>> [Ken] It appears the current Linux iSER initiator does not send the HELLO
>> message when
>> the connection transits to full feature phase. The stgt target also ignores
>> this
>> message (if it were to appear).
[Ken] The IBTA document does not mention the HELLO/HELLOREPLY messages.
Implementing this message exchange gives a distinction between the current
implementations
and those that will correctly calculate the write_va (as per Pete Wyckoff's
option 3).
>> [Ken] Both of these implementations use a non-conformant iSER header (they
>> add
>> write_va and read_va fields, which incidentally do not appear to be used).
>> Are
>> these changes documented anywhere in the IB domain, or are these variations
>> needed for another reason?
>
> [Erez] Take a look at the iSER for IB annex:
> http://www.infinibandta.org/members/spec/Annex_iSER.PDF
[Ken] Ouch. That link requires a username/password. Looks like it is only
available to members
of the Infiniband Trade Association. Fortunately I gained access to it with
username "open" and
password "standard". ;-)
[Ken] Neither of these implementations send or examine the iSER CM REQ/REP
message private
data. The document doesn't define what action to take when this message is
absent. Interestingly,
when the target reports that "ZBVA shall be used for this connection" and "the
target shall issue
Send with Invalidate as needed" then it appears the iSER header specified in
RFC5046 should be
used for control-type PDUs. Is there any plan to conform with the list of
requirements for IBTA
compliance?
Cheers,
Ken
_________________________________________________________________
Never miss another e-mail with Hotmail on your mobile.
http://www.livelife.ninemsn.com.au/article.aspx?id=343869
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
