Re: iSCSI padding should be 0

["Julian Satran" <Julian_Satran@il.ibm.com>]

Is the software padding more expensive than checking if there is a CRC (or
any other digest)?
On the other hand CRCs don't require the bytes to be 0 they can be
arbitrary values.
The 0 was brought in to avoid "leakage" (security) by somebody on the list.
We can choose to revert to arbitrary and leave the burden of cleaning to
the applications for which security is a concern.

Julo

"Eddy Quicksall" <ESQuicksall@hotmail.com> on 27-07-2001 22:05:59

Please respond to "Eddy Quicksall" <ESQuicksall@hotmail.com>

To:   Julian Satran/Haifa/IBM@IBMIL
cc:   "ips" <ips@ece.cmu.edu>
Subject:  Re: iSCSI padding should be 0




I saw one objection to this by Michael Fischer
[Michael_Fischer@adaptec.com]. He pointed out that if there is no CRC then
why require the padding to be 0. I agree with his point.

The problem is with software only implementations ... if they use the
sockets send function and if they are sending from a ULP buffer and if the
data being sent needs padding, they will have to either copy to another
buffer or do an extra tiny send for the pad.

So, my thinking is that we say:

    iSCSI PDUs are padded to an integer number of 4 byte words. If CRC is
being used, the padding MUST be 0. If CRC is not being used, the content of
the padding is unpredictable and irrelevent.

What do you think?

Eddy
----- Original Message -----
From: "Julian_Satran/Haifa/IBM%IBMIL" <julian_satran@il.ibm.com>
To: <eddy_quicksall@ivivity.com>
Cc: "ips" <ips@ece.cmu.edu>
Sent: Friday, July 27, 2001 11:45 AM
Subject: Re: iSCSI padding should be 0


>
> Perhaps we should say MUST be sent as 0 and keep quiet about what the
> receiver should  do (check for 0 - we don't want that).
>
> Thanks,Julo
>
> "Eddy Quicksall" <eddy_quicksall@ivivity.com> on 27-07-2001 18:18:33
>
> Please respond to eddy_quicksall@ivivity.com
>
> To:   Julian Satran/Haifa/IBM@IBMIL, ips@ece.cmu.edu
> cc:
> Subject:  iSCSI padding should be 0
>
>
>
>
> Julian,
>
> Section 2.1 says the padding should be 0. I guess that is correct because
> one may not use CRC and therefore may not want to set them to 0. Wouldn't
> it
> be better if section 2.1 was more specific and mentioned when they must
be
> 0
> if there is a CRC. Also, I noticed at the UNH plug fest that at least one
> person thought "should" meant "must". Therefore, I don't think it should
> say
> "should" ... I think it should not mention the 0'ness unless there is a
CRC
> present.
>
> Also,
>
>         transmission.  Padding bytes, when presents in a segment covered
by
> a
>         CRC, have to be set to 0 and are included in the CRC.
>
> should say "when present in".
>
> Eddy_Quicksall@iVivity.com
>
>
>
>
>