Re: [PATCH -next 7/8] block, bfq: cleanup bfq_bfqq_update_budg_for_activation()

[Jan Kara]

On Sat 14-05-22 17:05:21, Yu Kuai wrote:
> It will only be called from bfq_bfqq_handle_idle_busy_switch() in
> specific code branch, there is no need to precaculate
> 'bfqq_wants_to_preempt' each time bfq_bfqq_handle_idle_busy_switch()
> is caleld.
> 
> Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>

Please see below:

> @@ -1816,14 +1807,6 @@ static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd,
>  		  (bfqq->bic || RQ_BIC(rq)->stably_merged) &&
>  		   (*interactive || soft_rt)));
>  
> -	/*
> -	 * Using the last flag, update budget and check whether bfqq
> -	 * may want to preempt the in-service queue.
> -	 */
> -	bfqq_wants_to_preempt =
> -		bfq_bfqq_update_budg_for_activation(bfqd, bfqq,
> -						    arrived_in_time);
> -
>  	/*
>  	 * If bfqq happened to be activated in a burst, but has been
>  	 * idle for much more than an interactive queue, then we
...
> @@ -1918,7 +1900,7 @@ static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data *bfqd,
>  	 * (2) this switch of bfqq to busy changes the scenario.
>  	 */
>  	if (bfqd->in_service_queue &&
> -	    ((bfqq_wants_to_preempt &&
> +	    ((bfq_bfqq_update_budg_for_activation(bfqd, bfqq) &&
>  	      bfqq->wr_coeff >= bfqd->in_service_queue->wr_coeff) ||
>  	     bfq_bfqq_higher_class_or_weight(bfqq, bfqd->in_service_queue) ||
>  	     !bfq_better_to_idle(bfqd->in_service_queue)) &&

So these changes are actually wrong because
bfq_bfqq_update_budg_for_activation() relies on
bfq_bfqq_non_blocking_wait_rq() but bfq_add_bfqq_busy() clears that. And
bfq_add_bfqq_busy() is called between the place where
bfq_bfqq_update_budg_for_activation() was called previously and now so your
patch breaks this logic.

								Honza
-- 
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR