Re: [PATCH 2/3] libmultipath: pgcmp(): compare number of paths

[Benjamin Marzinski <bmarzins@redhat.com>]

On Tue, Nov 26, 2024 at 12:15:06PM +0100, Martin Wilck wrote:
> On Mon, 2024-11-25 at 16:06 -0500, Benjamin Marzinski wrote:
> > On Mon, Nov 25, 2024 at 03:32:23PM +0100, Martin Wilck wrote:
> > > pathcmp() makes sure that all paths in pgp have a match in cpgp,
> > > but not
> > > vice-versa. Check the number of paths, too.
> > 
> > This looks fine. But looking at it made we a nervous about cpgp->id. 
> > We
> > only calculate that in pgcmp() (and just for mpp, not cmpp) and in
> > disassemble_map(). But we clearly can have pathgroup changes after
> > that,
> > as the last patch has shown. To be safe, we should either skip the
> > whole
> > pgp->id thing (it's not a huge time savings) or recalculate it for
> > all
> > of cmpp's path groups before we start the loops in pgcmp().
> 
> Agreed. Like before, I am wondering if we should drop pgp->id in 0.11.0
> already or postpone.
> 
> IMO my patch is a valid fix for a minor issue which needs redesign in a
> future release.

Yep. Like I said, your patch looks fine. I just noticed the gpg->id
issue while looking at it.

-Ben

> 
> Martin