fix: remove write-lock convoy in getLocal + fix mutateState CAS bug

UniversalCache.getLocal(): when a cached token expires, the RLock fast
path (line 385-398) previously fell through to c.mu.Lock() (write lock).
Under Yaegi, the write-lock holder takes 10-100ms for LRU manipulation,
and Go's RWMutex writer-priority blocks ALL new RLock callers. A single
expired-token event turned every concurrent request from read-parallel
into write-serialized — the convoy that produced the 737-goroutine
pileup at 0x400275a608 (pprof captured at /tmp/traefik-spike-1779663149).

Fix: return (nil, false) immediately on expiry for Token/JWK/Session
cache types. The periodic cleanup goroutine handles eviction. Write lock
is never taken on the read path for these cache types.

refreshAttemptTracker.mutateState(): the CAS loop used
t.state.CompareAndSwap(t.state.Load(), next) — a second Load that can
see a different value from a concurrent writer, silently overwriting
their update. Fixed to CompareAndSwap(cur, next) using the snapshot we
computed the mutation from.
This commit is contained in:
2026-05-25 00:06:47 +01:00
parent 5f9c574f95
commit f821b8829b
2 changed files with 11 additions and 3 deletions
+1 -1
View File
@@ -498,7 +498,7 @@ func (t *refreshAttemptTracker) mutateState(mutate func(cur *attemptState) *atte
if next == nil {
return cur
}
if t.state.CompareAndSwap(t.state.Load(), next) {
if t.state.CompareAndSwap(cur, next) {
return next
}
}
+10 -2
View File
@@ -396,8 +396,16 @@ func (c *UniversalCache) getLocal(key string) (interface{}, bool) {
return value, true
}
c.mu.RUnlock()
// Expired — fall through to the write-locked slow path below to
// remove the entry under exclusive access.
// Expired — return miss immediately. The periodic cleanup goroutine
// will evict the stale entry. NEVER fall through to the write-locked
// slow path for Token/JWK/Session caches: under Yaegi the write Lock
// at line 403 costs 10-100ms per acquisition, and Go's RWMutex
// writer-priority semantics block ALL new RLock callers while a Lock
// is pending. A single expired-token event turns every concurrent
// request from read-parallel into write-serialized — the exact
// convoy that produced the 737-goroutine pileup at 0x400275a608.
atomic.AddInt64(&c.misses, 1)
return nil, false
}
c.mu.Lock()