Files
socktop-webterm/tests/reaper_tests.rs
T
jasonwitty 1114482046
Build and Deploy to K3s / test (push) Successful in 2m6s
Build and Deploy to K3s / lint (push) Successful in 1m2s
Build and Deploy to K3s / build-and-push (push) Successful in 9m23s
Build and Deploy to K3s / deploy (push) Successful in 2m13s
Fix actix worker wedge on session teardown; grant CAP_KILL; bump to 0.3.12
Terminal::stopping did child.kill() (error ignored) and then a blocking
child.wait() on the actix worker thread. Since the 0.3.11 privilege drop,
sessions run as `demo` and the server had no CAP_KILL, so kill() failed
with EPERM and wait() blocked forever. With two workers, exactly every
other request to :8082 then timed out — 5 days of readiness/liveness
flapping on socktop.io and orphaned restricted-shell/socktop processes
piling up in the pods.

- Signal the session's whole process group (portable-pty setsid()s the
  child), not just the shell: a non-interactive bash defers signals while
  a foreground command runs, so the old SIGHUP/SIGKILL to the shell alone
  orphaned socktop anyway.
- Reap on a dedicated thread: SIGHUP, 2 s grace, SIGKILL, 10 s deadline,
  then a blocking wait on that thread only. Signal failures are logged.
- Drop the pty handles before handing off the child.
- kubernetes/03-deployment.yaml: add CAP_KILL (sessions still lose every
  cap via setpriv). CI never applies the manifest — it was patched live.
- tests/reaper_tests.rs pins the non-blocking return, process-group kill,
  and SIGHUP→SIGKILL escalation.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-08-29 17:53:53 -07:00

145 lines
4.8 KiB
Rust

// Copyright (c) 2024 Jason Witty <jasonpwitty+socktop@proton.me>.
// All rights reserved.
//
// Regression tests for session teardown (`webterm::reap_child`).
//
// Background: `Terminal::stopping` used to call `child.kill()` then a blocking
// `child.wait()` on the actix worker thread. When the signal was refused
// (sessions run as another uid, server without CAP_KILL) or only reached the
// shell (a non-interactive bash defers signals while a foreground command
// runs), the child survived and the worker hung forever — every other request
// to the server then timed out. These tests pin the fixed behaviour: the whole
// process group is signalled, SIGHUP escalates to SIGKILL, and the caller is
// never blocked.
use std::io::Read;
use std::time::{Duration, Instant};
use portable_pty::{native_pty_system, CommandBuilder, PtySize};
/// True while `pid` exists and is not a zombie (kill(pid, 0) also succeeds on
/// zombies, so read /proc directly).
fn alive(pid: i32) -> bool {
match std::fs::read_to_string(format!("/proc/{pid}/status")) {
Ok(s) => !s
.lines()
.any(|l| l.starts_with("State:") && l.contains("Z (zombie)")),
Err(_) => false,
}
}
fn wait_gone(pid: i32, timeout: Duration) -> bool {
let start = Instant::now();
while start.elapsed() < timeout {
if !alive(pid) {
return true;
}
std::thread::sleep(Duration::from_millis(50));
}
!alive(pid)
}
/// Spawn `script` under `sh -c` in a fresh pty. The script must print the pid
/// of its long-running grandchild as its first line; that pid is returned along
/// with the pty child handle.
fn spawn_in_pty(script: &str) -> (Box<dyn portable_pty::Child + Send>, i32) {
let pty = native_pty_system();
let pair = pty
.openpty(PtySize {
rows: 24,
cols: 80,
pixel_width: 0,
pixel_height: 0,
})
.expect("openpty");
let mut cmd = CommandBuilder::new("sh");
cmd.arg("-c");
cmd.arg(script);
let child = pair.slave.spawn_command(cmd).expect("spawn");
drop(pair.slave);
let mut reader = pair.master.try_clone_reader().expect("reader");
let mut buf = Vec::new();
let mut byte = [0u8; 1];
let deadline = Instant::now() + Duration::from_secs(5);
while Instant::now() < deadline {
match reader.read(&mut byte) {
Ok(1) if byte[0] == b'\n' => break,
Ok(1) => buf.push(byte[0]),
_ => break,
}
}
let grandchild: i32 = String::from_utf8_lossy(&buf)
.trim()
.parse()
.expect("grandchild pid line");
assert!(
alive(grandchild),
"grandchild {grandchild} should be running"
);
// Keep the master alive for the duration of the test (mirrors the
// server's reader thread holding a cloned fd), so the kernel does not
// hang up the pty for us and the reaper alone has to do the work.
std::mem::forget(pair.master);
std::mem::forget(reader);
(child, grandchild)
}
#[test]
fn reap_child_returns_immediately_and_kills_whole_process_group() {
// sh waits on a foreground sleep: signalling sh alone would be deferred.
let (child, sleeper) = spawn_in_pty("sleep 30 & echo $!; wait");
let shell = child.process_id().expect("pid") as i32;
let t0 = Instant::now();
let reaper = webterm::reap_child(child);
assert!(
t0.elapsed() < Duration::from_millis(500),
"reap_child must not block the caller (took {:?})",
t0.elapsed()
);
reaper.join().expect("reaper thread");
assert!(
wait_gone(shell, Duration::from_secs(2)),
"shell {shell} survived"
);
assert!(
wait_gone(sleeper, Duration::from_secs(2)),
"grandchild {sleeper} was orphaned instead of killed with its group"
);
}
#[test]
fn reap_child_escalates_to_sigkill_when_sighup_is_ignored() {
// `trap '' HUP` makes sh ignore SIGHUP and the ignored disposition is
// inherited across exec, so sleep ignores it too — exactly the shape of a
// TUI that swallows HUP.
let (child, sleeper) = spawn_in_pty("trap '' HUP; sleep 30 & echo $!; wait");
let shell = child.process_id().expect("pid") as i32;
let t0 = Instant::now();
webterm::reap_child(child).join().expect("reaper thread");
let took = t0.elapsed();
assert!(
took >= Duration::from_secs(2),
"should have waited out the SIGHUP grace period (took {took:?})"
);
assert!(
took < Duration::from_secs(8),
"SIGKILL escalation should finish well inside the deadline (took {took:?})"
);
assert!(
wait_gone(shell, Duration::from_secs(2)),
"shell {shell} survived SIGKILL"
);
assert!(
wait_gone(sleeper, Duration::from_secs(2)),
"grandchild {sleeper} survived SIGKILL"
);
}