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

[Tejun Heo]

Hello,

Sorry about the delay.  Was traveling.

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.

> 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.
...
> Signed-off-by: Tang Chen <tangchen@xxxxxxxxxxxxxx>
> Signed-off-by: Miao Xie <miaox@xxxxxxxxxxxxxx>

After separating out the fix part into a separate patch, please add
Cc: stable@xxxxxxxxxxxxxxxx

>  void cmci_rediscover(int dying)
>  {
> -	int banks;
> -	int cpu;
> -	cpumask_var_t old;
> +	int cpu, banks;
>  
>  	if (!cmci_supported(&banks))
>  		return;
> -	if (!alloc_cpumask_var(&old, GFP_KERNEL))
> -		return;
> -	cpumask_copy(old, &current->cpus_allowed);
>  
>  	for_each_online_cpu(cpu) {
> -		if (cpu == dying)
> -			continue;
> -		if (set_cpus_allowed_ptr(current, cpumask_of(cpu)))
> +		BUG_ON(cpu == dying);
> +
> +		if (cpu == smp_processor_id()) {
> +			cmci_rediscover_work_func(NULL);
>  			continue;
> -		/* Recheck banks in case CPUs don't all have the same */
> -		if (cmci_supported(&banks))
> -			cmci_discover(banks, 0);
> -	}
> +		}
>  
> -	set_cpus_allowed_ptr(current, old);
> -	free_cpumask_var(old);
> +		work_on_cpu(cpu, cmci_rediscover_work_func, NULL);
> +	}

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

Thanks!

-- 
tejun
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/