SAS SP to SL_IR handoff

Elliott, Robert (Server Storage) Elliott at hp.com
Mon Mar 10 20:38:27 PST 2003


* From the T10 Reflector (t10 at t10.org), posted by:
* "Elliott, Robert (Server Storage)" <Elliott at hp.com>
*
This is a multi-part message in MIME format.

------_=_NextPart_001_01C2E788.0ED8A041
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable

Maxtor found an issue with the handoff from the phy layer (SP)
completing the speed negotiation sequence to the link layer (SL_IR)
starting the identification sequence.=20

Due to COMSAS detection uncertainty and clock tolerance differences
through the speed negotiation windows, one phy can stll be in its SNTT
(speed negotiation transmit time) when the other phy sends its IDENTIFY
address frame.  As described in sas-r03e and earlier, it will miss the
IDENTIFY since SL_IR has not yet been started.


Two solutions were discussed:=20
1. delay sending IDENTIFY for nn microseconds.  This give the slowest
phy time to finish its SNTT before it has to worry about an IDENTIFY.
We would have to pick a time big enough to solve the problem without
running into the 1 ms IDENTIFY timeout.

2. start the IDENTIFY receiver as soon as ALIGN(1)s are detected during
the last speed negotiation window.  Start the transmitter at the end of
SNTT or when an IDENTIFY is detected.

On Monday, the SAS WG chose #2, which is the more tolerant solution =
(but
slightly harder to document).=20

Changes include:=20
SP12:SAS_AwaitSNW sends a Start_IR_Receiver confirmation to SL_IR_RIF
and SL_IR_IRC. That confirmation replaces Phy Layer Ready going into
those two state machines, enabling them to start watching for IDENTIFY
address frames and HARD_RESETs.

If SP12:Await_SNW or SP13:SAS_Pass receive an Stop SNTT request from
SL_IR_IRC during the final speed negotiation window, they transition to
SP15:SAS_PHY_Ready.  This short circuits the SNTT (the other phy must
have received enough ALIGNs and left its SNTT if an IDENTIFY or
HARD_RESET showed up, so further sending of ALIGN(1)s is unnecessary).
When SP reaches SP15:SAS_PHY_Ready, it still sends Phy Layer Ready =
which
wakes up SL_IR_TIR as currently defined. This takes zero time so the 1
ms IDENTIFY timeout can easily be honored.


Including a short delay before sending might still be a good idea for
conservative design but won't be mentioned.=20

---=20
Rob Elliott, HP Server Storage=20
elliott at hp.com=20


------_=_NextPart_001_01C2E788.0ED8A041
Content-Type: text/html;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">

SAS SP to SL_IR handoff Maxtor found an issue with the handoff = from the phy layer (SP) completing the speed negotiation sequence to = the link layer (SL_IR) starting the identification sequence. = Due to COMSAS detection uncertainty = and clock tolerance differences through the speed negotiation windows, = one phy can stll be in its SNTT (speed negotiation transmit time) when = the other phy sends its IDENTIFY address frame.  As described in = sas-r03e and earlier, it will miss the IDENTIFY since SL_IR has not yet = been started. 
Two solutions were discussed: 
1. delay sending IDENTIFY for nn = microseconds.  This give the slowest phy time to finish its SNTT = before it has to worry about an IDENTIFY.  We would have to pick a = time big enough to solve the problem without running into the 1 ms = IDENTIFY timeout. 2. start the IDENTIFY receiver as soon = as ALIGN(1)s are detected during the last speed negotiation = window.  Start the transmitter at the end of SNTT or when an = IDENTIFY is detected. On Monday, the SAS WG chose #2, which = is the more tolerant solution (but slightly harder to document). Changes include: 
SP12:SAS_AwaitSNW sends a = Start_IR_Receiver confirmation to SL_IR_RIF and SL_IR_IRC. That = confirmation replaces Phy Layer Ready going into those two state = machines, enabling them to start watching for IDENTIFY address frames = and HARD_RESETs. If SP12:Await_SNW or SP13:SAS_Pass = receive an Stop SNTT request from SL_IR_IRC during the final speed = negotiation window, they transition to SP15:SAS_PHY_Ready.  This = short circuits the SNTT (the other phy must have received enough ALIGNs = and left its SNTT if an IDENTIFY or HARD_RESET showed up, so further = sending of ALIGN(1)s is unnecessary).  When SP reaches = SP15:SAS_PHY_Ready, it still sends Phy Layer Ready which wakes up = SL_IR_TIR as currently defined. This takes zero time so the 1 ms = IDENTIFY timeout can easily be honored. 
Including a short delay before sending = might still be a good idea for conservative design but won't be = mentioned. --- 
Rob Elliott, HP Server Storage 
elliott at hp.com 
------_=_NextPart_001_01C2E788.0ED8A041--




More information about the T10 mailing list