On Tue, 2018-01-30 at 12:34 -0600, Benjamin Marzinski wrote: > On Tue, Jan 30, 2018 at 05:51:03PM +0100, Martin Wilck wrote: > > > > It just occured to me while reviewing this patch once more that > > this > > might cause problems if the path WWID changes, as recently > > discussed.Perhaps you should at least skip paths with the > > "wwid_changed" attribute? > > I thought that we agreed that updating pp->udev was a bad idea if the > wwid changed. I'm still not sure that we need to update the udev > device > at all. For things that might reasonably change, we should grab them > from sysfs or from the current uevent (without necessarily updating > the > one attached to the path). For things that aren't supposed to change, > we > can get them through libudev, and then they'll be cached for the > future. > > But at any rate, as long as we don't update pp->udev when the new > udev > device has a different wwid, this shouldn't be a problem. right? Allright, the only thing that matters here is pp->hwe. So yes, we may be ok. I have a rather bad feeling about these "zombie" paths being carried around in multipathd, and parts of the code relying on that being the case, but I guess I can't offer a better solution. Thanks, Martin -- Dr. Martin Wilck <mwilck@xxxxxxxx>, Tel. +49 (0)911 74053 2107 SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton HRB 21284 (AG Nürnberg) -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
