Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

scx_loader: Avoid race condition with start_scheduler #1225

Merged
merged 1 commit into from
Jan 22, 2025
Merged

scx_loader: Avoid race condition with start_scheduler #1225

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions Cargo.lock
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

1 change: 1 addition & 0 deletions rust/scx_loader/Cargo.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@ nix = { features = ["process", "signal"], default-features = false, version = "0
serde = { version = "1.0", features = ["derive"] }
sysinfo = "0.31.4"
tokio = { version = "1.39", features = ["macros", "sync", "rt-multi-thread", "process"] }
tokio-util = "0.7"
toml = "0.8.19"
zbus = { version = "5", features = ["tokio"], default-features = false }
zvariant = "5.1"
Expand Down
140 changes: 67 additions & 73 deletions rust/scx_loader/src/main.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,9 +10,8 @@ mod logger;
use scx_loader::dbus::LoaderClientProxy;
use scx_loader::*;

use std::process::ExitStatus;
use std::process::Stdio;
use std::sync::atomic::AtomicU32;
use std::sync::atomic::Ordering;
use std::sync::Arc;

use anyhow::Context;
Expand Down Expand Up @@ -384,44 +383,45 @@ async fn worker_loop(
}

async fn handle_child_process(mut rx: tokio::sync::mpsc::Receiver<RunnerMessage>) -> Result<()> {
let child_id = Arc::new(AtomicU32::new(0));
let mut task: Option<tokio::task::JoinHandle<Result<Option<ExitStatus>>>> = None;
let mut cancel_token = Arc::new(tokio_util::sync::CancellationToken::new());

while let Some(message) = rx.recv().await {
match message {
RunnerMessage::Switch((scx_sched, sched_args)) => {
// stop the sched if its running
if let Err(stop_err) = stop_scheduler(child_id.clone()).await {
log::error!("Failed to stop previous scheduler: {stop_err}");
}
stop_scheduler(&mut task, &mut cancel_token).await;

// overwise start scheduler
if let Err(sched_err) =
start_scheduler(scx_sched, sched_args, child_id.clone()).await
{
log::error!("Scheduler exited with err: {sched_err}");
} else {
log::debug!("Scheduler exited");
match start_scheduler(scx_sched, sched_args, cancel_token.clone()).await {
Ok(handle) => {
task = Some(handle);
log::debug!("Scheduler started");
}
Err(err) => {
log::error!("Failed to start scheduler: {err}");
}
}
}
RunnerMessage::Start((scx_sched, sched_args)) => {
// check if sched is running or not
if child_id.load(Ordering::Relaxed) != 0 {
if task.is_some() {
log::error!("Scheduler wasn't finished yet. Stop already running scheduler!");
continue;
}
// overwise start scheduler
if let Err(sched_err) =
start_scheduler(scx_sched, sched_args, child_id.clone()).await
{
log::error!("Scheduler exited with err: {sched_err}");
} else {
log::debug!("Scheduler exited");
match start_scheduler(scx_sched, sched_args, cancel_token.clone()).await {
Ok(handle) => {
task = Some(handle);
log::debug!("Scheduler started");
}
Err(err) => {
log::error!("Failed to start scheduler: {err}");
}
}
}
RunnerMessage::Stop => {
if let Err(stop_err) = stop_scheduler(child_id.clone()).await {
log::error!("Failed to stop scheduler: {stop_err}");
}
stop_scheduler(&mut task, &mut cancel_token).await;
}
}
}
Expand All @@ -433,35 +433,51 @@ async fn handle_child_process(mut rx: tokio::sync::mpsc::Receiver<RunnerMessage>
async fn start_scheduler(
scx_crate: SupportedSched,
args: Vec<String>,
child_id: Arc<AtomicU32>,
) -> Result<()> {
cancel_token: Arc<tokio_util::sync::CancellationToken>,
) -> Result<tokio::task::JoinHandle<Result<Option<ExitStatus>>>> {
// Ensure the child process exit is handled correctly in the runtime
tokio::spawn(async move {
let handle = tokio::spawn(async move {
let mut retries = 0u32;
let max_retries = 5u32;

let mut last_status: Option<ExitStatus> = None;

while retries < max_retries {
let child = spawn_scheduler(scx_crate.clone(), args.clone(), child_id.clone()).await;
let child = spawn_scheduler(scx_crate.clone(), args.clone()).await;

let mut failed = false;
if let Ok(mut child) = child {
let status = child
.wait()
.await
.expect("child process encountered an error");

if !status.success() {
failed = true;
}
log::debug!("Child process exited with status: {status:?}");
tokio::select! {
status = child.wait() => {
let status = status.expect("child process encountered an error");
last_status = Some(status);
if !status.success() {
failed = true;
}
log::debug!("Child process exited with status: {status:?}");
}

_ = cancel_token.cancelled() => {
log::debug!("Received cancellation signal");
// Send SIGINT
if let Some(child_id) = child.id() {
nix::sys::signal::kill(
nix::unistd::Pid::from_raw(child_id as i32),
nix::sys::signal::SIGINT,
).context("Failed to send termination signal to the child")?;
}
let status = child.wait().await.expect("child process encountered an error");
last_status = Some(status);
break;
}
};
} else {
log::debug!("Failed to spawn child process");
failed = true;
}

// retrying if failed, otherwise exit
if !failed {
child_id.store(0, Ordering::Relaxed);
break;
}

Expand All @@ -472,19 +488,16 @@ async fn start_scheduler(
max_retries,
);
}
child_id.store(0, Ordering::Relaxed);

Ok(last_status)
});

Ok(())
Ok(handle)
}

/// Starts the scheduler as a child process and returns child object to manage lifecycle by the
/// caller.
async fn spawn_scheduler(
scx_crate: SupportedSched,
args: Vec<String>,
child_id: Arc<AtomicU32>,
) -> Result<Child> {
async fn spawn_scheduler(scx_crate: SupportedSched, args: Vec<String>) -> Result<Child> {
let sched_bin_name: &str = scx_crate.into();
log::info!("starting {sched_bin_name} command");

Expand All @@ -500,38 +513,19 @@ async fn spawn_scheduler(
// spawn process
let child = cmd.spawn().expect("failed to spawn command");

// NOTE: unsafe because the child might not exist, when we will try to stop it
// set child id
child_id.store(
child
.id()
.ok_or(anyhow::anyhow!("Failed to get child id"))?,
Ordering::Relaxed,
);

Ok(child)
}

async fn stop_scheduler(child_id: Arc<AtomicU32>) -> Result<()> {
// if child_proc is 0, then we assume the child process is terminated
let child_proc = child_id.load(Ordering::Relaxed);
if child_proc == 0 {
return Ok(());
async fn stop_scheduler(
task: &mut Option<tokio::task::JoinHandle<Result<Option<ExitStatus>>>>,
cancel_token: &mut Arc<tokio_util::sync::CancellationToken>,
) {
if let Some(task) = task.take() {
log::debug!("Stopping already running scheduler..");
cancel_token.cancel();
let status = task.await;
log::debug!("Scheduler was stopped with status: {:?}", status);
// Create a new cancellation token
*cancel_token = Arc::new(tokio_util::sync::CancellationToken::new());
}

// send SIGINT signal to child
log::debug!("Stopping already running scheduler..");
nix::sys::signal::kill(
nix::unistd::Pid::from_raw(child_proc as i32),
nix::sys::signal::SIGINT,
)
.context("Failed to send termination signal to the child")?;

// Wait for the child process to exit and child_id to become 0
while child_id.load(Ordering::Relaxed) != 0 {
tokio::time::sleep(Duration::from_secs(1)).await;
}
log::debug!("Scheduler was stopped");

Ok(())
}