Comment # 27 on bug 1134353 from Andreas Herrmann

Test runs of Paolo's script don't show anomalies on the test system
where I had gathered the original results.
(Maybe test file sizes in this script are too small to cause
a difference.)

I've also rerun io-fio-randread-async-seqwrite-ext4 from mmtests
with 5.2.2 (1) to ensure that performance differences are still
visible and (2) to double-check that each test run was done under
similar test conditions (ie. ensuring reboot between each test)
and (3) to see whether mitigations=auto vs. mitigations=off
has significant influence.

Those tests confirm what's in attachment 809495 [details] for
"(3) io-fio-randread-async-seqwrite-ext4-nosecure".

Mitigations enabled or disabled didn't make a difference
(wrt bfq vs. mq-deadline).

Probably I'll rerun the other fio-randread tests with 5.2.2
also.