Comment # 18 on bug 1030310 from Mel Gorman

> Those blank entries for read are somewhat of a reporting bug but occur due
> to no samples being recorded. A manual check verifies. 304 samples with the
> vanilla kernel and 0 with the patches applied.
> 

A rerun with the same patches and configuration captured some samples -- 64
samples and the latency was fine. It's very possible this is entirely a
limitation of the benchmark and it should report sample times more frequently
with more samples captured overall so it still runs for a decent length of
time.