Reacquire lock when runtime exits

This commit adds changes so that the main process almost immediately
acquires a lock on the data file when runtime is dropped. This is just
an added precaution to try and ensure that no other process does
something silly with the data file.

The descriptor is cloned for this using `FileLock::try_clone`

.gitignore

@@ -1,6 +1,6 @@
 /target
 /.vscode
-data.bin
+*.bin
 /server/snapshots
 snapstore.bin
 snapstore.partmap

server/src/coredb/mod.rs

@@ -52,7 +52,7 @@ macro_rules! flush_db {
         crate::coredb::CoreDB::flush_db(&$db, None)
     };
     ($db:expr, $file:expr) => {
-        crate::coredb::CoreDB::flush_db(&$db, Some($file))
+        crate::coredb::CoreDB::flush_db(&$db, Some(&mut $file))
     };
 }
 
@@ -286,7 +286,7 @@ impl CoreDB {
         bgsave: BGSave,
         snapshot_cfg: SnapshotConfig,
         restore_file: Option<PathBuf>,
-    ) -> TResult<(Self, Option<flock::FileLock>)> {
+    ) -> TResult<(Self, Option<flock::FileLock>, flock::FileLock)> {
         let coretable = diskstore::get_saved(restore_file)?;
         let mut background_tasks: usize = 0;
         if !bgsave.is_disabled() {
@@ -324,12 +324,13 @@ impl CoreDB {
         ));
         let lock = flock::FileLock::lock("data.bin")
             .map_err(|e| format!("Failed to acquire lock on data file with error '{}'", e))?;
+        let cloned_descriptor = lock.try_clone()?;
         if bgsave.is_disabled() {
-            Ok((db, Some(lock)))
+            Ok((db, Some(lock), cloned_descriptor))
         } else {
             // Spawn the BGSAVE service in a separate task
             tokio::spawn(diskstore::bgsave_scheduler(db.clone(), bgsave, lock));
-            Ok((db, None))
+            Ok((db, None, cloned_descriptor))
         }
     }
     /// Create an empty in-memory table
@@ -366,7 +367,7 @@ impl CoreDB {
         self.shared.table.read()
     }
     /// Flush the contents of the in-memory table onto disk
-    pub fn flush_db(&self, file: Option<flock::FileLock>) -> TResult<()> {
+    pub fn flush_db(&self, file: Option<&mut flock::FileLock>) -> TResult<()> {
         let data = match self.acquire_write() {
             Some(wlock) => wlock,
             None => return Err("Can no longer flush data; coretable is poisoned".into()),

server/src/dbnet/mod.rs

@@ -44,7 +44,8 @@ use crate::config::PortConfig;
 use crate::config::SnapshotConfig;
 use crate::config::SslOpts;
 use crate::dbnet::tcp::Listener;
-use crate::diskstore::snapshot::DIR_REMOTE_SNAPSHOT;
+use crate::diskstore::{self, snapshot::DIR_REMOTE_SNAPSHOT};
+use diskstore::flock;
 mod tcp;
 use crate::CoreDB;
 use libsky::TResult;
@@ -315,16 +316,17 @@ pub async fn run(
     snapshot_cfg: SnapshotConfig,
     sig: impl Future,
     restore_filepath: Option<PathBuf>,
-) -> CoreDB {
+) -> (CoreDB, flock::FileLock) {
     let (signal, _) = broadcast::channel(1);
     let (terminate_tx, terminate_rx) = mpsc::channel(1);
-    let (db, lock) = match CoreDB::new(bgsave_cfg, snapshot_cfg, restore_filepath) {
-        Ok((db, lock)) => (db, lock),
-        Err(e) => {
-            log::error!("ERROR: {}", e);
-            process::exit(0x100);
-        }
-    };
+    let (db, lock, cloned_descriptor) =
+        match CoreDB::new(bgsave_cfg, snapshot_cfg, restore_filepath) {
+            Ok((db, lock, cloned_descriptor)) => (db, lock, cloned_descriptor),
+            Err(e) => {
+                log::error!("ERROR: {}", e);
+                process::exit(0x100);
+            }
+        };
     match fs::create_dir_all(&*DIR_REMOTE_SNAPSHOT) {
         Ok(_) => (),
         Err(e) => match e.kind() {
@@ -391,7 +393,7 @@ pub async fn run(
         log::error!("Failed to release lock on data file with '{}'", e);
         process::exit(0x100);
     }
-    db
+    (db, cloned_descriptor)
 }
 
 /// This is a **test only** function

server/src/diskstore/flock.rs

@@ -97,6 +97,18 @@ impl FileLock {
         // Now write to the file
         self.file.write_all(bytes)
     }
+    pub fn try_clone(&self) -> Result<Self> {
+        Ok(Self {
+            file: self.file.try_clone()?,
+            unlocked: false,
+        })
+    }
+    /// Reacquire a lock
+    pub fn reacquire(&mut self) -> Result<()> {
+        Self::_lock(&self.file)?;
+        self.unlocked = false;
+        Ok(())
+    }
 }
 
 impl Drop for FileLock {
@@ -146,6 +158,17 @@ mod tests {
         file2.unlock().unwrap();
         drop(file2);
     }
+    #[cfg(windows)]
+    #[test]
+    fn test_release_one_acquire_second() {
+        let mut file = FileLock::lock("data5.bin").unwrap();
+        let mut cloned = file.try_clone().unwrap();
+        file.write(&[1, 2, 3]).unwrap();
+        drop(file);
+        cloned.reacquire().unwrap();
+        cloned.write(&[4, 5, 6]).unwrap();
+        cloned.unlock().unwrap();
+    }
 }
 
 #[cfg(windows)]

server/src/main.rs

@@ -77,10 +77,10 @@ fn main() {
         .enable_all()
         .build()
         .unwrap();
-    let db = runtime.block_on(async {
+    let (db, mut descriptor) = runtime.block_on(async {
         let (tcplistener, bgsave_config, snapshot_config, restore_filepath) =
             check_args_and_get_cfg().await;
-        let db = run(
+        let (db, descriptor) = run(
             tcplistener,
             bgsave_config,
             snapshot_config,
@@ -88,7 +88,7 @@ fn main() {
             restore_filepath,
         )
         .await;
-        db
+        (db, descriptor)
     });
     // Make sure all background workers terminate
     drop(runtime);
@@ -97,14 +97,19 @@ fn main() {
         1,
         "Maybe the compiler reordered the drop causing more than one instance of CoreDB to live at this point"
     );
-    if let Err(e) = flush_db!(db) {
+    // Try to acquire lock almost immediately
+    if let Err(e) = descriptor.reacquire() {
+        log::error!("Failed to reacquire lock on data file with error: '{}'", e);
+        panic!("FATAL: data file relocking failure");
+    }
+    if let Err(e) = flush_db!(db, descriptor) {
         log::error!("Failed to flush data to disk with '{}'", e);
         loop {
             // Keep looping until we successfully write the in-memory table to disk
             log::warn!("Press enter to try again...");
             io::stdout().flush().unwrap();
             io::stdin().read(&mut [0]).unwrap();
-            if let Ok(_) = flush_db!(db) {
+            if let Ok(_) = flush_db!(db, descriptor) {
                 log::info!("Successfully saved data to disk");
                 break;
             } else {