rpms/kernel/F-13 sched-fix-over-scheduling-bug.patch, NONE, 1.1 kernel.spec, 1.2080, 1.2081

Chuck Ebbert cebbert at fedoraproject.org
Tue Jul 6 13:47:19 UTC 2010


Author: cebbert

Update of /cvs/pkgs/rpms/kernel/F-13
In directory cvs01.phx2.fedoraproject.org:/tmp/cvs-serv28548

Modified Files:
	kernel.spec 
Added Files:
	sched-fix-over-scheduling-bug.patch 
Log Message:
sched-fix-over-scheduling-bug.patch: fix scheduler bug with CGROUPS

sched-fix-over-scheduling-bug.patch:
 sched.c |    3 ---
 1 file changed, 3 deletions(-)

--- NEW FILE sched-fix-over-scheduling-bug.patch ---
From: Alex,Shi <alex.shi at intel.com>
Date: Thu, 17 Jun 2010 06:08:13 +0000 (+0800)
Subject: sched: Fix over-scheduling bug
X-Git-Url: http://git.kernel.org/?p=linux%2Fkernel%2Fgit%2Ftorvalds%2Flinux-2.6.git;a=commitdiff_plain;h=3c93717cfa51316e4dbb471e7c0f9d243359d5f8

sched: Fix over-scheduling bug

Commit e70971591 ("sched: Optimize unused cgroup configuration") introduced
an imbalanced scheduling bug.  [[ in 2.6.32-rc1 ]]

If we do not use CGROUP, function update_h_load won't update h_load. When the
system has a large number of tasks far more than logical CPU number, the
incorrect cfs_rq[cpu]->h_load value will cause load_balance() to pull too
many tasks to the local CPU from the busiest CPU. So the busiest CPU keeps
going in a round robin. That will hurt performance.

The issue was found originally by a scientific calculation workload that
developed by Yanmin. With that commit, the workload performance drops
about 40%.

 CPU  before    after

 00   : 2       : 7
 01   : 1       : 7
 02   : 11      : 6
 03   : 12      : 7
 04   : 6       : 6
 05   : 11      : 7
 06   : 10      : 6
 07   : 12      : 7
 08   : 11      : 6
 09   : 12      : 6
 10   : 1       : 6
 11   : 1       : 6
 12   : 6       : 6
 13   : 2       : 6
 14   : 2       : 6
 15   : 1       : 6

Reviewed-by: Yanmin zhang <yanmin.zhang at intel.com>
Signed-off-by: Alex Shi <alex.shi at intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra at chello.nl>
LKML-Reference: <1276754893.9452.5442.camel at debian>
Signed-off-by: Ingo Molnar <mingo at elte.hu>
---

diff --git a/kernel/sched.c b/kernel/sched.c
index 2aaceeb..6c9e7c8 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -1657,9 +1657,6 @@ static void update_shares(struct sched_domain *sd)
 
 static void update_h_load(long cpu)
 {
-	if (root_task_group_empty())
-		return;
-
 	walk_tg_tree(tg_load_down, tg_nop, (void *)cpu);
 }
 


Index: kernel.spec
===================================================================
RCS file: /cvs/pkgs/rpms/kernel/F-13/kernel.spec,v
retrieving revision 1.2080
retrieving revision 1.2081
diff -u -p -r1.2080 -r1.2081
--- kernel.spec	6 Jul 2010 08:17:19 -0000	1.2080
+++ kernel.spec	6 Jul 2010 13:47:17 -0000	1.2081
@@ -859,6 +859,8 @@ Patch13010: rt2x00-rt2800-Make-rt30xx-an
 # iwlwifi: cancel scan watchdog in iwl_bg_abort_scan
 Patch13020: iwlwifi-cancel-scan-watchdog-in-iwl_bg_abort_scan.patch
 
+Patch13030: sched-fix-over-scheduling-bug.patch
+
 %endif
 
 BuildRoot: %{_tmppath}/kernel-%{KVERREL}-root
@@ -1564,6 +1566,9 @@ ApplyPatch rt2x00-rt2800-Make-rt30xx-and
 # iwlwifi: cancel scan watchdog in iwl_bg_abort_scan
 ApplyPatch iwlwifi-cancel-scan-watchdog-in-iwl_bg_abort_scan.patch
 
+# fix performance problem with CGROUPS
+ApplyPatch sched-fix-over-scheduling-bug.patch
+
 # END OF PATCH APPLICATIONS
 
 %endif
@@ -2214,6 +2219,9 @@ fi
 # and build.
 
 %changelog
+* Tue Jul 06 2010 Chuck Ebbert <cebbert at redhat.com> 2.6.33.6-144
+- sched-fix-over-scheduling-bug.patch: fix scheduler bug with CGROUPS
+
 * Tue Jul 06 2010 Chuck Ebbert <cebbert at redhat.com> 2.6.33.6-143
 - Linux 2.6.33.6
 



More information about the scm-commits mailing list