feat(init): watchdog: pinger thread PINGs servers (framework answers, no server code), main kills after 5s unanswered -> crash path; not pinged: terminal, pm (handlers wait for user); give-up drains queued ping; spawntest hang check

crates/apps/spawntest/src/main.rs

@@ -200,6 +200,22 @@ fn restarted_after_crashes() -> Result<(), &'static str> {
     Ok(())
 }
 
+/// A server stuck in a handler is noticed by the watchdog, killed (which
+/// releases its caller) and restarted.
+fn restarted_after_hanging() -> Result<(), &'static str> {
+    let asked = syscall::get_time().raw();
+    // Comes back, with an error, once the hung server is killed.
+    let _ = syscall::request(FLAKY, [flaky::op::HANG, 0, 0, 0]);
+    let hung_ms = (syscall::get_time().raw() - asked) / 10_000;
+    if hung_ms < 4_500 {
+        return Err("the hung server was stopped before the watchdog's deadline");
+    }
+    if flaky()? != 1 {
+        return Err("the hung server was not restarted");
+    }
+    Ok(())
+}
+
 fn check() -> Result<(), &'static str> {
     let mut fs = Storage::attach(STORAGE, STORAGE_BUF_VA).map_err(|_| "cannot attach to storage")?;
     write_a_file(&mut fs)?;
@@ -232,6 +248,9 @@ fn check() -> Result<(), &'static str> {
 
     restarted_after_crashes()?;
     libkos::println!("spawntest: a crashed server is restarted, a crash loop backs off: ok");
+
+    restarted_after_hanging()?;
+    libkos::println!("spawntest: a hung server is noticed, stopped and restarted: ok");
     Ok(())
 }