RE: The limitation of scsi target and lun supported under Linux?

Matthew Jacob (mjacob@feral.com)
Tue, 28 Sep 1999 10:52:20 -0700 (PDT)


On Tue, 28 Sep 1999, WANG,YIDING (HP-SanJose,ex1) wrote:

> The scsi support depends on two sources. One from OS target driver
> and another from spec. Even wide scsi support 16 targets, most Unix
> OSes support 32 and more. For example, Solaris support 128 target and

Wrong. 65535.

> 32 lun each.

Wrong. 256 (I've done this much with Fibre Channel).

Both these are driven by the bits available in the struct scsi_address
structure.

Further, the NDI framework which everything inside Solaris is shifting to
completely separates addresses out from limitations as the dev_info_t is
constructed by the nexus driver and the address can be arbitrary. In
theory this means that even without changing the target drivers the
a_target, a_lun and a_sublun fields could be recycled to be a flat 32 bit
integer which means you get 4GB seperate addresses *per HBA instance*.

>
> I am doing Fibre Channel device support which using SCSI III protocal.
> Currently many
> vendor requires large number device support. For example, EMC asks at least
> 256 LUN is
> supported.

Again, this is trivial. Try doing at least first level SCCLUN with 65535
luns.

> for scsi target support and max. 8 LUN is supported, still I have question
> of the limitation of target and lun current Linux can support? This is due
> to my feeling that there should be a limitation for target support and 8 lun
> support is too few. Also Benjamin mentioned 127 lun supoort although I
> don't know where the number comes from.

The problem is more serious in what you can dig up for device nodes than
actual internal scsi addressing.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/