Handle device EOF and I/O errors in the main event loop #88

[SajjadPourali]

#88

[Copilot]

🔵 Needs a closer look

Existing stream tasks are not notified or terminated on EOF or read errors, which can leave accepted streams waiting indefinitely.

Pull request overview

Handles device EOF and I/O errors in the IP stack’s main event loop.

Changes:

• Treats zero-byte reads as EOF.
• Logs and propagates device read errors.
• Adds EOF regression coverage.

File summaries

File	Summary
src/lib.rs	Updates event-loop termination handling and adds EOF coverage.

Review details

Suppressed comments (2)

src/lib.rs:326

• Returning here drops the sessions map, but it does not stop already accepted stream tasks. sessions only owns a cloned packet sender, while each IpStackTcpStream/IpStackUdpStream retains another sender for its receive channel, so those tasks are not notified that the device has ended; an application awaiting an existing stream can remain pending indefinitely after EOF (and after a read error). Propagate a stack-shutdown signal to each session or explicitly abort/close the session tasks before returning.

                        Ok(0) => {
                            log::info!("Device EOF, stopping IP stack");
                            return Ok(());

src/lib.rs:331

• This error path has the same lifecycle problem as the EOF path: returning from the loop drops the event loop and its sessions senders, but does not notify or terminate existing stream tasks. A stream already returned by accept() can therefore wait forever after a device read error; use the same session-shutdown mechanism here before returning.

                        Err(e) => {
                            log::error!("Device read error: {e}");
                            return Err(e.into());

• Files reviewed: 1/1 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.