hardcoding ra_rnti in case of long prach for benetel
This commit is contained in:
@@ -248,8 +248,6 @@ void nr_initiate_ra_proc(module_id_t module_idP,
|
|||||||
uint8_t ul_carrier_id = 0; // 0 for NUL 1 for SUL
|
uint8_t ul_carrier_id = 0; // 0 for NUL 1 for SUL
|
||||||
NR_SearchSpace_t *ss;
|
NR_SearchSpace_t *ss;
|
||||||
int UE_id = 0;
|
int UE_id = 0;
|
||||||
// ra_rnti from 5.1.3 in 38.321
|
|
||||||
uint16_t ra_rnti=1+symbol+(slotP*14)+(freq_index*14*80)+(ul_carrier_id*14*80*8);
|
|
||||||
|
|
||||||
uint16_t msg2_frame, msg2_slot,monitoring_slot_period,monitoring_offset;
|
uint16_t msg2_frame, msg2_slot,monitoring_slot_period,monitoring_offset;
|
||||||
gNB_MAC_INST *nr_mac = RC.nrmac[module_idP];
|
gNB_MAC_INST *nr_mac = RC.nrmac[module_idP];
|
||||||
@@ -259,6 +257,16 @@ void nr_initiate_ra_proc(module_id_t module_idP,
|
|||||||
NR_ServingCellConfigCommon_t *scc = cc->ServingCellConfigCommon;
|
NR_ServingCellConfigCommon_t *scc = cc->ServingCellConfigCommon;
|
||||||
NR_RA_t *ra = &cc->ra[0];
|
NR_RA_t *ra = &cc->ra[0];
|
||||||
|
|
||||||
|
uint16_t ra_rnti;
|
||||||
|
|
||||||
|
// ra_rnti from 5.1.3 in 38.321
|
||||||
|
// FK: in case of long PRACH the phone seems to expect the subframe number instead of the slot number here.
|
||||||
|
if (scc->uplinkConfigCommon->initialUplinkBWP->rach_ConfigCommon->choice.setup->prach_RootSequenceIndex.present==NR_RACH_ConfigCommon__prach_RootSequenceIndex_PR_l839)
|
||||||
|
ra_rnti=1+symbol+(9/*slotP*/*14)+(freq_index*14*80)+(ul_carrier_id*14*80*8);
|
||||||
|
else
|
||||||
|
ra_rnti=1+symbol+(slotP*14)+(freq_index*14*80)+(ul_carrier_id*14*80*8);
|
||||||
|
|
||||||
|
|
||||||
// if the preamble received correspond to one of the listed
|
// if the preamble received correspond to one of the listed
|
||||||
// the UE sent a RACH either for starting RA procedure or RA procedure failed and UE retries
|
// the UE sent a RACH either for starting RA procedure or RA procedure failed and UE retries
|
||||||
int pr_found=0;
|
int pr_found=0;
|
||||||
|
|||||||
Reference in New Issue
Block a user