fix(bpf_collector): fix command name in case of kernel processes

  if the cgroupID for the pid is 1, it is considered as kernel process
  and bpf cpu time for each such process is aggregated under the same
  ProcessStat

  But the command which shows up in metric is from the first pid which
  created the entry in ProcessStats. It could be a kworker, or
  rcu_preempt, or some other kernel process. Which is misleading because
  it shows that process consuming most of bpf cpu time.

  This commit:
  - uses "kernel_processes" as the command for pid 1 kernel processes
  - sets the pid as 1 instead of the pid of the process that created the
    entry

Signed-off-by: Vimal Kumar <[email protected]>

pkg/collector/resourceutilization/bpf/process_bpf_collector.go

@@ -110,17 +110,19 @@ func UpdateProcessBPFMetrics(bpfExporter bpf.Exporter, processStats map[uint64]*
 		}
 
 		mapKey := ct.Pid
+		process := comm
 		if ct.CgroupId == 1 && config.EnabledEBPFCgroupID() {
 			// we aggregate all kernel process to minimize overhead
 			// all kernel process has cgroup id as 1 and pid 1 is also a kernel process
 			mapKey = 1
+			process = "kernel_processes"
 		}
 
 		bpfSupportedMetrics := bpfExporter.SupportedMetrics()
 		var ok bool
 		var pStat *stats.ProcessStats
 		if pStat, ok = processStats[mapKey]; !ok {
-			pStat = stats.NewProcessStats(ct.Pid, ct.CgroupId, containerID, vmID, comm, bpfSupportedMetrics)
+			pStat = stats.NewProcessStats(mapKey, ct.CgroupId, containerID, vmID, process, bpfSupportedMetrics)
 			processStats[mapKey] = pStat
 		} else if pStat.Command == "" {
 			pStat.Command = comm