[compiler-rt] [scudo] Reduce thread contention in secondary cache releases. (PR #102984)

[Christopher Ferris via llvm-commits]

================
@@ -351,9 +351,13 @@ class MapAllocatorCache {
     for (MemMapT &EvictMemMap : EvictionMemMaps)
       unmapCallBack(EvictMemMap);
 
-    if (Interval >= 0) {
+    // If a thread already holds the mutex, the current thread can
----------------
cferris1000 wrote:

I don't think this is strictly true, right? Specifically, in the retrieve function you could hold the lock while trying to find an entry in the cache.

Is there a worry that you could skip these checks while doing a retrieve?

https://github.com/llvm/llvm-project/pull/102984