Re: Query regarding SCTP ordering queue handling.

[sandeep kandula <[e-mail]>]

Hi Randall Stewart
please find answers in line.

On Tue, Nov 23, 2010 at 5:58 PM, Randy Stewart <[e-mail]> wrote:

> Michael:
> I agree with your assessment .. but I do have a query for Sandeep (I assume
> this is how you interpreted his request but it is not clear)...
>
>
> On Nov 22, 2010, at 11:48 PM, Michael Tüxen wrote:
>
> > On Nov 23, 2010, at 8:11 AM, sandeep kandula wrote:
> >> We have scenario as explained below, please let me know how SCTP has to
> handle this situation.
> >>
> >>      • 1.       There is an association established between two nodes
> (say Node A and Node B) and traffic is flowing between these nodes.
> >>      • 2.       Node A sent a data chunk with TSN 25 to Node B. Even
> though this is complete chunk, due to error at sending side, the B-bit was
> sent while sending before the checksum is calculated.
>
> So you would normally have the B and E bit set, so the E bit is NOT set
> would be how I would guess
> you mean?
>
> [SANDEEP] : This was done for testing purpose.
>


> >>      • 3.       On reception of chunk (TSN 25) at Node B, it will be
> keep this chunk in ordering queue as it is expecting the other fragments and
> will not provide it to SCTP user. Node B will acknowledge the TSN 25 with a
> SACK.
> >>      • 4.       As B-bit was set due to error at Node-A, it will not
> transmit any other fragment of the packets to Node-B.
> >>      • 5.       Node-B will be waiting for infinitely for other
> fragments of the packets and it will be able to deliver it user.
> >> Is this the expected behavior of the SCTP at Node-B? If not, can anyone
> please tell what shall be correct behavior of the SCTP at Node-B? How SCTP
> can overcome from such situations?
> > Node B is behaving correctly. To overcome this situation, fix Node A.
> >
> > Please note, that the situation might change if Node A sends another
> message.
> > Then Node B might be able to figure out that Node A is violating the
> protocol
> > specification and might (depending on the implementation) send an ABORT.
> It
> > might include an error cause indicating a protocol violation.
>
> I know that the BSD implementation would send an ABORT if say TSN 26 were
> sent with B bit set. Since
> it would recognize the peer (A) has violated the protocol sending to chunks
> with B=1/E=0 set in a row.
>
> R
>
> >
> > Best regards
> > Michael
> >>
> >> Regards,
> >> V Sandeep Kandula.
> >
> >
>
> -----
> Randall Stewart
> [e-mail]
>
>
>
>
>


-- 

*Regards,
V Sandeep Kandula.*