[LSF/MM/BPF ATTEND][LSF/MM/BPF TOPIC] : blktests: status, expansion plan for the storage stack test framework

[James Bottomley]

On Fri, 2026-02-13 at 09:30 -0800, Bart Van Assche wrote:
> On 2/11/26 11:57 PM, Johannes Thumshirn wrote:
> > One thing that comes to my mind (and that I always wanted to do for
> > fstests but didn't for $REASONS) is adding per-test code coverage
> > information.
> 
> Code coverage information is useful but it's important to keep in
> mind that 100% code coverage (which is very hard to achieve) does not
> guarantee code correctness. There are many state machines in the
> block layer and also in block drivers. Code coverage information does
> not reveal what percentage of the states of state machines has been
> triggered.

This is not an either/or.  Usually our functional tests try to cover
the state machine (although often requiring error injection).  However,
a lot of our bugs hide in error legs and code coverage at least assures
us we've looked for them.

Regards,

James