|
Message
From: H. Peter Anvin<hpa@z...>
Date: Thu Apr 12 02:21:24 CEST 2007
Subject: [oc] OCIDEC and 48-bit LBA?
Chris Edmiston wrote:
> Has anyone used the OCIDEC core and attempted to use 48-bit LBA
> addressing? The first few pages of the documentation for OCIDEC says
> it's ATAPI-5 compliant, and LBA 48-bit appears to have been introduced
> in ATAPI-6, but from what I can tell it's all a matter of register usage
> and interpretation. If that's the case, I'm thinking I can simply use
> the registers in that mannor without changing the core. Does anyone
> know if this is true?
>
> I'm also curious if OCIDEC-3 is what most people are using for access to
> an IDE HDD, because it looks as if it hasn't been updated in years and
> may never be. Is there something else out there that works better (SATA
> is an option as well)?
First of all, 48-bit LBA applies to ATA, *not* to ATAPI (which is
basically SCSI commands carried over an ATA transport.)
ATA is pretty much designed so that the HBA can be very dumb. All
problems with 48-bit LBA I've seen have been with "smart" cores than do
more than just provide a basic bus interface.
Modulo bugs, I don't see any reason why a current single-port ATA HBA
should ever have to be updated, because ATA is unlikely to progress any
further, having been supplanted by SATA.
It's worth noting that although there are SATA controllers which mimic
ATA controllers, they are *very* different standards, and those
controllers have complex emulation in hardware. Furthermore, native
SATA controllers have much better performance. They do, however,
require very different (and much more complex) driver support.
-hpa
|
 |
Browse
