Attention is currently required from: neels, pespin. laforge has posted comments on this change. ( https://gerrit.osmocom.org/c/osmo-upf/+/27631 )
Change subject: libosmo-pfcp: implement PFCP header and msg handling ......................................................................
Patch Set 3: Code-Review+1
(3 comments)
File include/osmocom/pfcp/pfcp_msg.h:
https://gerrit.osmocom.org/c/osmo-upf/+/27631/comment/1d5c28ce_fe655651 PS3, Line 69: static inline uint32_t osmo_pfcp_next_seq(uint32_t *seq_state)
This function will return 1 on first call most probably. […]
I would expect no constraint whatsoevre on the starting / initial value of the sequence number. Basically every command must have a unique serial number so we can match responses to requests, and the receiver can distinguish re-transmission of a command (same seq) from a new command (different seq)
https://gerrit.osmocom.org/c/osmo-upf/+/27631/comment/ca80ed75_e149434f PS3, Line 72: (*seq_state) &= 0xffffff; the interesting question is why we use a uint32_t variable if we only use 16 bits of it?
https://gerrit.osmocom.org/c/osmo-upf/+/27631/comment/f78a1104_ebf16f72 PS3, Line 110: struct osmo_fsm_inst *peer_fi;
Looks like these 2 groups can be in a union?
peers and sessins are orthogonal concepts, why do you think a message is only either part of a session or related to a peer? shouldn't it be both at the same time?