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
use crate::{RtckError, RtckResult};
use log::*;
use nix::unistd::Pid;
use std::path::PathBuf;
/// Terminate the hypervisor by sending SIGTERM
/// Note: this command will kill firecracker itself
#[cfg(target_os = "linux")]
fn terminate(pid: Pid) -> RtckResult<()> {
use nix::sys::signal::{kill, Signal};
// the hypervisor occupies the pid by opening fd to it (procfs).
// so kill -9 to this pid is safe.
kill(pid, Signal::SIGTERM).map_err(|e| {
let msg = format!("Fail to terminate pid {pid}: {e}");
error!("{msg}");
RtckError::Hypervisor(msg)
})
}
/// Terminate the hypervisor by sending SIGKILL
/// Note: this command will kill firecracker itself.
#[cfg(target_os = "linux")]
fn kill(pid: Pid) -> RtckResult<()> {
// the hypervisor occupies the pid by opening fd to it (procfs).
// so kill -9 to this pid is safe.
use nix::sys::signal::{kill, Signal};
kill(pid, Signal::SIGKILL)
// kill -9 should not trigger this error since SIGKILL is not blockable
.map_err(|e| {
let msg = format!("Fail to kill pid {pid}: {e}");
error!("{msg}");
RtckError::Hypervisor(msg)
})
}
pub enum Rollback {
Jailing {
clear: bool,
instance_dir: PathBuf,
},
StopProcess {
pid: u32,
},
RemoveSocket {
path: PathBuf,
},
#[allow(unused)]
RemoveFile {
path: PathBuf,
},
Umount {
mount_point: PathBuf,
},
Chown {
path: PathBuf,
original_uid: u32,
original_gid: u32,
},
}
impl Rollback {
fn rollback(self) {
match self {
Rollback::Jailing {
clear,
instance_dir,
} => {
// remove the instance directory
if clear {
info!("Removing instance dir {:?}", instance_dir);
let _ = std::fs::remove_dir_all(instance_dir);
}
}
Rollback::StopProcess { pid } => {
info!("Terminating process {pid}");
use nix::{
sys::wait::{waitpid, WaitPidFlag, WaitStatus},
unistd::Pid,
};
let pid = Pid::from_raw(pid as i32);
// first check the status of process.
// if user has give up this microVM voluntarily
loop {
match waitpid(pid, Some(WaitPidFlag::WNOHANG)) {
Ok(WaitStatus::Exited(_, exit_status)) => {
warn!("Process {pid} has exited {exit_status}");
return;
}
Ok(WaitStatus::Signaled(_, signal, core_dumped)) => {
warn!(
"Process {pid} has been signaled {signal}, core_dumped: {core_dumped}"
);
return;
}
Ok(WaitStatus::StillAlive)
| Ok(WaitStatus::PtraceEvent(_, _, _))
| Ok(WaitStatus::PtraceSyscall(_))
| Ok(WaitStatus::Continued(_)) => break, // terminate it!
Err(nix::errno::Errno::ECHILD) => {
error!("No such process {}", pid);
return;
}
Err(nix::errno::Errno::EINTR) => {
warn!("Checking status interrupted, trying again...");
continue;
}
Err(nix::errno::Errno::EINVAL) => {
error!(
"Checking status invalid arguments, send a terminate signal anyway"
);
break;
}
_ => {
error!("Fatal error! unreachable");
return;
}
}
}
if self::terminate(pid).is_err() {
let _ = self::kill(pid);
}
loop {
match waitpid(pid, None) {
Ok(WaitStatus::Exited(_, status)) => {
info!("Process {} exited with status {}", pid, status);
break;
}
Ok(WaitStatus::Signaled(_, signal, _)) => {
warn!("Process {} was killed by signal {}", pid, signal);
break;
}
Ok(_) => {
// other WaitStatus，e.g. Stopped、Continued
if terminate(pid).is_err() {
let _ = kill(pid);
}
std::thread::sleep(std::time::Duration::from_millis(100));
}
Err(nix::errno::Errno::ECHILD) => {
error!("No such process {}", pid);
break;
}
Err(err) => {
error!("Error occurred: {}", err);
break;
}
}
}
}
Rollback::RemoveSocket { path } => {
// removal failure is not a big deal so ignore possible error
info!("Removing socket {:?}", path);
let _ = std::fs::remove_file(path);
}
Rollback::RemoveFile { path } => {
// removal failure is not a big deal so ignore possible error
info!("Removing file {:?}", path);
let _ = std::fs::remove_file(path);
}
Rollback::Umount { mount_point } => {
info!("Umount {mount_point:?}");
use nix::mount::{umount2, MntFlags};
let _ = umount2(&mount_point, MntFlags::MNT_FORCE).map_err(|e| {
RtckError::Hypervisor(format!("Fail to umount the kernel dir, errno = {}", e))
});
}
Rollback::Chown {
path,
original_uid,
original_gid,
} => {
info!(
"Change onwership of {:?} back to ({}:{})",
path, original_uid, original_gid
);
use nix::unistd::{Gid, Uid};
let _ = nix::unistd::chown(
&path,
Some(Uid::from_raw(original_uid)),
Some(Gid::from_raw(original_gid)),
);
}
}
}
}
/// Stack that manages rollbacks.
pub struct RollbackStack {
pub stack: Vec<Rollback>,
}
impl RollbackStack {
pub fn new() -> Self {
RollbackStack { stack: Vec::new() }
}
pub fn push(&mut self, rollback: Rollback) {
self.stack.push(rollback);
}
pub fn insert_1(&mut self, rollback: Rollback) {
self.stack.insert(self.stack.len() - 1, rollback);
}
pub fn rollback_all(&mut self) {
while let Some(op) = self.stack.pop() {
op.rollback();
}
}
}
impl Drop for RollbackStack {
fn drop(&mut self) {
self.rollback_all();
}
}