Re: [PATCH-tip v7 09/20] locking/rwsem: Always release wait_lock before waking up tasks

[Waiman Long]

On 5/3/19 9:37 AM, Peter Zijlstra wrote:
> On Sun, Apr 28, 2019 at 05:25:46PM -0400, Waiman Long wrote:
>
>> +			/*
>> +			 * This waiter may have become first in the wait
>> +			 * list after re-acquring the wait_lock. The
>> +			 * rwsem_first_waiter() test in the main while
>> +			 * loop below will correctly detect that. We do
>> +			 * need to reload count to perform proper trylock
>> +			 * and avoid missed wakeup.
>> +			 */
>> +			count = atomic_long_read(&sem->count);
>> +		}
>>  	} else {
>>  		count = atomic_long_add_return(RWSEM_FLAG_WAITERS, &sem->count);
>>  	}
> I've been eyeing that count usage for the past few patches, and this
> here makes me think we should get rid of it.
>
> --- a/kernel/locking/rwsem.c
> +++ b/kernel/locking/rwsem.c
> @@ -400,13 +400,14 @@ static void __rwsem_mark_wake(struct rw_
>   * If wstate is WRITER_HANDOFF, it will make sure that either the handoff
>   * bit is set or the lock is acquired with handoff bit cleared.
>   */
> -static inline bool rwsem_try_write_lock(long count, struct rw_semaphore *sem,
> +static inline bool rwsem_try_write_lock(struct rw_semaphore *sem,
>  					enum writer_wait_state wstate)
>  {
> -	long new;
> +	long count, new;
>  
>  	lockdep_assert_held(&sem->wait_lock);
>  
> +	count = atomic_long_read(&sem->count);
>  	do {
>  		bool has_handoff = !!(count & RWSEM_FLAG_HANDOFF);
>  
> @@ -760,25 +761,16 @@ rwsem_down_write_slowpath(struct rw_sema
>  			wake_up_q(&wake_q);
>  			wake_q_init(&wake_q);	/* Used again, reinit */
>  			raw_spin_lock_irq(&sem->wait_lock);
> -			/*
> -			 * This waiter may have become first in the wait
> -			 * list after re-acquring the wait_lock. The
> -			 * rwsem_first_waiter() test in the main while
> -			 * loop below will correctly detect that. We do
> -			 * need to reload count to perform proper trylock
> -			 * and avoid missed wakeup.
> -			 */
> -			count = atomic_long_read(&sem->count);
>  		}
>  	} else {
> -		count = atomic_long_add_return(RWSEM_FLAG_WAITERS, &sem->count);
> +		atomic_long_or(RWSEM_FLAG_WAITERS, &sem->count);
>  	}
>  
>  wait:
>  	/* wait until we successfully acquire the lock */
>  	set_current_state(state);
>  	for (;;) {
> -		if (rwsem_try_write_lock(count, sem, wstate))
> +		if (rwsem_try_write_lock(sem, wstate))
>  			break;
>  
>  		raw_spin_unlock_irq(&sem->wait_lock);
> @@ -819,7 +811,6 @@ rwsem_down_write_slowpath(struct rw_sema
>  		}
>  
>  		raw_spin_lock_irq(&sem->wait_lock);
> -		count = atomic_long_read(&sem->count);
>  	}
>  	__set_current_state(TASK_RUNNING);
>  	list_del(&waiter.list);

Yes, this is an alternative way of doing it.

Cheers,
Longman