Safeguard suibase-daemon from being started without file lock.

rust/suibase/Cargo.toml

@@ -10,7 +10,7 @@ members = ["crates/suibase-daemon",
 [workspace.package]
 # Bump 'version' for the daemon to self-restart after an update.
 # (this is not the Suibase package version, it is specifically for the Rust crates).
-version = "0.0.14"
+version = "0.0.15"
 edition = "2021"
 
 [workspace.dependencies]

rust/suibase/crates/suibase-daemon/src/main.rs

@@ -91,6 +91,36 @@ impl Command {
     pub async fn execute(self, globals: Globals) -> Result<(), anyhow::Error> {
         match self {
             Command::Run {} => {
+                // Verify that this process is the one running under file lock ~/tmp/.suibase-daemon.lock
+                // If not, exit with an exit code of 13 (which is detected as "do no restart" by Suibase run-daemon.sh)
+                // This is a safeguard against a user trying to start suibase-daemon without the proper locking.
+
+                // Get the pid of this process.
+                let my_pid = std::process::id();
+
+                // Call the bash script "~/suibase/scripts/verify-suibase-daemon-lock $my_pid"
+                let home_dir = home::home_dir().expect("Failed to get home directory");
+                let script_path = home_dir.join("suibase/scripts/common/verify-suibase-daemon-lock.sh");
+
+                // Call the bash script "~/suibase/scripts/verify-suibase-daemon-lock $my_pid"
+                // Returns OK when the process is the one running under file lock ~/tmp/.suibase-daemon.lock
+                let output = std::process::Command::new("/bin/bash")
+                    .arg("-c")
+                    .arg(format!("{} {}", script_path.display(), my_pid))
+                    .output();
+
+                let mut force_exit = true;
+
+                if let Ok(output) = output {
+                    if output.status.success() && output.stdout.starts_with(b"OK") {
+                        force_exit = false;
+                    }
+                }
+
+                if force_exit {
+                    std::process::exit(13);
+                }
+
                 // Create mpsc channels (internal messaging between threads).
                 //
                 // The AdminController handles events about configuration changes

scripts/common/run-daemon.sh

@@ -43,9 +43,7 @@ esac
 
 
 force_stop_all_services() {
-  # This is called in rare cases.
-  #
-  # It will force stop all processes for localnet, but not touch
+  # This will force stop all processes for localnet, but not touch
   # the "user_request" to preserve the user intent.
   #
   # It will create a "force_stop" key-value in each .state workdir
@@ -55,7 +53,6 @@ force_stop_all_services() {
   # It is assumed that the daemon will delete the force_stop,
   # and resume the services as configured by "user_request".
 
-
   update_SUI_FAUCET_PROCESS_PID_var
   if [ -n "$SUI_FAUCET_PROCESS_PID" ]; then
     set_key_value "localnet" "force_stop" "true"
@@ -67,6 +64,22 @@ force_stop_all_services() {
     set_key_value "localnet" "force_stop" "true"
     stop_sui_process
   fi
+
+  # Wait until all process confirm stopped (or timeout).
+  count=0
+  while [ $count -lt 10 ]; do
+    if is_suibase_daemon_running; then
+      # Already running? then do nothing.
+      exit 0
+    fi
+    update_SUI_FAUCET_PROCESS_PID_var
+    update_SUI_PROCESS_PID_var
+    if [ -z "$SUI_FAUCET_PROCESS_PID" ] && [ -z "$SUI_PROCESS_PID" ]; then
+      break
+    fi
+    sleep 1
+    count=$((count + 1))
+  done
 }
 export -f force_stop_all_services
 
@@ -132,33 +145,37 @@ main() {
     # The suibase-daemon is responsible to properly re-start the child processes/services.
     if [ "$PARAM_NAME" = "suibase" ]; then
       if [ -f "$_LOCKFILE" ]; then
-        local _BLOCKED=true
         for i in 1 2 3; do
           if is_suibase_daemon_running; then
-            _BLOCKED=false
-            break
+            # Already running? then do nothing.
+            exit 0
           fi
 
           if ! [ -f "$_LOCKFILE" ]; then
-            _BLOCKED=false
             break
           fi
 
-          sleep 1
-        done
-
-        if [ $_BLOCKED = true ]; then
           # This is to recover when the lockfile exists, but the suibase-daemon
           # is NOT running (e.g. was killed). Killing only the daemon is problematic
           # when its child process are left running. This is why all potential
           # child services are stopped here.
           force_stop_all_services
-        fi
+        done
       fi
     fi
 
     # shellcheck disable=SC2086,SC2016
-    try_locked_command "$_LOCKFILE" /bin/sh -uec 'while true; do "$@" > $0 2>&1; echo "Restarting process" > $0 2>&1; sleep 1; done' "$_LOG" $_CMD_LINE
+    try_locked_command "$_LOCKFILE" /bin/sh -uec '
+    while true; do
+      "$@" > $0 2>&1
+      exit_status=$?
+      if [ $exit_status -eq 13 ]; then
+        echo "Process exited with status 13. Exiting loop." > $0 2>&1
+        break
+      fi
+      echo "Restarting process" > $0 2>&1
+      sleep 1
+    done' "$_LOG" $_CMD_LINE
   fi
 
 }

scripts/common/verify-suibase-daemon-lock.sh

@@ -0,0 +1,40 @@
+#!/bin/bash
+
+# Source '__globals.sh'.
+SUIBASE_DIR="$HOME/suibase"
+SCRIPT_COMMON_CALLER="$(readlink -f "$0")"
+WORKDIR="none"
+
+# Validate that the parameter is a number
+if [ -z "$1" ]; then
+  echo "ERROR: Missing PID to check"
+  exit 1
+fi
+
+if ! [[ "$1" =~ ^[0-9]+$ ]]; then
+  echo "ERROR: Invalid PID: $1"
+  exit 1
+fi
+
+# shellcheck source=SCRIPTDIR/__globals.sh
+source "$SUIBASE_DIR/scripts/common/__globals.sh" "$SCRIPT_COMMON_CALLER" "$WORKDIR"
+trap cleanup EXIT
+
+# shellcheck source=SCRIPTDIR/__suibase-daemon.sh
+source "$SUIBASE_DIR/scripts/common/__suibase-daemon.sh"
+
+# Get the PID of suibase (under the lock file)
+update_SUIBASE_DAEMON_PID_var
+
+if [ -z "$SUIBASE_DAEMON_PID" ]; then
+  echo "ERROR: suibase daemon not running under lock"
+  exit 1
+fi
+
+if [ "$1" != "$SUIBASE_DAEMON_PID" ]; then
+  echo "ERROR: PID $1 is not the suibase-daemon PID under proper lock ($SUIBASE_DAEMON_PID)"
+  exit 1
+fi
+
+echo "OK"
+exit 0