Report unexpected trailing bytes at the end of INIT and OPEN SYN

[jcos1]

Describe the bug

What happens: When there is a stray byte at the end of the OPEN SYN Zenohd will reply with an OPEN ACK It will not report anything about the trailing byte. When sending other messages (KEEP_ALIVE [1,0,4] or PUT-FRAME) it will not process these messages. Zenohd will also not report anything about those messages. It will not recover (discarding that trailing byte). It will still send KEEP_ALIVE messages. After a while zenohd will time out.

What I expect to happen: I think zenohd should report any unexpected bytes. Perhaps it should also try to recover.

To reproduce

1. Start Zenohd
2. Do the handshake but add a trailing byte at the end of the OPEN SYN

For example, OPEN SYN: [38,00,66,255,0,1 + COOKIE + 00]

zenohd config changes:

  "id": "7",
...
 "connect_scouted": null,
...
 "batching": {
       "enabled": false,
...
"sequence_number_resolution": "8bit",
...
"shared_memory": {
  "enabled": false
...
 "qos": {
     "enabled": false

To reproduce:

use tokio::time::Duration;
use tokio::net::TcpStream;
use tokio::io::{AsyncReadExt, AsyncWriteExt};
use tokio::time::sleep;

#[tokio::main]
async fn main() -> tokio::io::Result<()> {

    let mut buffer: [u8; 1024] = [0; 1024];  

    let mut stream = TcpStream::connect("127.0.0.1:7447").await?;
    println!("Connection ok");
    
    let z_init = [4,0, 1_u8,9,2,1];     
    println!("Sending INIT SYN {:x?}", z_init);
    stream.write_all(&z_init).await?;
    stream.flush().await?;
    println!("Waiting for INIT ACK");
    
    // INIT ACK        
    let bytes_read = stream.read(&mut buffer).await?;    
    println!("Received INIT ACK: {:x?}", &buffer[..bytes_read]);
    // Store the cookie received with the INIT ACK 
    let mut z_cookie = buffer[9..43].to_vec(); 
    println!("Cookie: {:x?}", z_cookie);

    // OPEN SYN
    let mut z_open = vec![38_u8,0, 66, 255, 0,1];         
    z_open.append(&mut z_cookie);  
    let mut bad_byte = vec![00]; // This shouldn't be here
    z_open.append(&mut bad_byte);  
    println!("Sending OPEN SYN {:x?}", z_open);      
    stream.write_all(&z_open).await?;    
    stream.flush().await?;

    println!("Waiting for OPEN ACK");
    let bytes_read = stream.read(&mut buffer).await?;    
    println!("OPEN ACK: {:x?}", &buffer[..bytes_read]);

    // KEEP ALIVE
    let z_keep_alive = [1_u8,0,4];
    println!("Sending KEEP_ALIVE{:x?}", z_keep_alive);
    stream.write_all(&z_keep_alive).await?;
    stream.flush().await?;
    let bytes_read = stream.read(&mut buffer).await?;    
    println!("KEEP_ALIVE ACK: {:x?}", &buffer[..bytes_read]);

    let z_put = [9u8,0,5,15,61,0,1,65,1,1,65]; 
    println!("Sending FRAME: {:x?}", z_put);
    stream.write_all(&z_put).await?;
    stream.flush().await?;
    let bytes_read = stream.read(&mut buffer).await?;    
    println!("PUT reply: {:x?}", &buffer[..bytes_read]); // it replies with a keep alive
    
    sleep(Duration::from_millis(2000)).await;

    // KEEP ALIVE
    let z_keep_alive = [1_u8,0,4];
    println!("Sending KEEP_ALIVE{:x?}", z_keep_alive);
    stream.write_all(&z_keep_alive).await?;
    stream.flush().await?;
    let bytes_read = stream.read(&mut buffer).await?;    
    println!("KEEP_ALIVE ACK: {:x?}", &buffer[..bytes_read]);

    Ok(())
}

System info

• Zenohd v1.2.0 on windows
• The actual client sending the stray byte is a PLC, but a code snippet with the same result is shown above.

[Mallets]

I don't think this is a bug... the protocol doesn't allow for trailing bytes nor bytes misalignment. When the protocol is not respected than undefined behaviour is to be expected.

Keeping in mind this kind of behaviour affects only the connection being established because the PLC is not respecting the protocol.

Is this happening with zenoh-pico or with a custom implementation of the protocol?

[jcos1]

This is happening with a custom implementation of the protocol.

Perhaps it can be marked as a feature request?

It would help during debugging and troubleshooting.

[kydos]

I don't think this is a bug... the protocol doesn't allow for trailing bytes nor bytes misalignment. When the protocol is not respected than undefined behaviour is to be expected.

Trailing bytes will be interpreted as the beginning of the next message, and with highly likelihood that will be considered as either an unknown or ill-formed message and the session will be closed. If the trailing bytes end-up representing a valid Zenoh message, then those will be interpreted in the context of the current state of the protocol. Thus I am not sure it is accurate to talk about undefined behaviour.

[jcos1]

That's correct. It keeps reading bytes until the session times out because zenohd can't figure out what to do with the bytes it is receiving.

A nice feature would be to make zenohd print out any unprocessed (=received but not interpreted) bytes (when RUST_LOG is DEBUG or TRACE)