Re: [PATCH] Radix-tree pagecache for 2.5

[Alan Cox]

> Mathematically the hashtable complexity is O(N). But probabilistically
> with the tuning we do on the hashtable size, the collisions will be
> nearly zero for most buckets for of most workloads. Despite the worst
> case is with all the pagecache and swapcache queued in a single linked
> list :).

Providing it handles the worst case. Some of the hash table inputs appear
to be user controllable so an end user can set out to get worst case 
behaviour 8(