Attention is currently required from: pespin, msuraev.

neels has posted comments on this change. ( https://gerrit.osmocom.org/c/osmo-hnbgw/+/32323 )

Change subject: cnpool: allow separate cs7 for IuPS and IuCS ......................................................................

Patch Set 9:

(2 comments)

File include/osmocom/hnbgw/hnbgw.h:

https://gerrit.osmocom.org/c/osmo-hnbgw/+/32323/comment/2e47297f_08df39f5 PS8, Line 87: DECLARE_HASHTABLE(hnbgw_context_map_by_conn_id, 6);

maybe add a comment what's going to be added in here (struct hnbgw_cnlink?).

it would be pretty much exactly the name of the table:

/* hash table of struct hnbgw_context_map by SCCP conn ID */

but ok if it helps i can do it

https://gerrit.osmocom.org/c/osmo-hnbgw/+/32323/comment/8c1542b3_125575af PS8, Line 133: if (!cnlink->hnbgw_sccp_user)

can this really happen? Is it expected to happen? I find it confusing when seeing all these checks w […]

In places where it's established to be safe i can just use cnlink->hnbgw_sccp_usser->ss7->sccp directly.

This function's reason to exist is to avoid NULL dereferences. I want to call this function in *any* godforsaken corner case from early init to late cleanup, and never run into a NULL deref. I want to get the osmo_sccp_instance, or NULL if there isn't any, for whichever weird reason there might be.

In the branch's final state, this function is called in two places:

cnlink_sccp_addr_to_str(): return a string for logging -- if it is early link startup logging there might be no SCCP instance, or anything leading up to the SCCP instance, assigned yet.

tx_reset_ack(): the SCCP link should be stable, but we're in the middle of a RESET operation, so it could be some state transition, and rarely happens. I decided to just do the extra check.

I thought I'd need it more often, but two places is already enough to not dup all these checks.