[PATCH 1/2] cpuset: use rebuild_sched_domains() in cpuset_hotplug_workfn()

[PATCH 1/2] cpuset: use rebuild_sched_domains() in cpuset_hotplug_workfn()

[Li Zefan]

From: Li Zhong <zhong-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>

In cpuset_hotplug_workfn(), partition_sched_domains() is called without
hotplug lock held, which is actually needed (stated in the function
header of partition_sched_domains()).

This patch tries to use rebuild_sched_domains() to solve the above
issue, and makes the code looks a little simpler.

Signed-off-by: Li Zhong <zhong-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
Signed-off-by: Li Zefan <lizefan-hv44wF8Li93QT0dZR+AlfA@public.gmane.org>
---
 kernel/cpuset.c | 13 ++-----------
 1 file changed, 2 insertions(+), 11 deletions(-)

diff --git a/kernel/cpuset.c b/kernel/cpuset.c
index 943968d..b0f18ba 100644
--- a/kernel/cpuset.c
+++ b/kernel/cpuset.c
@@ -2184,17 +2184,8 @@ static void cpuset_hotplug_workfn(struct work_struct *work)
 	flush_workqueue(cpuset_propagate_hotplug_wq);
 
 	/* rebuild sched domains if cpus_allowed has changed */
-	if (cpus_updated) {
-		struct sched_domain_attr *attr;
-		cpumask_var_t *doms;
-		int ndoms;
-
-		mutex_lock(&cpuset_mutex);
-		ndoms = generate_sched_domains(&doms, &attr);
-		mutex_unlock(&cpuset_mutex);
-
-		partition_sched_domains(ndoms, doms, attr);
-	}
+	if (cpus_updated)
+		rebuild_sched_domains();
 }
 
 void cpuset_update_active_cpus(bool cpu_online)
-- 
1.8.0.2

[PATCH 2/2] cpuset: fix cpu hotplug vs rebuild_sched_domains() race

[Li Zefan]

rebuild_sched_domains() might pass doms with offlined cpu to
partition_sched_domains(), which results in an oops:

general protection fault: 0000 [#1] SMP
...
RIP: 0010:[<ffffffff81077a1e>]  [<ffffffff81077a1e>] get_group+0x6e/0x90
...
Call Trace:
 [<ffffffff8107f07c>] build_sched_domains+0x70c/0xcb0
 [<ffffffff8107f2a7>] ? build_sched_domains+0x937/0xcb0
 [<ffffffff81173f64>] ? kfree+0xe4/0x1b0
 [<ffffffff8107f6e0>] ? partition_sched_domains+0xc0/0x470
 [<ffffffff8107f905>] partition_sched_domains+0x2e5/0x470
 [<ffffffff8107f6e0>] ? partition_sched_domains+0xc0/0x470
 [<ffffffff810c9007>] ? generate_sched_domains+0xc7/0x530
 [<ffffffff810c94a8>] rebuild_sched_domains_locked+0x38/0x70
 [<ffffffff810cb4a4>] cpuset_write_resmask+0x1a4/0x500
 [<ffffffff810c8700>] ? cpuset_mount+0xe0/0xe0
 [<ffffffff810c7f50>] ? cpuset_read_u64+0x100/0x100
 [<ffffffff810be890>] ? cgroup_iter_next+0x90/0x90
 [<ffffffff810cb300>] ? cpuset_css_offline+0x70/0x70
 [<ffffffff810c1a73>] cgroup_file_write+0x133/0x2e0
 [<ffffffff8118995b>] vfs_write+0xcb/0x130
 [<ffffffff8118a174>] sys_write+0x64/0xa0

Reported-by: Li Zhong <zhong-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
Signed-off-by: Li Zefan <lizefan-hv44wF8Li93QT0dZR+AlfA@public.gmane.org>
---
 kernel/cpuset.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/kernel/cpuset.c b/kernel/cpuset.c
index b0f18ba..ef05901 100644
--- a/kernel/cpuset.c
+++ b/kernel/cpuset.c
@@ -769,12 +769,20 @@ static void rebuild_sched_domains_locked(void)
 	lockdep_assert_held(&cpuset_mutex);
 	get_online_cpus();
 
+	/*
+	 * We have raced with CPU hotplug. Don't do anything to avoid
+	 * passing doms with offlined cpu to partition_sched_domains().
+	 * Anyways, hotplug work item will rebuild sched domains.
+	 */
+	if (!cpumask_equal(top_cpuset.cpus_allowed, cpu_active_mask))
+		goto out;
+
 	/* Generate domain masks and attrs */
 	ndoms = generate_sched_domains(&doms, &attr);
 
 	/* Have scheduler rebuild the domains */
 	partition_sched_domains(ndoms, doms, attr);
-
+out:
 	put_online_cpus();
 }
 #else /* !CONFIG_SMP */
-- 
1.8.0.2

Re: [PATCH 2/2] cpuset: fix cpu hotplug vs rebuild_sched_domains() race

[Tejun Heo]

On Sat, Apr 27, 2013 at 05:53:48PM +0800, Li Zefan wrote:
> rebuild_sched_domains() might pass doms with offlined cpu to
> partition_sched_domains(), which results in an oops:
> 
> general protection fault: 0000 [#1] SMP
> ...
> RIP: 0010:[<ffffffff81077a1e>]  [<ffffffff81077a1e>] get_group+0x6e/0x90
> ...
> Call Trace:
>  [<ffffffff8107f07c>] build_sched_domains+0x70c/0xcb0
>  [<ffffffff8107f2a7>] ? build_sched_domains+0x937/0xcb0
>  [<ffffffff81173f64>] ? kfree+0xe4/0x1b0
>  [<ffffffff8107f6e0>] ? partition_sched_domains+0xc0/0x470
>  [<ffffffff8107f905>] partition_sched_domains+0x2e5/0x470
>  [<ffffffff8107f6e0>] ? partition_sched_domains+0xc0/0x470
>  [<ffffffff810c9007>] ? generate_sched_domains+0xc7/0x530
>  [<ffffffff810c94a8>] rebuild_sched_domains_locked+0x38/0x70
>  [<ffffffff810cb4a4>] cpuset_write_resmask+0x1a4/0x500
>  [<ffffffff810c8700>] ? cpuset_mount+0xe0/0xe0
>  [<ffffffff810c7f50>] ? cpuset_read_u64+0x100/0x100
>  [<ffffffff810be890>] ? cgroup_iter_next+0x90/0x90
>  [<ffffffff810cb300>] ? cpuset_css_offline+0x70/0x70
>  [<ffffffff810c1a73>] cgroup_file_write+0x133/0x2e0
>  [<ffffffff8118995b>] vfs_write+0xcb/0x130
>  [<ffffffff8118a174>] sys_write+0x64/0xa0
> 
> Reported-by: Li Zhong <zhong-23VcF4HTsmIX0ybBhKVfKdBPR1lH4CV8@public.gmane.org>
> Signed-off-by: Li Zefan <lizefan-hv44wF8Li93QT0dZR+AlfA@public.gmane.org>

Applied 1-2 to cgroup/for-3.10.

Thanks.

-- 
tejun