Re: [PATCH] Fix race between oom kill and task exit

[Johannes Weiner]

Cc William and azur who might have encountered this problem.

On Thu, Nov 28, 2013 at 05:09:16AM +0000, Ma, Xindong wrote:
> From: Leon Ma <xindong.ma@xxxxxxxxx>
> Date: Thu, 28 Nov 2013 12:46:09 +0800
> Subject: [PATCH] Fix race between oom kill and task exit
> 
> There is a race between oom kill and task exit. Scenario is:
>    TASK  A                      TASK  B
> TASK B is selected to oom kill
> in oom_kill_process()
> check PF_EXITING of TASK B
>                             task call do_exit()
>                             task set PF_EXITING flag
>                             write_lock_irq(&tasklist_lock);
>                             remove TASK B from thread group in __unhash_process()
>                             write_unlock_irq(&tasklist_lock);
> read_lock(&tasklist_lock);
> traverse threads of TASK B
> read_unlock(&tasklist_lock);
> 
> After that, the following traversal of threads in TASK B will not end because TASK B is not in the thread group:
> do {
> ....
> } while_each_thread(p, t);
> 
> Signed-off-by: Leon Ma <xindong.ma@xxxxxxxxx>
> Signed-off-by: xiaobing tu <xiaobing.tu@xxxxxxxxx>
> ---
>  mm/oom_kill.c |   20 ++++++++++----------
>  1 files changed, 10 insertions(+), 10 deletions(-)
> 
> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
> index 1e4a600..32ec88d 100644
> --- a/mm/oom_kill.c
> +++ b/mm/oom_kill.c
> @@ -412,16 +412,6 @@ void oom_kill_process(struct task_struct *p, gfp_t gfp_mask, int order,
>  	static DEFINE_RATELIMIT_STATE(oom_rs, DEFAULT_RATELIMIT_INTERVAL,
>  					      DEFAULT_RATELIMIT_BURST);
>  
> -	/*
> -	 * If the task is already exiting, don't alarm the sysadmin or kill
> -	 * its children or threads, just set TIF_MEMDIE so it can die quickly
> -	 */
> -	if (p->flags & PF_EXITING) {
> -		set_tsk_thread_flag(p, TIF_MEMDIE);
> -		put_task_struct(p);
> -		return;
> -	}
> -
>  	if (__ratelimit(&oom_rs))
>  		dump_header(p, gfp_mask, order, memcg, nodemask);
>  
> @@ -437,6 +427,16 @@ void oom_kill_process(struct task_struct *p, gfp_t gfp_mask, int order,
>  	 * still freeing memory.
>  	 */
>  	read_lock(&tasklist_lock);
> +	/*
> +	 * If the task is already exiting, don't alarm the sysadmin or kill
> +	 * its children or threads, just set TIF_MEMDIE so it can die quickly
> +	 */
> +	if (p->flags & PF_EXITING) {
> +		set_tsk_thread_flag(p, TIF_MEMDIE);
> +		put_task_struct(p);
> +		read_unlock(&tasklist_lock);
> +		return;
> +	}
>  	do {
>  		list_for_each_entry(child, &t->children, sibling) {
>  			unsigned int child_points;
> -- 
> 1.7.4.1
> 
--
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/