Re: NOHZ: WARNING: at arch/x86/kernel/smp.c:123 native_smp_send_reschedule, round 2

[Michael Wang <wangyun@linux.vnet.ibm.com>]

On 05/20/2013 05:09 PM, Viresh Kumar wrote:
> On 20 May 2013 14:26, Michael Wang <wangyun@linux.vnet.ibm.com> wrote:
>> On 05/20/2013 03:25 PM, Michael Wang wrote:
>>> Yeah, that's right, I guess the issue is, although the policy->cpus is
>>> correct at a given time, after get cpu from it, it's possible to be
>>> changed, unless we disabled preempt or irq, or hotplug before we use it...
>>>
>>> Like such issue cases:
>>>                               get x from policy->cpus
>>>       DOWN notifier
>>>       change policy->cpus
>>>       do offline x
>>>                               send ipi to x
>>>
>>> Will that happen?
> 
> Sorry I am not sure. :(
> 
> I can see mutex being used in cpufreq_governor.c which should take care
> of race conditions...
> 
>> May be we could do some test to confirm it?
>>
>> diff --git a/drivers/cpufreq/cpufreq_governor.c
>> b/drivers/cpufreq/cpufreq_governor.c
>> index 443442d..449be88 100644
>> --- a/drivers/cpufreq/cpufreq_governor.c
>> +++ b/drivers/cpufreq/cpufreq_governor.c
>> @@ -26,6 +26,7 @@
>>  #include <linux/tick.h>
>>  #include <linux/types.h>
>>  #include <linux/workqueue.h>
>> +#include <linux/cpu.h>
>>
>>  #include "cpufreq_governor.h"
>>
>> @@ -180,8 +181,10 @@ void gov_queue_work(struct dbs_data *dbs_data,
>> struct cpufreq_policy *policy,
>>         if (!all_cpus) {
>>                 __gov_queue_work(smp_processor_id(), dbs_data, delay);
>>         } else {
>> +               get_online_cpus();
>>                 for_each_cpu(i, policy->cpus)
>>                         __gov_queue_work(i, dbs_data, delay);
>> +               put_online_cpus();
>>         }
>>  }
>>  EXPORT_SYMBOL_GPL(gov_queue_work);
>>
>> This is supposed to make WARN disappear, if it works, then BINGO :)
> 
> Let people test it and then we can talk :)

Agree :)

Borislav, would you like to take a try?

If this fix cause other troubles, you could try get_cpu() or disable irq
also.

Regards,
Michael Wang

>