riscv: IT WORKS

finally. once i'm sure it's 100% working i can probably like, develop this addon finally

native/projects/riscv/README.md

@@ -9,3 +9,7 @@ This is based off [cnlohr/mini-rv32ima](https://github.com/cnlohr/mini-rv32ima),
 - Moved *ALL* device and MMIO code to seperate interfaces
 
 Depends on lucore.
+
+# Usage
+
+TBD (if this is moved to another repo). See the riscv_test_harness project.

native/projects/riscv/include/riscv/CPU.hpp

@@ -6,6 +6,7 @@ namespace riscv {
 
 	/// The CPU core.
 	struct CPU : Bus::Device {
+		BasicType Type() const override { return BasicType::Cpu; } 
 		bool Clocked() const override { return true; }
 		void Clock() override;
 
@@ -15,6 +16,17 @@ namespace riscv {
 
 		void TimerInterrupt();
 
+		constexpr CPU() {
+			Reset();
+		}
+
+		constexpr void Reset() {
+			// Initalize some state. We're cool like that :)
+			pc = 0x80000000;
+			gpr[10] = 0x0; // HART id
+			extraflags |= 3; // Start in Machine mode
+		}
+
 		// TODO: Handlers for CSR read/write (if we need it?)
 
 		/// CPU state

native/projects/riscv/include/riscv/CPUTypes.hpp

@@ -1,5 +1,6 @@
 #pragma once
 #include <riscv/Types.hpp>
+#include <string_view>
 
 namespace riscv {
 
@@ -25,41 +26,78 @@ namespace riscv {
 	};
 
 	enum class Gpr : u8 {
-		X0, // zero
+		Zero,
-		X1,
+		Ra,
-		X2,
+		Sp,
-		X3,
+		Gp,
-		X4,
+		Tp,
-		X5,
+		T0,
-		X6,
+		T1,
-		X7,
+		T2,
-		X8,
+		S0, // also `fp`
-		X9,
+		S1,
-		X10,
+		A0,
-		X11,
+		A1,
-		X12,
+		A2,
-		X13,
+		A3,
-		X14,
+		A4,
-		X15,
+		A5,
-		X16,
+		A6,
-		X17,
+		A7,
-		X18,
+		S2,
-		X19,
+		S3,
-		X20,
+		S4,
-		X21,
+		S5,
-		X22,
+		S6,
-		X23,
+		S7,
-		X24,
+		S8,
-		X25,
+		S9,
-		X26,
+		S10,
-		X27,
+		S11,
-		X28,
+		T3,
-		X29,
+		T4,
-		X30,
+		T5,
-		X31,
+		T6
-		PC
 	};
 
+	constexpr std::string_view RegName(Gpr gpr) {
+		std::string_view table[] = {
+			"zero",
+			"ra",
+			"sp",
+			"gp",
+			"tp",
+			"t0",
+			"t1",
+			"t2",
+			"s0",
+			"s1",
+			"a0",
+			"a1",
+			"a2",
+			"a3",
+			"a4",
+			"a5",
+			"a6",
+			"a7",
+			"s2",
+			"s3",
+			"s4",
+			"s5",
+			"s6",
+			"s7",
+			"s8",
+			"s9",
+			"s10",
+			"s11",
+			"t3",
+			"t4",
+			"t5",
+			"t6"
+		};
+		return table[static_cast<usize>(gpr)];
+	}
+
 	/// Container for GPRs which can be Statically Typed or not depending on use case.
 	/// Pretty cool, huh?
 	struct GeneralPurposeRegisters {

native/projects/riscv/include/riscv/Devices/RamDevice.hpp

@@ -14,6 +14,9 @@ namespace riscv::devices {
 		Address Base() const override;
 		Address Size() const override;
 
+		u8* Raw() const {
+			return memory;
+		}
 
 		u8 PeekByte(Address address) override;
 		u16 PeekShort(Address address) override;

native/projects/riscv/include/riscv/Devices/SysconDevice.hpp

@@ -11,6 +11,7 @@ namespace riscv::devices {
 
 		SysconDevice(System* system);
 
+		Address Base() const override { return BASE_ADDRESS; }
 		Address Size() const override { return sizeof(u32); } // I think this is right?
 
 		u32 Peek(Address address) override;

native/projects/riscv/include/riscv/System.hpp

@@ -1,8 +1,8 @@
 #include <riscv/Bus.hpp>
 #include <riscv/CPU.hpp>
+#include <riscv/Devices/ClntDevice.hpp>
 #include <riscv/Devices/RamDevice.hpp>
 #include <riscv/Devices/SysconDevice.hpp>
-#include <riscv/Devices/ClntDevice.hpp>
 
 namespace riscv {
 
@@ -10,25 +10,13 @@ namespace riscv {
 	struct System {
 		/// Create a basic system with the basic periphials created.
 		/// All other periphials should be managed by the creator of this System
-		static System* WithMemory(Address ramSize);
+		static System* Create(Address ramSize);
 
 		~System();
 
-		void AddDeviceToBus(Bus::Device* device);
+		void Step();
 
-		/// returns false if the cpu broke execution
+		// TODO: callbacks for SYSCON PowerOff and Reboot.
-		bool Step();
-
-		CPU* GetCPU();
-		Bus* GetBus();
-
-	   private:
-		/// How many Cycle() calls will the bus get
-		/// (also decides ipsRate)
-		u32 cycleRate;
-
-		/// How many instructions will the CPU execute each step
-		u32 ipsRate;
 
 		Bus* bus;
 
@@ -37,6 +25,9 @@ namespace riscv {
 		devices::RamDevice* ram;
 		devices::SysconDevice* syscon;
 		devices::ClntDevice* clnt;
+
+	   private:
+		System() = default;
 	};
 
 } // namespace riscv

native/projects/riscv/src/Bus.cpp

@@ -14,6 +14,8 @@ namespace riscv {
 		if(!device)
 			return false;
 
+		device->Attached(this);
+
 		if(device->IsA<CPU*>()) {
 			// Return early to avoid putting the CPU pointer inside the devices vector.
 			// We do not actually own the CPU.
@@ -119,7 +121,7 @@ namespace riscv {
 	}
 
 	Bus::Device* Bus::FindDeviceForAddress(Address address) const {
-		auto try_find_device = [&](auto container, Address address) {
+		auto try_find_device = [&](const auto& container, Address address) {
 			return std::find_if(container.begin(), container.end(), [&](const auto& pair) {
 				return
 				// We can shorcut region checking if the requested addess matches base address.

native/projects/riscv/src/CPU.cpp

@@ -3,11 +3,12 @@
 #include <riscv/Bus.hpp>
 #include <riscv/CPU.hpp>
 
-#include "riscv/CPUTypes.hpp"
+#include <lucore/Logger.hpp>
 
 namespace riscv {
 
-	constexpr static Address RamImageOffset = 0x80000000;
+	// Not needed
+	//constexpr static Address RamImageOffset = 0x80000000;
 
 	void CPU::Clock() {
 		// do the thing
@@ -47,31 +48,35 @@ namespace riscv {
 
 		u32 rdid = 0;
 		u32 rval = 0;
-		u32 pc = this->pc;
 		u32 cycle = this->cyclel;
 
 		if(interruptsInFlight()) {
 			Trap(0x80000007);
 		} else {
 			for(u32 iInst = 0; iInst < instCount; ++iInst) {
-				auto ofs_pc = pc - RamImageOffset;
+				rdid = 0; // force it to gpr 0 (zero), which is not writable
 				cycle++;
 
-				if(ofs_pc & 3) {
+				//lucore::LogInfo("[CPU] pc @ 0x{:08x}", pc);
+
+				if((pc & 3)) {
+					lucore::LogWarning("[CPU] misaligned jump target.. 0x{:08x}", pc);
 					Trap(TrapCode::InstructionAddressMisaligned);
 					break;
 				} else {
-					auto ir = bus->PeekWord(ofs_pc);
+					auto ir = bus->PeekWord(pc);
 					if(trapped) {
 						// Overwrite the trap that the bus generated. This might not really work out
 						// in practice but should at least kind-of replicate behaviour (our address
 						// space is emulated using the [Bus] class, so there is no heap write issue
 						// that could be caused by leaving it unbound).
 						Trap(TrapCode::InstructionAccessFault);
-						rval = ofs_pc + RamImageOffset;
+						rval = pc;
 						break;
 					}
 
+					//lucore::LogInfo("[CPU] fetch 0x{:08x} 0x{:08x}", pc, ir);
+
 					rdid = (ir >> 7) & 0x1f;
 
 					// Do the thing!
@@ -91,6 +96,7 @@ namespace riscv {
 							if(reladdy & 0x00100000)
 								reladdy |= 0xffe00000;
 							rval = pc + 4;
+							//lucore::LogInfo("j/al 0x{:08x}", pc + reladdy);
 							pc = pc + reladdy - 4;
 							break;
 						}
@@ -100,6 +106,7 @@ namespace riscv {
 							i32 imm_se = imm | ((imm & 0x800) ? 0xfffff000 : 0);
 							rval = pc + 4;
 							pc = ((gpr[((ir >> 15) & 0x1f)] + imm_se) & ~1) - 4;
+							//lucore::LogInfo("jalr {}, 0x{:08x}", RegName(static_cast<Gpr>(((ir >> 15) & 0x1f))), ((gpr[((ir >> 15) & 0x1f)] + imm_se) & ~1) - 4);
 							break;
 						}
 
@@ -153,8 +160,6 @@ namespace riscv {
 							i32 imm_se = imm | ((imm & 0x800) ? 0xfffff000 : 0);
 							u32 rsval = rs1 + imm_se;
 
-							rsval -= RamImageOffset;
-
 							switch((ir >> 12) & 0x7) {
 								// LB, LH, LW, LBU, LHU
 								case 0:
@@ -178,7 +183,7 @@ namespace riscv {
 							}
 
 							if(trapped) {
-								rval = rsval + RamImageOffset;
+								rval = rsval;// + RamImageOffset;
 							}
 							break;
 						}
@@ -188,7 +193,7 @@ namespace riscv {
 							u32 addy = ((ir >> 7) & 0x1f) | ((ir & 0xfe000000) >> 20);
 							if(addy & 0x800)
 								addy |= 0xfffff000;
-							addy += rs1 - RamImageOffset;
+							addy += rs1;
 							rdid = 0;
 
 							switch((ir >> 12) & 0x7) {
@@ -200,6 +205,7 @@ namespace riscv {
 									bus->PokeShort(addy, rs2);
 									break;
 								case 2:
+									//lucore::LogInfo("storeWord(0x{:08x}, 0x{:08x})", addy, rs2);
 									bus->PokeWord(addy, rs2);
 									break;
 								default:
@@ -207,7 +213,7 @@ namespace riscv {
 							}
 
 							if(trapped) {
-								rval = addy + RamImageOffset;
+								rval = addy;// + RamImageOffset;
 							}
 							break;
 						}
@@ -457,13 +463,13 @@ namespace riscv {
 
 							// rs1 -= MINIRV32_RAM_IMAGE_OFFSET;
 
-							rs1 -= RamImageOffset;
+							//rs1 -= RamImageOffset;
 
 							// We don't implement load/store from UART or CLNT with RV32A here.
 
 							rval = bus->PeekWord(rs1);
 							if(trapped) {
-								rval = rs1 + RamImageOffset;
+								rval = rs1;// + RamImageOffset;
 								break;
 							}
 
@@ -523,6 +529,7 @@ namespace riscv {
 					break;
 
 				if(rdid) {
+					//lucore::LogInfo("writing register {} -> 0x{:08x}", RegName(static_cast<Gpr>(rdid)), rval);
 					gpr[rdid] = rval;
 				}
 

native/projects/riscv/src/System.cpp

@@ -0,0 +1,36 @@
+#include <riscv/System.hpp>
+
+namespace riscv {
+
+	System* System::Create(Address ramSize) {
+		auto* system = new System;
+
+		system->bus = new Bus();
+		system->cpu = new CPU();
+		system->ram = new devices::RamDevice(0x80000000, ramSize);
+		system->clnt = new devices::ClntDevice();
+		system->syscon = new devices::SysconDevice(system);
+
+		system->cpu->Reset();
+
+		system->bus->AttachDevice(system->cpu);
+		system->bus->AttachDevice(system->clnt);
+		system->bus->AttachDevice(system->syscon);
+		system->bus->AttachDevice(system->ram);
+		
+		return system;
+	}
+
+	System::~System() {
+		delete cpu;
+		delete bus; // the rest of the device pointers will be deleted by the bus.
+	}
+
+	void System::Step() {
+		// Clock the bus, it'll do everything else.
+		bus->Clock();
+
+		// Later: handling for invalid cases!
+	}
+
+} // namespace riscv

native/projects/riscv_test_harness/main.cpp

@@ -1,5 +1,10 @@
-//! A test harness for testing the riscv library.
+//! A test harness for testing if the riscv library actually works.
 #include <riscv/System.hpp>
+#include <lucore/StdoutSink.hpp>
+
+#include <cstdio> // I know, I know, but this is a test program. Yell later :)
+#include <lucore/Assert.hpp>
+#include <lucore/Logger.hpp>
 
 /// simple 16550 UART implementation
 struct SimpleUartDevice : public riscv::Bus::MmioDevice {
@@ -7,29 +12,55 @@ struct SimpleUartDevice : public riscv::Bus::MmioDevice {
 
 	riscv::Address Base() const override { return BASE_ADDRESS; }
 
-	riscv::Address Size() const override { return 5; }
+	riscv::Address Size() const override { return 12; } // for now
 
-	// TODO: emulate properly
 	u32 Peek(riscv::Address address) override {
 		switch(address) {
 			case BASE_ADDRESS:
-				break;
+				return 0x60; // active, but no keyboard input
 			case BASE_ADDRESS + 5:
-				break;
+				return '\0';
 		}
 
 		return 0;
 	}
 
 	void Poke(riscv::Address address, u32 value) override {
-		if(address == BASE_ADDRESS) { // write to data buffer
+		if(address == BASE_ADDRESS) {
-			printf("%c\n", value);
+			char c = value & 0x000000ff;
+			//lucore::LogInfo("[UART] got data buffer poke of char {:02x} @ pc 0x{:08x}", c, bus->GetCPU()->pc);
+			//printf("%c", c);
+			fputc(c, stderr);
 		}
 	}
 };
 
-int main() {
+int main(int argc, char** argv) {
-	auto system = riscv::System::WithMemory(128 * 1024);
+	lucore::LoggerAttachStdout();
-	system->AddDeviceToBus(new SimpleUartDevice);
+
+	LUCORE_CHECK(argc == 2, "this test harness expects one argument (the file to load into riscv memory and execute). got {} arguments", argc);
+	
+	// 128 KB of ram. Won't be enough to boot linux but should be good enough to test most baremetal apps
+	auto system = riscv::System::Create(128 * 1024);
+
+	// Attach our UART device
+	system->bus->AttachDevice(new SimpleUartDevice);
+
+	auto fp = std::fopen(argv[1], "rb");
+	LUCORE_CHECK(fp, "could not open file \"{}\"", argv[1]);
+
+	fseek(fp, 0, SEEK_END);
+	auto len = ftell(fp);
+	fseek(fp, 0, SEEK_SET);
+
+	fread(system->ram->Raw(), 1, len, fp);
+	fclose(fp);
+
+	// Do the thing now!
+	while(true) {
+		system->Step();
+	}
+
+	delete system;
 	return 0;
 }

native/projects/riscv_test_harness/test/Makefile

@@ -0,0 +1,51 @@
+PROJECT = test
+
+# where your rv32 toolchain is
+TCPATH = /home/lily/bin/riscv/bin
+PREFIX = $(TCPATH)/riscv32-unknown-elf
+
+CC = $(PREFIX)-gcc
+CXX = $(PREFIX)-g++
+
+CCFLAGS = -ffreestanding -fno-stack-protector
+CCFLAGS += -static -static-libgcc -fdata-sections -ffunction-sections
+CCFLAGS += -g -Os -march=rv32ima -mabi=ilp32 
+
+CXXFLAGS = $(CCFLAGS) -std=c++20 -fno-exceptions -fno-rtti
+
+LDFLAGS = -T binary.ld -nostdlib -Wl,--gc-sections
+
+OBJS := start.o main.o
+
+.PHONY: all test clean
+
+all: $(PROJECT).bin $(PROJECT).debug.txt
+
+test: $(PROJECT).bin $(PROJECT).debug.txt
+	../../../../build/projects/riscv_test_harness/rvtest $<
+
+clean:
+	rm $(PROJECT).elf $(PROJECT).bin $(PROJECT).debug.txt $(OBJS)
+
+$(PROJECT).elf: $(OBJS)
+	$(CC) $(CCFLAGS) $(LDFLAGS) -o $@ $(OBJS) 
+
+$(PROJECT).bin : $(PROJECT).elf
+	$(PREFIX)-objcopy $^ -O binary $@
+
+$(PROJECT).debug.txt : $(PROJECT).elf
+	$(PREFIX)-objdump -t $^ > $@
+	$(PREFIX)-objdump -S $^ >> $@
+
+# Compile rules
+
+%.o: %.cpp
+	$(CXX) -c $(CXXFLAGS) $< -o $@
+
+%.o: %.c
+	$(CC) -c $(CCFLAGS) $< -o $@
+
+%.o: %.S
+	$(CC) -x assembler-with-cpp -march=rv32ima -mabi=ilp32 -c $< -o $@
+
+

native/projects/riscv_test_harness/test/README.md

@@ -0,0 +1,5 @@
+# what
+
+This is a simple bare-metal riscv program that is able to run on the test harness that does some testing.
+
+It is barebones, I know.

native/projects/riscv_test_harness/test/binary.ld

@@ -0,0 +1,73 @@
+__heap_size = 0x1000;
+__stack_size = 0x1000;
+
+ENTRY(_start)
+
+SECTIONS
+{
+	. = 0x80000000;
+	.text : ALIGN(16) {
+		__TEXT_BEGIN__ = .;
+		*(.initial_jump)
+		*(.entry.text)
+		*(.init.literal)
+		*(.init)
+		*(.text)
+		*(.literal .text .literal.* .text.* .stub)
+		*(.out_jump.literal.*)
+		*(.out_jump.*)
+		__TEXT_END__ = .;
+	}
+
+	/* If we're on a newer compiler */
+	/DISCARD/ :
+	{
+		*(.interp)
+		*(.dynsym)
+		*(.dynstr)
+		*(.header)
+	} : phdr
+
+	.data : ALIGN(16) {
+		__DATA_BEGIN__ = .;
+		*(.rodata)
+		*(.rodata.*)
+		*(.gnu.linkonce.r.*)
+		*(.rodata1)
+		*(.dynsbss)
+		*(.gnu.linkonce.sb.*)
+		*(.scommon)
+		*(.gnu.linkonce.sb2.*)
+		*(.sbss)
+		*(.sbss.*)
+		*(.sbss2)
+		*(.sbss2.*)
+		*(.dynbss)
+		*(.data)
+		*(.data.*)
+		*(.got)
+		*(.got.*)
+		__DATA_END__ = .;
+	}
+
+	.bss : ALIGN( 16 ) {
+		__BSS_BEGIN__ = .;
+		*(.bss) /* Tricky: BSS needs to be allocated but not sent. GCC Will not populate these for calculating data size */
+		*(.bss.*)
+		__BSS_END__ = .;
+	}
+
+	.heap : ALIGN( 16 ) {
+		_sheap = .;
+		. = . + __heap_size;
+		_eheap = .;
+	}
+
+	.stack : ALIGN( 16 ) {
+		_estack = .;
+		. = . + __stack_size;
+		_sstack = .;
+	}
+}
+
+

native/projects/riscv_test_harness/test/main.c

@@ -0,0 +1,54 @@
+// a simple test program
+#include <stdint.h>
+
+uint32_t strlen(const char* str) {
+	if(!str)
+		return 0;
+	const char* c = str;
+	while(*c++)
+		;
+	return c - str;
+}
+
+#define UART_BASE 0x10000000
+#define UART_DATA *(volatile uint32_t*)UART_BASE
+#define UART_STATUS UART_DATA
+
+void putc(char c) {
+	UART_DATA = (uint32_t)c;
+}
+
+__attribute__((noinline)) void puts(const char* str) {
+	const uint32_t length = strlen(str);
+	for(uint32_t i = 0; i < length; ++i)
+		putc(str[i]);
+}
+
+static uint32_t value = 0;
+static uint16_t shortvalue = 0;
+static uint8_t bytevalue = 0;
+
+#define COUNTER_TEST(var, max)                        \
+	for(int i = 0; i < max; ++i) {                      \
+		puts(#var " is (before modification): "); \
+		putc("0123456789"[var]);                    \
+		putc('\n');                                   \
+                                                      \
+		var = i;                                    \
+		puts(#var " is (after  modification): "); \
+		putc("0123456789"[var]);                    \
+		putc('\n');                                   \
+	}
+
+void main() {
+	puts("hello world I guess\n");
+
+#if 1
+	COUNTER_TEST(value, 9);
+	COUNTER_TEST(shortvalue, 9);
+	COUNTER_TEST(bytevalue, 9);
+#endif
+
+	// loop forever
+	for(;;);
+}

native/projects/riscv_test_harness/test/start.S

@@ -0,0 +1,13 @@
+# Simple bare-metal RISCV startup code.
+
+.section .initial_jump
+.global _start
+
+.extern main
+
+.align 4
+_start:
+	la sp, _sstack	# set up C stack
+	addi sp,sp,-16  # ...
+	sw ra,12(sp)    # ...
+	jal ra, main	# jump to C code!