Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756053Ab0GGNal (ORCPT ); Wed, 7 Jul 2010 09:30:41 -0400 Received: from a.mx.secunet.com ([195.81.216.161]:47355 "EHLO a.mx.secunet.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755531Ab0GGNaj (ORCPT ); Wed, 7 Jul 2010 09:30:39 -0400 Date: Wed, 7 Jul 2010 15:32:39 +0200 From: Steffen Klassert To: Herbert Xu Cc: Dan Kruchinin , Andrew Morton , linux-crypto@vger.kernel.org, linux-kernel@vger.kernel.org Subject: [PATCH 5/6] padata: simplify serialization mechanism Message-ID: <20100707133239.GA10072@secunet.com> References: <20100707132915.GV10072@secunet.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20100707132915.GV10072@secunet.com> User-Agent: Mutt/1.5.20 (2009-06-14) X-OriginalArrivalTime: 07 Jul 2010 13:30:37.0341 (UTC) FILETIME=[95CA84D0:01CB1DD8] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5017 Lines: 172 We count the number of processed objects on a percpu basis, so we need to go through all the percpu reorder queues to calculate the sequence number of the next object that needs serialization. This patch changes this to count the number of processed objects global. So we can calculate the sequence number and the percpu reorder queue of the next object that needs serialization without searching through the percpu reorder queues. This avoids some accesses to memory of foreign cpus. Signed-off-by: Steffen Klassert --- include/linux/padata.h | 6 ++-- kernel/padata.c | 71 +++++++++++++----------------------------------- 2 files changed, 22 insertions(+), 55 deletions(-) diff --git a/include/linux/padata.h b/include/linux/padata.h index e4c17f9..8844b85 100644 --- a/include/linux/padata.h +++ b/include/linux/padata.h @@ -67,7 +67,6 @@ struct padata_list { * @pwork: work struct for parallelization. * @swork: work struct for serialization. * @pd: Backpointer to the internal control structure. - * @num_obj: Number of objects that are processed by this cpu. * @cpu_index: Index of the cpu. */ struct padata_queue { @@ -77,7 +76,6 @@ struct padata_queue { struct work_struct pwork; struct work_struct swork; struct parallel_data *pd; - atomic_t num_obj; int cpu_index; }; @@ -93,6 +91,7 @@ struct padata_queue { * @max_seq_nr: Maximal used sequence number. * @cpumask: cpumask in use. * @lock: Reorder lock. + * @processed: Number of already processed objects. * @timer: Reorder timer. */ struct parallel_data { @@ -103,7 +102,8 @@ struct parallel_data { atomic_t refcnt; unsigned int max_seq_nr; cpumask_var_t cpumask; - spinlock_t lock; + spinlock_t lock ____cacheline_aligned; + unsigned int processed; struct timer_list timer; }; diff --git a/kernel/padata.c b/kernel/padata.c index ae8defc..450d67d 100644 --- a/kernel/padata.c +++ b/kernel/padata.c @@ -170,79 +170,47 @@ EXPORT_SYMBOL(padata_do_parallel); */ static struct padata_priv *padata_get_next(struct parallel_data *pd) { - int cpu, num_cpus, empty, calc_seq_nr; - int seq_nr, next_nr, overrun, next_overrun; + int cpu, num_cpus; + int next_nr, next_index; struct padata_queue *queue, *next_queue; struct padata_priv *padata; struct padata_list *reorder; - empty = 0; - next_nr = -1; - next_overrun = 0; - next_queue = NULL; - num_cpus = cpumask_weight(pd->cpumask); - for_each_cpu(cpu, pd->cpumask) { - queue = per_cpu_ptr(pd->queue, cpu); - reorder = &queue->reorder; - - /* - * Calculate the seq_nr of the object that should be - * next in this reorder queue. - */ - overrun = 0; - calc_seq_nr = (atomic_read(&queue->num_obj) * num_cpus) - + queue->cpu_index; - - if (unlikely(calc_seq_nr > pd->max_seq_nr)) { - calc_seq_nr = calc_seq_nr - pd->max_seq_nr - 1; - overrun = 1; - } - - if (!list_empty(&reorder->list)) { - padata = list_entry(reorder->list.next, - struct padata_priv, list); - - seq_nr = padata->seq_nr; - BUG_ON(calc_seq_nr != seq_nr); - } else { - seq_nr = calc_seq_nr; - empty++; - } - - if (next_nr < 0 || seq_nr < next_nr - || (next_overrun && !overrun)) { - next_nr = seq_nr; - next_overrun = overrun; - next_queue = queue; - } + /* + * Calculate the percpu reorder queue and the sequence + * number of the next object. + */ + next_nr = pd->processed; + next_index = next_nr % num_cpus; + cpu = padata_index_to_cpu(pd, next_index); + next_queue = per_cpu_ptr(pd->queue, cpu); + + if (unlikely(next_nr > pd->max_seq_nr)) { + next_nr = next_nr - pd->max_seq_nr - 1; + next_index = next_nr % num_cpus; + cpu = padata_index_to_cpu(pd, next_index); + next_queue = per_cpu_ptr(pd->queue, cpu); + pd->processed = 0; } padata = NULL; - if (empty == num_cpus) - goto out; - reorder = &next_queue->reorder; if (!list_empty(&reorder->list)) { padata = list_entry(reorder->list.next, struct padata_priv, list); - if (unlikely(next_overrun)) { - for_each_cpu(cpu, pd->cpumask) { - queue = per_cpu_ptr(pd->queue, cpu); - atomic_set(&queue->num_obj, 0); - } - } + BUG_ON(next_nr != padata->seq_nr); spin_lock(&reorder->lock); list_del_init(&padata->list); atomic_dec(&pd->reorder_objects); spin_unlock(&reorder->lock); - atomic_inc(&next_queue->num_obj); + pd->processed++; goto out; } @@ -430,7 +398,6 @@ static struct parallel_data *padata_alloc_pd(struct padata_instance *pinst, INIT_WORK(&queue->pwork, padata_parallel_worker); INIT_WORK(&queue->swork, padata_serial_worker); - atomic_set(&queue->num_obj, 0); } num_cpus = cpumask_weight(pd->cpumask); -- 1.5.6.5 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/