This message is from the T13 list server.

[EMAIL PROTECTED] wrote:
1) Why was is necessary to add, in ATA-8 and not in SATA 1.x or 
ATA/ATAPI-7, the use of ICRC to report errors in PIO data transfer 
commands? If this is needed now, why was it not needed in SATA 1.x or in

ATA/ATAPI-7? SATAIO should publish the proposal for this change/addition

to the SATA specificaton. And SATAIO should publish the meeting minutes 
when this was discussed so that everyone can understand why this 
addition error reporting method is needed now, and why it was not 
previously included in SATA 1.x or ATA/ATAPI-7, and what the risks are 
when using older SATA hosts and devices that do not implement this new 
error reporting feature.



<BMD> To clarify, the original request for updating the T13 material
with the ICRC information specific to this did propose both ATA/8
changes AND errata material for ATA/7 - see doc e05133r0.  As a matter
of review and direction from T13, this proposal was updated to only make
the changes in ATA/8.  The review of this document was done in the June
plenary meeting which has minutes associated with it.  The updates are
included in the latest ATA8-ACS document available from T13.  This
request for changes arose so that appropriate reference could be made in
Serial ATA specifications assuming that ATA included the appropriate
material.  It was felt that this change was relevant to the ATA
documentation, which could be referenced by the SATA.  The appropriate
reference is included in the Serial ATA Revision 2.5 Specification which
is available on the SATA-IO website (http://www.sata-io.org/spec.asp).

Regards,
Brian Dees

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Hale
Landis
Sent: Thursday, December 15, 2005 9:43 PM
To: T13 List Server
Subject: Re: [t13] SATA PIO data-out with Data FIS error

This message is from the T13 list server.


[EMAIL PROTECTED] wrote:
> This message is from the T13 list server.
> Yes, it's all a conspiracy.  Get off your soapbox please.

Answer the questions below and I'll be happy to go away. I don't 
understand why SATAIO is unable to answer basic questions about error 
detection and reporting. It makes me wonder what SATAIO is hidding from 
the public. I don't understand why you should be so defensive when 
someone attempts to talk about possible SATA technical issues. I would 
think a technical discussion of the issues would be best for everyone.

> Where is the data integrity issues?? There is no cover up.

What ever the reasons are, SATA has serious problems, corrupted commands

and/or command parameters that have cause the wrong command to be 
executed by the device with no error indication or the wrong data to be 
read or written by the device with no error indication.

> If there is a CRC error on a command FIS, it is retried.  If it
continues
> to fail, the host obviously knows about it, nothing is 'hidden'.

Then the SATA bus traces I have seen that show a corrupted command FIS 
received by a device with none of the retries you describe really never 
happens?

> If a bus error occurs during data transfer, the device reports a CRC
error
> to the host whether it is DMA, PIO, or whatever.

There is nothing in ATA/ATAPI-7 saying that the ICRC error is used by 
PIO data transfer commands. In ATA/ATAPI-7 ICRC is used only by DMA 
commands. SATAIO needs to explain how a CRC error in a PIO data transfer

was detected and reported to the host before this recent change to the 
definition of ICRC in ATA-8.

> The only issue is the
> last data packet on a PIO transfer.  The device cannot report the
error to
> the host since status goes out before the data.  But, the host still
knows
> about it because the R_ERR is still seen by the host.  Again nothing
is
> hidden.

What if the host doesn't know about SATA devices and it doesn't know to 
look at the R_ERR bit? What if the host controller doesn't make the 
R_ERR available to the host?

***

Here is the most important questions that need to be answered with 
respect to SATA error detection and reporting:

1) Why was is necessary to add, in ATA-8 and not in SATA 1.x or 
ATA/ATAPI-7, the use of ICRC to report errors in PIO data transfer 
commands? If this is needed now, why was it not needed in SATA 1.x or in

ATA/ATAPI-7? SATAIO should publish the proposal for this change/addition

to the SATA specificaton. And SATAIO should publish the meeting minutes 
when this was discussed so that everyone can understand why this 
addition error reporting method is needed now, and why it was not 
previously included in SATA 1.x or ATA/ATAPI-7, and what the risks are 
when using older SATA hosts and devices that do not implement this new 
error reporting feature.

2) Millions of SATA devices have been shipped that use SATA-to-PATA 
bridge devices. How do these SATA-to-PATA bridge devices handle 
corruption of command or data FIS on the SATA interface? How is an error

on the SATA interface reported to the PATA interface side of the bridge 
device? Does the SATA side of the bridge report the error to the host or

does the PATA device report the error to the host? Does any SATA-to-PATA

bridge use the ICRC method of reporting an error on the SATA interface?

3) SATAIO should describe how all types of errors are detected and 
reported in all versions of the SATA specifications and all SATA 
implementations (with and without PATA/SATA bridge devices in the 
interface between host an device). SATAIO should describe how all types 
of errors are detected and reported by SATA host controllers that 
emulate the PATA programming interface. What SATA interface errors are 
not reported to a PATA host that thinks it is working with a PATA host 
and device hardware?

***

> Unlike PATA, SATA has CRC checking on all transfers.  Likewise, errors
are
> detected on both sides of the cable.  Nothing is ever hidden.  PATA
does
> not have this protection.  Especially for PIO transfers.

I would agree that SATA can offer better error detection and reporting 
than PATA. But millions of SATA hosts and devices have been shipped that

fail to conform to the statement "errors are detected on both sides of 
the cable".

> If there are data integrity issues with SATA, it has never, I repeat,
never
> been reported back to device manufacturers.

This is odd, since such errors have been reported to me many times over 
the last 2 years. And I have been inside one manufacturer's test lab and

seen SATA bus traces that show unreported data corruption of SATA 
packets resulting in the wrong data written to the device's media or the

wrong data read from the device's media. And no, I'm not talking about 
data corruption caused by drive firmware errors - these are simple data 
corrutpion issues on the SATA interface cable.

> Please discontinue your 'secret society' comments and slanderous
assault on
> an interface just because you personally were not invited to attend.
These
> types of comments belong on a Blog site, not this forum.

Why do you think I care if I was or was not invited to attend the secret

society meetings? I don't care, I never did care. If I had wanted to 
attend it seems that all I need to do was join the secret society. But 
that would have required me to sign away all my rights to discuss this 
I/O interface - not something I was willing to do.

I happen to think that PATA and SATA data integrity issues are an 
appropriate topic to discuss on the T13 email list. Do you think data 
and system integrity are important topics that should be discussed?

And I'm sorry, but what should we call an organization that holds its 
discussions in secret and under NDA and does not make the minutes of its

meetings public or make most of its documents public? Why do you object 
to calling it a secret society?

Hale

-- 

++ Hale Landis ++ www.ata-atapi.com ++

Reply via email to