGHT+1);
-        if(UI.nMouseY <= nY + nHeight+0 && UI.nMouseY >= nY-0 && UI.nMouseX <= nX + nWidth + 0 && UI.nMouseX >= nX - 0)
+        nTextWidth = (2+nTextWidth) * (MICROPROFILE_TEXT_WIDTH+1);
+        nTextHeight = nNumLines * (MICROPROFILE_TEXT_HEIGHT+1);
+        if(UI.nMouseY <= nY + nTextHeight+0 && UI.nMouseY >= nY-0 && UI.nMouseX <= nX + nTextWidth + 0 && UI.nMouseX >= nX - 0)
         {
             UI.nActiveMenu = nMenu;
         }
@@ -2455,21 +2454,21 @@ inline void MicroProfileDrawMenu(uint32_t nWidth, uint32_t nHeight)
         {
             UI.nActiveMenu = UINT32_MAX;
         }
-        MicroProfileDrawBox(nX, nY, nX + nWidth, nY + nHeight, 0xff000000|g_nMicroProfileBackColors[1]);
+        MicroProfileDrawBox(nX, nY, nX + nTextWidth, nY + nTextHeight, 0xff000000|g_nMicroProfileBackColors[1]);
         for(int i = 0; i < nNumLines; ++i)
         {
-            bool bSelected = false;
-            const char* pString = CB(i, &bSelected);
+            bool bSelected2 = false;
+            const char* pString2 = CB(i, &bSelected2);
             if(UI.nMouseY >= nY && UI.nMouseY < nY + MICROPROFILE_TEXT_HEIGHT + 1)
             {
                 if(UI.nMouseLeft || UI.nMouseRight)
                 {
                     CBClick[nMenu](i);
                 }
-                MicroProfileDrawBox(nX, nY, nX + nWidth, nY + MICROPROFILE_TEXT_HEIGHT + 1, 0xff888888);
+                MicroProfileDrawBox(nX, nY, nX + nTextWidth, nY + MICROPROFILE_TEXT_HEIGHT + 1, 0xff888888);
             }
-            int nLen = snprintf(buffer, SBUF_SIZE-1, "%c %s", bSelected ? '*' : ' ' ,pString);
-            MicroProfileDrawText(nX, nY, UINT32_MAX, buffer, nLen);
+            int nTextLen = snprintf(buffer, SBUF_SIZE-1, "%c %s", bSelected2 ? '*' : ' ' ,pString2);
+            MicroProfileDrawText(nX, nY, UINT32_MAX, buffer, nTextLen);
             nY += MICROPROFILE_TEXT_HEIGHT+1;
         }
     }
@@ -2605,7 +2604,7 @@ inline void MicroProfileDrawCustom(uint32_t nWidth, uint32_t nHeight)
             for(uint32_t i = 0; i < nCount; ++i)
             {
                 nOffsetY += (1+MICROPROFILE_TEXT_HEIGHT);
-                uint32_t nWidth = MicroProfileMin(nMaxWidth, (uint32_t)(nMaxWidth * pMs[i] * fRcpReference));
+                nWidth = MicroProfileMin(nMaxWidth, (uint32_t)(nMaxWidth * pMs[i] * fRcpReference));
                 MicroProfileDrawBox(nMaxOffsetX, nOffsetY, nMaxOffsetX+nWidth, nOffsetY+MICROPROFILE_TEXT_HEIGHT, pColors[i]|0xff000000);
             }
         }

src/CMakeLists.txt

@@ -65,6 +65,10 @@ if (MSVC)
         /we4305 # 'context': truncation from 'type1' to 'type2'
         /we4388 # 'expression': signed/unsigned mismatch
         /we4389 # 'operator': signed/unsigned mismatch
+        /we4456 # Declaration of 'identifier' hides previous local declaration
+        /we4457 # Declaration of 'identifier' hides function parameter
+        /we4458 # Declaration of 'identifier' hides class member
+        /we4459 # Declaration of 'identifier' hides global declaration
         /we4505 # 'function': unreferenced local function has been removed
         /we4547 # 'operator': operator before comma has no effect; expected operator with side-effect
         /we4549 # 'operator1': operator before comma has no effect; did you intend 'operator2'?
@@ -92,6 +96,7 @@ else()
         -Werror=missing-declarations
         -Werror=missing-field-initializers
         -Werror=reorder
+        -Werror=shadow
         -Werror=sign-compare
         -Werror=switch
         -Werror=uninitialized

src/audio_core/CMakeLists.txt

@@ -49,9 +49,6 @@ if (NOT MSVC)
     target_compile_options(audio_core PRIVATE
         -Werror=conversion
         -Werror=ignored-qualifiers
-        -Werror=shadow
-        -Werror=unused-parameter
-        -Werror=unused-variable
 
         $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
         $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>

src/audio_core/command_generator.cpp

@@ -429,7 +429,7 @@ void CommandGenerator::GenerateDataSourceCommand(ServerVoiceInfo& voice_info, Vo
                                   in_params.node_id);
             break;
         default:
-            UNREACHABLE_MSG("Unimplemented sample format={}", in_params.sample_format);
+            ASSERT_MSG(false, "Unimplemented sample format={}", in_params.sample_format);
         }
     }
 }
@@ -1312,7 +1312,7 @@ void CommandGenerator::DecodeFromWaveBuffers(ServerVoiceInfo& voice_info, std::s
                                 samples_to_read - samples_read, channel, temp_mix_offset);
                 break;
             default:
-                UNREACHABLE_MSG("Unimplemented sample format={}", in_params.sample_format);
+                ASSERT_MSG(false, "Unimplemented sample format={}", in_params.sample_format);
             }
 
             temp_mix_offset += samples_decoded;

src/audio_core/effect_context.cpp

@@ -50,7 +50,7 @@ EffectBase* EffectContext::RetargetEffect(std::size_t i, EffectType effect) {
         effects[i] = std::make_unique<EffectBiquadFilter>();
         break;
     default:
-        UNREACHABLE_MSG("Unimplemented effect {}", effect);
+        ASSERT_MSG(false, "Unimplemented effect {}", effect);
         effects[i] = std::make_unique<EffectStubbed>();
     }
     return GetInfo(i);
@@ -104,7 +104,7 @@ void EffectI3dl2Reverb::Update(EffectInfo::InParams& in_params) {
     auto& params = GetParams();
     const auto* reverb_params = reinterpret_cast<I3dl2ReverbParams*>(in_params.raw.data());
     if (!ValidChannelCountForEffect(reverb_params->max_channels)) {
-        UNREACHABLE_MSG("Invalid reverb max channel count {}", reverb_params->max_channels);
+        ASSERT_MSG(false, "Invalid reverb max channel count {}", reverb_params->max_channels);
         return;
     }
 

src/audio_core/splitter_context.cpp

@@ -483,7 +483,7 @@ bool NodeStates::DepthFirstSearch(EdgeMatrix& edge_matrix) {
                     // Add more work
                     index_stack.push(j);
                 } else if (node_state == NodeStates::State::InFound) {
-                    UNREACHABLE_MSG("Node start marked as found");
+                    ASSERT_MSG(false, "Node start marked as found");
                     ResetState();
                     return false;
                 }

src/audio_core/voice_context.cpp

@@ -114,7 +114,7 @@ void ServerVoiceInfo::UpdateParameters(const VoiceInfo::InParams& voice_in,
         in_params.current_playstate = ServerPlayState::Play;
         break;
     default:
-        UNREACHABLE_MSG("Unknown playstate {}", voice_in.play_state);
+        ASSERT_MSG(false, "Unknown playstate {}", voice_in.play_state);
         break;
     }
 
@@ -410,7 +410,7 @@ bool ServerVoiceInfo::UpdateParametersForCommandGeneration(
         return in_params.should_depop;
     }
     default:
-        UNREACHABLE_MSG("Invalid playstate {}", in_params.current_playstate);
+        ASSERT_MSG(false, "Invalid playstate {}", in_params.current_playstate);
     }
 
     return false;

src/common/assert.cpp

@@ -6,8 +6,13 @@
 
 #include "common/settings.h"
 
-void assert_handle_failure() {
+void assert_fail_impl() {
     if (Settings::values.use_debug_asserts) {
         Crash();
     }
 }
+
+[[noreturn]] void unreachable_impl() {
+    Crash();
+    throw std::runtime_error("Unreachable code");
+}

src/common/assert.h

@@ -9,44 +9,43 @@
 // Sometimes we want to try to continue even after hitting an assert.
 // However touching this file yields a global recompilation as this header is included almost
 // everywhere. So let's just move the handling of the failed assert to a single cpp file.
-void assert_handle_failure();
 
-// For asserts we'd like to keep all the junk executed when an assert happens away from the
-// important code in the function. One way of doing this is to put all the relevant code inside a
-// lambda and force the compiler to not inline it. Unfortunately, MSVC seems to have no syntax to
-// specify __declspec on lambda functions, so what we do instead is define a noinline wrapper
-// template that calls the lambda. This seems to generate an extra instruction at the call-site
-// compared to the ideal implementation (which wouldn't support ASSERT_MSG parameters), but is good
-// enough for our purposes.
-template <typename Fn>
-#if defined(_MSC_VER)
-[[msvc::noinline]]
-#elif defined(__GNUC__)
-[[gnu::cold, gnu::noinline]]
+void assert_fail_impl();
+[[noreturn]] void unreachable_impl();
+
+#ifdef _MSC_VER
+#define YUZU_NO_INLINE __declspec(noinline)
+#else
+#define YUZU_NO_INLINE __attribute__((noinline))
 #endif
-static void
-assert_noinline_call(const Fn& fn) {
-    fn();
-    assert_handle_failure();
-}
 
 #define ASSERT(_a_)                                                                                \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([] { LOG_CRITICAL(Debug, "Assertion Failed!"); });                \
+    ([&]() YUZU_NO_INLINE {                                                                        \
+        if (!(_a_)) [[unlikely]] {                                                                 \
+            LOG_CRITICAL(Debug, "Assertion Failed!");                                              \
+            assert_fail_impl();                                                                    \
         }                                                                                          \
-    while (0)
+    }())
 
 #define ASSERT_MSG(_a_, ...)                                                                       \
-    do                                                                                             \
-        if (!(_a_)) {                                                                              \
-            assert_noinline_call([&] { LOG_CRITICAL(Debug, "Assertion Failed!\n" __VA_ARGS__); }); \
+    ([&]() YUZU_NO_INLINE {                                                                        \
+        if (!(_a_)) [[unlikely]] {                                                                 \
+            LOG_CRITICAL(Debug, "Assertion Failed!\n" __VA_ARGS__);                                \
+            assert_fail_impl();                                                                    \
         }                                                                                          \
-    while (0)
+    }())
+
+#define UNREACHABLE()                                                                              \
+    do {                                                                                           \
+        LOG_CRITICAL(Debug, "Unreachable code!");                                                  \
+        unreachable_impl();                                                                        \
+    } while (0)
 
-#define UNREACHABLE() assert_noinline_call([] { LOG_CRITICAL(Debug, "Unreachable code!"); })
 #define UNREACHABLE_MSG(...)                                                                       \
-    assert_noinline_call([&] { LOG_CRITICAL(Debug, "Unreachable code!\n" __VA_ARGS__); })
+    do {                                                                                           \
+        LOG_CRITICAL(Debug, "Unreachable code!\n" __VA_ARGS__);                                    \
+        unreachable_impl();                                                                        \
+    } while (0)
 
 #ifdef _DEBUG
 #define DEBUG_ASSERT(_a_) ASSERT(_a_)

src/common/bounded_threadsafe_queue.h

@@ -0,0 +1,167 @@
+// SPDX-FileCopyrightText: Copyright (c) 2020 Erik Rigtorp <erik@rigtorp.se>
+// SPDX-License-Identifier: MIT
+
+#pragma once
+
+#include <atomic>
+#include <bit>
+#include <condition_variable>
+#include <memory>
+#include <mutex>
+#include <new>
+#include <stop_token>
+#include <type_traits>
+#include <utility>
+
+namespace Common {
+
+#if defined(__cpp_lib_hardware_interference_size)
+constexpr size_t hardware_interference_size = std::hardware_destructive_interference_size;
+#else
+constexpr size_t hardware_interference_size = 64;
+#endif
+
+#ifdef _MSC_VER
+#pragma warning(push)
+#pragma warning(disable : 4324)
+#endif
+
+template <typename T, size_t capacity = 0x400>
+class MPSCQueue {
+public:
+    explicit MPSCQueue() : allocator{std::allocator<Slot<T>>()} {
+        // Allocate one extra slot to prevent false sharing on the last slot
+        slots = allocator.allocate(capacity + 1);
+        // Allocators are not required to honor alignment for over-aligned types
+        // (see http://eel.is/c++draft/allocator.requirements#10) so we verify
+        // alignment here
+        if (reinterpret_cast<uintptr_t>(slots) % alignof(Slot<T>) != 0) {
+            allocator.deallocate(slots, capacity + 1);
+            throw std::bad_alloc();
+        }
+        for (size_t i = 0; i < capacity; ++i) {
+            std::construct_at(&slots[i]);
+        }
+        static_assert(std::has_single_bit(capacity), "capacity must be an integer power of 2");
+        static_assert(alignof(Slot<T>) == hardware_interference_size,
+                      "Slot must be aligned to cache line boundary to prevent false sharing");
+        static_assert(sizeof(Slot<T>) % hardware_interference_size == 0,
+                      "Slot size must be a multiple of cache line size to prevent "
+                      "false sharing between adjacent slots");
+        static_assert(sizeof(MPSCQueue) % hardware_interference_size == 0,
+                      "Queue size must be a multiple of cache line size to "
+                      "prevent false sharing between adjacent queues");
+    }
+
+    ~MPSCQueue() noexcept {
+        for (size_t i = 0; i < capacity; ++i) {
+            std::destroy_at(&slots[i]);
+        }
+        allocator.deallocate(slots, capacity + 1);
+    }
+
+    // The queue must be both non-copyable and non-movable
+    MPSCQueue(const MPSCQueue&) = delete;
+    MPSCQueue& operator=(const MPSCQueue&) = delete;
+
+    MPSCQueue(MPSCQueue&&) = delete;
+    MPSCQueue& operator=(MPSCQueue&&) = delete;
+
+    void Push(const T& v) noexcept {
+        static_assert(std::is_nothrow_copy_constructible_v<T>,
+                      "T must be nothrow copy constructible");
+        emplace(v);
+    }
+
+    template <typename P, typename = std::enable_if_t<std::is_nothrow_constructible_v<T, P&&>>>
+    void Push(P&& v) noexcept {
+        emplace(std::forward<P>(v));
+    }
+
+    void Pop(T& v, std::stop_token stop) noexcept {
+        auto const tail = tail_.fetch_add(1);
+        auto& slot = slots[idx(tail)];
+        if (!slot.turn.test()) {
+            std::unique_lock lock{cv_mutex};
+            cv.wait(lock, stop, [&slot] { return slot.turn.test(); });
+        }
+        v = slot.move();
+        slot.destroy();
+        slot.turn.clear();
+        slot.turn.notify_one();
+    }
+
+private:
+    template <typename U = T>
+    struct Slot {
+        ~Slot() noexcept {
+            if (turn.test()) {
+                destroy();
+            }
+        }
+
+        template <typename... Args>
+        void construct(Args&&... args) noexcept {
+            static_assert(std::is_nothrow_constructible_v<U, Args&&...>,
+                          "T must be nothrow constructible with Args&&...");
+            std::construct_at(reinterpret_cast<U*>(&storage), std::forward<Args>(args)...);
+        }
+
+        void destroy() noexcept {
+            static_assert(std::is_nothrow_destructible_v<U>, "T must be nothrow destructible");
+            std::destroy_at(reinterpret_cast<U*>(&storage));
+        }
+
+        U&& move() noexcept {
+            return reinterpret_cast<U&&>(storage);
+        }
+
+        // Align to avoid false sharing between adjacent slots
+        alignas(hardware_interference_size) std::atomic_flag turn{};
+        struct aligned_store {
+            struct type {
+                alignas(U) unsigned char data[sizeof(U)];
+            };
+        };
+        typename aligned_store::type storage;
+    };
+
+    template <typename... Args>
+    void emplace(Args&&... args) noexcept {
+        static_assert(std::is_nothrow_constructible_v<T, Args&&...>,
+                      "T must be nothrow constructible with Args&&...");
+        auto const head = head_.fetch_add(1);
+        auto& slot = slots[idx(head)];
+        slot.turn.wait(true);
+        slot.construct(std::forward<Args>(args)...);
+        slot.turn.test_and_set();
+        cv.notify_one();
+    }
+
+    constexpr size_t idx(size_t i) const noexcept {
+        return i & mask;
+    }
+
+    static constexpr size_t mask = capacity - 1;
+
+    // Align to avoid false sharing between head_ and tail_
+    alignas(hardware_interference_size) std::atomic<size_t> head_{0};
+    alignas(hardware_interference_size) std::atomic<size_t> tail_{0};
+
+    std::mutex cv_mutex;
+    std::condition_variable_any cv;
+
+    Slot<T>* slots;
+    [[no_unique_address]] std::allocator<Slot<T>> allocator;
+
+    static_assert(std::is_nothrow_copy_assignable_v<T> || std::is_nothrow_move_assignable_v<T>,
+                  "T must be nothrow copy or move assignable");
+
+    static_assert(std::is_nothrow_destructible_v<T>, "T must be nothrow destructible");
+};
+
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif
+
+} // namespace Common

src/common/detached_tasks.cpp

@@ -33,9 +33,9 @@ void DetachedTasks::AddTask(std::function<void()> task) {
     ++instance->count;
     std::thread([task{std::move(task)}]() {
         task();
-        std::unique_lock lock{instance->mutex};
+        std::unique_lock thread_lock{instance->mutex};
         --instance->count;
-        std::notify_all_at_thread_exit(instance->cv, std::move(lock));
+        std::notify_all_at_thread_exit(instance->cv, std::move(thread_lock));
     }).detach();
 }
 

src/common/page_table.h

@@ -15,6 +15,9 @@ enum class PageType : u8 {
     Unmapped,
     /// Page is mapped to regular memory. This is the only type you can get pointers to.
     Memory,
+    /// Page is mapped to regular memory, but inaccessible from CPU fastmem and must use
+    /// the callbacks.
+    DebugMemory,
     /// Page is mapped to regular memory, but also needs to check for rasterizer cache flushing and
     /// invalidation
     RasterizerCachedMemory,

src/common/param_package.cpp

@@ -76,7 +76,7 @@ std::string ParamPackage::Serialize() const {
 std::string ParamPackage::Get(const std::string& key, const std::string& default_value) const {
     auto pair = data.find(key);
     if (pair == data.end()) {
-        LOG_DEBUG(Common, "key '{}' not found", key);
+        LOG_TRACE(Common, "key '{}' not found", key);
         return default_value;
     }
 
@@ -86,7 +86,7 @@ std::string ParamPackage::Get(const std::string& key, const std::string& default
 int ParamPackage::Get(const std::string& key, int default_value) const {
     auto pair = data.find(key);
     if (pair == data.end()) {
-        LOG_DEBUG(Common, "key '{}' not found", key);
+        LOG_TRACE(Common, "key '{}' not found", key);
         return default_value;
     }
 
@@ -101,7 +101,7 @@ int ParamPackage::Get(const std::string& key, int default_value) const {
 float ParamPackage::Get(const std::string& key, float default_value) const {
     auto pair = data.find(key);
     if (pair == data.end()) {
-        LOG_DEBUG(Common, "key {} not found", key);
+        LOG_TRACE(Common, "key {} not found", key);
         return default_value;
     }
 

src/common/settings.cpp

@@ -147,7 +147,7 @@ void UpdateRescalingInfo() {
         info.down_shift = 0;
         break;
     default:
-        UNREACHABLE();
+        ASSERT(false);
         info.up_scale = 1;
         info.down_shift = 0;
     }

src/common/settings.h

@@ -496,7 +496,7 @@ struct Values {
 
     // Renderer
     RangedSetting<RendererBackend> renderer_backend{
-        RendererBackend::OpenGL, RendererBackend::OpenGL, RendererBackend::Vulkan, "backend"};
+        RendererBackend::Vulkan, RendererBackend::OpenGL, RendererBackend::Vulkan, "backend"};
     BasicSetting<bool> renderer_debug{false, "debug"};
     BasicSetting<bool> renderer_shader_feedback{false, "shader_feedback"};
     BasicSetting<bool> enable_nsight_aftermath{false, "nsight_aftermath"};

src/core/CMakeLists.txt

@@ -743,16 +743,11 @@ if (MSVC)
         /we4244 # 'conversion': conversion from 'type1' to 'type2', possible loss of data
         /we4245 # 'conversion': conversion from 'type1' to 'type2', signed/unsigned mismatch
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
-        /we4456 # Declaration of 'identifier' hides previous local declaration
-        /we4457 # Declaration of 'identifier' hides function parameter
-        /we4458 # Declaration of 'identifier' hides class member
-        /we4459 # Declaration of 'identifier' hides global declaration
     )
 else()
     target_compile_options(core PRIVATE
         -Werror=conversion
         -Werror=ignored-qualifiers
-        -Werror=shadow
 
         $<$<CXX_COMPILER_ID:GNU>:-Werror=class-memaccess>
         $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>

src/core/arm/arm_interface.cpp

@@ -95,7 +95,7 @@ void ARM_Interface::Run() {
     using Kernel::SuspendType;
 
     while (true) {
-        Kernel::KThread* current_thread{system.Kernel().CurrentScheduler()->GetCurrentThread()};
+        Kernel::KThread* current_thread{Kernel::GetCurrentThreadPointer(system.Kernel())};
         Dynarmic::HaltReason hr{};
 
         // Notify the debugger and go to sleep if a step was performed
@@ -107,6 +107,7 @@ void ARM_Interface::Run() {
         }
 
         // Otherwise, run the thread.
+        system.EnterDynarmicProfile();
         if (current_thread->GetStepState() == StepState::StepPending) {
             hr = StepJit();
 
@@ -116,11 +117,19 @@ void ARM_Interface::Run() {
         } else {
             hr = RunJit();
         }
+        system.ExitDynarmicProfile();
 
         // Notify the debugger and go to sleep if a breakpoint was hit.
         if (Has(hr, breakpoint)) {
+            RewindBreakpointInstruction();
             system.GetDebugger().NotifyThreadStopped(current_thread);
-            current_thread->RequestSuspend(Kernel::SuspendType::Debug);
+            current_thread->RequestSuspend(SuspendType::Debug);
+            break;
+        }
+        if (Has(hr, watchpoint)) {
+            RewindBreakpointInstruction();
+            system.GetDebugger().NotifyThreadWatchpoint(current_thread, *HaltedWatchpoint());
+            current_thread->RequestSuspend(SuspendType::Debug);
             break;
         }
 
@@ -134,4 +143,36 @@ void ARM_Interface::Run() {
     }
 }
 
+void ARM_Interface::LoadWatchpointArray(const WatchpointArray& wp) {
+    watchpoints = ℘
+}
+
+const Kernel::DebugWatchpoint* ARM_Interface::MatchingWatchpoint(
+    VAddr addr, u64 size, Kernel::DebugWatchpointType access_type) const {
+    if (!watchpoints) {
+        return nullptr;
+    }
+
+    const VAddr start_address{addr};
+    const VAddr end_address{addr + size};
+
+    for (size_t i = 0; i < Core::Hardware::NUM_WATCHPOINTS; i++) {
+        const auto& watch{(*watchpoints)[i]};
+
+        if (end_address <= watch.start_address) {
+            continue;
+        }
+        if (start_address >= watch.end_address) {
+            continue;
+        }
+        if ((access_type & watch.type) == Kernel::DebugWatchpointType::None) {
+            continue;
+        }
+
+        return &watch;
+    }
+
+    return nullptr;
+}
+
 } // namespace Core

src/core/arm/arm_interface.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <span>
 #include <vector>
 
 #include <dynarmic/interface/halt_reason.h>
@@ -19,13 +20,16 @@ struct PageTable;
 
 namespace Kernel {
 enum class VMAPermission : u8;
-}
+enum class DebugWatchpointType : u8;
+struct DebugWatchpoint;
+} // namespace Kernel
 
 namespace Core {
 class System;
 class CPUInterruptHandler;
 
 using CPUInterrupts = std::array<CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES>;
+using WatchpointArray = std::array<Kernel::DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS>;
 
 /// Generic ARMv8 CPU interface
 class ARM_Interface {
@@ -170,6 +174,7 @@ public:
     virtual void SaveContext(ThreadContext64& ctx) = 0;
     virtual void LoadContext(const ThreadContext32& ctx) = 0;
     virtual void LoadContext(const ThreadContext64& ctx) = 0;
+    void LoadWatchpointArray(const WatchpointArray& wp);
 
     /// Clears the exclusive monitor's state.
     virtual void ClearExclusiveState() = 0;
@@ -198,18 +203,24 @@ public:
     static constexpr Dynarmic::HaltReason break_loop = Dynarmic::HaltReason::UserDefined2;
     static constexpr Dynarmic::HaltReason svc_call = Dynarmic::HaltReason::UserDefined3;
     static constexpr Dynarmic::HaltReason breakpoint = Dynarmic::HaltReason::UserDefined4;
+    static constexpr Dynarmic::HaltReason watchpoint = Dynarmic::HaltReason::UserDefined5;
 
 protected:
     /// System context that this ARM interface is running under.
     System& system;
     CPUInterrupts& interrupt_handlers;
+    const WatchpointArray* watchpoints;
     bool uses_wall_clock;
 
     static void SymbolicateBacktrace(Core::System& system, std::vector<BacktraceEntry>& out);
+    const Kernel::DebugWatchpoint* MatchingWatchpoint(
+        VAddr addr, u64 size, Kernel::DebugWatchpointType access_type) const;
 
     virtual Dynarmic::HaltReason RunJit() = 0;
     virtual Dynarmic::HaltReason StepJit() = 0;
     virtual u32 GetSvcNumber() const = 0;
+    virtual const Kernel::DebugWatchpoint* HaltedWatchpoint() const = 0;
+    virtual void RewindBreakpointInstruction() = 0;
 };
 
 } // namespace Core

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -29,45 +29,62 @@ using namespace Common::Literals;
 class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
 public:
     explicit DynarmicCallbacks32(ARM_Dynarmic_32& parent_)
-        : parent{parent_}, memory(parent.system.Memory()) {}
+        : parent{parent_},
+          memory(parent.system.Memory()), debugger_enabled{parent.system.DebuggerEnabled()} {}
 
     u8 MemoryRead8(u32 vaddr) override {
+        CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
         return memory.Read8(vaddr);
     }
     u16 MemoryRead16(u32 vaddr) override {
+        CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
         return memory.Read16(vaddr);
     }
     u32 MemoryRead32(u32 vaddr) override {
+        CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
         return memory.Read32(vaddr);
     }
     u64 MemoryRead64(u32 vaddr) override {
+        CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
         return memory.Read64(vaddr);
     }
 
     void MemoryWrite8(u32 vaddr, u8 value) override {
-        memory.Write8(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
+            memory.Write8(vaddr, value);
+        }
     }
     void MemoryWrite16(u32 vaddr, u16 value) override {
-        memory.Write16(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
+            memory.Write16(vaddr, value);
+        }
     }
     void MemoryWrite32(u32 vaddr, u32 value) override {
-        memory.Write32(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
+            memory.Write32(vaddr, value);
+        }
     }
     void MemoryWrite64(u32 vaddr, u64 value) override {
-        memory.Write64(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
+            memory.Write64(vaddr, value);
+        }
     }
 
     bool MemoryWriteExclusive8(u32 vaddr, u8 value, u8 expected) override {
-        return memory.WriteExclusive8(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive8(vaddr, value, expected);
     }
     bool MemoryWriteExclusive16(u32 vaddr, u16 value, u16 expected) override {
-        return memory.WriteExclusive16(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive16(vaddr, value, expected);
     }
     bool MemoryWriteExclusive32(u32 vaddr, u32 value, u32 expected) override {
-        return memory.WriteExclusive32(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive32(vaddr, value, expected);
     }
     bool MemoryWriteExclusive64(u32 vaddr, u64 value, u64 expected) override {
-        return memory.WriteExclusive64(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive64(vaddr, value, expected);
     }
 
     void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
@@ -77,8 +94,8 @@ public:
     }
 
     void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
-        if (parent.system.DebuggerEnabled()) {
-            parent.jit.load()->Regs()[15] = pc;
+        if (debugger_enabled) {
+            parent.SaveContext(parent.breakpoint_context);
             parent.jit.load()->HaltExecution(ARM_Interface::breakpoint);
             return;
         }
@@ -117,9 +134,26 @@ public:
         return std::max<s64>(parent.system.CoreTiming().GetDowncount(), 0);
     }
 
+    bool CheckMemoryAccess(VAddr addr, u64 size, Kernel::DebugWatchpointType type) {
+        if (!debugger_enabled) {
+            return true;
+        }
+
+        const auto match{parent.MatchingWatchpoint(addr, size, type)};
+        if (match) {
+            parent.SaveContext(parent.breakpoint_context);
+            parent.jit.load()->HaltExecution(ARM_Interface::watchpoint);
+            parent.halted_watchpoint = match;
+            return false;
+        }
+
+        return true;
+    }
+
     ARM_Dynarmic_32& parent;
     Core::Memory::Memory& memory;
     std::size_t num_interpreted_instructions{};
+    bool debugger_enabled{};
     static constexpr u64 minimum_run_cycles = 1000U;
 };
 
@@ -154,6 +188,11 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
     config.code_cache_size = 512_MiB;
     config.far_code_offset = 400_MiB;
 
+    // Allow memory fault handling to work
+    if (system.DebuggerEnabled()) {
+        config.check_halt_on_memory_access = true;
+    }
+
     // null_jit
     if (!page_table) {
         // Don't waste too much memory on null_jit
@@ -248,6 +287,14 @@ u32 ARM_Dynarmic_32::GetSvcNumber() const {
     return svc_swi;
 }
 
+const Kernel::DebugWatchpoint* ARM_Dynarmic_32::HaltedWatchpoint() const {
+    return halted_watchpoint;
+}
+
+void ARM_Dynarmic_32::RewindBreakpointInstruction() {
+    LoadContext(breakpoint_context);
+}
+
 ARM_Dynarmic_32::ARM_Dynarmic_32(System& system_, CPUInterrupts& interrupt_handlers_,
                                  bool uses_wall_clock_, ExclusiveMonitor& exclusive_monitor_,
                                  std::size_t core_index_)

src/core/arm/dynarmic/arm_dynarmic_32.h

@@ -72,6 +72,8 @@ protected:
     Dynarmic::HaltReason RunJit() override;
     Dynarmic::HaltReason StepJit() override;
     u32 GetSvcNumber() const override;
+    const Kernel::DebugWatchpoint* HaltedWatchpoint() const override;
+    void RewindBreakpointInstruction() override;
 
 private:
     std::shared_ptr<Dynarmic::A32::Jit> MakeJit(Common::PageTable* page_table) const;
@@ -98,6 +100,10 @@ private:
 
     // SVC callback
     u32 svc_swi{};
+
+    // Watchpoint info
+    const Kernel::DebugWatchpoint* halted_watchpoint;
+    ThreadContext32 breakpoint_context;
 };
 
 } // namespace Core

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -29,55 +29,76 @@ using namespace Common::Literals;
 class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
     explicit DynarmicCallbacks64(ARM_Dynarmic_64& parent_)
-        : parent{parent_}, memory(parent.system.Memory()) {}
+        : parent{parent_},
+          memory(parent.system.Memory()), debugger_enabled{parent.system.DebuggerEnabled()} {}
 
     u8 MemoryRead8(u64 vaddr) override {
+        CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Read);
         return memory.Read8(vaddr);
     }
     u16 MemoryRead16(u64 vaddr) override {
+        CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Read);
         return memory.Read16(vaddr);
     }
     u32 MemoryRead32(u64 vaddr) override {
+        CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Read);
         return memory.Read32(vaddr);
     }
     u64 MemoryRead64(u64 vaddr) override {
+        CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Read);
         return memory.Read64(vaddr);
     }
     Vector MemoryRead128(u64 vaddr) override {
+        CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Read);
         return {memory.Read64(vaddr), memory.Read64(vaddr + 8)};
     }
 
     void MemoryWrite8(u64 vaddr, u8 value) override {
-        memory.Write8(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write)) {
+            memory.Write8(vaddr, value);
+        }
     }
     void MemoryWrite16(u64 vaddr, u16 value) override {
-        memory.Write16(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write)) {
+            memory.Write16(vaddr, value);
+        }
     }
     void MemoryWrite32(u64 vaddr, u32 value) override {
-        memory.Write32(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write)) {
+            memory.Write32(vaddr, value);
+        }
     }
     void MemoryWrite64(u64 vaddr, u64 value) override {
-        memory.Write64(vaddr, value);
+        if (CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write)) {
+            memory.Write64(vaddr, value);
+        }
     }
     void MemoryWrite128(u64 vaddr, Vector value) override {
-        memory.Write64(vaddr, value[0]);
-        memory.Write64(vaddr + 8, value[1]);
+        if (CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Write)) {
+            memory.Write64(vaddr, value[0]);
+            memory.Write64(vaddr + 8, value[1]);
+        }
     }
 
     bool MemoryWriteExclusive8(u64 vaddr, std::uint8_t value, std::uint8_t expected) override {
-        return memory.WriteExclusive8(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 1, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive8(vaddr, value, expected);
     }
     bool MemoryWriteExclusive16(u64 vaddr, std::uint16_t value, std::uint16_t expected) override {
-        return memory.WriteExclusive16(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 2, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive16(vaddr, value, expected);
     }
     bool MemoryWriteExclusive32(u64 vaddr, std::uint32_t value, std::uint32_t expected) override {
-        return memory.WriteExclusive32(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 4, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive32(vaddr, value, expected);
     }
     bool MemoryWriteExclusive64(u64 vaddr, std::uint64_t value, std::uint64_t expected) override {
-        return memory.WriteExclusive64(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 8, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive64(vaddr, value, expected);
     }
     bool MemoryWriteExclusive128(u64 vaddr, Vector value, Vector expected) override {
-        return memory.WriteExclusive128(vaddr, value, expected);
+        return CheckMemoryAccess(vaddr, 16, Kernel::DebugWatchpointType::Write) &&
+               memory.WriteExclusive128(vaddr, value, expected);
     }
 
     void InterpreterFallback(u64 pc, std::size_t num_instructions) override {
@@ -118,8 +139,8 @@ public:
         case Dynarmic::A64::Exception::Yield:
             return;
         default:
-            if (parent.system.DebuggerEnabled()) {
-                parent.jit.load()->SetPC(pc);
+            if (debugger_enabled) {
+                parent.SaveContext(parent.breakpoint_context);
                 parent.jit.load()->HaltExecution(ARM_Interface::breakpoint);
                 return;
             }
@@ -160,10 +181,27 @@ public:
         return parent.system.CoreTiming().GetClockTicks();
     }
 
+    bool CheckMemoryAccess(VAddr addr, u64 size, Kernel::DebugWatchpointType type) {
+        if (!debugger_enabled) {
+            return true;
+        }
+
+        const auto match{parent.MatchingWatchpoint(addr, size, type)};
+        if (match) {
+            parent.SaveContext(parent.breakpoint_context);
+            parent.jit.load()->HaltExecution(ARM_Interface::watchpoint);
+            parent.halted_watchpoint = match;
+            return false;
+        }
+
+        return true;
+    }
+
     ARM_Dynarmic_64& parent;
     Core::Memory::Memory& memory;
     u64 tpidrro_el0 = 0;
     u64 tpidr_el0 = 0;
+    bool debugger_enabled{};
     static constexpr u64 minimum_run_cycles = 1000U;
 };
 
@@ -214,6 +252,11 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
     config.code_cache_size = 512_MiB;
     config.far_code_offset = 400_MiB;
 
+    // Allow memory fault handling to work
+    if (system.DebuggerEnabled()) {
+        config.check_halt_on_memory_access = true;
+    }
+
     // null_jit
     if (!page_table) {
         // Don't waste too much memory on null_jit
@@ -308,6 +351,14 @@ u32 ARM_Dynarmic_64::GetSvcNumber() const {
     return svc_swi;
 }
 
+const Kernel::DebugWatchpoint* ARM_Dynarmic_64::HaltedWatchpoint() const {
+    return halted_watchpoint;
+}
+
+void ARM_Dynarmic_64::RewindBreakpointInstruction() {
+    LoadContext(breakpoint_context);
+}
+
 ARM_Dynarmic_64::ARM_Dynarmic_64(System& system_, CPUInterrupts& interrupt_handlers_,
                                  bool uses_wall_clock_, ExclusiveMonitor& exclusive_monitor_,
                                  std::size_t core_index_)

src/core/arm/dynarmic/arm_dynarmic_64.h

@@ -66,6 +66,8 @@ protected:
     Dynarmic::HaltReason RunJit() override;
     Dynarmic::HaltReason StepJit() override;
     u32 GetSvcNumber() const override;
+    const Kernel::DebugWatchpoint* HaltedWatchpoint() const override;
+    void RewindBreakpointInstruction() override;
 
 private:
     std::shared_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable* page_table,
@@ -91,6 +93,10 @@ private:
 
     // SVC callback
     u32 svc_swi{};
+
+    // Breakpoint info
+    const Kernel::DebugWatchpoint* halted_watchpoint;
+    ThreadContext64 breakpoint_context;
 };
 
 } // namespace Core

src/core/core.cpp

@@ -138,7 +138,6 @@ struct System::Impl {
 
         kernel.Suspend(false);
         core_timing.SyncPause(false);
-        cpu_manager.Pause(false);
         is_paused = false;
 
         return status;
@@ -150,25 +149,22 @@ struct System::Impl {
 
         core_timing.SyncPause(true);
         kernel.Suspend(true);
-        cpu_manager.Pause(true);
         is_paused = true;
 
         return status;
     }
 
-    std::unique_lock<std::mutex> StallCPU() {
+    std::unique_lock<std::mutex> StallProcesses() {
         std::unique_lock<std::mutex> lk(suspend_guard);
         kernel.Suspend(true);
         core_timing.SyncPause(true);
-        cpu_manager.Pause(true);
         return lk;
     }
 
-    void UnstallCPU() {
+    void UnstallProcesses() {
         if (!is_paused) {
             core_timing.SyncPause(false);
             kernel.Suspend(false);
-            cpu_manager.Pause(false);
         }
     }
 
@@ -334,6 +330,8 @@ struct System::Impl {
             gpu_core->NotifyShutdown();
         }
 
+        kernel.ShutdownCores();
+        cpu_manager.Shutdown();
         debugger.reset();
         services.reset();
         service_manager.reset();
@@ -493,12 +491,18 @@ void System::Shutdown() {
     impl->Shutdown();
 }
 
-std::unique_lock<std::mutex> System::StallCPU() {
-    return impl->StallCPU();
+void System::DetachDebugger() {
+    if (impl->debugger) {
+        impl->debugger->NotifyShutdown();
+    }
 }
 
-void System::UnstallCPU() {
-    impl->UnstallCPU();
+std::unique_lock<std::mutex> System::StallProcesses() {
+    return impl->StallProcesses();
+}
+
+void System::UnstallProcesses() {
+    impl->UnstallProcesses();
 }
 
 void System::InitializeDebugger() {

src/core/core.h

@@ -160,8 +160,11 @@ public:
     /// Shutdown the emulated system.
     void Shutdown();
 
-    std::unique_lock<std::mutex> StallCPU();
-    void UnstallCPU();
+    /// Forcibly detach the debugger if it is running.
+    void DetachDebugger();
+
+    std::unique_lock<std::mutex> StallProcesses();
+    void UnstallProcesses();
 
     /**
      * Initialize the debugger.

src/core/cpu_manager.cpp

@@ -21,23 +21,24 @@ CpuManager::~CpuManager() = default;
 
 void CpuManager::ThreadStart(std::stop_token stop_token, CpuManager& cpu_manager,
                              std::size_t core) {
-    cpu_manager.RunThread(stop_token, core);
+    cpu_manager.RunThread(core);
 }
 
 void CpuManager::Initialize() {
-    running_mode = true;
-    if (is_multicore) {
-        for (std::size_t core = 0; core < Core::Hardware::NUM_CPU_CORES; core++) {
-            core_data[core].host_thread = std::jthread(ThreadStart, std::ref(*this), core);
-        }
-    } else {
-        core_data[0].host_thread = std::jthread(ThreadStart, std::ref(*this), 0);
+    num_cores = is_multicore ? Core::Hardware::NUM_CPU_CORES : 1;
+    gpu_barrier = std::make_unique<Common::Barrier>(num_cores + 1);
+
+    for (std::size_t core = 0; core < num_cores; core++) {
+        core_data[core].host_thread = std::jthread(ThreadStart, std::ref(*this), core);
     }
 }
 
 void CpuManager::Shutdown() {
-    running_mode = false;
-    Pause(false);
+    for (std::size_t core = 0; core < num_cores; core++) {
+        if (core_data[core].host_thread.joinable()) {
+            core_data[core].host_thread.join();
+        }
+    }
 }
 
 std::function<void(void*)> CpuManager::GetGuestThreadStartFunc() {
@@ -48,8 +49,8 @@ std::function<void(void*)> CpuManager::GetIdleThreadStartFunc() {
     return IdleThreadFunction;
 }
 
-std::function<void(void*)> CpuManager::GetSuspendThreadStartFunc() {
-    return SuspendThreadFunction;
+std::function<void(void*)> CpuManager::GetShutdownThreadStartFunc() {
+    return ShutdownThreadFunction;
 }
 
 void CpuManager::GuestThreadFunction(void* cpu_manager_) {
@@ -79,17 +80,12 @@ void CpuManager::IdleThreadFunction(void* cpu_manager_) {
     }
 }
 
-void CpuManager::SuspendThreadFunction(void* cpu_manager_) {
-    CpuManager* cpu_manager = static_cast<CpuManager*>(cpu_manager_);
-    if (cpu_manager->is_multicore) {
-        cpu_manager->MultiCoreRunSuspendThread();
-    } else {
-        cpu_manager->SingleCoreRunSuspendThread();
-    }
+void CpuManager::ShutdownThreadFunction(void* cpu_manager) {
+    static_cast<CpuManager*>(cpu_manager)->ShutdownThread();
 }
 
-void* CpuManager::GetStartFuncParamater() {
-    return static_cast<void*>(this);
+void* CpuManager::GetStartFuncParameter() {
+    return this;
 }
 
 ///////////////////////////////////////////////////////////////////////////////
@@ -99,7 +95,7 @@ void* CpuManager::GetStartFuncParamater() {
 void CpuManager::MultiCoreRunGuestThread() {
     auto& kernel = system.Kernel();
     kernel.CurrentScheduler()->OnThreadStart();
-    auto* thread = kernel.CurrentScheduler()->GetCurrentThread();
+    auto* thread = kernel.CurrentScheduler()->GetSchedulerCurrentThread();
     auto& host_context = thread->GetHostContext();
     host_context->SetRewindPoint(GuestRewindFunction, this);
     MultiCoreRunGuestLoop();
@@ -110,12 +106,10 @@ void CpuManager::MultiCoreRunGuestLoop() {
 
     while (true) {
         auto* physical_core = &kernel.CurrentPhysicalCore();
-        system.EnterDynarmicProfile();
         while (!physical_core->IsInterrupted()) {
             physical_core->Run();
             physical_core = &kernel.CurrentPhysicalCore();
         }
-        system.ExitDynarmicProfile();
         {
             Kernel::KScopedDisableDispatch dd(kernel);
             physical_core->ArmInterface().ClearExclusiveState();
@@ -131,58 +125,6 @@ void CpuManager::MultiCoreRunIdleThread() {
     }
 }
 
-void CpuManager::MultiCoreRunSuspendThread() {
-    auto& kernel = system.Kernel();
-    kernel.CurrentScheduler()->OnThreadStart();
-    while (true) {
-        auto core = kernel.CurrentPhysicalCoreIndex();
-        auto& scheduler = *kernel.CurrentScheduler();
-        Kernel::KThread* current_thread = scheduler.GetCurrentThread();
-        Common::Fiber::YieldTo(current_thread->GetHostContext(), *core_data[core].host_context);
-        ASSERT(scheduler.ContextSwitchPending());
-        ASSERT(core == kernel.CurrentPhysicalCoreIndex());
-        scheduler.RescheduleCurrentCore();
-    }
-}
-
-void CpuManager::MultiCorePause(bool paused) {
-    if (!paused) {
-        bool all_not_barrier = false;
-        while (!all_not_barrier) {
-            all_not_barrier = true;
-            for (const auto& data : core_data) {
-                all_not_barrier &= !data.is_running.load() && data.initialized.load();
-            }
-        }
-        for (auto& data : core_data) {
-            data.enter_barrier->Set();
-        }
-        if (paused_state.load()) {
-            bool all_barrier = false;
-            while (!all_barrier) {
-                all_barrier = true;
-                for (const auto& data : core_data) {
-                    all_barrier &= data.is_paused.load() && data.initialized.load();
-                }
-            }
-            for (auto& data : core_data) {
-                data.exit_barrier->Set();
-            }
-        }
-    } else {
-        /// Wait until all cores are paused.
-        bool all_barrier = false;
-        while (!all_barrier) {
-            all_barrier = true;
-            for (const auto& data : core_data) {
-                all_barrier &= data.is_paused.load() && data.initialized.load();
-            }
-        }
-        /// Don't release the barrier
-    }
-    paused_state = paused;
-}
-
 ///////////////////////////////////////////////////////////////////////////////
 ///                             SingleCore                                   ///
 ///////////////////////////////////////////////////////////////////////////////
@@ -190,7 +132,7 @@ void CpuManager::MultiCorePause(bool paused) {
 void CpuManager::SingleCoreRunGuestThread() {
     auto& kernel = system.Kernel();
     kernel.CurrentScheduler()->OnThreadStart();
-    auto* thread = kernel.CurrentScheduler()->GetCurrentThread();
+    auto* thread = kernel.CurrentScheduler()->GetSchedulerCurrentThread();
     auto& host_context = thread->GetHostContext();
     host_context->SetRewindPoint(GuestRewindFunction, this);
     SingleCoreRunGuestLoop();
@@ -200,12 +142,10 @@ void CpuManager::SingleCoreRunGuestLoop() {
     auto& kernel = system.Kernel();
     while (true) {
         auto* physical_core = &kernel.CurrentPhysicalCore();
-        system.EnterDynarmicProfile();
         if (!physical_core->IsInterrupted()) {
             physical_core->Run();
             physical_core = &kernel.CurrentPhysicalCore();
         }
-        system.ExitDynarmicProfile();
         kernel.SetIsPhantomModeForSingleCore(true);
         system.CoreTiming().Advance();
         kernel.SetIsPhantomModeForSingleCore(false);
@@ -228,25 +168,11 @@ void CpuManager::SingleCoreRunIdleThread() {
     }
 }
 
-void CpuManager::SingleCoreRunSuspendThread() {
-    auto& kernel = system.Kernel();
-    kernel.CurrentScheduler()->OnThreadStart();
-    while (true) {
-        auto core = kernel.GetCurrentHostThreadID();
-        auto& scheduler = *kernel.CurrentScheduler();
-        Kernel::KThread* current_thread = scheduler.GetCurrentThread();
-        Common::Fiber::YieldTo(current_thread->GetHostContext(), *core_data[0].host_context);
-        ASSERT(scheduler.ContextSwitchPending());
-        ASSERT(core == kernel.GetCurrentHostThreadID());
-        scheduler.RescheduleCurrentCore();
-    }
-}
-
 void CpuManager::PreemptSingleCore(bool from_running_enviroment) {
     {
         auto& kernel = system.Kernel();
         auto& scheduler = kernel.Scheduler(current_core);
-        Kernel::KThread* current_thread = scheduler.GetCurrentThread();
+        Kernel::KThread* current_thread = scheduler.GetSchedulerCurrentThread();
         if (idle_count >= 4 || from_running_enviroment) {
             if (!from_running_enviroment) {
                 system.CoreTiming().Idle();
@@ -258,7 +184,7 @@ void CpuManager::PreemptSingleCore(bool from_running_enviroment) {
         }
         current_core.store((current_core + 1) % Core::Hardware::NUM_CPU_CORES);
         system.CoreTiming().ResetTicks();
-        scheduler.Unload(scheduler.GetCurrentThread());
+        scheduler.Unload(scheduler.GetSchedulerCurrentThread());
 
         auto& next_scheduler = kernel.Scheduler(current_core);
         Common::Fiber::YieldTo(current_thread->GetHostContext(), *next_scheduler.ControlContext());
@@ -267,47 +193,21 @@ void CpuManager::PreemptSingleCore(bool from_running_enviroment) {
     // May have changed scheduler
     {
         auto& scheduler = system.Kernel().Scheduler(current_core);
-        scheduler.Reload(scheduler.GetCurrentThread());
-        if (!scheduler.IsIdle()) {
-            idle_count = 0;
-        }
+        scheduler.Reload(scheduler.GetSchedulerCurrentThread());
+        idle_count = 0;
     }
 }
 
-void CpuManager::SingleCorePause(bool paused) {
-    if (!paused) {
-        bool all_not_barrier = false;
-        while (!all_not_barrier) {
-            all_not_barrier = !core_data[0].is_running.load() && core_data[0].initialized.load();
-        }
-        core_data[0].enter_barrier->Set();
-        if (paused_state.load()) {
-            bool all_barrier = false;
-            while (!all_barrier) {
-                all_barrier = core_data[0].is_paused.load() && core_data[0].initialized.load();
-            }
-            core_data[0].exit_barrier->Set();
-        }
-    } else {
-        /// Wait until all cores are paused.
-        bool all_barrier = false;
-        while (!all_barrier) {
-            all_barrier = core_data[0].is_paused.load() && core_data[0].initialized.load();
-        }
-        /// Don't release the barrier
-    }
-    paused_state = paused;
+void CpuManager::ShutdownThread() {
+    auto& kernel = system.Kernel();
+    auto core = is_multicore ? kernel.CurrentPhysicalCoreIndex() : 0;
+    auto* current_thread = kernel.GetCurrentEmuThread();
+
+    Common::Fiber::YieldTo(current_thread->GetHostContext(), *core_data[core].host_context);
+    UNREACHABLE();
 }
 
-void CpuManager::Pause(bool paused) {
-    if (is_multicore) {
-        MultiCorePause(paused);
-    } else {
-        SingleCorePause(paused);
-    }
-}
-
-void CpuManager::RunThread(std::stop_token stop_token, std::size_t core) {
+void CpuManager::RunThread(std::size_t core) {
     /// Initialization
     system.RegisterCoreThread(core);
     std::string name;
@@ -320,45 +220,24 @@ void CpuManager::RunThread(std::stop_token stop_token, std::size_t core) {
     Common::SetCurrentThreadName(name.c_str());
     Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
     auto& data = core_data[core];
-    data.enter_barrier = std::make_unique<Common::Event>();
-    data.exit_barrier = std::make_unique<Common::Event>();
     data.host_context = Common::Fiber::ThreadToFiber();
-    data.is_running = false;
-    data.initialized = true;
-    const bool sc_sync = !is_async_gpu && !is_multicore;
-    bool sc_sync_first_use = sc_sync;
 
     // Cleanup
     SCOPE_EXIT({
         data.host_context->Exit();
-        data.enter_barrier.reset();
-        data.exit_barrier.reset();
-        data.initialized = false;
         MicroProfileOnThreadExit();
     });
 
-    /// Running
-    while (running_mode) {
-        data.is_running = false;
-        data.enter_barrier->Wait();
-        if (sc_sync_first_use) {
-            system.GPU().ObtainContext();
-            sc_sync_first_use = false;
-        }
+    // Running
+    gpu_barrier->Sync();
 
-        // Emulation was stopped
-        if (stop_token.stop_requested()) {
-            return;
-        }
-
-        auto current_thread = system.Kernel().CurrentScheduler()->GetCurrentThread();
-        data.is_running = true;
-        Common::Fiber::YieldTo(data.host_context, *current_thread->GetHostContext());
-        data.is_running = false;
-        data.is_paused = true;
-        data.exit_barrier->Wait();
-        data.is_paused = false;
+    if (!is_async_gpu && !is_multicore) {
+        system.GPU().ObtainContext();
     }
+
+    auto* current_thread = system.Kernel().CurrentScheduler()->GetIdleThread();
+    Kernel::SetCurrentThread(system.Kernel(), current_thread);
+    Common::Fiber::YieldTo(data.host_context, *current_thread->GetHostContext());
 }
 
 } // namespace Core

src/core/cpu_manager.h

@@ -43,15 +43,17 @@ public:
         is_async_gpu = is_async;
     }
 
+    void OnGpuReady() {
+        gpu_barrier->Sync();
+    }
+
     void Initialize();
     void Shutdown();
 
-    void Pause(bool paused);
-
     static std::function<void(void*)> GetGuestThreadStartFunc();
     static std::function<void(void*)> GetIdleThreadStartFunc();
-    static std::function<void(void*)> GetSuspendThreadStartFunc();
-    void* GetStartFuncParamater();
+    static std::function<void(void*)> GetShutdownThreadStartFunc();
+    void* GetStartFuncParameter();
 
     void PreemptSingleCore(bool from_running_enviroment = true);
 
@@ -63,43 +65,34 @@ private:
     static void GuestThreadFunction(void* cpu_manager);
     static void GuestRewindFunction(void* cpu_manager);
     static void IdleThreadFunction(void* cpu_manager);
-    static void SuspendThreadFunction(void* cpu_manager);
+    static void ShutdownThreadFunction(void* cpu_manager);
 
     void MultiCoreRunGuestThread();
     void MultiCoreRunGuestLoop();
     void MultiCoreRunIdleThread();
-    void MultiCoreRunSuspendThread();
-    void MultiCorePause(bool paused);
 
     void SingleCoreRunGuestThread();
     void SingleCoreRunGuestLoop();
     void SingleCoreRunIdleThread();
-    void SingleCoreRunSuspendThread();
-    void SingleCorePause(bool paused);
 
     static void ThreadStart(std::stop_token stop_token, CpuManager& cpu_manager, std::size_t core);
 
-    void RunThread(std::stop_token stop_token, std::size_t core);
+    void ShutdownThread();
+    void RunThread(std::size_t core);
 
     struct CoreData {
         std::shared_ptr<Common::Fiber> host_context;
-        std::unique_ptr<Common::Event> enter_barrier;
-        std::unique_ptr<Common::Event> exit_barrier;
-        std::atomic<bool> is_running;
-        std::atomic<bool> is_paused;
-        std::atomic<bool> initialized;
         std::jthread host_thread;
     };
 
-    std::atomic<bool> running_mode{};
-    std::atomic<bool> paused_state{};
-
+    std::unique_ptr<Common::Barrier> gpu_barrier{};
     std::array<CoreData, Core::Hardware::NUM_CPU_CORES> core_data{};
 
     bool is_async_gpu{};
     bool is_multicore{};
     std::atomic<std::size_t> current_core{};
     std::size_t idle_count{};
+    std::size_t num_cores{};
     static constexpr std::size_t max_cycle_runs = 5;
 
     System& system;

src/core/crypto/key_manager.cpp

@@ -140,7 +140,6 @@ u64 GetSignatureTypeDataSize(SignatureType type) {
         return 0x3C;
     }
     UNREACHABLE();
-    return 0;
 }
 
 u64 GetSignatureTypePaddingSize(SignatureType type) {
@@ -155,7 +154,6 @@ u64 GetSignatureTypePaddingSize(SignatureType type) {
         return 0x40;
     }
     UNREACHABLE();
-    return 0;
 }
 
 SignatureType Ticket::GetSignatureType() const {

src/core/debugger/debugger.cpp

@@ -42,6 +42,18 @@ static std::span<const u8> ReceiveInto(Readable& r, Buffer& buffer) {
     return received_data;
 }
 
+enum class SignalType {
+    Stopped,
+    Watchpoint,
+    ShuttingDown,
+};
+
+struct SignalInfo {
+    SignalType type;
+    Kernel::KThread* thread;
+    const Kernel::DebugWatchpoint* watchpoint;
+};
+
 namespace Core {
 
 class DebuggerImpl : public DebuggerBackend {
@@ -56,17 +68,23 @@ public:
         ShutdownServer();
     }
 
-    bool NotifyThreadStopped(Kernel::KThread* thread) {
-        std::scoped_lock lk{connection_lock};
+    bool SignalDebugger(SignalInfo signal_info) {
+        {
+            std::scoped_lock lk{connection_lock};
 
-        if (stopped) {
-            // Do not notify the debugger about another event.
-            // It should be ignored.
-            return false;
+            if (stopped) {
+                // Do not notify the debugger about another event.
+                // It should be ignored.
+                return false;
+            }
+
+            // Set up the state.
+            stopped = true;
+            info = signal_info;
         }
-        stopped = true;
 
-        boost::asio::write(signal_pipe, boost::asio::buffer(&thread, sizeof(thread)));
+        // Write a single byte into the pipe to wake up the debug interface.
+        boost::asio::write(signal_pipe, boost::asio::buffer(&stopped, sizeof(stopped)));
         return true;
     }
 
@@ -96,7 +114,7 @@ private:
         connection_thread = std::jthread([&, port](std::stop_token stop_token) {
             try {
                 // Initialize the listening socket and accept a new client.
-                tcp::endpoint endpoint{boost::asio::ip::address_v4::loopback(), port};
+                tcp::endpoint endpoint{boost::asio::ip::address_v4::any(), port};
                 tcp::acceptor acceptor{io_context, endpoint};
 
                 acceptor.async_accept(client_socket, [](const auto&) {});
@@ -124,12 +142,9 @@ private:
         Common::SetCurrentThreadName("yuzu:Debugger");
 
         // Set up the client signals for new data.
-        AsyncReceiveInto(signal_pipe, active_thread, [&](auto d) { PipeData(d); });
+        AsyncReceiveInto(signal_pipe, pipe_data, [&](auto d) { PipeData(d); });
         AsyncReceiveInto(client_socket, client_data, [&](auto d) { ClientData(d); });
 
-        // Stop the emulated CPU.
-        AllCoreStop();
-
         // Set the active thread.
         UpdateActiveThread();
 
@@ -142,9 +157,33 @@ private:
     }
 
     void PipeData(std::span<const u8> data) {
-        AllCoreStop();
-        UpdateActiveThread();
-        frontend->Stopped(active_thread);
+        switch (info.type) {
+        case SignalType::Stopped:
+        case SignalType::Watchpoint:
+            // Stop emulation.
+            PauseEmulation();
+
+            // Notify the client.
+            active_thread = info.thread;
+            UpdateActiveThread();
+
+            if (info.type == SignalType::Watchpoint) {
+                frontend->Watchpoint(active_thread, *info.watchpoint);
+            } else {
+                frontend->Stopped(active_thread);
+            }
+
+            break;
+        case SignalType::ShuttingDown:
+            frontend->ShuttingDown();
+
+            // Wait for emulation to shut down gracefully now.
+            signal_pipe.close();
+            client_socket.shutdown(boost::asio::socket_base::shutdown_both);
+            LOG_INFO(Debug_GDBStub, "Shut down server");
+
+            break;
+        }
     }
 
     void ClientData(std::span<const u8> data) {
@@ -156,32 +195,29 @@ private:
                     std::scoped_lock lk{connection_lock};
                     stopped = true;
                 }
-                AllCoreStop();
+                PauseEmulation();
                 UpdateActiveThread();
                 frontend->Stopped(active_thread);
                 break;
             }
             case DebuggerAction::Continue:
-                active_thread->SetStepState(Kernel::StepState::NotStepping);
-                ResumeInactiveThreads();
-                AllCoreResume();
+                MarkResumed([&] { ResumeEmulation(); });
                 break;
             case DebuggerAction::StepThreadUnlocked:
-                active_thread->SetStepState(Kernel::StepState::StepPending);
-                ResumeInactiveThreads();
-                AllCoreResume();
+                MarkResumed([&] {
+                    active_thread->SetStepState(Kernel::StepState::StepPending);
+                    active_thread->Resume(Kernel::SuspendType::Debug);
+                    ResumeEmulation(active_thread);
+                });
                 break;
-            case DebuggerAction::StepThreadLocked:
-                active_thread->SetStepState(Kernel::StepState::StepPending);
-                SuspendInactiveThreads();
-                AllCoreResume();
+            case DebuggerAction::StepThreadLocked: {
+                MarkResumed([&] {
+                    active_thread->SetStepState(Kernel::StepState::StepPending);
+                    active_thread->Resume(Kernel::SuspendType::Debug);
+                });
                 break;
+            }
             case DebuggerAction::ShutdownEmulation: {
-                // Suspend all threads and release any locks held
-                active_thread->RequestSuspend(Kernel::SuspendType::Debug);
-                SuspendInactiveThreads();
-                AllCoreResume();
-
                 // Spawn another thread that will exit after shutdown,
                 // to avoid a deadlock
                 Core::System* system_ref{&system};
@@ -193,33 +229,33 @@ private:
         }
     }
 
-    void AllCoreStop() {
-        if (!suspend) {
-            suspend = system.StallCPU();
+    void PauseEmulation() {
+        // Put all threads to sleep on next scheduler round.
+        for (auto* thread : ThreadList()) {
+            thread->RequestSuspend(Kernel::SuspendType::Debug);
+        }
+
+        // Signal an interrupt so that scheduler will fire.
+        system.Kernel().InterruptAllPhysicalCores();
+    }
+
+    void ResumeEmulation(Kernel::KThread* except = nullptr) {
+        // Wake up all threads.
+        for (auto* thread : ThreadList()) {
+            if (thread == except) {
+                continue;
+            }
+
+            thread->SetStepState(Kernel::StepState::NotStepping);
+            thread->Resume(Kernel::SuspendType::Debug);
         }
     }
 
-    void AllCoreResume() {
+    template <typename Callback>
+    void MarkResumed(Callback&& cb) {
+        std::scoped_lock lk{connection_lock};
         stopped = false;
-        system.UnstallCPU();
-        suspend.reset();
-    }
-
-    void SuspendInactiveThreads() {
-        for (auto* thread : ThreadList()) {
-            if (thread != active_thread) {
-                thread->RequestSuspend(Kernel::SuspendType::Debug);
-            }
-        }
-    }
-
-    void ResumeInactiveThreads() {
-        for (auto* thread : ThreadList()) {
-            if (thread != active_thread) {
-                thread->Resume(Kernel::SuspendType::Debug);
-                thread->SetStepState(Kernel::StepState::NotStepping);
-            }
-        }
+        cb();
     }
 
     void UpdateActiveThread() {
@@ -227,8 +263,6 @@ private:
         if (std::find(threads.begin(), threads.end(), active_thread) == threads.end()) {
             active_thread = threads[0];
         }
-        active_thread->Resume(Kernel::SuspendType::Debug);
-        active_thread->SetStepState(Kernel::StepState::NotStepping);
     }
 
     const std::vector<Kernel::KThread*>& ThreadList() {
@@ -244,9 +278,10 @@ private:
     boost::asio::io_context io_context;
     boost::process::async_pipe signal_pipe;
     boost::asio::ip::tcp::socket client_socket;
-    std::optional<std::unique_lock<std::mutex>> suspend;
 
+    SignalInfo info;
     Kernel::KThread* active_thread;
+    bool pipe_data;
     bool stopped;
 
     std::array<u8, 4096> client_data;
@@ -263,7 +298,18 @@ Debugger::Debugger(Core::System& system, u16 port) {
 Debugger::~Debugger() = default;
 
 bool Debugger::NotifyThreadStopped(Kernel::KThread* thread) {
-    return impl && impl->NotifyThreadStopped(thread);
+    return impl && impl->SignalDebugger(SignalInfo{SignalType::Stopped, thread, nullptr});
+}
+
+bool Debugger::NotifyThreadWatchpoint(Kernel::KThread* thread,
+                                      const Kernel::DebugWatchpoint& watch) {
+    return impl && impl->SignalDebugger(SignalInfo{SignalType::Watchpoint, thread, &watch});
+}
+
+void Debugger::NotifyShutdown() {
+    if (impl) {
+        impl->SignalDebugger(SignalInfo{SignalType::ShuttingDown, nullptr, nullptr});
+    }
 }
 
 } // namespace Core

src/core/debugger/debugger.h

@@ -9,7 +9,8 @@
 
 namespace Kernel {
 class KThread;
-}
+struct DebugWatchpoint;
+} // namespace Kernel
 
 namespace Core {
 class System;
@@ -35,6 +36,16 @@ public:
      */
     bool NotifyThreadStopped(Kernel::KThread* thread);
 
+    /**
+     * Notify the debugger that a shutdown is being performed now and disconnect.
+     */
+    void NotifyShutdown();
+
+    /*
+     * Notify the debugger that the given thread has stopped due to hitting a watchpoint.
+     */
+    bool NotifyThreadWatchpoint(Kernel::KThread* thread, const Kernel::DebugWatchpoint& watch);
+
 private:
     std::unique_ptr<DebuggerImpl> impl;
 };

src/core/debugger/debugger_interface.h

@@ -11,7 +11,8 @@
 
 namespace Kernel {
 class KThread;
-}
+struct DebugWatchpoint;
+} // namespace Kernel
 
 namespace Core {
 
@@ -66,6 +67,16 @@ public:
      */
     virtual void Stopped(Kernel::KThread* thread) = 0;
 
+    /**
+     * Called when emulation is shutting down.
+     */
+    virtual void ShuttingDown() = 0;
+
+    /*
+     * Called when emulation has stopped on a watchpoint.
+     */
+    virtual void Watchpoint(Kernel::KThread* thread, const Kernel::DebugWatchpoint& watch) = 0;
+
     /**
      * Called when new data is asynchronously received on the client socket.
      * A list of actions to perform is returned.

src/core/debugger/gdbstub.cpp

@@ -106,10 +106,29 @@ GDBStub::~GDBStub() = default;
 
 void GDBStub::Connected() {}
 
+void GDBStub::ShuttingDown() {}
+
 void GDBStub::Stopped(Kernel::KThread* thread) {
     SendReply(arch->ThreadStatus(thread, GDB_STUB_SIGTRAP));
 }
 
+void GDBStub::Watchpoint(Kernel::KThread* thread, const Kernel::DebugWatchpoint& watch) {
+    const auto status{arch->ThreadStatus(thread, GDB_STUB_SIGTRAP)};
+
+    switch (watch.type) {
+    case Kernel::DebugWatchpointType::Read:
+        SendReply(fmt::format("{}rwatch:{:x};", status, watch.start_address));
+        break;
+    case Kernel::DebugWatchpointType::Write:
+        SendReply(fmt::format("{}watch:{:x};", status, watch.start_address));
+        break;
+    case Kernel::DebugWatchpointType::ReadOrWrite:
+    default:
+        SendReply(fmt::format("{}awatch:{:x};", status, watch.start_address));
+        break;
+    }
+}
+
 std::vector<DebuggerAction> GDBStub::ClientData(std::span<const u8> data) {
     std::vector<DebuggerAction> actions;
     current_command.insert(current_command.end(), data.begin(), data.end());
@@ -276,44 +295,124 @@ void GDBStub::ExecuteCommand(std::string_view packet, std::vector<DebuggerAction
     case 'c':
         actions.push_back(DebuggerAction::Continue);
         break;
-    case 'Z': {
-        const auto addr_sep{std::find(command.begin(), command.end(), ',') - command.begin() + 1};
-        const size_t addr{static_cast<size_t>(strtoll(command.data() + addr_sep, nullptr, 16))};
-
-        if (system.Memory().IsValidVirtualAddress(addr)) {
-            replaced_instructions[addr] = system.Memory().Read32(addr);
-            system.Memory().Write32(addr, arch->BreakpointInstruction());
-            system.InvalidateCpuInstructionCacheRange(addr, sizeof(u32));
-
-            SendReply(GDB_STUB_REPLY_OK);
-        } else {
-            SendReply(GDB_STUB_REPLY_ERR);
-        }
+    case 'Z':
+        HandleBreakpointInsert(command);
         break;
-    }
-    case 'z': {
-        const auto addr_sep{std::find(command.begin(), command.end(), ',') - command.begin() + 1};
-        const size_t addr{static_cast<size_t>(strtoll(command.data() + addr_sep, nullptr, 16))};
-
-        const auto orig_insn{replaced_instructions.find(addr)};
-        if (system.Memory().IsValidVirtualAddress(addr) &&
-            orig_insn != replaced_instructions.end()) {
-            system.Memory().Write32(addr, orig_insn->second);
-            system.InvalidateCpuInstructionCacheRange(addr, sizeof(u32));
-            replaced_instructions.erase(addr);
-
-            SendReply(GDB_STUB_REPLY_OK);
-        } else {
-            SendReply(GDB_STUB_REPLY_ERR);
-        }
+    case 'z':
+        HandleBreakpointRemove(command);
         break;
-    }
     default:
         SendReply(GDB_STUB_REPLY_EMPTY);
         break;
     }
 }
 
+enum class BreakpointType {
+    Software = 0,
+    Hardware = 1,
+    WriteWatch = 2,
+    ReadWatch = 3,
+    AccessWatch = 4,
+};
+
+void GDBStub::HandleBreakpointInsert(std::string_view command) {
+    const auto type{static_cast<BreakpointType>(strtoll(command.data(), nullptr, 16))};
+    const auto addr_sep{std::find(command.begin(), command.end(), ',') - command.begin() + 1};
+    const auto size_sep{std::find(command.begin() + addr_sep, command.end(), ',') -
+                        command.begin() + 1};
+    const size_t addr{static_cast<size_t>(strtoll(command.data() + addr_sep, nullptr, 16))};
+    const size_t size{static_cast<size_t>(strtoll(command.data() + size_sep, nullptr, 16))};
+
+    if (!system.Memory().IsValidVirtualAddressRange(addr, size)) {
+        SendReply(GDB_STUB_REPLY_ERR);
+        return;
+    }
+
+    bool success{};
+
+    switch (type) {
+    case BreakpointType::Software:
+        replaced_instructions[addr] = system.Memory().Read32(addr);
+        system.Memory().Write32(addr, arch->BreakpointInstruction());
+        system.InvalidateCpuInstructionCacheRange(addr, sizeof(u32));
+        success = true;
+        break;
+    case BreakpointType::WriteWatch:
+        success = system.CurrentProcess()->InsertWatchpoint(system, addr, size,
+                                                            Kernel::DebugWatchpointType::Write);
+        break;
+    case BreakpointType::ReadWatch:
+        success = system.CurrentProcess()->InsertWatchpoint(system, addr, size,
+                                                            Kernel::DebugWatchpointType::Read);
+        break;
+    case BreakpointType::AccessWatch:
+        success = system.CurrentProcess()->InsertWatchpoint(
+            system, addr, size, Kernel::DebugWatchpointType::ReadOrWrite);
+        break;
+    case BreakpointType::Hardware:
+    default:
+        SendReply(GDB_STUB_REPLY_EMPTY);
+        return;
+    }
+
+    if (success) {
+        SendReply(GDB_STUB_REPLY_OK);
+    } else {
+        SendReply(GDB_STUB_REPLY_ERR);
+    }
+}
+
+void GDBStub::HandleBreakpointRemove(std::string_view command) {
+    const auto type{static_cast<BreakpointType>(strtoll(command.data(), nullptr, 16))};
+    const auto addr_sep{std::find(command.begin(), command.end(), ',') - command.begin() + 1};
+    const auto size_sep{std::find(command.begin() + addr_sep, command.end(), ',') -
+                        command.begin() + 1};
+    const size_t addr{static_cast<size_t>(strtoll(command.data() + addr_sep, nullptr, 16))};
+    const size_t size{static_cast<size_t>(strtoll(command.data() + size_sep, nullptr, 16))};
+
+    if (!system.Memory().IsValidVirtualAddressRange(addr, size)) {
+        SendReply(GDB_STUB_REPLY_ERR);
+        return;
+    }
+
+    bool success{};
+
+    switch (type) {
+    case BreakpointType::Software: {
+        const auto orig_insn{replaced_instructions.find(addr)};
+        if (orig_insn != replaced_instructions.end()) {
+            system.Memory().Write32(addr, orig_insn->second);
+            system.InvalidateCpuInstructionCacheRange(addr, sizeof(u32));
+            replaced_instructions.erase(addr);
+            success = true;
+        }
+        break;
+    }
+    case BreakpointType::WriteWatch:
+        success = system.CurrentProcess()->RemoveWatchpoint(system, addr, size,
+                                                            Kernel::DebugWatchpointType::Write);
+        break;
+    case BreakpointType::ReadWatch:
+        success = system.CurrentProcess()->RemoveWatchpoint(system, addr, size,
+                                                            Kernel::DebugWatchpointType::Read);
+        break;
+    case BreakpointType::AccessWatch:
+        success = system.CurrentProcess()->RemoveWatchpoint(
+            system, addr, size, Kernel::DebugWatchpointType::ReadOrWrite);
+        break;
+    case BreakpointType::Hardware:
+    default:
+        SendReply(GDB_STUB_REPLY_EMPTY);
+        return;
+    }
+
+    if (success) {
+        SendReply(GDB_STUB_REPLY_OK);
+    } else {
+        SendReply(GDB_STUB_REPLY_ERR);
+    }
+}
+
 // Structure offsets are from Atmosphere
 // See osdbg_thread_local_region.os.horizon.hpp and osdbg_thread_type.os.horizon.hpp
 

src/core/debugger/gdbstub.h

@@ -23,6 +23,8 @@ public:
 
     void Connected() override;
     void Stopped(Kernel::KThread* thread) override;
+    void ShuttingDown() override;
+    void Watchpoint(Kernel::KThread* thread, const Kernel::DebugWatchpoint& watch) override;
     std::vector<DebuggerAction> ClientData(std::span<const u8> data) override;
 
 private:
@@ -30,6 +32,8 @@ private:
     void ExecuteCommand(std::string_view packet, std::vector<DebuggerAction>& actions);
     void HandleVCont(std::string_view command, std::vector<DebuggerAction>& actions);
     void HandleQuery(std::string_view command);
+    void HandleBreakpointInsert(std::string_view command);
+    void HandleBreakpointRemove(std::string_view command);
     std::vector<char>::const_iterator CommandEnd() const;
     std::optional<std::string> DetachCommand();
     Kernel::KThread* GetThreadByID(u64 thread_id);

src/core/debugger/gdbstub_arch.h

@@ -15,6 +15,7 @@ namespace Core {
 
 class GDBStubArch {
 public:
+    virtual ~GDBStubArch() = default;
     virtual std::string GetTargetXML() const = 0;
     virtual std::string RegRead(const Kernel::KThread* thread, size_t id) const = 0;
     virtual void RegWrite(Kernel::KThread* thread, size_t id, std::string_view value) const = 0;

src/core/file_sys/content_archive.cpp

@@ -419,7 +419,7 @@ std::optional<Core::Crypto::Key128> NCA::GetKeyAreaKey(NCASectionCryptoType type
         Core::Crypto::Mode::ECB);
     cipher.Transcode(key_area.data(), key_area.size(), key_area.data(), Core::Crypto::Op::Decrypt);
 
-    Core::Crypto::Key128 out;
+    Core::Crypto::Key128 out{};
     if (type == NCASectionCryptoType::XTS) {
         std::copy(key_area.begin(), key_area.begin() + 0x10, out.begin());
     } else if (type == NCASectionCryptoType::CTR || type == NCASectionCryptoType::BKTR) {

src/core/file_sys/nca_patch.cpp

@@ -50,7 +50,7 @@ std::pair<std::size_t, std::size_t> SearchBucketEntry(u64 offset, const BlockTyp
             low = mid + 1;
         }
     }
-    UNREACHABLE_MSG("Offset could not be found in BKTR block.");
+    ASSERT_MSG(false, "Offset could not be found in BKTR block.");
     return {0, 0};
 }
 } // Anonymous namespace

src/core/file_sys/registered_cache.cpp

@@ -108,7 +108,7 @@ ContentRecordType GetCRTypeFromNCAType(NCAContentType type) {
         // TODO(DarkLordZach): Peek at NCA contents to differentiate Manual and Legal.
         return ContentRecordType::HtmlDocument;
     default:
-        UNREACHABLE_MSG("Invalid NCAContentType={:02X}", type);
+        ASSERT_MSG(false, "Invalid NCAContentType={:02X}", type);
         return ContentRecordType{};
     }
 }

src/core/file_sys/vfs_real.cpp

@@ -144,7 +144,7 @@ VirtualFile RealVfsFilesystem::MoveFile(std::string_view old_path_, std::string_
             LOG_ERROR(Service_FS, "Failed to open path {} in order to re-cache it", new_path);
         }
     } else {
-        UNREACHABLE();
+        ASSERT(false);
         return nullptr;
     }
 

src/core/frontend/applets/controller.cpp

@@ -65,7 +65,7 @@ void DefaultControllerApplet::ReconfigureControllers(std::function<void()> callb
             controller->SetNpadStyleIndex(Core::HID::NpadStyleIndex::Handheld);
             controller->Connect(true);
         } else {
-            UNREACHABLE_MSG("Unable to add a new controller based on the given parameters!");
+            ASSERT_MSG(false, "Unable to add a new controller based on the given parameters!");
         }
     }
 

src/core/hardware_properties.h

@@ -25,6 +25,9 @@ constexpr std::array<s32, Common::BitSize<u64>()> VirtualToPhysicalCoreMap{
     0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3,
 };
 
+// Cortex-A57 supports 4 memory watchpoints
+constexpr u64 NUM_WATCHPOINTS = 4;
+
 } // namespace Hardware
 
 } // namespace Core

src/core/hid/hid_core.cpp

@@ -48,7 +48,7 @@ EmulatedController* HIDCore::GetEmulatedController(NpadIdType npad_id_type) {
         return handheld.get();
     case NpadIdType::Invalid:
     default:
-        UNREACHABLE_MSG("Invalid NpadIdType={}", npad_id_type);
+        ASSERT_MSG(false, "Invalid NpadIdType={}", npad_id_type);
         return nullptr;
     }
 }
@@ -77,7 +77,7 @@ const EmulatedController* HIDCore::GetEmulatedController(NpadIdType npad_id_type
         return handheld.get();
     case NpadIdType::Invalid:
     default:
-        UNREACHABLE_MSG("Invalid NpadIdType={}", npad_id_type);
+        ASSERT_MSG(false, "Invalid NpadIdType={}", npad_id_type);
         return nullptr;
     }
 }

src/core/hle/kernel/hle_ipc.h

@@ -141,7 +141,7 @@ public:
         if (index < DomainHandlerCount()) {
             domain_handlers[index] = nullptr;
         } else {
-            UNREACHABLE_MSG("Unexpected handler index {}", index);
+            ASSERT_MSG(false, "Unexpected handler index {}", index);
         }
     }
 

src/core/hle/kernel/init/init_slab_setup.cpp

@@ -244,7 +244,7 @@ void InitializeSlabHeaps(Core::System& system, KMemoryLayout& memory_layout) {
             FOREACH_SLAB_TYPE(INITIALIZE_SLAB_HEAP)
             // If we somehow get an invalid type, abort.
         default:
-            UNREACHABLE_MSG("Unknown slab type: {}", slab_types[i]);
+            ASSERT_MSG(false, "Unknown slab type: {}", slab_types[i]);
         }
 
         // If we've hit the end of a gap, free it.

src/core/hle/kernel/k_address_arbiter.cpp

@@ -234,7 +234,7 @@ ResultCode KAddressArbiter::SignalAndModifyByWaitingCountIfEqual(VAddr addr, s32
 
 ResultCode KAddressArbiter::WaitIfLessThan(VAddr addr, s32 value, bool decrement, s64 timeout) {
     // Prepare to wait.
-    KThread* cur_thread = kernel.CurrentScheduler()->GetCurrentThread();
+    KThread* cur_thread = GetCurrentThreadPointer(kernel);
     ThreadQueueImplForKAddressArbiter wait_queue(kernel, std::addressof(thread_tree));
 
     {
@@ -287,7 +287,7 @@ ResultCode KAddressArbiter::WaitIfLessThan(VAddr addr, s32 value, bool decrement
 
 ResultCode KAddressArbiter::WaitIfEqual(VAddr addr, s32 value, s64 timeout) {
     // Prepare to wait.
-    KThread* cur_thread = kernel.CurrentScheduler()->GetCurrentThread();
+    KThread* cur_thread = GetCurrentThreadPointer(kernel);
     ThreadQueueImplForKAddressArbiter wait_queue(kernel, std::addressof(thread_tree));
 
     {

src/core/hle/kernel/k_address_arbiter.h

@@ -35,7 +35,7 @@ public:
         case Svc::SignalType::SignalAndModifyByWaitingCountIfEqual:
             return SignalAndModifyByWaitingCountIfEqual(addr, value, count);
         }
-        UNREACHABLE();
+        ASSERT(false);
         return ResultUnknown;
     }
 
@@ -49,7 +49,7 @@ public:
         case Svc::ArbitrationType::WaitIfEqual:
             return WaitIfEqual(addr, value, timeout);
         }
-        UNREACHABLE();
+        ASSERT(false);
         return ResultUnknown;
     }
 

src/core/hle/kernel/k_address_space_info.cpp

@@ -84,7 +84,7 @@ u64 KAddressSpaceInfo::GetAddressSpaceStart(std::size_t width, Type type) {
         ASSERT(IsAllowedIndexForAddress(AddressSpaceIndices39Bit[index]));
         return AddressSpaceInfos[AddressSpaceIndices39Bit[index]].address;
     }
-    UNREACHABLE();
+    ASSERT(false);
     return 0;
 }
 
@@ -101,7 +101,7 @@ std::size_t KAddressSpaceInfo::GetAddressSpaceSize(std::size_t width, Type type)
         ASSERT(IsAllowed39BitType(type));
         return AddressSpaceInfos[AddressSpaceIndices39Bit[index]].size;
     }
-    UNREACHABLE();
+    ASSERT(false);
     return 0;
 }
 

src/core/hle/kernel/k_auto_object.h

@@ -18,7 +18,7 @@ namespace Kernel {
 class KernelCore;
 class KProcess;
 
-#define KERNEL_AUTOOBJECT_TRAITS(CLASS, BASE_CLASS)                                                \
+#define KERNEL_AUTOOBJECT_TRAITS_IMPL(CLASS, BASE_CLASS, ATTRIBUTE)                                \
                                                                                                    \
 private:                                                                                           \
     friend class ::Kernel::KClassTokenGenerator;                                                   \
@@ -40,16 +40,19 @@ public:
     static constexpr const char* GetStaticTypeName() {                                             \
         return TypeName;                                                                           \
     }                                                                                              \
-    virtual TypeObj GetTypeObj() const {                                                           \
+    virtual TypeObj GetTypeObj() ATTRIBUTE {                                                       \
         return GetStaticTypeObj();                                                                 \
     }                                                                                              \
-    virtual const char* GetTypeName() const {                                                      \
+    virtual const char* GetTypeName() ATTRIBUTE {                                                  \
         return GetStaticTypeName();                                                                \
     }                                                                                              \
                                                                                                    \
 private:                                                                                           \
     constexpr bool operator!=(const TypeObj& rhs)
 
+#define KERNEL_AUTOOBJECT_TRAITS(CLASS, BASE_CLASS)                                                \
+    KERNEL_AUTOOBJECT_TRAITS_IMPL(CLASS, BASE_CLASS, const override)
+
 class KAutoObject {
 protected:
     class TypeObj {
@@ -82,7 +85,7 @@ protected:
     };
 
 private:
-    KERNEL_AUTOOBJECT_TRAITS(KAutoObject, KAutoObject);
+    KERNEL_AUTOOBJECT_TRAITS_IMPL(KAutoObject, KAutoObject, const);
 
 public:
     explicit KAutoObject(KernelCore& kernel_) : kernel(kernel_) {

src/core/hle/kernel/k_class_token.h

@@ -49,6 +49,7 @@ private:
                 }
             }
         }
+        UNREACHABLE();
     }();
 
     template <typename T>

src/core/hle/kernel/k_code_memory.cpp

@@ -27,23 +27,18 @@ ResultCode KCodeMemory::Initialize(Core::DeviceMemory& device_memory, VAddr addr
     auto& page_table = m_owner->PageTable();
 
     // Construct the page group.
-    m_page_group =
-        KPageLinkedList(page_table.GetPhysicalAddr(addr), Common::DivideUp(size, PageSize));
+    m_page_group = {};
 
     // Lock the memory.
-    R_TRY(page_table.LockForCodeMemory(addr, size))
+    R_TRY(page_table.LockForCodeMemory(&m_page_group, addr, size))
 
     // Clear the memory.
-    //
-    // FIXME: this ends up clobbering address ranges outside the scope of the mapping within
-    // guest memory, and is not specifically required if the guest program is correctly
-    // written, so disable until this is further investigated.
-    //
-    // for (const auto& block : m_page_group.Nodes()) {
-    //     std::memset(device_memory.GetPointer(block.GetAddress()), 0xFF, block.GetSize());
-    // }
+    for (const auto& block : m_page_group.Nodes()) {
+        std::memset(device_memory.GetPointer(block.GetAddress()), 0xFF, block.GetSize());
+    }
 
     // Set remaining tracking members.
+    m_owner->Open();
     m_address = addr;
     m_is_initialized = true;
     m_is_owner_mapped = false;
@@ -57,8 +52,14 @@ void KCodeMemory::Finalize() {
     // Unlock.
     if (!m_is_mapped && !m_is_owner_mapped) {
         const size_t size = m_page_group.GetNumPages() * PageSize;
-        m_owner->PageTable().UnlockForCodeMemory(m_address, size);
+        m_owner->PageTable().UnlockForCodeMemory(m_address, size, m_page_group);
     }
+
+    // Close the page group.
+    m_page_group = {};
+
+    // Close our reference to our owner.
+    m_owner->Close();
 }
 
 ResultCode KCodeMemory::Map(VAddr address, size_t size) {
@@ -118,7 +119,8 @@ ResultCode KCodeMemory::MapToOwner(VAddr address, size_t size, Svc::MemoryPermis
         k_perm = KMemoryPermission::UserReadExecute;
         break;
     default:
-        break;
+        // Already validated by ControlCodeMemory svc
+        UNREACHABLE();
     }
 
     // Map the memory.

src/core/hle/kernel/k_condition_variable.cpp

@@ -106,7 +106,7 @@ KConditionVariable::KConditionVariable(Core::System& system_)
 KConditionVariable::~KConditionVariable() = default;
 
 ResultCode KConditionVariable::SignalToAddress(VAddr addr) {
-    KThread* owner_thread = kernel.CurrentScheduler()->GetCurrentThread();
+    KThread* owner_thread = GetCurrentThreadPointer(kernel);
 
     // Signal the address.
     {
@@ -147,7 +147,7 @@ ResultCode KConditionVariable::SignalToAddress(VAddr addr) {
 }
 
 ResultCode KConditionVariable::WaitForAddress(Handle handle, VAddr addr, u32 value) {
-    KThread* cur_thread = kernel.CurrentScheduler()->GetCurrentThread();
+    KThread* cur_thread = GetCurrentThreadPointer(kernel);
     ThreadQueueImplForKConditionVariableWaitForAddress wait_queue(kernel);
 
     // Wait for the address.

src/core/hle/kernel/k_interrupt_manager.cpp

@@ -15,8 +15,7 @@ void HandleInterrupt(KernelCore& kernel, s32 core_id) {
         return;
     }
 
-    auto& scheduler = kernel.Scheduler(core_id);
-    auto& current_thread = *scheduler.GetCurrentThread();
+    auto& current_thread = GetCurrentThread(kernel);
 
     // If the user disable count is set, we may need to pin the current thread.
     if (current_thread.GetUserDisableCount() && !process->GetPinnedThread(core_id)) {
@@ -26,7 +25,7 @@ void HandleInterrupt(KernelCore& kernel, s32 core_id) {
         process->PinCurrentThread(core_id);
 
         // Set the interrupt flag for the thread.
-        scheduler.GetCurrentThread()->SetInterruptFlag();
+        GetCurrentThread(kernel).SetInterruptFlag();
     }
 }
 

src/core/hle/kernel/k_memory_manager.cpp

@@ -29,7 +29,7 @@ constexpr KMemoryManager::Pool GetPoolFromMemoryRegionType(u32 type) {
     } else if ((type | KMemoryRegionType_DramSystemNonSecurePool) == type) {
         return KMemoryManager::Pool::SystemNonSecure;
     } else {
-        UNREACHABLE_MSG("InvalidMemoryRegionType for conversion to Pool");
+        ASSERT_MSG(false, "InvalidMemoryRegionType for conversion to Pool");
         return {};
     }
 }

src/core/hle/kernel/k_page_table.cpp

@@ -35,7 +35,7 @@ constexpr std::size_t GetAddressSpaceWidthFromType(FileSys::ProgramAddressSpaceT
     case FileSys::ProgramAddressSpaceType::Is39Bit:
         return 39;
     default:
-        UNREACHABLE();
+        ASSERT(false);
         return {};
     }
 }
@@ -65,7 +65,6 @@ ResultCode KPageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_
     std::size_t alias_region_size{GetSpaceSize(KAddressSpaceInfo::Type::Alias)};
     std::size_t heap_region_size{GetSpaceSize(KAddressSpaceInfo::Type::Heap)};
 
-    ASSERT(start <= code_addr);
     ASSERT(code_addr < code_addr + code_size);
     ASSERT(code_addr + code_size - 1 <= end - 1);
 
@@ -128,7 +127,7 @@ ResultCode KPageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_
     const std::size_t needed_size{
         (alias_region_size + heap_region_size + stack_region_size + kernel_map_region_size)};
     if (alloc_size < needed_size) {
-        UNREACHABLE();
+        ASSERT(false);
         return ResultOutOfMemory;
     }
 
@@ -542,6 +541,95 @@ ResultCode KPageTable::MakePageGroup(KPageLinkedList& pg, VAddr addr, size_t num
     return ResultSuccess;
 }
 
+bool KPageTable::IsValidPageGroup(const KPageLinkedList& pg_ll, VAddr addr, size_t num_pages) {
+    ASSERT(this->IsLockedByCurrentThread());
+
+    const size_t size = num_pages * PageSize;
+    const auto& pg = pg_ll.Nodes();
+    const auto& memory_layout = system.Kernel().MemoryLayout();
+
+    // Empty groups are necessarily invalid.
+    if (pg.empty()) {
+        return false;
+    }
+
+    // We're going to validate that the group we'd expect is the group we see.
+    auto cur_it = pg.begin();
+    PAddr cur_block_address = cur_it->GetAddress();
+    size_t cur_block_pages = cur_it->GetNumPages();
+
+    auto UpdateCurrentIterator = [&]() {
+        if (cur_block_pages == 0) {
+            if ((++cur_it) == pg.end()) {
+                return false;
+            }
+
+            cur_block_address = cur_it->GetAddress();
+            cur_block_pages = cur_it->GetNumPages();
+        }
+        return true;
+    };
+
+    // Begin traversal.
+    Common::PageTable::TraversalContext context;
+    Common::PageTable::TraversalEntry next_entry;
+    if (!page_table_impl.BeginTraversal(next_entry, context, addr)) {
+        return false;
+    }
+
+    // Prepare tracking variables.
+    PAddr cur_addr = next_entry.phys_addr;
+    size_t cur_size = next_entry.block_size - (cur_addr & (next_entry.block_size - 1));
+    size_t tot_size = cur_size;
+
+    // Iterate, comparing expected to actual.
+    while (tot_size < size) {
+        if (!page_table_impl.ContinueTraversal(next_entry, context)) {
+            return false;
+        }
+
+        if (next_entry.phys_addr != (cur_addr + cur_size)) {
+            const size_t cur_pages = cur_size / PageSize;
+
+            if (!IsHeapPhysicalAddress(memory_layout, cur_addr)) {
+                return false;
+            }
+
+            if (!UpdateCurrentIterator()) {
+                return false;
+            }
+
+            if (cur_block_address != cur_addr || cur_block_pages < cur_pages) {
+                return false;
+            }
+
+            cur_block_address += cur_size;
+            cur_block_pages -= cur_pages;
+            cur_addr = next_entry.phys_addr;
+            cur_size = next_entry.block_size;
+        } else {
+            cur_size += next_entry.block_size;
+        }
+
+        tot_size += next_entry.block_size;
+    }
+
+    // Ensure we compare the right amount for the last block.
+    if (tot_size > size) {
+        cur_size -= (tot_size - size);
+    }
+
+    if (!IsHeapPhysicalAddress(memory_layout, cur_addr)) {
+        return false;
+    }
+
+    if (!UpdateCurrentIterator()) {
+        return false;
+    }
+
+    return cur_block_address == cur_addr && cur_block_pages == (cur_size / PageSize);
+}
+
 ResultCode KPageTable::UnmapProcessMemory(VAddr dst_addr, std::size_t size,
                                           KPageTable& src_page_table, VAddr src_addr) {
     KScopedLightLock lk(general_lock);
@@ -1341,7 +1429,7 @@ ResultCode KPageTable::SetProcessMemoryPermission(VAddr addr, std::size_t size,
             new_state = KMemoryState::AliasCodeData;
             break;
         default:
-            UNREACHABLE();
+            ASSERT(false);
         }
     }
 
@@ -1687,22 +1775,22 @@ ResultCode KPageTable::UnlockForDeviceAddressSpace(VAddr addr, std::size_t size)
     return ResultSuccess;
 }
 
-ResultCode KPageTable::LockForCodeMemory(VAddr addr, std::size_t size) {
+ResultCode KPageTable::LockForCodeMemory(KPageLinkedList* out, VAddr addr, std::size_t size) {
     return this->LockMemoryAndOpen(
-        nullptr, nullptr, addr, size, KMemoryState::FlagCanCodeMemory,
-        KMemoryState::FlagCanCodeMemory, KMemoryPermission::All, KMemoryPermission::UserReadWrite,
-        KMemoryAttribute::All, KMemoryAttribute::None,
+        out, nullptr, addr, size, KMemoryState::FlagCanCodeMemory, KMemoryState::FlagCanCodeMemory,
+        KMemoryPermission::All, KMemoryPermission::UserReadWrite, KMemoryAttribute::All,
+        KMemoryAttribute::None,
         static_cast<KMemoryPermission>(KMemoryPermission::NotMapped |
                                        KMemoryPermission::KernelReadWrite),
         KMemoryAttribute::Locked);
 }
 
-ResultCode KPageTable::UnlockForCodeMemory(VAddr addr, std::size_t size) {
-    return this->UnlockMemory(addr, size, KMemoryState::FlagCanCodeMemory,
-                              KMemoryState::FlagCanCodeMemory, KMemoryPermission::None,
-                              KMemoryPermission::None, KMemoryAttribute::All,
-                              KMemoryAttribute::Locked, KMemoryPermission::UserReadWrite,
-                              KMemoryAttribute::Locked, nullptr);
+ResultCode KPageTable::UnlockForCodeMemory(VAddr addr, std::size_t size,
+                                           const KPageLinkedList& pg) {
+    return this->UnlockMemory(
+        addr, size, KMemoryState::FlagCanCodeMemory, KMemoryState::FlagCanCodeMemory,
+        KMemoryPermission::None, KMemoryPermission::None, KMemoryAttribute::All,
+        KMemoryAttribute::Locked, KMemoryPermission::UserReadWrite, KMemoryAttribute::Locked, &pg);
 }
 
 ResultCode KPageTable::InitializeMemoryLayout(VAddr start, VAddr end) {
@@ -1734,9 +1822,7 @@ void KPageTable::AddRegionToPages(VAddr start, std::size_t num_pages,
     VAddr addr{start};
     while (addr < start + (num_pages * PageSize)) {
         const PAddr paddr{GetPhysicalAddr(addr)};
-        if (!paddr) {
-            UNREACHABLE();
-        }
+        ASSERT(paddr != 0);
         page_linked_list.AddBlock(paddr, 1);
         addr += PageSize;
     }
@@ -1767,7 +1853,7 @@ ResultCode KPageTable::Operate(VAddr addr, std::size_t num_pages, const KPageLin
             system.Memory().MapMemoryRegion(page_table_impl, addr, size, node.GetAddress());
             break;
         default:
-            UNREACHABLE();
+            ASSERT(false);
         }
 
         addr += size;
@@ -1798,7 +1884,7 @@ ResultCode KPageTable::Operate(VAddr addr, std::size_t num_pages, KMemoryPermiss
     case OperationType::ChangePermissionsAndRefresh:
         break;
     default:
-        UNREACHABLE();
+        ASSERT(false);
     }
     return ResultSuccess;
 }
@@ -1835,7 +1921,6 @@ VAddr KPageTable::GetRegionAddress(KMemoryState state) const {
         return code_region_start;
     default:
         UNREACHABLE();
-        return {};
     }
 }
 
@@ -1871,7 +1956,6 @@ std::size_t KPageTable::GetRegionSize(KMemoryState state) const {
         return code_region_end - code_region_start;
     default:
         UNREACHABLE();
-        return {};
     }
 }
 
@@ -2125,7 +2209,7 @@ ResultCode KPageTable::UnlockMemory(VAddr addr, size_t size, KMemoryState state_
 
     // Check the page group.
     if (pg != nullptr) {
-        UNIMPLEMENTED_MSG("PageGroup support is unimplemented!");
+        R_UNLESS(this->IsValidPageGroup(*pg, addr, num_pages), ResultInvalidMemoryRegion);
     }
 
     // Decide on new perm and attr.

src/core/hle/kernel/k_page_table.h

@@ -72,8 +72,8 @@ public:
                                           KMemoryPermission perm, PAddr map_addr = 0);
     ResultCode LockForDeviceAddressSpace(VAddr addr, std::size_t size);
     ResultCode UnlockForDeviceAddressSpace(VAddr addr, std::size_t size);
-    ResultCode LockForCodeMemory(VAddr addr, std::size_t size);
-    ResultCode UnlockForCodeMemory(VAddr addr, std::size_t size);
+    ResultCode LockForCodeMemory(KPageLinkedList* out, VAddr addr, std::size_t size);
+    ResultCode UnlockForCodeMemory(VAddr addr, std::size_t size, const KPageLinkedList& pg);
     ResultCode MakeAndOpenPageGroup(KPageLinkedList* out, VAddr address, size_t num_pages,
                                     KMemoryState state_mask, KMemoryState state,
                                     KMemoryPermission perm_mask, KMemoryPermission perm,
@@ -178,6 +178,7 @@ private:
                             const KPageLinkedList* pg);
 
     ResultCode MakePageGroup(KPageLinkedList& pg, VAddr addr, size_t num_pages);
+    bool IsValidPageGroup(const KPageLinkedList& pg, VAddr addr, size_t num_pages);
 
     bool IsLockedByCurrentThread() const {
         return general_lock.IsLockedByCurrentThread();

src/core/hle/kernel/k_port.cpp

@@ -60,7 +60,7 @@ ResultCode KPort::EnqueueSession(KServerSession* session) {
     if (auto session_ptr = server.GetSessionRequestHandler().lock()) {
         session_ptr->ClientConnected(server.AcceptSession());
     } else {
-        UNREACHABLE();
+        ASSERT(false);
     }
 
     return ResultSuccess;

src/core/hle/kernel/k_process.cpp

@@ -57,18 +57,13 @@ void SetupMainThread(Core::System& system, KProcess& owner_process, u32 priority
     thread->GetContext64().cpu_registers[0] = 0;
     thread->GetContext32().cpu_registers[1] = thread_handle;
     thread->GetContext64().cpu_registers[1] = thread_handle;
-    thread->DisableDispatch();
 
-    auto& kernel = system.Kernel();
-    // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires
-    {
-        KScopedSchedulerLock lock{kernel};
-        thread->SetState(ThreadState::Runnable);
-
-        if (system.DebuggerEnabled()) {
-            thread->RequestSuspend(SuspendType::Debug);
-        }
+    if (system.DebuggerEnabled()) {
+        thread->RequestSuspend(SuspendType::Debug);
     }
+
+    // Run our thread.
+    void(thread->Run());
 }
 } // Anonymous namespace
 
@@ -181,7 +176,8 @@ void KProcess::PinCurrentThread(s32 core_id) {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
 
     // Get the current thread.
-    KThread* cur_thread = kernel.Scheduler(static_cast<std::size_t>(core_id)).GetCurrentThread();
+    KThread* cur_thread =
+        kernel.Scheduler(static_cast<std::size_t>(core_id)).GetSchedulerCurrentThread();
 
     // If the thread isn't terminated, pin it.
     if (!cur_thread->IsTerminationRequested()) {
@@ -198,7 +194,8 @@ void KProcess::UnpinCurrentThread(s32 core_id) {
     ASSERT(kernel.GlobalSchedulerContext().IsLocked());
 
     // Get the current thread.
-    KThread* cur_thread = kernel.Scheduler(static_cast<std::size_t>(core_id)).GetCurrentThread();
+    KThread* cur_thread =
+        kernel.Scheduler(static_cast<std::size_t>(core_id)).GetSchedulerCurrentThread();
 
     // Unpin it.
     cur_thread->Unpin();
@@ -275,11 +272,15 @@ void KProcess::RemoveSharedMemory(KSharedMemory* shmem, [[maybe_unused]] VAddr a
     shmem->Close();
 }
 
-void KProcess::RegisterThread(const KThread* thread) {
+void KProcess::RegisterThread(KThread* thread) {
+    KScopedLightLock lk{list_lock};
+
     thread_list.push_back(thread);
 }
 
-void KProcess::UnregisterThread(const KThread* thread) {
+void KProcess::UnregisterThread(KThread* thread) {
+    KScopedLightLock lk{list_lock};
+
     thread_list.remove(thread);
 }
 
@@ -297,6 +298,50 @@ ResultCode KProcess::Reset() {
     return ResultSuccess;
 }
 
+ResultCode KProcess::SetActivity(ProcessActivity activity) {
+    // Lock ourselves and the scheduler.
+    KScopedLightLock lk{state_lock};
+    KScopedLightLock list_lk{list_lock};
+    KScopedSchedulerLock sl{kernel};
+
+    // Validate our state.
+    R_UNLESS(status != ProcessStatus::Exiting, ResultInvalidState);
+    R_UNLESS(status != ProcessStatus::Exited, ResultInvalidState);
+
+    // Either pause or resume.
+    if (activity == ProcessActivity::Paused) {
+        // Verify that we're not suspended.
+        if (is_suspended) {
+            return ResultInvalidState;
+        }
+
+        // Suspend all threads.
+        for (auto* thread : GetThreadList()) {
+            thread->RequestSuspend(SuspendType::Process);
+        }
+
+        // Set ourselves as suspended.
+        SetSuspended(true);
+    } else {
+        ASSERT(activity == ProcessActivity::Runnable);
+
+        // Verify that we're suspended.
+        if (!is_suspended) {
+            return ResultInvalidState;
+        }
+
+        // Resume all threads.
+        for (auto* thread : GetThreadList()) {
+            thread->Resume(SuspendType::Process);
+        }
+
+        // Set ourselves as resumed.
+        SetSuspended(false);
+    }
+
+    return ResultSuccess;
+}
+
 ResultCode KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata,
                                       std::size_t code_size) {
     program_id = metadata.GetTitleID();
@@ -350,7 +395,7 @@ ResultCode KProcess::LoadFromMetadata(const FileSys::ProgramMetadata& metadata,
         break;
 
     default:
-        UNREACHABLE();
+        ASSERT(false);
     }
 
     // Create TLS region
@@ -377,11 +422,11 @@ void KProcess::PrepareForTermination() {
     ChangeStatus(ProcessStatus::Exiting);
 
     const auto stop_threads = [this](const std::vector<KThread*>& in_thread_list) {
-        for (auto& thread : in_thread_list) {
+        for (auto* thread : in_thread_list) {
             if (thread->GetOwnerProcess() != this)
                 continue;
 
-            if (thread == kernel.CurrentScheduler()->GetCurrentThread())
+            if (thread == GetCurrentThreadPointer(kernel))
                 continue;
 
             // TODO(Subv): When are the other running/ready threads terminated?
@@ -536,6 +581,52 @@ ResultCode KProcess::DeleteThreadLocalRegion(VAddr addr) {
     return ResultSuccess;
 }
 
+bool KProcess::InsertWatchpoint(Core::System& system, VAddr addr, u64 size,
+                                DebugWatchpointType type) {
+    const auto watch{std::find_if(watchpoints.begin(), watchpoints.end(), [&](const auto& wp) {
+        return wp.type == DebugWatchpointType::None;
+    })};
+
+    if (watch == watchpoints.end()) {
+        return false;
+    }
+
+    watch->start_address = addr;
+    watch->end_address = addr + size;
+    watch->type = type;
+
+    for (VAddr page = Common::AlignDown(addr, PageSize); page < addr + size; page += PageSize) {
+        debug_page_refcounts[page]++;
+        system.Memory().MarkRegionDebug(page, PageSize, true);
+    }
+
+    return true;
+}
+
+bool KProcess::RemoveWatchpoint(Core::System& system, VAddr addr, u64 size,
+                                DebugWatchpointType type) {
+    const auto watch{std::find_if(watchpoints.begin(), watchpoints.end(), [&](const auto& wp) {
+        return wp.start_address == addr && wp.end_address == addr + size && wp.type == type;
+    })};
+
+    if (watch == watchpoints.end()) {
+        return false;
+    }
+
+    watch->start_address = 0;
+    watch->end_address = 0;
+    watch->type = DebugWatchpointType::None;
+
+    for (VAddr page = Common::AlignDown(addr, PageSize); page < addr + size; page += PageSize) {
+        debug_page_refcounts[page]--;
+        if (!debug_page_refcounts[page]) {
+            system.Memory().MarkRegionDebug(page, PageSize, false);
+        }
+    }
+
+    return true;
+}
+
 void KProcess::LoadModule(CodeSet code_set, VAddr base_addr) {
     const auto ReprotectSegment = [&](const CodeSet::Segment& segment,
                                       Svc::MemoryPermission permission) {
@@ -556,9 +647,10 @@ bool KProcess::IsSignaled() const {
 }
 
 KProcess::KProcess(KernelCore& kernel_)
-    : KAutoObjectWithSlabHeapAndContainer{kernel_},
-      page_table{std::make_unique<KPageTable>(kernel_.System())}, handle_table{kernel_},
-      address_arbiter{kernel_.System()}, condition_var{kernel_.System()}, state_lock{kernel_} {}
+    : KAutoObjectWithSlabHeapAndContainer{kernel_}, page_table{std::make_unique<KPageTable>(
+                                                        kernel_.System())},
+      handle_table{kernel_}, address_arbiter{kernel_.System()}, condition_var{kernel_.System()},
+      state_lock{kernel_}, list_lock{kernel_} {}
 
 KProcess::~KProcess() = default;
 

src/core/hle/kernel/k_process.h

@@ -7,6 +7,7 @@
 #include <array>
 #include <cstddef>
 #include <list>
+#include <map>
 #include <string>
 #include "common/common_types.h"
 #include "core/hle/kernel/k_address_arbiter.h"
@@ -63,6 +64,25 @@ enum class ProcessStatus {
     DebugBreak,
 };
 
+enum class ProcessActivity : u32 {
+    Runnable,
+    Paused,
+};
+
+enum class DebugWatchpointType : u8 {
+    None = 0,
+    Read = 1 << 0,
+    Write = 1 << 1,
+    ReadOrWrite = Read | Write,
+};
+DECLARE_ENUM_FLAG_OPERATORS(DebugWatchpointType);
+
+struct DebugWatchpoint {
+    VAddr start_address;
+    VAddr end_address;
+    DebugWatchpointType type;
+};
+
 class KProcess final : public KAutoObjectWithSlabHeapAndContainer<KProcess, KWorkerTask> {
     KERNEL_AUTOOBJECT_TRAITS(KProcess, KSynchronizationObject);
 
@@ -282,17 +302,17 @@ public:
     u64 GetTotalPhysicalMemoryUsedWithoutSystemResource() const;
 
     /// Gets the list of all threads created with this process as their owner.
-    const std::list<const KThread*>& GetThreadList() const {
+    std::list<KThread*>& GetThreadList() {
         return thread_list;
     }
 
     /// Registers a thread as being created under this process,
     /// adding it to this process' thread list.
-    void RegisterThread(const KThread* thread);
+    void RegisterThread(KThread* thread);
 
     /// Unregisters a thread from this process, removing it
     /// from this process' thread list.
-    void UnregisterThread(const KThread* thread);
+    void UnregisterThread(KThread* thread);
 
     /// Clears the signaled state of the process if and only if it's signaled.
     ///
@@ -347,6 +367,8 @@ public:
 
     void DoWorkerTaskImpl();
 
+    ResultCode SetActivity(ProcessActivity activity);
+
     void PinCurrentThread(s32 core_id);
     void UnpinCurrentThread(s32 core_id);
     void UnpinThread(KThread* thread);
@@ -367,6 +389,19 @@ public:
     // Frees a used TLS slot identified by the given address
     ResultCode DeleteThreadLocalRegion(VAddr addr);
 
+    ///////////////////////////////////////////////////////////////////////////////////////////////
+    // Debug watchpoint management
+
+    // Attempts to insert a watchpoint into a free slot. Returns false if none are available.
+    bool InsertWatchpoint(Core::System& system, VAddr addr, u64 size, DebugWatchpointType type);
+
+    // Attempts to remove the watchpoint specified by the given parameters.
+    bool RemoveWatchpoint(Core::System& system, VAddr addr, u64 size, DebugWatchpointType type);
+
+    const std::array<DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS>& GetWatchpoints() const {
+        return watchpoints;
+    }
+
 private:
     void PinThread(s32 core_id, KThread* thread) {
         ASSERT(0 <= core_id && core_id < static_cast<s32>(Core::Hardware::NUM_CPU_CORES));
@@ -442,7 +477,7 @@ private:
     std::array<u64, RANDOM_ENTROPY_SIZE> random_entropy{};
 
     /// List of threads that are running with this process as their owner.
-    std::list<const KThread*> thread_list;
+    std::list<KThread*> thread_list;
 
     /// List of shared memory that are running with this process as their owner.
     std::list<KSharedMemoryInfo*> shared_memory_list;
@@ -471,10 +506,13 @@ private:
     std::array<KThread*, Core::Hardware::NUM_CPU_CORES> running_threads{};
     std::array<u64, Core::Hardware::NUM_CPU_CORES> running_thread_idle_counts{};
     std::array<KThread*, Core::Hardware::NUM_CPU_CORES> pinned_threads{};
+    std::array<DebugWatchpoint, Core::Hardware::NUM_WATCHPOINTS> watchpoints{};
+    std::map<VAddr, u64> debug_page_refcounts;
 
     KThread* exception_thread{};
 
     KLightLock state_lock;
+    KLightLock list_lock;
 
     using TLPTree =
         Common::IntrusiveRedBlackTreeBaseTraits<KThreadLocalPage>::TreeType<KThreadLocalPage>;

src/core/hle/kernel/k_scheduler.cpp

@@ -317,7 +317,7 @@ void KScheduler::RotateScheduledQueue(s32 cpu_core_id, s32 priority) {
 
     {
         KThread* best_thread = priority_queue.GetScheduledFront(cpu_core_id);
-        if (best_thread == GetCurrentThread()) {
+        if (best_thread == GetCurrentThreadPointer(kernel)) {
             best_thread = priority_queue.GetScheduledNext(cpu_core_id, best_thread);
         }
 
@@ -424,7 +424,7 @@ void KScheduler::YieldWithoutCoreMigration(KernelCore& kernel) {
     ASSERT(kernel.CurrentProcess() != nullptr);
 
     // Get the current thread and process.
-    KThread& cur_thread = Kernel::GetCurrentThread(kernel);
+    KThread& cur_thread = GetCurrentThread(kernel);
     KProcess& cur_process = *kernel.CurrentProcess();
 
     // If the thread's yield count matches, there's nothing for us to do.
@@ -463,7 +463,7 @@ void KScheduler::YieldWithCoreMigration(KernelCore& kernel) {
     ASSERT(kernel.CurrentProcess() != nullptr);
 
     // Get the current thread and process.
-    KThread& cur_thread = Kernel::GetCurrentThread(kernel);
+    KThread& cur_thread = GetCurrentThread(kernel);
     KProcess& cur_process = *kernel.CurrentProcess();
 
     // If the thread's yield count matches, there's nothing for us to do.
@@ -551,7 +551,7 @@ void KScheduler::YieldToAnyThread(KernelCore& kernel) {
     ASSERT(kernel.CurrentProcess() != nullptr);
 
     // Get the current thread and process.
-    KThread& cur_thread = Kernel::GetCurrentThread(kernel);
+    KThread& cur_thread = GetCurrentThread(kernel);
     KProcess& cur_process = *kernel.CurrentProcess();
 
     // If the thread's yield count matches, there's nothing for us to do.
@@ -642,7 +642,7 @@ KScheduler::~KScheduler() {
     ASSERT(!idle_thread);
 }
 
-KThread* KScheduler::GetCurrentThread() const {
+KThread* KScheduler::GetSchedulerCurrentThread() const {
     if (auto result = current_thread.load(); result) {
         return result;
     }
@@ -654,7 +654,7 @@ u64 KScheduler::GetLastContextSwitchTicks() const {
 }
 
 void KScheduler::RescheduleCurrentCore() {
-    ASSERT(GetCurrentThread()->GetDisableDispatchCount() == 1);
+    ASSERT(GetCurrentThread(system.Kernel()).GetDisableDispatchCount() == 1);
 
     auto& phys_core = system.Kernel().PhysicalCore(core_id);
     if (phys_core.IsInterrupted()) {
@@ -665,7 +665,7 @@ void KScheduler::RescheduleCurrentCore() {
     if (state.needs_scheduling.load()) {
         Schedule();
     } else {
-        GetCurrentThread()->EnableDispatch();
+        GetCurrentThread(system.Kernel()).EnableDispatch();
         guard.Unlock();
     }
 }
@@ -710,6 +710,7 @@ void KScheduler::Reload(KThread* thread) {
     Core::ARM_Interface& cpu_core = system.ArmInterface(core_id);
     cpu_core.LoadContext(thread->GetContext32());
     cpu_core.LoadContext(thread->GetContext64());
+    cpu_core.LoadWatchpointArray(thread->GetOwnerProcess()->GetWatchpoints());
     cpu_core.SetTlsAddress(thread->GetTLSAddress());
     cpu_core.SetTPIDR_EL0(thread->GetTPIDR_EL0());
     cpu_core.ClearExclusiveState();
@@ -717,13 +718,18 @@ void KScheduler::Reload(KThread* thread) {
 
 void KScheduler::SwitchContextStep2() {
     // Load context of new thread
-    Reload(GetCurrentThread());
+    Reload(GetCurrentThreadPointer(system.Kernel()));
 
     RescheduleCurrentCore();
 }
 
+void KScheduler::Schedule() {
+    ASSERT(GetCurrentThread(system.Kernel()).GetDisableDispatchCount() == 1);
+    this->ScheduleImpl();
+}
+
 void KScheduler::ScheduleImpl() {
-    KThread* previous_thread = GetCurrentThread();
+    KThread* previous_thread = GetCurrentThreadPointer(system.Kernel());
     KThread* next_thread = state.highest_priority_thread;
 
     state.needs_scheduling.store(false);
@@ -761,6 +767,7 @@ void KScheduler::ScheduleImpl() {
     old_context = &previous_thread->GetHostContext();
 
     // Set the new thread.
+    SetCurrentThread(system.Kernel(), next_thread);
     current_thread.store(next_thread);
 
     guard.Unlock();
@@ -804,6 +811,7 @@ void KScheduler::SwitchToCurrent() {
                 }
             }
             auto thread = next_thread ? next_thread : idle_thread;
+            SetCurrentThread(system.Kernel(), thread);
             Common::Fiber::YieldTo(switch_fiber, *thread->GetHostContext());
         } while (!is_switch_pending());
     }
@@ -829,6 +837,7 @@ void KScheduler::Initialize() {
     idle_thread = KThread::Create(system.Kernel());
     ASSERT(KThread::InitializeIdleThread(system, idle_thread, core_id).IsSuccess());
     idle_thread->SetName(fmt::format("IdleThread:{}", core_id));
+    idle_thread->EnableDispatch();
 }
 
 KScopedSchedulerLock::KScopedSchedulerLock(KernelCore& kernel)

src/core/hle/kernel/k_scheduler.h

@@ -48,18 +48,13 @@ public:
     void Reload(KThread* thread);
 
     /// Gets the current running thread
-    [[nodiscard]] KThread* GetCurrentThread() const;
+    [[nodiscard]] KThread* GetSchedulerCurrentThread() const;
 
     /// Gets the idle thread
     [[nodiscard]] KThread* GetIdleThread() const {
         return idle_thread;
     }
 
-    /// Returns true if the scheduler is idle
-    [[nodiscard]] bool IsIdle() const {
-        return GetCurrentThread() == idle_thread;
-    }
-
     /// Gets the timestamp for the last context switch in ticks.
     [[nodiscard]] u64 GetLastContextSwitchTicks() const;
 
@@ -149,10 +144,7 @@ private:
 
     void RotateScheduledQueue(s32 cpu_core_id, s32 priority);
 
-    void Schedule() {
-        ASSERT(GetCurrentThread()->GetDisableDispatchCount() == 1);
-        this->ScheduleImpl();
-    }
+    void Schedule();
 
     /// Switches the CPU's active thread context to that of the specified thread
     void ScheduleImpl();

src/core/hle/kernel/k_server_session.cpp

@@ -97,13 +97,13 @@ ResultCode KServerSession::HandleDomainSyncRequest(Kernel::HLERequestContext& co
                          "object_id {} is too big! This probably means a recent service call "
                          "to {} needed to return a new interface!",
                          object_id, name);
-            UNREACHABLE();
+            ASSERT(false);
             return ResultSuccess; // Ignore error if asserts are off
         }
         if (auto strong_ptr = manager->DomainHandler(object_id - 1).lock()) {
             return strong_ptr->HandleSyncRequest(*this, context);
         } else {
-            UNREACHABLE();
+            ASSERT(false);
             return ResultSuccess;
         }
 

src/core/hle/kernel/k_thread.cpp

@@ -133,7 +133,7 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
         UNIMPLEMENTED();
         break;
     default:
-        UNREACHABLE_MSG("KThread::Initialize: Unknown ThreadType {}", static_cast<u32>(type));
+        ASSERT_MSG(false, "KThread::Initialize: Unknown ThreadType {}", static_cast<u32>(type));
         break;
     }
     thread_type = type;
@@ -225,7 +225,7 @@ ResultCode KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_s
     // Setup the stack parameters.
     StackParameters& sp = GetStackParameters();
     sp.cur_thread = this;
-    sp.disable_count = 0;
+    sp.disable_count = 1;
     SetInExceptionHandler();
 
     // Set thread ID.
@@ -267,15 +267,15 @@ ResultCode KThread::InitializeDummyThread(KThread* thread) {
 ResultCode KThread::InitializeIdleThread(Core::System& system, KThread* thread, s32 virt_core) {
     return InitializeThread(thread, {}, {}, {}, IdleThreadPriority, virt_core, {}, ThreadType::Main,
                             Core::CpuManager::GetIdleThreadStartFunc(),
-                            system.GetCpuManager().GetStartFuncParamater());
+                            system.GetCpuManager().GetStartFuncParameter());
 }
 
 ResultCode KThread::InitializeHighPriorityThread(Core::System& system, KThread* thread,
                                                  KThreadFunction func, uintptr_t arg,
                                                  s32 virt_core) {
     return InitializeThread(thread, func, arg, {}, {}, virt_core, nullptr, ThreadType::HighPriority,
-                            Core::CpuManager::GetSuspendThreadStartFunc(),
-                            system.GetCpuManager().GetStartFuncParamater());
+                            Core::CpuManager::GetShutdownThreadStartFunc(),
+                            system.GetCpuManager().GetStartFuncParameter());
 }
 
 ResultCode KThread::InitializeUserThread(Core::System& system, KThread* thread,
@@ -284,7 +284,7 @@ ResultCode KThread::InitializeUserThread(Core::System& system, KThread* thread,
     system.Kernel().GlobalSchedulerContext().AddThread(thread);
     return InitializeThread(thread, func, arg, user_stack_top, prio, virt_core, owner,
                             ThreadType::User, Core::CpuManager::GetGuestThreadStartFunc(),
-                            system.GetCpuManager().GetStartFuncParamater());
+                            system.GetCpuManager().GetStartFuncParameter());
 }
 
 void KThread::PostDestroy(uintptr_t arg) {
@@ -382,7 +382,7 @@ void KThread::FinishTermination() {
         for (std::size_t i = 0; i < static_cast<std::size_t>(Core::Hardware::NUM_CPU_CORES); ++i) {
             KThread* core_thread{};
             do {
-                core_thread = kernel.Scheduler(i).GetCurrentThread();
+                core_thread = kernel.Scheduler(i).GetSchedulerCurrentThread();
             } while (core_thread == this);
         }
     }
@@ -631,7 +631,7 @@ ResultCode KThread::SetCoreMask(s32 core_id_, u64 v_affinity_mask) {
             s32 thread_core;
             for (thread_core = 0; thread_core < static_cast<s32>(Core::Hardware::NUM_CPU_CORES);
                  ++thread_core) {
-                if (kernel.Scheduler(thread_core).GetCurrentThread() == this) {
+                if (kernel.Scheduler(thread_core).GetSchedulerCurrentThread() == this) {
                     thread_is_current = true;
                     break;
                 }
@@ -748,6 +748,19 @@ void KThread::Continue() {
     KScheduler::OnThreadStateChanged(kernel, this, old_state);
 }
 
+void KThread::WaitUntilSuspended() {
+    // Make sure we have a suspend requested.
+    ASSERT(IsSuspendRequested());
+
+    // Loop until the thread is not executing on any core.
+    for (std::size_t i = 0; i < static_cast<std::size_t>(Core::Hardware::NUM_CPU_CORES); ++i) {
+        KThread* core_thread{};
+        do {
+            core_thread = kernel.Scheduler(i).GetSchedulerCurrentThread();
+        } while (core_thread == this);
+    }
+}
+
 ResultCode KThread::SetActivity(Svc::ThreadActivity activity) {
     // Lock ourselves.
     KScopedLightLock lk(activity_pause_lock);
@@ -809,7 +822,7 @@ ResultCode KThread::SetActivity(Svc::ThreadActivity activity) {
                 // Check if the thread is currently running.
                 // If it is, we'll need to retry.
                 for (auto i = 0; i < static_cast<s32>(Core::Hardware::NUM_CPU_CORES); ++i) {
-                    if (kernel.Scheduler(i).GetCurrentThread() == this) {
+                    if (kernel.Scheduler(i).GetSchedulerCurrentThread() == this) {
                         thread_is_current = true;
                         break;
                     }
@@ -1014,8 +1027,6 @@ ResultCode KThread::Run() {
         // Set our state and finish.
         SetState(ThreadState::Runnable);
 
-        DisableDispatch();
-
         return ResultSuccess;
     }
 }
@@ -1164,6 +1175,10 @@ std::shared_ptr<Common::Fiber>& KThread::GetHostContext() {
     return host_context;
 }
 
+void SetCurrentThread(KernelCore& kernel, KThread* thread) {
+    kernel.SetCurrentEmuThread(thread);
+}
+
 KThread* GetCurrentThreadPointer(KernelCore& kernel) {
     return kernel.GetCurrentEmuThread();
 }

src/core/hle/kernel/k_thread.h

@@ -106,6 +106,7 @@ enum class StepState : u32 {
     StepPerformed, ///< Thread has stepped, waiting to be scheduled again
 };
 
+void SetCurrentThread(KernelCore& kernel, KThread* thread);
 [[nodiscard]] KThread* GetCurrentThreadPointer(KernelCore& kernel);
 [[nodiscard]] KThread& GetCurrentThread(KernelCore& kernel);
 [[nodiscard]] s32 GetCurrentCoreId(KernelCore& kernel);
@@ -207,6 +208,8 @@ public:
 
     void Continue();
 
+    void WaitUntilSuspended();
+
     constexpr void SetSyncedIndex(s32 index) {
         synced_index = index;
     }

src/core/hle/kernel/kernel.cpp

@@ -76,7 +76,7 @@ struct KernelCore::Impl {
         InitializeMemoryLayout();
         Init::InitializeKPageBufferSlabHeap(system);
         InitializeSchedulers();
-        InitializeSuspendThreads();
+        InitializeShutdownThreads();
         InitializePreemption(kernel);
 
         RegisterHostThread();
@@ -143,9 +143,9 @@ struct KernelCore::Impl {
         CleanupObject(system_resource_limit);
 
         for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            if (suspend_threads[core_id]) {
-                suspend_threads[core_id]->Close();
-                suspend_threads[core_id] = nullptr;
+            if (shutdown_threads[core_id]) {
+                shutdown_threads[core_id]->Close();
+                shutdown_threads[core_id] = nullptr;
             }
 
             schedulers[core_id]->Finalize();
@@ -212,7 +212,9 @@ struct KernelCore::Impl {
         system_resource_limit = KResourceLimit::Create(system.Kernel());
         system_resource_limit->Initialize(&core_timing);
 
-        const auto [total_size, kernel_size] = memory_layout->GetTotalAndKernelMemorySizes();
+        const auto sizes{memory_layout->GetTotalAndKernelMemorySizes()};
+        const auto total_size{sizes.first};
+        const auto kernel_size{sizes.second};
 
         // If setting the default system values fails, then something seriously wrong has occurred.
         ASSERT(system_resource_limit->SetLimitValue(LimitableResource::PhysicalMemory, total_size)
@@ -245,13 +247,13 @@ struct KernelCore::Impl {
         system.CoreTiming().ScheduleEvent(time_interval, preemption_event);
     }
 
-    void InitializeSuspendThreads() {
+    void InitializeShutdownThreads() {
         for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            suspend_threads[core_id] = KThread::Create(system.Kernel());
-            ASSERT(KThread::InitializeHighPriorityThread(system, suspend_threads[core_id], {}, {},
+            shutdown_threads[core_id] = KThread::Create(system.Kernel());
+            ASSERT(KThread::InitializeHighPriorityThread(system, shutdown_threads[core_id], {}, {},
                                                          core_id)
                        .IsSuccess());
-            suspend_threads[core_id]->SetName(fmt::format("SuspendThread:{}", core_id));
+            shutdown_threads[core_id]->SetName(fmt::format("SuspendThread:{}", core_id));
         }
     }
 
@@ -329,6 +331,8 @@ struct KernelCore::Impl {
         return is_shutting_down.load(std::memory_order_relaxed);
     }
 
+    static inline thread_local KThread* current_thread{nullptr};
+
     KThread* GetCurrentEmuThread() {
         // If we are shutting down the kernel, none of this is relevant anymore.
         if (IsShuttingDown()) {
@@ -339,7 +343,12 @@ struct KernelCore::Impl {
         if (thread_id >= Core::Hardware::NUM_CPU_CORES) {
             return GetHostDummyThread();
         }
-        return schedulers[thread_id]->GetCurrentThread();
+
+        return current_thread;
+    }
+
+    void SetCurrentEmuThread(KThread* thread) {
+        current_thread = thread;
     }
 
     void DeriveInitialMemoryLayout() {
@@ -766,7 +775,7 @@ struct KernelCore::Impl {
     std::weak_ptr<ServiceThread> default_service_thread;
     Common::ThreadWorker service_threads_manager;
 
-    std::array<KThread*, Core::Hardware::NUM_CPU_CORES> suspend_threads;
+    std::array<KThread*, Core::Hardware::NUM_CPU_CORES> shutdown_threads;
     std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES> interrupts{};
     std::array<std::unique_ptr<Kernel::KScheduler>, Core::Hardware::NUM_CPU_CORES> schedulers{};
 
@@ -917,6 +926,12 @@ const KAutoObjectWithListContainer& KernelCore::ObjectListContainer() const {
     return *impl->global_object_list_container;
 }
 
+void KernelCore::InterruptAllPhysicalCores() {
+    for (auto& physical_core : impl->cores) {
+        physical_core.Interrupt();
+    }
+}
+
 void KernelCore::InvalidateAllInstructionCaches() {
     for (auto& physical_core : impl->cores) {
         physical_core.ArmInterface().ClearInstructionCache();
@@ -1016,6 +1031,10 @@ KThread* KernelCore::GetCurrentEmuThread() const {
     return impl->GetCurrentEmuThread();
 }
 
+void KernelCore::SetCurrentEmuThread(KThread* thread) {
+    impl->SetCurrentEmuThread(thread);
+}
+
 KMemoryManager& KernelCore::MemoryManager() {
     return *impl->memory_manager;
 }
@@ -1064,22 +1083,29 @@ const Kernel::KSharedMemory& KernelCore::GetHidBusSharedMem() const {
     return *impl->hidbus_shared_mem;
 }
 
-void KernelCore::Suspend(bool in_suspention) {
-    const bool should_suspend = exception_exited || in_suspention;
-    {
-        KScopedSchedulerLock lock(*this);
-        const auto state = should_suspend ? ThreadState::Runnable : ThreadState::Waiting;
-        for (u32 core_id = 0; core_id < Core::Hardware::NUM_CPU_CORES; core_id++) {
-            impl->suspend_threads[core_id]->SetState(state);
-            impl->suspend_threads[core_id]->SetWaitReasonForDebugging(
-                ThreadWaitReasonForDebugging::Suspended);
-            if (!should_suspend) {
-                impl->suspend_threads[core_id]->DisableDispatch();
+void KernelCore::Suspend(bool suspended) {
+    const bool should_suspend{exception_exited || suspended};
+    const auto activity = should_suspend ? ProcessActivity::Paused : ProcessActivity::Runnable;
+
+    for (auto* process : GetProcessList()) {
+        process->SetActivity(activity);
+
+        if (should_suspend) {
+            // Wait for execution to stop
+            for (auto* thread : process->GetThreadList()) {
+                thread->WaitUntilSuspended();
             }
         }
     }
 }
 
+void KernelCore::ShutdownCores() {
+    for (auto* thread : impl->shutdown_threads) {
+        void(thread->Run());
+    }
+    InterruptAllPhysicalCores();
+}
+
 bool KernelCore::IsMulticore() const {
     return impl->is_multicore;
 }

src/core/hle/kernel/kernel.h

@@ -184,6 +184,8 @@ public:
 
     const std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES>& Interrupts() const;
 
+    void InterruptAllPhysicalCores();
+
     void InvalidateAllInstructionCaches();
 
     void InvalidateCpuInstructionCacheRange(VAddr addr, std::size_t size);
@@ -224,6 +226,9 @@ public:
     /// Gets the current host_thread/guest_thread pointer.
     KThread* GetCurrentEmuThread() const;
 
+    /// Sets the current guest_thread pointer.
+    void SetCurrentEmuThread(KThread* thread);
+
     /// Gets the current host_thread handle.
     u32 GetCurrentHostThreadID() const;
 
@@ -269,12 +274,15 @@ public:
     /// Gets the shared memory object for HIDBus services.
     const Kernel::KSharedMemory& GetHidBusSharedMem() const;
 
-    /// Suspend/unsuspend the OS.
-    void Suspend(bool in_suspention);
+    /// Suspend/unsuspend all processes.
+    void Suspend(bool suspend);
 
-    /// Exceptional exit the OS.
+    /// Exceptional exit all processes.
     void ExceptionalExit();
 
+    /// Notify emulated CPU cores to shut down.
+    void ShutdownCores();
+
     bool IsMulticore() const;
 
     bool IsShuttingDown() const;

src/core/hle/kernel/svc.cpp

@@ -15,6 +15,7 @@
 #include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/core_timing.h"
+#include "core/debugger/debugger.h"
 #include "core/hle/kernel/k_client_port.h"
 #include "core/hle/kernel/k_client_session.h"
 #include "core/hle/kernel/k_code_memory.h"
@@ -326,7 +327,6 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
 
     LOG_TRACE(Kernel_SVC, "called handle=0x{:08X}({})", handle, session->GetName());
 
-    auto thread = kernel.CurrentScheduler()->GetCurrentThread();
     {
         KScopedSchedulerLock lock(kernel);
 
@@ -336,7 +336,7 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
         session->SendSyncRequest(&GetCurrentThread(kernel), system.Memory(), system.CoreTiming());
     }
 
-    return thread->GetWaitResult();
+    return GetCurrentThread(kernel).GetWaitResult();
 }
 
 static ResultCode SendSyncRequest32(Core::System& system, Handle handle) {
@@ -623,10 +623,16 @@ static void Break(Core::System& system, u32 reason, u64 info1, u64 info2) {
 
         handle_debug_buffer(info1, info2);
 
-        auto* const current_thread = system.Kernel().CurrentScheduler()->GetCurrentThread();
+        auto* const current_thread = GetCurrentThreadPointer(system.Kernel());
         const auto thread_processor_id = current_thread->GetActiveCore();
         system.ArmInterface(static_cast<std::size_t>(thread_processor_id)).LogBacktrace();
     }
+
+    if (system.DebuggerEnabled()) {
+        auto* thread = system.Kernel().GetCurrentEmuThread();
+        system.GetDebugger().NotifyThreadStopped(thread);
+        thread->RequestSuspend(Kernel::SuspendType::Debug);
+    }
 }
 
 static void Break32(Core::System& system, u32 reason, u32 info1, u32 info2) {
@@ -877,7 +883,7 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, Handle
 
         const auto& core_timing = system.CoreTiming();
         const auto& scheduler = *system.Kernel().CurrentScheduler();
-        const auto* const current_thread = scheduler.GetCurrentThread();
+        const auto* const current_thread = GetCurrentThreadPointer(system.Kernel());
         const bool same_thread = current_thread == thread.GetPointerUnsafe();
 
         const u64 prev_ctx_ticks = scheduler.GetLastContextSwitchTicks();
@@ -1096,7 +1102,7 @@ static ResultCode GetThreadContext(Core::System& system, VAddr out_context, Hand
         if (thread->GetRawState() != ThreadState::Runnable) {
             bool current = false;
             for (auto i = 0; i < static_cast<s32>(Core::Hardware::NUM_CPU_CORES); ++i) {
-                if (thread.GetPointerUnsafe() == kernel.Scheduler(i).GetCurrentThread()) {
+                if (thread.GetPointerUnsafe() == kernel.Scheduler(i).GetSchedulerCurrentThread()) {
                     current = true;
                     break;
                 }
@@ -1719,11 +1725,12 @@ static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_ha
 /// Exits the current process
 static void ExitProcess(Core::System& system) {
     auto* current_process = system.Kernel().CurrentProcess();
-    UNIMPLEMENTED();
 
     LOG_INFO(Kernel_SVC, "Process {} exiting", current_process->GetProcessID());
     ASSERT_MSG(current_process->GetStatus() == ProcessStatus::Running,
                "Process has already exited");
+
+    system.Exit();
 }
 
 static void ExitProcess32(Core::System& system) {
@@ -1843,7 +1850,7 @@ static ResultCode StartThread32(Core::System& system, Handle thread_handle) {
 static void ExitThread(Core::System& system) {
     LOG_DEBUG(Kernel_SVC, "called, pc=0x{:08X}", system.CurrentArmInterface().GetPC());
 
-    auto* const current_thread = system.Kernel().CurrentScheduler()->GetCurrentThread();
+    auto* const current_thread = GetCurrentThreadPointer(system.Kernel());
     system.GlobalSchedulerContext().RemoveThread(current_thread);
     current_thread->Exit();
     system.Kernel().UnregisterInUseObject(current_thread);
@@ -1876,7 +1883,7 @@ static void SleepThread(Core::System& system, s64 nanoseconds) {
         KScheduler::YieldToAnyThread(kernel);
     } else {
         // Nintendo does nothing at all if an otherwise invalid value is passed.
-        UNREACHABLE_MSG("Unimplemented sleep yield type '{:016X}'!", nanoseconds);
+        ASSERT_MSG(false, "Unimplemented sleep yield type '{:016X}'!", nanoseconds);
     }
 }
 
@@ -2530,7 +2537,7 @@ static ResultCode GetThreadList(Core::System& system, u32* out_num_threads, VAdd
         return ResultOutOfRange;
     }
 
-    const auto* const current_process = system.Kernel().CurrentProcess();
+    auto* const current_process = system.Kernel().CurrentProcess();
     const auto total_copy_size = out_thread_ids_size * sizeof(u64);
 
     if (out_thread_ids_size > 0 &&
@@ -2982,11 +2989,10 @@ static const FunctionDef* GetSVCInfo64(u32 func_num) {
 }
 
 void Call(Core::System& system, u32 immediate) {
-    system.ExitDynarmicProfile();
     auto& kernel = system.Kernel();
     kernel.EnterSVCProfile();
 
-    auto* thread = kernel.CurrentScheduler()->GetCurrentThread();
+    auto* thread = GetCurrentThreadPointer(kernel);
     thread->SetIsCallingSvc();
 
     const FunctionDef* info = system.CurrentProcess()->Is64BitProcess() ? GetSVCInfo64(immediate)
@@ -3007,8 +3013,6 @@ void Call(Core::System& system, u32 immediate) {
         auto* host_context = thread->GetHostContext().get();
         host_context->Rewind();
     }
-
-    system.EnterDynarmicProfile();
 }
 
 } // namespace Kernel::Svc

src/core/hle/service/am/am.cpp

@@ -686,7 +686,7 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system_,
         {66, &ICommonStateGetter::SetCpuBoostMode, "SetCpuBoostMode"},
         {67, nullptr, "CancelCpuBoostMode"},
         {68, nullptr, "GetBuiltInDisplayType"},
-        {80, nullptr, "PerformSystemButtonPressingIfInFocus"},
+        {80, &ICommonStateGetter::PerformSystemButtonPressingIfInFocus, "PerformSystemButtonPressingIfInFocus"},
         {90, nullptr, "SetPerformanceConfigurationChangedNotification"},
         {91, nullptr, "GetCurrentPerformanceConfiguration"},
         {100, nullptr, "SetHandlingHomeButtonShortPressedEnabled"},
@@ -826,6 +826,16 @@ void ICommonStateGetter::SetCpuBoostMode(Kernel::HLERequestContext& ctx) {
     apm_sys->SetCpuBoostMode(ctx);
 }
 
+void ICommonStateGetter::PerformSystemButtonPressingIfInFocus(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto system_button{rp.PopEnum<SystemButtonType>()};
+
+    LOG_WARNING(Service_AM, "(STUBBED) called, system_button={}", system_button);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
 void ICommonStateGetter::SetRequestExitToLibraryAppletAtExecuteNextProgramEnabled(
     Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_AM, "(STUBBED) called");

src/core/hle/service/am/am.h

@@ -220,6 +220,18 @@ private:
         Docked = 1,
     };
 
+    // This is nn::am::service::SystemButtonType
+    enum class SystemButtonType {
+        None,
+        HomeButtonShortPressing,
+        HomeButtonLongPressing,
+        PowerButtonShortPressing,
+        PowerButtonLongPressing,
+        ShutdownSystem,
+        CaptureButtonShortPressing,
+        CaptureButtonLongPressing,
+    };
+
     void GetEventHandle(Kernel::HLERequestContext& ctx);
     void ReceiveMessage(Kernel::HLERequestContext& ctx);
     void GetCurrentFocusState(Kernel::HLERequestContext& ctx);
@@ -234,6 +246,7 @@ private:
     void EndVrModeEx(Kernel::HLERequestContext& ctx);
     void GetDefaultDisplayResolution(Kernel::HLERequestContext& ctx);
     void SetCpuBoostMode(Kernel::HLERequestContext& ctx);
+    void PerformSystemButtonPressingIfInFocus(Kernel::HLERequestContext& ctx);
     void SetRequestExitToLibraryAppletAtExecuteNextProgramEnabled(Kernel::HLERequestContext& ctx);
 
     std::shared_ptr<AppletMessageQueue> msg_queue;

src/core/hle/service/am/applets/applet_controller.cpp

@@ -178,7 +178,7 @@ ResultCode Controller::GetStatus() const {
 }
 
 void Controller::ExecuteInteractive() {
-    UNREACHABLE_MSG("Attempted to call interactive execution on non-interactive applet.");
+    ASSERT_MSG(false, "Attempted to call interactive execution on non-interactive applet.");
 }
 
 void Controller::Execute() {

src/core/hle/service/am/applets/applet_error.cpp

@@ -156,7 +156,7 @@ ResultCode Error::GetStatus() const {
 }
 
 void Error::ExecuteInteractive() {
-    UNREACHABLE_MSG("Unexpected interactive applet data!");
+    ASSERT_MSG(false, "Unexpected interactive applet data!");
 }
 
 void Error::Execute() {

src/core/hle/service/am/applets/applet_general_backend.cpp

@@ -76,7 +76,7 @@ ResultCode Auth::GetStatus() const {
 }
 
 void Auth::ExecuteInteractive() {
-    UNREACHABLE_MSG("Unexpected interactive applet data.");
+    ASSERT_MSG(false, "Unexpected interactive applet data.");
 }
 
 void Auth::Execute() {
@@ -175,7 +175,7 @@ ResultCode PhotoViewer::GetStatus() const {
 }
 
 void PhotoViewer::ExecuteInteractive() {
-    UNREACHABLE_MSG("Unexpected interactive applet data.");
+    ASSERT_MSG(false, "Unexpected interactive applet data.");
 }
 
 void PhotoViewer::Execute() {

src/core/hle/service/am/applets/applet_mii_edit.cpp

@@ -67,7 +67,7 @@ ResultCode MiiEdit::GetStatus() const {
 }
 
 void MiiEdit::ExecuteInteractive() {
-    UNREACHABLE_MSG("Attempted to call interactive execution on non-interactive applet.");
+    ASSERT_MSG(false, "Attempted to call interactive execution on non-interactive applet.");
 }
 
 void MiiEdit::Execute() {

src/core/hle/service/am/applets/applet_profile_select.cpp

@@ -44,7 +44,7 @@ ResultCode ProfileSelect::GetStatus() const {
 }
 
 void ProfileSelect::ExecuteInteractive() {
-    UNREACHABLE_MSG("Attempted to call interactive execution on non-interactive applet.");
+    ASSERT_MSG(false, "Attempted to call interactive execution on non-interactive applet.");
 }
 
 void ProfileSelect::Execute() {

src/core/hle/service/am/applets/applet_software_keyboard.cpp

@@ -71,7 +71,7 @@ void SoftwareKeyboard::Initialize() {
         InitializeBackground(applet_mode);
         break;
     default:
-        UNREACHABLE_MSG("Invalid LibraryAppletMode={}", applet_mode);
+        ASSERT_MSG(false, "Invalid LibraryAppletMode={}", applet_mode);
         break;
     }
 }

src/core/hle/service/am/applets/applet_web_browser.cpp

@@ -279,7 +279,7 @@ void WebBrowser::Initialize() {
         InitializeLobby();
         break;
     default:
-        UNREACHABLE_MSG("Invalid ShimKind={}", web_arg_header.shim_kind);
+        ASSERT_MSG(false, "Invalid ShimKind={}", web_arg_header.shim_kind);
         break;
     }
 }
@@ -320,7 +320,7 @@ void WebBrowser::Execute() {
         ExecuteLobby();
         break;
     default:
-        UNREACHABLE_MSG("Invalid ShimKind={}", web_arg_header.shim_kind);
+        ASSERT_MSG(false, "Invalid ShimKind={}", web_arg_header.shim_kind);
         WebBrowserExit(WebExitReason::EndButtonPressed);
         break;
     }

src/core/hle/service/filesystem/fsp_srv.cpp

@@ -899,7 +899,7 @@ void FSP_SRV::OpenSaveDataFileSystem(Kernel::HLERequestContext& ctx) {
     case FileSys::SaveDataSpaceId::TemporaryStorage:
     case FileSys::SaveDataSpaceId::ProperSystem:
     case FileSys::SaveDataSpaceId::SafeMode:
-        UNREACHABLE();
+        ASSERT(false);
     }
 
     auto filesystem = std::make_shared<IFileSystem>(system, std::move(dir.Unwrap()),

src/core/hle/service/glue/notif.cpp

@@ -1,6 +1,11 @@
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include <algorithm>
+#include <cstring>
+
+#include "common/assert.h"
+#include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/service/glue/notif.h"
 
@@ -9,11 +14,11 @@ namespace Service::Glue {
 NOTIF_A::NOTIF_A(Core::System& system_) : ServiceFramework{system_, "notif:a"} {
     // clang-format off
     static const FunctionInfo functions[] = {
-        {500, nullptr, "RegisterAlarmSetting"},
-        {510, nullptr, "UpdateAlarmSetting"},
+        {500, &NOTIF_A::RegisterAlarmSetting, "RegisterAlarmSetting"},
+        {510, &NOTIF_A::UpdateAlarmSetting, "UpdateAlarmSetting"},
         {520, &NOTIF_A::ListAlarmSettings, "ListAlarmSettings"},
-        {530, nullptr, "LoadApplicationParameter"},
-        {540, nullptr, "DeleteAlarmSetting"},
+        {530, &NOTIF_A::LoadApplicationParameter, "LoadApplicationParameter"},
+        {540, &NOTIF_A::DeleteAlarmSetting, "DeleteAlarmSetting"},
         {1000, &NOTIF_A::Initialize, "Initialize"},
     };
     // clang-format on
@@ -23,21 +28,132 @@ NOTIF_A::NOTIF_A(Core::System& system_) : ServiceFramework{system_, "notif:a"} {
 
 NOTIF_A::~NOTIF_A() = default;
 
-void NOTIF_A::ListAlarmSettings(Kernel::HLERequestContext& ctx) {
-    // Returns an array of AlarmSetting
-    constexpr s32 alarm_count = 0;
+void NOTIF_A::RegisterAlarmSetting(Kernel::HLERequestContext& ctx) {
+    const auto alarm_setting_buffer_size = ctx.GetReadBufferSize(0);
+    const auto application_parameter_size = ctx.GetReadBufferSize(1);
 
-    LOG_WARNING(Service_NOTIF, "(STUBBED) called");
+    ASSERT_MSG(alarm_setting_buffer_size == sizeof(AlarmSetting),
+               "alarm_setting_buffer_size is not 0x40 bytes");
+    ASSERT_MSG(application_parameter_size <= sizeof(ApplicationParameter),
+               "application_parameter_size is bigger than 0x400 bytes");
+
+    AlarmSetting new_alarm{};
+    memcpy(&new_alarm, ctx.ReadBuffer(0).data(), sizeof(AlarmSetting));
+
+    // TODO: Count alarms per game id
+    if (alarms.size() >= max_alarms) {
+        LOG_ERROR(Service_NOTIF, "Alarm limit reached");
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultUnknown);
+        return;
+    }
+
+    new_alarm.alarm_setting_id = last_alarm_setting_id++;
+    alarms.push_back(new_alarm);
+
+    // TODO: Save application parameter data
+
+    LOG_WARNING(Service_NOTIF,
+                "(STUBBED) called, application_parameter_size={}, setting_id={}, kind={}, muted={}",
+                application_parameter_size, new_alarm.alarm_setting_id, new_alarm.kind,
+                new_alarm.muted);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+    rb.Push(new_alarm.alarm_setting_id);
+}
+
+void NOTIF_A::UpdateAlarmSetting(Kernel::HLERequestContext& ctx) {
+    const auto alarm_setting_buffer_size = ctx.GetReadBufferSize(0);
+    const auto application_parameter_size = ctx.GetReadBufferSize(1);
+
+    ASSERT_MSG(alarm_setting_buffer_size == sizeof(AlarmSetting),
+               "alarm_setting_buffer_size is not 0x40 bytes");
+    ASSERT_MSG(application_parameter_size <= sizeof(ApplicationParameter),
+               "application_parameter_size is bigger than 0x400 bytes");
+
+    AlarmSetting alarm_setting{};
+    memcpy(&alarm_setting, ctx.ReadBuffer(0).data(), sizeof(AlarmSetting));
+
+    const auto alarm_it = GetAlarmFromId(alarm_setting.alarm_setting_id);
+    if (alarm_it != alarms.end()) {
+        LOG_DEBUG(Service_NOTIF, "Alarm updated");
+        *alarm_it = alarm_setting;
+        // TODO: Save application parameter data
+    }
+
+    LOG_WARNING(Service_NOTIF,
+                "(STUBBED) called, application_parameter_size={}, setting_id={}, kind={}, muted={}",
+                application_parameter_size, alarm_setting.alarm_setting_id, alarm_setting.kind,
+                alarm_setting.muted);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+}
+
+void NOTIF_A::ListAlarmSettings(Kernel::HLERequestContext& ctx) {
+    LOG_INFO(Service_NOTIF, "called, alarm_count={}", alarms.size());
+
+    // TODO: Only return alarms of this game id
+    ctx.WriteBuffer(alarms);
 
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(ResultSuccess);
-    rb.Push(alarm_count);
+    rb.Push(static_cast<u32>(alarms.size()));
+}
+
+void NOTIF_A::LoadApplicationParameter(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto alarm_setting_id{rp.Pop<AlarmSettingId>()};
+
+    const auto alarm_it = GetAlarmFromId(alarm_setting_id);
+    if (alarm_it == alarms.end()) {
+        LOG_ERROR(Service_NOTIF, "Invalid alarm setting id={}", alarm_setting_id);
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(ResultUnknown);
+        return;
+    }
+
+    // TODO: Read application parameter related to this setting id
+    ApplicationParameter application_parameter{};
+
+    LOG_WARNING(Service_NOTIF, "(STUBBED) called, alarm_setting_id={}", alarm_setting_id);
+
+    ctx.WriteBuffer(application_parameter);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
+    rb.Push(static_cast<u32>(application_parameter.size()));
+}
+
+void NOTIF_A::DeleteAlarmSetting(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto alarm_setting_id{rp.Pop<AlarmSettingId>()};
+
+    std::erase_if(alarms, [alarm_setting_id](const AlarmSetting& alarm) {
+        return alarm.alarm_setting_id == alarm_setting_id;
+    });
+
+    LOG_INFO(Service_NOTIF, "called, alarm_setting_id={}", alarm_setting_id);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(ResultSuccess);
 }
 
 void NOTIF_A::Initialize(Kernel::HLERequestContext& ctx) {
+    // TODO: Load previous alarms from config
+
     LOG_WARNING(Service_NOTIF, "(STUBBED) called");
     IPC::ResponseBuilder rb{ctx, 2};
     rb.Push(ResultSuccess);
 }
 
+std::vector<NOTIF_A::AlarmSetting>::iterator NOTIF_A::GetAlarmFromId(
+    AlarmSettingId alarm_setting_id) {
+    return std::find_if(alarms.begin(), alarms.end(),
+                        [alarm_setting_id](const AlarmSetting& alarm) {
+                            return alarm.alarm_setting_id == alarm_setting_id;
+                        });
+}
+
 } // namespace Service::Glue

src/core/hle/service/glue/notif.h

@@ -3,6 +3,10 @@
 
 #pragma once
 
+#include <array>
+#include <vector>
+
+#include "common/uuid.h"
 #include "core/hle/service/service.h"
 
 namespace Core {
@@ -17,8 +21,52 @@ public:
     ~NOTIF_A() override;
 
 private:
+    static constexpr std::size_t max_alarms = 8;
+
+    // This is nn::notification::AlarmSettingId
+    using AlarmSettingId = u16;
+    static_assert(sizeof(AlarmSettingId) == 0x2, "AlarmSettingId is an invalid size");
+
+    using ApplicationParameter = std::array<u8, 0x400>;
+    static_assert(sizeof(ApplicationParameter) == 0x400, "ApplicationParameter is an invalid size");
+
+    struct DailyAlarmSetting {
+        s8 hour;
+        s8 minute;
+    };
+    static_assert(sizeof(DailyAlarmSetting) == 0x2, "DailyAlarmSetting is an invalid size");
+
+    struct WeeklyScheduleAlarmSetting {
+        INSERT_PADDING_BYTES(0xA);
+        std::array<DailyAlarmSetting, 0x7> day_of_week;
+    };
+    static_assert(sizeof(WeeklyScheduleAlarmSetting) == 0x18,
+                  "WeeklyScheduleAlarmSetting is an invalid size");
+
+    // This is nn::notification::AlarmSetting
+    struct AlarmSetting {
+        AlarmSettingId alarm_setting_id;
+        u8 kind;
+        u8 muted;
+        INSERT_PADDING_BYTES(0x4);
+        Common::UUID account_id;
+        u64 application_id;
+        INSERT_PADDING_BYTES(0x8);
+        WeeklyScheduleAlarmSetting schedule;
+    };
+    static_assert(sizeof(AlarmSetting) == 0x40, "AlarmSetting is an invalid size");
+
+    void RegisterAlarmSetting(Kernel::HLERequestContext& ctx);
+    void UpdateAlarmSetting(Kernel::HLERequestContext& ctx);
     void ListAlarmSettings(Kernel::HLERequestContext& ctx);
+    void LoadApplicationParameter(Kernel::HLERequestContext& ctx);
+    void DeleteAlarmSetting(Kernel::HLERequestContext& ctx);
     void Initialize(Kernel::HLERequestContext& ctx);
+
+    std::vector<AlarmSetting>::iterator GetAlarmFromId(AlarmSettingId alarm_setting_id);
+
+    std::vector<AlarmSetting> alarms{};
+    AlarmSettingId last_alarm_setting_id{};
 };
 
 } // namespace Service::Glue

src/core/hle/service/hid/controllers/gesture.cpp

@@ -61,6 +61,7 @@ void Controller_Gesture::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
     }
 
     last_update_timestamp = shared_memory->gesture_lifo.timestamp;
+    UpdateGestureSharedMemory(gesture, time_difference);
 }
 
 void Controller_Gesture::ReadTouchInput() {
@@ -94,8 +95,7 @@ bool Controller_Gesture::ShouldUpdateGesture(const GestureProperties& gesture,
     return false;
 }
 
-void Controller_Gesture::UpdateGestureSharedMemory(u8* data, std::size_t size,
-                                                   GestureProperties& gesture,
+void Controller_Gesture::UpdateGestureSharedMemory(GestureProperties& gesture,
                                                    f32 time_difference) {
     GestureType type = GestureType::Idle;
     GestureAttribute attributes{};

src/core/hle/service/hid/controllers/gesture.h

@@ -107,8 +107,7 @@ private:
     bool ShouldUpdateGesture(const GestureProperties& gesture, f32 time_difference);
 
     // Updates the shared memory to the next state
-    void UpdateGestureSharedMemory(u8* data, std::size_t size, GestureProperties& gesture,
-                                   f32 time_difference);
+    void UpdateGestureSharedMemory(GestureProperties& gesture, f32 time_difference);
 
     // Initializes new gesture
     void NewGesture(GestureProperties& gesture, GestureType& type, GestureAttribute& attributes);

src/core/hle/service/hid/controllers/npad.cpp

@@ -160,7 +160,7 @@ void Controller_NPad::InitNewlyAddedController(Core::HID::NpadIdType npad_id) {
     shared_memory->system_properties.raw = 0;
     switch (controller_type) {
     case Core::HID::NpadStyleIndex::None:
-        UNREACHABLE();
+        ASSERT(false);
         break;
     case Core::HID::NpadStyleIndex::ProController:
         shared_memory->style_tag.fullkey.Assign(1);
@@ -422,7 +422,7 @@ void Controller_NPad::OnUpdate(const Core::Timing::CoreTiming& core_timing) {
         libnx_state.connection_status.is_connected.Assign(1);
         switch (controller_type) {
         case Core::HID::NpadStyleIndex::None:
-            UNREACHABLE();
+            ASSERT(false);
             break;
         case Core::HID::NpadStyleIndex::ProController:
         case Core::HID::NpadStyleIndex::NES:
@@ -597,7 +597,7 @@ void Controller_NPad::OnMotionUpdate(const Core::Timing::CoreTiming& core_timing
 
         switch (controller_type) {
         case Core::HID::NpadStyleIndex::None:
-            UNREACHABLE();
+            ASSERT(false);
             break;
         case Core::HID::NpadStyleIndex::ProController:
         case Core::HID::NpadStyleIndex::Pokeball:
@@ -856,7 +856,7 @@ void Controller_NPad::VibrateController(
     }
 
     if (vibration_device_handle.device_index == Core::HID::DeviceIndex::None) {
-        UNREACHABLE_MSG("DeviceIndex should never be None!");
+        ASSERT_MSG(false, "DeviceIndex should never be None!");
         return;
     }
 

src/core/hle/service/hid/hid.cpp

@@ -1441,7 +1441,7 @@ void Hid::GetVibrationDeviceInfo(Kernel::HLERequestContext& ctx) {
         break;
     case Core::HID::DeviceIndex::None:
     default:
-        UNREACHABLE_MSG("DeviceIndex should never be None!");
+        ASSERT_MSG(false, "DeviceIndex should never be None!");
         vibration_device_info.position = Core::HID::VibrationDevicePosition::None;
         break;
     }

src/core/hle/service/ldr/ldr.cpp

@@ -347,7 +347,7 @@ public:
         }
 
         if (!succeeded) {
-            UNREACHABLE_MSG("Out of address space!");
+            ASSERT_MSG(false, "Out of address space!");
             return Kernel::ResultOutOfMemory;
         }
 

src/core/hle/service/mii/mii_manager.cpp

@@ -290,7 +290,7 @@ MiiStoreData BuildRandomStoreData(Age age, Gender gender, Race race, const Commo
     u8 glasses_type{};
     while (glasses_type_start < glasses_type_info.values[glasses_type]) {
         if (++glasses_type >= glasses_type_info.values_count) {
-            UNREACHABLE();
+            ASSERT(false);
             break;
         }
     }

src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp

@@ -150,9 +150,9 @@ NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector
     event.event->GetWritableEvent().Clear();
     if (events_interface.failed[event_id]) {
         {
-            auto lk = system.StallCPU();
+            auto lk = system.StallProcesses();
             gpu.WaitFence(params.syncpt_id, target_value);
-            system.UnstallCPU();
+            system.UnstallProcesses();
         }
         std::memcpy(output.data(), &params, sizeof(params));
         events_interface.failed[event_id] = false;

src/core/hle/service/nvdrv/syncpoint_manager.cpp

@@ -23,7 +23,7 @@ u32 SyncpointManager::AllocateSyncpoint() {
             return syncpoint_id;
         }
     }
-    UNREACHABLE_MSG("No more available syncpoints!");
+    ASSERT_MSG(false, "No more available syncpoints!");
     return {};
 }
 

src/core/hle/service/nvflinger/buffer_queue_producer.cpp

@@ -659,7 +659,7 @@ Status BufferQueueProducer::Query(NativeWindow what, s32* out_value) {
         value = core->consumer_usage_bit;
         break;
     default:
-        UNREACHABLE();
+        ASSERT(false);
         return Status::BadValue;
     }
 

src/core/hle/service/time/standard_user_system_clock_core.cpp

@@ -48,12 +48,12 @@ ResultCode StandardUserSystemClockCore::GetClockContext(Core::System& system,
 }
 
 ResultCode StandardUserSystemClockCore::Flush(const SystemClockContext&) {
-    UNREACHABLE();
+    UNIMPLEMENTED();
     return ERROR_NOT_IMPLEMENTED;
 }
 
 ResultCode StandardUserSystemClockCore::SetClockContext(const SystemClockContext&) {
-    UNREACHABLE();
+    UNIMPLEMENTED();
     return ERROR_NOT_IMPLEMENTED;
 }
 

src/core/hle/service/time/time_manager.cpp

@@ -111,7 +111,7 @@ struct TimeManager::Impl final {
                               FileSys::VirtualFile& vfs_file) {
         if (time_zone_content_manager.GetTimeZoneManager().SetDeviceLocationNameWithTimeZoneRule(
                 location_name, vfs_file) != ResultSuccess) {
-            UNREACHABLE();
+            ASSERT(false);
             return;
         }
 
@@ -155,7 +155,7 @@ struct TimeManager::Impl final {
         } else {
             if (standard_local_system_clock_core.SetCurrentTime(system_, posix_time) !=
                 ResultSuccess) {
-                UNREACHABLE();
+                ASSERT(false);
                 return;
             }
         }
@@ -170,7 +170,7 @@ struct TimeManager::Impl final {
 
         if (standard_network_system_clock_core.SetSystemClockContext(clock_context) !=
             ResultSuccess) {
-            UNREACHABLE();
+            ASSERT(false);
             return;
         }
 
@@ -183,7 +183,7 @@ struct TimeManager::Impl final {
                                       Clock::SteadyClockTimePoint steady_clock_time_point) {
         if (standard_user_system_clock_core.SetAutomaticCorrectionEnabled(
                 system_, is_automatic_correction_enabled) != ResultSuccess) {
-            UNREACHABLE();
+            ASSERT(false);
             return;
         }
 
@@ -203,7 +203,7 @@ struct TimeManager::Impl final {
         if (GetStandardLocalSystemClockCore()
                 .SetCurrentTime(system_, timespan.ToSeconds())
                 .IsError()) {
-            UNREACHABLE();
+            ASSERT(false);
             return;
         }
     }

src/core/hle/service/time/time_zone_manager.cpp

@@ -279,7 +279,7 @@ static constexpr int TransitionTime(int year, Rule rule, int offset) {
         break;
     }
     default:
-        UNREACHABLE();
+        ASSERT(false);
     }
     return value + rule.transition_time + offset;
 }

src/core/loader/nso.cpp

@@ -128,11 +128,10 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::KProcess& process, Core::
 
     // Apply patches if necessary
     if (pm && (pm->HasNSOPatch(nso_header.build_id) || Settings::values.dump_nso)) {
-        std::vector<u8> pi_header;
-        pi_header.insert(pi_header.begin(), reinterpret_cast<u8*>(&nso_header),
-                         reinterpret_cast<u8*>(&nso_header) + sizeof(NSOHeader));
-        pi_header.insert(pi_header.begin() + sizeof(NSOHeader), program_image.data(),
-                         program_image.data() + program_image.size());
+        std::vector<u8> pi_header(sizeof(NSOHeader) + program_image.size());
+        std::memcpy(pi_header.data(), &nso_header, sizeof(NSOHeader));
+        std::memcpy(pi_header.data() + sizeof(NSOHeader), program_image.data(),
+                    program_image.size());
 
         pi_header = pm->PatchNSO(pi_header, nso_file.GetName());
 

src/core/memory.cpp

@@ -67,6 +67,16 @@ struct Memory::Impl {
         return system.DeviceMemory().GetPointer(paddr) + vaddr;
     }
 
+    [[nodiscard]] u8* GetPointerFromDebugMemory(VAddr vaddr) const {
+        const PAddr paddr{current_page_table->backing_addr[vaddr >> PAGE_BITS]};
+
+        if (paddr == 0) {
+            return {};
+        }
+
+        return system.DeviceMemory().GetPointer(paddr) + vaddr;
+    }
+
     u8 Read8(const VAddr addr) {
         return Read<u8>(addr);
     }
@@ -187,6 +197,12 @@ struct Memory::Impl {
                 on_memory(copy_amount, mem_ptr);
                 break;
             }
+            case Common::PageType::DebugMemory: {
+                DEBUG_ASSERT(pointer);
+                u8* const mem_ptr{GetPointerFromDebugMemory(current_vaddr)};
+                on_memory(copy_amount, mem_ptr);
+                break;
+            }
             case Common::PageType::RasterizerCachedMemory: {
                 u8* const host_ptr{GetPointerFromRasterizerCachedMemory(current_vaddr)};
                 on_rasterizer(current_vaddr, copy_amount, host_ptr);
@@ -316,6 +332,58 @@ struct Memory::Impl {
             });
     }
 
+    void MarkRegionDebug(VAddr vaddr, u64 size, bool debug) {
+        if (vaddr == 0) {
+            return;
+        }
+
+        // Iterate over a contiguous CPU address space, marking/unmarking the region.
+        // The region is at a granularity of CPU pages.
+
+        const u64 num_pages = ((vaddr + size - 1) >> PAGE_BITS) - (vaddr >> PAGE_BITS) + 1;
+        for (u64 i = 0; i < num_pages; ++i, vaddr += PAGE_SIZE) {
+            const Common::PageType page_type{
+                current_page_table->pointers[vaddr >> PAGE_BITS].Type()};
+            if (debug) {
+                // Switch page type to debug if now debug
+                switch (page_type) {
+                case Common::PageType::Unmapped:
+                    ASSERT_MSG(false, "Attempted to mark unmapped pages as debug");
+                    break;
+                case Common::PageType::RasterizerCachedMemory:
+                case Common::PageType::DebugMemory:
+                    // Page is already marked.
+                    break;
+                case Common::PageType::Memory:
+                    current_page_table->pointers[vaddr >> PAGE_BITS].Store(
+                        nullptr, Common::PageType::DebugMemory);
+                    break;
+                default:
+                    UNREACHABLE();
+                }
+            } else {
+                // Switch page type to non-debug if now non-debug
+                switch (page_type) {
+                case Common::PageType::Unmapped:
+                    ASSERT_MSG(false, "Attempted to mark unmapped pages as non-debug");
+                    break;
+                case Common::PageType::RasterizerCachedMemory:
+                case Common::PageType::Memory:
+                    // Don't mess with already non-debug or rasterizer memory.
+                    break;
+                case Common::PageType::DebugMemory: {
+                    u8* const pointer{GetPointerFromDebugMemory(vaddr & ~PAGE_MASK)};
+                    current_page_table->pointers[vaddr >> PAGE_BITS].Store(
+                        pointer - (vaddr & ~PAGE_MASK), Common::PageType::Memory);
+                    break;
+                }
+                default:
+                    UNREACHABLE();
+                }
+            }
+        }
+    }
+
     void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
         if (vaddr == 0) {
             return;
@@ -342,6 +410,7 @@ struct Memory::Impl {
                     // It is not necessary for a process to have this region mapped into its address
                     // space, for example, a system module need not have a VRAM mapping.
                     break;
+                case Common::PageType::DebugMemory:
                 case Common::PageType::Memory:
                     current_page_table->pointers[vaddr >> PAGE_BITS].Store(
                         nullptr, Common::PageType::RasterizerCachedMemory);
@@ -360,6 +429,7 @@ struct Memory::Impl {
                     // It is not necessary for a process to have this region mapped into its address
                     // space, for example, a system module need not have a VRAM mapping.
                     break;
+                case Common::PageType::DebugMemory:
                 case Common::PageType::Memory:
                     // There can be more than one GPU region mapped per CPU region, so it's common
                     // that this area is already unmarked as cached.
@@ -460,6 +530,8 @@ struct Memory::Impl {
         case Common::PageType::Memory:
             ASSERT_MSG(false, "Mapped memory page without a pointer @ 0x{:016X}", vaddr);
             return nullptr;
+        case Common::PageType::DebugMemory:
+            return GetPointerFromDebugMemory(vaddr);
         case Common::PageType::RasterizerCachedMemory: {
             u8* const host_ptr{GetPointerFromRasterizerCachedMemory(vaddr)};
             on_rasterizer();
@@ -591,7 +663,8 @@ bool Memory::IsValidVirtualAddress(const VAddr vaddr) const {
         return false;
     }
     const auto [pointer, type] = page_table.pointers[page].PointerType();
-    return pointer != nullptr || type == Common::PageType::RasterizerCachedMemory;
+    return pointer != nullptr || type == Common::PageType::RasterizerCachedMemory ||
+           type == Common::PageType::DebugMemory;
 }
 
 bool Memory::IsValidVirtualAddressRange(VAddr base, u64 size) const {
@@ -707,4 +780,8 @@ void Memory::RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
     impl->RasterizerMarkRegionCached(vaddr, size, cached);
 }
 
+void Memory::MarkRegionDebug(VAddr vaddr, u64 size, bool debug) {
+    impl->MarkRegionDebug(vaddr, size, debug);
+}
+
 } // namespace Core::Memory

src/core/memory.h

@@ -446,6 +446,17 @@ public:
      */
     void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached);
 
+    /**
+     * Marks each page within the specified address range as debug or non-debug.
+     * Debug addresses are not accessible from fastmem pointers.
+     *
+     * @param vaddr The virtual address indicating the start of the address range.
+     * @param size  The size of the address range in bytes.
+     * @param debug Whether or not any pages within the address range should be
+     *              marked as debug or non-debug.
+     */
+    void MarkRegionDebug(VAddr vaddr, u64 size, bool debug);
+
 private:
     Core::System& system;