Turn your code inside-out: programming (and debugging) bare metal with async Rust

Rust has gotten a lot of attention in the embedded space lately, but there's an under-appreciated language feature that can be uniquely enabling for writing drivers and resource-constrained code: async. The async transform takes code written in the usual way -- with loops, function calls, and the like -- and inverts it into an explicit state machine that can be stepped in response to events. This gives developers the benefits of both simple "straight-line" code, such as readability and structured control flow, combined with the benefits of writing explicit state machines by hand, such as lower resource usage compared to threads or nested interrupts for concurrency. But more importantly, it provides something that's hard to achieve in explicit state machines: automatic composition of state machines, with type and lifetime checking.

This talk will propose an embedded-specific way of viewing Rust's async feature, look at how to apply it to solve certain common firmware problems, and touch on the elephant in the room: debugger support. Examples will use the author's open-source async embedded OS, lilos, and a proof-of-concept debugger for embedded async Rust.

Rust has gotten a lot of attention in the embedded space lately, but there's an under-appreciated language feature that can be uniquely enabling for writing drivers and resource-constrained code: async. The async transform takes code written in the usual way -- with loops, function calls, and the like -- and inverts it into an explicit state machine that can be stepped in response to events. This gives developers the benefits of both simple "straight-line" code, such as readability and structured control flow, combined with the benefits of writing explicit state machines by hand, such as lower resource usage compared to threads or nested interrupts for concurrency. But more importantly, it provides something that's hard to achieve in explicit state machines: automatic composition of state machines, with type and lifetime checking.

This talk will propose an embedded-specific way of viewing Rust's async feature, look at how to apply it to solve certain common firmware problems, and touch on the elephant in the room: debugger support. Examples will use the author's open-source async embedded OS, lilos, and a proof-of-concept debugger for embedded async Rust.