jit: add support for relr-type relocations

Do not allocate DeviceLocal buffers as mapped

dist/qt_themes/default/style.qss

@@ -120,6 +120,10 @@ QWidget#connectedControllers {
     background: transparent;
 }
 
+QWidget#closeButtons {
+    background: transparent;
+}
+
 QWidget#playersSupported,
 QWidget#controllersSupported,
 QWidget#controllerSupported1,

dist/qt_themes/qdarkstyle/style.qss

@@ -1380,6 +1380,10 @@ QWidget#connectedControllers {
     background: transparent;
 }
 
+QWidget#closeButtons {
+    background: transparent;
+}
+
 QWidget#playersSupported,
 QWidget#controllersSupported,
 QWidget#controllerSupported1,

dist/qt_themes/qdarkstyle_midnight_blue/style.qss

@@ -2305,6 +2305,10 @@ QWidget#connectedControllers {
   background: transparent;
 }
 
+QWidget#closeButtons {
+    background: transparent;
+}
+
 QWidget#playersSupported,
 QWidget#controllersSupported,
 QWidget#controllerSupported1,

src/audio_core/adsp/apps/audio_renderer/audio_renderer.cpp

@@ -77,6 +77,7 @@ void AudioRenderer::Wait() {
                   "{}, got {}",
                   Message::RenderResponse, msg);
     }
+    PostDSPClearCommandBuffer();
 }
 
 void AudioRenderer::Send(Direction dir, u32 message) {
@@ -96,6 +97,14 @@ void AudioRenderer::SetCommandBuffer(s32 session_id, CpuAddr buffer, u64 size, u
     command_buffers[session_id].reset_buffer = reset;
 }
 
+void AudioRenderer::PostDSPClearCommandBuffer() noexcept {
+    for (auto& buffer : command_buffers) {
+        buffer.buffer = 0;
+        buffer.size = 0;
+        buffer.reset_buffer = false;
+    }
+}
+
 u32 AudioRenderer::GetRemainCommandCount(s32 session_id) const noexcept {
     return command_buffers[session_id].remaining_command_count;
 }

src/audio_core/adsp/apps/audio_renderer/audio_renderer.h

@@ -85,6 +85,8 @@ private:
      */
     void CreateSinkStreams();
 
+    void PostDSPClearCommandBuffer() noexcept;
+
     /// Core system
     Core::System& system;
     /// The output sink the AudioRenderer will send samples to

src/common/elf.h

@@ -211,6 +211,11 @@ struct Elf64_Rela {
     Elf64_Sxword r_addend; /* Addend */
 };
 
+/* RELR relocation table entry */
+
+using Elf32_Relr = Elf32_Word;
+using Elf64_Relr = Elf64_Xword;
+
 /* How to extract and insert information held in the r_info field.  */
 
 static inline u32 Elf32RelSymIndex(Elf32_Word r_info) {
@@ -328,6 +333,9 @@ constexpr u32 ElfDtFiniArray = 26;   /* Array with addresses of fini fct */
 constexpr u32 ElfDtInitArraySz = 27; /* Size in bytes of DT_INIT_ARRAY */
 constexpr u32 ElfDtFiniArraySz = 28; /* Size in bytes of DT_FINI_ARRAY */
 constexpr u32 ElfDtSymtabShndx = 34; /* Address of SYMTAB_SHNDX section */
+constexpr u32 ElfDtRelrsz = 35;      /* Size of RELR relative relocations */
+constexpr u32 ElfDtRelr = 36;        /* Address of RELR relative relocations */
+constexpr u32 ElfDtRelrent = 37;     /* Size of one RELR relative relocation */
 
 } // namespace ELF
 } // namespace Common

src/core/hle/service/jit/jit_context.cpp

@@ -156,6 +156,8 @@ public:
 
     bool LoadNRO(std::span<const u8> data) {
         local_memory.clear();
+
+        relocbase = local_memory.size();
         local_memory.insert(local_memory.end(), data.begin(), data.end());
 
         if (FixupRelocations()) {
@@ -181,8 +183,8 @@ public:
         // https://refspecs.linuxbase.org/elf/gabi4+/ch5.dynamic.html
         // https://refspecs.linuxbase.org/elf/gabi4+/ch4.reloc.html
         VAddr dynamic_offset{mod_offset + callbacks->MemoryRead32(mod_offset + 4)};
-        VAddr rela_dyn = 0;
-        size_t num_rela = 0;
+        VAddr rela_dyn = 0, relr_dyn = 0;
+        size_t num_rela = 0, num_relr = 0;
         while (true) {
             const auto dyn{callbacks->ReadMemory<Elf64_Dyn>(dynamic_offset)};
             dynamic_offset += sizeof(Elf64_Dyn);
@@ -196,6 +198,12 @@ public:
             if (dyn.d_tag == ElfDtRelasz) {
                 num_rela = dyn.d_un.d_val / sizeof(Elf64_Rela);
             }
+            if (dyn.d_tag == ElfDtRelr) {
+                relr_dyn = dyn.d_un.d_ptr;
+            }
+            if (dyn.d_tag == ElfDtRelrsz) {
+                num_relr = dyn.d_un.d_val / sizeof(Elf64_Relr);
+            }
         }
 
         for (size_t i = 0; i < num_rela; i++) {
@@ -207,6 +215,29 @@ public:
             callbacks->MemoryWrite64(rela.r_offset, contents + rela.r_addend);
         }
 
+        VAddr relr_where = 0;
+        for (size_t i = 0; i < num_relr; i++) {
+            const auto relr{callbacks->ReadMemory<Elf64_Relr>(relr_dyn + i * sizeof(Elf64_Relr))};
+            const auto incr{[&](VAddr where) {
+                callbacks->MemoryWrite64(where, callbacks->MemoryRead64(where) + relocbase);
+            }};
+
+            if ((relr & 1) == 0) {
+                // where pointer
+                relr_where = relocbase + relr;
+                incr(relr_where);
+                relr_where += sizeof(Elf64_Addr);
+            } else {
+                // bitmap
+                for (int bit = 1; bit < 64; bit++) {
+                    if ((relr & (1ULL << bit)) != 0) {
+                        incr(relr_where + i * sizeof(Elf64_Addr));
+                    }
+                }
+                relr_where += 63 * sizeof(Elf64_Addr);
+            }
+        }
+
         return true;
     }
 
@@ -313,6 +344,7 @@ public:
     Core::Memory::Memory& memory;
     VAddr top_of_stack;
     VAddr heap_pointer;
+    VAddr relocbase;
 };
 
 void DynarmicCallbacks64::CallSVC(u32 swi) {

src/video_core/buffer_cache/buffer_cache.h

@@ -544,7 +544,7 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
         it++;
     }
 
-    boost::container::small_vector<std::pair<BufferCopy, BufferId>, 1> downloads;
+    boost::container::small_vector<std::pair<BufferCopy, BufferId>, 16> downloads;
     u64 total_size_bytes = 0;
     u64 largest_copy = 0;
     for (const IntervalSet& intervals : committed_ranges) {
@@ -914,6 +914,11 @@ void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
 
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const bool is_written = ((channel_state->written_storage_buffers[stage] >> index) & 1) != 0;
+
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         if constexpr (NEEDS_BIND_STORAGE_INDEX) {
             runtime.BindStorageBuffer(stage, binding_index, buffer, offset, size, is_written);
             ++binding_index;
@@ -931,6 +936,11 @@ void BufferCache<P>::BindHostGraphicsTextureBuffers(size_t stage) {
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
+        const bool is_written = ((channel_state->written_texture_buffers[stage] >> index) & 1) != 0;
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const PixelFormat format = binding.format;
         if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
@@ -962,6 +972,8 @@ void BufferCache<P>::BindHostTransformFeedbackBuffers() {
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
+        MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+
         const u32 offset = buffer.Offset(binding.cpu_addr);
         host_bindings.buffers.push_back(&buffer);
         host_bindings.offsets.push_back(offset);
@@ -1011,6 +1023,11 @@ void BufferCache<P>::BindHostComputeStorageBuffers() {
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const bool is_written =
             ((channel_state->written_compute_storage_buffers >> index) & 1) != 0;
+
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         if constexpr (NEEDS_BIND_STORAGE_INDEX) {
             runtime.BindComputeStorageBuffer(binding_index, buffer, offset, size, is_written);
             ++binding_index;
@@ -1028,6 +1045,12 @@ void BufferCache<P>::BindHostComputeTextureBuffers() {
         const u32 size = binding.size;
         SynchronizeBuffer(buffer, binding.cpu_addr, size);
 
+        const bool is_written =
+            ((channel_state->written_compute_texture_buffers >> index) & 1) != 0;
+        if (is_written) {
+            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, size);
+        }
+
         const u32 offset = buffer.Offset(binding.cpu_addr);
         const PixelFormat format = binding.format;
         if constexpr (SEPARATE_IMAGE_BUFFERS_BINDINGS) {
@@ -1201,16 +1224,11 @@ void BufferCache<P>::UpdateUniformBuffers(size_t stage) {
 
 template <class P>
 void BufferCache<P>::UpdateStorageBuffers(size_t stage) {
-    const u32 written_mask = channel_state->written_storage_buffers[stage];
     ForEachEnabledBit(channel_state->enabled_storage_buffers[stage], [&](u32 index) {
         // Resolve buffer
         Binding& binding = channel_state->storage_buffers[stage][index];
         const BufferId buffer_id = FindBuffer(binding.cpu_addr, binding.size);
         binding.buffer_id = buffer_id;
-        // Mark buffer as written if needed
-        if (((written_mask >> index) & 1) != 0) {
-            MarkWrittenBuffer(buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
@@ -1219,10 +1237,6 @@ void BufferCache<P>::UpdateTextureBuffers(size_t stage) {
     ForEachEnabledBit(channel_state->enabled_texture_buffers[stage], [&](u32 index) {
         Binding& binding = channel_state->texture_buffers[stage][index];
         binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
-        // Mark buffer as written if needed
-        if (((channel_state->written_texture_buffers[stage] >> index) & 1) != 0) {
-            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
@@ -1252,7 +1266,6 @@ void BufferCache<P>::UpdateTransformFeedbackBuffer(u32 index) {
         .size = size,
         .buffer_id = buffer_id,
     };
-    MarkWrittenBuffer(buffer_id, *cpu_addr, size);
 }
 
 template <class P>
@@ -1279,10 +1292,6 @@ void BufferCache<P>::UpdateComputeStorageBuffers() {
         // Resolve buffer
         Binding& binding = channel_state->compute_storage_buffers[index];
         binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
-        // Mark as written if needed
-        if (((channel_state->written_compute_storage_buffers >> index) & 1) != 0) {
-            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
@@ -1291,18 +1300,11 @@ void BufferCache<P>::UpdateComputeTextureBuffers() {
     ForEachEnabledBit(channel_state->enabled_compute_texture_buffers, [&](u32 index) {
         Binding& binding = channel_state->compute_texture_buffers[index];
         binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
-        // Mark as written if needed
-        if (((channel_state->written_compute_texture_buffers >> index) & 1) != 0) {
-            MarkWrittenBuffer(binding.buffer_id, binding.cpu_addr, binding.size);
-        }
     });
 }
 
 template <class P>
 void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 size) {
-    if (memory_tracker.IsRegionCpuModified(cpu_addr, size)) {
-        SynchronizeBuffer(slot_buffers[buffer_id], cpu_addr, size);
-    }
     memory_tracker.MarkRegionAsGpuModified(cpu_addr, size);
 
     const IntervalType base_interval{cpu_addr, cpu_addr + size};

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -1436,7 +1436,6 @@ void QueryCacheRuntime::Barriers(bool is_prebarrier) {
         .srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
         .dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
     };
-    impl->scheduler.RequestOutsideRenderPassOperationContext();
     if (is_prebarrier) {
         impl->scheduler.Record([](vk::CommandBuffer cmdbuf) {
             cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,

src/video_core/renderer_vulkan/vk_render_pass_cache.cpp

@@ -19,7 +19,7 @@ VkAttachmentDescription AttachmentDescription(const Device& device, PixelFormat
                                               VkSampleCountFlagBits samples) {
     using MaxwellToVK::SurfaceFormat;
     return {
-        .flags = {},
+        .flags = VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT,
         .format = SurfaceFormat(device, FormatType::Optimal, true, format).format,
         .samples = samples,
         .loadOp = VK_ATTACHMENT_LOAD_OP_LOAD,

src/video_core/texture_cache/formatter.cpp

@@ -10,23 +10,19 @@
 #include "video_core/texture_cache/image_info.h"
 #include "video_core/texture_cache/image_view_base.h"
 #include "video_core/texture_cache/render_targets.h"
-#include "video_core/texture_cache/samples_helper.h"
 
 namespace VideoCommon {
 
 std::string Name(const ImageBase& image) {
     const GPUVAddr gpu_addr = image.gpu_addr;
     const ImageInfo& info = image.info;
-    u32 width = info.size.width;
-    u32 height = info.size.height;
+    const u32 width = info.size.width;
+    const u32 height = info.size.height;
     const u32 depth = info.size.depth;
     const u32 num_layers = image.info.resources.layers;
     const u32 num_levels = image.info.resources.levels;
     std::string resource;
     if (image.info.num_samples > 1) {
-        const auto [samples_x, samples_y] = VideoCommon::SamplesLog2(image.info.num_samples);
-        width >>= samples_x;
-        height >>= samples_y;
         resource += fmt::format(":{}xMSAA", image.info.num_samples);
     }
     if (num_layers > 1) {

src/video_core/texture_cache/samples_helper.h

@@ -24,7 +24,7 @@ namespace VideoCommon {
         return {2, 2};
     }
     ASSERT_MSG(false, "Invalid number of samples={}", num_samples);
-    return {0, 0};
+    return {1, 1};
 }
 
 [[nodiscard]] inline int NumSamples(Tegra::Texture::MsaaMode msaa_mode) {

src/video_core/vulkan_common/vulkan_memory_allocator.cpp

@@ -66,9 +66,10 @@ struct Range {
     switch (usage) {
     case MemoryUsage::Upload:
     case MemoryUsage::Stream:
-        return VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
+        return VMA_ALLOCATION_CREATE_MAPPED_BIT |
+               VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT;
     case MemoryUsage::Download:
-        return VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT;
+        return VMA_ALLOCATION_CREATE_MAPPED_BIT | VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT;
     case MemoryUsage::DeviceLocal:
         return {};
     }
@@ -252,8 +253,7 @@ vk::Image MemoryAllocator::CreateImage(const VkImageCreateInfo& ci) const {
 
 vk::Buffer MemoryAllocator::CreateBuffer(const VkBufferCreateInfo& ci, MemoryUsage usage) const {
     const VmaAllocationCreateInfo alloc_ci = {
-        .flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | VMA_ALLOCATION_CREATE_MAPPED_BIT |
-                 MemoryUsageVmaFlags(usage),
+        .flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT | MemoryUsageVmaFlags(usage),
         .usage = MemoryUsageVma(usage),
         .requiredFlags = 0,
         .preferredFlags = MemoryUsagePreferedVmaFlags(usage),

src/yuzu/applets/qt_controller.cpp

@@ -23,6 +23,7 @@
 #include "yuzu/configuration/configure_vibration.h"
 #include "yuzu/configuration/input_profiles.h"
 #include "yuzu/main.h"
+#include "yuzu/util/controller_navigation.h"
 
 namespace {
 
@@ -132,6 +133,8 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
         ui->checkboxPlayer7Connected, ui->checkboxPlayer8Connected,
     };
 
+    ui->labelError->setVisible(false);
+
     // Setup/load everything prior to setting up connections.
     // This avoids unintentionally changing the states of elements while loading them in.
     SetSupportedControllers();
@@ -143,6 +146,8 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
 
     LoadConfiguration();
 
+    controller_navigation = new ControllerNavigation(system.HIDCore(), this);
+
     for (std::size_t i = 0; i < NUM_PLAYERS; ++i) {
         SetExplainText(i);
         UpdateControllerIcon(i);
@@ -151,6 +156,8 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
 
         connect(player_groupboxes[i], &QGroupBox::toggled, [this, i](bool checked) {
             if (checked) {
+                // Hide eventual error message about number of controllers
+                ui->labelError->setVisible(false);
                 for (std::size_t index = 0; index <= i; ++index) {
                     connected_controller_checkboxes[index]->setChecked(checked);
                 }
@@ -199,6 +206,12 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
     connect(ui->buttonBox, &QDialogButtonBox::accepted, this,
             &QtControllerSelectorDialog::ApplyConfiguration);
 
+    connect(controller_navigation, &ControllerNavigation::TriggerKeyboardEvent,
+            [this](Qt::Key key) {
+                QKeyEvent* event = new QKeyEvent(QEvent::KeyPress, key, Qt::NoModifier);
+                QCoreApplication::postEvent(this, event);
+            });
+
     // Enhancement: Check if the parameters have already been met before disconnecting controllers.
     // If all the parameters are met AND only allows a single player,
     // stop the constructor here as we do not need to continue.
@@ -217,6 +230,7 @@ QtControllerSelectorDialog::QtControllerSelectorDialog(
 }
 
 QtControllerSelectorDialog::~QtControllerSelectorDialog() {
+    controller_navigation->UnloadController();
     system.HIDCore().DisableAllControllerConfiguration();
 }
 
@@ -291,6 +305,31 @@ void QtControllerSelectorDialog::CallConfigureInputProfileDialog() {
     dialog.exec();
 }
 
+void QtControllerSelectorDialog::keyPressEvent(QKeyEvent* evt) {
+    const auto num_connected_players = static_cast<int>(
+        std::count_if(player_groupboxes.begin(), player_groupboxes.end(),
+                      [](const QGroupBox* player) { return player->isChecked(); }));
+
+    const auto min_supported_players = parameters.enable_single_mode ? 1 : parameters.min_players;
+    const auto max_supported_players = parameters.enable_single_mode ? 1 : parameters.max_players;
+
+    if ((evt->key() == Qt::Key_Enter || evt->key() == Qt::Key_Return) && !parameters_met) {
+        // Display error message when trying to validate using "Enter" and "OK" button is disabled
+        ui->labelError->setVisible(true);
+        return;
+    } else if (evt->key() == Qt::Key_Left && num_connected_players > min_supported_players) {
+        // Remove a player if possible
+        connected_controller_checkboxes[num_connected_players - 1]->setChecked(false);
+        return;
+    } else if (evt->key() == Qt::Key_Right && num_connected_players < max_supported_players) {
+        // Add a player, if possible
+        ui->labelError->setVisible(false);
+        connected_controller_checkboxes[num_connected_players]->setChecked(true);
+        return;
+    }
+    QDialog::keyPressEvent(evt);
+}
+
 bool QtControllerSelectorDialog::CheckIfParametersMet() {
     // Here, we check and validate the current configuration against all applicable parameters.
     const auto num_connected_players = static_cast<int>(

src/yuzu/applets/qt_controller.h

@@ -34,6 +34,8 @@ class HIDCore;
 enum class NpadStyleIndex : u8;
 } // namespace Core::HID
 
+class ControllerNavigation;
+
 class QtControllerSelectorDialog final : public QDialog {
     Q_OBJECT
 
@@ -46,6 +48,8 @@ public:
 
     int exec() override;
 
+    void keyPressEvent(QKeyEvent* evt) override;
+
 private:
     // Applies the current configuration.
     void ApplyConfiguration();
@@ -110,6 +114,8 @@ private:
 
     Core::System& system;
 
+    ControllerNavigation* controller_navigation = nullptr;
+
     // This is true if and only if all parameters are met. Otherwise, this is false.
     // This determines whether the "OK" button can be clicked to exit the applet.
     bool parameters_met{false};

src/yuzu/applets/qt_controller.ui

@@ -2624,13 +2624,53 @@
           </spacer>
          </item>
          <item alignment="Qt::AlignBottom">
-          <widget class="QDialogButtonBox" name="buttonBox">
-           <property name="enabled">
-            <bool>true</bool>
-           </property>
-           <property name="standardButtons">
-            <set>QDialogButtonBox::Cancel|QDialogButtonBox::Ok</set>
-           </property>
+          <widget class="QWidget" name="closeButtons" native="true">
+           <layout class="QVBoxLayout" name="verticalLayout_46">
+            <property name="spacing">
+             <number>7</number>
+            </property>
+            <property name="leftMargin">
+             <number>0</number>
+            </property>
+            <property name="topMargin">
+             <number>0</number>
+            </property>
+            <property name="rightMargin">
+             <number>0</number>
+            </property>
+            <property name="bottomMargin">
+             <number>0</number>
+            </property>
+            <item>
+             <widget class="QLabel" name="labelError">
+              <property name="enabled">
+               <bool>true</bool>
+              </property>
+              <property name="styleSheet">
+               <string notr="true">QLabel { color : red; }</string>
+              </property>
+              <property name="text">
+               <string>Not enough controllers</string>
+              </property>
+              <property name="alignment">
+               <set>Qt::AlignCenter</set>
+              </property>
+              <property name="margin">
+               <number>0</number>
+              </property>
+             </widget>
+            </item>
+            <item>
+             <widget class="QDialogButtonBox" name="buttonBox">
+              <property name="enabled">
+               <bool>true</bool>
+              </property>
+              <property name="standardButtons">
+               <set>QDialogButtonBox::Cancel|QDialogButtonBox::Ok</set>
+              </property>
+             </widget>
+            </item>
+           </layout>
           </widget>
          </item>
         </layout>