[lvm-devel] [PATCH 05/13] Add lvm_vg_open().

[Dave Wysochanski]

On Tue, 2009-02-03 at 00:28 +0100, Petr Rockai wrote:
> Dave Wysochanski <dwysocha at redhat.com> writes:
> > RFC - lvm_vg_open().  The new vg_read() function is a step in the right
> > direction but still contains a lot of new options / flags that will
> > be new to people and we can hopefully simplify.  lvm_vg_open() was
> > modelled after 'open' which gives us a potential model that others are
> > familiar with.  Can we use the modes and flags of 'open' in a similar way
> > or do we need to introduce new flags/concepts to people?  For example,
> > can we use a some of the open flags such as O_CREAT and O_EXCL to
> > implement READ_CHECK_EXISTENCE?
> I am currently leaning towards using a completely separate entrypoint for the
> existence check function. It might have been a one bit too much flag abuse,
> that one. Count it as laziness on my part.
> 
> I am not entirely convinced about the open metaphor. For one, the locking
> semantics don't go well with that -- files are racy by default and need to be
> quite carefully locked to prevent awful things from happening. We don't want to
> make that implication. Maybe working a little on the vg_read interface would
> make it acceptable to a newcomer, without causing too much confusion?
>

Well, I'm not convinced of either direction really.  At this point I
think I will continue down the open path and see where I get stuck.
Refining your vg_read() patches is good - both efforts should lead to a
better interface in the long run, whatever we choose.