feat: Final RV64 CPU with MMU and kernel

Author: TonyCrane

.gitignore

@@ -1,4 +1,6 @@
 .DS_Store
 
 src/core.vvp
-src/dump.vcd
+src/dump.vcd
+
+src/dut_changes.txt

README.md

@@ -6,7 +6,9 @@
 
 ## 实验进度
 
-系统贯通课程会逐步实现一个 RISC-V 五级流水线 CPU，并实现异常处理、分支预测、Cache 等功能。本 repo 通过分支、tag 等来记录实验进度，保存各阶段成果。
+系统贯通课程会逐步实现一个 RISC-V 五级流水线 CPU，并实现异常处理、分支预测、Cache、MMU 等功能，并在其上运行自己编写的简易 kernel。
+
+本 repo 通过分支、tag 等来记录实验进度，保存各阶段成果。
 
 - [x] 系统 Ⅰ lab5-1/lab5-2：单周期 CPU
 - [x] extra：单周期 CPU with 特权指令/异常处理
@@ -18,6 +20,12 @@
     - 使用了提供的实验框架而非自己的，就不放在 repo 里了
 - [x] 系统 Ⅲ lab2：流水线 CPU with Cache
     - 使用了提供的实验框架而非自己的，就不放在 repo 里了
+- [x] 系统 Ⅲ lab Xpart：软硬件贯通实验，主要部分是实现 MMU 以及调试 kernel
+    - RV64IZicsr 全部指令（除去 fence ebreak wfi）
+    - 包含 Supervisor 和 User 两个特权级
+    - 实现了 Bare 和 Sv39 两种分页模式
+    - 支持串口输出
+    - 展示 slides 在：[slides.tonycrane.cc/sys3-xpart-pre](https://slides.tonycrane.cc/sys3-xpart-pre/)
 
 ## 实验环境
 

src/CPU.v

@@ -6,19 +6,25 @@ module Core (
     input  wire        step,
     input  wire        debug_mode,
     input  wire [4:0]  debug_reg_addr, // register address
-    output wire [31:0] chip_debug_out0,
-    output wire [31:0] chip_debug_out1,
-    output wire [31:0] chip_debug_out2,
-    output wire [31:0] chip_debug_out3
+    output wire [63:0] chip_debug_out0,
+    output wire [63:0] chip_debug_out1,
+    output wire [63:0] chip_debug_out2,
+    output wire [63:0] chip_debug_out3,
+    output wire [7:0]  sim_uart_char_out,
+    output wire        sim_uart_char_valid
 );
     wire rst, mem_write, mem_clk, cpu_clk;
-    wire [31:0] inst, core_data_in, addr_out, core_data_out, pc_out;
+    wire [31:0] inst;
+    wire [63:0] core_data_in, addr_out, core_data_out, pc_out;
     reg  [31:0] clk_div;
-    wire [31:0] debug_reg;
+    wire [63:0] debug_reg;
+    wire [63:0] satp;
+    wire stall, mem_read;
+    wire [2:0]  width;
 
     assign rst = ~aresetn;
 
-    CPU cpu(
+    CPU cpu (
         .clk(cpu_clk),
         .rst(rst),
         .inst(inst),
@@ -27,8 +33,12 @@ module Core (
         .data_out(core_data_out),    // data to data memory
         .pc_out(pc_out),             // connect to instruction memory
         .mem_write(mem_write),
+        .mem_read(mem_read),
         .debug_reg_addr(debug_reg_addr),
-        .debug_reg(debug_reg)
+        .debug_reg(debug_reg),
+        .satp(satp),
+        .data_width(width),
+        .stall(stall)
     );
 
     always @(posedge clk) begin
@@ -38,21 +48,24 @@ module Core (
     assign mem_clk = ~clk_div[0];
     assign cpu_clk = debug_mode ? clk_div[0] : step;
 
-    ROM rom_unit (
-        .address(pc_out / 4),
-        .out(inst)
-    );
-    
-    RAM ram_unit (
-        .clk(mem_clk),
-        .we(mem_write),
-        .address(addr_out / 4),
+    MMU mmu (
+        .clk(mem_clk), .rst(rst),
+        .inst_addr(pc_out),
+        .data_addr(addr_out),
         .write_data(core_data_out),
-        .read_data(core_data_in)
+        .mem_write(mem_write),
+        .mem_read(mem_read),
+        .width(width),
+        .satp(satp),
+        .inst_out(inst),
+        .read_data(core_data_in),
+        .stall_pipeline(stall),
+        .sim_uart_char_out(sim_uart_char_out),
+        .sim_uart_char_valid(sim_uart_char_valid)
     );
 
     assign chip_debug_out0 = pc_out;
     assign chip_debug_out1 = addr_out;
-    assign chip_debug_out2 = inst;
+    assign chip_debug_out2 = {32'b0, inst};
     assign chip_debug_out3 = debug_reg;
 endmodule

src/Core.v

@@ -2,23 +2,28 @@
 
 module CoreSim;
     reg clk, rst;
-    Core core(
+    Core core (
         .clk(clk),
         .aresetn(~rst),
         .step(1'b0),
         .debug_mode(1'b1),
         .debug_reg_addr(5'b0)
     );
 
-    initial begin
-        $dumpvars(0, CoreSim);
-        #700000 $finish;
-    end
+    // initial begin
+    //     $dumpvars(0, CoreSim);
+    //     // #700000 $finish;
+    //     // #2000 $finish;
+    //     // #1000000000 $finish;
+    // end
 
     initial begin
         clk = 0;
         rst = 1;
         #2 rst = 0;
     end
     always #1 clk = ~clk;
+    // always #5000000 begin
+    //     $display("pc: %h", core.cpu.MEM_WB_pc);
+    // end
 endmodule

src/CoreSim.v

@@ -0,0 +1,35 @@
+`timescale 1ns / 1ps
+
+module MMUSim;
+    reg clk, rst;
+    wire [31:0] inst_out;
+    wire [63:0] read_data;
+    wire stall_pipeline;
+
+    MMU mmu (
+        .clk(clk), .rst(rst),
+        .inst_addr(64'hffffffe000200000),
+        .data_addr(64'hffffffe00020b000),
+        // .data_addr(64'h800),
+        .write_data(64'h0000000000000000),
+        .mem_write(1'b0),
+        .mem_read(1'b1),
+        .width(3'b011),
+        .satp(64'h8000000000080202),
+        .inst_out(inst_out),
+        .read_data(read_data),
+        .stall_pipeline(stall_pipeline)
+    );
+
+    initial begin
+        $dumpvars(0, MMUSim);
+        #200 $finish;
+    end
+
+    initial begin
+        clk = 0;
+        rst = 1;
+        #2 rst = 0;
+    end
+    always #1 clk = ~clk;
+endmodule

src/MMUSim.v

@@ -6,13 +6,27 @@ all: compile simulate
 
 compile:
 	iverilog -o core.vvp -y . \
-		-y components -y memory -y utils \
+		-y components -y utils \
 		-I headers \
 		CoreSim.v
 
 simulate:
 	vvp -n core.vvp
+
+open:
 	$(GTKWAVE) dump.vcd wave.gtkw
 
+mmu: compile-mmu simulate-mmu
+
+compile-mmu:
+	iverilog -o core.vvp -y . \
+		-y components -y utils \
+		-I headers \
+		MMUSim.v
+
+simulate-mmu:
+	vvp -n core.vvp
+	$(GTKWAVE) dump.vcd wave_mmu.gtkw
+
 clean:
 	rm -f core.vvp dump.vcd

src/Makefile

@@ -10,30 +10,44 @@ module Top(
     output [ 2:0]  rgb1,
     output [ 2:0]  rgb2,
     output [ 7:0]  num_csn,
-    output [ 7:0]  num_an
+    output [ 7:0]  num_an,
+    output         UART_TXD
 );
     logic aresetn;
     logic step;
 
-    logic [31:0] chip_debug_out0;
-    logic [31:0] chip_debug_out1;
-    logic [31:0] chip_debug_out2;
-    logic [31:0] chip_debug_out3;
+    logic [63:0] chip_debug_out0;
+    logic [63:0] chip_debug_out1;
+    logic [63:0] chip_debug_out2;
+    logic [63:0] chip_debug_out3;
+
+    logic [7:0] uart_data, sim_uart_char;
+    logic uart_send, uart_ready, sim_uart_char_valid;
+    logic [31:0] clk_div;
+    logic clk_cpu;
+
+    always @(posedge clk) begin
+        if (!resetn) clk_div <= 0;
+        else clk_div <= clk_div + 1;
+    end
+    assign clk_cpu = clk_div[3];
 
     Core chip_inst(
-        .clk(clk),
+        .clk(clk_cpu),
         .aresetn(aresetn),
         .step(step),
         .debug_mode(switch[15]),
         .debug_reg_addr(switch[11:7]),
         .chip_debug_out0(chip_debug_out0),
         .chip_debug_out1(chip_debug_out1),
         .chip_debug_out2(chip_debug_out2),
-        .chip_debug_out3(chip_debug_out3)
+        .chip_debug_out3(chip_debug_out3),
+        .sim_uart_char_out(sim_uart_char),
+        .sim_uart_char_valid(sim_uart_char_valid),
     );
 
     IO_Manager io_manager_inst(
-        .clk(clk),
+        .clk(clk_cpu),
         .resetn(resetn),
 
         // to chip
@@ -54,9 +68,27 @@ module Top(
         .debug1({16'b0, switch[15:0]}),
         .debug2({12'b0, 3'b0, button[4], 3'b0, button[3], 3'b0, button[2], 3'b0, button[1], 3'b0, button[0]}),
         .debug3(32'h12345678),
-        .debug4(chip_debug_out0),
-        .debug5(chip_debug_out1),
-        .debug6(chip_debug_out2),
-        .debug7(chip_debug_out3)
+        .debug4(chip_debug_out0[31:0]),
+        .debug5(chip_debug_out1[31:0]),
+        .debug6(chip_debug_out2[31:0]),
+        .debug7(chip_debug_out3[31:0])
+    );
+
+    UART_TX_CTRL uart_tx_ctrl (
+        .SEND(uart_send),
+        .DATA(uart_data),
+        .CLK(clk),
+        .READY(uart_ready),
+        .UART_TX(UART_TXD)
+    );
+
+    uart_buffer UART_BUFF (
+        .clk(clk_cpu),
+        .rst(~aresetn),
+        .ready(uart_ready),
+        .sim_uart_char_valid(sim_uart_char_valid),
+        .sim_uart_char(sim_uart_char),
+        .send(uart_send),
+        .datao(uart_data)
     );
 endmodule