lkml.org 
[lkml]   [2014]   [Apr]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH 4/6] workqueue: commit worker to pool's concurrency setting atomically.
Date
workers' concurrency setting need to be coordinate with pool's
concurrency setting when create_worker()/destroy_worker()/cpu_inline()
cpu_offline().

But create_worker() handles it non-atomically(not in a single pool->lock).
This patch makes the behavior atomically.

Now bind_list is used for coordinating workers' cpumask with the pool.
worker_idr is used for coordinating workers' concurrency with the pool.

cpumask is coordinated at first and then concurrency.

We don't want to remove worker_idr and re-use bind_list.
if we do so:
1) the locking will become much complex.
2) after removing worker_idr, we need to add a ida back
we do not save any thing.

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
---
kernel/workqueue.c | 15 +++------------
1 files changed, 3 insertions(+), 12 deletions(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 6c38aed..3a6be02 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -63,10 +63,6 @@ enum {
* While DISASSOCIATED, the cpu may be offline and all workers have
* %WORKER_UNBOUND set and concurrency management disabled, and may
* be executing on any CPU. The pool behaves as an unbound one.
- *
- * Note that DISASSOCIATED should be flipped only while holding
- * manager_mutex to avoid changing binding state while
- * create_worker() is in progress.
*/
POOL_MANAGE_WORKERS = 1 << 0, /* need to manage workers */
POOL_DISASSOCIATED = 1 << 2, /* cpu can't serve workers */
@@ -1745,16 +1741,10 @@ static struct worker *create_worker(struct worker_pool *pool)

bind_worker(worker, pool);

- /*
- * The caller is responsible for ensuring %POOL_DISASSOCIATED
- * remains stable across this function. See the comments above the
- * flag definition for details.
- */
+ /* successful, commit the worker to the pool's concurrency setting */
+ spin_lock_irq(&pool->lock);
if (pool->flags & POOL_DISASSOCIATED)
worker->flags |= WORKER_UNBOUND;
-
- /* successful, commit the pointer to idr */
- spin_lock_irq(&pool->lock);
idr_replace(&pool->worker_idr, worker, worker->id);
spin_unlock_irq(&pool->lock);

@@ -1841,6 +1831,7 @@ static void destroy_worker(struct worker *worker)

list_del_init(&worker->entry);
worker->flags |= WORKER_DIE;
+ /* release @id and leave pool's concurrency setting */
idr_remove(&pool->worker_idr, worker->id);
wake_up_process(worker->task);
}
--
1.7.4.4


\
 
 \ /
  Last update: 2014-04-12 13:21    [W:0.025 / U:0.368 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site