fix(sdap): skip unmapped-QFI drop check on UE DL path

[alexjiao2021]

Ping UE failed with 3rd-party gNB, UE will report error as below:
[SDAP] Dropping UL SDAP PDU with unmapped QFI=5 (ue=0 pdu_session=1 drb=2)

gNB has configuration for SDAP as below:

                                drb-ToAddModList: 1 item
                                    Item 0
                                        DRB-ToAddMod
                                            cnAssociation: sdap-Config (1)
                                                sdap-Config
                                                    pdu-Session: 1
                                                    sdap-HeaderDL: absent (1)
                                                    sdap-HeaderUL: present (0)
                                                    ..1. .... defaultDRB: True
                                                    mappedQoS-FlowsToAdd: 1 item
                                                        Item 0
                                                            QFI: 2
                                            drb-Identity: 2

Complete gNB RRC log in pcap format:
nr_rrc.zip
UE log:
nr_ue_ping_fail.log

This PR fixed the issue, details are in commit message.

[durantabot]

CI Build: #184 | Failed on the following stages:

• OAI-FLEXRIC-RAN-Integration-Test
• RAN-VRT-Sim-Test-5G
• RAN-RF-Sim-Test-5G

[durantabot]

CI Build: #185 | Failed on the following stages:

• RAN-VRT-Sim-Test-5G
• RAN-RF-Sim-Test-5G

[GuidoCasati]

Hi @alexjiao2021, thanks for reporting this.

We have a similar scenario in https://gitlab.eurecom.fr/oai/cn5g/oai-cn5g-upf/-/work_items/36 in the CN to gNB N3 path: the simple-switch UPF sends DL GTP-U with default qfi = 0x05 until an UL packet reclassifies the session.

In this case, it also happen that QFI=5 is received but for a different reason: gNB sends DL without SDAP header, the UE SDAP sees raw IPv4 SDUs and QFI=5 comes from IPv4 header byte 0x45.

This is actually a bug, since entity->enable_sdap, is true if either UL or DL header is present, and your capture shows that it can be not the case since we have UL header only.

However, the fix is not enough, as it would still parse IP header as an SDAP header, so we have bogus qfi, and also parsing should be done only when RRC says header is present. I'll think about a more proper fix, which I believe should do the header parsing on a per-direction role.

[alexjiao2021]

However, the fix is not enough, as it would still parse IP header as an SDAP header, so we have bogus qfi, and also parsing should be done only when RRC says header is present. I'll think about a more proper fix, which I believe should do the header parsing on a per-direction role.

Thanks @GuidoCasati , once you have a more proper fix ready, I'll close this PR.

[GuidoCasati]

However, the fix is not enough, as it would still parse IP header as an SDAP header, so we have bogus qfi, and also parsing should be done only when RRC says header is present. I'll think about a more proper fix, which I believe should do the header parsing on a per-direction role.

Thanks @GuidoCasati , once you have a more proper fix ready, I'll close this PR.

hi @alexjiao2021, could you please test #249 and see if it solve the issue?