According to Jason Thorpe <thorpej@nas.nasa.gov> :
> 
> ...and Curt's "man, that'd take a long time to reboot" comment, and the
> other comment about how some won't spin up again until a power cycle
> made me think it wasn't such a good idea :-)
> 
> I do have a change to the SCSI disk driver to perform a SYNCHRONIZE CACHE
> on shutdown ... I'm waiting to hear back from Eric to see if it helps his
> particular situation.
> 
Jason, it is easier to suppress existing code with a single kernel 
compile option or, preferably, selectively per drive, than it is to add 
the code to each new release of the kernel.

My vote is for the former.

++L