Flow Control and Read Streaming Proposal

Mark_Heath at notes.seagate.com Mark_Heath at notes.seagate.com
Tue May 2 14:15:08 PDT 2000


* From the T10 Reflector (t10 at t10.org), posted by:
* Mark_Heath at notes.seagate.com
*


Re:

>> In proposal 00-142 you use a mode page bit to indicate if a target is to
>> use read streaming or not. This will not work because, in most cases,
there
>> is only one mode page for all initiators. Because of this if you have
one
>> initiator that supports read streaming and one that does not and the
>> initiator that supports read streaming turnings it on then the target
will
>> use read streaming with all the initiators. So what do you think will
>> happen when the target starts streaming read data to an initiator that
does
>> not support read streaming? I think it will detect a protocol error and
hit
>> the bus with a reset.

>> One way around this would be to not do read streaming but still use the
P0
>> to indicate flow control for reads. When the flow control bit is set the
>> target would assert the P0 line when it knew it would not be sending any
>> more packets for that nexus but it would still send the L_Q between each
>> packet. Initiators that knew this was going to happen could ignore the
L_Q
>> (we would have to make a rule that the target could not change the L_Q
>> information in this mode) and initiators that know nothing about the
mode
>> would continue to operate even though they know nothing about the mode
page

I suggest that read streaming be enabled/disabled via a new option bit in
the PPR message instead of via a mode page bit.  This would allow the read
streaming feature to be conveniently enabled/disabled on a per-initiator
basis.  [My apologies if this idea has already been discussed and
rejected!]

Regards,

Mark A. Heath
Seagate Technology, Inc.

*
* 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