On Sun, Jul 12, 2020 at 12:05:37AM +0700, Robert Elz wrote: > Just to make things clear here, the LUN you're talking about is not > the scsi unit number (which is what I think Martin was referring to) > but a sub-device number within a single scsi ID. Right?
Correct. I should have written "SCSI target" instead of "SCSI bus" in my commit message to avoid confusion. In both of the use cases I highlighted (Proxmox and Linode), the disks are always on SCSI ID 0. However, the "problematic" disks are on LUN 1 instead of LUN 0. Out of curiosity, I just added a "scsi2" disk to the Proxmox VM. It has been placed on LUN 2, and NetBSD does not pick it up even with this patch... FreeBSD does (although the earlier LUNs are clearly causing unexpected output from the "pass" attachments), so I guess I might be looking at this a bit more. It would be nice to have (at least debug) output about the LUN that terminates the scan loop. Will probably open a ticket with Proxmox, too, in an attempt to put a stop to this unnecessary wasteful skipping of perfectly good LUN numbers. + Kimmo