Merge pull request #123 from coastalwhite/r0-in-asm

Implement r0 crate in assembly

CHANGELOG.md

@@ -16,6 +16,7 @@ and this project adheres to [Semantic Versioning](http://semver.org/).
 - Use inline assembly instead of pre-compiled blobs
 - Removed bors in favor of GitHub Merge Queue
 - `start_trap_rust` is now marked as `unsafe`
+- Implement `r0` as inline assembly
 
 ## [v0.11.0] - 2023-01-18
 

Cargo.toml

@@ -15,7 +15,6 @@ s-mode = []
 single-hart = []
 
 [dependencies]
-r0 = "1.0.0"
 riscv = "0.10"
 riscv-rt-macros = { path = "macros", version = "0.2.0" }
 

build.rs

@@ -5,9 +5,17 @@ use std::env;
 use std::fs;
 use std::path::PathBuf;
 
+fn add_linker_script(bytes: &[u8]) {
+    let out_dir = PathBuf::from(env::var("OUT_DIR").unwrap());
+
+    // Put the linker script somewhere the linker can find it
+    fs::write(out_dir.join("link.x"), bytes).unwrap();
+    println!("cargo:rustc-link-search={}", out_dir.display());
+    println!("cargo:rerun-if-changed=link.x");
+}
+
 fn main() {
     let target = env::var("TARGET").unwrap();
-    let out_dir = PathBuf::from(env::var("OUT_DIR").unwrap());
     let _name = env::var("CARGO_PKG_NAME").unwrap();
 
     // set configuration flags depending on the target
@@ -17,19 +25,16 @@ fn main() {
         match target.bits {
             32 => {
                 println!("cargo:rustc-cfg=riscv32");
+                add_linker_script(include_bytes!("link-rv32.x"));
             }
             64 => {
                 println!("cargo:rustc-cfg=riscv64");
+                add_linker_script(include_bytes!("link-rv64.x"));
             }
             _ => panic!("Unsupported bit width"),
         }
         if target.has_extension('m') {
             println!("cargo:rustc-cfg=riscvm"); // we can expose extensions this way
         }
     }
-
-    // Put the linker script somewhere the linker can find it
-    fs::write(out_dir.join("link.x"), include_bytes!("link.x")).unwrap();
-    println!("cargo:rustc-link-search={}", out_dir.display());
-    println!("cargo:rerun-if-changed=link.x");
 }

link.x → link-rv32.x

@@ -84,7 +84,7 @@ SECTIONS
     _edata = .;
   } > REGION_DATA AT > REGION_RODATA
 
-  .bss (NOLOAD) :
+  .bss (NOLOAD) : ALIGN(4)
   {
     _sbss = .;
     *(.sbss .sbss.* .bss .bss.*);

link-rv64.x

@@ -0,0 +1,174 @@
+PROVIDE(_stext = ORIGIN(REGION_TEXT));
+PROVIDE(_stack_start = ORIGIN(REGION_STACK) + LENGTH(REGION_STACK));
+PROVIDE(_max_hart_id = 0);
+PROVIDE(_hart_stack_size = 2K);
+PROVIDE(_heap_size = 0);
+
+PROVIDE(UserSoft = DefaultHandler);
+PROVIDE(SupervisorSoft = DefaultHandler);
+PROVIDE(MachineSoft = DefaultHandler);
+PROVIDE(UserTimer = DefaultHandler);
+PROVIDE(SupervisorTimer = DefaultHandler);
+PROVIDE(MachineTimer = DefaultHandler);
+PROVIDE(UserExternal = DefaultHandler);
+PROVIDE(SupervisorExternal = DefaultHandler);
+PROVIDE(MachineExternal = DefaultHandler);
+
+PROVIDE(DefaultHandler = DefaultInterruptHandler);
+PROVIDE(ExceptionHandler = DefaultExceptionHandler);
+
+/* # Pre-initialization function */
+/* If the user overrides this using the `#[pre_init]` attribute or by creating a `__pre_init` function,
+   then the function this points to will be called before the RAM is initialized. */
+PROVIDE(__pre_init = default_pre_init);
+
+/* A PAC/HAL defined routine that should initialize custom interrupt controller if needed. */
+PROVIDE(_setup_interrupts = default_setup_interrupts);
+
+/* # Multi-processing hook function
+   fn _mp_hook() -> bool;
+
+   This function is called from all the harts and must return true only for one hart,
+   which will perform memory initialization. For other harts it must return false
+   and implement wake-up in platform-dependent way (e.g. after waiting for a user interrupt).
+*/
+PROVIDE(_mp_hook = default_mp_hook);
+
+/* # Start trap function override
+  By default uses the riscv crates default trap handler
+  but by providing the `_start_trap` symbol external crates can override.
+*/
+PROVIDE(_start_trap = default_start_trap);
+
+SECTIONS
+{
+  .text.dummy (NOLOAD) :
+  {
+    /* This section is intended to make _stext address work */
+    . = ABSOLUTE(_stext);
+  } > REGION_TEXT
+
+  .text _stext :
+  {
+    /* Put reset handler first in .text section so it ends up as the entry */
+    /* point of the program. */
+    KEEP(*(.init));
+    KEEP(*(.init.rust));
+    . = ALIGN(4);
+    *(.trap);
+    *(.trap.rust);
+    *(.text.abort);
+    *(.text .text.*);
+  } > REGION_TEXT
+
+  .rodata : ALIGN(4)
+  {
+    *(.srodata .srodata.*);
+    *(.rodata .rodata.*);
+
+    /* 4-byte align the end (VMA) of this section.
+       This is required by LLD to ensure the LMA of the following .data
+       section will have the correct alignment. */
+    . = ALIGN(4);
+  } > REGION_RODATA
+
+  .data : ALIGN(8)
+  {
+    _sidata = LOADADDR(.data);
+    _sdata = .;
+    /* Must be called __global_pointer$ for linker relaxations to work. */
+    PROVIDE(__global_pointer$ = . + 0x800);
+    *(.sdata .sdata.* .sdata2 .sdata2.*);
+    *(.data .data.*);
+    . = ALIGN(8);
+    _edata = .;
+  } > REGION_DATA AT > REGION_RODATA
+
+  .bss (NOLOAD) : ALIGN(8)
+  {
+    _sbss = .;
+    *(.sbss .sbss.* .bss .bss.*);
+    . = ALIGN(8);
+    _ebss = .;
+  } > REGION_BSS
+
+  /* fictitious region that represents the memory available for the heap */
+  .heap (NOLOAD) :
+  {
+    _sheap = .;
+    . += _heap_size;
+    . = ALIGN(4);
+    _eheap = .;
+  } > REGION_HEAP
+
+  /* fictitious region that represents the memory available for the stack */
+  .stack (NOLOAD) :
+  {
+    _estack = .;
+    . = ABSOLUTE(_stack_start);
+    _sstack = .;
+  } > REGION_STACK
+
+  /* fake output .got section */
+  /* Dynamic relocations are unsupported. This section is only used to detect
+     relocatable code in the input files and raise an error if relocatable code
+     is found */
+  .got (INFO) :
+  {
+    KEEP(*(.got .got.*));
+  }
+
+  .eh_frame (INFO) : { KEEP(*(.eh_frame)) }
+  .eh_frame_hdr (INFO) : { *(.eh_frame_hdr) }
+}
+
+/* Do not exceed this mark in the error messages above                                    | */
+ASSERT(ORIGIN(REGION_TEXT) % 4 == 0, "
+ERROR(riscv-rt): the start of the REGION_TEXT must be 4-byte aligned");
+
+ASSERT(ORIGIN(REGION_RODATA) % 4 == 0, "
+ERROR(riscv-rt): the start of the REGION_RODATA must be 4-byte aligned");
+
+ASSERT(ORIGIN(REGION_DATA) % 8 == 0, "
+ERROR(riscv-rt): the start of the REGION_DATA must be 8-byte aligned");
+
+ASSERT(ORIGIN(REGION_HEAP) % 4 == 0, "
+ERROR(riscv-rt): the start of the REGION_HEAP must be 4-byte aligned");
+
+ASSERT(ORIGIN(REGION_TEXT) % 4 == 0, "
+ERROR(riscv-rt): the start of the REGION_TEXT must be 4-byte aligned");
+
+ASSERT(ORIGIN(REGION_STACK) % 4 == 0, "
+ERROR(riscv-rt): the start of the REGION_STACK must be 4-byte aligned");
+
+ASSERT(_stext % 4 == 0, "
+ERROR(riscv-rt): `_stext` must be 4-byte aligned");
+
+ASSERT(_sdata % 8 == 0 && _edata % 8 == 0, "
+BUG(riscv-rt): .data is not 8-byte aligned");
+
+ASSERT(_sidata % 8 == 0, "
+BUG(riscv-rt): the LMA of .data is not 8-byte aligned");
+
+ASSERT(_sbss % 8 == 0 && _ebss % 8 == 0, "
+BUG(riscv-rt): .bss is not 8-byte aligned");
+
+ASSERT(_sheap % 4 == 0, "
+BUG(riscv-rt): start of .heap is not 4-byte aligned");
+
+ASSERT(_stext + SIZEOF(.text) < ORIGIN(REGION_TEXT) + LENGTH(REGION_TEXT), "
+ERROR(riscv-rt): The .text section must be placed inside the REGION_TEXT region.
+Set _stext to an address smaller than 'ORIGIN(REGION_TEXT) + LENGTH(REGION_TEXT)'");
+
+ASSERT(SIZEOF(.stack) > (_max_hart_id + 1) * _hart_stack_size, "
+ERROR(riscv-rt): .stack section is too small for allocating stacks for all the harts.
+Consider changing `_max_hart_id` or `_hart_stack_size`.");
+
+ASSERT(SIZEOF(.got) == 0, "
+.got section detected in the input files. Dynamic relocations are not
+supported. If you are linking to C code compiled using the `gcc` crate
+then modify your build script to compile the C code _without_ the
+-fPIC flag. See the documentation of the `gcc::Config.fpic` method for
+details.");
+
+/* Do not exceed this mark in the error messages above                                    | */

src/lib.rs

@@ -366,6 +366,8 @@
 #[cfg(riscv)]
 mod asm;
 
+use core::sync::atomic::{compiler_fence, Ordering};
+
 #[cfg(feature = "s-mode")]
 use riscv::register::{scause as xcause, stvec as xtvec, stvec::TrapMode as xTrapMode};
 
@@ -378,19 +380,6 @@ pub use riscv_rt_macros::{entry, pre_init};
 #[doc(hidden)]
 pub static __ONCE__: () = ();
 
-extern "C" {
-    // Boundaries of the .bss section
-    static mut _ebss: u32;
-    static mut _sbss: u32;
-
-    // Boundaries of the .data section
-    static mut _edata: u32;
-    static mut _sdata: u32;
-
-    // Initial values of the .data section (stored in Flash)
-    static _sidata: u32;
-}
-
 /// Rust entry point (_start_rust)
 ///
 /// Zeros bss section, initializes data section and calls main. This function never returns.
@@ -424,8 +413,93 @@ pub unsafe extern "C" fn start_rust(a0: usize, a1: usize, a2: usize) -> ! {
     if _mp_hook(hartid) {
         __pre_init();
 
-        r0::zero_bss(&mut _sbss, &mut _ebss);
-        r0::init_data(&mut _sdata, &mut _edata, &_sidata);
+        // Initialize RAM
+        // 1. Copy over .data from flash to RAM
+        // 2. Zero out .bss
+
+        #[cfg(target_arch = "riscv32")]
+        core::arch::asm!(
+            "
+                // Copy over .data
+                la      {start},_sdata
+                la      {end},_edata
+                la      {input},_sidata
+
+            	bgeu    {start},{end},2f
+            1:
+            	lw      {a},0({input})
+            	addi    {input},{input},4
+            	sw      {a},0({start})
+            	addi    {start},{start},4
+            	bltu    {start},{end},1b
+
+            2:
+                li      {a},0
+                li      {input},0
+
+                // Zero out .bss
+            	la      {start},_sbss
+            	la      {end},_ebss
+
+            	bgeu    {start},{end},3f
+            2:
+            	sw      zero,0({start})
+            	addi    {start},{start},4
+            	bltu    {start},{end},2b
+
+            3:
+                li      {start},0
+                li      {end},0
+            ",
+            start = out(reg) _,
+            end = out(reg) _,
+            input = out(reg) _,
+            a = out(reg) _,
+        );
+
+        #[cfg(target_arch = "riscv64")]
+        core::arch::asm!(
+            "
+                // Copy over .data
+                la      {start},_sdata
+                la      {end},_edata
+                la      {input},_sidata
+
+                bgeu    {start},{end},2f
+
+            1: // .data Main Loop
+            	ld      {a},0({input})
+            	addi    {input},{input},8
+            	sd      {a},0({start})
+            	addi    {start},{start},8
+            	bltu    {start},{end},1b
+            
+            2: // .data zero registers
+                li      {a},0
+                li      {input},0
+
+            	la      {start},_sbss
+            	la      {end},_ebss
+            
+                bgeu    {start},{end},4f
+
+            3: // .bss main loop
+            	sd      zero,0({start})
+            	addi    {start},{start},8
+            	bltu    {start},{end},3b
+
+            4: // .bss zero registers
+                //    Zero out used registers
+                li      {start},0
+                li      {end},0
+        ",
+            start = out(reg) _,
+            end = out(reg) _,
+            input = out(reg) _,
+            a = out(reg) _,
+        );
+
+        compiler_fence(Ordering::SeqCst);
     }
 
     // TODO: Enable FPU when available