Implement a basic MMU configuration

.vscode/launch.json

@@ -33,6 +33,31 @@
       ],
       "preLaunchTask": "Run QEMU"
     },
+    {
+      "name": "Attach to QEMU (AArch64) wo. window",
+      "type": "cppdbg",
+      "request": "launch",
+      "program": "${workspaceFolder}/target/aarch64-unknown-none/debug/nova",
+      "miDebuggerServerAddress": "localhost:1234",
+      "miDebuggerPath": "gdb",
+      "cwd": "${workspaceFolder}",
+      "stopAtEntry": true,
+      "externalConsole": false,
+      "MIMode": "gdb",
+      "setupCommands": [
+        {
+          "description": "Enable pretty-printing for gdb",
+          "text": "-enable-pretty-printing",
+          "ignoreFailures": true
+        },
+        {
+          "description": "Show assembly on stop",
+          "text": "set disassemble-next-line on",
+          "ignoreFailures": true
+        }
+      ],
+      "preLaunchTask": "Run QEMU wo window"
+    },
 
     {
       "name": "Attach LLDB",

.vscode/tasks.json

@@ -14,9 +14,38 @@
     {
       "label": "Run QEMU",
       "type": "shell",
-      "command": "llvm-objcopy -O binary target/aarch64-unknown-none/debug/nova target/aarch64-unknown-none/debug/kernel8.img && qemu-system-aarch64 -M raspi3b -cpu cortex-a53 -serial stdio -sd sd.img -kernel ${workspaceFolder}/target/aarch64-unknown-none/debug/kernel8.img -S -s -m 1024",
+      "command": "llvm-objcopy -O binary target/aarch64-unknown-none/debug/nova target/aarch64-unknown-none/debug/kernel8.img && echo Starting QEMU&qemu-system-aarch64 -M raspi3b -cpu cortex-a53 -serial stdio -sd sd.img -kernel ${workspaceFolder}/target/aarch64-unknown-none/debug/kernel8.img -S -s -m 1024",
       "isBackground": true,
-      "dependsOn": ["Build"]
+      "dependsOn": ["Build"],
+      "problemMatcher": {
+                "pattern": {
+                    "regexp": "^(Starting QEMU)",
+                    "line": 1,
+                },
+                "background": {
+                    "activeOnStart": true,
+                    "beginsPattern": "^(Starting QEMU)",
+                    "endsPattern": "^(Starting QEMU)"
+                }
+            }
+    },
+        {
+      "label": "Run QEMU wo window",
+      "type": "shell",
+      "command": "llvm-objcopy -O binary target/aarch64-unknown-none/debug/nova target/aarch64-unknown-none/debug/kernel8.img && echo Starting QEMU&qemu-system-aarch64 -M raspi3b -cpu cortex-a53 -display none -serial stdio -sd sd.img -kernel ${workspaceFolder}/target/aarch64-unknown-none/debug/kernel8.img -S -s -m 1024",
+      "isBackground": true,
+      "dependsOn": ["Build"],
+      "problemMatcher": {
+                "pattern": {
+                    "regexp": "^(Starting QEMU)",
+                    "line": 1,
+                },
+                "background": {
+                    "activeOnStart": true,
+                    "beginsPattern": "^(Starting QEMU)",
+                    "endsPattern": "^(Starting QEMU)"
+                }
+            }
     }
   ]
 }

link.ld

@@ -10,7 +10,7 @@ SECTIONS {
         *(.rodata .rodata.*)
     }
 
-    .data : {
+    .data ALIGN(2M) : {
         _data = .;
         *(.data .data.*)
     }
@@ -27,28 +27,39 @@ SECTIONS {
         KEEP(*(.vector_table))
     }
 
+    .translation_table_l1 ALIGN(4096) : {
+        __translation_table_l1_start = .;
+        . += 4096;
+        __translation_table_l1_end = .;
+    }
+
+    .translation_table_l2 ALIGN(4096) : {
+        __translation_table_l2_start = .;
+        . += 4096;
+        __translation_table_l2_end = .;
+    }
+
     .heap : ALIGN(16)
     {
     __heap_start = .;
-    . += 0x10000; #10kB
+    . += 100K; #100kB
     __heap_end = .;
     }
 
     .stack : ALIGN(16)
     {
     __stack_start = .;
-    . += 0x10000; #10kB stack
+    . += 10K; #10kB stack
     __stack_end = .;
     }
 
-    .stack_el0 : ALIGN(16)
+    .stack_el0 : ALIGN(2M)
     {
     __stack_start_el0 = .;
-    . += 0x10000; #10kB stack
+    . += 10K; #10kB stack
     __stack_end_el0 = .;
     }
 
-
     _end = .;
 }
 

src/aarch64/mmu.rs

@@ -1,17 +1,67 @@
-use core::arch::asm;
+use core::ptr::write_volatile;
 
-pub fn init_mmu() {
+use crate::{println, PERIPHERAL_BASE};
-    let ips = 0b000 << 32;
 
-    // 4KB granularity
+unsafe extern "C" {
-    let tg0 = 0b00 << 14;
+    static mut __translation_table_l1_start: u64;
-    let tg1 = 0b00 << 30;
+    static mut __translation_table_l2_start: u64;
-
+    static __stack_start_el0: u64;
-    //64-25 = 29 bits of VA
+    static _data: u64;
-    // FFFF_FF80_0000_0000 start address
+}
-    let t0sz = 25;
+
-
+pub fn init_translation_table() {
-    let tcr_el1: u64 = ips | tg0 | tg1 | t0sz;
+    unsafe {
-
+        write_volatile(
-    unsafe { asm!("msr TCR_EL1, {0:x}", in(reg) tcr_el1) };
+            &raw mut __translation_table_l1_start,
+            table_descriptor_entry(&raw mut __translation_table_l2_start as u64),
+        );
+        println!("{}", &raw mut __translation_table_l2_start as u64);
+
+        for i in 0..512 {
+            let addr = 0x0 + (i as u64 * 2 * 1024 * 1024);
+
+            let descriptor = if addr < &_data as *const _ as u64 {
+                block_descriptor_entry(addr, NORMAL_MEM, USER_AP | DISALLOW_KERNEL_AP)
+            } else if addr < PERIPHERAL_BASE as u64 {
+                block_descriptor_entry(addr, NORMAL_MEM, KERNEL_AP)
+            } else {
+                block_descriptor_entry(addr, DEVICE_MEM, USER_AP)
+            };
+
+            write_volatile(
+                (&raw mut __translation_table_l2_start).byte_add(8 * i),
+                descriptor,
+            );
+        }
+    }
+}
+
+const BLOCK: u64 = 0b01;
+const TABLE: u64 = 0b11;
+
+const USER_AP: u64 = 1 << 6;
+const KERNEL_AP: u64 = 0 << 7;
+const DISALLOW_KERNEL_AP: u64 = 1 << 7;
+const ACCESS_FLAG: u64 = 1 << 10;
+const INNER_SHAREABILITY: u64 = 0b11 << 8;
+
+const NORMAL_MEM: u64 = 0 << 2;
+const DEVICE_MEM: u64 = 1 << 2;
+
+pub fn block_descriptor_entry(addr: u64, mair_index: u64, additional_flags: u64) -> u64 {
+    let pxn = 0 << 53; // allow EL1 execution
+    let uxn = 0 << 54; // allow EL0 execution
+
+    (addr & 0x0000_FFFF_FFE0_0000)
+        | BLOCK
+        | mair_index
+        | ACCESS_FLAG
+        | pxn
+        | uxn
+        | INNER_SHAREABILITY
+        | additional_flags
+}
+
+pub fn table_descriptor_entry(addr: u64) -> u64 {
+    0 | (addr & 0x0000_FFFF_FFFF_F000) | TABLE
 }

src/aarch64/registers.rs

@@ -52,6 +52,8 @@ psr!(SPSR_EL1, u32);
 
 psr!(ELR_EL1, u32);
 
+psr!(SCTLR_EL1, u32);
+
 pub fn read_exception_source_el() -> u32 {
     read_spsr_el1() & 0b1111
 }

src/configuration.rs

@@ -1,16 +1,37 @@
-static SCTLR_EL1_MMU_DISABLED: u64 = 0; //M
+const SCTLR_EL1_MMU_ENABLED: u64 = 1; //M
-static SCTLR_EL1_DATA_CACHE_DISABLED: u64 = 0 << 2; //C
+const SCTLR_EL1_DATA_CACHE_DISABLED: u64 = 0 << 2; //C
-static SCTLR_EL1_INSTRUCTION_CACHE_DISABLED: u64 = 0 << 12; //I
+const SCTLR_EL1_INSTRUCTION_CACHE_DISABLED: u64 = 0 << 12; //I
-static SCTLR_EL1_LITTLE_ENDIAN_EL0: u64 = 0 << 24; //E0E
+const SCTLR_EL1_LITTLE_ENDIAN_EL0: u64 = 0 << 24; //E0E
-static SCTLR_EL1_LITTLE_ENDIAN_EL1: u64 = 0 << 25; //EE
+const SCTLR_EL1_LITTLE_ENDIAN_EL1: u64 = 0 << 25; //EE
+const SCTLR_EL1_SPAN: u64 = 1 << 23; //SPAN
 
 #[allow(clippy::identity_op)]
-static SCTLR_EL1_RES: u64 = (0 << 6) | (1 << 11) | (0 << 17) | (1 << 20) | (1 << 22); //Res0 & Res1
+const SCTLR_EL1_RES: u64 = (0 << 6) | (1 << 11) | (0 << 17) | (1 << 20) | (1 << 22); //Res0 & Res1
 
 #[no_mangle]
-pub static SCTLR_EL1_CONF: u64 = SCTLR_EL1_MMU_DISABLED
+pub static SCTLR_EL1_CONF: u64 = SCTLR_EL1_MMU_ENABLED
     | SCTLR_EL1_DATA_CACHE_DISABLED
     | SCTLR_EL1_INSTRUCTION_CACHE_DISABLED
     | SCTLR_EL1_LITTLE_ENDIAN_EL0
     | SCTLR_EL1_LITTLE_ENDIAN_EL1
-    | SCTLR_EL1_RES;
+    | SCTLR_EL1_RES
+    | SCTLR_EL1_SPAN;
+
+// TODO: Document magic numbers
+
+const TG0: u64 = 0b00 << 14;
+const TG1: u64 = 0b10 << 30;
+
+const T0SZ: u64 = 27;
+const T1SZ: u64 = 27 << 16;
+
+const SH0: u64 = 0b11 << 12;
+const SH1: u64 = 0b11 << 28;
+
+const IPS: u64 = 0b000 << 32;
+const EPD1: u64 = 0b1 << 23;
+
+const AS: u64 = 0b1 << 36;
+const TBI0: u64 = 0b1 << 38;
+#[no_mangle]
+pub static TCR_EL1_CONF: u64 = IPS | TG0 | TG1 | T0SZ | T1SZ | SH0 | SH1 | EPD1 | AS | TBI0;

src/interrupt_handlers.rs

@@ -5,7 +5,7 @@ use alloc::vec::Vec;
 use crate::{
     aarch64::registers::{
         daif::{mask_all, unmask_irq},
-        read_esr_el1, read_exception_source_el,
+        read_elr_el1, read_esr_el1, read_exception_source_el,
     },
     get_current_el,
     peripherals::{
@@ -118,15 +118,17 @@ unsafe extern "C" fn rust_synchronous_interrupt_imm_lower_aarch64() {
     println!("--------Sync Exception in EL{}--------", source_el);
     println!("Exception escalated to EL {}", get_current_el());
     println!("Current EL: {}", get_current_el());
-    let esr = EsrElX::from(read_esr_el1());
+    let esr: EsrElX = EsrElX::from(read_esr_el1());
-    println!("{:?}", EsrElX::from(esr));
+    println!("{:?}", esr);
-    println!("Return register address: {:#x}", read_esr_el1());
+    println!("Return address: {:#x}", read_elr_el1());
 
     match esr.ec {
         0b100100 => {
             println!("Cause: Data Abort from a lower Exception level");
         }
-        _ => {}
+        _ => {
+            println!("Unknown Error Code: {:b}", esr.ec);
+        }
     }
     println!("-------------------------------------");
 
@@ -136,7 +138,9 @@ unsafe extern "C" fn rust_synchronous_interrupt_imm_lower_aarch64() {
 fn clear_interrupt_for_source(source: IRQSource) {
     match source {
         IRQSource::UartInt => clear_uart_interrupt_state(),
-        _ => {}
+        _ => {
+            todo!()
+        }
     }
 }
 
@@ -212,6 +216,7 @@ pub fn get_irq_pending_sources() -> u64 {
     pending
 }
 
+#[inline(always)]
 pub fn initialize_interrupt_handler() {
     unsafe { INTERRUPT_HANDLERS = Some(Vec::new()) };
 }

src/main.rs

@@ -12,7 +12,10 @@ extern crate alloc;
 
 use alloc::boxed::Box;
 use nova::{
-    aarch64::registers::{daif, read_id_aa64mmfr0_el1, read_tcr_el1},
+    aarch64::{
+        mmu::init_translation_table,
+        registers::{daif, read_id_aa64mmfr0_el1},
+    },
     framebuffer::{FrameBuffer, BLUE, GREEN, RED},
     get_current_el, init_heap,
     interrupt_handlers::{enable_irq_source, IRQSource},
@@ -33,6 +36,7 @@ global_asm!(include_str!("vector.S"));
 extern "C" {
     fn el2_to_el1();
     fn el1_to_el0();
+    fn configure_mmu_el1();
     static mut __bss_start: u32;
     static mut __bss_end: u32;
 }
@@ -63,7 +67,14 @@ pub extern "C" fn main() -> ! {
     println!("Exception level: {}", get_current_el());
 
     unsafe {
-        asm!("mrs x0, SCTLR_EL1",);
+        init_heap();
+        init_translation_table();
+        configure_mmu_el1();
+    };
+
+    println!("AA64 {:064b}", read_id_aa64mmfr0_el1());
+
+    unsafe {
         el2_to_el1();
     }
 
@@ -82,14 +93,10 @@ unsafe fn zero_bss() {
 #[no_mangle]
 pub extern "C" fn kernel_main() -> ! {
     nova::initialize_kernel();
-    println!("Kernel Main");
     println!("Exception Level: {}", get_current_el());
     daif::unmask_all();
 
     unsafe {
-        init_heap();
-        println!("{:b}", read_id_aa64mmfr0_el1());
-        println!("{:b}", read_tcr_el1());
         el1_to_el0();
     };
 

src/peripherals/uart.rs

@@ -118,11 +118,13 @@ fn uart_fifo_enable(enable: bool) {
     unsafe { write_address(UART0_LCRH, lcrh) };
 }
 
+#[inline(always)]
 fn uart_enable_rx_interrupt() {
     unsafe { write_address(UART0_IMSC, UART0_IMSC_RXIM) };
 }
 
 /// Set UART word length and set FIFO status
+#[inline(always)]
 fn uart_set_lcrh(wlen: u32, enable_fifo: bool) {
     let mut value = (wlen & 0b11) << 5;
     if enable_fifo {
@@ -131,10 +133,12 @@ fn uart_set_lcrh(wlen: u32, enable_fifo: bool) {
     unsafe { write_address(UART0_LCRH, value) };
 }
 
+#[inline(always)]
 pub fn read_uart_data() -> char {
     (unsafe { read_address(UART0_DR) } & 0xFF) as u8 as char
 }
 
+#[inline(always)]
 pub fn clear_uart_interrupt_state() {
     unsafe {
         write_address(UART0_ICR, 1 << 4);

src/vector.S

@@ -1,5 +1,5 @@
 
-.global vector_table
+.global v_table
 .extern irq_handler
 
 .macro    ventry    label
@@ -7,7 +7,7 @@
 b    \label
 .endm
 
-.section .vector_table, "ax"
+.section .vector_table  , "ax"
 vector_table:
     ventry .
     ventry .
@@ -21,12 +21,18 @@ vector_table:
 
     ventry synchronous_interrupt_imm_lower_aarch64
     ventry irq_handler
+    ventry .
+    ventry .
+
+    ventry .
+    ventry .
+    ventry .
+    ventry .
 
 
 .align 4
 .global el2_to_el1
 el2_to_el1:
-
     mov x0, #(1 << 31)
     msr HCR_EL2, x0
 
@@ -46,9 +52,13 @@ el2_to_el1:
     adr x0, vector_table
     msr VBAR_EL1, x0
 
-    // Disable MMU
+    isb
-    ldr x0, =SCTLR_EL1_CONF
+
-    msr sctlr_el1, x0
+    adrp x0, SCTLR_EL1_CONF
+    ldr  x1, [x0, :lo12:SCTLR_EL1_CONF]
+    msr SCTLR_EL1, x0
+
+    isb
 
     // SIMD should not be trapped
     mrs x0, CPACR_EL1
@@ -56,9 +66,33 @@ el2_to_el1:
     orr x0,x0, x1
     msr CPACR_EL1,x0
 
+    isb
+
     // Return to EL1
     eret
 
+.align 4
+.global configure_mmu_el1
+configure_mmu_el1:
+    adrp x0, TCR_EL1_CONF
+    ldr  x1, [x0, :lo12:TCR_EL1_CONF]
+    msr TCR_EL1, x0
+    isb
+
+    mov x0, #0x04FF
+    msr MAIR_EL1, x0
+    isb
+
+    ldr x0, =__translation_table_l1_start
+    msr TTBR0_EL1, x0
+    msr TTBR1_EL1, x0
+
+    tlbi vmalle1
+    dsb ish
+    isb
+
+    ret
+
 .align 4
 .global el1_to_el0
 el1_to_el0:
@@ -75,6 +109,8 @@ el1_to_el0:
     ldr x0, =__stack_end_el0
     msr SP_EL0, x0
 
+    isb
+
     // Return to EL0
     eret
 
@@ -174,3 +210,10 @@ synchronous_interrupt_no_el_change:
     add sp, sp, #176
 
     eret
+
+.align 4
+.global pan
+pan:
+    mov x0, #0
+    msr S3_0_C4_C2_3, x0
+    ret

tools/start_simulator.sh

@@ -1,3 +1,5 @@
+set -e
+
 cargo build --target aarch64-unknown-none --release
 cd "$(dirname "$0")"
 

tools/start_simulator_debug.sh

@@ -1,3 +1,5 @@
+set -e
+
 cargo build --target aarch64-unknown-none
 
 cd "$(dirname "$0")"