Change in osmo-bts[master]: part 2 of: fix SAPIs for handover, osmo-bts-sysmo

[neels]

neels has uploaded this change for review. ( https://gerrit.osmocom.org/c/osmo-bts/+/21275 )


Change subject: part 2 of: fix SAPIs for handover, osmo-bts-sysmo
......................................................................

part 2 of: fix SAPIs for handover, osmo-bts-sysmo

Change-Id: I30e2ca003a3b15766d9d9dcf529fb2157523758f
---
M src/osmo-bts-sysmo/oml.c
1 file changed, 25 insertions(+), 17 deletions(-)



  git pull ssh://gerrit.osmocom.org:29418/osmo-bts refs/changes/75/21275/1

diff --git a/src/osmo-bts-sysmo/oml.c b/src/osmo-bts-sysmo/oml.c
index 6121f76..a4dc31e 100644
--- a/src/osmo-bts-sysmo/oml.c
+++ b/src/osmo-bts-sysmo/oml.c
@@ -657,10 +657,6 @@
 #endif
 };
 
-static const struct sapi_dir ho_sapis[] = {
-	{ GsmL1_Sapi_Rach,	GsmL1_Dir_RxUplink },
-};
-
 struct lchan_sapis {
 	const struct sapi_dir *sapis;
 	unsigned int num_sapis;
@@ -693,11 +689,6 @@
 	},
 };
 
-static const struct lchan_sapis sapis_for_ho = {
-	.sapis = ho_sapis,
-	.num_sapis = ARRAY_SIZE(ho_sapis),
-};
-
 static int mph_send_activate_req(struct gsm_lchan *lchan, struct sapi_cmd *cmd);
 static int mph_send_deactivate_req(struct gsm_lchan *lchan, struct sapi_cmd *cmd);
 static int mph_send_config_ciphering(struct gsm_lchan *lchan, struct sapi_cmd *cmd);
@@ -1176,10 +1167,10 @@
 			"%s Trying to activate lchan, but commands in queue\n",
 			gsm_lchan_name(lchan));
 
-	/* override the regular SAPIs if this is the first hand-over
-	 * related activation of the LCHAN */
+	/* For handover, always start the main channel immediately. lchan->dl_sacch.want_active indicates whether dl
+	 * SACCH should be activated. Also, for HO, start the RACH SAPI. */
 	if (lchan->ho.active == HANDOVER_ENABLED)
-		s4l = &sapis_for_ho;
+		enqueue_sapi_act_cmd(lchan, GsmL1_Sapi_Rach, GsmL1_Dir_RxUplink);
 
 	for (i = 0; i < s4l->num_sapis; i++) {
 		int sapi = s4l->sapis[i].sapi;
@@ -1192,6 +1183,14 @@
 			fl1h->alive_prim_cnt = 0;
 			osmo_timer_schedule(&fl1h->alive_timer, 5, 0);
 		}
+
+		/* For handover, possibly postpone activating the dl SACCH until the HO RACH is received. */
+		if (sapi == GsmL1_Sapi_Sacch && dir == GsmL1_Dir_TxDownlink) {
+			if (!lchan->dl_sacch.want_active)
+				continue;
+			else
+				lchan->dl_sacch.activated = true;
+		}
 		enqueue_sapi_act_cmd(lchan, sapi, dir);
 	}
 
@@ -1880,11 +1879,20 @@
 	/* Give up listening to RACH bursts */
 	release_sapi_ul_rach(lchan);
 
-	/* Activate the normal SAPIs */
-	for (i = 0; i < s4l->num_sapis; i++) {
-		int sapi = s4l->sapis[i].sapi;
-		int dir = s4l->sapis[i].dir;
-		enqueue_sapi_act_cmd(lchan, sapi, dir);
+	/* All the normal SAPIs have already been activated, only DL SACCH may still be missing.
+	 *
+	 * Note: theoretically, it would only be necessary to activate the DL SACCH when it is not active yet, i.e.
+	 * dl_sacch.activated == false. With repeated HO RACH received, we shouldn't need to re-send the SAPI activation
+	 * every time. However, tests with sysmoBTS show that when sending this SAPI activation only once, the lchan
+	 * will release some seconds after a handover, with error messages indicating
+	 * "Lost SACCH block, faking meas reports and ms pwr".
+	 * When re-sending the SAPI activation for every RACH received, the problem goes away.
+	 * Before introducing lchan->dl_sacch.want_active, this code here would activate all SAPIs for the main channel,
+	 * which would also repeat for each RACH received. Now we only activate DL SACCH here.
+	 */
+	if (lchan->dl_sacch.want_active) {
+		enqueue_sapi_act_cmd(lchan, GsmL1_Sapi_Sacch, GsmL1_Dir_TxDownlink);
+		lchan->dl_sacch.activated = true;
 	}
 
 	return 0;

-- 
To view, visit https://gerrit.osmocom.org/c/osmo-bts/+/21275
To unsubscribe, or for help writing mail filters, visit https://gerrit.osmocom.org/settings

Gerrit-Project: osmo-bts
Gerrit-Branch: master
Gerrit-Change-Id: I30e2ca003a3b15766d9d9dcf529fb2157523758f
Gerrit-Change-Number: 21275
Gerrit-PatchSet: 1
Gerrit-Owner: neels <nhofmeyr at sysmocom.de>
Gerrit-MessageType: newchange
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osmocom.org/pipermail/gerrit-log/attachments/20201121/2cb3dae6/attachment.htm>