Wireshark-dev: Re: [Wireshark-dev] ZigBee APS re-assemble with re-used sequence number
From: Jaap Keuter <[email protected]>
Date: Thu, 2 Aug 2018 20:03:58 +0200

Not burdened by any ZigBee domain knowledge I would say that a seq# rollover would require a clearing of the non-reassembled fragments. But not all of them because we could still be in the process of reassembling the part of the stream with the not-yet rolled over seq#. A sliding window of non-reassembled fragments, of about half the seq# range, moved forward by the next received seq#, could be sufficient. All in all this would be an extension of the generic reassembly routenes, assuming they are used...


On 2 Aug 2018, at 12:17, Kenneth Soerensen <[email protected]> wrote:


Any idea how we can fix this?

The packet re-assembler is confused by ZigBee APS re-using sequence numbers, which makes it hard to distinguish what fragments belong to specific re-assembled packets.