This message is from the T13 list server.
Hale,
There is NO conspiracy on PIO CRCs.... It is just as simple as this:
1) SATA PIO transfers have a capability that PATA PIO transfers do not,
they have a CRC that is part of the transfer, and
2) so CRC errors can be detected (notice the word detected)
3) to take advantage of this "improvement" a reporting mechanism was
needed, thus the proposal of using iCRC on PIO transfers.
It was NOT that SATA needed it, it was that SATA could improve over
PATA.
Re: SATA 1.0 spec
So does T13 write errata for ATA-1 for all the changes that are going
into ATA-8, NO, its call spec evolution... same for SATA 2.5
Re: Host compatibility
Yes, you will need drivers that know about this new feature to take
advantage of it.
This is true for EVERY new feature that comes about.
Re: PIO error rate
The bit error rate for PIO transfers is exactly the same as the other
transfers as you are using the same physical transport layer... no more,
no less.
Jeff
Jeff Wolford Email: [EMAIL PROTECTED]
Master Architect
Storage Interface and Tools - Business PC Group
Voice: (281) 514-9465, Pager: (800) 973-5739
Hewlett-Packard Corporation
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
> Behalf Of Hale Landis
> Sent: Saturday, December 17, 2005 10:32 AM
> To: T13 List Server
> Subject: Re: [t13] SATA PIO data-out with Data FIS error
>
> This message is from the T13 list server.
>
>
> Dees, Brian M wrote:
> > This message is from the T13 list server.
>
> Thanks again for the information... but...
>
> > a) why not in original SATA or ATA/ATAPI-7
> >
> > Specifically for Serial ATA, the SATA Revision 2.5
> specification is a
> > single integration of Serial ATA 1.0a in addition to much of the
> > Serial ATA II material, including errata and design guides.
> The ICRC
> > update was simply applied to SATA Revision 2.5 with the
> understanding
> > that this new specification would replace the older set of
> > documentation, which leads to little need to update the old
> > documentation when it is getting replaced.
>
> This new use of ICRC is a sigificant change in SATA error
> reporting. If it is needed now (in SATA 2.5) why was it not
> needed in prior versions of the SATA spec (or in
> ATA/ATAPI-7)? What is the justification for adding it now if
> it was apparently not needed before?
>
> > b) the risk of using SATA that does not have this feature,
> > c) how are SATA CRC errors report by original SATA devices that did
> > not implement this feature?
> >
> > The change was specific to allowing ICRC usage for
> reporting interface
> > CRC errors with Multi-word DMA and PIO protocols, previously it was
> > only allowed with the Ultra DMA protocol. Since this
> behavior was not
> > allowed prior to the change for either PATA or SATA, the impact of
> > reporting interface CRC errors with these other protocols
> should not
> > be specific to the transport.
>
> I'm still trying to understand how the reporting of SATA CRC
> errors was done prior to this change in the use of ICRC. If
> the original methods of reporting SATA CRC errors was
> adequate then why this change to ICRC now?
>
> And I really want to know what risk there is when using
> original SATA that does not include this new use of ICRC to
> report SATA CRC errors.
>
> It would really be nice if SATAIO would publish the original
> proposal that caused this change to the SATA 2.x spec and
> also publish the meeting minutes where this change was
> discussed and approved so that everyone can why this change
> was needed and what problem(s) it is solving and perhaps give
> some insight into the risks to data integrity that exist
> prior to implementation of this change.
>
> Are there millions of SATA devices in use that do not
> implement this use of ICRC to report SATA interface CRC
> errors - do those devices represent a risk (no matter how
> small) to the data stored and accessed on those devices?
>
> As I've said before, many people have seen data corruption
> problems with SATA that could be directly related to a
> failure of the device or host to detect and report SATA CRC
> errors to the host software. This change to ICRC would appear
> to be an attempt to fix such problems in future SATA
> implementions, is that what it is?
>
> Hale
>
> --
>
> ++ Hale Landis ++ [EMAIL PROTECTED] ++
>
>
