With the current HEAD (27 Aug 25) and no slicing (SD blank), a real gNB (SageRAN-based) and real UEs (data-centric modems and phones) we’re seeing successful connections with Mediatek-based UEs and immediate disconnection of new PDU sessions initiated by Qualcomm-based UEs. The Qualcomm-based UEs send PDU Session Release Request: Regular Deactivation immediately after receiving PDU Session Establishment Accept from Free5GC.
I have compared the PDU Session Establishment Accept as seen by the UE between successful connections using a different core and the dropped connections using Free5GC and it could be that the message is slightly malformed (and this is where a bit more expertise would help).
Successful message (as seen from the UE):
Free5GC message (as seen from the UE):
You’ll see the Free5GC version includes the SD field but leaves it blank (rather than omitting the field) and includes (unnecessary?) HPLMNID but with MNC and MCC set to zero.
This may all be the result of configuration errors on my part but I’ve experimented without success and although I think I can see where the message is built (gsm_build.go ?) I can’t see exactly where the decision to include SD and HPLMNID is made. If someone can point me in the right direction I’ll happily experiment and report back.