lkml.org 
[lkml]   [1998]   [Aug]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: sector size of 2068?
On Thu, 27 Aug 1998, Brandon S. Allbery KF8NH wrote:

> In message <Pine.LNX.3.95.980827132719.204B-100000@chaos.analogic.com>,
> "Richard B. Johnson" writes:
> +-----
> | On Mon, 24 Aug 1998, Peter wrote:
> | > What would need to be changed to make a SCSI disk with sector_size
> | > reporting back as 2068 work?
> |
> | In the SCSI controller BIOS setup, low-level format the drive using
> | your new controller. The drive was apparently low-level formatted on
> +--->8
>
> We've been through that already. His question is, why can't we use it with
> 2068? (The reason he asked being that more space is "wasted" with 512-byte
> sectors due to sector mark overhead.)

Not really, but its nice to think someone believes I'd come up with
something like that. The reason I asked was because its on a sparcstation
and there's no low-level format, and thus is unusable.

Asside from that, the answer was to put together a 486 and have adaptec's
controller have fun with it. Works fine, but i'm a little bit perplexed
as to its result. It being a "9.1 Gig scsi drive". The drive is a
ST410800W in case you're wondering, and the full details are at
(http://www.seagate.com/support/disc/specs/st410800w.shtml).

What interests me about the resulting behavior is what fdisk says:
Device Flag Begin Start End Blocks Id System
/dev/sdb1 0 0 4923 8839246+ 83 Linux native
/dev/sdb3 0 0 4923 8839246+ 5 Whole disk

Is 8.43 gig really how much is left when you partition off a 9.09 gig
drive? That just doesn't seem sane. Somebody tell me if I'm crazy.

> The short answer is that most of the kernel assumes that disk sector sizes
> are a power of 2 and therefore can be represented via bit shifting 2068
> isn't a power of 2, so isn't eligible. 2048 might well be doable, though,
> if you really wanted it....
>
> | optimization to select the proper head, etc. Therefore I doubt
> | that the "real" block-size is 2068. In practice, it's usually
> | 512 or a multiple thereof.
> +--->8
>
> It has been suggested that they chose a format that makes the disk look as
> big as possible. That it's not *usable* in that format (to my knowledge, no
> other OS does 2068-byte sectors either) isn't significant to marketroids
> pushed to "demonstrate" their numbers....

This sounds reasonable, and probably very likely.

-- Peter

"What is a magician but a practicing theorist?"
-- Kenobi




-
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.altern.org/andrebalsa/doc/lkml-faq.html

\
 
 \ /
  Last update: 2005-03-22 13:44    [from the cache]
©2003-2011 Jasper Spaans