replacing call to RRC processing timers for msg3 dcch
This commit is contained in:
@@ -1781,8 +1781,16 @@ void nr_generate_Msg4(module_id_t module_idP, int CC_id, frame_t frameP, sub_fra
|
||||
harq->round = 0;
|
||||
harq->ndi ^= 1;
|
||||
|
||||
sched_ctrl->rrc_processing_timer = 1; // No need to wait at this point in time, setting it to 1 just to enter in function nr_mac_update_timers
|
||||
LOG_I(NR_MAC, "(%d.%d) Activating RRC processing timer for UE %04x with %d ms\n", frameP, slotP, UE->rnti, 1);
|
||||
if(!UE->CellGroup)
|
||||
// In the specific scenario where UE correctly received MSG4 (assuming it decoded RRCsetup with CellGroup) and gNB didn't correctly received an ACK,
|
||||
// the UE would already have CG but the UE-dedicated gNB structure wouldn't (because RA didn't complete on gNB side).
|
||||
uper_decode(NULL,
|
||||
&asn_DEF_NR_CellGroupConfig, //might be added prefix later
|
||||
(void **)&UE->CellGroup,
|
||||
(uint8_t *)UE->cg_buf,
|
||||
(UE->enc_rval.encoded+7)/8, 0, 0);
|
||||
process_CellGroup(UE->CellGroup, sched_ctrl);
|
||||
configure_UE_BWP(nr_mac, scc, sched_ctrl, NULL, UE);
|
||||
|
||||
// Reset uplink failure flags/counters/timers at MAC so gNB will resume again scheduling resources for this UE
|
||||
UE->UE_sched_ctrl.pusch_consecutive_dtx_cnt = 0;
|
||||
|
||||
Reference in New Issue
Block a user