Reason About Deadlocks, Synchronization, False Sharing, and Privilege Transitions

## Interview Prompt Explain the four conditions required for deadlock and practical ways to prevent or avoid it. Compare mutexes, condition variables, semaphores, atomics, and spinlocks; identify when a spinlock is appropriate. Then explain false sharing, thread-local storage, the ways execution enters kernel mode, and common causes of a segmentation fault. ### Constraints & Assumptions - Ground each mechanism in a concrete multithreaded systems example. - Distinguish blocking synchronization from busy waiting. - Separate a protection fault from a language-level explanation of undefined behavior. ### Clarifying Questions to Ask - Is the system single-core, multicore, or oversubscribed? - Are interrupt context and user-space locking both in scope? - Does deadlock prevention need to be static, dynamic, or both? ### What a Strong Answer Covers - Mutual exclusion, hold-and-wait, no preemption, and circular wait stated correctly. - Prevention strategies tied to specific conditions, including global lock order and all-at-once acquisition. - Accurate synchronization trade-offs and the narrow conditions where spinning beats sleeping. - Cache-line-level explanation of false sharing and a measured mitigation. - Correct system-call, interrupt, and exception paths plus realistic segmentation-fault causes. ### Follow-up Questions - How would you diagnose a deadlock in a live process? - Why can padding structures reduce false sharing but hurt cache locality? - How does priority inversion change the lock design?