[T10] Question: About the number of identify
roweber at ieee.org
Mon Feb 29 19:27:57 PST 2016
I am not an SPL expert, but maybe one is not needed here.
Based on SPL-4 r06, subclause 126.96.36.199.1 "State description" provides a
strong hint as to what is expected.
"Upon entry into this state, this state shall send either one or three
Transmit IDENTIFY Address Frame
messages to the SL_IR transmitter.
"NOTE 30 - Phys compliant with SAS-1.1 only transmitted one Transmit
IDENTIFY Address Frame message."
Thus, compliance with SAS-1.1 requires the sending of only one IDENTIFY
Address Frame ...
By implication, SAS-2 and later might have a preference for sending
three IDENTIFY Address Frame ... s ...
but this author cannot find any explicit statement to this effect in a
wide variety of standards than have been investigated (including the
SAS-2 revision in which the "one or three" wording first appeared).
The cited text can be traced back to T10/07-090r3 which was incorporated
in SAS-2 r13 in November 2007.
T10/07-090r3 contains one page of introductory text in which includes
the following sentence: "To better tolerate single-bit errors, the
IDENTIFY address frame should be sent 3 times rather than just 1 time."
Note: The proposal's introduction is not now (and as far as I can tell,
never has been) part of any SAS or SPL standard.
I leave it to the SPL wizards to decide whether valuable information in
T10/07-090r3 is already in SPL-4 or even needs to be in SPL-4.
All the best,
On 2/26/2016 1:35 AM, Guojian (F) wrote:
> We know that identify frame can be sent one or three to the other side
> after speed negotiation completed.
> What I want to know is what is the consideration? Just ensure the
> other side can receive identify?
> Is there any chance that a SAS PHY will miss the first identify frame?
> Now I met a issue. When the host send just one identify frame, the HDD
> will occur receiving identify expired with little probability. This
> issue will be fixed when the host send three identify frame.
> Attached Picture:
> T10 mailing list
> T10 at t10.org
More information about the T10