sched: Fix asymmetric scheduling for POWER7
authorVaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Wed, 30 Oct 2013 03:12:42 +0000 (08:42 +0530)
committerJiri Slaby <jslaby@suse.cz>
Wed, 12 Mar 2014 12:25:36 +0000 (13:25 +0100)
commit 2042abe7977222ef606306faa2dce8fd51e98e65 upstream.

Asymmetric scheduling within a core is a scheduler loadbalancing
feature that is triggered when SD_ASYM_PACKING flag is set.  The goal
for the load balancer is to move tasks to lower order idle SMT threads
within a core on a POWER7 system.

In nohz_kick_needed(), we intend to check if our sched domain (core)
is completely busy or we have idle cpu.

The following check for SD_ASYM_PACKING:

    (cpumask_first_and(nohz.idle_cpus_mask, sched_domain_span(sd)) < cpu)

already covers the case of checking if the domain has an idle cpu,
because cpumask_first_and() will not yield any set bits if this domain
has no idle cpu.

Hence, nr_busy check against group weight can be removed.

Reported-by: Michael Neuling <michael.neuling@au1.ibm.com>
Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Preeti U Murthy <preeti@linux.vnet.ibm.com>
Tested-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: vincent.guittot@linaro.org
Cc: bitbucket@online.de
Cc: benh@kernel.crashing.org
Cc: anton@samba.org
Cc: Morten.Rasmussen@arm.com
Cc: pjt@google.com
Link: http://lkml.kernel.org/r/20131030031242.23426.13019.stgit@preeti.in.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
kernel/sched/fair.c

index 411732334906bd172ad59d55921a088911f0d92a..68648f652dca44fbe511a7181049e5ff1028c938 100644 (file)
@@ -5840,7 +5840,7 @@ static inline int nohz_kick_needed(struct rq *rq, int cpu)
                if (sd->flags & SD_SHARE_PKG_RESOURCES && nr_busy > 1)
                        goto need_kick_unlock;
 
-               if (sd->flags & SD_ASYM_PACKING && nr_busy != sg->group_weight
+               if (sd->flags & SD_ASYM_PACKING
                    && (cpumask_first_and(nohz.idle_cpus_mask,
                                          sched_domain_span(sd)) < cpu))
                        goto need_kick_unlock;