Julia Evans

Day 39: Writing malloc wrong, for fun

My major achievement for today is writing the following five lines of code:

let a: ~u8 = ~('A' as u8);
stdio::putc(*a);
let b: ~u8 = ~('B' as u8);
stdio::putc(*a);
stdio::putc(*b);

and having them do the wrong thing. One would normally expect this to print “AAB”. But for me, right now, until I stop goofing off, it prints “ABB”. Why is that?

Well, it’s because my malloc implementation looks like this:

static mut base: uint = 0x200000;
pub extern "C" fn malloc(len: uint) -> *mut u8 {
    unsafe {
        let ret: uint = base;
        return base as *mut u8;
   }
}

This means that every time I allocate memory, I get the same pointer back, and so a and b will always be equal no matter what. And for that matter any variable I create will always have the same value. This is of course a terrible idea in real life, but it is really fun.

Here’s my real malloc function (that causes the above code to print “AAB”, like it should):

pub extern "C" fn malloc(len: uint) -> *mut u8 {
    unsafe {
        let ret: uint = base;
        base += len + size_of::<uint>();

        // Align next allocation to 4-byte boundary.
        if(base % 4 != 0) {
            base += 4 - (base % 4);
        }

        *(base as *mut uint) = len;

        return (ret + size_of::<uint>()) as *mut u8;
    }
}

pub extern "C" fn free(ptr: *mut u8) {
    // meh.
}

The hardest part about this was not actually writing malloc. Writing malloc is easy, as long as you never need to free memory. I also just wrote this by copying it from a C implementation. You just need to keep a counter and keep incrementing it.

The hard part was getting the type of the function right, because Rust :). This is entirely made up for by being able to play silly memory games.

Day 40: 12 things I learned today about linkers. Day 41: Linkers are upsetting