RE: iSCSI: using the C bit

[Eddy Quicksall <eddy_quicksall@ivivity.com>]

Ahh, thanks ... I missed that.

 

Eddy

-----Original Message-----
From: Carlos Rimola [mailto:crimola@silverbacksystems.com]
Sent: Thursday, January 16, 2003 8:36 PM
To: Eddy Quicksall; ips@ece.cmu.edu
Subject: Re: iSCSI: using the C bit

At 06:24 PM 1/16/2003 -0500, Eddy Quicksall wrote:

I want to be sure I understand the use of the C bit in a login. It seems to
me that the initiator could split a key into two parts: 1st part at the end
of request A and 2nd part at the beginning of request A+1.
 
e.g., the end of request A (with C bit set) could be "Immediate" and the
beginning of request A+1 could be "Data=yes" (the full key=value pair being
"ImmediateData=yes").
 
If this is done, the target will not process the partial key and may send
his own key saying "ImmediateData=no". But, the initiator has already
constructed the negotiation and request A+1 would violate the protocol
because ImmediateData has already been negotiated.
 
Comments?
 
Eddy

Eddy,

This scenario - the target sending his own key while the Initiator is still sending partial chunks (by setting C Bit = 1) would be a violation of the rule specified in 5.2 (Text Mode Negotiation) which applies during Login as well as Text Req/Resp:

A target receiving a Text or Login Request with the C bit set to 1
MUST answer with a Text or Login Response with no data segment
(DataSegmentLength 0). An initiator receiving a Text or Login
Response with the C bit set to 1 MUST answer with a Text or Login
Request with no data segment (DataSegmentLength 0).


The target can begin sending its own keys only when a PDU with the C Bit set to 0 arrives.  At this point, all keys will be complete and the above condition is avoided.

Regards,

Carlos Rimola
Silverback Systems