laforge has submitted this change and it was merged. (
https://gerrit.osmocom.org/c/osmo-bsc/+/15408 )
Change subject: bsc_subscr_conn_fsm: Cleanly clear BSSAP conn if associated
channel closed during WAIT_CC
......................................................................
bsc_subscr_conn_fsm: Cleanly clear BSSAP conn if associated channel closed
during WAIT_CC
TTCN3 BSC_Tests.TC_ms_rel_ind_does_not_cause_bssmap_reset seems to
sometimes run into a race condition on the order of messages received by
osmo-bsc comming from MSC and BTS.
Usual (expected) scenario):
BTS->BSC EST IND
BSC->MSC CL3 Info
BSC<-MSC CC
BTS->BSC REL IND
BTS<-BSC DEACT SACCH
BSC->MSC ClearRequest
BSC<-MSC ClearCommand
BSC->MSC ClearComplete
BTS<-BSC RF Chan Release
BTS->BSC RF Chan Release ACK
Sometimes CC message and REL IND message are received swapped (because they
are sent by different components asynchronously in TTCN3).
As a result, osmo-bsc was failing to go into CLEARING state and was
unable to send the ClearRequest because CC was still not received.
So the idea is to stay in WAIT_CC until CC is received, then check if
the lchan was dropped and in that case go into clearing state.
Change-Id: Id1abf5ee44c60925b478123409f26bd29006202b
---
M src/osmo-bsc/bsc_subscr_conn_fsm.c
1 file changed, 28 insertions(+), 5 deletions(-)
Approvals:
Jenkins Builder: Verified
laforge: Looks good to me, approved
diff --git a/src/osmo-bsc/bsc_subscr_conn_fsm.c
b/src/osmo-bsc/bsc_subscr_conn_fsm.c
index f8784f9..60f035d 100644
--- a/src/osmo-bsc/bsc_subscr_conn_fsm.c
+++ b/src/osmo-bsc/bsc_subscr_conn_fsm.c
@@ -332,8 +332,19 @@
struct gsm_subscriber_connection *conn = fi->priv;
switch (event) {
case GSCON_EV_A_CONN_CFM:
- /* MSC has confirmed the connection, we now change into the
- * active state and wait there for further operations */
+ /* MSC has confirmed the connection */
+
+ if (!conn->lchan) {
+ /* If associated lchan was released while we were
waiting for the
+ confirmed connection, then instead simply drop the
connection */
+ LOGPFSML(fi, LOGL_INFO,
+ "Connection confirmed but lchan was dropped
previously, clearing conn\n");
+ osmo_fsm_inst_state_chg(conn->fi, ST_CLEARING, 60, 999);
+ gscon_bssmap_clear(conn,
GSM0808_CAUSE_EQUIPMENT_FAILURE);
+ break;
+ }
+
+ /* We now change into the active state and wait there for
further operations. */
conn_fsm_state_chg(ST_ACTIVE);
/* if there's user payload, forward it just like EV_MT_DTAP */
/* FIXME: Question: if there's user payload attached to the CC,
forward it like EV_MT_DTAP? */
@@ -589,7 +600,7 @@
[ST_WAIT_CC] = {
.name = "WAIT_CC",
.in_event_mask = S(GSCON_EV_A_CONN_CFM),
- .out_state_mask = S(ST_ACTIVE),
+ .out_state_mask = S(ST_ACTIVE) | S(ST_CLEARING),
.action = gscon_fsm_wait_cc,
},
[ST_ACTIVE] = {
@@ -651,10 +662,22 @@
lchan_forget_conn(conn->lchan);
conn->lchan = NULL;
}
+ /* If the conn has no lchan anymore, it was released by the BTS and
needs to Clear towards MSC. */
if (!conn->lchan) {
- if (conn->fi->state != ST_CLEARING)
+ switch (conn->fi->state) {
+ case ST_WAIT_CC:
+ /* The SCCP connection was not yet confirmed by a CC,
the BSSAP is not fully established
+ yet so we cannot release it. First wait for the CC,
and release in gscon_fsm_wait_cc(). */
+ break;
+ default:
+ /* Ensure that the FSM is in ST_CLEARING. */
osmo_fsm_inst_state_chg(conn->fi, ST_CLEARING, 60, 999);
- gscon_bssmap_clear(conn, GSM0808_CAUSE_EQUIPMENT_FAILURE);
+ /* fall thru, omit an error log if already in
ST_CLEARING */
+ case ST_CLEARING:
+ /* Request a Clear Command from the MSC. */
+ gscon_bssmap_clear(conn,
GSM0808_CAUSE_EQUIPMENT_FAILURE);
+ break;
+ }
}
}
--
To view, visit https://gerrit.osmocom.org/c/osmo-bsc/+/15408
To unsubscribe, or for help writing mail filters, visit
https://gerrit.osmocom.org/settings
Gerrit-Project: osmo-bsc
Gerrit-Branch: master
Gerrit-Change-Id: Id1abf5ee44c60925b478123409f26bd29006202b
Gerrit-Change-Number: 15408
Gerrit-PatchSet: 4
Gerrit-Owner: pespin <pes...@sysmocom.de>
Gerrit-Reviewer: Jenkins Builder
Gerrit-Reviewer: fixeria <axilira...@gmail.com>
Gerrit-Reviewer: laforge <lafo...@gnumonks.org>
Gerrit-Reviewer: neels <nhofm...@sysmocom.de>
Gerrit-Reviewer: pespin <pes...@sysmocom.de>
Gerrit-MessageType: merged
