blog_os/_posts/2015-09-02-setup-rust.md

layout, title, category

layout	title	category
post	Setup Rust	rust-os

In the previous posts we created a [minimal Multiboot kernel][multiboot post] and [switched to Long Mode][long mode post]. Now we can finally switch to sweet Rust code. [Rust] is a beautiful high-level language without runtime. It allows us to not link the standard library and write bare metal code. Unfortunately the setup is not quite hassle-free yet.

This blog post tries to setup Rust step-by-step and point out the different problems. If you have any questions, problems, or suggestions please [file an issue] or create a comment at the bottom. The code from this post is in a [Github repository], too.

[multiboot post]: {{ site.url }}{{ page.previous.previous.url }} [long mode post]: {{ site.url }}{{ page.previous.url }} [Rust]: https://www.rust-lang.org/ [file an issue]: https://github.com/phil-opp/phil-opp.github.io/issues [Github repository]: https://github.com/phil-opp/blogOS/tree/setup_rust

Installing Rust

We need a nightly compiler, as we need to use many unstable features. To manage Rust installations I highly recommend brson's multirust. It allows you to install nightly, beta, and stable compilers side-by-side and makes it easy to update them. After installing you can just run multirust update nightly to install or update to the latest Rust nightly.

Creating a Rust project

Normally you would call cargo new when you want to create a new project folder. We can't use it because our folder already exists, so we need to do it manually. Fortunately we only need to add a cargo configuration file named Cargo.toml:

[package]
name = "blog_os"
version = "0.1.0"
authors = ["Philipp Oppermann <dev@phil-opp.com>"]

[lib]
crate-type = ["staticlib"]

The package section contains basic project metadata and is identical to the Cargo.toml created by cargo new blog_os. The lib section specifies that we want to build a static library, i.e. a library that contains all of its dependencies.

Now we place our root source file in src/lib.rs:

#![feature(no_std, lang_items)]
#![no_std]

#[no_mangle]
pub extern fn rust_main() {}

#[lang = "eh_personality"] extern fn eh_personality() {}
#[lang = "panic_fmt"] extern fn panic_fmt() -> ! {loop{}}

Let's break it down:

• #! defines an attribute of the current module. Since we are at the root module, they apply to the crate itself.
• The features attribute is used to allow the specified feature-gated attributes in this crate. You can't do that in a stable/beta compiler, so this is one reason we need a Rust nighly.
• The no_std attribute prevents the automatic linking of the standard library. We can't use std because it relies on operating system features like files, system calls, and various device drivers. Remember that currently the only “feature” of our OS is printing OKAY :).
• A # without a ! afterwards defines an attribute for the following item (a function in our case).
• The no_mangle attribute disables the automatic name mangling that Rust uses to get unique function names. We want to do a call rust_main from our assembly code, so this function name must stay as it is.
• We mark our main function as extern to make it compatible to the standard C calling convention.
• The lang attribute defines a Rust language item.
• The eh_personality function is used for Rust's unwinding on panic!. We can leave it empty since we don't have any unwinding support in our OS yet.
• The panic_fmt function is the entry point on panic. Right now we can't do anything useful, so we just make sure that it doesn't return (required by the ! return type).

Building Rust

We can now build it using cargo build. It creates a static library at target/debug/libblog_os.a, which can be linked with our assembly kernel. To build and link the rust library on make, we extend our Makefile(full file):

# ...
rust_os := target/debug/libblog_os.a
# ...
$(kernel): cargo $(rust_os) $(assembly_object_files) $(linker_script)
       @ld -n -T $(linker_script) -o $(kernel) $(assembly_object_files) $(rust_os)

cargo:
       @cargo build

We added a new cargo target that just executes cargo build and modified the $(kernel) target to link the created static lib .

But now cargo build is executed on every make, even if no source file was changed. And the ISO is recreated on every make iso/make run, too. We could try to avoid this by adding dependencies on all rust source and cargo configuration files to the cargo target, but the ISO creation takes only half a second on my machine and most of the time we will have changed a Rust file when we run make. So we keep it simple for now and let cargo do the bookkeeping of changed files (it does it anyway).

Calling Rust

Now we can call the main method in long_mode_start:

bits 64
long_mode_start:
    ; call the rust main
    extern rust_main     ; new
    call rust_main       ; new

    ; print `OKAY` to screen
    mov rax, 0x2f592f412f4b2f4f
    mov qword [0xb8000], rax
    hlt

By defining rust_main as extern we tell nasm that the function is defined in another file. As the linker takes care of linking them together, we'll get a linker error if we have a typo in the name or forget to mark the rust function as pub extern.

When we've done everything right, we still see the green OKAY when executing make run. That means that we successfully called the Rust function and returned back to assembly.

Testing

Now we can try some Rust code:

pub extern fn rust_main() {
    let x = ["Hello", " ", "World", "!"];
}

When we test it using make run, it fails with undefined reference to 'memcpy'. This function is one of the basic functions of the C library (libc). Usually the libc crate is linked to every Rust program together with the standard library but we opted out through #![no_std]. We could try to fix this by adding the libc crate as extern crate. But libc is just a wrapper for the system libc, for example glibc on Linux, so this won't work for us. Instead we need to recreate the basic libc functions such as memcpy, memmove, memset, and memcmp in Rust.

rlibc

Fortunately there already is a crate for that: rlibc. When we look at its source code we see that it contains no magic, just some raw pointer operations in a while loop. To add it as a dependency we just need to add two lines to the Cargo.toml:

...
[dependencies]
rlibc = "*"

and an extern crate in our src/lib.rs:

...
extern crate rlibc;

#[no_mangle]
pub extern fn rust_main() {
...

Now make run doesn't complain about memcpy anymore. Instead it will show a pile of new errors:

target/debug/libblog_os.a(core-35017696.0.o): In function `ops::f32.Rem::rem::hfcbbcbe5711a6e6emxm':
core.0.rs:(.text._ZN3ops7f32.Rem3rem20hfcbbcbe5711a6e6emxmE+0x1): undefined reference to `fmodf'
target/debug/libblog_os.a(core-35017696.0.o): In function `ops::f64.Rem::rem::hbf225030671c7a35Txm':
core.0.rs:(.text._ZN3ops7f64.Rem3rem20hbf225030671c7a35TxmE+0x1): undefined reference to `fmod'
...

--gc-sections

The new errors are linker errors about missing fmod and fmodf functions. These functions are used for the modulo operation (%) on floating point numbers in libcore. The core library is added implicitly when using #![no_std] and provides basic standard library features like Option or Iterator. According to the documentation it is “dependency-free”. But it actually has some dependencies, for example on fmod and fmodf.

So how do we fix this problem? We don't use any floating point operations, so we could just provide our own implementations of fmod and fmodf that just do a loop{}. But there's a better way that doesn't fail silently when we use floats some day. We tell the linker to remove unused sections. That's generally a good idea as it reduces kernel size. And we don't have any references to fmod and fmodf anymore until we use floating point modulo. The magic linker flag is --gc-sections, which stands for “garbage collect sections”. Let's add it to the $(kernel) target in our Makefile:

$(kernel): cargo $(rust_os) $(assembly_object_files) $(linker_script)
	@ld -n --gc-sections -T $(linker_script) -o $(kernel) $(assembly_object_files) $(rust_os)

Now we can do a make run again and… it doesn't boot anymore:

GRUB error: no multiboot header found.

What happened? Well, the linker removed unused sections. And since we don't use the Multiboot section anywhere ld removes it, too. So we need to tell the linker explicitely that it should keep this section. The KEEP command does exactly that, so we add it to the linker.ld:

.boot :
{
    /* ensure that the multiboot header is at the beginning */
    KEEP(*(.multiboot))
}

Now everything should work again (the green OKAY). But there is another linking issue, which is triggered by some other example code.

no-landing-pads

The following snippet still fails:

    ...
    let test = (0..3).flat_map(|x| 0..x).zip(0..);

The error is a linker error again (hence the ugly error message):

target/debug/libblog_os.a(blog_os.0.o): In function `blog_os::iter::Iterator::zip<core::iter::FlatMap<core::ops::Range<i32>, core::ops::Range<i32>, closure>,core::ops::RangeFrom<i32>>':
/home/.../src/libcore/iter.rs:223: undefined reference to `_Unwind_Resume'

So the linker can't find a function named _Unwind_Resume that is referenced in iter.rs:223 in libcore. This reference is inserted by the compiler, it's not really in the libcore source. The inserted code is a so-called landing pad that is used for exception handling. The easiest way of fixing this problem is to disable the landing pad creation since we don't supports panics anyway right now. We can do this by passing a -Z no-landing-pads flag to rustc. To do this we replace the cargo build command in our Makefile:

cargo:
    @cargo rustc -- -Z no-landing-pads

Now we fixed all linking issues.

The final problem

Unfortunately there is one last problem left that gets triggered by the following code:

let mut a = 42;
a += 1;

When we add that code to rust_main and test it using make run, the OS will constantly reboot itself. Let's try to debug it.

Debugging

Such a boot loop is most likely caused by some CPU exception. When these exceptions aren't handled, a Triple Fault occurs and the processor resets itself. We can look at generated CPU interrupts/exceptions using QEMU:

> qemu-system-x86_64 -d int -no-reboot -hda build/os-x86_64.iso
SMM: enter
...
SMM: after RSM
...
check_exception old: 0xffffffff new 0x6
     0: v=06 e=0000 i=0 cpl=0 IP=0008:0000000000100200 pc=0000000000100200
     SP=0010:0000000000102fd0 env->regs[R_EAX]=0000000080010010
...
check_exception old: 0xffffffff new 0xd
     1: v=0d e=0062 i=0 cpl=0 IP=0008:0000000000100200 pc=0000000000100200
     SP=0010:0000000000102fd0 env->regs[R_EAX]=0000000080010010
...
check_exception old: 0xd new 0xd
     2: v=08 e=0000 i=0 cpl=0 IP=0008:0000000000100200 pc=0000000000100200
     SP=0010:0000000000102fd0 env->regs[R_EAX]=0000000080010010
...
check_exception old: 0x8 new 0xd

Let me first explain the QEMU arguments: The -d int logs CPU interrupts to the console and the -no-reboot flag closes QEMU instead of constant rebooting. But what does the cryptical output mean? I already removed most of it as we don't need it here. Let's break down the rest:

• The first two blocks, SMM: enter and SMM: after RSM are created before our OS boots, so we just ignore them.
• The next block, check_exception old: 0xffffffff new 0x6 is the interesting one. It says: “a new CPU exception with number 0xe occurred“.
• The last blocks indicate further exceptions. They were thrown because we didn't handle the 0x6 exception, so we're going to ignore them, too.

So let's look at the first exception: old:0xffffffff means that the CPU wasn't handling an interrupt when the exception occurred. The new exception has number 0x6. By looking at an exception table we learn that 0x6 indicates a Invalid Opcode fault. So the lastly executed instruction was invalid. The register dump tells us that the current instruction was 0x100200 (through IP (instruction pointer) or pc (program counter)). Therefore the instruction at 0x100200 seems to be invalid. We can look at it using objdump:

> objdump -D build/kernel-x86_64.bin | grep "100200:"
100200:	0f 28 05 49 01 00 00 	movaps 0x149(%rip),%xmm0 ...

Through objdump -D we disassemble our whole kernel and grep picks the relevant line. The instruction at 0x100200 seems to be a valid movaps instruction. It's a SSE instruction that moves 128 bit between memory and SSE-registers (e.g. xmm0). But why the Invalid Opcode exception? The answer is hidden behind the movaps documentation: The section Protected Mode Exceptions lists the conditions for the various exceptions. The short code of the Invalid Opcode is #UD, so the exception occurs

For an unmasked Streaming SIMD Extensions 2 instructions numeric exception (CR4.OSXMMEXCPT =0). If EM in CR0 is set. If OSFXSR in CR4 is 0. If CPUID feature flag SSE2 is 0.

The rough translation of this cryptic definition is: If SSE isn't enabled. So apparently Rust uses SSE instructions by default and we didn't enable SSE before. So the fix for this bug is enabling SSE.

Enabling SSE

To enable SSE, assembly code is needed again. We want to add a function that tests if SSE is available and enables it then. Else we want to print an error message. But we can't use our existing error function because it uses (now invalid) 32-bit instructions. So we need a new one (in long_mode_init.asm):

; Prints `ERROR: ` and the given error code to screen and hangs.
; parameter: error code (in ascii) in al
error:
    mov rbx, 0x4f4f4f524f524f45
    mov [0xb8000], rbx
    mov rbx, 0x4f204f204f3a4f52
    mov [0xb8008], rbx
    mov byte [0xb800e], al
    hlt
    jmp error

It's the nearly the same as the 32-bit code in the [previous post][32-bit error function] (instead of ERR: we print ERROR: here). Now we can add a function that checks for SSE and enables it:

; Check for SSE and enable it. If it's not supported throw error "a".
setup_SSE:
    ; check for SSE
    mov rax, 0x1
    cpuid
    test edx, 1<<25
    jz .no_SSE

    ; enable SSE
    mov rax, cr0
    and ax, 0xFFFB      ; clear coprocessor emulation CR0.EM
    or ax, 0x2          ; set coprocessor monitoring  CR0.MP
    mov cr0, rax
    mov rax, cr4
    or ax, 3 << 9       ; set CR4.OSFXSR and CR4.OSXMMEXCPT at the same time
    mov cr4, rax

    ret
.no_SSE:
    mov al, "a"
    jmp error

The code is from the great [OSDev Wiki][osdev sse] again. Notice that it sets/unsets exactly the bits that can cause the Invalid Opcode exception.

When we insert a call setup_SSE right before calling rust_main, our Rust code will finally work.

[32-bit error function]: {{ site.url }}{{ page.previous.url }}#some-tests [osdev sse]: http://wiki.osdev.org/SSE#Checking_for_SSE

“OS returned!”

Now that we're editing assembly anyway, we should change the OKAY message to something more meaningful. My suggestion is a red OS returned!:

...
call rust_main

.os_returned:
    ; rust main returned, print `OS returned!`
    mov rax, 0x4f724f204f534f4f
    mov [0xb8000], rax
    mov rax, 0x4f724f754f744f65
    mov [0xb8008], rax
    mov rax, 0x4f214f644f654f6e
    mov [0xb8010], rax
    hlt

Ok that's enough assembly for now, let's switch back to Rust and print the traditional Hello World!.

Hello World!

Of course we could just write some magic bytes to memory like we did in assembly. But I chose some code that highlights some of Rust's features:

#![feature(no_std, lang_items)]
#![feature(core_slice_ext, core_str_ext, core_intrinsics)]
#![no_std]

extern crate rlibc;

use core::intrinsics::offset;

#[no_mangle]
pub extern fn rust_main() {
    // ATTENTION: we have a very small stack and no guard page
    let x = ["Hello", " ", "World", "!"];
    let screen_pointer = 0xb8000 as *const u16;

    for (byte, i) in x.iter().flat_map(|s| s.bytes()).zip(0..) {
        let c = 0x1f00 | (byte as u16);
        unsafe {
            let screen_char = offset(screen_pointer, i) as *mut u16;
            *screen_char = c
        }
    }

    loop{}
}

Let's break it down:

• the core_* features are needed because many functions from the core library are still unstable
• use imports a function, in our case offset that can be used for C-like pointer arithmetic
• Normally, Rust tries to avoid stack overflows through guard pages: The page below the stack isn't mapped and such a stack overflow triggers a page fault. But we can't unmap this page as we use just one big page right now, so we need to be very careful to not blow our small stack. Fortunately our stack is just above the page tables. A stack overflow would overwrite an important entry at some point and thus provoke a page fault (that crashes our kernel).
• x is an immutable array of &str
• screen_pointer is a raw pointer to the VGA text buffer
• We iterate over the elements of x, convert every &str to a byte iterator (s.bytes()), and join these iterators (through flat_map). Then we convert the byte iterator to a (byte, index) operator through the zip method.
• By casting the byte as u16 and OR-ing the colors in (0x1f is white text on blue background), we convert it to a screen character for the VGA buffer.
• Now we just need to write the screen_char to the right place in memory. Rust doesn't know the VGA buffer and can't guarantee that writing to it is safe (we could corrupt important data). So we need to tell Rust that we know what we're doing by using an unsafe block.
• offset is like pointer arithmetic in C: It calculates the address of the i-th u16 (starting at 0xb8000) and converts it to a pointer.
• To be able to write to it, we must cast the *const u16 to a *mut u16

What's next?

Until now we write magic bits to memory every time we want to print something. It's time to end this hackery. In the [next post] we create a VGA buffer module that allows us to print strings in different colors through a nice interface.

[next post]: {{ site.url }}{{ page.next.url }}