dumbgoy/soulseek-rs-modded
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

fix: survive busy sessions without exhausting file descriptors

Browse Source 
The peer listener died with "Too many open files (os error 24)" on any
long-lived non-daemon session: only the daemon raised RLIMIT_NOFILE, and
peer sockets were unbounded — no idle reaping, no deadlines on accepted
transfer streams, and one peer could hold two connections under split
registry identities.

- raise the open-file limit at the top of run(), for every entry point,
  and warn from Client::connect when the soft limit is below 1024
- reap peer control connections idle for five minutes; a clean
  disconnect no longer discards the peer's queued uploads
- give listener-supplied download streams and upload serving the same
  read/write deadlines dialled sockets already had
- register inbound peers under their plain username so the registry
  holds one connection per peer, and announce them like pierced ones
- hold the registry lock across actor spawn + insert and surface
  thread-spawn failure, so an instantly-dead peer cannot leave a
  zombie entry
This commit is contained in:
Michel de Graaf
2026-08-09 22:34:36 +02:00
parent d11326f149
commit 0bbb559f68
17 changed files with 442 additions and 132 deletions
Download Patch File Download Diff File Expand all files Collapse all files
soulseek-rs-lib/src/client/connection.rs
+33 -1
View File
@@ -2,7 +2,7 @@ use super::{
Arc, Client, ClientContext, ClientOperation, ConnectionType, DownloadPeer,
DownloadStatus, Listen, Peer, PeerRegistry, Receiver, Result, RwLock,
RwLockExt, Sender, ServerActor, ServerMessage, Shares, SoulseekRs,
TcpStream, error, info, mpsc, thread, trace,
TcpStream, error, info, mpsc, thread, trace, warn,
};
/// Ceiling on the wait for a login verdict. Generous enough for a slow server
@@ -10,8 +10,40 @@ use super::{
const LOGIN_RESPONSE_TIMEOUT: std::time::Duration =
std::time::Duration::from_secs(45);
/// Soft open-file limit below which a session should expect "too many open
/// files": every peer holds a socket, and one busy search talks to hundreds
/// of them.
const COMFORTABLE_OPEN_FILES: u64 = 1024;
/// Say up front when the process cannot hold a busy session's sockets, so an
/// operator reads a warning with a number in it instead of an EMFILE storm an
/// hour later. The embedding process owns its rlimits; the library only looks.
#[cfg(unix)]
fn warn_if_open_file_limit_is_low() {
// SAFETY: getrlimit only fills the struct it is handed.
let limit = unsafe {
let mut limit = std::mem::zeroed::<libc::rlimit>();
if libc::getrlimit(libc::RLIMIT_NOFILE, &raw mut limit) != 0 {
return;
}
limit.rlim_cur
};
if limit < COMFORTABLE_OPEN_FILES {
warn!(
"only {limit} open files allowed; a busy session wants \
{COMFORTABLE_OPEN_FILES} or more — expect \"too many open \
files\" under load"
);
}
}
#[cfg(not(unix))]
fn warn_if_open_file_limit_is_low() {}
impl Client {
pub fn connect(&mut self) -> Result<()> {
warn_if_open_file_limit_is_low();
let (sender, message_reader): (
Sender<ClientOperation>,
Receiver<ClientOperation>,
soulseek-rs-lib/src/client/mod.rs
+1 -1
View File
@@ -248,7 +248,7 @@ pub enum ClientOperation {
query: String,
},
/// A peer queued one of our shared files; `requester_key` is the registry
/// key of the peer actor (may carry a `:direct` suffix).
/// key of the peer actor — the peer's username.
QueueUpload {
requester_key: String,
filename: String,
soulseek-rs-lib/src/client/operations.rs
+18 -15
View File
@@ -1,4 +1,3 @@
use super::uploads::downloader_of;
use super::{
Arc, BROKER_CONNECT_TIMEOUT, Client, ClientContext, ClientOperation,
ConnectionType, Download, DownloadPeer, DownloadStatus, Peer, PeerMessage,
@@ -75,6 +74,14 @@ impl Client {
let _ = handle.stop();
}
}
// Only an error is evidence the peer is gone. A clean
// close — our idle reaper, or a remote client tidying
// an idle socket while it waits in our queue — must
// not throw away everything that peer has queued: the
// connection comes back, the queue cannot. A departed
// peer wedging uploads shut is still covered: its
// erroring transfer fails within the socket deadlines
// and the error branch here frees its slots.
if let Some(error) = error {
warn!(
"[client] Peer {} disconnected with error: {:?}",
@@ -85,12 +92,11 @@ impl Client {
&username,
None,
);
Self::release_upload_slots(
&client_context,
&username,
);
}
// A peer that leaves must not keep occupying an
// upload slot: with a slot cap, a couple of
// departed peers would otherwise wedge uploads
// shut for the rest of the session.
Self::release_upload_slots(&client_context, &username);
}
ClientOperation::DownloadFromPeer(token, peer, allowed) => {
let maybe_download = match client_context.read_safe() {
@@ -440,10 +446,11 @@ impl Client {
}
}
ClientOperation::PeerConnected(username) => {
// An outbound control connection just handshook.
// Flush any downloads that were queued for this
// peer while we were still connecting. Collect
// under a read guard, then act without it held.
// A control connection just came up — one we dialled,
// or an inbound one the listener registered. Flush any
// downloads that were queued for this peer while it
// was unreachable. Collect under a read guard, then
// act without it held.
let (registry, files): (
Option<PeerRegistry>,
Vec<String>,
@@ -547,8 +554,6 @@ impl Client {
filename,
} => {
// The peer served next may not be this one.
let downloader =
downloader_of(&requester_key).to_string();
match client_context.write_safe() {
Ok(mut ctx) => {
let Some(file) = ctx.shares.get(&filename)
@@ -563,7 +568,6 @@ impl Client {
let real_path = file.real_path.clone();
ctx.enqueue_upload(
&requester_key,
&downloader,
&filename,
real_path,
size,
@@ -580,12 +584,11 @@ impl Client {
requester_key,
filename,
} => {
let downloader = downloader_of(&requester_key);
let (registry, place) = match client_context.read_safe()
{
Ok(ctx) => (
ctx.peer_registry.clone(),
ctx.place_in_queue(downloader, &filename),
ctx.place_in_queue(&requester_key, &filename),
),
Err(_) => continue,
};
soulseek-rs-lib/src/client/tests.rs
+68
View File
@@ -327,3 +327,71 @@ fn upload_speed_is_reported_only_while_running() {
"unexpected rate {rate}"
);
}
// A clean disconnect — the idle reaper, or a remote client tidying an idle
// socket while it waits in our queue — must keep the peer's queued uploads;
// only an error disconnect is evidence the peer is gone.
#[test]
fn a_clean_disconnect_keeps_queued_uploads_an_error_drops_them() {
let client = Client::new("test-user", "test-password");
let (ops, ops_rx) = mpsc::channel();
Client::listen_to_client_operations(
ops_rx,
client.context.clone(),
"me".to_string(),
);
client.context.write().unwrap().enqueue_upload(
"amy",
"@@share\\f.mp3",
std::path::PathBuf::from("/tmp/f.mp3"),
4096,
);
ops.send(ClientOperation::PeerDisconnected(
1,
"amy".to_string(),
None,
))
.unwrap();
// The loop is serial, so once this fence op is visible the disconnect
// before it has been handled.
ops.send(ClientOperation::OwnPrivileges(7)).unwrap();
let deadline = Instant::now() + Duration::from_secs(5);
while client.own_privilege_seconds() != Some(7) {
assert!(Instant::now() < deadline, "ops loop never caught up");
thread::sleep(Duration::from_millis(10));
}
assert_eq!(
client
.context
.read()
.unwrap()
.place_in_queue("amy", "@@share\\f.mp3"),
Some(1),
"a clean disconnect must keep the queue"
);
ops.send(ClientOperation::PeerDisconnected(
1,
"amy".to_string(),
Some(crate::error::SoulseekRs::NotConnected),
))
.unwrap();
let deadline = Instant::now() + Duration::from_secs(5);
loop {
let place = client
.context
.read()
.unwrap()
.place_in_queue("amy", "@@share\\f.mp3");
if place.is_none() {
break;
}
assert!(
Instant::now() < deadline,
"an error disconnect must drop the queue"
);
thread::sleep(Duration::from_millis(10));
}
}
soulseek-rs-lib/src/client/upload_queue.rs
+3 -8
View File
@@ -31,9 +31,7 @@ pub struct UploadOffer {
/// A file a peer has asked us for and is waiting on.
#[derive(Debug, Clone)]
pub struct QueuedUpload {
/// Registry key of the asking peer's actor, which may carry a `:direct`
/// suffix — not the same string as `downloader`.
pub requester_key: String,
/// The asking peer's username — also the registry key of its actor.
pub downloader: String,
pub virtual_path: String,
pub real_path: PathBuf,
@@ -98,7 +96,6 @@ impl ClientContext {
/// for a file we do not have never reaches the queue.
pub fn enqueue_upload(
&mut self,
requester_key: &str,
downloader: &str,
filename: &str,
real_path: std::path::PathBuf,
@@ -106,7 +103,6 @@ impl ClientContext {
) {
self.upload_seq += 1;
self.upload_queue.push(QueuedUpload {
requester_key: requester_key.to_string(),
downloader: downloader.to_string(),
virtual_path: filename.to_string(),
real_path,
@@ -134,7 +130,7 @@ impl ClientContext {
let job = self.upload_queue.remove(index);
let token = next_token();
offers.push(UploadOffer {
requester_key: job.requester_key,
requester_key: job.downloader.clone(),
token,
virtual_path: job.virtual_path.clone(),
size: job.size,
@@ -259,7 +255,6 @@ mod tests {
fn queued(user: &str, privileged: bool, seq: u64) -> QueuedUpload {
QueuedUpload {
requester_key: user.to_string(),
downloader: user.to_string(),
virtual_path: format!("@@share\\{user}.mp3"),
real_path: PathBuf::from("/tmp/x.mp3"),
@@ -365,7 +360,7 @@ mod context_tests {
}
fn ask(ctx: &mut ClientContext, user: &str, file: &str) {
ctx.enqueue_upload(user, user, file, PathBuf::from("/tmp/x"), 4096);
ctx.enqueue_upload(user, file, PathBuf::from("/tmp/x"), 4096);
}
/// Who the pump offered to, by the order it offered them.
soulseek-rs-lib/src/client/uploads.rs
-8
View File
@@ -5,14 +5,6 @@ use super::{
use crate::types::UploadStatus;
use std::sync::atomic::{AtomicBool, AtomicU64};
/// The peer behind a registry key, which may carry a `:direct` suffix the
/// server never sees.
pub fn downloader_of(requester_key: &str) -> &str {
requester_key
.strip_suffix(":direct")
.unwrap_or(requester_key)
}
impl Client {
/// Whether the server listed `username` as privileged. Privileged peers are
/// served before everyone else.