Subject: Re: Another changer, another changer problem
To: None <current-users@netbsd.org>
From: Greg A. Woods <woods@most.weird.com>
List: current-users
Date: 10/06/1998 13:00:54
[ On Tue, October 6, 1998 at 16:02:33 (+0200), Christoph Badura wrote: ]
> Subject: Re: Another changer, another changer problem
>
> woods@weird.com (Greg A. Woods) writes:
> 
> >I want
> >to rename RAW_PART to RAW_DEV.  The raw disk should not be a "partition".
> 
> I guess I must have overlooked it in all the verbiage, but how do you plan
> to deal with devices that are unable to report their size if you don't store
> the size of RAW_{PART,DEV} in the disklabel?

The most critical use of a true "raw" disk is to access the first few
sectors (eg. the first "cylinder"-full, whatever that may mean) of the
device and to send special commands to the disk.  Except for device
types where the driver needs to know how many heads the disk has, and
similar issues, the exact geometry is not necessary.  Where the driver
does need to know this kind of stuff then obviously the driver's own
in-core copy of the label must be at least partly complete.  In some
operating systems the disk (aka volume) label is kept separately from
the partition (aka slice) labels to facilitate this very problem.

As for SCSI disks that are not yet labeled I'd simply rely on the device
to report errors is the user attempts to access the raw disk beyond it's
capacity.  If this doesn't work, then too bad -- there could be a quirk
table to warn the user that data stored or retrieved beyond a certain
limit is suspect (of course for storing the driver could verify the data
again and see if it matches what was stored, which might help a bit).

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>