laforge submitted this change.

View Change

octoi: Disconnect the link when >= 7500 underruns/s

This situation is the result of the peer clock being continuously
too slow compared to the local clock, leading to RIFO underruns at
[virtually] all of the 8000/s E1 frames.  As the current code doesn't
recover from this, we might as well disconnect and re-start for
recovery.

Change-Id: Ie3fffa1c1c20962b40320c8cc088c140b8d64e77
---
M src/octoi/e1oip.h
M src/octoi/octoi_clnt_fsm.c
M src/octoi/octoi_srv_fsm.c
3 files changed, 45 insertions(+), 5 deletions(-)

diff --git a/src/octoi/e1oip.h b/src/octoi/e1oip.h
index 1a184d2..99b88f0 100644
--- a/src/octoi/e1oip.h
+++ b/src/octoi/e1oip.h
@@ -59,6 +59,13 @@
 	/* TODO: statistics (RTT, frame loss, std deviation, alarms */
 };
 
+/* get the rate of the given counter during the last second */
+static inline uint64_t iline_ctr_get_rate_1s(struct e1oip_line *iline, unsigned int idx)
+{
+	const struct rate_ctr *ctr = rate_ctr_group_get_ctr(iline->ctrs, idx);
+	return ctr->intv[RATE_CTR_INTV_SEC].rate;
+}
+
 struct e1oip_line *e1oip_line_alloc(struct octoi_peer *peer);
 void e1oip_line_set_name(struct e1oip_line *line, const char *name);
 void e1oip_line_reset(struct e1oip_line *iline);
diff --git a/src/octoi/octoi_clnt_fsm.c b/src/octoi/octoi_clnt_fsm.c
index f2fe989..a3d0518 100644
--- a/src/octoi/octoi_clnt_fsm.c
+++ b/src/octoi/octoi_clnt_fsm.c
@@ -34,6 +34,7 @@
 	CLNT_ST_ACCEPTED,
 	CLNT_ST_REJECTED,
 	CLNT_ST_REDIRECTED,
+	CLNT_ST_WAIT_RECONNECT,
 };
 
 struct clnt_state {
@@ -179,7 +180,8 @@
 		.name = "ACCEPTED",
 		.in_event_mask = S(OCTOI_CLNT_EV_RX_AUTH_REQ) |
 				 S(OCTOI_EV_RX_TDM_DATA),
-		.out_state_mask = S(CLNT_ST_INIT),
+		.out_state_mask = S(CLNT_ST_INIT) |
+				  S(CLNT_ST_WAIT_RECONNECT),
 		.action = clnt_st_accepted,
 		.onenter = clnt_st_accepted_onenter,
 		.onleave = clnt_st_accepted_onleave,
@@ -196,6 +198,11 @@
 		.out_state_mask = S(CLNT_ST_SVC_REQ_SENT),
 		.action = clnt_st_redirected,
 	},
+	[CLNT_ST_WAIT_RECONNECT] = {
+		.name = "WAIT_RECONNECT",
+		.in_event_mask = 0,
+		.out_state_mask = S(CLNT_ST_INIT),
+	},
 };
 
 static void clnt_allstate_action(struct osmo_fsm_inst *fi, uint32_t event, void *data)
@@ -235,6 +242,9 @@
 				     PACKAGE_NAME, PACKAGE_VERSION, st->capability_flags);
 		osmo_fsm_inst_state_chg(fi, CLNT_ST_SVC_REQ_SENT, 10, 0);
 		break;
+	case CLNT_ST_WAIT_RECONNECT:
+		LOGPFSML(fi, LOGL_INFO, "Re-starting connection\n");
+		osmo_fsm_inst_state_chg(fi, CLNT_ST_INIT, 0, 0);
 	}
 	return 0;
 }
@@ -257,6 +267,7 @@
 	struct osmo_fsm_inst *fi = data;
 	struct clnt_state *st = fi->priv;
 	struct timespec ts;
+	uint64_t rate;
 
 	clock_gettime(CLOCK_MONOTONIC, &ts);
 
@@ -264,8 +275,19 @@
 		LOGPFSML(fi, LOGL_NOTICE, "No TDM data received for >= 3 seconds, declaring peer dead\n");
 		osmo_fsm_inst_state_chg(fi, CLNT_ST_INIT, 0, 0);
 		osmo_fsm_inst_dispatch(fi, OCTOI_CLNT_EV_REQUEST_SERVICE, NULL);
-	} else
-		osmo_timer_schedule(&st->rx_alive_timer, 3, 0);
+		return;
+	}
+
+	rate = iline_ctr_get_rate_1s(st->peer->iline, LINE_CTR_E1oIP_UNDERRUN);
+	if (rate > 7500) {
+		LOGPFSML(fi, LOGL_ERROR, "More than 7500 RIFO underruns per second: "
+			 "Your clock appears to be too fast. Disconnecting.\n");
+		osmo_fsm_inst_state_chg(fi, CLNT_ST_WAIT_RECONNECT, 10, 0);
+		osmo_fsm_inst_dispatch(fi, OCTOI_CLNT_EV_REQUEST_SERVICE, NULL);
+		return;
+	}
+
+	osmo_timer_schedule(&st->rx_alive_timer, 3, 0);
 }
 
 
diff --git a/src/octoi/octoi_srv_fsm.c b/src/octoi/octoi_srv_fsm.c
index 3990a16..5f51003 100644
--- a/src/octoi/octoi_srv_fsm.c
+++ b/src/octoi/octoi_srv_fsm.c
@@ -345,14 +345,25 @@
 	struct osmo_fsm_inst *fi = data;
 	struct srv_state *st = fi->priv;
 	struct timespec ts;
+	uint64_t rate;
 
 	clock_gettime(CLOCK_MONOTONIC, &ts);
 
 	if (ts.tv_sec - st->peer->last_rx_tdm > 3) {
 		LOGPFSML(fi, LOGL_NOTICE, "No TDM data received for >= 3 seconds, declaring peer dead\n");
 		osmo_fsm_inst_term(fi, OSMO_FSM_TERM_TIMEOUT, NULL);
-	} else
-		osmo_timer_schedule(&st->rx_alive_timer, 3, 0);
+		return;
+	}
+
+	rate = iline_ctr_get_rate_1s(st->peer->iline, LINE_CTR_E1oIP_UNDERRUN);
+	if (rate > 7500) {
+		LOGPFSML(fi, LOGL_ERROR, "More than 7500 RIFO underruns per second: "
+			 "Peer clock is too slow. Disconnecting.\n");
+		osmo_fsm_inst_term(fi, OSMO_FSM_TERM_ERROR, NULL);
+		return;
+	}
+
+	osmo_timer_schedule(&st->rx_alive_timer, 3, 0);
 }
 
 /* call-back function for every received OCTOI socket message for given peer */

To view, visit change 27839. To unsubscribe, or for help writing mail filters, visit settings.