fixes for the thread blocking problem
This commit is contained in:
+10
-3
@@ -423,11 +423,16 @@ int wakeup_txfh(PHY_VARS_gNB *gNB,gNB_L1_rxtx_proc_t *proc,int frame_tx,int slot
|
||||
|
||||
// note this should depend on the numerology used by the TX L1 thread, set here for 500us slot time
|
||||
VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_UE_GAIN_CONTROL,1);
|
||||
waitret=wait_on_condition(&proc->mutex_RUs_tx,&proc->cond_RUs,&proc->instance_cnt_RUs,"wakeup_txfh");
|
||||
AssertFatal(release_thread(&proc->mutex_RUs_tx,&proc->instance_cnt_RUs,"wakeup_txfh")==0, "error releaseing gNB lock on RUs\n");
|
||||
AssertFatal((ret = pthread_mutex_lock(&proc->mutex_RUs_tx))==0,"mutex_lock returns %d\n",ret);
|
||||
while (proc->instance_cnt_RUs < 0) {
|
||||
pthread_cond_wait(&proc->cond_RUs,&proc->mutex_RUs_tx); // this unlocks mutex_rxtx while waiting and then locks it again
|
||||
}
|
||||
proc->instance_cnt_RUs = -1;
|
||||
VCD_SIGNAL_DUMPER_DUMP_VARIABLE_BY_NAME(VCD_SIGNAL_DUMPER_VARIABLES_FRAME_NUMBER_RX0_UE,proc->instance_cnt_RUs);
|
||||
AssertFatal((ret = pthread_mutex_unlock(&proc->mutex_RUs_tx))==0,"mutex_unlock returns %d\n",ret);
|
||||
VCD_SIGNAL_DUMPER_DUMP_FUNCTION_BY_NAME(VCD_SIGNAL_DUMPER_FUNCTIONS_UE_GAIN_CONTROL,0);
|
||||
|
||||
VCD_SIGNAL_DUMPER_DUMP_VARIABLE_BY_NAME(VCD_SIGNAL_DUMPER_VARIABLES_FRAME_NUMBER_RX0_UE,proc->instance_cnt_RUs);
|
||||
|
||||
|
||||
if (waitret == ETIMEDOUT) {
|
||||
LOG_W(PHY,"Dropping TX slot (%d.%d) because FH is blocked more than 1 slot times (500us)\n",frame_tx,slot_tx);
|
||||
@@ -465,6 +470,8 @@ int wakeup_txfh(PHY_VARS_gNB *gNB,gNB_L1_rxtx_proc_t *proc,int frame_tx,int slot
|
||||
return(-1);
|
||||
}
|
||||
|
||||
AssertFatal((ret=pthread_mutex_lock(&ru_proc->mutex_gNBs))==0,"mutex_lock returned %d\n",ret);
|
||||
|
||||
ru_proc->instance_cnt_gNBs = 0;
|
||||
ru_proc->timestamp_tx = timestamp_tx;
|
||||
ru_proc->tti_tx = slot_tx;
|
||||
|
||||
Reference in New Issue
Block a user