refactor!: get rid of the pidfile and find a better way to get the ru…

…nning pid #48

Cargo.toml

@@ -35,6 +35,7 @@ tracing-subscriber = { version = "0.3.18", optional = true }
 chrono = { version = "0.4.38", optional = false }
 blake3 = "1.5.4"
 serde_repr = "0.1.19"
+sysinfo = "0.33.1"
 
 [[bin]] # client
 name = "netpulse"

src/bins/netpulsed.rs

@@ -29,7 +29,7 @@ use std::sync::atomic::AtomicBool;
 
 use getopts::Options;
 use netpulse::common::{
-    confirm, exec_cmd_for_user, getpid, getpid_running, init_logging, print_usage, root_guard,
+    confirm, exec_cmd_for_user, getpid_running, init_logging, print_usage, root_guard,
     setup_panic_handler,
 };
 use netpulse::errors::RunError;
@@ -38,6 +38,7 @@ use netpulse::{DAEMON_PID_FILE, DAEMON_USER};
 use nix::errno::Errno;
 use nix::sys::signal::{self, Signal};
 use nix::unistd::Pid;
+use sysinfo::{Process, System};
 use tracing::{debug, error, info, trace};
 
 mod daemon;
@@ -144,35 +145,20 @@ fn setup_general(skip_checks: bool) -> Result<(), RunError> {
 }
 
 fn setup_systemd(skip_checks: bool) -> Result<(), RunError> {
-    let mut is_running: bool = getpid_running().is_some();
-    info!("netpulsed is running: {is_running}");
-    let mut stop_requested = false;
-
-    while is_running {
-        if !stop_requested && is_running {
-            println!("netpulsed.service needs to be stopped if it's running.");
-            println!(
-                "To stop the running netpulsed.service (using systemd), run the following as root:"
-            );
-            println!("  systemctl stop netpulsed.service");
-            if !confirm("Do this automatically now?") {
-                stop_requested = true;
-                continue;
-            }
-            exec_cmd_for_user(
-                Command::new("systemctl")
-                    .arg("stop")
-                    .arg("netpulsed.service"),
-                skip_checks,
-            );
-            stop_requested = true;
-            println!(
-                "waiting until netpulsed is no longer running (pid: {:?})",
-                getpid()
-            );
+    if let Some(pid) = getpid_running() {
+        let s = System::new_all();
+        info!("daemon runs with pid {pid}");
+        let process = s
+            .process(pid)
+            .expect("process for the pid of the daemon not found");
+        if !skip_checks || !confirm("terminate the daemon now?") {
+            println!("stopping setup");
+            std::process::exit(0);
         }
-        is_running = getpid_running().is_some();
-        std::thread::sleep(std::time::Duration::from_millis(50));
+        process
+            .kill_with(sysinfo::Signal::Term)
+            .expect("SIGTERM does not exist on this platform");
+        process.wait(); // wait until the daemon has stopped
     }
 
     setup_general(skip_checks)?;
@@ -227,15 +213,9 @@ fn setup_systemd(skip_checks: bool) -> Result<(), RunError> {
 }
 
 fn infod() {
-    match getpid() {
+    match getpid_running() {
         Some(pid) => {
-            if pid_runs(pid) {
-                println!("netpulsed is running with pid {pid}")
-            } else {
-                println!(
-                    "the pid file exists with pid {pid}, but no process with that pid is running"
-                )
-            }
+            println!("netpulsed is running with pid {pid}")
         }
         None => println!("netpulsed is not running"),
     }
@@ -253,7 +233,8 @@ fn endd() {
             println!("netpulsed is not running");
             return;
         }
-        Some(raw) => Pid::from_raw(raw),
+        Some(raw) => Pid::from_raw(raw.as_u32() as i32), // this is weird, sysinfo has another raw type
+                                                         // for pids than nix
     };
 
     match signal::kill(pid, Signal::SIGTERM) {

src/common.rs

@@ -38,10 +38,9 @@ use std::io::{self, Write as _};
 use std::process::Command;
 use std::str::FromStr;
 
-use crate::DAEMON_PID_FILE;
-
 use getopts::Options;
-use tracing::{error, trace};
+use sysinfo::{Pid, System};
+use tracing::{debug, error, trace, warn};
 use tracing_subscriber::FmtSubscriber;
 
 /// Environment variable name for configuring log level
@@ -241,64 +240,31 @@ pub fn exec_cmd_for_user(cmd: &mut Command, skip_checks: bool) {
     }
 }
 
-/// Checks if the netpulse daemon is currently running.
-///
-/// # Returns
-///
-/// * `Some(pid)` - Daemon is running with the given PID
-/// * `None` - Daemon is not running
-///
-/// Checks both PID file existence and process existence.
-pub fn getpid_running() -> Option<i32> {
-    getpid().filter(|p| pid_runs(*p))
-}
-
-/// Checks if a process with the given PID exists.
-///
-/// # Arguments
-///
-/// * `pid` - Process ID to check
-///
-/// # Returns
-///
-/// * `true` if process exists
-/// * `false` if process does not exist
-///
-/// # Panics
-///
-/// Panics if unable to check process existence (e.g., permission denied)
-pub fn pid_runs(pid: i32) -> bool {
-    std::fs::exists(format!("/proc/{pid}")).expect("could not check if the process exists")
-}
+/// Get the pid of the running netpulsed daemon
+pub fn getpid_running() -> Option<Pid> {
+    let s = System::new_all();
+    let mut processes: Vec<&sysinfo::Process> =
+        s.processes_by_exact_name("netpulsed".as_ref()).collect();
 
-/// Reads the daemon's PID from its PID file.
-///
-/// # Returns
-///
-/// * `Some(pid)` - Successfully read PID from file
-/// * `None` - If PID file doesn't exist or contains invalid data
-///
-/// # Panics
-///
-/// Panics if:
-/// - Unable to check PID file existence
-/// - PID file exists but can't be read
-pub fn getpid() -> Option<i32> {
-    if !std::fs::exists(DAEMON_PID_FILE).expect("couldn't check if the pid file exists") {
+    if processes.is_empty() {
         None
+    } else if processes.len() == 1 {
+        Some(processes[0].pid())
     } else {
-        let pid_raw = std::fs::read_to_string(DAEMON_PID_FILE)
-            .expect("pid file does not exist")
-            .trim()
-            .to_string();
-        let pid = match pid_raw.parse() {
-            Ok(pid) => pid,
-            Err(err) => {
-                error!("Error while parsing the pid from file ('{pid_raw}'): {err}");
-                return None;
-            }
-        };
-        Some(pid)
+        // FIXME: for some reason, this shows many processes even if only two are running.
+        // Expected:
+        // 1. the real daemon
+        // 2. this current process, which is also netpulsed
+        // BUT:
+        // For some absurd reason, there are 18 processes running
+        // Upstream Issue: https://github.com/GuillaumeGomez/sysinfo/issues/1449
+        warn!("netpulsed is running multiple times ({})", processes.len());
+        processes.sort_by_key(|a| a.pid());
+        debug!(
+            "listing netpulsed processes: {:?}",
+            processes.iter().map(|p| p.pid()).collect::<Vec<_>>()
+        );
+        Some(processes[0].pid())
     }
 }