Re: [PATCH v3 1/2] rcu: Call rcu_nocb_rdp_deoffload() directly after rcuog/op kthreads spawn failed

[Frederic Weisbecker]

On Fri, Mar 18, 2022 at 04:07:18PM +0800, Zqiang wrote:
> @@ -1009,10 +1015,33 @@ static long rcu_nocb_rdp_deoffload(void *arg)
>  	 */
>  	rcu_segcblist_set_flags(cblist, SEGCBLIST_RCU_CORE);
>  	invoke_rcu_core();
> -	ret = rdp_offload_toggle(rdp, false, flags);
> -	swait_event_exclusive(rdp->nocb_state_wq,
> -			      !rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB |
> -							SEGCBLIST_KTHREAD_GP));
> +	wake_gp = rdp_offload_toggle(rdp, false, flags);
> +
> +	mutex_lock(&rdp_gp->nocb_gp_kthread_mutex);
> +	if (rdp_gp->nocb_gp_kthread) {
> +		if (wake_gp)
> +			wake_up_process(rdp_gp->nocb_gp_kthread);
> +
> +		if (rdp->nocb_cb_kthread) {
> +			condition = SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP;
> +		} else {
> +			/*
> +			 *If rcuop kthread spawn failed, direct remove SEGCBLIST_KTHREAD_CB
> +			 *just wait SEGCBLIST_KTHREAD_GP to be cleared.
> +			 */
> +			condition = SEGCBLIST_KTHREAD_GP;
> +			rcu_segcblist_clear_flags(&rdp->cblist,
> SEGCBLIST_KTHREAD_CB);

You may be running concurrently againt nocb_gp_wait() ->
nocb_gp_update_state_deoffloading() -> rcu_segcblist_clear_flags(cblist,
SEGCBLIST_KTHREAD_GP)

So you need to protect the flags clear with rcu_nocb lock.




> +		}
> +		swait_event_exclusive(rdp->nocb_state_wq,
> +					!rcu_segcblist_test_flags(cblist, condition));
> +	} else {
> +		rcu_nocb_lock_irqsave(rdp, flags);
> +		rcu_segcblist_clear_flags(&rdp->cblist,
> +				SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP);
> +		rcu_nocb_unlock_irqrestore(rdp, flags);

Like you're doing here.

Thanks!