From d2fc42d65f33558342a8947c6e8229e481035b9f Mon Sep 17 00:00:00 2001
From: guessi <guessi@gmail.com>
Date: Wed, 17 Jun 2026 19:42:05 +0800
Subject: [PATCH] fix: replace efs-proxy bind panic with graceful error exit

efs-proxy panicked with an unhelpful backtrace when it failed to bind
to 127.0.0.1:<port>. This manifested as a hard crash with no actionable
error message in environments like Bottlerocket/EKS containers where
the loopback bind can fail (e.g. permission denied, address in use).

Change Controller::new to return Result<Self, io::Error> and have
main() exit with code 1 and a clear error message instead of panicking.
The mount helper already detects a non-zero efs-proxy exit and surfaces
it to the user via mount.log.

Fixes #323
---
 src/proxy/src/controller.rs | 56 ++++++++++++++++++++++++++-----------
 src/proxy/src/main.rs       | 18 ++++++++++--
 2 files changed, 55 insertions(+), 19 deletions(-)

diff --git a/src/proxy/src/controller.rs b/src/proxy/src/controller.rs
index cce95db8..ca0a1175 100644
--- a/src/proxy/src/controller.rs
+++ b/src/proxy/src/controller.rs
@@ -110,12 +110,17 @@ impl<S: ProxyStream> Controller<S> {
         partition_finder: Arc<impl PartitionFinder<S> + Sync + Send + 'static>,
         status_reporter: StatusReporter,
         cw_publisher: Option<Arc<dyn CloudWatchClient>>,
-    ) -> Self {
-        let Ok(listener) = TcpListener::bind(listen_addr).await else {
-            panic!("Failed to bind {}", listen_addr);
-        };
+    ) -> Result<Self, std::io::Error> {
+        let listener = TcpListener::bind(listen_addr).await.map_err(|e| {
+            error!(
+                "Failed to bind to {}: {}. \
+                Ensure the address is reachable and the port is not already in use.",
+                listen_addr, e
+            );
+            e
+        })?;
 
-        Self {
+        Ok(Self {
             listener,
             partition_finder,
             proxy_id: ProxyIdentifier::new(),
@@ -125,7 +130,7 @@ impl<S: ProxyStream> Controller<S> {
             status_reporter,
             proxy_config,
             cw_publisher,
-        }
+        })
     }
 
     pub async fn run<T: RpcClient, V: S3ClientBuilder>(
@@ -218,7 +223,12 @@ impl<S: ProxyStream> Controller<S> {
             }
 
             // Skip channel init if read bypass is not requested
-            let channel_init_config = if !self.proxy_config.nested_config.read_bypass_config.requested {
+            let channel_init_config = if !self
+                .proxy_config
+                .nested_config
+                .read_bypass_config
+                .requested
+            {
                 ChannelInitConfig::default()
             } else {
                 let configs = vec![ChannelConfigArgs::AWSFILE_READ_BYPASS_V2(
@@ -503,10 +513,8 @@ impl<S: ProxyStream> Controller<S> {
 mod tests {
     use super::*;
     use crate::{
-        aws::cw_publisher::LogLevel,
-        config::channel_init_config::ChannelInitConfig,
-        proxy_builder::ProxyBuilder,
-        status_reporter::create_status_channel,
+        aws::cw_publisher::LogLevel, config::channel_init_config::ChannelInitConfig,
+        proxy_builder::ProxyBuilder, status_reporter::create_status_channel,
     };
     use std::sync::atomic::AtomicU64;
     use tokio::net::TcpStream;
@@ -599,12 +607,7 @@ mod tests {
         )
         .await;
 
-        let mut state = IncarnationState::new(
-            ProxyIdentifier::new(),
-            None,
-            events_tx,
-            1,
-        );
+        let mut state = IncarnationState::new(ProxyIdentifier::new(), None, events_tx, 1);
 
         let period_secs = 10;
         let publisher_clone = mock_publisher.clone();
@@ -658,4 +661,23 @@ mod tests {
         let proxy = handle.await.unwrap();
         let _ = proxy.shutdown().await;
     }
+
+    #[tokio::test]
+    async fn test_controller_new_bind_failure_returns_err() {
+        // Occupy a port so the second bind attempt fails.
+        let occupied = TcpListener::bind("127.0.0.1:0").await.unwrap();
+        let addr = occupied.local_addr().unwrap();
+
+        let (_status_requester, status_reporter) = create_status_channel();
+        let result = Controller::new(
+            &addr.to_string(),
+            ProxyConfig::default(),
+            Arc::new(MockPartitionFinder),
+            status_reporter,
+            None,
+        )
+        .await;
+
+        assert!(result.is_err(), "expected Err when port is already bound");
+    }
 }
diff --git a/src/proxy/src/main.rs b/src/proxy/src/main.rs
index d1432c6a..c18f1f40 100644
--- a/src/proxy/src/main.rs
+++ b/src/proxy/src/main.rs
@@ -100,7 +100,14 @@ async fn main() {
             status_reporter,
             cw_publisher.clone(),
         )
-        .await;
+        .await
+        .unwrap_or_else(|_| {
+            let p = std::path::Path::new(&proxy_config.pid_file_path);
+            if p.exists() {
+                let _ = std::fs::remove_file(p);
+            }
+            std::process::exit(1);
+        });
         tokio::spawn(controller.run(
             sigterm_cancellation_token.clone(),
             AwsFileRpcClient,
@@ -116,7 +123,14 @@ async fn main() {
             status_reporter,
             cw_publisher.clone(),
         )
-        .await;
+        .await
+        .unwrap_or_else(|_| {
+            let p = std::path::Path::new(&proxy_config.pid_file_path);
+            if p.exists() {
+                let _ = std::fs::remove_file(p);
+            }
+            std::process::exit(1);
+        });
         tokio::spawn(controller.run(
             sigterm_cancellation_token.clone(),
             AwsFileRpcClient,