rust-osdev/spinning_top

A simple spinlock crate based on the abstractions provided by the `lock_api` crate.

spinning_top's Issues

Reviews appreciated!

This crate is quite new and hasn't been reviewed by anyone yet. I would appreciate any reviews on the code especially on the unsafe implementation of the RawMutex trait:

spinning_top/src/spinlock.rs

Lines 29 to 57 in 5e78baa

	unsafe impl RawMutex for RawSpinlock {
	const INIT: RawSpinlock = RawSpinlock {
	locked: AtomicBool::new(false),
	};
	// A spinlock guard can be sent to another thread and unlocked there
	type GuardMarker = GuardSend;
	fn lock(&self) {
	while !self.try_lock() {
	// Wait until the lock looks unlocked before retrying
	// Code from https://github.com/mvdnes/spin-rs/commit/d3e60d19adbde8c8e9d3199c7c51e51ee5a20bf6
	while self.locked.load(Ordering::Relaxed) {
	// Tell the CPU that we're inside a busy-wait loop
	spin_loop_hint();
	}
	}
	}
	fn try_lock(&self) -> bool {
	self.locked
	.compare_exchange(false, true, Ordering::Acquire, Ordering::Relaxed)
	.is_ok()
	}
	fn unlock(&self) {
	self.locked.store(false, Ordering::Release);
	}
	}

Readers-writer lock

Basically, reviving #6.

Would you be interested in a PR adding a simple spinning RwLock? If concerned with minimalism, it could be behind a feature flag.

Add an RwLock equivalent to this crate

It would be nice if there was a spinlock equivalent of an RwLock in this crate as well. In my crate, I have the need of a read-write lock in a #![no_std] context. I am willing to implement this pull request.

spin-rs is now maintained

Hello! I've taken over maintenance of spin-rs. It now has support for lock_api and Send/Sync guards among other things. I doubt it'll have much bearing on spinning_top, but I thought you'd be interested in this (I'm told that spin-rs's lack of maintenance was the primary driving force behind this crate's development).

const fn Spinlock::new

It would be cool to have a way to construct a Spinlock in a const context. This would be super useful for a bunch of stuff I'm currently doing.

Exponential backoff

Would you be interested in a PR adding an exponential-backoff feature as a performance enhancement when running with a larger number of processors?

Add `arc_lock` feature

After #18, it might make sense to provide type aliases for all new lock_api guards, under the arc_lock feature flag.