George.Penokie at lsi.com
Wed Apr 1 06:06:19 PDT 2009
Formatted message: <a href="http://www.t10.org/cgi-bin/ac.pl?t=r&f=r0904010_f.htm">HTML-formatted message</a>
That is an error in SAS-2 that will be fixed in revision 16 per the below
note that went out two weeks ago from Bill Martin.
Rob & T10:
At the SAS protocol meeting today, it was discovered that in the SP11 state
there is an error in 22.214.171.124.3.
In 126.96.36.199.2 where the protocol for speed negotiation windows is defined
there are the following two statements:
In Figure 154 it states -
If the phy's receiver achieves dword synchronization at the SNW rate within
SNLT, its transmitter transmits ALIGN (1)s at the SNW rate for the remainder
of the SNTT.
After table 99 it states -
If the phy supports the SNW, then after RCDT it shall attempt to attain dword
synchronization on an incoming series of dwords (e.g., ALIGN (0) or ALIGN (1)
primitives) at that rate for the SNLT:
a) if the phy achieves dword synchronization within
the SNLT, then it shall change from transmitting ALIGN (0) primitives to
transmitting ALIGN (1) primitives for the remainder of the SNTT (i.e., the
remainder of the SNW time). The point at which the phy achieves dword
synchronization is called the actual lock time (ALT); or
b) if the phy does not achieve dword synchronization
within the SNLT, then it shall continue transmitting ALIGN (0) primitives for
the remainder of the SNTT (i.e., the remainder of the SNW time).
At the end of the SNTT:
a) if the phy is both transmitting and receiving
ALIGN (1) primitives, then it shall consider the SNW to be valid; or
b) if the phy is not both transmitting and receiving
ALIGN (1) primitives, then it shall consider the SNW to be invalid. The phy
shall disable SSC (see 5.7.6) during SNW-1, SNW-2, and Final-SNW.
The first list is describing state SP10 and item a) is the transition from
SP10 to SP11. The second list is describing SP11 and item a) is the
transition from SP11 to SP12. The issue is that the transition from SP11 to
SP12 is taken "if phy is both transmitting and receiving ALIGN (1)
primitives" by the end of SNTT; however 188.8.131.52.3 states:
184.108.40.206.3 Transition SP11:SAS_AwaitALIGN1 to SP12:SAS_AwaitSNW
This transition shall occur if this state receives an ALIGN Received (1)
message before the SNLT timer expires. This indicates that the attached phy
has been able to achieve dword synchronization in the current SNW.
This transition is based on receiving the ALIGN Received (1) message before
the SNLT timer expires not the SNTT timer. It was agreed by those present at
the SAS protocol working group meeting today that this should be changed to
SNTT in the version of SAS-2 that is published for public review.
Bye for now,
3033 41st St. NW
Rochester, MN 55901
george.penokie at lsi.com
From: Craig Stoops [mailto:craig at expertio.com]
Sent: Tuesday, March 31, 2009 7:31 PM
To: t10 at t10.org
Cc: Penokie, George
Subject: 09-063r1 comment
In your posting of 09-063r1 today, section 220.127.116.11 has the timer changed
|from SNLT to SNTT from the original 2008 proposal.
1) Is this in fact correct that the timer is changed
2) If so, there is still a reference to SNLT in 18.104.22.168.3 (a), I assume this
should also be SNTT.
If SNLT in this section is correct, then SNLT is missing initialization
(starting) and any consequent actions on SNLT expiring in the PS states.
Secondly, regarding section 22.214.171.124.3 "footnote" of:
"Receipt of the ALIGN Received (1) message indicates that the connected phy
has been able to achieve dword synchronization with the previously negotiated
1) Does this only apply to the 126.96.36.199.3 transition and not to the 188.8.131.52.3
transition to ALIGN1 state? Ie ALIGN0 state can transition to ALIGN
regardless of DWS sync?
If so, I would like to suggest that maybe just 184.108.40.206.3 "a" should just
"a) If the DWS is synced, and this state receives an ALIGN Received (1)
message before the SNTT timer expires..."
because other places in the phy spec the ALIGN Received message does not
imply the DWS is synced. I think it would be clearer and more consistent.
ExpertIO, Inc. - "Your Storage Protocol Verification Experts"
More information about the T10