From 1308b5bcf33c1d0778d06d7eb32c1138199f4ead Mon Sep 17 00:00:00 2001
From: Aiden McClelland <me@drbonez.dev>
Date: Mon, 11 Mar 2024 16:10:59 -0600
Subject: [PATCH] wait up to 30s for ip address

---
 Makefile                                      |  1 +
 core/startos/src/lxc/mod.rs                   | 38 ++++++++++++-------
 .../src/service/persistent_container.rs       |  2 +-
 3 files changed, 26 insertions(+), 15 deletions(-)

diff --git a/Makefile b/Makefile
index cb3d976e1..bc1edc6fe 100644
--- a/Makefile
+++ b/Makefile
@@ -148,6 +148,7 @@ update-overlay: $(ALL_TARGETS)
 	$(call ssh,"sudo systemctl start startd")
 
 wormhole: core/target/$(ARCH)-unknown-linux-musl/release/startbox
+	@echo "Paste the following command into the shell of your start-os server:"
 	@wormhole send core/target/$(ARCH)-unknown-linux-musl/release/startbox 2>&1 | awk -Winteractive '/wormhole receive/ { printf "sudo /usr/lib/startos/scripts/chroot-and-upgrade \"cd /usr/bin && rm startbox && wormhole receive --accept-file %s && chmod +x startbox\"\n", $$3 }'
 
 update: $(ALL_TARGETS)
diff --git a/core/startos/src/lxc/mod.rs b/core/startos/src/lxc/mod.rs
index 374952f98..f77837238 100644
--- a/core/startos/src/lxc/mod.rs
+++ b/core/startos/src/lxc/mod.rs
@@ -38,6 +38,7 @@ const LXC_CONTAINER_DIR: &str = "/var/lib/lxc";
 const RPC_DIR: &str = "media/startos/rpc"; // must not be absolute path
 pub const CONTAINER_RPC_SERVER_SOCKET: &str = "service.sock"; // must not be absolute path
 pub const HOST_RPC_SERVER_SOCKET: &str = "host.sock"; // must not be absolute path
+const CONTAINER_DHCP_TIMEOUT: Duration = Duration::from_secs(30);
 
 pub struct LxcManager {
     containers: Mutex<Vec<Weak<InternedString>>>,
@@ -110,7 +111,6 @@ impl LxcManager {
 pub struct LxcContainer {
     manager: Weak<LxcManager>,
     rootfs: OverlayGuard,
-    ip: Ipv4Addr,
     guid: Arc<InternedString>,
     rpc_bind: TmpMountGuard,
     config: LxcConfig,
@@ -171,20 +171,9 @@ impl LxcContainer {
             .arg(&*guid)
             .invoke(ErrorKind::Lxc)
             .await?;
-        let ip = String::from_utf8(
-            Command::new("lxc-info")
-                .arg("--name")
-                .arg(&*guid)
-                .arg("-iH")
-                .invoke(ErrorKind::Docker)
-                .await?,
-        )?
-        .trim()
-        .parse()?;
         Ok(Self {
             manager: Arc::downgrade(manager),
             rootfs,
-            ip,
             guid: Arc::new(guid),
             rpc_bind,
             config,
@@ -196,8 +185,29 @@ impl LxcContainer {
         self.rootfs.path()
     }
 
-    pub fn ip(&self) -> Ipv4Addr {
-        self.ip
+    pub async fn ip(&self) -> Result<Ipv4Addr, Error> {
+        let start = Instant::now();
+        loop {
+            let output = String::from_utf8(
+                Command::new("lxc-info")
+                    .arg("--name")
+                    .arg(&*self.guid)
+                    .arg("-iH")
+                    .invoke(ErrorKind::Docker)
+                    .await?,
+            )?;
+            let out_str = output.trim();
+            if !out_str.is_empty() {
+                return Ok(out_str.parse()?);
+            }
+            if start.elapsed() > CONTAINER_DHCP_TIMEOUT {
+                return Err(Error::new(
+                    eyre!("Timed out waiting for container to acquire DHCP lease"),
+                    ErrorKind::Timeout,
+                ));
+            }
+            tokio::time::sleep(Duration::from_millis(100)).await;
+        }
     }
 
     pub fn rpc_dir(&self) -> &Path {
diff --git a/core/startos/src/service/persistent_container.rs b/core/startos/src/service/persistent_container.rs
index 523b5f19a..a3756b37a 100644
--- a/core/startos/src/service/persistent_container.rs
+++ b/core/startos/src/service/persistent_container.rs
@@ -180,7 +180,7 @@ impl PersistentContainer {
         }
         let net_service = ctx
             .net_controller
-            .create_service(s9pk.as_manifest().id.clone(), lxc_container.ip())
+            .create_service(s9pk.as_manifest().id.clone(), lxc_container.ip().await?)
             .await?;
         Ok(Self {
             s9pk,