RE: [t13] Queuing

2005-01-28 Thread Mark Overby



There are no restrictions on ordering for completion of 
successful requests. When a drive decides to process and complete a request is 
implementation dependent (andjustly so).

NCQ is the same.


From: [EMAIL PROTECTED] 
[mailto:[EMAIL PROTECTED] On Behalf Of Dennis MooreSent: 
Friday, January 28, 2005 11:29 AMTo: Forum@t13.orgSubject: 
[t13] Queuing

As I recall when overlap/queuing was proposed for 
ATA devices (X3T10-95-258) that the handling of command reordering was left to 
the drive with no restrictions. It was the equivalent of SCSI task attribute 
SIMPLE and mode UNRESTRICTED. I can't find anything in current documentation 
that confirms or denies that. Are drives that support queuing today doing 
unrestricted reordering? Are SATA NCQ devices going to do unrestricted 
reordering? Any help is most appreciated.
Thanks,
[EMAIL PROTECTED]



RE: [t13] Queuing

2005-01-28 Thread Mudama, Eric
This message is from the T13 list server.



All Maxtor SATA/PATA drives operate in an equivalent to restricted
ordered mode, whereby time-coherent data integrity is enforced and
guaranteed.

For example, given this set of queued overlaps with the oldest command
on the bottom:

   r3r3r3
 w2
 w1w1w1w1

In that case, the data returned as part of r3 is guaranteed to look
like:

   w1w2w1

We will never return 'w1w1w1' for r3 in the above circumstance, nor will
we ever return 'w0' or whatever the prior state of those blocks were
prior to the w1 command.

--eric

PS: We know the specification doesn't require this behavior, but this is
how we chose to implement queueing on our desktop products.  I obviously
can't speak to what the other vendors are doing.





 -Original Message-
 From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On 
 Behalf Of Jeff Garzik
 Sent: Friday, January 28, 2005 12:53 PM
 To: Dennis Moore
 Cc: Forum@t13.org
 Subject: Re: [t13] Queuing
 
 
 This message is from the T13 list server.
 
 
 Dennis Moore wrote:
  As I recall when overlap/queuing was proposed for ATA devices 
  (X3T10-95-258) that the handling of command reordering was 
 left to the 
  drive with no restrictions. It was the equivalent of SCSI 
 task attribute 
  SIMPLE and mode UNRESTRICTED. I can't find anything in current 
  documentation that confirms or denies that. Are drives that support 
  queuing today doing unrestricted reordering? Are SATA NCQ 
 devices going 
  to do unrestricted reordering? Any help is most appreciated.
 
 
 In the lack of further information, one must -assume- that 
 the devices 
 are doing unrestricted reordering.
 
 And I think it's likely, given that unrestricted ordering 
 gives the ATA 
 device the most flexibility in optimizing the outstanding ATA 
 commands.
 
   Jeff