Number of target mappings returned by the SNIA library

[PrasannaKumari]

Hi,

I have a query regarding the number of target mappings returned by the SNIA library. What I have noticed is that the number of target mappings returned by the SNIA library is more than the (number of LUNs presented to the host * number of discovered ports through which the LUNs can be accessed) . I have noticed that few target mapping entries are getting created with a fcp lun value of zero for each discovered port when there is actually no LUN for that target mapping. Any reason for this?

Thanks,

Prasanna

[Matti_Kurkela]

In SCSI, all SCSI targets must have LUN 0, and may have other LUNs. When sending a SCSI Inquiry operation to an unknown device, the standard procedure is to first inquire LUN 0, then examine the results to see if it has any other LUNs, then probe the other LUNs only if they exist.

When FibreChannel is used as a transport for the SCSI protocol, I guess this behaviour maps to FC ports and their LUNs.

It might also be that the SNIA library simply presents SCSI-compatible semantics: if a port is discoverable but has no LUN 0, the SNIA library might internally handle it in FC-specific way, but present a "fake" LUN 0 that only exists  to satisfy any SCSI-style inquiries directed to LUN 0... including the list of LUNs this port actually has available.

By the way: "the SNIA library" is ambiguous. At least some Qlogic FC driver versions include Qlogic's own SNIA library. You might want to specify the name and version of the Linux distribution, the type of the FC HBA, and the name and version of the FC driver package (if other than the standard driver included in the distribution) to get more accurate answers.