https://bugs.wireshark.org/bugzilla/show_bug.cgi?id=13202
--- Comment #5 from Parav Pandit <paravpan...@yahoo.com> --- Hi Chuck, I debugged further there are X issues. 1. packet-rpcrdma.c invokes upper level dissector (nfs), which performs conversation lookup based on bidirectional property (tuple of sip, dip, sport, dport). This tuple is not setup by packet-infiniband.c because src_qp doesn't exist in packet. So NFS dissector fails to load. Thats why you see Infiniband packets showing up. 2. packet-rpcrdma.c is calling find_or_create_conversation(pinfo) during dissect_rpcordma_heur(), which is incorrect for Infiniband. Because for proto_ib conversation, we store service_id etc for a conversation. Creating conversation by upper protocol without setting protocol data won't find it. This issue got exposed because duplicate conversation entries doesn't exist anymore. So now to avoid every dissectors above infiniband to perform lookup based on out-of-band src_qp, (which I am doing in nvme dissector), it needs to move to packet-infiniband.c By doing so, pinfo sport will contain src_qp and tuple lookup by nfs, and dissectors above infiniband can succeed. I have tested this in local sandbox. I will send patch soon for packet-infiniband.c and packet-rpcrdma.c -- You are receiving this mail because: You are watching all bug changes.
___________________________________________________________________________ Sent via: Wireshark-bugs mailing list <wireshark-bugs@wireshark.org> Archives: https://www.wireshark.org/lists/wireshark-bugs Unsubscribe: https://www.wireshark.org/mailman/options/wireshark-bugs mailto:wireshark-bugs-requ...@wireshark.org?subject=unsubscribe
