Here's another example of crosslinked entries: ╭────[virtual]─────────────────────────────────────────────────────────────[-]─╮ │╭────────────────────────────────────────────────────────────────────────────╮│ ││Bind devices to the new aggregate. To be eligible, a device must either be ││ ││unpartitioned, or be a partition having the appropriate component type. The ││ ││device furthermore must not have a valid filesystem signature. ││ │╰──────────────────────────╭────────────────────select aggregate components───╮ │ active└┤Linux mdadm │╭ ╮ │ │ md127┌───────────────│ 1 sdo1 pci-0000:04:00.0-scsi-0:0:0:0-part1 │ │ raid6│ ∾∾∾∾∾∾∾∾∾∾∾∾ 1│ 2 sdn pci-0000:04:00.0-scsi-0:0:0:0 │ │ 1-degraded└┤Linux mdadm │ 3 │ │ loop4 n/a │ 4 │ │ loop6 n/a │╰ ╯ │ │ loop7 n/a ╰──────────────────'C' confirms setup, ⎋esc returns╯
so, i rebuilt using both udev 85 (the most recent pure udev release) and udev 189 (the most recent release since the merge with systemd). in both cases, straight SATA devices are not listed in /dev/disk/by-path at all. examining the diff i posted above, apparently the crosslinking had been noticed by udev maintainers, and thus they disabled by-path link generation for all ATA devices(!) as i noted earlier, it seems we can construct unambiguous, meaningful ATA links. do you agree? if so, what would your feelings be on such a patch? i've been running this down by myself, but in order to proceed further, i'd at this point appreciate a reply from the package maintainer. Marco, could you look at this and record your thoughts? thanks so much. --nick