Nanosecond resolution timers that used CLOCK_PROCESS_CPUTIME_ID such as
clock_nanosleep would not work since the values were never added to the
account group.

-ck
---
 kernel/sched/bfs.c |    3 +++
 1 file changed, 3 insertions(+)

Index: linux-3.7-bfs/kernel/sched/bfs.c
===================================================================
--- linux-3.7-bfs.orig/kernel/sched/bfs.c	2013-01-29 01:22:38.020760109 +1100
+++ linux-3.7-bfs/kernel/sched/bfs.c	2013-01-29 01:22:38.419755120 +1100
@@ -2510,6 +2510,9 @@ update_cpu_clock(struct rq *rq, struct t
 			pc_user_time(rq, p, account_pc, account_ns);
 	}
 
+	if (p != idle)
+		account_group_exec_runtime(p, account_ns);
+
 ts_account:
 	/* time_slice accounting is done in usecs to avoid overflow on 32bit */
 	if (rq->rq_policy != SCHED_FIFO && p != idle) {