Blocking nr_scheduler_response untill L1 tx finishes
This commit is contained in:
@@ -178,6 +178,13 @@ void nr_schedule_response(NR_Sched_Rsp_t *Sched_INFO){
|
||||
|
||||
gNB = RC.gNB[Mod_id];
|
||||
|
||||
// we wait for the L1 Tx processing to finish before copying MAC PDUs to PHY
|
||||
{
|
||||
notifiedFIFO_elt_t *res;
|
||||
res = pullTpool(gNB->resp_L1_tx, gNB->threadPool);
|
||||
pushNotifiedFIFO(gNB->resp_L1_tx,res);
|
||||
}
|
||||
|
||||
uint8_t number_dl_pdu = (DL_req==NULL) ? 0 : DL_req->dl_tti_request_body.nPDUs;
|
||||
uint8_t number_ul_dci_pdu = (UL_dci_req==NULL) ? 0 : UL_dci_req->numPdus;
|
||||
uint8_t number_ul_tti_pdu = (UL_tti_req==NULL) ? 0 : UL_tti_req->n_pdus;
|
||||
|
||||
Reference in New Issue
Block a user