Submit: Section 4.4 describes one technique, namely lock-free comparisons in the dentry cache. Why does this protocol allow for better scalability than the original implementation that uses a spinlock per dentry? Would using a scalable lock instead of a spinlock allow for better scalability too?