Re: [PATCH 6/8] RCU, workqueue: Implement rcu_work

[Lai Jiangshan]

On Tue, Mar 20, 2018 at 12:45 AM, Tejun Heo <tj@xxxxxxxxxx> wrote:
> Hello, Lai.
>
> On Fri, Mar 16, 2018 at 02:01:35PM +0800, Lai Jiangshan wrote:
>> > +bool flush_rcu_work(struct rcu_work *rwork)
>> > +{
>> > +       if (test_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(&rwork->work))) {
>> > +               rcu_barrier();
>> > +               flush_work(&rwork->work);
>> > +               return true;
>>
>> A possible tiny improvement: check if it was already queued on wq.
>> For example:
>>
>>        if (test_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(&rwork->work))) {
>>                if (!flush_work(&rwork->work)) {
>>                       rcu_barrier();
>>                       flush_work(&rwork->work);
>>                }
>>                return true;
>
> But this breaks the guarantee that flush_work waits for the latest
> queueing instance.  Please consider the following scenario.

Oh, I'm sorry I was wrong. It is so evident that
"flush_work(&rwork->work) return true"
doesn't equals to "it has been queued on wq".

To detect if "it has been queued on wq" requires
a bunch of code. It is not worthy to complicate
this function.

>
>
>  1. rcu-work is queued
>  2. rcu-work starts executing
>                                 3. rcu-work is queued again
>                                 4. rcu-work is flushed
>  5. execution finishes
>                                 6. flush finishes
>                                 7. execution finishes
>
> 6 should happen after 7 but it didn't.
>
> Thanks.
>
> --
> tejun