From b6cac2d133c975e16a66f88afa093f564ae66441 Mon Sep 17 00:00:00 2001
From: DavidVentura <davidventura27@gmail.com>
Date: Sat, 27 Jan 2024 16:24:35 +0100
Subject: [PATCH] add audio support

---
 README.md   |  4 ++--
 handlers.ts | 10 ++++++++--
 server.ts   | 11 ++++++++---
 3 files changed, 18 insertions(+), 7 deletions(-)

diff --git a/README.md b/README.md
index 0989a3a..72879cc 100644
--- a/README.md
+++ b/README.md
@@ -18,7 +18,7 @@ The hooks used with frida are at `frida-hooks.js`, but it's mostly a playground
 There's also a pretty crappy Wireshark dissector at `dissector.lua`. You can install it with `make install-wireshark-dissector`.
 
 ## Running
-To execute the server, run `make run`; JPEG files will be created in a folder named `captures`.
+To execute the server, run `make run`; JPEG files and `audio.pcm` will be created in a folder named `captures`.
 
 There's no live-stream server built into this project yet.
 
@@ -73,7 +73,7 @@ To start a stream, a single _control packet_ is sent.
 
 The received stream is broken up into 1028 byte payloads, along with a sequence number.
 
-Stitching the payloads together yields JPEG frames for video, and (TBD) for audio.
+Stitching the payloads together yields JPEG frames for video, and 8KHz A-law PCM for audio.
 
 ```mermaid
 ---
diff --git a/handlers.ts b/handlers.ts
index 92130e4..9e5c5fb 100644
--- a/handlers.ts
+++ b/handlers.ts
@@ -1,6 +1,7 @@
 import { Commands, CommandsByValue, ControlCommands } from "./datatypes.js";
 import { create_P2pRdy, SendStartVideo, SendUsrChk } from "./impl.js";
 import { Session } from "./server.js";
+import { u16_swap } from "./utils.js";
 
 let curImage = null;
 
@@ -66,12 +67,17 @@ const deal_with_data = (session: Session, dv: DataView) => {
   }
 
   if (audio) {
-    // TODO audio pkt
+    // "stream_head_t->type == 0x06" per pdf
+    if (dv.add(12).readU8() == 0x06) {
+      const audio_len = u16_swap(dv.add(8 + 16).readU16());
+      const audio_buf = dv.add(32 + 8).readByteArray(audio_len).buffer; // 8 for pkt header, 32 for `stream_head_t`
+      session.eventEmitter.emit("audio", Buffer.from(audio_buf));
+    }
   } else {
     const data = dv.add(8).readByteArray(pkt_len - 4);
     if (is_new_image) {
       if (curImage != null) {
-        session.frameEmitter.emit("frame", curImage);
+        session.eventEmitter.emit("frame", curImage);
       }
       curImage = Buffer.from(data.buffer);
     } else {
diff --git a/server.ts b/server.ts
index fc2371a..448b562 100644
--- a/server.ts
+++ b/server.ts
@@ -11,7 +11,7 @@ export type Session = {
   broadcast: (msg: DataView) => void;
   outgoingCommandId: number;
   ticket: number[];
-  frameEmitter: EventEmitter;
+  eventEmitter: EventEmitter;
 };
 
 export type PacketHandler = (session: Session, dv: DataView) => void;
@@ -50,7 +50,7 @@ const makeSession = (cb: msgCb, connCb: connCb, options?: opt): Session => {
   const session: Session = {
     outgoingCommandId: 0,
     ticket: [0, 0, 0, 0],
-    frameEmitter: new EventEmitter(),
+    eventEmitter: new EventEmitter(),
     send: (msg: DataView) => {
       const raw = msg.readU16();
       const cmd = CommandsByValue[raw];
@@ -117,7 +117,8 @@ const s = makeSession(
 );
 
 let cur_image_index = 0;
-s.frameEmitter.on("frame", (frame: Buffer) => {
+const audioFd = createWriteStream(`captures/audio.pcm`);
+s.eventEmitter.on("frame", (frame: Buffer) => {
   const fname = `captures/${cur_image_index.toString().padStart(4, "0")}.jpg`;
   let cur_image = createWriteStream(fname);
   cur_image_index++;
@@ -125,3 +126,7 @@ s.frameEmitter.on("frame", (frame: Buffer) => {
   cur_image.close();
   console.log("got an entire frame", frame.length);
 });
+
+s.eventEmitter.on("audio", (frame: Buffer) => {
+  audioFd.write(frame);
+});