Re: [PATCH] Do not change worker's running cpu in cmci_rediscover().

[Tang Chen <tangchen@cn.fujitsu.com>]

On 10/17/2012 07:08 AM, Tejun Heo wrote:
> On Thu, Sep 27, 2012 at 05:19:46PM +0800, Tang Chen wrote:
>> 1. cmci_rediscover() is only called by the CPU_POST_DEAD event handler, which
>> means the corresponding cpu has already dead. As a result, it won't be accessed
>> in the for_each_online_cpu loop.
>> So, we could change the if(cpu == dying) statement into a BUG_ON().
>
> Let's please move this to a separate patch and use WARN_ON_ONCE()
> instead.

OK, no problem. :)

>
>> 2. cmci_rediscover() used set_cpus_allowed_ptr() to change the current process's
>> running cpu, and migrate itself to the dest cpu. But worker processes are not
>> allowed to be migrated. If current is a worker, the worker will be migrated to
>> another cpu, but the corresponding  worker_pool is still on the original cpu.
>>
>> In this case, the following BUG_ON in try_to_wake_up_local() will be triggered:
>> BUG_ON(rq != this_rq());
>>
>> This will cause the kernel panic.
>>
>> This patch removes the set_cpus_allowed_ptr() call, and put the cmci rediscover
>> jobs onto all the other cpus using system_wq. This could bring some delay for
>> the jobs.

> After separating out the fix part into a separate patch, please add
> Cc: stable@vger.kernel.org.

OK, I will.

>
> The change looks correct to me but fails to apply to the current
> mainline.

I'll update it and resend it soon. :)
Thanks.

>
> Thanks!
>