pespin has submitted this change. ( https://gerrit.osmocom.org/c/osmo-iuh/+/40708?usp=email )

Change subject: iu_client: Initial N-NOTICE.ind support (Routing Failure of SCCP CL messages) ......................................................................

iu_client: Initial N-NOTICE.ind support (Routing Failure of SCCP CL messages)

ITU Q.714 2.8: """ When an end node is informed of a routing failure, this information is forwarded towards the SCCP user by using the N-DISCONNECT primitive (refer to reason for release in 2.1.1.2.4/Q.711) or the N-NOTICE primitive (refer to reason for return in 2.2.2.2.4/Q.711) """

We are already handling N-DISCONNECT.ind for CO messages, but N-NOTICE.ind for CL messages was not being handled.

If CL messages are not arriving to the peer (rnc), then reset the link and mark the peer as disconnected, until a new RESET can successfully fo through.

There's currently no way to submit the indication to upper layers (app). It's not worth changing the API now since anyway sole user of this API (osmo-sgsn) will actually merged the implementation directly and get rid of this layer at some point.

Related: OS#5487 Related: OS#3403 Related: OS#5917 Change-Id: Ifc59cdcb9ef910274b0580e9baba2b3a7b5e5354 --- M src/iu_client.c 1 file changed, 44 insertions(+), 0 deletions(-)

Approvals: Jenkins Builder: Verified lynxis lazus: Looks good to me, approved osmith: Looks good to me, but someone else must approve

diff --git a/src/iu_client.c b/src/iu_client.c index 75be457..e1b7af3 100644 --- a/src/iu_client.c +++ b/src/iu_client.c @@ -35,6 +35,7 @@ #include <osmocom/gprs/gprs_msgb.h> #include <osmocom/sigtran/sccp_sap.h> #include <osmocom/sigtran/sccp_helpers.h> +#include <osmocom/sccp/sccp_types.h> #include <osmocom/ranap/ranap_common_cn.h> #include <osmocom/ranap/ranap_ies_defs.h> #include <osmocom/ranap/ranap_msg_factory.h> @@ -1017,6 +1018,44 @@ ranap_iu_free_ue(ue); }

+static void handle_notice_ind(struct osmo_ss7_instance *cs7, const struct osmo_scu_notice_param *ni) +{ + struct ranap_iu_rnc *rnc; + + LOGPIU(LOGL_DEBUG, "(calling_addr=%s) N-NOTICE.ind cause=%u='%s' importance=%u\n", + osmo_sccp_addr_dump(&ni->calling_addr), + ni->cause, osmo_sccp_return_cause_name(ni->cause), + ni->importance); + + switch (ni->cause) { + case SCCP_RETURN_CAUSE_SUBSYSTEM_CONGESTION: + case SCCP_RETURN_CAUSE_NETWORK_CONGESTION: + /* Transient failures (hopefully), keep going. */ + return; + default: + break; + } + + /* Messages are not arriving to RNC. Signal to user that all related ue_ctx are invalid. */ + llist_for_each_entry(rnc, &rnc_list, entry) { + struct ranap_ue_conn_ctx *ue_ctx, *ue_ctx_tmp; + if (osmo_sccp_addr_ri_cmp(&rnc->sccp_addr, &ni->calling_addr)) + continue; + LOGPIU(LOGL_NOTICE, + "RNC %s now unreachable: N-NOTICE.ind cause=%u='%s' importance=%u\n", + osmo_rnc_id_name(&rnc->rnc_id), + ni->cause, osmo_sccp_return_cause_name(ni->cause), + ni->importance); + llist_for_each_entry_safe(ue_ctx, ue_ctx_tmp, &ue_conn_ctx_list, list) { + if (ue_ctx->rnc != rnc) + continue; + ue_ctx_link_invalidated_free(ue_ctx); + } + /* TODO: ideally we'd have some event to submit to upper + * layer to inform about peer availability change... */ + } +} + static void handle_pcstate_ind(struct osmo_ss7_instance *cs7, const struct osmo_scu_pcstate_param *pcst) { struct osmo_sccp_addr rem_addr; @@ -1209,6 +1248,11 @@ osmo_hexdump(msgb_l2(oph->msg), msgb_l2len(oph->msg))); rc = ranap_cn_rx_cl(cn_ranap_handle_cl, prim, msgb_l2(oph->msg), msgb_l2len(oph->msg)); break; + case OSMO_PRIM(OSMO_SCU_PRIM_N_NOTICE, PRIM_OP_INDICATION): + LOGPIU(LOGL_DEBUG, "N-NOTICE.ind(%s)\n", + osmo_hexdump(msgb_l2(oph->msg), msgb_l2len(oph->msg))); + handle_notice_ind(osmo_sccp_get_ss7(sccp), &prim->u.notice); + break; case OSMO_PRIM(OSMO_SCU_PRIM_N_PCSTATE, PRIM_OP_INDICATION): handle_pcstate_ind(osmo_sccp_get_ss7(sccp), &prim->u.pcstate); break;