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
//! ShimHandler — concrete VmHandler for a libkrun shim subprocess.
pub use a3s_box_core::vmm::{VmHandler, VmMetrics, DEFAULT_SHUTDOWN_TIMEOUT_MS};
use a3s_box_core::error::Result;
use std::process::Child;
use std::sync::Mutex;
use sysinfo::{Pid, System};
/// Handler for a running VM subprocess (shim process).
///
/// Provides lifecycle operations (stop, metrics, status) for a VM identified by PID.
pub struct ShimHandler {
pid: u32,
box_id: String,
/// Child process handle for proper lifecycle management.
/// When we spawn the process, we keep the Child to properly wait() on stop.
/// When we attach to an existing process, this is None.
process: Option<Child>,
/// Shared System instance for CPU metrics calculation across calls.
/// CPU usage requires comparing snapshots over time, so we must reuse the same System.
metrics_sys: Mutex<System>,
/// Exit code of the shim process, set when stop() collects the exit status.
exit_code: Option<i32>,
}
impl ShimHandler {
/// Create a handler for a spawned VM with process ownership.
///
/// This constructor takes ownership of the Child process handle for proper
/// lifecycle management (clean shutdown with wait()).
pub fn from_child(process: Child, box_id: String) -> Self {
let pid = process.id();
Self {
pid,
box_id,
process: Some(process),
metrics_sys: Mutex::new(System::new()),
exit_code: None,
}
}
/// Create a handler for an existing VM (attach mode).
///
/// Used when reconnecting to a running box. We don't have a Child handle,
/// so we manage the process by PID only.
pub fn from_pid(pid: u32, box_id: String) -> Self {
Self {
pid,
box_id,
process: None,
metrics_sys: Mutex::new(System::new()),
exit_code: None,
}
}
/// Get the box ID.
pub fn box_id(&self) -> &str {
&self.box_id
}
}
impl VmHandler for ShimHandler {
fn pid(&self) -> u32 {
self.pid
}
#[cfg(unix)]
fn stop(&mut self, signal: i32, timeout_ms: u64) -> Result<()> {
// Graceful shutdown: send configured signal first, wait, then SIGKILL if needed.
// This gives libkrun time to flush its virtio-blk buffers to disk.
if let Some(mut process) = self.process.take() {
// Step 1: Send configured stop signal for graceful shutdown
let pid = process.id();
tracing::debug!(pid, box_id = %self.box_id, signal, "Sending stop signal to VM process");
unsafe {
libc::kill(pid as i32, signal);
}
// Step 2: Wait with timeout for process to exit
let start = std::time::Instant::now();
loop {
match process.try_wait() {
Ok(Some(status)) => {
tracing::debug!(pid, ?status, "VM process exited gracefully");
self.exit_code = status.code();
return Ok(());
}
Ok(None) => {
// Still running, check timeout
if start.elapsed().as_millis() > timeout_ms as u128 {
tracing::warn!(
pid,
timeout_ms,
"VM process did not exit gracefully, sending SIGKILL"
);
let _ = process.kill();
if let Ok(status) = process.wait() {
self.exit_code = status.code();
}
return Ok(());
}
// Brief sleep before checking again
std::thread::sleep(std::time::Duration::from_millis(50));
}
Err(e) => {
tracing::warn!(pid, error = %e, "Error checking process status, forcing kill");
let _ = process.kill();
let _ = process.wait();
return Ok(());
}
}
}
} else {
// Attached mode: use configured signal then SIGKILL with polling
tracing::debug!(pid = self.pid, box_id = %self.box_id, signal, "Sending stop signal to attached VM process");
unsafe {
libc::kill(self.pid as i32, signal);
}
// Poll for exit with timeout
let start = std::time::Instant::now();
loop {
let mut status: i32 = 0;
let result = unsafe { libc::waitpid(self.pid as i32, &mut status, libc::WNOHANG) };
if result > 0 {
tracing::debug!(pid = self.pid, "VM process exited gracefully");
return Ok(());
}
if result < 0 {
// Error - process may not be our child (common in attached mode)
let exists = unsafe { libc::kill(self.pid as i32, 0) } == 0;
if !exists {
return Ok(()); // Already dead
}
}
if start.elapsed().as_millis() > timeout_ms as u128 {
tracing::warn!(
pid = self.pid,
timeout_ms,
"VM process did not exit gracefully, sending SIGKILL"
);
unsafe {
libc::kill(self.pid as i32, libc::SIGKILL);
}
return Ok(());
}
std::thread::sleep(std::time::Duration::from_millis(50));
}
}
}
#[cfg(windows)]
fn stop(&mut self, _signal: i32, timeout_ms: u64) -> Result<()> {
// Windows version: use Child::kill() or terminate process by PID
if let Some(mut process) = self.process.take() {
tracing::debug!(pid = self.pid, box_id = %self.box_id, "Terminating VM process");
// Try graceful wait first
let start = std::time::Instant::now();
loop {
match process.try_wait() {
Ok(Some(status)) => {
tracing::debug!(pid = self.pid, ?status, "VM process exited");
self.exit_code = status.code();
return Ok(());
}
Ok(None) => {
if start.elapsed().as_millis() > timeout_ms as u128 {
tracing::warn!(pid = self.pid, "VM process did not exit, forcing kill");
let _ = process.kill();
if let Ok(status) = process.wait() {
self.exit_code = status.code();
}
return Ok(());
}
std::thread::sleep(std::time::Duration::from_millis(50));
}
Err(e) => {
tracing::warn!(pid = self.pid, error = %e, "Error checking process, forcing kill");
let _ = process.kill();
let _ = process.wait();
return Ok(());
}
}
}
} else {
// Attached mode: just check if process exists
tracing::debug!(pid = self.pid, "Checking attached VM process");
let start = std::time::Instant::now();
while start.elapsed().as_millis() <= timeout_ms as u128 {
if !self.is_running() {
return Ok(());
}
std::thread::sleep(std::time::Duration::from_millis(50));
}
tracing::warn!(pid = self.pid, "Attached VM process still running after timeout");
Ok(())
}
}
fn metrics(&self) -> VmMetrics {
let pid = Pid::from_u32(self.pid);
// Use the shared System instance for stateful CPU tracking
let mut sys = match self.metrics_sys.lock() {
Ok(guard) => guard,
Err(e) => {
tracing::warn!(error = %e, "metrics_sys lock poisoned");
return VmMetrics::default();
}
};
// Refresh process info - this updates the internal state for delta calculation
sys.refresh_process(pid);
// Try to get process information
if let Some(proc_info) = sys.process(pid) {
return VmMetrics {
cpu_percent: Some(proc_info.cpu_usage()),
memory_bytes: Some(proc_info.memory()),
};
}
// Process not found or not running - return empty metrics
VmMetrics::default()
}
#[cfg(unix)]
fn is_running(&self) -> bool {
// Check if process exists by sending signal 0
unsafe { libc::kill(self.pid as i32, 0) == 0 }
}
#[cfg(windows)]
fn is_running(&self) -> bool {
// Use sysinfo to check if process exists
let mut sys = System::new();
sys.refresh_process(Pid::from_u32(self.pid));
sys.process(Pid::from_u32(self.pid)).is_some()
}
fn exit_code(&self) -> Option<i32> {
self.exit_code
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_vm_metrics_default() {
let m = VmMetrics::default();
assert!(m.cpu_percent.is_none());
assert!(m.memory_bytes.is_none());
}
#[test]
fn test_vm_metrics_clone() {
let m = VmMetrics {
cpu_percent: Some(50.0),
memory_bytes: Some(1024 * 1024),
};
let cloned = m.clone();
assert_eq!(cloned.cpu_percent, Some(50.0));
assert_eq!(cloned.memory_bytes, Some(1024 * 1024));
}
#[test]
fn test_shim_handler_from_pid() {
let handler = ShimHandler::from_pid(12345, "box-abc".to_string());
assert_eq!(handler.pid(), 12345);
assert_eq!(handler.box_id(), "box-abc");
assert_eq!(handler.exit_code(), None);
}
#[test]
fn test_shim_handler_is_running_nonexistent_pid() {
// PID 999999999 should not exist
let handler = ShimHandler::from_pid(999_999_999, "test".to_string());
assert!(!handler.is_running());
}
#[test]
fn test_shim_handler_metrics_nonexistent_pid() {
let handler = ShimHandler::from_pid(999_999_999, "test".to_string());
let m = handler.metrics();
// Non-existent process should return default metrics
assert!(m.cpu_percent.is_none() || m.cpu_percent == Some(0.0));
}
#[test]
fn test_shim_handler_is_running_current_process() {
// Current process PID should be running
let pid = std::process::id();
let handler = ShimHandler::from_pid(pid, "self".to_string());
assert!(handler.is_running());
}
#[test]
fn test_default_shutdown_timeout() {
assert_eq!(DEFAULT_SHUTDOWN_TIMEOUT_MS, 10_000);
}
#[test]
fn test_vm_metrics_debug() {
let m = VmMetrics {
cpu_percent: Some(25.5),
memory_bytes: Some(512),
};
let debug = format!("{:?}", m);
assert!(debug.contains("25.5"));
assert!(debug.contains("512"));
}
}