Re: [PATCH v2] wfcqueue: functions for local append and enqueue

[Mathieu Desnoyers]

* Eric Wong (normalperson@xxxxxxxx) wrote:
> With level-triggered epoll, append/enqueue operations to the
> local/locked queues increase performance by avoiding unnecessary atomic
> operations and barriers.  These are necessary to avoid performance
> regressions when looping through ep_send_events and appending many
> items to a local queue.
> 
> Signed-off-by: Eric Wong <normalperson@xxxxxxxx>
> Cc: Mathieu Desnoyers <mathieu.desnoyers@xxxxxxxxxxxx>
> Cc: Lai Jiangshan <laijs@xxxxxxxxxxxxxx>
> Cc: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
> Cc: Stephen Hemminger <shemminger@xxxxxxxxxx>
> Cc: Davide Libenzi <davidel@xxxxxxxxxxxxxxx>
> ---
>  I noticed the original __wfcq_append function was not in the
>  synchronization table, so I left out ___wfcq_append from the table
>  as well.
> 
>  include/linux/wfcqueue.h | 59 ++++++++++++++++++++++++++++++++++++++++++------
>  1 file changed, 52 insertions(+), 7 deletions(-)
> 
> diff --git a/include/linux/wfcqueue.h b/include/linux/wfcqueue.h
> index 9464a0c..800b1dd 100644
> --- a/include/linux/wfcqueue.h
> +++ b/include/linux/wfcqueue.h
> @@ -55,14 +55,16 @@
>   * [4] __wfcq_splice (source queue)
>   * [5] __wfcq_first
>   * [6] __wfcq_next
> + * [7] __wfcq_enqueue
>   *
> - *     [1] [2] [3] [4] [5] [6]
> - * [1]  -   -   -   -   -   -
> - * [2]  -   -   -   -   -   -
> - * [3]  -   -   X   X   X   X
> - * [4]  -   -   X   -   X   X
> - * [5]  -   -   X   X   -   -
> - * [6]  -   -   X   X   -   -
> + *     [1] [2] [3] [4] [5] [6] [7]
> + * [1]  -   -   -   -   -   -   X
> + * [2]  -   -   -   -   -   -   X
> + * [3]  -   -   X   X   X   X   X
> + * [4]  -   -   X   -   X   X   X
> + * [5]  -   -   X   X   -   -   X
> + * [6]  -   -   X   X   -   -   X
> + * [7]  X   X   X   X   X   X   X
>   *
>   * Besides locking, mutual exclusion of dequeue, splice and iteration
>   * can be ensured by performing all of those operations from a single
> @@ -205,6 +207,49 @@ static inline bool wfcq_enqueue(struct wfcq_head *head,
>  }
>  
>  /*
> + * ___wfcq_append: append one local queue to another local queue
> + *
> + * No memory barriers are issued.  Mutual exclusion is the responsibility
> + * of the caller.
> + *
> + * Returns false if the queue was empty prior to adding the node.
> + * Returns true otherwise.
> + */

__wfcq_append() and ___wfcq_append() are meant to be private to
wfcqueue.h. Therefore, the comment above should be removed, since it is
not part of the API.

I notice that I should have used ___wfcq_append() for the original
append function for consistency (other private helpers in this header
are prefixed with ___).

So maybe we should rename __wfcq_append to ___wfcq_append (making it
clear that it is a private helper), and introduce your helper as
___wfcq_append_local() (I don't care about having "local" in there since
it is not part of the exposed API).

> +static inline bool ___wfcq_append(struct wfcq_head *head,
> +		struct wfcq_tail *tail,
> +		struct wfcq_node *new_head,
> +		struct wfcq_node *new_tail)
> +{
> +	struct wfcq_node *old_tail;
> +
> +	old_tail = tail->p;
> +	tail->p = new_tail;
> +	old_tail->next = new_head;
> +
> +	/*
> +	 * Return false if queue was empty prior to adding the node,
> +	 * else return true.
> +	 */
> +	return old_tail != &head->node;
> +}
> +
> +/*
> + * __wfcq_enqueue: enqueue a node into a local queue

The concept of "local queue" is not clearly defined.

Perhaps it would be clearer to state:

 * __wfcq_enqueue: enqueue a node into a queue, requiring mutual exclusion.

Thoughts ?

Thanks,

Mathieu


> + *
> + * No memory barriers are issued.  Mutual exclusion is the responsibility
> + * of the caller.
> + *
> + * Returns false if the queue was empty prior to adding the node.
> + * Returns true otherwise.
> + */
> +static inline bool __wfcq_enqueue(struct wfcq_head *head,
> +		struct wfcq_tail *tail,
> +		struct wfcq_node *new_tail)
> +{
> +	return ___wfcq_append(head, tail, new_tail, new_tail);
> +}
> +
> +/*
>   * ___wfcq_busy_wait: busy-wait.
>   */
>  static inline void ___wfcq_busy_wait(void)
> -- 
> 1.8.2.rc3.2.geae6cf5
> 

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/