George,
It retry data needs to be re-written even
if it has been ACKed, then the SAS standard needs to clearly specify that
because I don’t see anywhere in the standard said the retried data need
to overwrite the old data – I don’t believe this is how some of the
existing implementations are implemented as you described, because it is very
difficult to walk thru the host buffer scatter gather list (usually this is how
host buffers are being implemented) at the retried data offset point when data
frame is retried by the target – it is not impossible, it is just added
complexity for no obvious reason. Don’t think this is a problem for the
target, I believe it is a problem for the initiator receives read data (retried
read data) from the target.
Pak
From:
owner-t10@t10.org [mailto:owner-t10@t10.org] On
Behalf Of Mignon Fernandez
Sent: Wednesday, July 12, 2006
2:16 AM
To: George Penokie
Cc: t10@t10.org
Subject: RE: SAS: transport layer:
ACK lost
Hello Mr. Penokie,
Thank you for your response. Yes, I
was assuming frame level retries are enabled and was referring to handling of
retried data in general (initiator’s or target’s point of view).
Regards,
Mignon
From: George Penokie [mailto:gop@us.ibm.com]
Sent: Thursday, July 06, 2006 2:35
PM
To:
Cc:
Subject: Re: SAS: transport layer:
ACK lost
Mignon,
From
your question it is not clear if you are asking from a targets point of view or
an initiators point of view. It's also not clear if you are asking about what
happens if frame level retries are enabled or disabled.
So
given that lack of knowledge the answer to you question is yes you have to
overwrite data that is retried even if it has been ACKed.
Bye
for now,
George Penokie
E-Mail: gop@us.ibm.com
Internal: 553-5208
External: 507-253-5208
|
" 06/21/2006
08:00 AM |
|
Hello,
If
retries are enabled, and an ACK gets lost, then the transmitter retries, is it
required for the receiver to overwrite the data that is retried even if it has ACK-ed
all data in the previous connection?
Thanks
in advance,