Re: [PATCH v7 2/3] sched/task: Add the put_task_struct_atomic_safe() function

[Valentin Schneider]

On 04/05/23 10:42, Peter Zijlstra wrote:
> On Tue, Apr 25, 2023 at 08:43:02AM -0300, Wander Lairson Costa wrote:
>> diff --git a/include/linux/sched/task.h b/include/linux/sched/task.h
>> index b597b97b1f8f..cf774b83b2ec 100644
>> --- a/include/linux/sched/task.h
>> +++ b/include/linux/sched/task.h
>> @@ -141,6 +141,41 @@ static inline void put_task_struct_many(struct task_struct *t, int nr)
>>
>>  void put_task_struct_rcu_user(struct task_struct *task);
>>
>> +extern void __delayed_put_task_struct(struct rcu_head *rhp);
>> +
>> +static inline void put_task_struct_atomic_safe(struct task_struct *task)
>> +{
>> +	if (IS_ENABLED(CONFIG_PREEMPT_RT)) {
>> +		/*
>> +		 * Decrement the refcount explicitly to avoid unnecessarily
>> +		 * calling call_rcu.
>> +		 */
>> +		if (refcount_dec_and_test(&task->usage))
>> +			/*
>> +			 * under PREEMPT_RT, we can't call put_task_struct
>> +			 * in atomic context because it will indirectly
>> +			 * acquire sleeping locks.
>> +			 * call_rcu() will schedule __delayed_put_task_struct()
>> +			 * to be called in process context.
>> +			 *
>> +			 * __put_task_struct() is called when
>> +			 * refcount_dec_and_test(&t->usage) succeeds.
>> +			 *
>> +			 * This means that it can't conflict with
>> +			 * put_task_struct_rcu_user() which abuses ->rcu the same
>> +			 * way; rcu_users has a reference so task->usage can't be
>> +			 * zero after rcu_users 1 -> 0 transition.
>> +			 *
>> +			 * delayed_free_task() also uses ->rcu, but it is only called
>> +			 * when it fails to fork a process. Therefore, there is no
>> +			 * way it can conflict with put_task_struct().
>> +			 */
>> +			call_rcu(&task->rcu, __delayed_put_task_struct);
>> +	} else {
>> +		put_task_struct(task);
>> +	}
>> +}
>
> Urgh.. that's plenty horrible. And I'm sure everybody plus kitchen sink
> has already asked why can't we just rcu free the thing unconditionally.
>
> Google only found me an earlier version of this same patch set, but I'm
> sure we've had that discussion many times over the past several years.
> The above and your follow up patch is just horrible.
>

So on v3/v4 we got to doing that unconditionally for PREEMPT_RT, but per
[1] Wander went back to hand-fixing the problematic callsites.

Now that I'm looking at it again, I couldn't find a concrete argument from
Oleg against doing this unconditionally - as Wander is pointing out in the
changelog and comments, reusing task_struct.rcu for that purpose is safe
(although not necessarily obviously so).

Is this just miscommunication, or is there a genuine issue with doing this
unconditionally? As argued before, I'd also much rather have this be an
unconditional call_rcu() (regardless of context or PREEMPT_RT).