CAM rescanner thread?

[mjacob at freebsd.org]

> The goal was to create sys/cam/scsi/scsi_probe.c and start divorcing the
> SCSI knowledge from the XPT.  Having it be a thread was just a side
> effect.  Unfortunately, I haven't been able to finish that work yet.
> It's getting closer, though.  But seriously, I saw little specific
> benefit to it being a separate thread rather than part of the camisr.
> Scanning/probing doesn't block, so it's not like it's blocking the
> camisr from processing other I/O.  It's just nice from a modularity
> standpoint.

You can do mallocs. It's true that scanning/probing doesn't block, but 
it really has to be started from process context.

I see an advantage in it as a response to initiating a rescan from Fibre 
Channel events. It makes the multipath stuff work usefully. D'ya mind if 
I put it in as an interim?