On Tue, Oct 24, 2017 at 9:18 AM, Li Bin <huawei.li...@huawei.com> wrote:
> When queue_work() is used in irq handler, there is a potential
> case that trigger NULL pointer dereference.
> ----------------------------------------------------------------
> worker_thread()
> |-spin_lock_irq()
> |-process_one_work()
> |-worker->current_pwq = pwq
> |-spin_unlock_irq()
> |-worker->current_func(work)
> |-spin_lock_irq()
> |-worker->current_pwq = NULL
> |-spin_unlock_irq()
>
> //interrupt here
> |-irq_handler
> |-__queue_work()
> //assuming that the wq is
> draining
> |-is_chained_work(wq)
> |-current_wq_worker()
> //Here, 'current' is
> the interrupted worker!
>
> |-current->current_pwq is NULL here!
I remember that softirq can be invoked when irq_eixt(),
and in this case the current->current_pwq is also NULL
if __queue_work() is called in the soft irq.
So in_task() might be better than !in_irq() for the fix?
> |-schedule()
> ----------------------------------------------------------------
>
> Avoid it by checking for irq context in current_wq_worker(), and
> if in irq context, we shouldn't use the 'current' to check the
> condition.
>
> Reported-by: Xiaofei Tan <tanxiao...@huawei.com>
> Signed-off-by: Li Bin <huawei.li...@huawei.com>
> ---
> kernel/workqueue_internal.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/workqueue_internal.h b/kernel/workqueue_internal.h
> index 8635417..d81cb9b 100644
> --- a/kernel/workqueue_internal.h
> +++ b/kernel/workqueue_internal.h
> @@ -9,6 +9,7 @@
>
> #include <linux/workqueue.h>
> #include <linux/kthread.h>
> +#include <linux/preempt.h>
>
> struct worker_pool;
>
> @@ -59,7 +60,7 @@ struct worker {
> */
> static inline struct worker *current_wq_worker(void)
> {
> - if (current->flags & PF_WQ_WORKER)
> + if (!in_irq() && (current->flags & PF_WQ_WORKER))
> return kthread_data(current);
> return NULL;
> }
> --
> 1.7.12.4
>
