]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
cpuset: Fix cpuset sched_relax_domain_level
authorJason Low <jason.low2@hp.com>
Fri, 13 Feb 2015 03:58:07 +0000 (11:58 +0800)
committerSasha Levin <sasha.levin@oracle.com>
Sat, 28 Mar 2015 13:42:58 +0000 (09:42 -0400)
[ Upstream commit 283cb41f426b723a0255702b761b0fc5d1b53a81 ]

The cpuset.sched_relax_domain_level can control how far we do
immediate load balancing on a system. However, it was found on recent
kernels that echo'ing a value into cpuset.sched_relax_domain_level
did not reduce any immediate load balancing.

The reason this occurred was because the update_domain_attr_tree() traversal
did not update for the "top_cpuset". This resulted in nothing being changed
when modifying the sched_relax_domain_level parameter.

This patch is able to address that problem by having update_domain_attr_tree()
allow updates for the root in the cpuset traversal.

Fixes: fc560a26acce ("cpuset: replace cpuset->stack_list with cpuset_for_each_descendant_pre()")
Cc: <stable@vger.kernel.org> # 3.9+
Signed-off-by: Jason Low <jason.low2@hp.com>
Signed-off-by: Zefan Li <lizefan@huawei.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Tested-by: Serge Hallyn <serge.hallyn@canonical.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
kernel/cpuset.c

index d4921624a089bbf922fc1bd10b3fc98d019cc054..672310e1597e565b3c90a8261e69d8d8bed67ab1 100644 (file)
@@ -538,9 +538,6 @@ static void update_domain_attr_tree(struct sched_domain_attr *dattr,
 
        rcu_read_lock();
        cpuset_for_each_descendant_pre(cp, pos_css, root_cs) {
-               if (cp == root_cs)
-                       continue;
-
                /* skip the whole subtree if @cp doesn't have any CPU */
                if (cpumask_empty(cp->cpus_allowed)) {
                        pos_css = css_rightmost_descendant(pos_css);