hack to avoid zombie UEs in the MAC layer

Here is the problem:
    Sometimes the UE has no PHY context but
    is still present in the MAC with 'ul_failure_timer' = 0 and
    'ul_out_of_sync' = 0. It seems wrong and the UE stays there forever. Let's
    start an UL out of sync procedure in this case.
    The root cause of this problem has to be found and corrected.
    In the meantime, this hack...

This has to be redone at some point.

openair2/LAYER2/MAC/eNB_scheduler_ulsch.c

@@ -762,8 +762,25 @@ void schedule_ulsch_rnti(module_id_t   module_idP,
         break;
       }
     }
-    if (drop_ue == 1)
+    if (drop_ue == 1) {
+      /* TODO: this is a hack. Sometimes the UE has no PHY context but
+       * is still present in the MAC with 'ul_failure_timer' = 0 and
+       * 'ul_out_of_sync' = 0. It seems wrong and the UE stays there forever. Let's
+       * start an UL out of sync procedure in this case.
+       * The root cause of this problem has to be found and corrected.
+       * In the meantime, this hack...
+       */
+      if (UE_list->UE_sched_ctrl[UE_id].ul_failure_timer == 0 &&
+          UE_list->UE_sched_ctrl[UE_id].ul_out_of_sync == 0) {
+        LOG_W(MAC,"[eNB %d] frame %d subframe %d, UE %d/%x CC %d: UE in weird state, let's put it 'out of sync'\n",
+              module_idP,frameP,subframeP,UE_id,rnti,CC_id);
+        // inform RRC of failure and clear timer
+        mac_eNB_rrc_ul_failure(module_idP,CC_id,frameP,subframeP,rnti);
+        UE_list->UE_sched_ctrl[UE_id].ul_failure_timer=0;
+        UE_list->UE_sched_ctrl[UE_id].ul_out_of_sync=1;
+      }
       continue;
+    }
 
     // loop over all active UL CC_ids for this UE
     for (n=0; n<UE_list->numactiveULCCs[UE_id]; n++) {