running-process 4.6.2

Subprocess and PTY runtime for the running-process project
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
use std::io;
use std::process::{Command, Stdio};
use std::sync::{Arc, Mutex};
use std::thread;
use std::time::{Duration, Instant};

use crate::helpers::{kill_drain_deadline, poll_until};

trait UnixChild: Send {
    fn kill(&mut self) -> io::Result<()>;
    fn wait(&mut self) -> io::Result<std::process::ExitStatus>;
    fn try_wait(&mut self) -> io::Result<Option<std::process::ExitStatus>>;
}

impl UnixChild for std::process::Child {
    fn kill(&mut self) -> io::Result<()> {
        std::process::Child::kill(self)
    }

    fn wait(&mut self) -> io::Result<std::process::ExitStatus> {
        std::process::Child::wait(self)
    }

    fn try_wait(&mut self) -> io::Result<Option<std::process::ExitStatus>> {
        std::process::Child::try_wait(self)
    }
}

pub struct SpawnedInner {
    child: Arc<Mutex<Option<Box<dyn UnixChild>>>>,
    pgid: i32,
}

impl SpawnedInner {
    pub fn kill(&self) -> io::Result<()> {
        // Try the child first, then the process group, to make sure
        // any siblings spawned inside go down too.
        let mut guard = self.child.lock().expect("child mutex poisoned");
        if let Some(child) = guard.as_mut() {
            let _ = child.kill();
        }
        drop(guard);
        unsafe {
            libc::killpg(self.pgid, libc::SIGKILL);
        }
        Ok(())
    }

    pub fn wait(&self) -> io::Result<i32> {
        let mut guard = self.child.lock().expect("child mutex poisoned");
        let Some(child) = guard.as_mut() else {
            return Err(io::Error::other("child handle absent"));
        };
        let status = child.wait()?;
        Ok(super::unix_exit_code(status))
    }

    pub fn try_wait(&self) -> io::Result<Option<i32>> {
        let mut guard = self.child.lock().expect("child mutex poisoned");
        let Some(child) = guard.as_mut() else {
            return Ok(None);
        };
        Ok(child.try_wait()?.map(super::unix_exit_code))
    }

    pub fn shutdown(&mut self) {
        self.shutdown_with_deadline(kill_drain_deadline());
    }

    fn shutdown_with_deadline(&mut self, deadline: Instant) {
        let group_signaled = unsafe { libc::killpg(self.pgid, libc::SIGKILL) == 0 };
        let Some(mut child) = self.child.lock().expect("child mutex poisoned").take() else {
            return;
        };
        if !group_signaled {
            let _ = child.kill();
        }
        match poll_until(deadline, Duration::from_millis(10), || child.try_wait()) {
            Ok(Some(_)) => {}
            Ok(None) | Err(_) => spawn_background_reaper(child),
        }
    }
}

fn spawn_background_reaper(mut child: Box<dyn UnixChild>) {
    thread::spawn(move || {
        // Once ownership is off the caller's teardown path, a blocking wait is
        // the most reliable terminal policy: it reaps exactly once without a
        // retry loop, spinning, or retaining the shared child mutex.
        let _ = child.wait();
    });
}

fn slot_to_stdio(slot: &super::StdioSource<'_>) -> io::Result<Stdio> {
    match slot {
        super::StdioSource::Null => Ok(Stdio::null()),
        super::StdioSource::Parent => Ok(Stdio::inherit()),
        super::StdioSource::Fd(fd) => {
            let owned = fd.try_clone_to_owned()?;
            Ok(Stdio::from(owned))
        }
        super::StdioSource::Pipe => Ok(Stdio::piped()),
        super::StdioSource::_Phantom(_) => unreachable!(),
    }
}

pub fn spawn_daemon(
    command: &mut Command,
    policy: super::EnvironmentPolicy,
) -> io::Result<super::DaemonChild> {
    use std::os::unix::process::CommandExt;

    apply_environment_policy(command, policy);

    command
        .stdin(Stdio::null())
        .stdout(Stdio::null())
        .stderr(Stdio::null());

    unsafe {
        command.pre_exec(|| {
            if libc::setsid() == -1 {
                // Already a session leader — not fatal.
            }
            close_extra_fds();
            Ok(())
        });
    }

    let child = command.spawn()?;
    let pid = child.id();
    Ok(super::DaemonChild { pid, child })
}

pub fn spawn(
    command: &mut Command,
    stdio: super::SpawnStdio<'_>,
    policy: super::EnvironmentPolicy,
) -> io::Result<super::SpawnedChild> {
    use std::os::unix::process::CommandExt;

    apply_environment_policy(command, policy);
    command.stdin(slot_to_stdio(&stdio.stdin)?);
    command.stdout(slot_to_stdio(&stdio.stdout)?);
    command.stderr(slot_to_stdio(&stdio.stderr)?);

    unsafe {
        command.pre_exec(|| {
            if libc::setpgid(0, 0) == -1 {
                return Err(io::Error::last_os_error());
            }
            #[cfg(target_os = "linux")]
            {
                if libc::prctl(libc::PR_SET_PDEATHSIG, libc::SIGKILL) == -1 {
                    return Err(io::Error::last_os_error());
                }
                if libc::getppid() == 1 {
                    libc::_exit(1);
                }
            }
            close_extra_fds();
            Ok(())
        });
    }

    let mut child = command.spawn()?;
    let pid = child.id();
    let pgid = pid as i32;

    let stdin = child.stdin.take();
    let stdout = child.stdout.take();
    let stderr = child.stderr.take();

    let child: Arc<Mutex<Option<Box<dyn UnixChild>>>> = Arc::new(Mutex::new(Some(Box::new(child))));

    // Drain watcher: wait for exit, then sleep `drain_timeout`. We
    // don't proactively close anything on Unix — Rust's ChildStdin/etc.
    // own their fds; once the child exits and the kernel ref-counts
    // its copies to zero, parent reads will EOF naturally.
    if let Some(timeout) = stdio.drain_timeout {
        let child_clone = Arc::clone(&child);
        thread::spawn(move || {
            // Borrow child for try_wait.  We do a polling loop so
            // shutdown() taking the inner Child during Drop doesn't
            // wedge us.
            loop {
                {
                    let mut guard = child_clone.lock().expect("child mutex poisoned");
                    match guard.as_mut() {
                        Some(c) => match c.try_wait() {
                            Ok(Some(_)) => break,
                            Ok(None) => {}
                            Err(_) => break,
                        },
                        None => return,
                    }
                }
                // #199: intentional — try_wait poll on the contained
                // child, 50ms cadence inside a bounded outer drain
                // loop. waitpid(WNOHANG)-equivalent semantics.
                thread::sleep(std::time::Duration::from_millis(50));
            }
            // #199: intentional — post-mortem pipe drain. Children's
            // write-ends of the captured stdio pipes are still being
            // closed by the kernel after exit; this gives readers a
            // chance to see the final bytes before the watcher
            // releases its keep-alive.
            thread::sleep(timeout);
        });
    }

    Ok(super::SpawnedChild {
        stdin,
        stdout,
        stderr,
        pid,
        inner: SpawnedInner { child, pgid },
    })
}

fn apply_environment_policy(command: &mut Command, policy: super::EnvironmentPolicy) {
    match policy {
        super::EnvironmentPolicy::Clear => {
            // Preserve explicit Command::env overrides while changing the
            // base to empty. Command::env_clear() also clears the override
            // map, so snapshot and restore it.
            let explicit: Vec<_> = command
                .get_envs()
                .filter_map(|(key, value)| {
                    value.map(|value| (key.to_os_string(), value.to_os_string()))
                })
                .collect();
            command.env_clear();
            command.envs(explicit);
        }
        super::EnvironmentPolicy::Inherit
        | super::EnvironmentPolicy::UserBaseline
        | super::EnvironmentPolicy::Auto => {
            // There is no stable Unix API equivalent to Windows
            // CreateEnvironmentBlock. UserBaseline conservatively falls
            // back to inheritance on Unix.
        }
    }
}

/// Async-signal-safe fd sweep used in pre_exec. See sanitized.rs (now
/// merged here) for the rationale.
unsafe fn close_extra_fds() {
    #[cfg(target_os = "linux")]
    {
        #[cfg(any(
            target_arch = "x86_64",
            target_arch = "aarch64",
            target_arch = "x86",
            target_arch = "arm",
            target_arch = "riscv64",
            target_arch = "powerpc64",
        ))]
        {
            const SYS_CLOSE_RANGE: libc::c_long = 436;
            let rc = libc::syscall(SYS_CLOSE_RANGE, 3u32, libc::c_uint::MAX, 0u32);
            if rc == 0 {
                return;
            }
        }
    }

    let dir = libc::opendir(c"/dev/fd".as_ptr());
    if !dir.is_null() {
        let dir_fd = libc::dirfd(dir);
        loop {
            let ent = libc::readdir(dir);
            if ent.is_null() {
                break;
            }
            let name_ptr = (*ent).d_name.as_ptr();
            let mut fd: libc::c_int = 0;
            let mut p = name_ptr;
            let mut ok = false;
            while *p != 0 {
                let c = *p as u8;
                if !c.is_ascii_digit() {
                    ok = false;
                    break;
                }
                fd = fd * 10 + (c - b'0') as libc::c_int;
                p = p.add(1);
                ok = true;
            }
            if !ok {
                continue;
            }
            if fd > 2 && fd != dir_fd {
                libc::close(fd);
            }
        }
        libc::closedir(dir);
        return;
    }

    let max = libc::sysconf(libc::_SC_OPEN_MAX);
    let max = if max < 0 { 4096 } else { max as libc::c_int };
    for fd in 3..max {
        libc::close(fd);
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use std::os::unix::process::ExitStatusExt;
    use std::sync::atomic::{AtomicUsize, Ordering};
    use std::sync::{Condvar, mpsc};

    struct FakeChild {
        wait_gate: Arc<(Mutex<bool>, Condvar)>,
        waits: Arc<AtomicUsize>,
        kills: Arc<AtomicUsize>,
    }

    impl UnixChild for FakeChild {
        fn kill(&mut self) -> io::Result<()> {
            self.kills.fetch_add(1, Ordering::SeqCst);
            Ok(())
        }

        fn wait(&mut self) -> io::Result<std::process::ExitStatus> {
            self.waits.fetch_add(1, Ordering::SeqCst);
            let (lock, condvar) = &*self.wait_gate;
            let mut released = lock.lock().expect("wait gate mutex poisoned");
            while !*released {
                released = condvar.wait(released).expect("wait gate mutex poisoned");
            }
            Ok(std::process::ExitStatus::from_raw(0))
        }

        fn try_wait(&mut self) -> io::Result<Option<std::process::ExitStatus>> {
            self.waits.fetch_add(1, Ordering::SeqCst);
            let released = *self.wait_gate.0.lock().expect("wait gate mutex poisoned");
            Ok(released.then(|| std::process::ExitStatus::from_raw(0)))
        }
    }

    struct BlockedFixture {
        inner: SpawnedInner,
        child: Arc<Mutex<Option<Box<dyn UnixChild>>>>,
        wait_gate: Arc<(Mutex<bool>, Condvar)>,
        waits: Arc<AtomicUsize>,
        kills: Arc<AtomicUsize>,
    }

    fn blocked_inner() -> BlockedFixture {
        let wait_gate = Arc::new((Mutex::new(false), Condvar::new()));
        let waits = Arc::new(AtomicUsize::new(0));
        let kills = Arc::new(AtomicUsize::new(0));
        let child: Arc<Mutex<Option<Box<dyn UnixChild>>>> =
            Arc::new(Mutex::new(Some(Box::new(FakeChild {
                wait_gate: Arc::clone(&wait_gate),
                waits: Arc::clone(&waits),
                kills: Arc::clone(&kills),
            }))));
        BlockedFixture {
            inner: SpawnedInner {
                child: Arc::clone(&child),
                pgid: i32::MAX,
            },
            child,
            wait_gate,
            waits,
            kills,
        }
    }

    fn release_wait(wait_gate: &Arc<(Mutex<bool>, Condvar)>) {
        let (lock, condvar) = &**wait_gate;
        *lock.lock().expect("wait gate mutex poisoned") = true;
        condvar.notify_all();
    }

    struct ShutdownOnDrop {
        inner: Option<SpawnedInner>,
        deadline: Instant,
    }

    impl Drop for ShutdownOnDrop {
        fn drop(&mut self) {
            self.inner
                .as_mut()
                .expect("test wrapper missing inner")
                .shutdown_with_deadline(self.deadline);
        }
    }

    #[test]
    fn drop_is_bounded_when_child_wait_does_not_complete() {
        // Regression for #619: SpawnedChild::drop delegates directly to
        // SpawnedInner::shutdown, modeled by this wrapper around the fake child.
        let BlockedFixture {
            inner, wait_gate, ..
        } = blocked_inner();
        let (tx, rx) = mpsc::channel();
        let started = Instant::now();
        let worker = thread::spawn(move || {
            drop(ShutdownOnDrop {
                inner: Some(inner),
                deadline: Instant::now() + Duration::from_millis(50),
            });
            let _ = tx.send(started.elapsed());
        });

        let timely = rx.recv_timeout(Duration::from_millis(100));
        release_wait(&wait_gate);
        let elapsed = timely
            .or_else(|_| rx.recv_timeout(Duration::from_secs(1)))
            .expect("shutdown did not unblock after releasing fake child");
        worker.join().expect("shutdown worker panicked");
        assert!(
            elapsed < Duration::from_millis(100),
            "Drop blocked for {elapsed:?} in child.wait()"
        );
    }

    #[test]
    fn shutdown_does_not_hold_child_mutex_while_reaping() {
        let BlockedFixture {
            mut inner,
            child,
            wait_gate,
            waits,
            ..
        } = blocked_inner();
        let worker = thread::spawn(move || {
            inner.shutdown_with_deadline(Instant::now() + Duration::from_millis(50));
        });
        let deadline = Instant::now() + Duration::from_secs(1);
        while waits.load(Ordering::SeqCst) == 0 && Instant::now() < deadline {
            thread::yield_now();
        }
        assert_eq!(waits.load(Ordering::SeqCst), 1, "fake wait never started");

        let child_mutex_available = child.try_lock().is_ok();
        release_wait(&wait_gate);
        worker.join().expect("shutdown worker panicked");
        assert!(
            child_mutex_available,
            "shutdown held the child mutex across reaping"
        );
    }

    struct ReadyChild {
        polls: Arc<AtomicUsize>,
        waits: Arc<AtomicUsize>,
    }

    impl UnixChild for ReadyChild {
        fn kill(&mut self) -> io::Result<()> {
            Ok(())
        }

        fn wait(&mut self) -> io::Result<std::process::ExitStatus> {
            self.waits.fetch_add(1, Ordering::SeqCst);
            Ok(std::process::ExitStatus::from_raw(0))
        }

        fn try_wait(&mut self) -> io::Result<Option<std::process::ExitStatus>> {
            self.polls.fetch_add(1, Ordering::SeqCst);
            Ok(Some(std::process::ExitStatus::from_raw(0)))
        }
    }

    #[test]
    fn shutdown_reaps_ready_child_exactly_once() {
        let polls = Arc::new(AtomicUsize::new(0));
        let waits = Arc::new(AtomicUsize::new(0));
        let child: Arc<Mutex<Option<Box<dyn UnixChild>>>> =
            Arc::new(Mutex::new(Some(Box::new(ReadyChild {
                polls: Arc::clone(&polls),
                waits: Arc::clone(&waits),
            }))));
        let mut inner = SpawnedInner {
            child,
            pgid: i32::MAX,
        };

        inner.shutdown_with_deadline(Instant::now() + Duration::from_secs(1));

        assert_eq!(polls.load(Ordering::SeqCst), 1);
        assert_eq!(waits.load(Ordering::SeqCst), 0);
    }

    #[test]
    fn shutdown_falls_back_to_direct_kill_when_group_signal_fails() {
        let BlockedFixture {
            mut inner,
            wait_gate,
            kills,
            ..
        } = blocked_inner();
        release_wait(&wait_gate);

        inner.shutdown_with_deadline(Instant::now() + Duration::from_secs(1));

        assert_eq!(kills.load(Ordering::SeqCst), 1);
    }
}