Re: [PATCH] workqueue: fix a workqueue kernel panic issue.

[Tejun Heo <tj@kernel.org>]

On Sun, Sep 21, 2014 at 08:30:32PM -0700, Yifan Zhang wrote:
> Hi Tejun,
> 
> What's do you think of this patch ? Any concern ?

Hmmm?  Haven't I already responded to this patch?

> -----Original Message-----
> From: Yifan Zhang [mailto:zhangyf@marvell.com] 
> Sent: 2014年9月17日 16:18
> To: Tejun Heo; Jing Xiang; linux-kernel@vger.kernel.org
> Cc: Yifan Zhang
> Subject: [PATCH] workqueue: fix a workqueue kernel panic issue.
> 
> if created workqueue in multi-thread unsynchronized,

Can you please elaborate?

> get_work_pwq() may return NULL, which cause kernel panic. Judge get_work_pwq() return value before use
> pwq->wq->flags.
> 
> Signed-off-by: Yifan Zhang <zhangyf@marvell.com>
> ---
>  kernel/workqueue.c |   12 +++++++++++-
>  1 file changed, 11 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/workqueue.c b/kernel/workqueue.c index 5dbe22a..d3ac87f 100644
> --- a/kernel/workqueue.c
> +++ b/kernel/workqueue.c
> @@ -1947,9 +1947,19 @@ __acquires(&pool->lock)  {
>  	struct pool_workqueue *pwq = get_work_pwq(work);
>  	struct worker_pool *pool = worker->pool;
> -	bool cpu_intensive = pwq->wq->flags & WQ_CPU_INTENSIVE;
> +	bool cpu_intensive;
>  	int work_color;
>  	struct worker *collision;
> +
> +	if (pwq == NULL) {
> +		pr_err("BUG: invalid struct work_struct.data %lu\n",
> +				atomic_long_read(&work->data));
> +		dump_stack();
> +		return;

I have difficult time seeing how the above piece of code would be
acceptable but maybe the situation you're trying to explain is weird
enough to justify it.

Thanks.

-- 
tejun