Selection with ATN in IU mode question
Elliott, Robert (Hou)
Robert.Elliott at COMPAQ.com
Tue Sep 19 12:44:51 PDT 2000
* From the T10 Reflector (t10 at t10.org), posted by:
* "Elliott, Robert (Hou)" <Robert.Elliott at compaq.com>
*
> * From the T10 Reflector (t10 at t10.org), posted by:
> * Ken Erickson <Ken.Erickson at quantum.com>
> *
> What is the expected behavior of a target that has just
> sent a Message Reject message followed by a WDTR w/
> narrow mode to an initiator while in the IU mode? To me
> it appears that since the target is no longer in IU mode
> that the target should go to a BUS FREE phase as it did
> when it entered the IU mode.
This is from SPI-3 section 10.6.2.2, selection with attention
with information units enabled.
I think it is trying to cover what happens when an Ultra 3+
IU-capable initiator is hot-swapped with an non-IU capable
initiator. That initiator will try sending an IDENTIFY
link control message. It is not obligated to follow that with
WDTR/SDTR; it may assume async mode. As soon as the target
detects this initiator discrepancy, the target needs to clear
all existing tasks and switch to non-IU mode. The WDTR/SDTR
negotiation does this. It needs to go BUS FREE for a few
reasons:
1) it rejected the IDENTIFY so has no LUN or tag information.
The initiator might assert ATN and resend, but I doubt it.
2) BUS FREE is customary whenever IU mode changes (or any PPR
is done)
3) BUS FREE is customary whenever an abort task set occurs.
> Also, why was it chosen that if the initiator sent a message
> while selecting a target with ATN and the message is not a
> PPR message or a TASK MANAGEMENT message that the target will
> respond by sending a WDTR. Why not send a PPR message? That
> is how the initiator started the IU process.
Targets are prohibited from initiating PPR because they don't
control domain validation. Imagine an Ultra 3 initiator
and Ultra 3 target connected via an Ultra 2 expander. PPR
negotiation will successfully enable DT despite the fact that
the expander is there, incapable of DT mode. Ultra 3 runs
the P_CRCA signal in the opposite direction of Ultra 2
during writes. After PPR, the subsequent data phase needs
to be a read, or bus contention will result on P_CRCA. If the
command is from an initiator running domain validation, it
will always be a read and work fine. If the command is any
arbitrary command (which it could be if the target is choosing
when to request PPR), it certainly could be a write and
cause trouble.
> Thanks,
> Ken Erickson
> Quantum Corp.
---
PC: Robert.Elliott at compaq.com
UNIX: relliott at unixmail.compaq.com
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org
More information about the T10
mailing list