From f40e31b0376638a7e4471b248cf37c3bc77d677b Mon Sep 17 00:00:00 2001
From: Jordan Doyle <jordan@doyle.la>
Date: Sat, 28 Aug 2021 01:37:56 +0100
Subject: [PATCH] Read messages off the wire from git clients

---
 Cargo.lock       |  2 ++
 Cargo.toml       |  2 ++
 src/main.rs      | 27 ++++++++++++++++++++++++++-
 src/git/codec.rs | 68 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 src/git/mod.rs   |  2 ++
 5 files changed, 95 insertions(+), 6 deletions(-)

diff --git a/Cargo.lock b/Cargo.lock
index c3ec7e1..656b3c2 100644
--- a/Cargo.lock
+++ a/Cargo.lock
@@ -189,11 +189,13 @@
  "anyhow",
  "async-trait",
  "axum",
+ "bytes",
  "env_logger",
  "futures",
  "thrussh",
  "thrussh-keys",
  "tokio",
+ "tokio-util",
  "tower",
  "tower-http",
 ]
diff --git a/Cargo.toml b/Cargo.toml
index ebde3e0..76ddaef 100644
--- a/Cargo.toml
+++ a/Cargo.toml
@@ -17,3 +17,5 @@
 thrussh-keys = "0.21"
 anyhow = "1"
 env_logger = "0.9"
+tokio-util = { version = "0.6", features = ["codec"] }
+bytes = "1"
diff --git a/src/main.rs b/src/main.rs
index 7782588..bb8bf81 100644
--- a/src/main.rs
+++ a/src/main.rs
@@ -7,6 +7,9 @@
 use thrussh::server::{Auth, Session};
 use thrussh::*;
 use thrussh_keys::*;
+use bytes::BytesMut;
+use crate::git::codec::GitCodec;
+use tokio_util::codec::Decoder;
 
 #[tokio::main]
 async fn main() {
@@ -26,13 +29,20 @@
 struct Server;
 
 impl server::Server for Server {
-    type Handler = Self;
-    fn new(&mut self, _: Option<std::net::SocketAddr>) -> Self {
-        self.clone()
+    type Handler = Handler;
+
+    fn new(&mut self, _: Option<std::net::SocketAddr>) -> Self::Handler {
+        Handler::default()
     }
+}
+
+#[derive(Default)]
+struct Handler {
+    codec: GitCodec,
+    input_bytes: BytesMut,
 }
 
-impl server::Handler for Server {
+impl server::Handler for Handler {
     type Error = anyhow::Error;
     type FutureAuth = futures::future::Ready<Result<(Self, server::Auth), anyhow::Error>>;
     type FutureUnit = futures::future::Ready<Result<(Self, Session), anyhow::Error>>;
@@ -101,9 +111,14 @@
         eprintln!("finished auth pubkey");
         self.finished_auth(server::Auth::Accept)
     }
+
+    fn data(mut self, _channel: ChannelId, data: &[u8], session: Session) -> Self::FutureUnit {
+        self.input_bytes.extend_from_slice(data);
+
+        while let Some(frame) = self.codec.decode(&mut self.input_bytes).unwrap() {
+            eprintln!("data: {:x?}", frame);
+        }
 
-    fn data(self, _channel: ChannelId, data: &[u8], session: Session) -> Self::FutureUnit {
-        eprintln!("got data: {:x?}", data);
         futures::future::ready(Ok((self, session)))
     }
 
diff --git a/src/git/codec.rs b/src/git/codec.rs
new file mode 100644
index 0000000..42e27b6 100644
--- /dev/null
+++ a/src/git/codec.rs
@@ -1,0 +1,68 @@
+use tokio_util::codec;
+use bytes::{Bytes, Buf};
+
+#[derive(Default)]
+pub struct GitCodec;
+
+impl codec::Decoder for GitCodec {
+    type Item = Bytes;
+    type Error = anyhow::Error;
+
+    fn decode(&mut self, src: &mut bytes::BytesMut) -> Result<Option<Self::Item>, Self::Error> {
+        if src.len() < 4 {
+            return Ok(None);
+        }
+
+        let mut length_bytes = [0u8; 4];
+        length_bytes.copy_from_slice(&src[..4]);
+        let length = u16::from_str_radix(std::str::from_utf8(&length_bytes)?, 16)? as usize;
+
+        if length == 0 // flush-pkt
+            || length == 1 // delim-pkt
+            || length == 2 // response-end-pkt
+        {
+            src.advance(4);
+            return self.decode(src);
+        }
+
+        if length > 65520 || length <= 4 {
+            return Err(std::io::Error::new(std::io::ErrorKind::InvalidData, "protocol abuse").into());
+        }
+
+        if src.len() < length {
+            src.reserve(length - src.len());
+            return Ok(None);
+        }
+
+        let mut bytes = src.split_to(length);
+        bytes.advance(4);
+
+        if bytes.ends_with(b"\n") {
+            bytes.truncate(bytes.len() - 1);
+        }
+
+        Ok(Some(bytes.freeze()))
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use tokio_util::codec::Decoder;
+    use bytes::BytesMut;
+    use std::fmt::Write;
+
+    #[test]
+    fn decode() {
+        let mut codec = super::GitCodec;
+
+        let mut bytes = BytesMut::new();
+
+        bytes.write_str("0015agent=git/2.32.0").unwrap();
+        let res = codec.decode(&mut bytes).unwrap();
+        assert_eq!(res, None);
+
+        bytes.write_char('\n').unwrap();
+        let res = codec.decode(&mut bytes).unwrap();
+        assert_eq!(res.as_deref(), Some("agent=git/2.32.0".as_bytes()));
+    }
+}
diff --git a/src/git/mod.rs b/src/git/mod.rs
index f464cbf..7e366b3 100644
--- a/src/git/mod.rs
+++ a/src/git/mod.rs
@@ -1,3 +1,5 @@
+pub mod codec;
+
 use thrussh::CryptoVec;
 
 pub const END_OF_MESSAGE: &'static [u8] = b"0000";
--
rgit 0.1.3