On Mon, 2024-11-11 at 12:17 -0500, Benjamin Marzinski wrote:
> On Sat, Nov 09, 2024 at 12:03:33AM +0100, Martin Wilck wrote:
> > On Fri, 2024-11-08 at 17:08 -0500, Benjamin Marzinski wrote:
> > >
> > > Given that the code already does the right thing without needing
> > > any
> > > changes to handle tableless maps makes me feel like just removing
> > > the
> > > final patch is the best idea.
> >
> > If we do this, do we still need the special case for DMP_BAD_DEV?
> > IMO
> > we just need to make sure that dm_get_maps() doesn't return an
> > error if
> > it encounters an empty map.
>
> Depends. If we fail attempting to create a device, we could have
> raced
> with something else attempting to create it. Just checking for
> dm_map_present() before we remove the device won't distinguish
> between
> these two cases at all.
I would like to keep things as simple as possible.
The main mistake that I made in bf3a4ad ("libmultipath: simplify
dm_get_maps()") was that I'd changed dm_get_maps() such that returned
error if anything went wrong for a single map. That was wrong, it
should just skip the map in question (IMO that even includes the
DMP_ERR case, but that's up to discussion). Just changing the behavior
of dm_get_maps() would fix the regressions reported on GitHub (correct
me if I'm wrong).
Wrt libmp_mapinfo:
