Re: [Sigtran] Abort chunk bundled with data chunk

[Ashwani Kathuria <ashwani.grps at gmail.com>]

Hi Ankit/Salil

As per protocol there should always be a check to validate the SCTP
packet and the validation should be
- to check that DATA chunks if bundled with conntrol chunks are not at
the start of the packet and also
- that DATA chunks should not be bundled with some specific control
chunks like ABORT, SHUTDOWN or SHUTDOWN ACK etc.

This is what protocol behavior should be. Now it's all on
implementation how they take care of performance.

--Ashwani Kathuria

On Thu, Oct 15, 2009 at 3:18 PM, Ankit Kumar Sharma
<ankisharma at gmail.com> wrote:
> Just wandering whether traversing a whole list of chunks after receiving
> every correct packet, is too much of a penalty to handle just one invalid
> case.
>
> If we reply with a SACK for the DATA(received in a packet containing ABORT),
> and then ABORT the association then it should also be ok. If we have
> received a DATA chunk in a packet, and then ABORT in a next packet then also
> we perform the same action.
>
> Just to add to more confusion, section 6.1 says that "The SCTP endpoint MUST
> always acknowledge the reception of each valid DATA chunk when the DATA
> chunk received is inside its receive window". Specification is saying about
> "valid DATA chunk" so is it talking about only syntactically correct DATA
> chunk?
>
> Cheers
>
> On Thu, Oct 15, 2009 at 10:06 AM, Salil Agrawal <Salil.Agrawal at ccpu.com>
> wrote:
>>
>> Thanks Ashwani,
>>
>> That means if we receive any DATA chunk in a packet then we first need
>> to check if this packet contains the ABORT or not before sending the
>> SACK. If packet is not containing the ABORT then only SACK will we sent
>> otherwise not.
>> Also if this is going to be expected behavior then it would impact the
>> protocol performance as we need to traverse the whole packet before
>> sending any SACK.
>>
>> Thanks,
>> Salil
>>
>>
>>
>> -----Original Message-----
>> From: Ashwani Kathuria [mailto:ashwani.grps at gmail.com]
>> Sent: Thursday, October 15, 2009 2:30 PM
>> To: Salil Agrawal
>> Cc: sigtran at ietf.org
>> Subject: Re: [Sigtran] Abort chunk bundled with data chunk
>>
>> No, SACK should not be sent in such case.
>>
>> On Thu, Oct 15, 2009 at 2:28 PM, Salil Agrawal <Salil.Agrawal at ccpu.com>
>> wrote:
>> > Hi Ashwani,
>> >
>> > The question is not that whether DATA chunk should be bundled with the
>> > ABORT or not. RFC is very clear about it that the sender should not do
>> > in that way but here the question is if we at the receiver end
>> receives
>> > the same then what should be the action at the receiver end (should we
>> > send the SACK for the DATA or not).
>> >
>> > Thanks,
>> > Salil
>> >
>> > -----Original Message-----
>> > From: Ashwani Kathuria [mailto:ashwani.grps at gmail.com]
>> > Sent: Thursday, October 15, 2009 2:11 PM
>> > To: Salil Agrawal
>> > Cc: sigtran at ietf.org
>> > Subject: Re: [Sigtran] Abort chunk bundled with data chunk
>> >
>> > Placing DATA chunk before control chunk (ABORT) makes it an Invalid
>> SCTP
>> > packet.
>> >
>> > --Ashwani Kathuria
>> >
>> > On Thu, Oct 15, 2009 at 1:46 PM, Salil Agrawal
>> <Salil.Agrawal at ccpu.com>
>> > wrote:
>> >> Thanks Ashwani,
>> >>
>> >> I understand that it does not make any sense to send the SACK, and
>> > abort
>> >> chunk should be the placed before the DATA chunk. But the question is
>> > if
>> >> the ABORT chunk is placed after the DATA chunk in the received packet
>> >> then should be send the SACK for the DATA or not.
>> >>
>> >> Thanks,
>> >> Salil
>> >>
>> >>
>> >> -----Original Message-----
>> >> From: Ashwani Kathuria [mailto:ashwani.grps at gmail.com]
>> >> Sent: Thursday, October 15, 2009 12:58 PM
>> >> To: Salil Agrawal
>> >> Cc: sigtran at ietf.org
>> >> Subject: Re: [Sigtran] Abort chunk bundled with data chunk
>> >>
>> >> Hi Salil,
>> >>
>> >> When you are bundling anything (control chunk) with DATA chunk it
>> >> should be placed before DATA chunks in the SCTP packets.
>> >> Now if you bundle the ABORT chunk with DATA chunk in a single SCTP
>> >> packet the ABORT chunk will be processed before DATA chunk and
>> >> association will be broken down.
>> >> So, that DATA chunk will have no significance as it's association is
>> >> now broken by the ABORT chunk placed before it.
>> >>
>> >> --Ashwani Kathuria
>> >>
>> >> On Thu, Oct 15, 2009 at 12:37 PM, Salil Agrawal
>> > <Salil.Agrawal at ccpu.com>
>> >> wrote:
>> >>> Hi,
>> >>>
>> >>> We have a confusion while understanding the SCTP RFC4960 section 9.1
>> >> it
>> >>> says "DATA chunks MUST NOT be bundled with ABORT" so sender must not
>> >>> bundle the data with abort.
>> >>>
>> >>> But the second paragraph states that "An endpoint MUST NOT respond
>> to
>> >>> any received packet that contains an ABORT chunk". Which means if
>> any
>> >>> packet is received at the receiver end where ABORT is bundled with
>> > the
>> >>> DATA it should not send any response. Does it mean that for the DATA
>> >>> chunk SACK should not be generated? Please clarify.
>> >>>
>> >>> Thanks,
>> >>> Salil Agrawal
>> >>> _______________________________________________
>> >>> Sigtran mailing list
>> >>> Sigtran at ietf.org
>> >>> https://www.ietf.org/mailman/listinfo/sigtran
>> >>>
>> >>
>> >
>> _______________________________________________
>> Sigtran mailing list
>> Sigtran at ietf.org
>> https://www.ietf.org/mailman/listinfo/sigtran
>
>