[PATCH] Fix segfault when using vgsplit in stacked environment

[Peter Rajnoha <prajnoha@redhat.com>]

Alasdair G Kergon wrote:
> It is meaningless to call pv_uses_vg() if the PV does not exist.

Well, it's some kind of an exception, so I thought it would be
better to check it right here than all the other places outside
the function.

> And the function currently returns 0 if the answer is unknown - it'll
> get even more confusing if it returns 0 for some types of 'unknown'
> and 1 for other types!

Hmm, isn't it the opposite way? Looking at the dev_manager_device_uses_vg
function that is called from pv_uses_vg and returs its value,
dev_manager_device_uses_vg should return 1 if uncertain. So I relied on
this at that time... I return 1 when checking for NULL value of pv->dev,
because we can't decide if PV uses VG in this situation, too.

Peter