Deadlock is a concurrency failure where two or more execution paths each hold a resource the others need, so none can continue. In practice, this often involves locks, mutexes, semaphores, or file and device resources. Because every participant is waiting, the system stops making forward progress instead of simply slowing down.
In cyber security, deadlocks matter because they can become a reliable denial-of-service condition. In kernel code, driver code, or high-traffic services, a bad lock order can freeze a subsystem, stall requests, or hang an entire machine. Defenders look for deadlocks with lock-order audits, stress testing, tracing, and watchdogs that detect stuck threads. Preventing them usually means strict lock ordering, shorter critical sections, avoiding nested locks when possible, and careful review of concurrent code paths.