[dm-devel] [PATCH 3/3] libmultipath: add ignore_udev_uid option

[Benjamin Marzinski]

On Wed, Sep 16, 2020 at 02:46:18PM +0000, Martin Wilck wrote:
> Hi Ben,
> 
> On Tue, 2020-09-15 at 16:45 -0500, Benjamin Marzinski wrote:
> > Setting this option to yes will force multipath to get the uid by
> > using
> > the fallback sysfs methods, instead of getting it from udev. This
> > will
> > cause devices that can't get their uid from the standard locations to
> > not get a uid. It will also disable uevent merging.
> > 
> > It will not stop uevents from being resent for device that failed to
> > get a WWID, although I'm on the fence about the benefit of this.
> > 
> > Signed-off-by: Benjamin Marzinski <bmarzins at redhat.com>
> 
> Can you explain how this differs from setting uid_attribute to the
> empty string (and leaving uid_attrs at the default, empty)?
> 
> The patch is alright, but the configuration of WWID determination is
> already sooo complicated... I'm not too happy about adding yet another
> option which complicates matters further. IMO we should rather attempt
> to make this easier for users (meaning less options, less combinations
> thereof, and less "x supersedes y but only if z is not set" kind of
> logic). This is not a nack, I just want to understand.

Ummm... Actually, I hadn't thought of using

overrides {
	uid_attribute ""
}

to test SID.  That works, so this patch is unnecessary. I hope you don't
object to me changing the code to log at level 3 if uid_attribute is
configured to "". In this case, using the fallback methods is
expected, so I don't see a need to log at a higher priority.

-Ben
 
> Regards
> Martin
> 
> --
> Dr. Martin Wilck <mwilck at suse.com>, Tel. +49 (0)911 74053 2107
> SUSE  Software Solutions Germany GmbH
> HRB 36809, AG Nürnberg GF: Felix
> Imendörffer
>