[ovs-dev] [ovs-dev,v18,4/9] Persist ovn flow tables.

[Ben Pfaff]

On Thu, Jun 23, 2016 at 04:45:40PM -0500, Ryan Moats wrote:
> > struct ovn_flow contains a struct uuid *.  I don't understand how the
> > function can guarantee that the pointer remains valid.  It seems like
> > the prudent thing to do would be to contain a struct uuid instead.
> > (I've noticed the same thing in previous patches; it worried me there
> > too but those seemed like less general cases.)

> If I remember correctly, I asked about the uuid pointer in the ovsdb
> record hanging around and you answered "yes it does."  However,
> I'm happy to
> (a) replace this with a struct uuid and copy/free information
> as needed, and
> (b) craft a patch to do the same with the previous uses.

It does hang around as long as tracking is going on.  It still makes me
nervous--it seems like we're trading 8 bytes of memory for some risk--so
I'd appreciate both (a) and (b).  Thanks.

> Will you be looking at the rest of the refactors or should I just resubmit
> those patches?

I looked at 5-8 and they seem straightforward, I expect to commit them
soon after #4.

I don't plan to review #9 in this round.

Please resubmit 4-9 when you are ready.

I'm glad that we're starting to get this in.