Redundant primitive sequences
Elliott, Robert (Server Storage)
Elliott at hp.com
Tue Feb 28 11:29:31 PST 2006
Formatted message: <A HREF="r0602281_f.htm">HTML-formatted message</A>
The figure is showing when the receiver detects the primitive sequence
under a variety of different scenarios (but not every possible
scenario). If a transmitter sends more than 6, they'll still just be
decoded as part of one primitive sequence.
There are so few redundant primitives - BREAK, BROADCAST, and HARD_RESET
- that some of the details are not important. After HARD_RESET, nothing
else matters. After BREAK, another BREAK has no effect. Identical
BROADCASTs are already subject to being combined by expanders, so
whether you decode 1 or two makes no difference.
If you're testing a transmitter, I'd flag it as a bug if it sent more
than 6 because there is no benefit from doing so and it was probably not
intended by the designer. In a real system, though, receivers won't
Rob Elliott, elliott at hp.com
Hewlett-Packard Industry Standard Server Storage Advanced Technology
From: Yamini [mailto:Yamini at medusalabs.com]
Sent: Tuesday, February 07, 2006 10:40 PM
To: t10 at t10.org; Elliott, Robert (Server Storage)
Subject: Redundant primitive sequences
I have a question about Primitive Sequences. The number of
primitives that a sender can send in a Redundant Primitive Sequence is
not clearly mentioned in the spec. I hope someone from the reflector can
clarify this for me.
'Figure 131 - redundant primitive sequence' in SAS spec
2.0(sas2r00.pdf) shows 10 back to back Primitives in a redundant
sequence. Is this legal?
The reason why ask is that the text preceeding this figure says:
22.214.171.124 Redundant primitive sequence
Primitives that form redundant primitive sequences (e.g.,
BROADCAST (CHANGE)) shall be sent six times consecutively. ALIGNs and
NOTIFYs may be sent inside primitive sequences as described in 126.96.36.199.
A receiver shall detect a redundant primitive sequence after the
identical primitive is received in three
consecutive dwords. After receiving a redundant primitive
sequence, a receiver shall not detect a second instance of the same
redundant primitive sequence until it has received six consecutive
dwords that are not
any of the following:
a) the original primitive; or
b) an ALIGN or NOTIFY."
I interpret it as the sender SHALL only send six consecutive
primitives. So is sending more than 6 primitives in a redundant sequence
an error? Also, is sending sequences without a six dword gap in between
sequences legal? The above section from the spec states clearly about
the receiver's requirements but not the sender's side.
Thanks in advance,
More information about the T10