The memory barriers in list_del_init_careful() and list_empty_careful()
in pairs already handle the proper ordering between data.got_token
and data.wq.entry. So remove the redundant explicit barriers. And also
change a "break" statement to "return" to avoid redundant calling of
finish_wait().

Signed-off-by: Muchun Song <songmuchun@bytedance.com>
---
 block/blk-rq-qos.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/block/blk-rq-qos.c b/block/blk-rq-qos.c
index dc510f493ba57..9b0aa7dd6779f 100644
--- a/block/blk-rq-qos.c
+++ b/block/blk-rq-qos.c
@@ -218,7 +218,6 @@ static int rq_qos_wake_function(struct wait_queue_entry *curr,
 		return -1;
 
 	data->got_token = true;
-	smp_wmb();
 	wake_up_process(data->task);
 	list_del_init_careful(&curr->entry);
 	return 1;
@@ -274,10 +273,9 @@ void rq_qos_wait(struct rq_wait *rqw, void *private_data,
 			 * which means we now have two. Put our local token
 			 * and wake anyone else potentially waiting for one.
 			 */
-			smp_rmb();
 			if (data.got_token)
 				cleanup_cb(rqw, private_data);
-			break;
+			return;
 		}
 		io_schedule();
 		has_sleeper = true;
-- 
2.20.1

On Mon, 21 Oct 2024 16:52:51 +0800, Muchun Song wrote:
> The memory barriers in list_del_init_careful() and list_empty_careful()
> in pairs already handle the proper ordering between data.got_token
> and data.wq.entry. So remove the redundant explicit barriers. And also
> change a "break" statement to "return" to avoid redundant calling of
> finish_wait().
> 
> 
> [...]

Applied, thanks!

[1/1] block: remove redundant explicit memory barrier from rq_qos waiter and waker
      commit: 904ebd2527c507752f5ddb358f887d2e0dab96a0

Best regards,
-- 
Jens Axboe

On 2024/10/21 16:52, Muchun Song wrote:
> The memory barriers in list_del_init_careful() and list_empty_careful()
> in pairs already handle the proper ordering between data.got_token
> and data.wq.entry. So remove the redundant explicit barriers. And also
> change a "break" statement to "return" to avoid redundant calling of
> finish_wait().
> 
> Signed-off-by: Muchun Song <songmuchun@bytedance.com>

Good catch! Just a small nit below, feel free to add:

Reviewed-by: Chengming Zhou <chengming.zhou@linux.dev>

> ---
>   block/blk-rq-qos.c | 4 +---
>   1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/block/blk-rq-qos.c b/block/blk-rq-qos.c
> index dc510f493ba57..9b0aa7dd6779f 100644
> --- a/block/blk-rq-qos.c
> +++ b/block/blk-rq-qos.c
> @@ -218,7 +218,6 @@ static int rq_qos_wake_function(struct wait_queue_entry *curr,
>   		return -1;
>   
>   	data->got_token = true;
> -	smp_wmb();
>   	wake_up_process(data->task);
>   	list_del_init_careful(&curr->entry);
>   	return 1;
> @@ -274,10 +273,9 @@ void rq_qos_wait(struct rq_wait *rqw, void *private_data,
>   			 * which means we now have two. Put our local token
>   			 * and wake anyone else potentially waiting for one.
>   			 */
> -			smp_rmb();
>   			if (data.got_token)
>   				cleanup_cb(rqw, private_data);
> -			break;
> +			return;
>   		}

Would it be better to move this acquire_inflight_cb() above out of
the do-while(1) since we rely on the waker to get inflight counter
for us?

Thanks.

>   		io_schedule();
>   		has_sleeper = true;

> On Oct 22, 2024, at 15:53, Chengming Zhou <chengming.zhou@linux.dev> wrote:
> 
> On 2024/10/21 16:52, Muchun Song wrote:
>> The memory barriers in list_del_init_careful() and list_empty_careful()
>> in pairs already handle the proper ordering between data.got_token
>> and data.wq.entry. So remove the redundant explicit barriers. And also
>> change a "break" statement to "return" to avoid redundant calling of
>> finish_wait().
>> Signed-off-by: Muchun Song <songmuchun@bytedance.com>
> 
> Good catch! Just a small nit below, feel free to add:
> 
> Reviewed-by: Chengming Zhou <chengming.zhou@linux.dev>
> 
>> ---
>>  block/blk-rq-qos.c | 4 +---
>>  1 file changed, 1 insertion(+), 3 deletions(-)
>> diff --git a/block/blk-rq-qos.c b/block/blk-rq-qos.c
>> index dc510f493ba57..9b0aa7dd6779f 100644
>> --- a/block/blk-rq-qos.c
>> +++ b/block/blk-rq-qos.c
>> @@ -218,7 +218,6 @@ static int rq_qos_wake_function(struct wait_queue_entry *curr,
>>   		return -1;
>>     	data->got_token = true;
>> - 	smp_wmb();
>>   	wake_up_process(data->task);
>>   	list_del_init_careful(&curr->entry);
>>   	return 1;
>> @@ -274,10 +273,9 @@ void rq_qos_wait(struct rq_wait *rqw, void *private_data,
>>  			 * which means we now have two. Put our local token
>>  			 * and wake anyone else potentially waiting for one.
>>  			 */
>> - 			smp_rmb();
>>   			if (data.got_token)
>>   				cleanup_cb(rqw, private_data);
>> - 			break;
>> + 			return;
>>   		}
> 
> Would it be better to move this acquire_inflight_cb() above out of
> the do-while(1) since we rely on the waker to get inflight counter
> for us?

I also noticed about this and I am working on this. Will send a separate
patch for this refactoring later.

Thanks.

> 
> Thanks.
> 
>>   io_schedule();
>>   has_sleeper = true;

On 10/21/24 2:52 AM, Muchun Song wrote:
> The memory barriers in list_del_init_careful() and list_empty_careful()
> in pairs already handle the proper ordering between data.got_token
> and data.wq.entry. So remove the redundant explicit barriers. And also
> change a "break" statement to "return" to avoid redundant calling of
> finish_wait().

Not sure why you didn't CC Omar on this one, as he literally just last
week fixed an issue related to this.

> 
> Signed-off-by: Muchun Song <songmuchun@bytedance.com>
> ---
>  block/blk-rq-qos.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
> 
> diff --git a/block/blk-rq-qos.c b/block/blk-rq-qos.c
> index dc510f493ba57..9b0aa7dd6779f 100644
> --- a/block/blk-rq-qos.c
> +++ b/block/blk-rq-qos.c
> @@ -218,7 +218,6 @@ static int rq_qos_wake_function(struct wait_queue_entry *curr,
>  		return -1;
>  
>  	data->got_token = true;
> -	smp_wmb();
>  	wake_up_process(data->task);
>  	list_del_init_careful(&curr->entry);
>  	return 1;
> @@ -274,10 +273,9 @@ void rq_qos_wait(struct rq_wait *rqw, void *private_data,
>  			 * which means we now have two. Put our local token
>  			 * and wake anyone else potentially waiting for one.
>  			 */
> -			smp_rmb();
>  			if (data.got_token)
>  				cleanup_cb(rqw, private_data);
> -			break;
> +			return;
>  		}
>  		io_schedule();
>  		has_sleeper = true;


-- 
Jens Axboe

> On Oct 21, 2024, at 21:45, Jens Axboe <axboe@kernel.dk> wrote:
> 
> On 10/21/24 2:52 AM, Muchun Song wrote:
>> The memory barriers in list_del_init_careful() and list_empty_careful()
>> in pairs already handle the proper ordering between data.got_token
>> and data.wq.entry. So remove the redundant explicit barriers. And also
>> change a "break" statement to "return" to avoid redundant calling of
>> finish_wait().
> 
> Not sure why you didn't CC Omar on this one, as he literally just last
> week fixed an issue related to this.

Hi Jens,

Yes. I only CC the author of patch of adding the barriers, I thought
they should be more confident about this. Thanks for your reminder.
I saw Omar's great fix. And thanks for you help me CC Omar. I think
he'll be also suitable for commenting on this patch.

Muchun,
Thanks.

On Tue, Oct 22, 2024 at 02:31:53PM +0800, Muchun Song wrote:
> 
> 
> > On Oct 21, 2024, at 21:45, Jens Axboe <axboe@kernel.dk> wrote:
> > 
> > On 10/21/24 2:52 AM, Muchun Song wrote:
> >> The memory barriers in list_del_init_careful() and list_empty_careful()
> >> in pairs already handle the proper ordering between data.got_token
> >> and data.wq.entry. So remove the redundant explicit barriers. And also
> >> change a "break" statement to "return" to avoid redundant calling of
> >> finish_wait().
> > 
> > Not sure why you didn't CC Omar on this one, as he literally just last
> > week fixed an issue related to this.
> 
> Hi Jens,
> 
> Yes. I only CC the author of patch of adding the barriers, I thought
> they should be more confident about this. Thanks for your reminder.
> I saw Omar's great fix. And thanks for you help me CC Omar. I think
> he'll be also suitable for commenting on this patch.
> 
> Muchun,
> Thanks.

Well there goes my streak of not reading memory-barriers.txt for a few
months...

This looks fine to me. wake_up_process() also implies a full memory
barrier, so I that smp_wmb() was extra redundant.

Reviewed-by: Omar Sandoval <osandov@fb.com>