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
use std::os::fd::{AsFd, AsRawFd, OwnedFd};
use std::path::Path;
use std::sync::Arc;
use tokio::io::unix::AsyncFd;
use crate::fid_parser::read_fid_events_cached;
use super::Monitor;
use super::channel::EventSender;
// ---- Reader supervision ----
/// Per-reader-task restart tracking for exponential backoff.
/// Restarts are capped at MAX_RESTARTS within BACKOFF_WINDOW.
pub(crate) struct ReaderState {
pub(crate) restart_count: u32,
pub(crate) last_restart: std::time::Instant,
/// Set when restart_reader gives up (backoff exhausted within window).
/// Reset when spawn_fd_reader attempts a new spawn (even if it later fails).
/// Used by health() for reliable alive/dead reporting.
pub(crate) gave_up: bool,
}
pub(crate) const MAX_RESTARTS: u32 = 3;
pub(crate) const BACKOFF_WINDOW: std::time::Duration = std::time::Duration::from_secs(60);
impl Monitor {
/// Duplicate a file descriptor, returning an owned fd.
/// The returned `OwnedFd` has independent lifetime from the source
/// and will be closed on drop.
pub(crate) fn dup_fd(fd: &impl AsFd) -> std::io::Result<OwnedFd> {
nix::unistd::dup(fd).map_err(std::io::Error::other)
}
/// Open a directory and return an owned fd.
/// The returned `OwnedFd` has the directory open and will be
/// closed on drop.
pub(crate) fn open_dir(path: &Path) -> std::io::Result<OwnedFd> {
nix::fcntl::open(
path,
nix::fcntl::OFlag::O_DIRECTORY,
nix::sys::stat::Mode::empty(),
)
.map_err(std::io::Error::other)
}
/// Spawn a tokio reader task for `group_idx` in `fs_groups`.
/// Both the fanotify fd and mount fd are duplicated so the reader task
/// owns independent copies, avoiding double-close with Monitor's OwnedFd.
pub(crate) fn spawn_fd_reader(&mut self, group_idx: usize) {
let tx = match self.event_tx.as_ref() {
Some(t) => t.clone(),
None => {
eprintln!("[ERROR] Cannot spawn reader: event_tx not initialized");
return;
}
};
let dc = match &self.shared_dir_cache {
Some(d) => d.clone(),
None => {
eprintln!("[ERROR] Cannot spawn reader: shared_dir_cache not initialized");
return;
}
};
let death_tx = self.reader_death_tx.clone();
let buf_size = self.buffer_size;
let debug = self.debug;
let group = &self.fs_groups[group_idx];
// Duplicate fds so the reader task owns independent copies
let owned_fan_fd = match Self::dup_fd(&group.fan_fd) {
Ok(fd) => fd,
Err(e) => {
eprintln!(
"[ERROR] Failed to dup fanotify fd {}: {}",
group.fan_fd.as_raw_fd(),
e
);
return;
}
};
let owned_mount_fd = match Self::dup_fd(&group.mount_fd) {
Ok(fd) => fd,
Err(e) => {
eprintln!(
"[ERROR] Failed to dup mount fd {}: {}",
group.mount_fd.as_raw_fd(),
e
);
// owned_fan_fd drops here, closing the dup'd fan fd
return;
}
};
let raw_fd = owned_fan_fd.as_raw_fd();
let mfds = Arc::new(vec![owned_mount_fd]);
if debug {
eprintln!(
"[DEBUG] spawning reader for group {} (fd {})",
group_idx, raw_fd
);
}
tokio::spawn(async move {
if debug {
eprintln!(
"[DEBUG] reader task spawned for group {} (fd {})",
group_idx, raw_fd
);
}
let afd = match AsyncFd::new(owned_fan_fd) {
Ok(a) => {
if debug {
eprintln!("[DEBUG] reader {} AsyncFd created, entering loop", raw_fd);
}
a
}
Err(e) => {
eprintln!("[ERROR] AsyncFd for fd {}: {}", raw_fd, e);
let _ = death_tx.send(group_idx);
return;
}
};
let mut buf = vec![0u8; buf_size];
loop {
let result = afd.readable().await;
let mut guard = match result {
Ok(g) => g,
Err(e) => {
eprintln!("[ERROR] fd {} readable: {}", raw_fd, e);
break;
}
};
let events = read_fid_events_cached(afd.get_ref(), &mfds, &dc, &mut buf);
if debug {
eprintln!(
"[DEBUG] fd {} reader: got {} event(s)",
raw_fd,
events.len()
);
}
if !events.is_empty() {
let send_err = match &tx {
EventSender::Unbounded(tx) => tx.send(events).is_err(),
EventSender::Bounded(tx) => tx.send(events).await.is_err(),
};
if send_err {
break;
}
// Edge-triggered epoll: retain readiness so the next
// readable().await resolves immediately if more events
// are still queued (e.g. DELETE → DELETE_SELF batch).
guard.retain_ready();
if debug {
eprintln!("[DEBUG] fd {} reader: retain_ready, looping", raw_fd);
}
} else {
if debug {
eprintln!("[DEBUG] fd {} reader: empty read, clear_ready", raw_fd);
}
guard.clear_ready();
}
}
if debug {
eprintln!(
"[DEBUG] Reader task for group {} (fd {}) exited",
group_idx, raw_fd
);
}
let _ = death_tx.send(group_idx);
});
// Track reader state for restart backoff
if let Some(state) = self
.reader_states
.get_mut(group_idx)
.and_then(|s| s.as_mut())
{
state.restart_count += 1;
state.last_restart = std::time::Instant::now();
state.gave_up = false;
} else {
// Ensure reader_states is large enough
if group_idx >= self.reader_states.len() {
self.reader_states.resize_with(group_idx + 1, || None);
}
self.reader_states[group_idx] = Some(ReaderState {
restart_count: 1,
last_restart: std::time::Instant::now(),
gave_up: false,
});
}
self.metrics.set_reader_groups(self.fs_groups.len() as i64);
}
/// Restart a reader task that has died.
///
/// Applies exponential backoff: up to MAX_RESTARTS within BACKOFF_WINDOW.
/// If the backoff limit is exceeded, logs a warning and gives up.
/// On success, the dead task's fds are re-duplicated from FsGroup and
/// a new reader is spawned.
pub(crate) fn restart_reader(&mut self, group_idx: usize) {
// Check backoff limits
let now = std::time::Instant::now();
let state = self.reader_states.get(group_idx).and_then(|s| s.as_ref());
if let Some(s) = state {
let in_window = now.duration_since(s.last_restart) < BACKOFF_WINDOW;
if in_window && s.restart_count >= MAX_RESTARTS {
eprintln!(
"[ERROR] Reader task for group {} has crashed {} times in \
the last {}s — giving up. fsmon daemon restart required.",
group_idx,
MAX_RESTARTS,
BACKOFF_WINDOW.as_secs(),
);
// Mark gave_up so health() reports accurate alive/dead status.
// This will be reset when spawn_fd_reader is called again.
if let Some(s) = self
.reader_states
.get_mut(group_idx)
.and_then(|s| s.as_mut())
{
s.gave_up = true;
}
return;
}
}
// Verify the FsGroup still exists (may have been removed during shutdown)
if group_idx >= self.fs_groups.len() {
eprintln!(
"[WARNING] Cannot restart reader for group {}: group no longer exists",
group_idx
);
return;
}
let dev_id = self.fs_groups[group_idx].dev_id;
eprintln!(
"[INFO] Restarting reader task for group {} (dev_id={})...",
group_idx, dev_id
);
self.spawn_fd_reader(group_idx);
}
}