doloro/simple-instant-stream
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: WHIP support

Browse Source
This commit is contained in:
doloro
2026-08-07 11:48:31 +01:00
parent c1435a70fb
commit 5f152e9db7
3 changed files with 349 additions and 189 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crates/server/src/webrtc_ingest.rs
+40 -28
View File
@@ -64,9 +64,7 @@ pub async fn handle_whip_injest_patch(
body: String,
) -> Result<impl IntoResponse, HttpError> {
// The slug is the stream_key_id (set by the POST handler's Location header).
let stream_key_id: i32 = slug
.parse()
.map_err(|e| {
let stream_key_id: i32 = slug.parse().map_err(|e| {
warn!(%slug, "Whip PATCH: bad slug: {:?}", e);
HttpError::NotFound
})?;
@@ -74,16 +72,8 @@ HttpError::NotFound
info!(stream_key_id, ct = ?headers.get(header::CONTENT_TYPE), "Whip PATCH: trickle candidate");
// Look up the trickle sender for this session.
let trickle_map = state
.appstate
.lock()
.await
.webrtc_proxy
.trickle_tx
.clone();
let tx = trickle_map
.get(&stream_key_id)
.ok_or_else(|| {
let trickle_map = state.appstate.lock().await.webrtc_proxy.trickle_tx.clone();
let tx = trickle_map.get(&stream_key_id).ok_or_else(|| {
warn!(stream_key_id, "Whip PATCH: no trickle channel for session");
HttpError::NotFound
})?;
@@ -218,8 +208,7 @@ pub async fn handle_whip_injest(
// Extract negotiated codec, PT, and profile from the answer
// so WHEP viewers can use the exact same codec config.
let (negotiated_codec, video_pt, video_profile) =
extract_negotiated_codec_info(&offer_answe);
let (negotiated_codec, video_pt, video_profile) = extract_negotiated_codec_info(&offer_answe);
info!(
?negotiated_codec,
video_pt,
@@ -231,7 +220,8 @@ pub async fn handle_whip_injest(
return err(StatusCode::NOT_ACCEPTABLE, "no common video codec");
}
let answer_sdp = offer_answe.to_sdp_string()
let answer_sdp = offer_answe
.to_sdp_string()
// Strip a=ice-options:trickle so OBS starts ICE immediately.
.replace("a=ice-options:trickle\r\n", "")
.replace("a=ice-options:trickle\n", "");
@@ -241,13 +231,17 @@ pub async fn handle_whip_injest(
let answer_sdp = answer_sdp
.replace("a=group:BUNDLE 0 1\r\n", "")
.replace("a=group:BUNDLE 0 1\n", "");
let answer_sdp = if let Some(cand_line) = answer_sdp
.lines()
.find(|l| l.starts_with("a=candidate:"))
{
let answer_sdp =
if let Some(cand_line) = answer_sdp.lines().find(|l| l.starts_with("a=candidate:")) {
// Insert the candidate line after the video m= line.
let cand_replacement = format!("m=video 9 UDP/TLS/RTP/SAVPF 96\r\nc=IN IP4 127.0.0.1\r\n{}\r\n", cand_line);
answer_sdp.replace("m=video 9 UDP/TLS/RTP/SAVPF 96\r\nc=IN IP4 0.0.0.0\r\n", &cand_replacement)
let cand_replacement = format!(
"m=video 9 UDP/TLS/RTP/SAVPF 96\r\nc=IN IP4 {}\r\n{}\r\n",
public_addr.ip(), cand_line
);
answer_sdp.replace(
"m=video 9 UDP/TLS/RTP/SAVPF 96\r\nc=IN IP4 0.0.0.0\r\n",
&cand_replacement,
)
} else {
answer_sdp
};
@@ -267,12 +261,21 @@ pub async fn handle_whip_injest(
.find(|l| l.starts_with("a=ice-pwd:"))
.and_then(|l| l.strip_prefix("a=ice-pwd:"))
.map(|s| s.trim().to_string());
info!(?ufrag, ?ice_pwd, "Whip: registering ICE credentials with proxy");
info!(
?ufrag,
?ice_pwd,
"Whip: registering ICE credentials with proxy"
);
let (socket, rx) = state.appstate.lock().await.webrtc_proxy.add_client(ufrag);
let stream_sessions = state.appstate.lock().await.stream_sessions.clone();
let (video_tx, video_rx) = broadcast::<Arc<VideoFrame>>(4);
let (audio_tx, audio_rx) = broadcast::<Arc<OpusAudioFrame>>(4);
let (mut video_tx, video_rx) = broadcast::<Arc<VideoFrame>>(32);
let (mut audio_tx, audio_rx) = broadcast::<Arc<OpusAudioFrame>>(32);
// Never block the ingest loop on slow/missing viewers: overwrite the
// oldest frame instead (same as the RTMP path; the viewer re-waits for a
// keyframe on Overflowed). The undrained rx below must not stall sends.
video_tx.set_overflow(true);
audio_tx.set_overflow(true);
stream_sessions.insert(
key.id,
@@ -288,7 +291,10 @@ pub async fn handle_whip_injest(
video_profile_level_id: video_profile,
},
);
info!(stream_key_id = key.id, "Whip: StreamSession inserted, spawning detach task");
info!(
stream_key_id = key.id,
"Whip: StreamSession inserted, spawning detach task"
);
// Trickle-ICE channel: OBS can send candidates via PATCH after the
// initial offer. We forward them to the Rtc task for add_remote_candidate.
@@ -326,7 +332,10 @@ pub async fn handle_whip_injest(
.header(header::LOCATION, &location)
// Provide a STUN server via Link header so OBS can gather
// ICE candidates even without explicit STUN configuration.
.header(header::LINK, "<stun:stun.l.google.com:19302>; rel=\"ice-server\"")
.header(
header::LINK,
"<stun:stun.l.google.com:19302>; rel=\"ice-server\"",
)
.body(axum::body::Body::from(answer_sdp))
.unwrap()
}
@@ -430,7 +439,10 @@ async fn detach_inject_rtc(
}
}
Event::Connected => {
info!(stream_key_id, "Whip DTLS+ICE connected, wiring broadcast channels");
info!(
stream_key_id,
"Whip DTLS+ICE connected, wiring broadcast channels"
);
if let Some(session) = sessions_ref.get(&stream_key_id) {
video_tx = Some(session.frame_channel.clone());
audio_tx = Some(session.audio_channel.clone());
crates/server/src/webrtc_proxy.rs
+46 -16
View File
@@ -36,19 +36,31 @@ impl WebrtcProxy {
let sock = UdpSocket::bind(format!("0.0.0.0:{}", config.proxy_port)).await?;
let port = sock.local_addr()?.port();
let public_ip = match env::var("PUBLIC_DOMAIN") {
Ok(domain) => {
let public_ip = match env::var("PUBLIC_DOMAIN")
.ok()
.filter(|s| !s.trim().is_empty())
{
Some(domain) => {
let ip = resolve_domain(&domain).await?;
info!(%domain, %ip, "resolved PUBLIC_DOMAIN for WebRTC candidates");
ip
}
Err(_) => {
None => {
if cfg!(debug_assertions) {
// For testing
info!(
"using ip 0.0.0.0 for WebRTC candidates (because we are in a debug build)"
);
// For testing — advertise a real interface IP. Loopback is
// unreachable from clients whose ICE stack pins its UDP
// sockets to a specific interface (OBS sets
// IP_UNICAST_IF), which makes checks to 127.0.0.1 vanish.
match default_iface_ipv4() {
Some(ip) => {
info!(%ip, "using default interface IP for WebRTC candidates (debug build)");
ip
}
None => {
warn!("no non-loopback IPv4 interface found, falling back to 127.0.0.1");
IpAddr::from([127, 0, 0, 1])
}
}
} else {
let ip = stun_public_ip().await?;
info!(%ip, "discovered public IP via STUN for WebRTC candidates");
@@ -85,7 +97,12 @@ impl WebrtcProxy {
// By addr
if let Some(tx) = by_addr.get(&from) {
debug!("proxy: routing {} bytes by addr {}:{} → channel", b, from.ip(), from.port());
debug!(
"proxy: routing {} bytes by addr {}:{} → channel",
b,
from.ip(),
from.port()
);
match tx.try_send((data, from)) {
Ok(_) => continue,
Err(e) => {
@@ -110,16 +127,22 @@ impl WebrtcProxy {
// might be a response to OUR STUN request (remote:local)
// or an incoming request from the remote peer (local:remote).
let part2_lookup = part2.clone();
let entry = by_ufrag.remove(&part1).or_else(|| {
part2_lookup.and_then(|p2| by_ufrag.remove(&p2))
});
let entry = by_ufrag
.remove(&part1)
.or_else(|| part2_lookup.and_then(|p2| by_ufrag.remove(&p2)));
let Some((_, tx)) = entry else {
warn!("STUN packet ({}/{:?}), isnt registored", part1, part2);
continue;
};
by_addr.insert(from, tx.clone());
info!("proxy: STUN match → promoted {} → ufrag={} (match was {}/{})", from, part1, part1, part2.as_deref().unwrap_or("-"));
info!(
"proxy: STUN match → promoted {} → ufrag={} (match was {}/{})",
from,
part1,
part1,
part2.as_deref().unwrap_or("-")
);
debug!("sending data");
if let Err(e) = tx.try_send((data, from)) {
match e {
@@ -162,10 +185,8 @@ impl WebrtcProxy {
let attr_len: u16 = u16::from_be_bytes(b[pos + 2..pos + 4].try_into().ok()?);
pos += 4;
if attr_type == 0x0006 {
let value = std::str::from_utf8(
b[pos..pos + (attr_len as usize)].try_into().ok()?,
)
.ok()?;
let value =
std::str::from_utf8(b[pos..pos + (attr_len as usize)].try_into().ok()?).ok()?;
let mut parts = value.split(':');
let first = parts.next()?.to_string();
let second = parts.next().map(|s| s.to_string());
@@ -177,6 +198,15 @@ impl WebrtcProxy {
}
}
/// IP of the interface holding the default route, via a UDP connect() trick:
/// connect() only does a route lookup (no packets sent), so the kernel binds
/// the source address the OS would use for outbound traffic.
fn default_iface_ipv4() -> Option<IpAddr> {
let sock = std::net::UdpSocket::bind("0.0.0.0:0").ok()?;
sock.connect("8.8.8.8:9").ok()?;
sock.local_addr().ok().map(|a| a.ip())
}
async fn resolve_domain(domain: &str) -> Result<std::net::IpAddr, Box<dyn Error>> {
let addr = tokio::net::lookup_host(format!("{}:0", domain))
.await?
crates/server/tests/whip_sdp_probe.rs
+118
View File
@@ -0,0 +1,118 @@
//! Temporary probe: replicate handle_whip_injest's Rtc setup, feed it a
//! realistic OBS/libdatachannel WHIP offer, print the answer SDP and check
//! whether the string-fixups in webrtc_ingest.rs actually match.
use std::{net::SocketAddr, time::Instant};
use str0m::{change::SdpOffer, net::Protocol, Candidate, Rtc};
fn obs_like_offer() -> String {
let mut s = String::new();
s.push_str("v=0\r\n");
s.push_str("o=- 4527835755568137757 2 IN IP4 127.0.0.1\r\n");
s.push_str("s=-\r\n");
s.push_str("t=0 0\r\n");
s.push_str("a=group:BUNDLE 0 1\r\n");
s.push_str("a=msid-semantic: WMS *\r\n");
s.push_str("m=audio 9 UDP/TLS/RTP/SAVPF 111\r\n");
s.push_str("c=IN IP4 0.0.0.0\r\n");
s.push_str("a=rtcp:9 IN IP4 0.0.0.0\r\n");
s.push_str("a=ice-ufrag:obs_ufrag_audio\r\n");
s.push_str("a=ice-pwd:obs_pwd_audio\r\n");
s.push_str("a=ice-options:trickle\r\n");
s.push_str("a=fingerprint:sha-256 5A:6B:7C:8D:9E:AF:B0:C1:D2:E3:F4:05:16:27:38:49:5A:6B:7C:8D:9E:AF:B0:C1:D2:E3:F4:05:16:27:38:49:5A\r\n");
s.push_str("a=setup:actpass\r\n");
s.push_str("a=mid:0\r\n");
s.push_str("a=sendrecv\r\n");
s.push_str("a=rtcp-mux\r\n");
s.push_str("a=rtpmap:111 opus/48000/2\r\n");
s.push_str("a=rtcp-fb:111 transport-cc\r\n");
s.push_str("a=fmtp:111 minptime=10;useinbandfec=1\r\n");
s.push_str("m=video 9 UDP/TLS/RTP/SAVPF 96\r\n");
s.push_str("c=IN IP4 0.0.0.0\r\n");
s.push_str("a=rtcp:9 IN IP4 0.0.0.0\r\n");
s.push_str("a=ice-ufrag:obs_ufrag_video\r\n");
s.push_str("a=ice-pwd:obs_pwd_video\r\n");
s.push_str("a=ice-options:trickle\r\n");
s.push_str("a=fingerprint:sha-256 5A:6B:7C:8D:9E:AF:B0:C1:D2:E3:F4:05:16:27:38:49:5A:6B:7C:8D:9E:AF:B0:C1:D2:E3:F4:05:16:27:38:49:5A\r\n");
s.push_str("a=setup:actpass\r\n");
s.push_str("a=mid:1\r\n");
s.push_str("a=sendrecv\r\n");
s.push_str("a=rtcp-mux\r\n");
s.push_str("a=rtpmap:96 H264/90000\r\n");
s.push_str("a=rtcp-fb:96 nack\r\n");
s.push_str("a=rtcp-fb:96 nack pli\r\n");
s.push_str("a=rtcp-fb:96 transport-cc\r\n");
s.push_str(
"a=fmtp:96 level-asymmetry-allowed=1;packetization-mode=1;profile-level-id=42e01f\r\n",
);
s
}
#[test]
fn probe_whip_answer() {
let offer_sdp = obs_like_offer();
let sdp_offer = SdpOffer::from_sdp_string(&offer_sdp).expect("parse offer");
// Mirror handle_whip_injest's builder setup.
let mut builder = Rtc::builder();
{
let cc = builder.codec_config();
cc.clear();
cc.enable_opus(true);
cc.enable_h264(true);
}
let mut rtc = builder.build(Instant::now());
let public_addr: SocketAddr = "203.0.113.7:6969".parse().unwrap();
let candidate = Candidate::host(public_addr, Protocol::Udp).unwrap();
rtc.add_local_candidate(candidate);
let answer = rtc.sdp_api().accept_offer(sdp_offer).expect("accept offer");
let answer_sdp = answer.to_sdp_string();
println!("=== RAW ANSWER ===");
println!("{answer_sdp}");
println!("=== END RAW ANSWER ===");
// Now replicate the fixups from webrtc_ingest.rs verbatim.
let answer_sdp = answer_sdp
.replace("a=ice-options:trickle\r\n", "")
.replace("a=ice-options:trickle\n", "");
let answer_sdp = answer_sdp
.replace("a=group:BUNDLE 0 1\r\n", "")
.replace("a=group:BUNDLE 0 1\n", "");
let answer_sdp =
if let Some(cand_line) = answer_sdp.lines().find(|l| l.starts_with("a=candidate:")) {
let cand_replacement = format!(
"m=video 9 UDP/TLS/RTP/SAVPF 96\r\nc=IN IP4 127.0.0.1\r\n{}\r\n",
cand_line
);
answer_sdp.replace(
"m=video 9 UDP/TLS/RTP/SAVPF 96\r\nc=IN IP4 0.0.0.0\r\n",
&cand_replacement,
)
} else {
answer_sdp
};
println!("=== FIXED ANSWER ===");
println!("{answer_sdp}");
println!("=== END FIXED ANSWER ===");
// Diagnostics
let video_mline = answer_sdp
.lines()
.find(|l| l.starts_with("m=video"))
.unwrap();
println!("video m-line after fixup: {video_mline}");
println!(
"has a=candidate after fixup: {}",
answer_sdp.lines().any(|l| l.starts_with("a=candidate:"))
);
println!(
"a=candidate lines: {:?}",
answer_sdp
.lines()
.filter(|l| l.starts_with("a=candidate:"))
.collect::<Vec<_>>()
);
}