scx-upstream/rust/scx_rustland_core
Andrea Righi 081d4bdb86 scx_rustland_core: add debugging to dispatch_direct_cpu()
Report dispatch_direct_cpu() events in the trace, like any other
dispatch-related event.

Tested-by: SoulHarsh007 <harsh.peshwani@outlook.com>
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
2024-06-25 14:02:39 +02:00
..
assets scx_rustland_core: add debugging to dispatch_direct_cpu() 2024-06-25 14:02:39 +02:00
src scx_rustland_core: include buddy-alloc and refactor allocator code 2024-06-19 14:44:04 +02:00
.gitignore rust: introduce scx_rustland_core crate 2024-02-28 17:49:44 +01:00
bindings.h rust: introduce scx_rustland_core crate 2024-02-28 17:49:44 +01:00
bpf_h rust: introduce scx_rustland_core crate 2024-02-28 17:49:44 +01:00
build.rs scx_utils: introduce Builder() 2024-02-28 17:49:44 +01:00
Cargo.toml scx_rustland_core: include buddy-alloc and refactor allocator code 2024-06-19 14:44:04 +02:00
LICENSE rust: introduce scx_rustland_core crate 2024-02-28 17:49:44 +01:00
meson.build Fetch and build bpftool by default 2024-03-11 10:00:01 -07:00
README.md scx_rustland_core: introduce per-task time slice 2024-03-03 15:06:56 +01:00

Framework to implement sched_ext schedulers running in user-space

sched_ext is a Linux kernel feature which enables implementing kernel thread schedulers in BPF and dynamically loading them.

This crate provides a generic layer that can be used to implement sched-ext schedulers that run in user-space.

It provides a generic BPF abstraction that is completely agnostic of the particular scheduling policy implemented in user-space.

Developers can use such abstraction to implement schedulers using pure Rust code, without having to deal with any internal kernel / BPF internal details.

API

The main BPF interface is provided by the BpfScheduler struct. When this object is initialized it will take care of registering and initializing the BPF component.

The scheduler then can use BpfScheduler instance to receive tasks (in the form of QueuedTask objects) and dispatch tasks (in the form of DispatchedTask objects), using respectively the methods dequeue_task() and dispatch_task().

Example usage (FIFO scheduler):

struct Scheduler<'a> {
    bpf: BpfScheduler<'a>,
}

impl<'a> Scheduler<'a> {
    fn init() -> Result<Self> {
        let topo = Topology::new().expect("Failed to build host topology");
        let bpf = BpfScheduler::init(5000, topo.nr_cpus() as i32, false, false, false)?;
        Ok(Self { bpf })
    }

    fn schedule(&mut self) {
        match self.bpf.dequeue_task() {
            Ok(Some(task)) => {
                // task.cpu < 0 is used to to notify an exiting task, in this
                // case we can simply ignore it.
                if task.cpu >= 0 {
                    let _ = self.bpf.dispatch_task(&DispatchedTask {
                        pid: task.pid,
                        cpu: task.cpu,
                        cpumask_cnt: task.cpumask_cnt,
                        slice_ns: 0,
                    });
                }
            }
            Ok(None) => {
                // Notify the BPF component that all tasks have been dispatched.
                self.bpf.update_tasks(Some(0), Some(0))?

                break;
            }
            Err(_) => {
                break;
            }
        }
    }

Moreover, a CPU ownership map (that keeps track of which PID runs on which CPU) can be accessed using the method get_cpu_pid(). This also allows to keep track of the idle and busy CPUs, with the corresponding PIDs associated to them.

BPF counters and statistics can be accessed using the methods nr_*_mut(), in particular nr_queued_mut() and nr_scheduled_mut() can be updated to notify the BPF component if the user-space scheduler has still some pending work to do or not.

Lastly, the methods exited() and shutdown_and_report() can be used respectively to test whether the BPF component exited, and to shutdown and report the exit message.