From 45fbf91beec0d36905e0a23401ee01dbd8e4e065 Mon Sep 17 00:00:00 2001
From: euzu <33094714+euzu@users.noreply.github.com>
Date: Wed, 28 Jan 2026 12:15:31 +0100
Subject: [PATCH] Feature/provider linup manager deadlock fix (#545)

Identified a critical deadlock path in  reconcile_connections where a DashMap iterator held internal shard locks while awaiting an asynchronous
---
 CHANGELOG.md                                     |  1 +
 backend/src/api/model/provider_lineup_manager.rs | 14 +++++++++-----
 2 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 8539a3317..135c84017 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -78,6 +78,7 @@
 - Added **TMDB** settings (Rate Limit, Cache Duration, Language) and **Metadata Formats** (NFO support) to Library configuration UI.
 
 ## 🚀 Performance & Stability
+- **Deadlock Resolution**: Fixed a potential deadlock in `ProviderLineupManager::reconcile_connections` by refactoring `DashMap` iterations to use snapshots, preventing internal shard locks from being held during async lock acquisition.
 - **Connection Reconciliation & GC**: Resolved a critical issue where provider connection counters could leak or become stale during hot reloads. Added automatic garbage collection for unused provider records to prevent logical memory buildup.
 - **Full Async Runtime**: Transitioned to `#[tokio::main]` and async I/O throughout the entire application.
 - **Non-Blocking Operations**: Cache persistence, playlist exports, and config saves moved to async tasks to prevent runtime stalls.
diff --git a/backend/src/api/model/provider_lineup_manager.rs b/backend/src/api/model/provider_lineup_manager.rs
index a02322105..0476d039b 100644
--- a/backend/src/api/model/provider_lineup_manager.rs
+++ b/backend/src/api/model/provider_lineup_manager.rs
@@ -631,11 +631,15 @@ impl ProviderLineupManager {
 
     pub async fn reconcile_connections(&self, mut counts: HashMap<Arc<str>, usize>) {
         // 1. Synchronize known providers from actual counts.
-        // This avoids a transient "zero state" by updating each provider in one step.
-        for entry in &self.provider_connections {
-            let name = entry.key();
-            let count = counts.remove(name).unwrap_or(0);
-            let mut conn = entry.value().write().await;
+        // We take a snapshot of the keys and locks to avoid holding the DashMap's internal
+        // shard locks while awaiting the RwLock of each provider. Holding both can lead to deadlocks.
+        let snapshot: Vec<_> = self.provider_connections.iter()
+            .map(|e| (e.key().clone(), Arc::clone(e.value())))
+            .collect();
+
+        for (name, conn_lock) in snapshot {
+            let count = counts.remove(&name).unwrap_or(0);
+            let mut conn = conn_lock.write().await;
             conn.current_connections = count;
             if count == 0 {
                 conn.granted_grace = false;