Re: Patches for SCSI scanning

[Christoph Hellwig <hch@infradead.org>]

> * Patch 2: Drop REPORT_LUN config option and introduce 
>   noreportlun and reportlun2 parameters (the latter allowing SCSI-2
>   devs to be scanned with REPORT_LUNS).

I'm all for this one.

> * Patch 3: Allow host adapters to avoid REPORT_LUNS.

Alson sounds okay, do you already have drivers that would set it?

> * Patch 4: allow_ghost_devices parameter

I still think this should be a blacklist flag as it's really a spec
violation from EMC if they really don't get it right.  And I still
haven't an answer why they can't work around it by setting those
magic devices online from userland.

> * Patch 5: inq_timeout parameter

Maybe this one should actually be a per-device tunable in sysfs?  Sure,
to actually make use of it for the root device you'd need an initrd but
the distros do that anyway.

> Sidenote: We should drop SCSI_FORCELUN and CONFIG_SCSI_MULTI_LUN, otherwise 
> people will misconfigure their kernel (such as the HP guy) and start to want 
> every multi-lun device listed.

Agreed.