From: Tejun Heo <htejun@gmail.com>
To: torvalds@linux-foundation.org, mingo@elte.hu,
linux-kernel@vger.kernel.org, jeff@garzik.org,
akpm@linux-foundation.org, rusty@rustcorp.com.au,
cl@linux-foundation.org, dhowells@redhat.com,
arjan@linux.intel.com, oleg@redhat.com, axboe@kernel.dk,
fweisbec@gmail.com, dwalker@codeaurora.org,
stefanr@s5r6.in-berlin.de, florian@mickler.org,
andi@firstfloor.org, mst@redhat.com, randy.dunlap@oracle.com
Subject: [PATCH 1/4] workqueue: use worker_set/clr_flags() only from worker itself
Date: Fri, 02 Jul 2010 10:33:24 +0200 [thread overview]
Message-ID: <4C2DA454.2010401@gmail.com> (raw)
In-Reply-To: <4C2DA42C.7090804@gmail.com>
worker_set/clr_flags() assume that if none of NOT_RUNNING flags is set
the worker must be contributing to nr_running which is only true if
the worker is actually running.
As when called from self, it is guaranteed that the worker is running,
those functions can be safely used from the worker itself and they
aren't necessary from other places anyway. Make the following changes
to fix the bug.
* Make worker_set/clr_flags() whine if not called from self.
* Convert all places which called those functions from other tasks to
manipulate flags directly.
* Make trustee_thread() directly clear nr_running after setting
WORKER_ROGUE on all workers. This is the only place where
nr_running manipulation is necessary outside of workers themselves.
* While at it, add sanity check for nr_running in worker_enter_idle().
Signed-off-by: Tejun Heo <tj@kernel.org>
---
kernel/workqueue.c | 47 ++++++++++++++++++++++++++++-------------------
1 files changed, 28 insertions(+), 19 deletions(-)
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 6fa847c..5587338 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -601,7 +601,7 @@ struct task_struct *wq_worker_sleeping(struct task_struct *task,
/**
* worker_set_flags - set worker flags and adjust nr_running accordingly
- * @worker: worker to set flags for
+ * @worker: self
* @flags: flags to set
* @wakeup: wakeup an idle worker if necessary
*
@@ -609,14 +609,16 @@ struct task_struct *wq_worker_sleeping(struct task_struct *task,
* nr_running becomes zero and @wakeup is %true, an idle worker is
* woken up.
*
- * LOCKING:
- * spin_lock_irq(gcwq->lock).
+ * CONTEXT:
+ * spin_lock_irq(gcwq->lock)
*/
static inline void worker_set_flags(struct worker *worker, unsigned int flags,
bool wakeup)
{
struct global_cwq *gcwq = worker->gcwq;
+ WARN_ON_ONCE(worker->task != current);
+
/*
* If transitioning into NOT_RUNNING, adjust nr_running and
* wake up an idle worker as necessary if requested by
@@ -639,19 +641,21 @@ static inline void worker_set_flags(struct worker *worker, unsigned int flags,
/**
* worker_clr_flags - clear worker flags and adjust nr_running accordingly
- * @worker: worker to set flags for
+ * @worker: self
* @flags: flags to clear
*
* Clear @flags in @worker->flags and adjust nr_running accordingly.
*
- * LOCKING:
- * spin_lock_irq(gcwq->lock).
+ * CONTEXT:
+ * spin_lock_irq(gcwq->lock)
*/
static inline void worker_clr_flags(struct worker *worker, unsigned int flags)
{
struct global_cwq *gcwq = worker->gcwq;
unsigned int oflags = worker->flags;
+ WARN_ON_ONCE(worker->task != current);
+
worker->flags &= ~flags;
/* if transitioning out of NOT_RUNNING, increment nr_running */
@@ -1073,7 +1077,8 @@ static void worker_enter_idle(struct worker *worker)
BUG_ON(!list_empty(&worker->entry) &&
(worker->hentry.next || worker->hentry.pprev));
- worker_set_flags(worker, WORKER_IDLE, false);
+ /* can't use worker_set_flags(), also called from start_worker() */
+ worker->flags |= WORKER_IDLE;
gcwq->nr_idle++;
worker->last_active = jiffies;
@@ -1086,6 +1091,10 @@ static void worker_enter_idle(struct worker *worker)
jiffies + IDLE_WORKER_TIMEOUT);
} else
wake_up_all(&gcwq->trustee_wait);
+
+ /* sanity check nr_running */
+ WARN_ON_ONCE(gcwq->nr_workers == gcwq->nr_idle &&
+ atomic_read(get_gcwq_nr_running(gcwq->cpu)));
}
/**
@@ -1270,7 +1279,7 @@ fail:
*/
static void start_worker(struct worker *worker)
{
- worker_set_flags(worker, WORKER_STARTED, false);
+ worker->flags |= WORKER_STARTED;
worker->gcwq->nr_workers++;
worker_enter_idle(worker);
wake_up_process(worker->task);
@@ -1300,7 +1309,7 @@ static void destroy_worker(struct worker *worker)
gcwq->nr_idle--;
list_del_init(&worker->entry);
- worker_set_flags(worker, WORKER_DIE, false);
+ worker->flags |= WORKER_DIE;
spin_unlock_irq(&gcwq->lock);
@@ -2979,10 +2988,10 @@ static int __cpuinit trustee_thread(void *__gcwq)
gcwq->flags |= GCWQ_MANAGING_WORKERS;
list_for_each_entry(worker, &gcwq->idle_list, entry)
- worker_set_flags(worker, WORKER_ROGUE, false);
+ worker->flags |= WORKER_ROGUE;
for_each_busy_worker(worker, i, pos, gcwq)
- worker_set_flags(worker, WORKER_ROGUE, false);
+ worker->flags |= WORKER_ROGUE;
/*
* Call schedule() so that we cross rq->lock and thus can
@@ -2995,12 +3004,12 @@ static int __cpuinit trustee_thread(void *__gcwq)
spin_lock_irq(&gcwq->lock);
/*
- * Sched callbacks are disabled now. gcwq->nr_running should
- * be zero and will stay that way, making need_more_worker()
- * and keep_working() always return true as long as the
- * worklist is not empty.
+ * Sched callbacks are disabled now. Zap nr_running. After
+ * this, nr_running stays zero and need_more_worker() and
+ * keep_working() are always true as long as the worklist is
+ * not empty.
*/
- WARN_ON_ONCE(atomic_read(get_gcwq_nr_running(gcwq->cpu)) != 0);
+ atomic_set(get_gcwq_nr_running(gcwq->cpu), 0);
spin_unlock_irq(&gcwq->lock);
del_timer_sync(&gcwq->idle_timer);
@@ -3046,7 +3055,7 @@ static int __cpuinit trustee_thread(void *__gcwq)
worker = create_worker(gcwq, false);
spin_lock_irq(&gcwq->lock);
if (worker) {
- worker_set_flags(worker, WORKER_ROGUE, false);
+ worker->flags |= WORKER_ROGUE;
start_worker(worker);
}
}
@@ -3085,8 +3094,8 @@ static int __cpuinit trustee_thread(void *__gcwq)
* operations. Use a separate flag to mark that
* rebinding is scheduled.
*/
- worker_set_flags(worker, WORKER_REBIND, false);
- worker_clr_flags(worker, WORKER_ROGUE);
+ worker->flags |= WORKER_REBIND;
+ worker->flags &= ~WORKER_ROGUE;
/* queue rebind_work, wq doesn't matter, use the default one */
if (test_and_set_bit(WORK_STRUCT_PENDING_BIT,
--
1.6.4.2
next prev parent reply other threads:[~2010-07-02 8:33 UTC|newest]
Thread overview: 98+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-06-28 21:03 [PATCHSET] workqueue: concurrency managed workqueue, take#6 Tejun Heo
2010-06-28 21:03 ` [PATCH 01/35] kthread: implement kthread_worker Tejun Heo
2010-06-28 21:03 ` [PATCH 02/35] ivtv: use kthread_worker instead of workqueue Tejun Heo
2010-07-05 17:11 ` Andy Walls
2010-07-06 7:01 ` Tejun Heo
2010-07-09 13:15 ` Andy Walls
2010-06-28 21:03 ` [PATCH 03/35] kthread: implement kthread_data() Tejun Heo
2010-06-28 21:03 ` [PATCH 04/35] acpi: use queue_work_on() instead of binding workqueue worker to cpu0 Tejun Heo
2010-06-28 21:03 ` [PATCH 05/35] workqueue: kill RT workqueue Tejun Heo
2010-06-28 21:03 ` [PATCH 06/35] workqueue: misc/cosmetic updates Tejun Heo
2010-06-28 21:03 ` [PATCH 07/35] workqueue: merge feature parameters into flags Tejun Heo
2010-06-28 21:03 ` [PATCH 08/35] workqueue: define masks for work flags and conditionalize STATIC flags Tejun Heo
2010-06-28 21:03 ` [PATCH 09/35] workqueue: separate out process_one_work() Tejun Heo
2010-06-28 21:03 ` [PATCH 10/35] workqueue: temporarily remove workqueue tracing Tejun Heo
2010-06-28 21:03 ` [PATCH 11/35] workqueue: kill cpu_populated_map Tejun Heo
2010-06-28 21:04 ` [PATCH 12/35] workqueue: update cwq alignement Tejun Heo
2010-06-28 22:47 ` Frederic Weisbecker
2010-06-29 7:39 ` Tejun Heo
2010-06-29 12:36 ` Frederic Weisbecker
2010-06-29 15:42 ` Tejun Heo
2010-06-29 15:47 ` Frederic Weisbecker
2010-06-29 15:51 ` Tejun Heo
2010-06-29 16:01 ` Frederic Weisbecker
2010-06-29 16:09 ` Tejun Heo
2010-06-29 16:17 ` Frederic Weisbecker
2010-07-06 14:22 ` Christoph Lameter
2010-07-06 14:26 ` Tejun Heo
2010-06-29 8:12 ` [PATCH UPDATED " Tejun Heo
2010-06-29 13:39 ` Frederic Weisbecker
2010-06-28 21:04 ` [PATCH 13/35] workqueue: reimplement workqueue flushing using color coded works Tejun Heo
2010-06-28 21:04 ` [PATCH 14/35] workqueue: introduce worker Tejun Heo
2010-06-28 21:04 ` [PATCH 15/35] workqueue: reimplement work flushing using linked works Tejun Heo
2010-06-28 21:04 ` [PATCH 16/35] workqueue: implement per-cwq active work limit Tejun Heo
2010-06-28 21:04 ` [PATCH 17/35] workqueue: reimplement workqueue freeze using max_active Tejun Heo
2010-06-28 21:04 ` [PATCH 18/35] workqueue: introduce global cwq and unify cwq locks Tejun Heo
2010-06-28 21:04 ` [PATCH 19/35] workqueue: implement worker states Tejun Heo
2010-06-28 21:04 ` [PATCH 20/35] workqueue: reimplement CPU hotplugging support using trustee Tejun Heo
2010-06-28 21:04 ` [PATCH 21/35] workqueue: make single thread workqueue shared worker pool friendly Tejun Heo
2010-06-28 21:04 ` [PATCH 22/35] workqueue: add find_worker_executing_work() and track current_cwq Tejun Heo
2010-06-28 21:04 ` [PATCH 23/35] workqueue: carry cpu number in work data once execution starts Tejun Heo
2010-06-28 21:04 ` [PATCH 24/35] workqueue: implement WQ_NON_REENTRANT Tejun Heo
2010-06-28 21:04 ` [PATCH 25/35] workqueue: use shared worklist and pool all workers per cpu Tejun Heo
2010-06-28 21:04 ` [PATCH 26/35] workqueue: implement worker_{set|clr}_flags() Tejun Heo
2010-06-28 21:04 ` [PATCH 27/35] workqueue: implement concurrency managed dynamic worker pool Tejun Heo
2010-07-09 9:11 ` Yong Zhang
2010-07-12 8:53 ` [PATCH] workqueue: fix locking in retry path of maybe_create_worker() Tejun Heo
2010-07-12 13:23 ` Yong Zhang
2010-07-14 9:37 ` Tejun Heo
2010-06-28 21:04 ` [PATCH 28/35] workqueue: increase max_active of keventd and kill current_is_keventd() Tejun Heo
2010-06-28 21:04 ` [PATCH 29/35] workqueue: s/__create_workqueue()/alloc_workqueue()/, and add system workqueues Tejun Heo
2010-06-28 21:04 ` [PATCH 30/35] workqueue: implement several utility APIs Tejun Heo
2010-06-28 21:04 ` [PATCH 31/35] workqueue: implement high priority workqueue Tejun Heo
2010-06-28 21:04 ` [PATCH 32/35] workqueue: implement cpu intensive workqueue Tejun Heo
2010-06-28 21:04 ` [PATCH 33/35] libata: take advantage of cmwq and remove concurrency limitations Tejun Heo
2010-06-28 22:32 ` Jeff Garzik
2010-06-29 7:00 ` Tejun Heo
2010-06-28 21:04 ` [PATCH 34/35] async: use workqueue for worker pool Tejun Heo
2010-06-28 22:55 ` Frederic Weisbecker
2010-06-29 7:25 ` Tejun Heo
2010-06-29 12:18 ` Frederic Weisbecker
2010-06-29 15:46 ` Tejun Heo
2010-06-29 15:52 ` Frederic Weisbecker
2010-06-29 15:55 ` Tejun Heo
2010-06-29 16:40 ` Arjan van de Ven
2010-06-29 16:59 ` Tejun Heo
2010-06-29 17:12 ` Tejun Heo
2010-06-29 18:08 ` Arjan van de Ven
2010-06-29 18:07 ` Arjan van de Ven
2010-06-29 18:15 ` Tejun Heo
2010-06-29 18:22 ` Arjan van de Ven
2010-06-29 18:34 ` Tejun Heo
2010-06-29 18:41 ` Arjan van de Ven
2010-06-29 18:59 ` Tejun Heo
2010-06-29 21:37 ` David Howells
2010-07-02 9:17 ` [PATCHSET] workqueue: implement and use WQ_UNBOUND Tejun Heo
2010-07-02 9:19 ` [PATCH 1/4] workqueue: prepare for WQ_UNBOUND implementation Tejun Heo
2010-07-02 9:24 ` [PATCH 3/4] workqueue: remove WQ_SINGLE_CPU and use WQ_UNBOUND instead Tejun Heo
2010-07-02 9:25 ` [PATCH 4/4] async: use workqueue for worker pool Tejun Heo
2010-07-02 15:09 ` Stefan Richter
2010-07-02 16:26 ` Tejun Heo
2010-07-02 16:25 ` [PATCH UPDATED " Tejun Heo
2010-07-02 9:28 ` [PATCH 2/4] workqueue: implement unbound workqueue Tejun Heo
2010-07-02 9:32 ` [PATCHSET] workqueue: implement and use WQ_UNBOUND Tejun Heo
2010-07-07 5:41 ` Tejun Heo
2010-07-14 9:39 ` Tejun Heo
2010-07-20 22:01 ` David Howells
2010-07-02 9:20 ` [PATCH 2/4] workqueue: implement unbound workqueue Tejun Heo
2010-06-28 21:04 ` [PATCH 35/35] pcrypt: use HIGHPRI and CPU_INTENSIVE workqueues for padata Tejun Heo
2010-06-28 23:18 ` [PATCHSET] workqueue: concurrency managed workqueue, take#6 Frederic Weisbecker
2010-06-29 7:05 ` Tejun Heo
2010-07-02 8:32 ` [PATCHSET] workqueue: fixes on top of cmwq take#6 Tejun Heo
2010-07-02 8:33 ` Tejun Heo [this message]
2010-07-02 8:34 ` [PATCH 2/4] workqueue: fix race condition in flush_workqueue() Tejun Heo
2010-07-02 8:35 ` [PATCH 3/4] workqueue: fix incorrect cpu number BUG_ON() in get_work_gcwq() Tejun Heo
2010-07-02 8:35 ` [PATCH 4/4] workqueue: fix worker management invocation without pending works Tejun Heo
2010-07-19 14:51 ` [PATCHSET] workqueue: concurrency managed workqueue, take#6 Tejun Heo
2010-07-21 13:23 ` David Howells
2010-07-21 14:52 ` Tejun Heo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4C2DA454.2010401@gmail.com \
--to=htejun@gmail.com \
--cc=akpm@linux-foundation.org \
--cc=andi@firstfloor.org \
--cc=arjan@linux.intel.com \
--cc=axboe@kernel.dk \
--cc=cl@linux-foundation.org \
--cc=dhowells@redhat.com \
--cc=dwalker@codeaurora.org \
--cc=florian@mickler.org \
--cc=fweisbec@gmail.com \
--cc=jeff@garzik.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=mst@redhat.com \
--cc=oleg@redhat.com \
--cc=randy.dunlap@oracle.com \
--cc=rusty@rustcorp.com.au \
--cc=stefanr@s5r6.in-berlin.de \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.