Kernel: Fix sse state saving

This was broken when I added SMP support. This patch makes sse kind of
dumb as it is saved and restored on every interrupt, but now it at least
works properly... I'll have to look into how sse can get optimized
nicely with SMP. Simple way would be pinning each thread to a specific
processor and doing pretty much what I had before, but sse thread saved
in processor rather than static global.

.vscode/c_cpp_properties.json

@@ -8,8 +8,9 @@
                 "${workspaceFolder}/userspace/libraries/*/include"
             ],
             "defines": [
-				"__arch=x86_64"
-			],
+                "__arch=x86_64",
+                "__enable_sse=1"
+            ],
             "compilerPath": "${workspaceFolder}/toolchain/local/bin/x86_64-banan_os-gcc",
             "cStandard": "c17",
             "cppStandard": "gnu++20",
@@ -23,9 +24,10 @@
                 "${workspaceFolder}/userspace/libraries/*/include"
             ],
             "defines": [
-				"__arch=x86_64",
-				"__is_kernel"
-			],
+                "__arch=x86_64",
+                "__is_kernel",
+                "__enable_sse=1"
+            ],
             "compilerPath": "${workspaceFolder}/toolchain/local/bin/x86_64-banan_os-gcc",
             "cStandard": "c17",
             "cppStandard": "gnu++20",
@@ -33,4 +35,4 @@
         }
     ],
     "version": 4
-}
+}

kernel/include/kernel/Thread.h

@@ -82,7 +82,6 @@ namespace Kernel
 #if __enable_sse
 		void save_sse();
 		void load_sse();
-		static Thread* sse_thread();
 #endif
 
 		void add_mutex() { m_mutex_count++; }

kernel/include/kernel/USB/HID/HIDDriver.h

@@ -75,12 +75,12 @@ namespace Kernel
 		};
 
 	public:
-		static BAN::ErrorOr<BAN::UniqPtr<USBHIDDriver>> create(USBDevice&, const USBDevice::InterfaceDescriptor&, uint8_t interface_index);
+		static BAN::ErrorOr<BAN::UniqPtr<USBHIDDriver>> create(USBDevice&, const USBDevice::InterfaceDescriptor&);
 
 		void handle_input_data(BAN::ConstByteSpan, uint8_t endpoint_id) override;
 
 	private:
-		USBHIDDriver(USBDevice&, const USBDevice::InterfaceDescriptor&, uint8_t interface_index);
+		USBHIDDriver(USBDevice&, const USBDevice::InterfaceDescriptor&);
 		~USBHIDDriver();
 
 		BAN::ErrorOr<void> initialize();
@@ -90,7 +90,6 @@ namespace Kernel
 	private:
 		USBDevice& m_device;
 		USBDevice::InterfaceDescriptor m_interface;
-		const uint8_t m_interface_index;
 
 		bool m_uses_report_id { false };
 		BAN::Vector<DeviceReport> m_device_inputs;

kernel/kernel/IDT.cpp

@@ -173,6 +173,10 @@ namespace Kernel
 
 		if (tid)
 		{
+#if __enable_sse
+			Thread::current().save_sse();
+#endif
+
 			if (isr == ISR::PageFault)
 			{
 				// Check if stack is OOB
@@ -218,31 +222,6 @@ namespace Kernel
 					}
 				}
 			}
-#if __enable_sse
-			else if (isr == ISR::DeviceNotAvailable)
-			{
-#if ARCH(x86_64)
-				asm volatile(
-					"movq %cr0, %rax;"
-					"andq $~(1 << 3), %rax;"
-					"movq %rax, %cr0;"
-				);
-#elif ARCH(i686)
-				asm volatile(
-					"movl %cr0, %eax;"
-					"andl $~(1 << 3), %eax;"
-					"movl %eax, %cr0;"
-				);
-#endif
-				if (auto* current = &Thread::current(); current != Thread::sse_thread())
-				{
-					if (auto* sse = Thread::sse_thread())
-						sse->save_sse();
-					current->load_sse();
-				}
-				goto done;
-			}
-#endif
 		}
 
 		Debug::s_debug_lock.lock();
@@ -334,7 +313,11 @@ namespace Kernel
 		ASSERT(Thread::current().state() != Thread::State::Terminated);
 
 done:
+#if __enable_sse
+		Thread::current().load_sse();
+#else
 		return;
+#endif
 	}
 
 	extern "C" void cpp_yield_handler(InterruptStack* interrupt_stack, InterruptRegisters* interrupt_registers)
@@ -357,6 +340,10 @@ done:
 			asm volatile("cli; 1: hlt; jmp 1b");
 		}
 
+#if __enable_sse
+		Thread::current().save_sse();
+#endif
+
 		if (!InterruptController::get().is_in_service(irq))
 			dprintln("spurious irq 0x{2H}", irq);
 		else
@@ -377,6 +364,10 @@ done:
 		Scheduler::get().reschedule_if_idling();
 
 		ASSERT(Thread::current().state() != Thread::State::Terminated);
+
+#if __enable_sse
+		Thread::current().load_sse();
+#endif
 	}
 
 	void IDT::register_interrupt_handler(uint8_t index, void (*handler)())

kernel/kernel/Terminal/VirtualTTY.cpp

@@ -364,6 +364,10 @@ namespace Kernel
 				ASSERT_NOT_REACHED();
 		}
 
+		bool old_show_cursor = m_show_cursor;
+		m_show_cursor = false;
+		set_cursor_position(m_column, m_row);
+
 		switch (codepoint)
 		{
 			case BEL: // TODO
@@ -410,10 +414,6 @@ namespace Kernel
 			for (uint32_t x = 0; x < m_width; x++)
 				m_buffer[(m_height - 1) * m_width + x] = { .foreground = m_foreground, .background = m_background, .codepoint = ' ' };
 
-			// hide cursor during scrolling
-			bool old_show_cursor = m_show_cursor;
-			m_show_cursor = false;
-			set_cursor_position(0, 0);
 			if (!m_terminal_driver->scroll(m_background))
 			{
 				// No fast scrolling, render the whole buffer to the screen
@@ -421,12 +421,12 @@ namespace Kernel
 					for (uint32_t x = 0; x < m_width; x++)
 						render_from_buffer(x, y);
 			}
-			m_show_cursor = old_show_cursor;
 
 			m_column = 0;
 			m_row--;
 		}
 
+		m_show_cursor = old_show_cursor;
 		set_cursor_position(m_column, m_row);
 	}
 

kernel/kernel/Thread.cpp

@@ -113,33 +113,8 @@ namespace Kernel
 		: m_tid(tid), m_process(process)
 	{
 #if __enable_sse
-	#if ARCH(x86_64)
-		uintptr_t cr0;
-		asm volatile(
-			"movq %%cr0, %%rax;"
-			"movq %%rax, %[cr0];"
-			"andq $~(1 << 3), %%rax;"
-			"movq %%rax, %%cr0;"
-			: [cr0]"=r"(cr0)
-			:: "rax"
-		);
+		// initializes sse storage to valid state
 		save_sse();
-		asm volatile("movq %0, %%cr0" :: "r"(cr0));
-	#elif ARCH(i686)
-		uintptr_t cr0;
-		asm volatile(
-			"movl %%cr0, %%eax;"
-			"movl %%eax, %[cr0];"
-			"andl $~(1 << 3), %%eax;"
-			"movl %%eax, %%cr0;"
-			: [cr0]"=r"(cr0)
-			:: "eax"
-		);
-		save_sse();
-		asm volatile("movl %0, %%cr0" :: "r"(cr0));
-	#else
-		#error
-	#endif
 #endif
 	}
 
@@ -482,8 +457,6 @@ namespace Kernel
 	}
 
 #if __enable_sse
-	static Thread* s_sse_thread = nullptr;
-
 	void Thread::save_sse()
 	{
 		asm volatile("fxsave %0" :: "m"(m_sse_storage));
@@ -492,12 +465,6 @@ namespace Kernel
 	void Thread::load_sse()
 	{
 		asm volatile("fxrstor %0" :: "m"(m_sse_storage));
-		s_sse_thread = this;
-	}
-
-	Thread* Thread::sse_thread()
-	{
-		return s_sse_thread;
 	}
 #endif
 

kernel/kernel/USB/Device.cpp

@@ -140,10 +140,8 @@ namespace Kernel
 				TRY(send_request(request, 0));
 			}
 
-			for (size_t j = 0; j < configuration.interfaces.size(); j++)
+			for (const auto& interface : configuration.interfaces)
 			{
-				const auto& interface = configuration.interfaces[j];
-
 				switch (static_cast<USB::InterfaceBaseClass>(interface.descriptor.bInterfaceClass))
 				{
 					case USB::InterfaceBaseClass::Audio:
@@ -153,7 +151,7 @@ namespace Kernel
 						dprintln_if(DEBUG_USB, "Found CommunicationAndCDCControl interface");
 						break;
 					case USB::InterfaceBaseClass::HID:
-						if (auto result = USBHIDDriver::create(*this, interface, j); !result.is_error())
+						if (auto result = USBHIDDriver::create(*this, interface); !result.is_error())
 							TRY(m_class_drivers.push_back(result.release_value()));
 						break;
 					case USB::InterfaceBaseClass::Physical:
@@ -217,12 +215,15 @@ namespace Kernel
 						dprintln_if(DEBUG_USB, "Invalid interface base class {2H}", interface.descriptor.bInterfaceClass);
 						break;
 				}
+			}
 
-				if (!m_class_drivers.empty())
-				{
-					dprintln("Successfully initialized USB interface");
-					return {};
-				}
+			if (!m_class_drivers.empty())
+			{
+				dprintln("Successfully initialized USB device with {}/{} interfaces",
+					m_class_drivers.size(),
+					configuration.interfaces.size()
+				);
+				return {};
 			}
 		}
 

kernel/kernel/USB/HID/HIDDriver.cpp

@@ -11,6 +11,11 @@
 namespace Kernel
 {
 
+	enum HIDRequest : uint8_t
+	{
+		SET_PROTOCOL = 0x0B,
+	};
+
 	enum class HIDDescriptorType : uint8_t
 	{
 		HID      = 0x21,
@@ -64,17 +69,16 @@ namespace Kernel
 
 	static BAN::ErrorOr<BAN::Vector<Collection>> parse_report_descriptor(BAN::ConstByteSpan report_data, bool& out_use_report_id);
 
-	BAN::ErrorOr<BAN::UniqPtr<USBHIDDriver>> USBHIDDriver::create(USBDevice& device, const USBDevice::InterfaceDescriptor& interface, uint8_t interface_index)
+	BAN::ErrorOr<BAN::UniqPtr<USBHIDDriver>> USBHIDDriver::create(USBDevice& device, const USBDevice::InterfaceDescriptor& interface)
 	{
-		auto result = TRY(BAN::UniqPtr<USBHIDDriver>::create(device, interface, interface_index));
+		auto result = TRY(BAN::UniqPtr<USBHIDDriver>::create(device, interface));
 		TRY(result->initialize());
 		return result;
 	}
 
-	USBHIDDriver::USBHIDDriver(USBDevice& device, const USBDevice::InterfaceDescriptor& interface, uint8_t interface_index)
+	USBHIDDriver::USBHIDDriver(USBDevice& device, const USBDevice::InterfaceDescriptor& interface)
 		: m_device(device)
 		, m_interface(interface)
-		, m_interface_index(interface_index)
 	{}
 
 	USBHIDDriver::~USBHIDDriver()
@@ -119,18 +123,17 @@ namespace Kernel
 		}
 
 		// If this device supports boot protocol, make sure it is not used
-		if (m_interface.endpoints.front().descriptor.bDescriptorType & 0x80)
+		if (m_interface.descriptor.bInterfaceSubClass == 0x01)
 		{
 			USBDeviceRequest request;
 			request.bmRequestType = USB::RequestType::HostToDevice | USB::RequestType::Class | USB::RequestType::Interface;
-			request.bRequest      = USB::Request::SET_INTERFACE;
+			request.bRequest      = HIDRequest::SET_PROTOCOL;
 			request.wValue        = 1; // report protocol
-			request.wIndex        = m_interface_index;
+			request.wIndex        = m_interface.descriptor.bInterfaceNumber;
 			request.wLength       = 0;
 			TRY(m_device.send_request(request, 0));
 		}
 
-
 		const auto& hid_descriptor = *reinterpret_cast<const HIDDescriptor*>(m_interface.misc_descriptors[hid_descriptor_index].data());
 		dprintln_if(DEBUG_HID, "HID descriptor ({} bytes)", m_interface.misc_descriptors[hid_descriptor_index].size());
 		dprintln_if(DEBUG_HID, "  bLength:         {}",       hid_descriptor.bLength);
@@ -139,6 +142,7 @@ namespace Kernel
 		dprintln_if(DEBUG_HID, "  bCountryCode:    {}",       hid_descriptor.bCountryCode);
 		dprintln_if(DEBUG_HID, "  bNumDescriptors: {}",       hid_descriptor.bNumDescriptors);
 
+		uint32_t report_descriptor_index = 0;
 		BAN::Vector<Collection> collections;
 		for (size_t i = 0; i < hid_descriptor.bNumDescriptors; i++)
 		{
@@ -160,8 +164,8 @@ namespace Kernel
 				USBDeviceRequest request;
 				request.bmRequestType = USB::RequestType::DeviceToHost | USB::RequestType::Standard | USB::RequestType::Interface;
 				request.bRequest      = USB::Request::GET_DESCRIPTOR;
-				request.wValue        = static_cast<uint16_t>(HIDDescriptorType::Report) << 8;
-				request.wIndex        = m_interface_index;
+				request.wValue        = (static_cast<uint16_t>(HIDDescriptorType::Report) << 8) | report_descriptor_index++;
+				request.wIndex        = m_interface.descriptor.bInterfaceNumber;
 				request.wLength       = descriptor.wItemLength;
 				auto transferred = TRY(m_device.send_request(request, dma_buffer->paddr()));