2005-04-16 16:20:36 -06:00
|
|
|
/*
|
2010-09-10 08:51:36 -06:00
|
|
|
* kernel/workqueue.c - generic async execution with shared worker pool
|
2005-04-16 16:20:36 -06:00
|
|
|
*
|
2010-09-10 08:51:36 -06:00
|
|
|
* Copyright (C) 2002 Ingo Molnar
|
2005-04-16 16:20:36 -06:00
|
|
|
*
|
2010-09-10 08:51:36 -06:00
|
|
|
* Derived from the taskqueue/keventd code by:
|
|
|
|
* David Woodhouse <dwmw2@infradead.org>
|
|
|
|
* Andrew Morton
|
|
|
|
* Kai Petzke <wpp@marie.physik.tu-berlin.de>
|
|
|
|
* Theodore Ts'o <tytso@mit.edu>
|
2005-04-16 16:20:36 -06:00
|
|
|
*
|
2010-09-10 08:51:36 -06:00
|
|
|
* Made to use alloc_percpu by Christoph Lameter.
|
2005-04-16 16:20:36 -06:00
|
|
|
*
|
2010-09-10 08:51:36 -06:00
|
|
|
* Copyright (C) 2010 SUSE Linux Products GmbH
|
|
|
|
* Copyright (C) 2010 Tejun Heo <tj@kernel.org>
|
2005-10-30 16:01:59 -07:00
|
|
|
*
|
2010-09-10 08:51:36 -06:00
|
|
|
* This is the generic async execution mechanism. Work items as are
|
|
|
|
* executed in process context. The worker pool is shared and
|
2013-08-20 18:50:39 -06:00
|
|
|
* automatically managed. There are two worker pools for each CPU (one for
|
|
|
|
* normal work items and the other for high priority ones) and some extra
|
|
|
|
* pools for workqueues which are not bound to any specific CPU - the
|
|
|
|
* number of these backing pools is dynamic.
|
2010-09-10 08:51:36 -06:00
|
|
|
*
|
|
|
|
* Please read Documentation/workqueue.txt for details.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
|
|
|
|
2011-05-23 12:51:41 -06:00
|
|
|
#include <linux/export.h>
|
2005-04-16 16:20:36 -06:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/signal.h>
|
|
|
|
#include <linux/completion.h>
|
|
|
|
#include <linux/workqueue.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/cpu.h>
|
|
|
|
#include <linux/notifier.h>
|
|
|
|
#include <linux/kthread.h>
|
2006-02-23 11:43:43 -07:00
|
|
|
#include <linux/hardirq.h>
|
2006-10-11 02:21:26 -06:00
|
|
|
#include <linux/mempolicy.h>
|
2006-12-06 21:34:49 -07:00
|
|
|
#include <linux/freezer.h>
|
2006-12-06 21:37:26 -07:00
|
|
|
#include <linux/kallsyms.h>
|
|
|
|
#include <linux/debug_locks.h>
|
2007-10-19 00:39:55 -06:00
|
|
|
#include <linux/lockdep.h>
|
2010-06-29 02:07:11 -06:00
|
|
|
#include <linux/idr.h>
|
2013-03-12 12:30:03 -06:00
|
|
|
#include <linux/jhash.h>
|
2012-12-17 08:01:23 -07:00
|
|
|
#include <linux/hashtable.h>
|
2013-03-12 12:30:00 -06:00
|
|
|
#include <linux/rculist.h>
|
2013-04-01 12:23:32 -06:00
|
|
|
#include <linux/nodemask.h>
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
#include <linux/moduleparam.h>
|
2013-04-30 16:27:22 -06:00
|
|
|
#include <linux/uaccess.h>
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-01-18 15:05:55 -07:00
|
|
|
#include "workqueue_internal.h"
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
enum {
|
2013-01-24 12:01:33 -07:00
|
|
|
/*
|
|
|
|
* worker_pool flags
|
2012-07-17 13:39:27 -06:00
|
|
|
*
|
2013-01-24 12:01:33 -07:00
|
|
|
* A bound pool is either associated or disassociated with its CPU.
|
2012-07-17 13:39:27 -06:00
|
|
|
* While associated (!DISASSOCIATED), all workers are bound to the
|
|
|
|
* CPU and none has %WORKER_UNBOUND set and concurrency management
|
|
|
|
* is in effect.
|
|
|
|
*
|
|
|
|
* While DISASSOCIATED, the cpu may be offline and all workers have
|
|
|
|
* %WORKER_UNBOUND set and concurrency management disabled, and may
|
2013-01-24 12:01:33 -07:00
|
|
|
* be executing on any CPU. The pool behaves as an unbound one.
|
2012-07-17 13:39:27 -06:00
|
|
|
*
|
2013-03-13 20:47:39 -06:00
|
|
|
* Note that DISASSOCIATED should be flipped only while holding
|
|
|
|
* manager_mutex to avoid changing binding state while
|
2013-01-24 12:01:33 -07:00
|
|
|
* create_worker() is in progress.
|
2012-07-17 13:39:27 -06:00
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
POOL_MANAGE_WORKERS = 1 << 0, /* need to manage workers */
|
2013-01-24 12:01:33 -07:00
|
|
|
POOL_DISASSOCIATED = 1 << 2, /* cpu can't serve workers */
|
2013-01-24 12:01:33 -07:00
|
|
|
POOL_FREEZING = 1 << 3, /* freeze in progress */
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/* worker flags */
|
|
|
|
WORKER_STARTED = 1 << 0, /* started */
|
|
|
|
WORKER_DIE = 1 << 1, /* die die die */
|
|
|
|
WORKER_IDLE = 1 << 2, /* is idle */
|
2010-06-29 02:07:14 -06:00
|
|
|
WORKER_PREP = 1 << 3, /* preparing to run works */
|
2010-06-29 02:07:15 -06:00
|
|
|
WORKER_CPU_INTENSIVE = 1 << 6, /* cpu intensive */
|
2010-07-02 02:03:51 -06:00
|
|
|
WORKER_UNBOUND = 1 << 7, /* worker is unbound */
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
WORKER_REBOUND = 1 << 8, /* worker was rebound */
|
2010-06-29 02:07:14 -06:00
|
|
|
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
WORKER_NOT_RUNNING = WORKER_PREP | WORKER_CPU_INTENSIVE |
|
|
|
|
WORKER_UNBOUND | WORKER_REBOUND,
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
NR_STD_WORKER_POOLS = 2, /* # standard pools per cpu */
|
2012-07-13 23:16:44 -06:00
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
UNBOUND_POOL_HASH_ORDER = 6, /* hashed by pool->attrs */
|
2010-06-29 02:07:12 -06:00
|
|
|
BUSY_WORKER_HASH_ORDER = 6, /* 64 pointers */
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
MAX_IDLE_WORKERS_RATIO = 4, /* 1/4 of busy can be idle */
|
|
|
|
IDLE_WORKER_TIMEOUT = 300 * HZ, /* keep idle ones for 5 mins */
|
|
|
|
|
2011-02-16 10:10:19 -07:00
|
|
|
MAYDAY_INITIAL_TIMEOUT = HZ / 100 >= 2 ? HZ / 100 : 2,
|
|
|
|
/* call for help after 10ms
|
|
|
|
(min two ticks) */
|
2010-06-29 02:07:14 -06:00
|
|
|
MAYDAY_INTERVAL = HZ / 10, /* and then every 100ms */
|
|
|
|
CREATE_COOLDOWN = HZ, /* time to breath after fail */
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Rescue workers are used only on emergencies and shared by
|
|
|
|
* all cpus. Give -20.
|
|
|
|
*/
|
|
|
|
RESCUER_NICE_LEVEL = -20,
|
2012-07-13 23:16:45 -06:00
|
|
|
HIGHPRI_NICE_LEVEL = -20,
|
2013-04-01 12:23:34 -06:00
|
|
|
|
|
|
|
WQ_NAME_LEN = 24,
|
2010-06-29 02:07:12 -06:00
|
|
|
};
|
2005-04-16 16:20:36 -06:00
|
|
|
|
|
|
|
/*
|
2010-06-29 02:07:10 -06:00
|
|
|
* Structure fields follow one of the following exclusion rules.
|
|
|
|
*
|
2010-08-24 06:22:47 -06:00
|
|
|
* I: Modifiable by initialization/destruction paths and read-only for
|
|
|
|
* everyone else.
|
2010-06-29 02:07:10 -06:00
|
|
|
*
|
2010-06-29 02:07:14 -06:00
|
|
|
* P: Preemption protected. Disabling preemption is enough and should
|
|
|
|
* only be modified and accessed from the local cpu.
|
|
|
|
*
|
2013-01-24 12:01:33 -07:00
|
|
|
* L: pool->lock protected. Access with pool->lock held.
|
2010-06-29 02:07:10 -06:00
|
|
|
*
|
2013-01-24 12:01:33 -07:00
|
|
|
* X: During normal operation, modification requires pool->lock and should
|
|
|
|
* be done only from local cpu. Either disabling preemption on local
|
|
|
|
* cpu or grabbing pool->lock is enough for read access. If
|
|
|
|
* POOL_DISASSOCIATED is set, it's identical to L.
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
2013-03-19 14:45:21 -06:00
|
|
|
* MG: pool->manager_mutex and pool->lock protected. Writes require both
|
|
|
|
* locks. Reads can happen under either lock.
|
|
|
|
*
|
2013-03-25 17:57:17 -06:00
|
|
|
* PL: wq_pool_mutex protected.
|
2013-03-13 20:47:40 -06:00
|
|
|
*
|
2013-03-25 17:57:17 -06:00
|
|
|
* PR: wq_pool_mutex protected for writes. Sched-RCU protected for reads.
|
2013-03-12 12:30:00 -06:00
|
|
|
*
|
2013-03-25 17:57:17 -06:00
|
|
|
* WQ: wq->mutex protected.
|
|
|
|
*
|
2013-03-25 17:57:19 -06:00
|
|
|
* WR: wq->mutex protected for writes. Sched-RCU protected for reads.
|
2013-03-13 20:47:40 -06:00
|
|
|
*
|
|
|
|
* MD: wq_mayday_lock protected.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
|
|
|
|
2013-01-18 15:05:55 -07:00
|
|
|
/* struct worker is defined in workqueue_internal.h */
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool {
|
2013-01-24 12:01:33 -07:00
|
|
|
spinlock_t lock; /* the pool lock */
|
2013-03-12 12:29:59 -06:00
|
|
|
int cpu; /* I: the associated cpu */
|
2013-04-01 12:23:34 -06:00
|
|
|
int node; /* I: the associated node ID */
|
2013-01-24 12:01:33 -07:00
|
|
|
int id; /* I: pool ID */
|
2012-07-12 15:46:37 -06:00
|
|
|
unsigned int flags; /* X: flags */
|
2012-07-12 15:46:37 -06:00
|
|
|
|
|
|
|
struct list_head worklist; /* L: list of pending works */
|
|
|
|
int nr_workers; /* L: total number of workers */
|
workqueue: reimplement idle worker rebinding
Currently rebind_workers() uses rebinds idle workers synchronously
before proceeding to requesting busy workers to rebind. This is
necessary because all workers on @worker_pool->idle_list must be bound
before concurrency management local wake-ups from the busy workers
take place.
Unfortunately, the synchronous idle rebinding is quite complicated.
This patch reimplements idle rebinding to simplify the code path.
Rather than trying to make all idle workers bound before rebinding
busy workers, we simply remove all to-be-bound idle workers from the
idle list and let them add themselves back after completing rebinding
(successful or not).
As only workers which finished rebinding can on on the idle worker
list, the idle worker list is guaranteed to have only bound workers
unless CPU went down again and local wake-ups are safe.
After the change, @worker_pool->nr_idle may deviate than the actual
number of idle workers on @worker_pool->idle_list. More specifically,
nr_idle may be non-zero while ->idle_list is empty. All users of
->nr_idle and ->idle_list are audited. The only affected one is
too_many_workers() which is updated to check %false if ->idle_list is
empty regardless of ->nr_idle.
After this patch, rebind_workers() no longer performs the nasty
idle-rebind retries which require temporary release of gcwq->lock, and
both unbinding and rebinding are atomic w.r.t. global_cwq->lock.
worker->idle_rebind and global_cwq->rebind_hold are now unnecessary
and removed along with the definition of struct idle_rebind.
Changed from V1:
1) remove unlikely from too_many_workers(), ->idle_list can be empty
anytime, even before this patch, no reason to use unlikely.
2) fix a small rebasing mistake.
(which is from rebasing the orignal fixing patch to for-next)
3) add a lot of comments.
4) clear WORKER_REBIND unconditionaly in idle_worker_rebind()
tj: Updated comments and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
2012-09-18 10:59:22 -06:00
|
|
|
|
|
|
|
/* nr_idle includes the ones off idle_list for rebinding */
|
2012-07-12 15:46:37 -06:00
|
|
|
int nr_idle; /* L: currently idle ones */
|
|
|
|
|
|
|
|
struct list_head idle_list; /* X: list of idle workers */
|
|
|
|
struct timer_list idle_timer; /* L: worker idle timeout */
|
|
|
|
struct timer_list mayday_timer; /* L: SOS timer for workers */
|
|
|
|
|
2013-03-13 17:51:36 -06:00
|
|
|
/* a workers is either on busy_hash or idle_list, or the manager */
|
2013-01-24 12:01:33 -07:00
|
|
|
DECLARE_HASHTABLE(busy_hash, BUSY_WORKER_HASH_ORDER);
|
|
|
|
/* L: hash of busy workers */
|
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
/* see manage_workers() for details on the two manager mutexes */
|
2013-03-12 12:30:00 -06:00
|
|
|
struct mutex manager_arb; /* manager arbitration */
|
2013-03-13 20:47:39 -06:00
|
|
|
struct mutex manager_mutex; /* manager exclusion */
|
2013-03-19 14:45:21 -06:00
|
|
|
struct idr worker_idr; /* MG: worker IDs and iteration */
|
2013-01-24 12:39:44 -07:00
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
struct workqueue_attrs *attrs; /* I: worker attributes */
|
2013-03-25 17:57:17 -06:00
|
|
|
struct hlist_node hash_node; /* PL: unbound_pool_hash node */
|
|
|
|
int refcnt; /* PL: refcnt for unbound pools */
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-01-24 12:39:44 -07:00
|
|
|
/*
|
|
|
|
* The current concurrency level. As it's likely to be accessed
|
|
|
|
* from other CPUs during try_to_wake_up(), put it in a separate
|
|
|
|
* cacheline.
|
|
|
|
*/
|
|
|
|
atomic_t nr_running ____cacheline_aligned_in_smp;
|
2013-03-12 12:30:03 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Destruction of pool is sched-RCU protected to allow dereferences
|
|
|
|
* from get_work_pool().
|
|
|
|
*/
|
|
|
|
struct rcu_head rcu;
|
2010-06-29 02:07:12 -06:00
|
|
|
} ____cacheline_aligned_in_smp;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/*
|
2013-02-13 20:29:12 -07:00
|
|
|
* The per-pool workqueue. While queued, the lower WORK_STRUCT_FLAG_BITS
|
|
|
|
* of work_struct->data are used for flags and the remaining high bits
|
|
|
|
* point to the pwq; thus, pwqs need to be aligned at two's power of the
|
|
|
|
* number of flag bits.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
struct pool_workqueue {
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool; /* I: the associated pool */
|
2010-06-29 02:07:10 -06:00
|
|
|
struct workqueue_struct *wq; /* I: the owning workqueue */
|
2010-06-29 02:07:11 -06:00
|
|
|
int work_color; /* L: current color */
|
|
|
|
int flush_color; /* L: flushing color */
|
2013-03-12 12:30:04 -06:00
|
|
|
int refcnt; /* L: reference count */
|
2010-06-29 02:07:11 -06:00
|
|
|
int nr_in_flight[WORK_NR_COLORS];
|
|
|
|
/* L: nr of in_flight works */
|
2010-06-29 02:07:12 -06:00
|
|
|
int nr_active; /* L: nr of active works */
|
2010-06-29 02:07:12 -06:00
|
|
|
int max_active; /* L: max active works */
|
2010-06-29 02:07:12 -06:00
|
|
|
struct list_head delayed_works; /* L: delayed works */
|
2013-03-25 17:57:17 -06:00
|
|
|
struct list_head pwqs_node; /* WR: node on wq->pwqs */
|
2013-03-13 20:47:40 -06:00
|
|
|
struct list_head mayday_node; /* MD: node on wq->maydays */
|
2013-03-12 12:30:04 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Release of unbound pwq is punted to system_wq. See put_pwq()
|
|
|
|
* and pwq_unbound_release_workfn() for details. pool_workqueue
|
|
|
|
* itself is also sched-RCU protected so that the first pwq can be
|
2013-03-25 17:57:18 -06:00
|
|
|
* determined without grabbing wq->mutex.
|
2013-03-12 12:30:04 -06:00
|
|
|
*/
|
|
|
|
struct work_struct unbound_release_work;
|
|
|
|
struct rcu_head rcu;
|
2013-03-12 12:29:57 -06:00
|
|
|
} __aligned(1 << WORK_STRUCT_FLAG_BITS);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
/*
|
|
|
|
* Structure used to wait for workqueue flush.
|
|
|
|
*/
|
|
|
|
struct wq_flusher {
|
2013-03-25 17:57:17 -06:00
|
|
|
struct list_head list; /* WQ: list of flushers */
|
|
|
|
int flush_color; /* WQ: flush color waiting for */
|
2010-06-29 02:07:11 -06:00
|
|
|
struct completion done; /* flush completion */
|
|
|
|
};
|
|
|
|
|
2013-03-12 12:30:05 -06:00
|
|
|
struct wq_device;
|
|
|
|
|
2005-04-16 16:20:36 -06:00
|
|
|
/*
|
2013-03-13 17:51:36 -06:00
|
|
|
* The externally visible workqueue. It relays the issued work items to
|
|
|
|
* the appropriate worker_pool through its pool_workqueues.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
|
|
|
struct workqueue_struct {
|
2013-03-25 17:57:17 -06:00
|
|
|
struct list_head pwqs; /* WR: all pwqs of this wq */
|
2013-03-25 17:57:17 -06:00
|
|
|
struct list_head list; /* PL: list of all workqueues */
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
struct mutex mutex; /* protects this wq */
|
|
|
|
int work_color; /* WQ: current work color */
|
|
|
|
int flush_color; /* WQ: current flush color */
|
2013-02-13 20:29:12 -07:00
|
|
|
atomic_t nr_pwqs_to_flush; /* flush in progress */
|
2013-03-25 17:57:17 -06:00
|
|
|
struct wq_flusher *first_flusher; /* WQ: first flusher */
|
|
|
|
struct list_head flusher_queue; /* WQ: flush waiters */
|
|
|
|
struct list_head flusher_overflow; /* WQ: flush overflow list */
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-13 20:47:40 -06:00
|
|
|
struct list_head maydays; /* MD: pwqs requesting rescue */
|
2010-06-29 02:07:14 -06:00
|
|
|
struct worker *rescuer; /* I: rescue worker */
|
|
|
|
|
2013-03-25 17:57:18 -06:00
|
|
|
int nr_drainers; /* WQ: drain in progress */
|
2013-03-25 17:57:19 -06:00
|
|
|
int saved_max_active; /* WQ: saved pwq max_active */
|
2013-03-12 12:30:05 -06:00
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
struct workqueue_attrs *unbound_attrs; /* WQ: only for unbound wqs */
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
struct pool_workqueue *dfl_pwq; /* WQ: only for unbound wqs */
|
2013-04-01 12:23:34 -06:00
|
|
|
|
2013-03-12 12:30:05 -06:00
|
|
|
#ifdef CONFIG_SYSFS
|
|
|
|
struct wq_device *wq_dev; /* I: for sysfs interface */
|
|
|
|
#endif
|
2007-10-19 00:39:55 -06:00
|
|
|
#ifdef CONFIG_LOCKDEP
|
2010-06-29 02:07:10 -06:00
|
|
|
struct lockdep_map lockdep_map;
|
2007-10-19 00:39:55 -06:00
|
|
|
#endif
|
2013-04-01 12:23:34 -06:00
|
|
|
char name[WQ_NAME_LEN]; /* I: workqueue name */
|
2013-04-01 12:23:35 -06:00
|
|
|
|
|
|
|
/* hot fields used during command issue, aligned to cacheline */
|
|
|
|
unsigned int flags ____cacheline_aligned; /* WQ: WQ_* flags */
|
|
|
|
struct pool_workqueue __percpu *cpu_pwqs; /* I: per-cpu pwqs */
|
2013-04-01 12:23:35 -06:00
|
|
|
struct pool_workqueue __rcu *numa_pwq_tbl[]; /* FR: unbound pwqs indexed by node */
|
2005-04-16 16:20:36 -06:00
|
|
|
};
|
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
static struct kmem_cache *pwq_cache;
|
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
static int wq_numa_tbl_len; /* highest possible NUMA node id + 1 */
|
|
|
|
static cpumask_var_t *wq_numa_possible_cpumask;
|
|
|
|
/* possible CPUs of each node */
|
|
|
|
|
2013-04-01 12:23:38 -06:00
|
|
|
static bool wq_disable_numa;
|
|
|
|
module_param_named(disable_numa, wq_disable_numa, bool, 0444);
|
|
|
|
|
2013-04-08 05:15:40 -06:00
|
|
|
/* see the comment above the definition of WQ_POWER_EFFICIENT */
|
|
|
|
#ifdef CONFIG_WQ_POWER_EFFICIENT_DEFAULT
|
|
|
|
static bool wq_power_efficient = true;
|
|
|
|
#else
|
|
|
|
static bool wq_power_efficient;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
module_param_named(power_efficient, wq_power_efficient, bool, 0444);
|
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
static bool wq_numa_enabled; /* unbound NUMA affinity enabled */
|
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/* buf for wq_update_unbound_numa_attrs(), protected by CPU hotplug exclusion */
|
|
|
|
static struct workqueue_attrs *wq_update_unbound_numa_attrs_buf;
|
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
static DEFINE_MUTEX(wq_pool_mutex); /* protects pools and workqueues list */
|
2013-03-13 20:47:40 -06:00
|
|
|
static DEFINE_SPINLOCK(wq_mayday_lock); /* protects wq->maydays list */
|
2013-03-13 20:47:40 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
static LIST_HEAD(workqueues); /* PL: list of all workqueues */
|
|
|
|
static bool workqueue_freezing; /* PL: have wqs started freezing? */
|
2013-03-13 20:47:40 -06:00
|
|
|
|
|
|
|
/* the per-cpu worker pools */
|
|
|
|
static DEFINE_PER_CPU_SHARED_ALIGNED(struct worker_pool [NR_STD_WORKER_POOLS],
|
|
|
|
cpu_worker_pools);
|
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
static DEFINE_IDR(worker_pool_idr); /* PR: idr of all pools */
|
2013-03-13 20:47:40 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
/* PL: hash of all unbound pools keyed by pool->attrs */
|
2013-03-12 12:30:03 -06:00
|
|
|
static DEFINE_HASHTABLE(unbound_pool_hash, UNBOUND_POOL_HASH_ORDER);
|
|
|
|
|
2013-03-13 17:51:36 -06:00
|
|
|
/* I: attributes used when instantiating standard unbound pools on demand */
|
2013-03-12 12:30:03 -06:00
|
|
|
static struct workqueue_attrs *unbound_std_wq_attrs[NR_STD_WORKER_POOLS];
|
|
|
|
|
2013-09-05 10:30:04 -06:00
|
|
|
/* I: attributes used when instantiating ordered pools on demand */
|
|
|
|
static struct workqueue_attrs *ordered_wq_attrs[NR_STD_WORKER_POOLS];
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
struct workqueue_struct *system_wq __read_mostly;
|
2013-05-06 15:44:55 -06:00
|
|
|
EXPORT_SYMBOL(system_wq);
|
2012-08-18 15:52:42 -06:00
|
|
|
struct workqueue_struct *system_highpri_wq __read_mostly;
|
2012-08-15 08:25:39 -06:00
|
|
|
EXPORT_SYMBOL_GPL(system_highpri_wq);
|
2012-08-18 15:52:42 -06:00
|
|
|
struct workqueue_struct *system_long_wq __read_mostly;
|
2010-06-29 02:07:14 -06:00
|
|
|
EXPORT_SYMBOL_GPL(system_long_wq);
|
2012-08-18 15:52:42 -06:00
|
|
|
struct workqueue_struct *system_unbound_wq __read_mostly;
|
2010-07-02 02:03:51 -06:00
|
|
|
EXPORT_SYMBOL_GPL(system_unbound_wq);
|
2012-08-18 15:52:42 -06:00
|
|
|
struct workqueue_struct *system_freezable_wq __read_mostly;
|
2011-02-21 01:52:50 -07:00
|
|
|
EXPORT_SYMBOL_GPL(system_freezable_wq);
|
2013-04-24 05:42:54 -06:00
|
|
|
struct workqueue_struct *system_power_efficient_wq __read_mostly;
|
|
|
|
EXPORT_SYMBOL_GPL(system_power_efficient_wq);
|
|
|
|
struct workqueue_struct *system_freezable_power_efficient_wq __read_mostly;
|
|
|
|
EXPORT_SYMBOL_GPL(system_freezable_power_efficient_wq);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-13 20:47:40 -06:00
|
|
|
static int worker_thread(void *__worker);
|
|
|
|
static void copy_workqueue_attrs(struct workqueue_attrs *to,
|
|
|
|
const struct workqueue_attrs *from);
|
|
|
|
|
2010-10-05 02:41:14 -06:00
|
|
|
#define CREATE_TRACE_POINTS
|
|
|
|
#include <trace/events/workqueue.h>
|
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
#define assert_rcu_or_pool_mutex() \
|
2013-03-13 20:47:40 -06:00
|
|
|
rcu_lockdep_assert(rcu_read_lock_sched_held() || \
|
2013-03-25 17:57:17 -06:00
|
|
|
lockdep_is_held(&wq_pool_mutex), \
|
|
|
|
"sched RCU or wq_pool_mutex should be held")
|
2013-03-13 20:47:40 -06:00
|
|
|
|
2013-03-25 17:57:18 -06:00
|
|
|
#define assert_rcu_or_wq_mutex(wq) \
|
2013-03-12 12:30:00 -06:00
|
|
|
rcu_lockdep_assert(rcu_read_lock_sched_held() || \
|
2013-03-25 17:57:19 -06:00
|
|
|
lockdep_is_held(&wq->mutex), \
|
2013-03-25 17:57:18 -06:00
|
|
|
"sched RCU or wq->mutex should be held")
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-03-19 14:45:21 -06:00
|
|
|
#ifdef CONFIG_LOCKDEP
|
|
|
|
#define assert_manager_or_pool_lock(pool) \
|
2013-03-19 13:28:21 -06:00
|
|
|
WARN_ONCE(debug_locks && \
|
|
|
|
!lockdep_is_held(&(pool)->manager_mutex) && \
|
2013-03-19 14:45:21 -06:00
|
|
|
!lockdep_is_held(&(pool)->lock), \
|
|
|
|
"pool->manager_mutex or ->lock should be held")
|
|
|
|
#else
|
|
|
|
#define assert_manager_or_pool_lock(pool) do { } while (0)
|
|
|
|
#endif
|
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
#define for_each_cpu_worker_pool(pool, cpu) \
|
|
|
|
for ((pool) = &per_cpu(cpu_worker_pools, cpu)[0]; \
|
|
|
|
(pool) < &per_cpu(cpu_worker_pools, cpu)[NR_STD_WORKER_POOLS]; \
|
2013-03-12 12:30:03 -06:00
|
|
|
(pool)++)
|
2012-07-13 23:16:44 -06:00
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
/**
|
|
|
|
* for_each_pool - iterate through all worker_pools in the system
|
|
|
|
* @pool: iteration cursor
|
2013-03-13 17:51:36 -06:00
|
|
|
* @pi: integer used for iteration
|
2013-03-12 12:30:00 -06:00
|
|
|
*
|
2013-03-25 17:57:17 -06:00
|
|
|
* This must be called either with wq_pool_mutex held or sched RCU read
|
|
|
|
* locked. If the pool needs to be used beyond the locking in effect, the
|
|
|
|
* caller is responsible for guaranteeing that the pool stays online.
|
2013-03-12 12:30:00 -06:00
|
|
|
*
|
|
|
|
* The if/else clause exists only for the lockdep assertion and can be
|
|
|
|
* ignored.
|
2013-03-12 12:29:58 -06:00
|
|
|
*/
|
2013-03-13 17:51:36 -06:00
|
|
|
#define for_each_pool(pool, pi) \
|
|
|
|
idr_for_each_entry(&worker_pool_idr, pool, pi) \
|
2013-03-25 17:57:17 -06:00
|
|
|
if (({ assert_rcu_or_pool_mutex(); false; })) { } \
|
2013-03-12 12:30:00 -06:00
|
|
|
else
|
2013-03-12 12:29:58 -06:00
|
|
|
|
2013-03-19 14:45:21 -06:00
|
|
|
/**
|
|
|
|
* for_each_pool_worker - iterate through all workers of a worker_pool
|
|
|
|
* @worker: iteration cursor
|
|
|
|
* @wi: integer used for iteration
|
|
|
|
* @pool: worker_pool to iterate workers of
|
|
|
|
*
|
|
|
|
* This must be called with either @pool->manager_mutex or ->lock held.
|
|
|
|
*
|
|
|
|
* The if/else clause exists only for the lockdep assertion and can be
|
|
|
|
* ignored.
|
|
|
|
*/
|
|
|
|
#define for_each_pool_worker(worker, wi, pool) \
|
|
|
|
idr_for_each_entry(&(pool)->worker_idr, (worker), (wi)) \
|
|
|
|
if (({ assert_manager_or_pool_lock((pool)); false; })) { } \
|
|
|
|
else
|
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
/**
|
|
|
|
* for_each_pwq - iterate through all pool_workqueues of the specified workqueue
|
|
|
|
* @pwq: iteration cursor
|
|
|
|
* @wq: the target workqueue
|
2013-03-12 12:30:00 -06:00
|
|
|
*
|
2013-03-25 17:57:18 -06:00
|
|
|
* This must be called either with wq->mutex held or sched RCU read locked.
|
2013-03-13 20:47:40 -06:00
|
|
|
* If the pwq needs to be used beyond the locking in effect, the caller is
|
|
|
|
* responsible for guaranteeing that the pwq stays online.
|
2013-03-12 12:30:00 -06:00
|
|
|
*
|
|
|
|
* The if/else clause exists only for the lockdep assertion and can be
|
|
|
|
* ignored.
|
2013-03-12 12:29:58 -06:00
|
|
|
*/
|
|
|
|
#define for_each_pwq(pwq, wq) \
|
2013-03-12 12:30:00 -06:00
|
|
|
list_for_each_entry_rcu((pwq), &(wq)->pwqs, pwqs_node) \
|
2013-03-25 17:57:18 -06:00
|
|
|
if (({ assert_rcu_or_wq_mutex(wq); false; })) { } \
|
2013-03-12 12:30:00 -06:00
|
|
|
else
|
2010-07-02 02:03:51 -06:00
|
|
|
|
2009-11-15 09:09:48 -07:00
|
|
|
#ifdef CONFIG_DEBUG_OBJECTS_WORK
|
|
|
|
|
|
|
|
static struct debug_obj_descr work_debug_descr;
|
|
|
|
|
2011-03-07 01:58:33 -07:00
|
|
|
static void *work_debug_hint(void *addr)
|
|
|
|
{
|
|
|
|
return ((struct work_struct *) addr)->func;
|
|
|
|
}
|
|
|
|
|
2009-11-15 09:09:48 -07:00
|
|
|
/*
|
|
|
|
* fixup_init is called when:
|
|
|
|
* - an active object is initialized
|
|
|
|
*/
|
|
|
|
static int work_fixup_init(void *addr, enum debug_obj_state state)
|
|
|
|
{
|
|
|
|
struct work_struct *work = addr;
|
|
|
|
|
|
|
|
switch (state) {
|
|
|
|
case ODEBUG_STATE_ACTIVE:
|
|
|
|
cancel_work_sync(work);
|
|
|
|
debug_object_init(work, &work_debug_descr);
|
|
|
|
return 1;
|
|
|
|
default:
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* fixup_activate is called when:
|
|
|
|
* - an active object is activated
|
|
|
|
* - an unknown object is activated (might be a statically initialized object)
|
|
|
|
*/
|
|
|
|
static int work_fixup_activate(void *addr, enum debug_obj_state state)
|
|
|
|
{
|
|
|
|
struct work_struct *work = addr;
|
|
|
|
|
|
|
|
switch (state) {
|
|
|
|
|
|
|
|
case ODEBUG_STATE_NOTAVAILABLE:
|
|
|
|
/*
|
|
|
|
* This is not really a fixup. The work struct was
|
|
|
|
* statically initialized. We just make sure that it
|
|
|
|
* is tracked in the object tracker.
|
|
|
|
*/
|
2010-06-29 02:07:10 -06:00
|
|
|
if (test_bit(WORK_STRUCT_STATIC_BIT, work_data_bits(work))) {
|
2009-11-15 09:09:48 -07:00
|
|
|
debug_object_init(work, &work_debug_descr);
|
|
|
|
debug_object_activate(work, &work_debug_descr);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
WARN_ON_ONCE(1);
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
case ODEBUG_STATE_ACTIVE:
|
|
|
|
WARN_ON(1);
|
|
|
|
|
|
|
|
default:
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* fixup_free is called when:
|
|
|
|
* - an active object is freed
|
|
|
|
*/
|
|
|
|
static int work_fixup_free(void *addr, enum debug_obj_state state)
|
|
|
|
{
|
|
|
|
struct work_struct *work = addr;
|
|
|
|
|
|
|
|
switch (state) {
|
|
|
|
case ODEBUG_STATE_ACTIVE:
|
|
|
|
cancel_work_sync(work);
|
|
|
|
debug_object_free(work, &work_debug_descr);
|
|
|
|
return 1;
|
|
|
|
default:
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct debug_obj_descr work_debug_descr = {
|
|
|
|
.name = "work_struct",
|
2011-03-07 01:58:33 -07:00
|
|
|
.debug_hint = work_debug_hint,
|
2009-11-15 09:09:48 -07:00
|
|
|
.fixup_init = work_fixup_init,
|
|
|
|
.fixup_activate = work_fixup_activate,
|
|
|
|
.fixup_free = work_fixup_free,
|
|
|
|
};
|
|
|
|
|
|
|
|
static inline void debug_work_activate(struct work_struct *work)
|
|
|
|
{
|
|
|
|
debug_object_activate(work, &work_debug_descr);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void debug_work_deactivate(struct work_struct *work)
|
|
|
|
{
|
|
|
|
debug_object_deactivate(work, &work_debug_descr);
|
|
|
|
}
|
|
|
|
|
|
|
|
void __init_work(struct work_struct *work, int onstack)
|
|
|
|
{
|
|
|
|
if (onstack)
|
|
|
|
debug_object_init_on_stack(work, &work_debug_descr);
|
|
|
|
else
|
|
|
|
debug_object_init(work, &work_debug_descr);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(__init_work);
|
|
|
|
|
|
|
|
void destroy_work_on_stack(struct work_struct *work)
|
|
|
|
{
|
|
|
|
debug_object_free(work, &work_debug_descr);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(destroy_work_on_stack);
|
|
|
|
|
2014-03-23 08:20:44 -06:00
|
|
|
void destroy_delayed_work_on_stack(struct delayed_work *work)
|
|
|
|
{
|
|
|
|
destroy_timer_on_stack(&work->timer);
|
|
|
|
debug_object_free(&work->work, &work_debug_descr);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(destroy_delayed_work_on_stack);
|
|
|
|
|
2009-11-15 09:09:48 -07:00
|
|
|
#else
|
|
|
|
static inline void debug_work_activate(struct work_struct *work) { }
|
|
|
|
static inline void debug_work_deactivate(struct work_struct *work) { }
|
|
|
|
#endif
|
|
|
|
|
2013-09-09 19:52:35 -06:00
|
|
|
/**
|
|
|
|
* worker_pool_assign_id - allocate ID and assing it to @pool
|
|
|
|
* @pool: the pool pointer of interest
|
|
|
|
*
|
|
|
|
* Returns 0 if ID in [0, WORK_OFFQ_POOL_NONE) is allocated and assigned
|
|
|
|
* successfully, -errno on failure.
|
|
|
|
*/
|
2013-01-24 12:01:33 -07:00
|
|
|
static int worker_pool_assign_id(struct worker_pool *pool)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
lockdep_assert_held(&wq_pool_mutex);
|
2013-03-13 20:47:40 -06:00
|
|
|
|
2013-09-09 19:52:35 -06:00
|
|
|
ret = idr_alloc(&worker_pool_idr, pool, 0, WORK_OFFQ_POOL_NONE,
|
|
|
|
GFP_KERNEL);
|
Linux 3.9-rc5
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
iQEcBAABAgAGBQJRWLTrAAoJEHm+PkMAQRiGe8oH/iMy48mecVWvxVZn74Tx3Cef
xmW/PnAIj28EhSPqK49N/Ow6AfQToFKf7AP0ge20KAf5teTq95AY+tH74DAANt8F
BjKXXTZiR5xwBvRkq7CR5wDcCvEcBAAz8fgTEd6SEDB2d2VXFf5eKdKUqt1avTCh
Z6Hup5kuwX+ddtwY2DCBXtp2n6fL0Rm5yLzY1A3OOBye1E7VyLTF7M5BR603Q44P
4kRLxn8+R7jy3hTuZIhAeoS8TKUoBwVk7DmKxEzrhTHZVOmvwE9lEHybRnIyOpd/
k1JnbRbiPsLsCVFOn10SQkGDAIk00lro3tuWP2C1ljERiD/OOh5Ui9nXYAhMkbI=
=q15K
-----END PGP SIGNATURE-----
Merge tag 'v3.9-rc5' into wq/for-3.10
Writeback conversion to workqueue will be based on top of wq/for-3.10
branch to take advantage of custom attrs and NUMA support for unbound
workqueues. Mainline currently contains two commits which result in
non-trivial merge conflicts with wq/for-3.10 and because
block/for-3.10/core is based on v3.9-rc3 which contains one of the
conflicting commits, we need a pre-merge-window merge anyway. Let's
pull v3.9-rc5 into wq/for-3.10 so that the block tree doesn't suffer
from workqueue merge conflicts.
The two conflicts and their resolutions:
* e68035fb65 ("workqueue: convert to idr_alloc()") in mainline changes
worker_pool_assign_id() to use idr_alloc() instead of the old idr
interface. worker_pool_assign_id() goes through multiple locking
changes in wq/for-3.10 causing the following conflict.
static int worker_pool_assign_id(struct worker_pool *pool)
{
int ret;
<<<<<<< HEAD
lockdep_assert_held(&wq_pool_mutex);
do {
if (!idr_pre_get(&worker_pool_idr, GFP_KERNEL))
return -ENOMEM;
ret = idr_get_new(&worker_pool_idr, pool, &pool->id);
} while (ret == -EAGAIN);
=======
mutex_lock(&worker_pool_idr_mutex);
ret = idr_alloc(&worker_pool_idr, pool, 0, 0, GFP_KERNEL);
if (ret >= 0)
pool->id = ret;
mutex_unlock(&worker_pool_idr_mutex);
>>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89
return ret < 0 ? ret : 0;
}
We want locking from the former and idr_alloc() usage from the
latter, which can be combined to the following.
static int worker_pool_assign_id(struct worker_pool *pool)
{
int ret;
lockdep_assert_held(&wq_pool_mutex);
ret = idr_alloc(&worker_pool_idr, pool, 0, 0, GFP_KERNEL);
if (ret >= 0) {
pool->id = ret;
return 0;
}
return ret;
}
* eb2834285c ("workqueue: fix possible pool stall bug in
wq_unbind_fn()") updated wq_unbind_fn() such that it has single
larger for_each_std_worker_pool() loop instead of two separate loops
with a schedule() call inbetween. wq/for-3.10 renamed
pool->assoc_mutex to pool->manager_mutex causing the following
conflict (earlier function body and comments omitted for brevity).
static void wq_unbind_fn(struct work_struct *work)
{
...
spin_unlock_irq(&pool->lock);
<<<<<<< HEAD
mutex_unlock(&pool->manager_mutex);
}
=======
mutex_unlock(&pool->assoc_mutex);
>>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89
schedule();
<<<<<<< HEAD
for_each_cpu_worker_pool(pool, cpu)
=======
>>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89
atomic_set(&pool->nr_running, 0);
spin_lock_irq(&pool->lock);
wake_up_worker(pool);
spin_unlock_irq(&pool->lock);
}
}
The resolution is mostly trivial. We want the control flow of the
latter with the rename of the former.
static void wq_unbind_fn(struct work_struct *work)
{
...
spin_unlock_irq(&pool->lock);
mutex_unlock(&pool->manager_mutex);
schedule();
atomic_set(&pool->nr_running, 0);
spin_lock_irq(&pool->lock);
wake_up_worker(pool);
spin_unlock_irq(&pool->lock);
}
}
Signed-off-by: Tejun Heo <tj@kernel.org>
2013-04-01 18:08:13 -06:00
|
|
|
if (ret >= 0) {
|
2013-03-13 15:59:38 -06:00
|
|
|
pool->id = ret;
|
Linux 3.9-rc5
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (GNU/Linux)
iQEcBAABAgAGBQJRWLTrAAoJEHm+PkMAQRiGe8oH/iMy48mecVWvxVZn74Tx3Cef
xmW/PnAIj28EhSPqK49N/Ow6AfQToFKf7AP0ge20KAf5teTq95AY+tH74DAANt8F
BjKXXTZiR5xwBvRkq7CR5wDcCvEcBAAz8fgTEd6SEDB2d2VXFf5eKdKUqt1avTCh
Z6Hup5kuwX+ddtwY2DCBXtp2n6fL0Rm5yLzY1A3OOBye1E7VyLTF7M5BR603Q44P
4kRLxn8+R7jy3hTuZIhAeoS8TKUoBwVk7DmKxEzrhTHZVOmvwE9lEHybRnIyOpd/
k1JnbRbiPsLsCVFOn10SQkGDAIk00lro3tuWP2C1ljERiD/OOh5Ui9nXYAhMkbI=
=q15K
-----END PGP SIGNATURE-----
Merge tag 'v3.9-rc5' into wq/for-3.10
Writeback conversion to workqueue will be based on top of wq/for-3.10
branch to take advantage of custom attrs and NUMA support for unbound
workqueues. Mainline currently contains two commits which result in
non-trivial merge conflicts with wq/for-3.10 and because
block/for-3.10/core is based on v3.9-rc3 which contains one of the
conflicting commits, we need a pre-merge-window merge anyway. Let's
pull v3.9-rc5 into wq/for-3.10 so that the block tree doesn't suffer
from workqueue merge conflicts.
The two conflicts and their resolutions:
* e68035fb65 ("workqueue: convert to idr_alloc()") in mainline changes
worker_pool_assign_id() to use idr_alloc() instead of the old idr
interface. worker_pool_assign_id() goes through multiple locking
changes in wq/for-3.10 causing the following conflict.
static int worker_pool_assign_id(struct worker_pool *pool)
{
int ret;
<<<<<<< HEAD
lockdep_assert_held(&wq_pool_mutex);
do {
if (!idr_pre_get(&worker_pool_idr, GFP_KERNEL))
return -ENOMEM;
ret = idr_get_new(&worker_pool_idr, pool, &pool->id);
} while (ret == -EAGAIN);
=======
mutex_lock(&worker_pool_idr_mutex);
ret = idr_alloc(&worker_pool_idr, pool, 0, 0, GFP_KERNEL);
if (ret >= 0)
pool->id = ret;
mutex_unlock(&worker_pool_idr_mutex);
>>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89
return ret < 0 ? ret : 0;
}
We want locking from the former and idr_alloc() usage from the
latter, which can be combined to the following.
static int worker_pool_assign_id(struct worker_pool *pool)
{
int ret;
lockdep_assert_held(&wq_pool_mutex);
ret = idr_alloc(&worker_pool_idr, pool, 0, 0, GFP_KERNEL);
if (ret >= 0) {
pool->id = ret;
return 0;
}
return ret;
}
* eb2834285c ("workqueue: fix possible pool stall bug in
wq_unbind_fn()") updated wq_unbind_fn() such that it has single
larger for_each_std_worker_pool() loop instead of two separate loops
with a schedule() call inbetween. wq/for-3.10 renamed
pool->assoc_mutex to pool->manager_mutex causing the following
conflict (earlier function body and comments omitted for brevity).
static void wq_unbind_fn(struct work_struct *work)
{
...
spin_unlock_irq(&pool->lock);
<<<<<<< HEAD
mutex_unlock(&pool->manager_mutex);
}
=======
mutex_unlock(&pool->assoc_mutex);
>>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89
schedule();
<<<<<<< HEAD
for_each_cpu_worker_pool(pool, cpu)
=======
>>>>>>> c67bf5361e7e66a0ff1f4caf95f89347d55dfb89
atomic_set(&pool->nr_running, 0);
spin_lock_irq(&pool->lock);
wake_up_worker(pool);
spin_unlock_irq(&pool->lock);
}
}
The resolution is mostly trivial. We want the control flow of the
latter with the rename of the former.
static void wq_unbind_fn(struct work_struct *work)
{
...
spin_unlock_irq(&pool->lock);
mutex_unlock(&pool->manager_mutex);
schedule();
atomic_set(&pool->nr_running, 0);
spin_lock_irq(&pool->lock);
wake_up_worker(pool);
spin_unlock_irq(&pool->lock);
}
}
Signed-off-by: Tejun Heo <tj@kernel.org>
2013-04-01 18:08:13 -06:00
|
|
|
return 0;
|
|
|
|
}
|
2013-03-12 12:30:00 -06:00
|
|
|
return ret;
|
2013-01-24 12:01:33 -07:00
|
|
|
}
|
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
/**
|
|
|
|
* unbound_pwq_by_node - return the unbound pool_workqueue for the given node
|
|
|
|
* @wq: the target workqueue
|
|
|
|
* @node: the node ID
|
|
|
|
*
|
|
|
|
* This must be called either with pwq_lock held or sched RCU read locked.
|
|
|
|
* If the pwq needs to be used beyond the locking in effect, the caller is
|
|
|
|
* responsible for guaranteeing that the pwq stays online.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: The unbound pool_workqueue for @node.
|
2013-04-01 12:23:35 -06:00
|
|
|
*/
|
|
|
|
static struct pool_workqueue *unbound_pwq_by_node(struct workqueue_struct *wq,
|
|
|
|
int node)
|
|
|
|
{
|
|
|
|
assert_rcu_or_wq_mutex(wq);
|
|
|
|
return rcu_dereference_raw(wq->numa_pwq_tbl[node]);
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
static unsigned int work_color_to_flags(int color)
|
|
|
|
{
|
|
|
|
return color << WORK_STRUCT_COLOR_SHIFT;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int get_work_color(struct work_struct *work)
|
|
|
|
{
|
|
|
|
return (*work_data_bits(work) >> WORK_STRUCT_COLOR_SHIFT) &
|
|
|
|
((1 << WORK_STRUCT_COLOR_BITS) - 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int work_next_color(int color)
|
|
|
|
{
|
|
|
|
return (color + 1) % WORK_NR_COLORS;
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2007-05-23 14:57:57 -06:00
|
|
|
/*
|
2013-02-13 20:29:12 -07:00
|
|
|
* While queued, %WORK_STRUCT_PWQ is set and non flag bits of a work's data
|
|
|
|
* contain the pointer to the queued pwq. Once execution starts, the flag
|
2013-01-24 12:01:33 -07:00
|
|
|
* is cleared and the high bits contain OFFQ flags and pool ID.
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
2013-02-13 20:29:12 -07:00
|
|
|
* set_work_pwq(), set_work_pool_and_clear_pending(), mark_work_canceling()
|
|
|
|
* and clear_work_data() can be used to set the pwq, pool or clear
|
2012-08-03 11:30:46 -06:00
|
|
|
* work->data. These functions should only be called while the work is
|
|
|
|
* owned - ie. while the PENDING bit is set.
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
2013-02-13 20:29:12 -07:00
|
|
|
* get_work_pool() and get_work_pwq() can be used to obtain the pool or pwq
|
2013-01-24 12:01:33 -07:00
|
|
|
* corresponding to a work. Pool is available once the work has been
|
2013-02-13 20:29:12 -07:00
|
|
|
* queued anywhere after initialization until it is sync canceled. pwq is
|
2013-01-24 12:01:33 -07:00
|
|
|
* available only while the work item is queued.
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
2012-08-03 11:30:46 -06:00
|
|
|
* %WORK_OFFQ_CANCELING is used to mark a work item which is being
|
|
|
|
* canceled. While being canceled, a work item may have its PENDING set
|
|
|
|
* but stay off timer and worklist for arbitrarily long and nobody should
|
|
|
|
* try to steal the PENDING bit.
|
2007-05-23 14:57:57 -06:00
|
|
|
*/
|
2010-06-29 02:07:13 -06:00
|
|
|
static inline void set_work_data(struct work_struct *work, unsigned long data,
|
|
|
|
unsigned long flags)
|
2006-11-22 07:54:49 -07:00
|
|
|
{
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(!work_pending(work));
|
2010-06-29 02:07:13 -06:00
|
|
|
atomic_long_set(&work->data, data | flags | work_static(work));
|
|
|
|
}
|
2006-11-22 07:54:49 -07:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
static void set_work_pwq(struct work_struct *work, struct pool_workqueue *pwq,
|
2010-06-29 02:07:13 -06:00
|
|
|
unsigned long extra_flags)
|
|
|
|
{
|
2013-02-13 20:29:12 -07:00
|
|
|
set_work_data(work, (unsigned long)pwq,
|
|
|
|
WORK_STRUCT_PENDING | WORK_STRUCT_PWQ | extra_flags);
|
2006-11-22 07:54:49 -07:00
|
|
|
}
|
|
|
|
|
2013-02-06 19:04:53 -07:00
|
|
|
static void set_work_pool_and_keep_pending(struct work_struct *work,
|
|
|
|
int pool_id)
|
|
|
|
{
|
|
|
|
set_work_data(work, (unsigned long)pool_id << WORK_OFFQ_POOL_SHIFT,
|
|
|
|
WORK_STRUCT_PENDING);
|
|
|
|
}
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
static void set_work_pool_and_clear_pending(struct work_struct *work,
|
|
|
|
int pool_id)
|
2010-06-29 02:07:13 -06:00
|
|
|
{
|
2012-08-13 18:08:19 -06:00
|
|
|
/*
|
|
|
|
* The following wmb is paired with the implied mb in
|
|
|
|
* test_and_set_bit(PENDING) and ensures all updates to @work made
|
|
|
|
* here are visible to and precede any updates by the next PENDING
|
|
|
|
* owner.
|
|
|
|
*/
|
|
|
|
smp_wmb();
|
2013-01-24 12:01:33 -07:00
|
|
|
set_work_data(work, (unsigned long)pool_id << WORK_OFFQ_POOL_SHIFT, 0);
|
2010-06-29 02:07:13 -06:00
|
|
|
}
|
2006-01-08 02:05:12 -07:00
|
|
|
|
2010-06-29 02:07:13 -06:00
|
|
|
static void clear_work_data(struct work_struct *work)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-01-24 12:01:33 -07:00
|
|
|
smp_wmb(); /* see set_work_pool_and_clear_pending() */
|
|
|
|
set_work_data(work, WORK_STRUCT_NO_POOL, 0);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
static struct pool_workqueue *get_work_pwq(struct work_struct *work)
|
2007-05-09 03:34:12 -06:00
|
|
|
{
|
2010-07-22 06:14:25 -06:00
|
|
|
unsigned long data = atomic_long_read(&work->data);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (data & WORK_STRUCT_PWQ)
|
2010-07-22 06:14:25 -06:00
|
|
|
return (void *)(data & WORK_STRUCT_WQ_DATA_MASK);
|
|
|
|
else
|
|
|
|
return NULL;
|
2010-04-23 09:40:40 -06:00
|
|
|
}
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
/**
|
|
|
|
* get_work_pool - return the worker_pool a given work was associated with
|
|
|
|
* @work: the work item of interest
|
|
|
|
*
|
2013-03-25 17:57:17 -06:00
|
|
|
* Pools are created and destroyed under wq_pool_mutex, and allows read
|
|
|
|
* access under sched-RCU read lock. As such, this function should be
|
|
|
|
* called under wq_pool_mutex or with preemption disabled.
|
2013-03-12 12:30:00 -06:00
|
|
|
*
|
|
|
|
* All fields of the returned pool are accessible as long as the above
|
|
|
|
* mentioned locking is in effect. If the returned pool needs to be used
|
|
|
|
* beyond the critical section, the caller is responsible for ensuring the
|
|
|
|
* returned pool is and stays online.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: The worker_pool @work was last associated with. %NULL if none.
|
2013-01-24 12:01:33 -07:00
|
|
|
*/
|
|
|
|
static struct worker_pool *get_work_pool(struct work_struct *work)
|
2006-11-22 07:54:49 -07:00
|
|
|
{
|
2010-07-22 06:14:25 -06:00
|
|
|
unsigned long data = atomic_long_read(&work->data);
|
2013-01-24 12:01:33 -07:00
|
|
|
int pool_id;
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
assert_rcu_or_pool_mutex();
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (data & WORK_STRUCT_PWQ)
|
|
|
|
return ((struct pool_workqueue *)
|
2013-01-24 12:01:33 -07:00
|
|
|
(data & WORK_STRUCT_WQ_DATA_MASK))->pool;
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
pool_id = data >> WORK_OFFQ_POOL_SHIFT;
|
|
|
|
if (pool_id == WORK_OFFQ_POOL_NONE)
|
2010-06-29 02:07:13 -06:00
|
|
|
return NULL;
|
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
return idr_find(&worker_pool_idr, pool_id);
|
2013-01-24 12:01:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* get_work_pool_id - return the worker pool ID a given work is associated with
|
|
|
|
* @work: the work item of interest
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return: The worker_pool ID @work was last associated with.
|
2013-01-24 12:01:33 -07:00
|
|
|
* %WORK_OFFQ_POOL_NONE if none.
|
|
|
|
*/
|
|
|
|
static int get_work_pool_id(struct work_struct *work)
|
|
|
|
{
|
2013-02-07 14:14:20 -07:00
|
|
|
unsigned long data = atomic_long_read(&work->data);
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (data & WORK_STRUCT_PWQ)
|
|
|
|
return ((struct pool_workqueue *)
|
2013-02-07 14:14:20 -07:00
|
|
|
(data & WORK_STRUCT_WQ_DATA_MASK))->pool->id;
|
2013-01-24 12:01:33 -07:00
|
|
|
|
2013-02-07 14:14:20 -07:00
|
|
|
return data >> WORK_OFFQ_POOL_SHIFT;
|
2013-01-24 12:01:33 -07:00
|
|
|
}
|
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
static void mark_work_canceling(struct work_struct *work)
|
|
|
|
{
|
2013-01-24 12:01:33 -07:00
|
|
|
unsigned long pool_id = get_work_pool_id(work);
|
2012-08-03 11:30:46 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
pool_id <<= WORK_OFFQ_POOL_SHIFT;
|
|
|
|
set_work_data(work, pool_id | WORK_OFFQ_CANCELING, WORK_STRUCT_PENDING);
|
2012-08-03 11:30:46 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
static bool work_is_canceling(struct work_struct *work)
|
|
|
|
{
|
|
|
|
unsigned long data = atomic_long_read(&work->data);
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
return !(data & WORK_STRUCT_PWQ) && (data & WORK_OFFQ_CANCELING);
|
2012-08-03 11:30:46 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/*
|
2012-07-13 23:16:45 -06:00
|
|
|
* Policy functions. These define the policies on how the global worker
|
|
|
|
* pools are managed. Unless noted otherwise, these functions assume that
|
2013-01-24 12:01:33 -07:00
|
|
|
* they're being called with pool->lock held.
|
2010-06-29 02:07:14 -06:00
|
|
|
*/
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool __need_more_worker(struct worker_pool *pool)
|
2007-05-09 03:34:17 -06:00
|
|
|
{
|
2013-01-24 12:39:44 -07:00
|
|
|
return !atomic_read(&pool->nr_running);
|
2007-05-09 03:34:17 -06:00
|
|
|
}
|
|
|
|
|
[PATCH] WorkStruct: Use direct assignment rather than cmpxchg()
Use direct assignment rather than cmpxchg() as the latter is unavailable
and unimplementable on some platforms and is actually unnecessary.
The use of cmpxchg() was to guard against two possibilities, neither of
which can actually occur:
(1) The pending flag may have been unset or may be cleared. However, given
where it's called, the pending flag is _always_ set. I don't think it
can be unset whilst we're in set_wq_data().
Once the work is enqueued to be actually run, the only way off the queue
is for it to be actually run.
If it's a delayed work item, then the bit can't be cleared by the timer
because we haven't started the timer yet. Also, the pending bit can't be
cleared by cancelling the delayed work _until_ the work item has had its
timer started.
(2) The workqueue pointer might change. This can only happen in two cases:
(a) The work item has just been queued to actually run, and so we're
protected by the appropriate workqueue spinlock.
(b) A delayed work item is being queued, and so the timer hasn't been
started yet, and so no one else knows about the work item or can
access it (the pending bit protects us).
Besides, set_wq_data() _sets_ the workqueue pointer unconditionally, so
it can be assigned instead.
So, replacing the set_wq_data() with a straight assignment would be okay
in most cases.
The problem is where we end up tangling with test_and_set_bit() emulated
using spinlocks, and even then it's not a problem _provided_
test_and_set_bit() doesn't attempt to modify the word if the bit was
set.
If that's a problem, then a bitops-proofed assignment will be required -
equivalent to atomic_set() vs other atomic_xxx() ops.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:33:26 -07:00
|
|
|
/*
|
2010-06-29 02:07:14 -06:00
|
|
|
* Need to wake up a worker? Called from anything but currently
|
|
|
|
* running workers.
|
2012-07-12 15:46:37 -06:00
|
|
|
*
|
|
|
|
* Note that, because unbound workers never contribute to nr_running, this
|
2013-01-24 12:01:34 -07:00
|
|
|
* function will always return %true for unbound pools as long as the
|
2012-07-12 15:46:37 -06:00
|
|
|
* worklist isn't empty.
|
[PATCH] WorkStruct: Use direct assignment rather than cmpxchg()
Use direct assignment rather than cmpxchg() as the latter is unavailable
and unimplementable on some platforms and is actually unnecessary.
The use of cmpxchg() was to guard against two possibilities, neither of
which can actually occur:
(1) The pending flag may have been unset or may be cleared. However, given
where it's called, the pending flag is _always_ set. I don't think it
can be unset whilst we're in set_wq_data().
Once the work is enqueued to be actually run, the only way off the queue
is for it to be actually run.
If it's a delayed work item, then the bit can't be cleared by the timer
because we haven't started the timer yet. Also, the pending bit can't be
cleared by cancelling the delayed work _until_ the work item has had its
timer started.
(2) The workqueue pointer might change. This can only happen in two cases:
(a) The work item has just been queued to actually run, and so we're
protected by the appropriate workqueue spinlock.
(b) A delayed work item is being queued, and so the timer hasn't been
started yet, and so no one else knows about the work item or can
access it (the pending bit protects us).
Besides, set_wq_data() _sets_ the workqueue pointer unconditionally, so
it can be assigned instead.
So, replacing the set_wq_data() with a straight assignment would be okay
in most cases.
The problem is where we end up tangling with test_and_set_bit() emulated
using spinlocks, and even then it's not a problem _provided_
test_and_set_bit() doesn't attempt to modify the word if the bit was
set.
If that's a problem, then a bitops-proofed assignment will be required -
equivalent to atomic_set() vs other atomic_xxx() ops.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:33:26 -07:00
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool need_more_worker(struct worker_pool *pool)
|
2006-11-22 07:54:49 -07:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
return !list_empty(&pool->worklist) && __need_more_worker(pool);
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
[PATCH] WorkStruct: Use direct assignment rather than cmpxchg()
Use direct assignment rather than cmpxchg() as the latter is unavailable
and unimplementable on some platforms and is actually unnecessary.
The use of cmpxchg() was to guard against two possibilities, neither of
which can actually occur:
(1) The pending flag may have been unset or may be cleared. However, given
where it's called, the pending flag is _always_ set. I don't think it
can be unset whilst we're in set_wq_data().
Once the work is enqueued to be actually run, the only way off the queue
is for it to be actually run.
If it's a delayed work item, then the bit can't be cleared by the timer
because we haven't started the timer yet. Also, the pending bit can't be
cleared by cancelling the delayed work _until_ the work item has had its
timer started.
(2) The workqueue pointer might change. This can only happen in two cases:
(a) The work item has just been queued to actually run, and so we're
protected by the appropriate workqueue spinlock.
(b) A delayed work item is being queued, and so the timer hasn't been
started yet, and so no one else knows about the work item or can
access it (the pending bit protects us).
Besides, set_wq_data() _sets_ the workqueue pointer unconditionally, so
it can be assigned instead.
So, replacing the set_wq_data() with a straight assignment would be okay
in most cases.
The problem is where we end up tangling with test_and_set_bit() emulated
using spinlocks, and even then it's not a problem _provided_
test_and_set_bit() doesn't attempt to modify the word if the bit was
set.
If that's a problem, then a bitops-proofed assignment will be required -
equivalent to atomic_set() vs other atomic_xxx() ops.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:33:26 -07:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/* Can I start working? Called from busy but !running workers. */
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool may_start_working(struct worker_pool *pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
return pool->nr_idle;
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Do I need to keep working? Called from currently running workers. */
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool keep_working(struct worker_pool *pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2013-01-24 12:39:44 -07:00
|
|
|
return !list_empty(&pool->worklist) &&
|
|
|
|
atomic_read(&pool->nr_running) <= 1;
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Do we need a new worker? Called from manager. */
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool need_to_create_worker(struct worker_pool *pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
return need_more_worker(pool) && !may_start_working(pool);
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
2006-11-22 07:54:49 -07:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/* Do I need to be the manager? */
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool need_to_manage_workers(struct worker_pool *pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
return need_to_create_worker(pool) ||
|
2012-07-12 15:46:37 -06:00
|
|
|
(pool->flags & POOL_MANAGE_WORKERS);
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Do we have too many workers and should some go away? */
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool too_many_workers(struct worker_pool *pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2013-03-12 12:30:00 -06:00
|
|
|
bool managing = mutex_is_locked(&pool->manager_arb);
|
2012-07-12 15:46:37 -06:00
|
|
|
int nr_idle = pool->nr_idle + managing; /* manager is considered idle */
|
|
|
|
int nr_busy = pool->nr_workers - nr_idle;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
workqueue: reimplement idle worker rebinding
Currently rebind_workers() uses rebinds idle workers synchronously
before proceeding to requesting busy workers to rebind. This is
necessary because all workers on @worker_pool->idle_list must be bound
before concurrency management local wake-ups from the busy workers
take place.
Unfortunately, the synchronous idle rebinding is quite complicated.
This patch reimplements idle rebinding to simplify the code path.
Rather than trying to make all idle workers bound before rebinding
busy workers, we simply remove all to-be-bound idle workers from the
idle list and let them add themselves back after completing rebinding
(successful or not).
As only workers which finished rebinding can on on the idle worker
list, the idle worker list is guaranteed to have only bound workers
unless CPU went down again and local wake-ups are safe.
After the change, @worker_pool->nr_idle may deviate than the actual
number of idle workers on @worker_pool->idle_list. More specifically,
nr_idle may be non-zero while ->idle_list is empty. All users of
->nr_idle and ->idle_list are audited. The only affected one is
too_many_workers() which is updated to check %false if ->idle_list is
empty regardless of ->nr_idle.
After this patch, rebind_workers() no longer performs the nasty
idle-rebind retries which require temporary release of gcwq->lock, and
both unbinding and rebinding are atomic w.r.t. global_cwq->lock.
worker->idle_rebind and global_cwq->rebind_hold are now unnecessary
and removed along with the definition of struct idle_rebind.
Changed from V1:
1) remove unlikely from too_many_workers(), ->idle_list can be empty
anytime, even before this patch, no reason to use unlikely.
2) fix a small rebasing mistake.
(which is from rebasing the orignal fixing patch to for-next)
3) add a lot of comments.
4) clear WORKER_REBIND unconditionaly in idle_worker_rebind()
tj: Updated comments and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
2012-09-18 10:59:22 -06:00
|
|
|
/*
|
|
|
|
* nr_idle and idle_list may disagree if idle rebinding is in
|
|
|
|
* progress. Never return %true if idle_list is empty.
|
|
|
|
*/
|
|
|
|
if (list_empty(&pool->idle_list))
|
|
|
|
return false;
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
return nr_idle > 2 && (nr_idle - 2) * MAX_IDLE_WORKERS_RATIO >= nr_busy;
|
2006-11-22 07:54:49 -07:00
|
|
|
}
|
|
|
|
|
2010-04-23 09:40:40 -06:00
|
|
|
/*
|
2010-06-29 02:07:14 -06:00
|
|
|
* Wake up functions.
|
|
|
|
*/
|
|
|
|
|
2010-06-29 02:07:13 -06:00
|
|
|
/* Return the first worker. Safe with preemption disabled */
|
2012-07-12 15:46:37 -06:00
|
|
|
static struct worker *first_worker(struct worker_pool *pool)
|
2010-06-29 02:07:13 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
if (unlikely(list_empty(&pool->idle_list)))
|
2010-06-29 02:07:13 -06:00
|
|
|
return NULL;
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
return list_first_entry(&pool->idle_list, struct worker, entry);
|
2010-06-29 02:07:13 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* wake_up_worker - wake up an idle worker
|
2012-07-12 15:46:37 -06:00
|
|
|
* @pool: worker pool to wake worker from
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
2012-07-12 15:46:37 -06:00
|
|
|
* Wake up the first idle worker of @pool.
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2010-06-29 02:07:13 -06:00
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
static void wake_up_worker(struct worker_pool *pool)
|
2010-06-29 02:07:13 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker *worker = first_worker(pool);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
|
|
|
if (likely(worker))
|
|
|
|
wake_up_process(worker->task);
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:13 -06:00
|
|
|
/**
|
2010-06-29 02:07:14 -06:00
|
|
|
* wq_worker_waking_up - a worker is waking up
|
|
|
|
* @task: task waking up
|
|
|
|
* @cpu: CPU @task is waking up to
|
|
|
|
*
|
|
|
|
* This function is called during try_to_wake_up() when a worker is
|
|
|
|
* being awoken.
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
|
|
|
* spin_lock_irq(rq->lock)
|
|
|
|
*/
|
2013-03-12 12:29:59 -06:00
|
|
|
void wq_worker_waking_up(struct task_struct *task, int cpu)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
|
|
|
struct worker *worker = kthread_data(task);
|
|
|
|
|
2012-10-26 08:03:49 -06:00
|
|
|
if (!(worker->flags & WORKER_NOT_RUNNING)) {
|
2013-01-24 12:01:33 -07:00
|
|
|
WARN_ON_ONCE(worker->pool->cpu != cpu);
|
2013-01-24 12:39:44 -07:00
|
|
|
atomic_inc(&worker->pool->nr_running);
|
2012-10-26 08:03:49 -06:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* wq_worker_sleeping - a worker is going to sleep
|
|
|
|
* @task: task going to sleep
|
|
|
|
* @cpu: CPU in question, must be the current CPU number
|
|
|
|
*
|
|
|
|
* This function is called during schedule() when a busy worker is
|
|
|
|
* going to sleep. Worker on the same cpu can be woken up by
|
|
|
|
* returning pointer to its task.
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
|
|
|
* spin_lock_irq(rq->lock)
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-06-29 02:07:14 -06:00
|
|
|
* Worker task on @cpu to wake up, %NULL if none.
|
|
|
|
*/
|
2013-03-12 12:29:59 -06:00
|
|
|
struct task_struct *wq_worker_sleeping(struct task_struct *task, int cpu)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
|
|
|
struct worker *worker = kthread_data(task), *to_wakeup = NULL;
|
2013-01-17 18:16:24 -07:00
|
|
|
struct worker_pool *pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-01-17 18:16:24 -07:00
|
|
|
/*
|
|
|
|
* Rescuers, which may not have all the fields set up like normal
|
|
|
|
* workers, also reach here, let's not access anything before
|
|
|
|
* checking NOT_RUNNING.
|
|
|
|
*/
|
workqueue: It is likely that WORKER_NOT_RUNNING is true
Running the annotate branch profiler on three boxes, including my
main box that runs firefox, evolution, xchat, and is part of the distcc farm,
showed this with the likelys in the workqueue code:
correct incorrect % Function File Line
------- --------- - -------- ---- ----
96 996253 99 wq_worker_sleeping workqueue.c 703
96 996247 99 wq_worker_waking_up workqueue.c 677
The likely()s in this case were assuming that WORKER_NOT_RUNNING will
most likely be false. But this is not the case. The reason is
(and shown by adding trace_printks and testing it) that most of the time
WORKER_PREP is set.
In worker_thread() we have:
worker_clr_flags(worker, WORKER_PREP);
[ do work stuff ]
worker_set_flags(worker, WORKER_PREP, false);
(that 'false' means not to wake up an idle worker)
The wq_worker_sleeping() is called from schedule when a worker thread
is putting itself to sleep. Which happens most of the time outside
of that [ do work stuff ].
The wq_worker_waking_up is called by the wakeup worker code, which
is also callod outside that [ do work stuff ].
Thus, the likely and unlikely used by those two functions are actually
backwards.
Remove the annotation and let gcc figure it out.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
2010-12-03 21:12:33 -07:00
|
|
|
if (worker->flags & WORKER_NOT_RUNNING)
|
2010-06-29 02:07:14 -06:00
|
|
|
return NULL;
|
|
|
|
|
2013-01-17 18:16:24 -07:00
|
|
|
pool = worker->pool;
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/* this can only happen on the local cpu */
|
2013-03-12 12:29:57 -06:00
|
|
|
if (WARN_ON_ONCE(cpu != raw_smp_processor_id()))
|
|
|
|
return NULL;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* The counterpart of the following dec_and_test, implied mb,
|
|
|
|
* worklist not empty test sequence is in insert_work().
|
|
|
|
* Please read comment there.
|
|
|
|
*
|
2012-07-17 13:39:27 -06:00
|
|
|
* NOT_RUNNING is clear. This means that we're bound to and
|
|
|
|
* running on the local cpu w/ rq lock held and preemption
|
|
|
|
* disabled, which in turn means that none else could be
|
2013-01-24 12:01:33 -07:00
|
|
|
* manipulating idle_list, so dereferencing idle_list without pool
|
2012-07-17 13:39:27 -06:00
|
|
|
* lock is safe.
|
2010-06-29 02:07:14 -06:00
|
|
|
*/
|
2013-01-24 12:39:44 -07:00
|
|
|
if (atomic_dec_and_test(&pool->nr_running) &&
|
|
|
|
!list_empty(&pool->worklist))
|
2012-07-12 15:46:37 -06:00
|
|
|
to_wakeup = first_worker(pool);
|
2010-06-29 02:07:14 -06:00
|
|
|
return to_wakeup ? to_wakeup->task : NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* worker_set_flags - set worker flags and adjust nr_running accordingly
|
2010-07-02 02:03:50 -06:00
|
|
|
* @worker: self
|
2010-06-29 02:07:13 -06:00
|
|
|
* @flags: flags to set
|
|
|
|
* @wakeup: wakeup an idle worker if necessary
|
|
|
|
*
|
2010-06-29 02:07:14 -06:00
|
|
|
* Set @flags in @worker->flags and adjust nr_running accordingly. If
|
|
|
|
* nr_running becomes zero and @wakeup is %true, an idle worker is
|
|
|
|
* woken up.
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
2010-07-02 02:03:50 -06:00
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock)
|
2010-06-29 02:07:13 -06:00
|
|
|
*/
|
|
|
|
static inline void worker_set_flags(struct worker *worker, unsigned int flags,
|
|
|
|
bool wakeup)
|
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2010-07-02 02:03:50 -06:00
|
|
|
WARN_ON_ONCE(worker->task != current);
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/*
|
|
|
|
* If transitioning into NOT_RUNNING, adjust nr_running and
|
|
|
|
* wake up an idle worker as necessary if requested by
|
|
|
|
* @wakeup.
|
|
|
|
*/
|
|
|
|
if ((flags & WORKER_NOT_RUNNING) &&
|
|
|
|
!(worker->flags & WORKER_NOT_RUNNING)) {
|
|
|
|
if (wakeup) {
|
2013-01-24 12:39:44 -07:00
|
|
|
if (atomic_dec_and_test(&pool->nr_running) &&
|
2012-07-12 15:46:37 -06:00
|
|
|
!list_empty(&pool->worklist))
|
2012-07-12 15:46:37 -06:00
|
|
|
wake_up_worker(pool);
|
2010-06-29 02:07:14 -06:00
|
|
|
} else
|
2013-01-24 12:39:44 -07:00
|
|
|
atomic_dec(&pool->nr_running);
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:13 -06:00
|
|
|
worker->flags |= flags;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2010-06-29 02:07:14 -06:00
|
|
|
* worker_clr_flags - clear worker flags and adjust nr_running accordingly
|
2010-07-02 02:03:50 -06:00
|
|
|
* @worker: self
|
2010-06-29 02:07:13 -06:00
|
|
|
* @flags: flags to clear
|
|
|
|
*
|
2010-06-29 02:07:14 -06:00
|
|
|
* Clear @flags in @worker->flags and adjust nr_running accordingly.
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
2010-07-02 02:03:50 -06:00
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock)
|
2010-06-29 02:07:13 -06:00
|
|
|
*/
|
|
|
|
static inline void worker_clr_flags(struct worker *worker, unsigned int flags)
|
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
unsigned int oflags = worker->flags;
|
|
|
|
|
2010-07-02 02:03:50 -06:00
|
|
|
WARN_ON_ONCE(worker->task != current);
|
|
|
|
|
2010-06-29 02:07:13 -06:00
|
|
|
worker->flags &= ~flags;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2011-01-11 07:58:49 -07:00
|
|
|
/*
|
|
|
|
* If transitioning out of NOT_RUNNING, increment nr_running. Note
|
|
|
|
* that the nested NOT_RUNNING is not a noop. NOT_RUNNING is mask
|
|
|
|
* of multiple flags, not a single flag.
|
|
|
|
*/
|
2010-06-29 02:07:14 -06:00
|
|
|
if ((flags & WORKER_NOT_RUNNING) && (oflags & WORKER_NOT_RUNNING))
|
|
|
|
if (!(worker->flags & WORKER_NOT_RUNNING))
|
2013-01-24 12:39:44 -07:00
|
|
|
atomic_inc(&pool->nr_running);
|
2010-06-29 02:07:13 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:13 -06:00
|
|
|
/**
|
|
|
|
* find_worker_executing_work - find worker which is executing a work
|
2013-01-24 12:01:33 -07:00
|
|
|
* @pool: pool of interest
|
2010-06-29 02:07:13 -06:00
|
|
|
* @work: work to find worker for
|
|
|
|
*
|
2013-01-24 12:01:33 -07:00
|
|
|
* Find a worker which is executing @work on @pool by searching
|
|
|
|
* @pool->busy_hash which is keyed by the address of @work. For a worker
|
2012-12-18 11:35:02 -07:00
|
|
|
* to match, its current execution should match the address of @work and
|
|
|
|
* its work function. This is to avoid unwanted dependency between
|
|
|
|
* unrelated work executions through a work item being recycled while still
|
|
|
|
* being executed.
|
|
|
|
*
|
|
|
|
* This is a bit tricky. A work item may be freed once its execution
|
|
|
|
* starts and nothing prevents the freed area from being recycled for
|
|
|
|
* another work item. If the same work item address ends up being reused
|
|
|
|
* before the original execution finishes, workqueue will identify the
|
|
|
|
* recycled work item as currently executing and make it wait until the
|
|
|
|
* current execution finishes, introducing an unwanted dependency.
|
|
|
|
*
|
2013-03-13 17:51:36 -06:00
|
|
|
* This function checks the work item address and work function to avoid
|
|
|
|
* false positives. Note that this isn't complete as one may construct a
|
|
|
|
* work function which can introduce dependency onto itself through a
|
|
|
|
* recycled work item. Well, if somebody wants to shoot oneself in the
|
|
|
|
* foot that badly, there's only so much we can do, and if such deadlock
|
|
|
|
* actually occurs, it should be easy to locate the culprit work function.
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2010-06-29 02:07:13 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
|
|
|
* Pointer to worker which is executing @work if found, %NULL
|
2010-06-29 02:07:13 -06:00
|
|
|
* otherwise.
|
2010-04-23 09:40:40 -06:00
|
|
|
*/
|
2013-01-24 12:01:33 -07:00
|
|
|
static struct worker *find_worker_executing_work(struct worker_pool *pool,
|
2010-06-29 02:07:13 -06:00
|
|
|
struct work_struct *work)
|
2010-04-23 09:40:40 -06:00
|
|
|
{
|
2012-12-17 08:01:23 -07:00
|
|
|
struct worker *worker;
|
|
|
|
|
hlist: drop the node parameter from iterators
I'm not sure why, but the hlist for each entry iterators were conceived
list_for_each_entry(pos, head, member)
The hlist ones were greedy and wanted an extra parameter:
hlist_for_each_entry(tpos, pos, head, member)
Why did they need an extra pos parameter? I'm not quite sure. Not only
they don't really need it, it also prevents the iterator from looking
exactly like the list iterator, which is unfortunate.
Besides the semantic patch, there was some manual work required:
- Fix up the actual hlist iterators in linux/list.h
- Fix up the declaration of other iterators based on the hlist ones.
- A very small amount of places were using the 'node' parameter, this
was modified to use 'obj->member' instead.
- Coccinelle didn't handle the hlist_for_each_entry_safe iterator
properly, so those had to be fixed up manually.
The semantic patch which is mostly the work of Peter Senna Tschudin is here:
@@
iterator name hlist_for_each_entry, hlist_for_each_entry_continue, hlist_for_each_entry_from, hlist_for_each_entry_rcu, hlist_for_each_entry_rcu_bh, hlist_for_each_entry_continue_rcu_bh, for_each_busy_worker, ax25_uid_for_each, ax25_for_each, inet_bind_bucket_for_each, sctp_for_each_hentry, sk_for_each, sk_for_each_rcu, sk_for_each_from, sk_for_each_safe, sk_for_each_bound, hlist_for_each_entry_safe, hlist_for_each_entry_continue_rcu, nr_neigh_for_each, nr_neigh_for_each_safe, nr_node_for_each, nr_node_for_each_safe, for_each_gfn_indirect_valid_sp, for_each_gfn_sp, for_each_host;
type T;
expression a,c,d,e;
identifier b;
statement S;
@@
-T b;
<+... when != b
(
hlist_for_each_entry(a,
- b,
c, d) S
|
hlist_for_each_entry_continue(a,
- b,
c) S
|
hlist_for_each_entry_from(a,
- b,
c) S
|
hlist_for_each_entry_rcu(a,
- b,
c, d) S
|
hlist_for_each_entry_rcu_bh(a,
- b,
c, d) S
|
hlist_for_each_entry_continue_rcu_bh(a,
- b,
c) S
|
for_each_busy_worker(a, c,
- b,
d) S
|
ax25_uid_for_each(a,
- b,
c) S
|
ax25_for_each(a,
- b,
c) S
|
inet_bind_bucket_for_each(a,
- b,
c) S
|
sctp_for_each_hentry(a,
- b,
c) S
|
sk_for_each(a,
- b,
c) S
|
sk_for_each_rcu(a,
- b,
c) S
|
sk_for_each_from
-(a, b)
+(a)
S
+ sk_for_each_from(a) S
|
sk_for_each_safe(a,
- b,
c, d) S
|
sk_for_each_bound(a,
- b,
c) S
|
hlist_for_each_entry_safe(a,
- b,
c, d, e) S
|
hlist_for_each_entry_continue_rcu(a,
- b,
c) S
|
nr_neigh_for_each(a,
- b,
c) S
|
nr_neigh_for_each_safe(a,
- b,
c, d) S
|
nr_node_for_each(a,
- b,
c) S
|
nr_node_for_each_safe(a,
- b,
c, d) S
|
- for_each_gfn_sp(a, c, d, b) S
+ for_each_gfn_sp(a, c, d) S
|
- for_each_gfn_indirect_valid_sp(a, c, d, b) S
+ for_each_gfn_indirect_valid_sp(a, c, d) S
|
for_each_host(a,
- b,
c) S
|
for_each_host_safe(a,
- b,
c, d) S
|
for_each_mesh_entry(a,
- b,
c, d) S
)
...+>
[akpm@linux-foundation.org: drop bogus change from net/ipv4/raw.c]
[akpm@linux-foundation.org: drop bogus hunk from net/ipv6/raw.c]
[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix warnings]
[akpm@linux-foudnation.org: redo intrusive kvm changes]
Tested-by: Peter Senna Tschudin <peter.senna@gmail.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2013-02-27 18:06:00 -07:00
|
|
|
hash_for_each_possible(pool->busy_hash, worker, hentry,
|
2012-12-18 11:35:02 -07:00
|
|
|
(unsigned long)work)
|
|
|
|
if (worker->current_work == work &&
|
|
|
|
worker->current_func == work->func)
|
2012-12-17 08:01:23 -07:00
|
|
|
return worker;
|
|
|
|
|
|
|
|
return NULL;
|
2010-04-23 09:40:40 -06:00
|
|
|
}
|
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
/**
|
|
|
|
* move_linked_works - move linked works to a list
|
|
|
|
* @work: start of series of works to be scheduled
|
|
|
|
* @head: target list to append @work to
|
|
|
|
* @nextp: out paramter for nested worklist walking
|
|
|
|
*
|
|
|
|
* Schedule linked works starting from @work to @head. Work series to
|
|
|
|
* be scheduled starts at @work and includes any consecutive work with
|
|
|
|
* WORK_STRUCT_LINKED set in its predecessor.
|
|
|
|
*
|
|
|
|
* If @nextp is not NULL, it's updated to point to the next work of
|
|
|
|
* the last scheduled work. This allows move_linked_works() to be
|
|
|
|
* nested inside outer list_for_each_entry_safe().
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2012-08-03 11:30:46 -06:00
|
|
|
*/
|
|
|
|
static void move_linked_works(struct work_struct *work, struct list_head *head,
|
|
|
|
struct work_struct **nextp)
|
|
|
|
{
|
|
|
|
struct work_struct *n;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Linked worklist will always end before the end of the list,
|
|
|
|
* use NULL for list head.
|
|
|
|
*/
|
|
|
|
list_for_each_entry_safe_from(work, n, NULL, entry) {
|
|
|
|
list_move_tail(&work->entry, head);
|
|
|
|
if (!(*work_data_bits(work) & WORK_STRUCT_LINKED))
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we're already inside safe list traversal and have moved
|
|
|
|
* multiple works to the scheduled queue, the next position
|
|
|
|
* needs to be updated.
|
|
|
|
*/
|
|
|
|
if (nextp)
|
|
|
|
*nextp = n;
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/**
|
|
|
|
* get_pwq - get an extra reference on the specified pool_workqueue
|
|
|
|
* @pwq: pool_workqueue to get
|
|
|
|
*
|
|
|
|
* Obtain an extra reference on @pwq. The caller should guarantee that
|
|
|
|
* @pwq has positive refcnt and be holding the matching pool->lock.
|
|
|
|
*/
|
|
|
|
static void get_pwq(struct pool_workqueue *pwq)
|
|
|
|
{
|
|
|
|
lockdep_assert_held(&pwq->pool->lock);
|
|
|
|
WARN_ON_ONCE(pwq->refcnt <= 0);
|
|
|
|
pwq->refcnt++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* put_pwq - put a pool_workqueue reference
|
|
|
|
* @pwq: pool_workqueue to put
|
|
|
|
*
|
|
|
|
* Drop a reference of @pwq. If its refcnt reaches zero, schedule its
|
|
|
|
* destruction. The caller should be holding the matching pool->lock.
|
|
|
|
*/
|
|
|
|
static void put_pwq(struct pool_workqueue *pwq)
|
|
|
|
{
|
|
|
|
lockdep_assert_held(&pwq->pool->lock);
|
|
|
|
if (likely(--pwq->refcnt))
|
|
|
|
return;
|
|
|
|
if (WARN_ON_ONCE(!(pwq->wq->flags & WQ_UNBOUND)))
|
|
|
|
return;
|
|
|
|
/*
|
|
|
|
* @pwq can't be released under pool->lock, bounce to
|
|
|
|
* pwq_unbound_release_workfn(). This never recurses on the same
|
|
|
|
* pool->lock as this path is taken only for unbound workqueues and
|
|
|
|
* the release work item is scheduled on a per-cpu workqueue. To
|
|
|
|
* avoid lockdep warning, unbound pool->locks are given lockdep
|
|
|
|
* subclass of 1 in get_unbound_pool().
|
|
|
|
*/
|
|
|
|
schedule_work(&pwq->unbound_release_work);
|
|
|
|
}
|
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
/**
|
|
|
|
* put_pwq_unlocked - put_pwq() with surrounding pool lock/unlock
|
|
|
|
* @pwq: pool_workqueue to put (can be %NULL)
|
|
|
|
*
|
|
|
|
* put_pwq() with locking. This function also allows %NULL @pwq.
|
|
|
|
*/
|
|
|
|
static void put_pwq_unlocked(struct pool_workqueue *pwq)
|
|
|
|
{
|
|
|
|
if (pwq) {
|
|
|
|
/*
|
|
|
|
* As both pwqs and pools are sched-RCU protected, the
|
|
|
|
* following lock operations are safe.
|
|
|
|
*/
|
|
|
|
spin_lock_irq(&pwq->pool->lock);
|
|
|
|
put_pwq(pwq);
|
|
|
|
spin_unlock_irq(&pwq->pool->lock);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
static void pwq_activate_delayed_work(struct work_struct *work)
|
2012-08-03 11:30:46 -06:00
|
|
|
{
|
2013-02-13 20:29:12 -07:00
|
|
|
struct pool_workqueue *pwq = get_work_pwq(work);
|
2012-08-03 11:30:46 -06:00
|
|
|
|
|
|
|
trace_workqueue_activate_work(work);
|
2013-02-13 20:29:12 -07:00
|
|
|
move_linked_works(work, &pwq->pool->worklist, NULL);
|
2012-08-03 11:30:46 -06:00
|
|
|
__clear_bit(WORK_STRUCT_DELAYED_BIT, work_data_bits(work));
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq->nr_active++;
|
2012-08-03 11:30:46 -06:00
|
|
|
}
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
static void pwq_activate_first_delayed(struct pool_workqueue *pwq)
|
workqueue: fix possible stall on try_to_grab_pending() of a delayed work item
Currently, when try_to_grab_pending() grabs a delayed work item, it
leaves its linked work items alone on the delayed_works. The linked
work items are always NO_COLOR and will cause future
cwq_activate_first_delayed() increase cwq->nr_active incorrectly, and
may cause the whole cwq to stall. For example,
state: cwq->max_active = 1, cwq->nr_active = 1
one work in cwq->pool, many in cwq->delayed_works.
step1: try_to_grab_pending() removes a work item from delayed_works
but leaves its NO_COLOR linked work items on it.
step2: Later on, cwq_activate_first_delayed() activates the linked
work item increasing ->nr_active.
step3: cwq->nr_active = 1, but all activated work items of the cwq are
NO_COLOR. When they finish, cwq->nr_active will not be
decreased due to NO_COLOR, and no further work items will be
activated from cwq->delayed_works. the cwq stalls.
Fix it by ensuring the target work item is activated before stealing
PENDING in try_to_grab_pending(). This ensures that all the linked
work items are activated without incorrectly bumping cwq->nr_active.
tj: Updated comment and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@kernel.org
2012-09-18 11:40:00 -06:00
|
|
|
{
|
2013-02-13 20:29:12 -07:00
|
|
|
struct work_struct *work = list_first_entry(&pwq->delayed_works,
|
workqueue: fix possible stall on try_to_grab_pending() of a delayed work item
Currently, when try_to_grab_pending() grabs a delayed work item, it
leaves its linked work items alone on the delayed_works. The linked
work items are always NO_COLOR and will cause future
cwq_activate_first_delayed() increase cwq->nr_active incorrectly, and
may cause the whole cwq to stall. For example,
state: cwq->max_active = 1, cwq->nr_active = 1
one work in cwq->pool, many in cwq->delayed_works.
step1: try_to_grab_pending() removes a work item from delayed_works
but leaves its NO_COLOR linked work items on it.
step2: Later on, cwq_activate_first_delayed() activates the linked
work item increasing ->nr_active.
step3: cwq->nr_active = 1, but all activated work items of the cwq are
NO_COLOR. When they finish, cwq->nr_active will not be
decreased due to NO_COLOR, and no further work items will be
activated from cwq->delayed_works. the cwq stalls.
Fix it by ensuring the target work item is activated before stealing
PENDING in try_to_grab_pending(). This ensures that all the linked
work items are activated without incorrectly bumping cwq->nr_active.
tj: Updated comment and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@kernel.org
2012-09-18 11:40:00 -06:00
|
|
|
struct work_struct, entry);
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq_activate_delayed_work(work);
|
workqueue: fix possible stall on try_to_grab_pending() of a delayed work item
Currently, when try_to_grab_pending() grabs a delayed work item, it
leaves its linked work items alone on the delayed_works. The linked
work items are always NO_COLOR and will cause future
cwq_activate_first_delayed() increase cwq->nr_active incorrectly, and
may cause the whole cwq to stall. For example,
state: cwq->max_active = 1, cwq->nr_active = 1
one work in cwq->pool, many in cwq->delayed_works.
step1: try_to_grab_pending() removes a work item from delayed_works
but leaves its NO_COLOR linked work items on it.
step2: Later on, cwq_activate_first_delayed() activates the linked
work item increasing ->nr_active.
step3: cwq->nr_active = 1, but all activated work items of the cwq are
NO_COLOR. When they finish, cwq->nr_active will not be
decreased due to NO_COLOR, and no further work items will be
activated from cwq->delayed_works. the cwq stalls.
Fix it by ensuring the target work item is activated before stealing
PENDING in try_to_grab_pending(). This ensures that all the linked
work items are activated without incorrectly bumping cwq->nr_active.
tj: Updated comment and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@kernel.org
2012-09-18 11:40:00 -06:00
|
|
|
}
|
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
/**
|
2013-02-13 20:29:12 -07:00
|
|
|
* pwq_dec_nr_in_flight - decrement pwq's nr_in_flight
|
|
|
|
* @pwq: pwq of interest
|
2012-08-03 11:30:46 -06:00
|
|
|
* @color: color of work which left the queue
|
|
|
|
*
|
|
|
|
* A work either has completed or is removed from pending queue,
|
2013-02-13 20:29:12 -07:00
|
|
|
* decrement nr_in_flight of its pwq and handle workqueue flushing.
|
2012-08-03 11:30:46 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2012-08-03 11:30:46 -06:00
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
static void pwq_dec_nr_in_flight(struct pool_workqueue *pwq, int color)
|
2012-08-03 11:30:46 -06:00
|
|
|
{
|
2013-03-12 12:30:04 -06:00
|
|
|
/* uncolored work items don't participate in flushing or nr_active */
|
2012-08-03 11:30:46 -06:00
|
|
|
if (color == WORK_NO_COLOR)
|
2013-03-12 12:30:04 -06:00
|
|
|
goto out_put;
|
2012-08-03 11:30:46 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq->nr_in_flight[color]--;
|
2012-08-03 11:30:46 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq->nr_active--;
|
|
|
|
if (!list_empty(&pwq->delayed_works)) {
|
2012-09-18 11:40:00 -06:00
|
|
|
/* one down, submit a delayed one */
|
2013-02-13 20:29:12 -07:00
|
|
|
if (pwq->nr_active < pwq->max_active)
|
|
|
|
pwq_activate_first_delayed(pwq);
|
2012-08-03 11:30:46 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/* is flush in progress and are we at the flushing tip? */
|
2013-02-13 20:29:12 -07:00
|
|
|
if (likely(pwq->flush_color != color))
|
2013-03-12 12:30:04 -06:00
|
|
|
goto out_put;
|
2012-08-03 11:30:46 -06:00
|
|
|
|
|
|
|
/* are there still in-flight works? */
|
2013-02-13 20:29:12 -07:00
|
|
|
if (pwq->nr_in_flight[color])
|
2013-03-12 12:30:04 -06:00
|
|
|
goto out_put;
|
2012-08-03 11:30:46 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
/* this pwq is done, clear flush_color */
|
|
|
|
pwq->flush_color = -1;
|
2012-08-03 11:30:46 -06:00
|
|
|
|
|
|
|
/*
|
2013-02-13 20:29:12 -07:00
|
|
|
* If this was the last pwq, wake up the first flusher. It
|
2012-08-03 11:30:46 -06:00
|
|
|
* will handle the rest.
|
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
if (atomic_dec_and_test(&pwq->wq->nr_pwqs_to_flush))
|
|
|
|
complete(&pwq->wq->first_flusher->done);
|
2013-03-12 12:30:04 -06:00
|
|
|
out_put:
|
|
|
|
put_pwq(pwq);
|
2012-08-03 11:30:46 -06:00
|
|
|
}
|
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
/**
|
2012-08-03 11:30:46 -06:00
|
|
|
* try_to_grab_pending - steal work item from worklist and disable irq
|
2012-08-03 11:30:46 -06:00
|
|
|
* @work: work item to steal
|
|
|
|
* @is_dwork: @work is a delayed_work
|
2012-08-03 11:30:46 -06:00
|
|
|
* @flags: place to store irq state
|
2012-08-03 11:30:46 -06:00
|
|
|
*
|
|
|
|
* Try to grab PENDING bit of @work. This function can handle @work in any
|
2013-07-31 15:59:24 -06:00
|
|
|
* stable state - idle, on timer or on worklist.
|
2012-08-03 11:30:46 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2012-08-03 11:30:46 -06:00
|
|
|
* 1 if @work was pending and we successfully stole PENDING
|
|
|
|
* 0 if @work was idle and we claimed PENDING
|
|
|
|
* -EAGAIN if PENDING couldn't be grabbed at the moment, safe to busy-retry
|
2012-08-03 11:30:46 -06:00
|
|
|
* -ENOENT if someone else is canceling @work, this state may persist
|
|
|
|
* for arbitrarily long
|
2012-08-03 11:30:46 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Note:
|
2012-08-03 11:30:46 -06:00
|
|
|
* On >= 0 return, the caller owns @work's PENDING bit. To avoid getting
|
2012-08-21 14:18:24 -06:00
|
|
|
* interrupted while holding PENDING and @work off queue, irq must be
|
|
|
|
* disabled on entry. This, combined with delayed_work->timer being
|
|
|
|
* irqsafe, ensures that we return -EAGAIN for finite short period of time.
|
2012-08-03 11:30:46 -06:00
|
|
|
*
|
|
|
|
* On successful return, >= 0, irq is disabled and the caller is
|
|
|
|
* responsible for releasing it using local_irq_restore(*@flags).
|
|
|
|
*
|
2012-08-21 14:18:24 -06:00
|
|
|
* This function is safe to call from any context including IRQ handler.
|
2012-08-03 11:30:46 -06:00
|
|
|
*/
|
2012-08-03 11:30:46 -06:00
|
|
|
static int try_to_grab_pending(struct work_struct *work, bool is_dwork,
|
|
|
|
unsigned long *flags)
|
2012-08-03 11:30:46 -06:00
|
|
|
{
|
2013-01-24 12:01:33 -07:00
|
|
|
struct worker_pool *pool;
|
2013-02-13 20:29:12 -07:00
|
|
|
struct pool_workqueue *pwq;
|
2012-08-03 11:30:46 -06:00
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
local_irq_save(*flags);
|
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
/* try to steal the timer if it exists */
|
|
|
|
if (is_dwork) {
|
|
|
|
struct delayed_work *dwork = to_delayed_work(work);
|
|
|
|
|
2012-08-21 14:18:24 -06:00
|
|
|
/*
|
|
|
|
* dwork->timer is irqsafe. If del_timer() fails, it's
|
|
|
|
* guaranteed that the timer is not queued anywhere and not
|
|
|
|
* running on the local CPU.
|
|
|
|
*/
|
2012-08-03 11:30:46 -06:00
|
|
|
if (likely(del_timer(&dwork->timer)))
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* try to claim PENDING the normal way */
|
2012-08-03 11:30:46 -06:00
|
|
|
if (!test_and_set_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(work)))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The queueing is in progress, or it is already queued. Try to
|
|
|
|
* steal it from ->worklist without clearing WORK_STRUCT_PENDING.
|
|
|
|
*/
|
2013-01-24 12:01:33 -07:00
|
|
|
pool = get_work_pool(work);
|
|
|
|
if (!pool)
|
2012-08-03 11:30:46 -06:00
|
|
|
goto fail;
|
2012-08-03 11:30:46 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock(&pool->lock);
|
workqueue: simplify is-work-item-queued-here test
Currently, determining whether a work item is queued on a locked pool
involves somewhat convoluted memory barrier dancing. It goes like the
following.
* When a work item is queued on a pool, work->data is updated before
work->entry is linked to the pending list with a wmb() inbetween.
* When trying to determine whether a work item is currently queued on
a pool pointed to by work->data, it locks the pool and looks at
work->entry. If work->entry is linked, we then do rmb() and then
check whether work->data points to the current pool.
This works because, work->data can only point to a pool if it
currently is or were on the pool and,
* If it currently is on the pool, the tests would obviously succeed.
* It it left the pool, its work->entry was cleared under pool->lock,
so if we're seeing non-empty work->entry, it has to be from the work
item being linked on another pool. Because work->data is updated
before work->entry is linked with wmb() inbetween, work->data update
from another pool is guaranteed to be visible if we do rmb() after
seeing non-empty work->entry. So, we either see empty work->entry
or we see updated work->data pointin to another pool.
While this works, it's convoluted, to put it mildly. With recent
updates, it's now guaranteed that work->data points to cwq only while
the work item is queued and that updating work->data to point to cwq
or back to pool is done under pool->lock, so we can simply test
whether work->data points to cwq which is associated with the
currently locked pool instead of the convoluted memory barrier
dancing.
This patch replaces the memory barrier based "are you still here,
really?" test with much simpler "does work->data points to me?" test -
if work->data points to a cwq which is associated with the currently
locked pool, the work item is guaranteed to be queued on the pool as
work->data can start and stop pointing to such cwq only under
pool->lock and the start and stop coincide with queue and dequeue.
tj: Rewrote the comments and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
2013-02-06 19:04:53 -07:00
|
|
|
/*
|
2013-02-13 20:29:12 -07:00
|
|
|
* work->data is guaranteed to point to pwq only while the work
|
|
|
|
* item is queued on pwq->wq, and both updating work->data to point
|
|
|
|
* to pwq on queueing and to pool on dequeueing are done under
|
|
|
|
* pwq->pool->lock. This in turn guarantees that, if work->data
|
|
|
|
* points to pwq which is associated with a locked pool, the work
|
workqueue: simplify is-work-item-queued-here test
Currently, determining whether a work item is queued on a locked pool
involves somewhat convoluted memory barrier dancing. It goes like the
following.
* When a work item is queued on a pool, work->data is updated before
work->entry is linked to the pending list with a wmb() inbetween.
* When trying to determine whether a work item is currently queued on
a pool pointed to by work->data, it locks the pool and looks at
work->entry. If work->entry is linked, we then do rmb() and then
check whether work->data points to the current pool.
This works because, work->data can only point to a pool if it
currently is or were on the pool and,
* If it currently is on the pool, the tests would obviously succeed.
* It it left the pool, its work->entry was cleared under pool->lock,
so if we're seeing non-empty work->entry, it has to be from the work
item being linked on another pool. Because work->data is updated
before work->entry is linked with wmb() inbetween, work->data update
from another pool is guaranteed to be visible if we do rmb() after
seeing non-empty work->entry. So, we either see empty work->entry
or we see updated work->data pointin to another pool.
While this works, it's convoluted, to put it mildly. With recent
updates, it's now guaranteed that work->data points to cwq only while
the work item is queued and that updating work->data to point to cwq
or back to pool is done under pool->lock, so we can simply test
whether work->data points to cwq which is associated with the
currently locked pool instead of the convoluted memory barrier
dancing.
This patch replaces the memory barrier based "are you still here,
really?" test with much simpler "does work->data points to me?" test -
if work->data points to a cwq which is associated with the currently
locked pool, the work item is guaranteed to be queued on the pool as
work->data can start and stop pointing to such cwq only under
pool->lock and the start and stop coincide with queue and dequeue.
tj: Rewrote the comments and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
2013-02-06 19:04:53 -07:00
|
|
|
* item is currently queued on that pool.
|
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq = get_work_pwq(work);
|
|
|
|
if (pwq && pwq->pool == pool) {
|
2013-02-06 19:04:53 -07:00
|
|
|
debug_work_deactivate(work);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* A delayed work item cannot be grabbed directly because
|
|
|
|
* it might have linked NO_COLOR work items which, if left
|
2013-02-13 20:29:12 -07:00
|
|
|
* on the delayed_list, will confuse pwq->nr_active
|
2013-02-06 19:04:53 -07:00
|
|
|
* management later on and cause stall. Make sure the work
|
|
|
|
* item is activated before grabbing.
|
|
|
|
*/
|
|
|
|
if (*work_data_bits(work) & WORK_STRUCT_DELAYED)
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq_activate_delayed_work(work);
|
2013-02-06 19:04:53 -07:00
|
|
|
|
|
|
|
list_del_init(&work->entry);
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq_dec_nr_in_flight(get_work_pwq(work), get_work_color(work));
|
2013-02-06 19:04:53 -07:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
/* work->data points to pwq iff queued, point to pool */
|
2013-02-06 19:04:53 -07:00
|
|
|
set_work_pool_and_keep_pending(work, pool->id);
|
|
|
|
|
|
|
|
spin_unlock(&pool->lock);
|
|
|
|
return 1;
|
2012-08-03 11:30:46 -06:00
|
|
|
}
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock(&pool->lock);
|
2012-08-03 11:30:46 -06:00
|
|
|
fail:
|
|
|
|
local_irq_restore(*flags);
|
|
|
|
if (work_is_canceling(work))
|
|
|
|
return -ENOENT;
|
|
|
|
cpu_relax();
|
2012-08-03 11:30:46 -06:00
|
|
|
return -EAGAIN;
|
2012-08-03 11:30:46 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
/**
|
2013-01-24 12:01:34 -07:00
|
|
|
* insert_work - insert a work into a pool
|
2013-02-13 20:29:12 -07:00
|
|
|
* @pwq: pwq @work belongs to
|
2010-06-29 02:07:10 -06:00
|
|
|
* @work: work to insert
|
|
|
|
* @head: insertion point
|
|
|
|
* @extra_flags: extra WORK_STRUCT_* flags to set
|
|
|
|
*
|
2013-02-13 20:29:12 -07:00
|
|
|
* Insert @work which belongs to @pwq after @head. @extra_flags is or'd to
|
2013-01-24 12:01:34 -07:00
|
|
|
* work_struct flags.
|
2010-06-29 02:07:10 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2010-06-29 02:07:10 -06:00
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
static void insert_work(struct pool_workqueue *pwq, struct work_struct *work,
|
|
|
|
struct list_head *head, unsigned int extra_flags)
|
implement flush_work()
A basic problem with flush_scheduled_work() is that it blocks behind _all_
presently-queued works, rather than just the work whcih the caller wants to
flush. If the caller holds some lock, and if one of the queued work happens
to want that lock as well then accidental deadlocks can occur.
One example of this is the phy layer: it wants to flush work while holding
rtnl_lock(). But if a linkwatch event happens to be queued, the phy code will
deadlock because the linkwatch callback function takes rtnl_lock.
So we implement a new function which will flush a *single* work - just the one
which the caller wants to free up. Thus we avoid the accidental deadlocks
which can arise from unrelated subsystems' callbacks taking shared locks.
flush_work() non-blockingly dequeues the work_struct which we want to kill,
then it waits for its handler to complete on all CPUs.
Add ->current_work to the "struct cpu_workqueue_struct", it points to
currently running "struct work_struct". When flush_work(work) detects
->current_work == work, it inserts a barrier at the _head_ of ->worklist
(and thus right _after_ that work) and waits for completition. This means
that the next work fired on that CPU will be this barrier, or another
barrier queued by concurrent flush_work(), so the caller of flush_work()
will be woken before any "regular" work has a chance to run.
When wait_on_work() unlocks workqueue_mutex (or whatever we choose to protect
against CPU hotplug), CPU may go away. But in that case take_over_work() will
move a barrier we queued to another CPU, it will be fired sometime, and
wait_on_work() will be woken.
Actually, we are doing cleanup_workqueue_thread()->kthread_stop() before
take_over_work(), so cwq->thread should complete its ->worklist (and thus
the barrier), because currently we don't check kthread_should_stop() in
run_workqueue(). But even if we did, everything should be ok.
[akpm@osdl.org: cleanup]
[akpm@osdl.org: add flush_work_keventd() wrapper]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:33:52 -06:00
|
|
|
{
|
2013-02-13 20:29:12 -07:00
|
|
|
struct worker_pool *pool = pwq->pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
/* we own @work, set data and link */
|
2013-02-13 20:29:12 -07:00
|
|
|
set_work_pwq(work, pwq, extra_flags);
|
2008-07-25 02:47:47 -06:00
|
|
|
list_add_tail(&work->entry, head);
|
2013-03-12 12:30:04 -06:00
|
|
|
get_pwq(pwq);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
/*
|
2013-03-13 17:51:36 -06:00
|
|
|
* Ensure either wq_worker_sleeping() sees the above
|
|
|
|
* list_add_tail() or we see zero nr_running to avoid workers lying
|
|
|
|
* around lazily while there are works to be processed.
|
2010-06-29 02:07:14 -06:00
|
|
|
*/
|
|
|
|
smp_mb();
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
if (__need_more_worker(pool))
|
|
|
|
wake_up_worker(pool);
|
implement flush_work()
A basic problem with flush_scheduled_work() is that it blocks behind _all_
presently-queued works, rather than just the work whcih the caller wants to
flush. If the caller holds some lock, and if one of the queued work happens
to want that lock as well then accidental deadlocks can occur.
One example of this is the phy layer: it wants to flush work while holding
rtnl_lock(). But if a linkwatch event happens to be queued, the phy code will
deadlock because the linkwatch callback function takes rtnl_lock.
So we implement a new function which will flush a *single* work - just the one
which the caller wants to free up. Thus we avoid the accidental deadlocks
which can arise from unrelated subsystems' callbacks taking shared locks.
flush_work() non-blockingly dequeues the work_struct which we want to kill,
then it waits for its handler to complete on all CPUs.
Add ->current_work to the "struct cpu_workqueue_struct", it points to
currently running "struct work_struct". When flush_work(work) detects
->current_work == work, it inserts a barrier at the _head_ of ->worklist
(and thus right _after_ that work) and waits for completition. This means
that the next work fired on that CPU will be this barrier, or another
barrier queued by concurrent flush_work(), so the caller of flush_work()
will be woken before any "regular" work has a chance to run.
When wait_on_work() unlocks workqueue_mutex (or whatever we choose to protect
against CPU hotplug), CPU may go away. But in that case take_over_work() will
move a barrier we queued to another CPU, it will be fired sometime, and
wait_on_work() will be woken.
Actually, we are doing cleanup_workqueue_thread()->kthread_stop() before
take_over_work(), so cwq->thread should complete its ->worklist (and thus
the barrier), because currently we don't check kthread_should_stop() in
run_workqueue(). But even if we did, everything should be ok.
[akpm@osdl.org: cleanup]
[akpm@osdl.org: add flush_work_keventd() wrapper]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:33:52 -06:00
|
|
|
}
|
|
|
|
|
2010-12-20 11:32:04 -07:00
|
|
|
/*
|
|
|
|
* Test whether @work is being queued from another work executing on the
|
2013-02-13 20:29:10 -07:00
|
|
|
* same workqueue.
|
2010-12-20 11:32:04 -07:00
|
|
|
*/
|
|
|
|
static bool is_chained_work(struct workqueue_struct *wq)
|
|
|
|
{
|
2013-02-13 20:29:10 -07:00
|
|
|
struct worker *worker;
|
|
|
|
|
|
|
|
worker = current_wq_worker();
|
|
|
|
/*
|
|
|
|
* Return %true iff I'm a worker execuing a work item on @wq. If
|
|
|
|
* I'm @worker, it's safe to dereference it without locking.
|
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
return worker && worker->current_pwq->wq == wq;
|
2010-12-20 11:32:04 -07:00
|
|
|
}
|
|
|
|
|
2013-03-12 12:29:59 -06:00
|
|
|
static void __queue_work(int cpu, struct workqueue_struct *wq,
|
2005-04-16 16:20:36 -06:00
|
|
|
struct work_struct *work)
|
|
|
|
{
|
2013-02-13 20:29:12 -07:00
|
|
|
struct pool_workqueue *pwq;
|
2013-03-12 12:30:04 -06:00
|
|
|
struct worker_pool *last_pool;
|
2010-06-29 02:07:12 -06:00
|
|
|
struct list_head *worklist;
|
2010-08-25 02:33:56 -06:00
|
|
|
unsigned int work_flags;
|
2012-08-15 08:25:37 -06:00
|
|
|
unsigned int req_cpu = cpu;
|
2012-08-03 11:30:45 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* While a work item is PENDING && off queue, a task trying to
|
|
|
|
* steal the PENDING will busy-loop waiting for it to either get
|
|
|
|
* queued or lose PENDING. Grabbing PENDING and queueing should
|
|
|
|
* happen with IRQ disabled.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(!irqs_disabled());
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2009-11-15 09:09:48 -07:00
|
|
|
debug_work_activate(work);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-09-08 23:13:58 -06:00
|
|
|
/* if draining, only works from the same workqueue are allowed */
|
2013-03-12 12:30:04 -06:00
|
|
|
if (unlikely(wq->flags & __WQ_DRAINING) &&
|
2010-12-20 11:32:04 -07:00
|
|
|
WARN_ON_ONCE(!is_chained_work(wq)))
|
2010-08-24 06:22:47 -06:00
|
|
|
return;
|
2013-03-12 12:30:04 -06:00
|
|
|
retry:
|
2013-04-01 12:23:35 -06:00
|
|
|
if (req_cpu == WORK_CPU_UNBOUND)
|
|
|
|
cpu = raw_smp_processor_id();
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/* pwq which will be used unless @work is executing elsewhere */
|
2013-04-01 12:23:35 -06:00
|
|
|
if (!(wq->flags & WQ_UNBOUND))
|
2013-03-12 12:30:00 -06:00
|
|
|
pwq = per_cpu_ptr(wq->cpu_pwqs, cpu);
|
2013-04-01 12:23:35 -06:00
|
|
|
else
|
|
|
|
pwq = unbound_pwq_by_node(wq, cpu_to_node(cpu));
|
workqueue: make all workqueues non-reentrant
By default, each per-cpu part of a bound workqueue operates separately
and a work item may be executing concurrently on different CPUs. The
behavior avoids some cross-cpu traffic but leads to subtle weirdities
and not-so-subtle contortions in the API.
* There's no sane usefulness in allowing a single work item to be
executed concurrently on multiple CPUs. People just get the
behavior unintentionally and get surprised after learning about it.
Most either explicitly synchronize or use non-reentrant/ordered
workqueue but this is error-prone.
* flush_work() can't wait for multiple instances of the same work item
on different CPUs. If a work item is executing on cpu0 and then
queued on cpu1, flush_work() can only wait for the one on cpu1.
Unfortunately, work items can easily cross CPU boundaries
unintentionally when the queueing thread gets migrated. This means
that if multiple queuers compete, flush_work() can't even guarantee
that the instance queued right before it is finished before
returning.
* flush_work_sync() was added to work around some of the deficiencies
of flush_work(). In addition to the usual flushing, it ensures that
all currently executing instances are finished before returning.
This operation is expensive as it has to walk all CPUs and at the
same time fails to address competing queuer case.
Incorrectly using flush_work() when flush_work_sync() is necessary
is an easy error to make and can lead to bugs which are difficult to
reproduce.
* Similar problems exist for flush_delayed_work[_sync]().
Other than the cross-cpu access concern, there's no benefit in
allowing parallel execution and it's plain silly to have this level of
contortion for workqueue which is widely used from core code to
extremely obscure drivers.
This patch makes all workqueues non-reentrant. If a work item is
executing on a different CPU when queueing is requested, it is always
queued to that CPU. This guarantees that any given work item can be
executing on one CPU at maximum and if a work item is queued and
executing, both are on the same CPU.
The only behavior change which may affect workqueue users negatively
is that non-reentrancy overrides the affinity specified by
queue_work_on(). On a reentrant workqueue, the affinity specified by
queue_work_on() is always followed. Now, if the work item is
executing on one of the CPUs, the work item will be queued there
regardless of the requested affinity. I've reviewed all workqueue
users which request explicit affinity, and, fortunately, none seems to
be crazy enough to exploit parallel execution of the same work item.
This adds an additional busy_hash lookup if the work item was
previously queued on a different CPU. This shouldn't be noticeable
under any sane workload. Work item queueing isn't a very
high-frequency operation and they don't jump across CPUs all the time.
In a micro benchmark to exaggerate this difference - measuring the
time it takes for two work items to repeatedly jump between two CPUs a
number (10M) of times with busy_hash table densely populated, the
difference was around 3%.
While the overhead is measureable, it is only visible in pathological
cases and the difference isn't huge. This change brings much needed
sanity to workqueue and makes its behavior consistent with timer. I
think this is the right tradeoff to make.
This enables significant simplification of workqueue API.
Simplification patches will follow.
Signed-off-by: Tejun Heo <tj@kernel.org>
2012-08-20 15:51:23 -06:00
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/*
|
|
|
|
* If @work was previously on a different pool, it might still be
|
|
|
|
* running there, in which case the work needs to be queued on that
|
|
|
|
* pool to guarantee non-reentrancy.
|
|
|
|
*/
|
|
|
|
last_pool = get_work_pool(work);
|
|
|
|
if (last_pool && last_pool != pwq->pool) {
|
|
|
|
struct worker *worker;
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
spin_lock(&last_pool->lock);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
worker = find_worker_executing_work(last_pool, work);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
if (worker && worker->current_pwq->wq == wq) {
|
|
|
|
pwq = worker->current_pwq;
|
2012-08-03 11:30:45 -06:00
|
|
|
} else {
|
2013-03-12 12:30:04 -06:00
|
|
|
/* meh... not running there, queue here */
|
|
|
|
spin_unlock(&last_pool->lock);
|
2013-02-13 20:29:12 -07:00
|
|
|
spin_lock(&pwq->pool->lock);
|
2012-08-03 11:30:45 -06:00
|
|
|
}
|
2010-07-02 02:03:51 -06:00
|
|
|
} else {
|
2013-02-13 20:29:12 -07:00
|
|
|
spin_lock(&pwq->pool->lock);
|
2010-06-29 02:07:13 -06:00
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/*
|
|
|
|
* pwq is determined and locked. For unbound pools, we could have
|
|
|
|
* raced with pwq release and it could already be dead. If its
|
|
|
|
* refcnt is zero, repeat pwq selection. Note that pwqs never die
|
2013-04-01 12:23:35 -06:00
|
|
|
* without another pwq replacing it in the numa_pwq_tbl or while
|
|
|
|
* work items are executing on it, so the retrying is guaranteed to
|
2013-03-12 12:30:04 -06:00
|
|
|
* make forward-progress.
|
|
|
|
*/
|
|
|
|
if (unlikely(!pwq->refcnt)) {
|
|
|
|
if (wq->flags & WQ_UNBOUND) {
|
|
|
|
spin_unlock(&pwq->pool->lock);
|
|
|
|
cpu_relax();
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
/* oops */
|
|
|
|
WARN_ONCE(true, "workqueue: per-cpu pwq for %s on cpu%d has 0 refcnt",
|
|
|
|
wq->name, cpu);
|
|
|
|
}
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
/* pwq determined, queue */
|
|
|
|
trace_workqueue_queue_work(req_cpu, pwq, work);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2012-04-13 13:06:58 -06:00
|
|
|
if (WARN_ON(!list_empty(&work->entry))) {
|
2013-02-13 20:29:12 -07:00
|
|
|
spin_unlock(&pwq->pool->lock);
|
2012-04-13 13:06:58 -06:00
|
|
|
return;
|
|
|
|
}
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq->nr_in_flight[pwq->work_color]++;
|
|
|
|
work_flags = work_color_to_flags(pwq->work_color);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (likely(pwq->nr_active < pwq->max_active)) {
|
2010-10-05 02:49:55 -06:00
|
|
|
trace_workqueue_activate_work(work);
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq->nr_active++;
|
|
|
|
worklist = &pwq->pool->worklist;
|
2010-08-25 02:33:56 -06:00
|
|
|
} else {
|
|
|
|
work_flags |= WORK_STRUCT_DELAYED;
|
2013-02-13 20:29:12 -07:00
|
|
|
worklist = &pwq->delayed_works;
|
2010-08-25 02:33:56 -06:00
|
|
|
}
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
insert_work(pwq, work, worklist, work_flags);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
spin_unlock(&pwq->pool->lock);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2006-07-30 04:03:42 -06:00
|
|
|
/**
|
2008-07-23 22:28:39 -06:00
|
|
|
* queue_work_on - queue work on specific cpu
|
|
|
|
* @cpu: CPU number to execute work on
|
2006-07-30 04:03:42 -06:00
|
|
|
* @wq: workqueue to use
|
|
|
|
* @work: work to queue
|
|
|
|
*
|
2008-07-23 22:28:39 -06:00
|
|
|
* We queue the work to a specific CPU, the caller must ensure it
|
|
|
|
* can't go away.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: %false if @work was already on a queue, %true otherwise.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
2012-08-03 11:30:44 -06:00
|
|
|
bool queue_work_on(int cpu, struct workqueue_struct *wq,
|
|
|
|
struct work_struct *work)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2012-08-03 11:30:44 -06:00
|
|
|
bool ret = false;
|
2012-08-03 11:30:45 -06:00
|
|
|
unsigned long flags;
|
2008-07-25 02:47:53 -06:00
|
|
|
|
2012-08-03 11:30:45 -06:00
|
|
|
local_irq_save(flags);
|
2008-07-23 22:28:39 -06:00
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
if (!test_and_set_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(work))) {
|
2010-06-29 02:07:10 -06:00
|
|
|
__queue_work(cpu, wq, work);
|
2012-08-03 11:30:44 -06:00
|
|
|
ret = true;
|
2008-07-23 22:28:39 -06:00
|
|
|
}
|
2008-07-25 02:47:53 -06:00
|
|
|
|
2012-08-03 11:30:45 -06:00
|
|
|
local_irq_restore(flags);
|
2005-04-16 16:20:36 -06:00
|
|
|
return ret;
|
|
|
|
}
|
2013-05-06 15:44:55 -06:00
|
|
|
EXPORT_SYMBOL(queue_work_on);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2012-08-03 11:30:45 -06:00
|
|
|
void delayed_work_timer_fn(unsigned long __data)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2006-11-22 07:54:01 -07:00
|
|
|
struct delayed_work *dwork = (struct delayed_work *)__data;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2012-08-21 14:18:24 -06:00
|
|
|
/* should have been called from irqsafe timer with irq already off */
|
2013-02-06 19:04:53 -07:00
|
|
|
__queue_work(dwork->cpu, dwork->wq, &dwork->work);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2013-01-24 05:36:31 -07:00
|
|
|
EXPORT_SYMBOL(delayed_work_timer_fn);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
static void __queue_delayed_work(int cpu, struct workqueue_struct *wq,
|
|
|
|
struct delayed_work *dwork, unsigned long delay)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2012-08-03 11:30:46 -06:00
|
|
|
struct timer_list *timer = &dwork->timer;
|
|
|
|
struct work_struct *work = &dwork->work;
|
|
|
|
|
|
|
|
WARN_ON_ONCE(timer->function != delayed_work_timer_fn ||
|
|
|
|
timer->data != (unsigned long)dwork);
|
2012-12-04 08:40:39 -07:00
|
|
|
WARN_ON_ONCE(timer_pending(timer));
|
|
|
|
WARN_ON_ONCE(!list_empty(&work->entry));
|
2012-08-03 11:30:46 -06:00
|
|
|
|
2012-12-01 17:23:42 -07:00
|
|
|
/*
|
|
|
|
* If @delay is 0, queue @dwork->work immediately. This is for
|
|
|
|
* both optimization and correctness. The earliest @timer can
|
|
|
|
* expire is on the closest next tick and delayed_work users depend
|
|
|
|
* on that there's no such delay when @delay is 0.
|
|
|
|
*/
|
|
|
|
if (!delay) {
|
|
|
|
__queue_work(cpu, wq, &dwork->work);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
timer_stats_timer_set_start_info(&dwork->timer);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-02-06 19:04:53 -07:00
|
|
|
dwork->wq = wq;
|
2012-08-08 10:38:42 -06:00
|
|
|
dwork->cpu = cpu;
|
2012-08-03 11:30:46 -06:00
|
|
|
timer->expires = jiffies + delay;
|
|
|
|
|
|
|
|
if (unlikely(cpu != WORK_CPU_UNBOUND))
|
|
|
|
add_timer_on(timer, cpu);
|
|
|
|
else
|
|
|
|
add_timer(timer);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2006-07-30 04:03:42 -06:00
|
|
|
/**
|
|
|
|
* queue_delayed_work_on - queue work on specific CPU after delay
|
|
|
|
* @cpu: CPU number to execute work on
|
|
|
|
* @wq: workqueue to use
|
2006-12-22 02:06:52 -07:00
|
|
|
* @dwork: work to queue
|
2006-07-30 04:03:42 -06:00
|
|
|
* @delay: number of jiffies to wait before queueing
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return: %false if @work was already on a queue, %true otherwise. If
|
2012-08-03 11:30:46 -06:00
|
|
|
* @delay is zero and @dwork is idle, it will be scheduled for immediate
|
|
|
|
* execution.
|
2006-07-30 04:03:42 -06:00
|
|
|
*/
|
2012-08-03 11:30:44 -06:00
|
|
|
bool queue_delayed_work_on(int cpu, struct workqueue_struct *wq,
|
|
|
|
struct delayed_work *dwork, unsigned long delay)
|
2006-06-28 14:50:33 -06:00
|
|
|
{
|
2006-11-22 07:54:01 -07:00
|
|
|
struct work_struct *work = &dwork->work;
|
2012-08-03 11:30:44 -06:00
|
|
|
bool ret = false;
|
2012-08-03 11:30:45 -06:00
|
|
|
unsigned long flags;
|
2006-06-28 14:50:33 -06:00
|
|
|
|
2012-08-03 11:30:45 -06:00
|
|
|
/* read the comment in __queue_work() */
|
|
|
|
local_irq_save(flags);
|
2006-06-28 14:50:33 -06:00
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
if (!test_and_set_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(work))) {
|
2012-08-03 11:30:46 -06:00
|
|
|
__queue_delayed_work(cpu, wq, dwork, delay);
|
2012-08-03 11:30:44 -06:00
|
|
|
ret = true;
|
2006-06-28 14:50:33 -06:00
|
|
|
}
|
2008-05-01 05:35:14 -06:00
|
|
|
|
2012-08-03 11:30:45 -06:00
|
|
|
local_irq_restore(flags);
|
2006-06-28 14:50:33 -06:00
|
|
|
return ret;
|
|
|
|
}
|
2013-05-06 15:44:55 -06:00
|
|
|
EXPORT_SYMBOL(queue_delayed_work_on);
|
2010-07-02 02:03:51 -06:00
|
|
|
|
2012-08-03 11:30:47 -06:00
|
|
|
/**
|
|
|
|
* mod_delayed_work_on - modify delay of or queue a delayed work on specific CPU
|
|
|
|
* @cpu: CPU number to execute work on
|
|
|
|
* @wq: workqueue to use
|
|
|
|
* @dwork: work to queue
|
|
|
|
* @delay: number of jiffies to wait before queueing
|
|
|
|
*
|
|
|
|
* If @dwork is idle, equivalent to queue_delayed_work_on(); otherwise,
|
|
|
|
* modify @dwork's timer so that it expires after @delay. If @delay is
|
|
|
|
* zero, @work is guaranteed to be scheduled immediately regardless of its
|
|
|
|
* current state.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return: %false if @dwork was idle and queued, %true if @dwork was
|
2012-08-03 11:30:47 -06:00
|
|
|
* pending and its timer was modified.
|
|
|
|
*
|
2012-08-21 14:18:24 -06:00
|
|
|
* This function is safe to call from any context including IRQ handler.
|
2012-08-03 11:30:47 -06:00
|
|
|
* See try_to_grab_pending() for details.
|
|
|
|
*/
|
|
|
|
bool mod_delayed_work_on(int cpu, struct workqueue_struct *wq,
|
|
|
|
struct delayed_work *dwork, unsigned long delay)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
int ret;
|
2010-07-02 02:03:51 -06:00
|
|
|
|
2012-08-03 11:30:47 -06:00
|
|
|
do {
|
|
|
|
ret = try_to_grab_pending(&dwork->work, true, &flags);
|
|
|
|
} while (unlikely(ret == -EAGAIN));
|
2007-05-09 03:34:16 -06:00
|
|
|
|
2012-08-03 11:30:47 -06:00
|
|
|
if (likely(ret >= 0)) {
|
|
|
|
__queue_delayed_work(cpu, wq, dwork, delay);
|
|
|
|
local_irq_restore(flags);
|
2006-06-28 14:50:33 -06:00
|
|
|
}
|
2012-08-03 11:30:47 -06:00
|
|
|
|
|
|
|
/* -ENOENT from try_to_grab_pending() becomes %true */
|
2006-06-28 14:50:33 -06:00
|
|
|
return ret;
|
|
|
|
}
|
2012-08-03 11:30:47 -06:00
|
|
|
EXPORT_SYMBOL_GPL(mod_delayed_work_on);
|
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/**
|
|
|
|
* worker_enter_idle - enter idle state
|
|
|
|
* @worker: worker which is entering idle state
|
|
|
|
*
|
|
|
|
* @worker is entering idle state. Update stats and idle timer if
|
|
|
|
* necessary.
|
|
|
|
*
|
|
|
|
* LOCKING:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
|
|
|
static void worker_enter_idle(struct worker *worker)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
if (WARN_ON_ONCE(worker->flags & WORKER_IDLE) ||
|
|
|
|
WARN_ON_ONCE(!list_empty(&worker->entry) &&
|
|
|
|
(worker->hentry.next || worker->hentry.pprev)))
|
|
|
|
return;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-07-02 02:03:50 -06:00
|
|
|
/* can't use worker_set_flags(), also called from start_worker() */
|
|
|
|
worker->flags |= WORKER_IDLE;
|
2012-07-12 15:46:37 -06:00
|
|
|
pool->nr_idle++;
|
2010-06-29 02:07:14 -06:00
|
|
|
worker->last_active = jiffies;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
|
|
|
/* idle_list is LIFO */
|
2012-07-12 15:46:37 -06:00
|
|
|
list_add(&worker->entry, &pool->idle_list);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2012-07-17 13:39:27 -06:00
|
|
|
if (too_many_workers(pool) && !timer_pending(&pool->idle_timer))
|
|
|
|
mod_timer(&pool->idle_timer, jiffies + IDLE_WORKER_TIMEOUT);
|
2010-07-02 02:03:50 -06:00
|
|
|
|
2012-05-14 16:04:50 -06:00
|
|
|
/*
|
2013-01-24 12:01:34 -07:00
|
|
|
* Sanity check nr_running. Because wq_unbind_fn() releases
|
2013-01-24 12:01:33 -07:00
|
|
|
* pool->lock between setting %WORKER_UNBOUND and zapping
|
2012-07-17 13:39:27 -06:00
|
|
|
* nr_running, the warning may trigger spuriously. Check iff
|
|
|
|
* unbind is not in progress.
|
2012-05-14 16:04:50 -06:00
|
|
|
*/
|
2013-01-24 12:01:33 -07:00
|
|
|
WARN_ON_ONCE(!(pool->flags & POOL_DISASSOCIATED) &&
|
2012-07-12 15:46:37 -06:00
|
|
|
pool->nr_workers == pool->nr_idle &&
|
2013-01-24 12:39:44 -07:00
|
|
|
atomic_read(&pool->nr_running));
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* worker_leave_idle - leave idle state
|
|
|
|
* @worker: worker which is leaving idle state
|
|
|
|
*
|
|
|
|
* @worker is leaving idle state. Update stats.
|
|
|
|
*
|
|
|
|
* LOCKING:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
|
|
|
static void worker_leave_idle(struct worker *worker)
|
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
if (WARN_ON_ONCE(!(worker->flags & WORKER_IDLE)))
|
|
|
|
return;
|
2010-06-29 02:07:13 -06:00
|
|
|
worker_clr_flags(worker, WORKER_IDLE);
|
2012-07-12 15:46:37 -06:00
|
|
|
pool->nr_idle--;
|
2010-06-29 02:07:12 -06:00
|
|
|
list_del_init(&worker->entry);
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/**
|
2013-02-19 13:17:02 -07:00
|
|
|
* worker_maybe_bind_and_lock - try to bind %current to worker_pool and lock it
|
|
|
|
* @pool: target worker_pool
|
|
|
|
*
|
|
|
|
* Bind %current to the cpu of @pool if it is associated and lock @pool.
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
|
|
|
* Works which are scheduled while the cpu is online must at least be
|
|
|
|
* scheduled to a worker which is bound to the cpu so that if they are
|
|
|
|
* flushed from cpu callbacks while cpu is going down, they are
|
|
|
|
* guaranteed to execute on the cpu.
|
|
|
|
*
|
2013-02-19 13:17:02 -07:00
|
|
|
* This function is to be used by unbound workers and rescuers to bind
|
2010-06-29 02:07:14 -06:00
|
|
|
* themselves to the target cpu and may race with cpu going down or
|
|
|
|
* coming online. kthread_bind() can't be used because it may put the
|
|
|
|
* worker to already dead cpu and set_cpus_allowed_ptr() can't be used
|
2013-01-24 12:01:34 -07:00
|
|
|
* verbatim as it's best effort and blocking and pool may be
|
2010-06-29 02:07:14 -06:00
|
|
|
* [dis]associated in the meantime.
|
|
|
|
*
|
2013-01-24 12:01:34 -07:00
|
|
|
* This function tries set_cpus_allowed() and locks pool and verifies the
|
2013-01-24 12:01:33 -07:00
|
|
|
* binding against %POOL_DISASSOCIATED which is set during
|
2012-07-17 13:39:26 -06:00
|
|
|
* %CPU_DOWN_PREPARE and cleared during %CPU_ONLINE, so if the worker
|
|
|
|
* enters idle state or fetches works without dropping lock, it can
|
|
|
|
* guarantee the scheduling requirement described in the first paragraph.
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* Might sleep. Called without any lock but returns with pool->lock
|
2010-06-29 02:07:14 -06:00
|
|
|
* held.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2013-01-24 12:01:34 -07:00
|
|
|
* %true if the associated pool is online (@worker is successfully
|
2010-06-29 02:07:14 -06:00
|
|
|
* bound), %false if offline.
|
|
|
|
*/
|
2013-02-19 13:17:02 -07:00
|
|
|
static bool worker_maybe_bind_and_lock(struct worker_pool *pool)
|
2013-01-24 12:01:33 -07:00
|
|
|
__acquires(&pool->lock)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
|
|
|
while (true) {
|
2007-10-19 00:39:55 -06:00
|
|
|
/*
|
2010-06-29 02:07:14 -06:00
|
|
|
* The following call may fail, succeed or succeed
|
|
|
|
* without actually migrating the task to the cpu if
|
|
|
|
* it races with cpu hotunplug operation. Verify
|
2013-01-24 12:01:33 -07:00
|
|
|
* against POOL_DISASSOCIATED.
|
2007-10-19 00:39:55 -06:00
|
|
|
*/
|
2013-01-24 12:01:33 -07:00
|
|
|
if (!(pool->flags & POOL_DISASSOCIATED))
|
2013-03-12 12:30:00 -06:00
|
|
|
set_cpus_allowed_ptr(current, pool->attrs->cpumask);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2013-01-24 12:01:33 -07:00
|
|
|
if (pool->flags & POOL_DISASSOCIATED)
|
2010-06-29 02:07:14 -06:00
|
|
|
return false;
|
2013-02-19 13:17:02 -07:00
|
|
|
if (task_cpu(current) == pool->cpu &&
|
2013-03-12 12:30:00 -06:00
|
|
|
cpumask_equal(¤t->cpus_allowed, pool->attrs->cpumask))
|
2010-06-29 02:07:14 -06:00
|
|
|
return true;
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2011-04-29 10:08:37 -06:00
|
|
|
/*
|
|
|
|
* We've raced with CPU hot[un]plug. Give it a breather
|
|
|
|
* and retry migration. cond_resched() is required here;
|
|
|
|
* otherwise, we might deadlock against cpu_stop trying to
|
|
|
|
* bring down the CPU on non-preemptive kernel.
|
|
|
|
*/
|
2010-06-29 02:07:14 -06:00
|
|
|
cpu_relax();
|
2011-04-29 10:08:37 -06:00
|
|
|
cond_resched();
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
static struct worker *alloc_worker(void)
|
|
|
|
{
|
|
|
|
struct worker *worker;
|
|
|
|
|
|
|
|
worker = kzalloc(sizeof(*worker), GFP_KERNEL);
|
2010-06-29 02:07:12 -06:00
|
|
|
if (worker) {
|
|
|
|
INIT_LIST_HEAD(&worker->entry);
|
2010-06-29 02:07:12 -06:00
|
|
|
INIT_LIST_HEAD(&worker->scheduled);
|
2010-06-29 02:07:14 -06:00
|
|
|
/* on creation a worker is in !idle && prep state */
|
|
|
|
worker->flags = WORKER_PREP;
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
2010-06-29 02:07:11 -06:00
|
|
|
return worker;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* create_worker - create a new workqueue worker
|
2012-07-12 15:46:37 -06:00
|
|
|
* @pool: pool the new worker will belong to
|
2010-06-29 02:07:11 -06:00
|
|
|
*
|
2012-07-12 15:46:37 -06:00
|
|
|
* Create a new worker which is bound to @pool. The returned worker
|
2010-06-29 02:07:11 -06:00
|
|
|
* can be started by calling start_worker() or destroyed using
|
|
|
|
* destroy_worker().
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
|
|
|
* Might sleep. Does GFP_KERNEL allocations.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-06-29 02:07:11 -06:00
|
|
|
* Pointer to the newly created worker.
|
|
|
|
*/
|
2012-07-17 13:39:27 -06:00
|
|
|
static struct worker *create_worker(struct worker_pool *pool)
|
2010-06-29 02:07:11 -06:00
|
|
|
{
|
|
|
|
struct worker *worker = NULL;
|
2010-07-02 02:03:51 -06:00
|
|
|
int id = -1;
|
2013-04-01 12:23:32 -06:00
|
|
|
char id_buf[16];
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
lockdep_assert_held(&pool->manager_mutex);
|
|
|
|
|
2013-03-19 14:45:21 -06:00
|
|
|
/*
|
|
|
|
* ID is needed to determine kthread name. Allocate ID first
|
|
|
|
* without installing the pointer.
|
|
|
|
*/
|
|
|
|
idr_preload(GFP_KERNEL);
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2013-03-19 14:45:21 -06:00
|
|
|
|
|
|
|
id = idr_alloc(&pool->worker_idr, NULL, 0, 0, GFP_NOWAIT);
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2013-03-19 14:45:21 -06:00
|
|
|
idr_preload_end();
|
|
|
|
if (id < 0)
|
|
|
|
goto fail;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
worker = alloc_worker();
|
|
|
|
if (!worker)
|
|
|
|
goto fail;
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
worker->pool = pool;
|
2010-06-29 02:07:11 -06:00
|
|
|
worker->id = id;
|
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
if (pool->cpu >= 0)
|
2013-04-01 12:23:32 -06:00
|
|
|
snprintf(id_buf, sizeof(id_buf), "%d:%d%s", pool->cpu, id,
|
|
|
|
pool->attrs->nice < 0 ? "H" : "");
|
2010-07-02 02:03:51 -06:00
|
|
|
else
|
2013-04-01 12:23:32 -06:00
|
|
|
snprintf(id_buf, sizeof(id_buf), "u%d:%d", pool->id, id);
|
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
worker->task = kthread_create_on_node(worker_thread, worker, pool->node,
|
2013-04-01 12:23:32 -06:00
|
|
|
"kworker/%s", id_buf);
|
2010-06-29 02:07:11 -06:00
|
|
|
if (IS_ERR(worker->task))
|
|
|
|
goto fail;
|
|
|
|
|
2013-11-14 04:56:18 -07:00
|
|
|
set_user_nice(worker->task, pool->attrs->nice);
|
|
|
|
|
|
|
|
/* prevent userland from meddling with cpumask of workqueue workers */
|
|
|
|
worker->task->flags |= PF_NO_SETAFFINITY;
|
|
|
|
|
2013-03-13 17:51:36 -06:00
|
|
|
/*
|
|
|
|
* set_cpus_allowed_ptr() will fail if the cpumask doesn't have any
|
|
|
|
* online CPUs. It'll be re-applied when any of the CPUs come up.
|
|
|
|
*/
|
2013-03-12 12:30:00 -06:00
|
|
|
set_cpus_allowed_ptr(worker->task, pool->attrs->cpumask);
|
2012-07-13 23:16:45 -06:00
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
/*
|
|
|
|
* The caller is responsible for ensuring %POOL_DISASSOCIATED
|
|
|
|
* remains stable across this function. See the comments above the
|
|
|
|
* flag definition for details.
|
|
|
|
*/
|
|
|
|
if (pool->flags & POOL_DISASSOCIATED)
|
2012-07-17 13:39:27 -06:00
|
|
|
worker->flags |= WORKER_UNBOUND;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-19 14:45:21 -06:00
|
|
|
/* successful, commit the pointer to idr */
|
|
|
|
spin_lock_irq(&pool->lock);
|
|
|
|
idr_replace(&pool->worker_idr, worker, worker->id);
|
|
|
|
spin_unlock_irq(&pool->lock);
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
return worker;
|
2013-03-19 14:45:21 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
fail:
|
|
|
|
if (id >= 0) {
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2013-03-19 14:45:21 -06:00
|
|
|
idr_remove(&pool->worker_idr, id);
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-06-29 02:07:11 -06:00
|
|
|
}
|
|
|
|
kfree(worker);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* start_worker - start a newly created worker
|
|
|
|
* @worker: worker to start
|
|
|
|
*
|
2013-01-24 12:01:34 -07:00
|
|
|
* Make the pool aware of @worker and start it.
|
2010-06-29 02:07:11 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2010-06-29 02:07:11 -06:00
|
|
|
*/
|
|
|
|
static void start_worker(struct worker *worker)
|
|
|
|
{
|
2010-07-02 02:03:50 -06:00
|
|
|
worker->flags |= WORKER_STARTED;
|
2012-07-12 15:46:37 -06:00
|
|
|
worker->pool->nr_workers++;
|
2010-06-29 02:07:12 -06:00
|
|
|
worker_enter_idle(worker);
|
2010-06-29 02:07:11 -06:00
|
|
|
wake_up_process(worker->task);
|
|
|
|
}
|
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
/**
|
|
|
|
* create_and_start_worker - create and start a worker for a pool
|
|
|
|
* @pool: the target pool
|
|
|
|
*
|
2013-03-13 20:47:39 -06:00
|
|
|
* Grab the managership of @pool and create and start a new worker for it.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: 0 on success. A negative error code otherwise.
|
2013-03-13 20:47:39 -06:00
|
|
|
*/
|
|
|
|
static int create_and_start_worker(struct worker_pool *pool)
|
|
|
|
{
|
|
|
|
struct worker *worker;
|
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_lock(&pool->manager_mutex);
|
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
worker = create_worker(pool);
|
|
|
|
if (worker) {
|
|
|
|
spin_lock_irq(&pool->lock);
|
|
|
|
start_worker(worker);
|
|
|
|
spin_unlock_irq(&pool->lock);
|
|
|
|
}
|
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_unlock(&pool->manager_mutex);
|
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
return worker ? 0 : -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
/**
|
|
|
|
* destroy_worker - destroy a workqueue worker
|
|
|
|
* @worker: worker to be destroyed
|
|
|
|
*
|
2013-01-24 12:01:34 -07:00
|
|
|
* Destroy @worker and adjust @pool stats accordingly.
|
2010-06-29 02:07:12 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock) which is released and regrabbed.
|
2010-06-29 02:07:11 -06:00
|
|
|
*/
|
|
|
|
static void destroy_worker(struct worker *worker)
|
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
lockdep_assert_held(&pool->manager_mutex);
|
|
|
|
lockdep_assert_held(&pool->lock);
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
/* sanity check frenzy */
|
2013-03-12 12:29:57 -06:00
|
|
|
if (WARN_ON(worker->current_work) ||
|
|
|
|
WARN_ON(!list_empty(&worker->scheduled)))
|
|
|
|
return;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
if (worker->flags & WORKER_STARTED)
|
2012-07-12 15:46:37 -06:00
|
|
|
pool->nr_workers--;
|
2010-06-29 02:07:12 -06:00
|
|
|
if (worker->flags & WORKER_IDLE)
|
2012-07-12 15:46:37 -06:00
|
|
|
pool->nr_idle--;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2014-02-15 07:02:28 -07:00
|
|
|
/*
|
|
|
|
* Once WORKER_DIE is set, the kworker may destroy itself at any
|
|
|
|
* point. Pin to ensure the task stays until we're done with it.
|
|
|
|
*/
|
|
|
|
get_task_struct(worker->task);
|
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
list_del_init(&worker->entry);
|
2010-07-02 02:03:50 -06:00
|
|
|
worker->flags |= WORKER_DIE;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-19 14:45:21 -06:00
|
|
|
idr_remove(&pool->worker_idr, worker->id);
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
kthread_stop(worker->task);
|
2014-02-15 07:02:28 -07:00
|
|
|
put_task_struct(worker->task);
|
2010-06-29 02:07:11 -06:00
|
|
|
kfree(worker);
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2010-06-29 02:07:11 -06:00
|
|
|
}
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
static void idle_worker_timeout(unsigned long __pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = (void *)__pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
if (too_many_workers(pool)) {
|
2010-06-29 02:07:14 -06:00
|
|
|
struct worker *worker;
|
|
|
|
unsigned long expires;
|
|
|
|
|
|
|
|
/* idle_list is kept in LIFO order, check the last one */
|
2012-07-12 15:46:37 -06:00
|
|
|
worker = list_entry(pool->idle_list.prev, struct worker, entry);
|
2010-06-29 02:07:14 -06:00
|
|
|
expires = worker->last_active + IDLE_WORKER_TIMEOUT;
|
|
|
|
|
|
|
|
if (time_before(jiffies, expires))
|
2012-07-12 15:46:37 -06:00
|
|
|
mod_timer(&pool->idle_timer, expires);
|
2010-06-29 02:07:14 -06:00
|
|
|
else {
|
|
|
|
/* it's been idle for too long, wake up manager */
|
2012-07-12 15:46:37 -06:00
|
|
|
pool->flags |= POOL_MANAGE_WORKERS;
|
2012-07-12 15:46:37 -06:00
|
|
|
wake_up_worker(pool);
|
2006-12-06 21:37:26 -07:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
2006-12-06 21:37:26 -07:00
|
|
|
|
2013-03-12 12:29:59 -06:00
|
|
|
static void send_mayday(struct work_struct *work)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2013-02-13 20:29:12 -07:00
|
|
|
struct pool_workqueue *pwq = get_work_pwq(work);
|
|
|
|
struct workqueue_struct *wq = pwq->wq;
|
2013-03-12 12:29:59 -06:00
|
|
|
|
2013-03-13 20:47:40 -06:00
|
|
|
lockdep_assert_held(&wq_mayday_lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
if (!wq->rescuer)
|
2013-03-12 12:29:59 -06:00
|
|
|
return;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
/* mayday mayday mayday */
|
2013-03-12 12:29:59 -06:00
|
|
|
if (list_empty(&pwq->mayday_node)) {
|
2014-04-18 09:04:16 -06:00
|
|
|
/*
|
|
|
|
* If @pwq is for an unbound wq, its base ref may be put at
|
|
|
|
* any time due to an attribute change. Pin @pwq until the
|
|
|
|
* rescuer is done with it.
|
|
|
|
*/
|
|
|
|
get_pwq(pwq);
|
2013-03-12 12:29:59 -06:00
|
|
|
list_add_tail(&pwq->mayday_node, &wq->maydays);
|
2010-06-29 02:07:14 -06:00
|
|
|
wake_up_process(wq->rescuer->task);
|
2013-03-12 12:29:59 -06:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
2013-01-24 12:01:34 -07:00
|
|
|
static void pool_mayday_timeout(unsigned long __pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = (void *)__pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
struct work_struct *work;
|
|
|
|
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_lock_irq(&wq_mayday_lock); /* for wq->maydays */
|
2013-03-12 12:29:59 -06:00
|
|
|
spin_lock(&pool->lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
if (need_to_create_worker(pool)) {
|
2010-06-29 02:07:14 -06:00
|
|
|
/*
|
|
|
|
* We've been trying to create a new worker but
|
|
|
|
* haven't been successful. We might be hitting an
|
|
|
|
* allocation deadlock. Send distress signals to
|
|
|
|
* rescuers.
|
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
list_for_each_entry(work, &pool->worklist, entry)
|
2010-06-29 02:07:14 -06:00
|
|
|
send_mayday(work);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-12 12:29:59 -06:00
|
|
|
spin_unlock(&pool->lock);
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_unlock_irq(&wq_mayday_lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
mod_timer(&pool->mayday_timer, jiffies + MAYDAY_INTERVAL);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/**
|
|
|
|
* maybe_create_worker - create a new worker if necessary
|
2012-07-12 15:46:37 -06:00
|
|
|
* @pool: pool to create a new worker for
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
2012-07-12 15:46:37 -06:00
|
|
|
* Create a new worker for @pool if necessary. @pool is guaranteed to
|
2010-06-29 02:07:14 -06:00
|
|
|
* have at least one idle worker on return from this function. If
|
|
|
|
* creating a new worker takes longer than MAYDAY_INTERVAL, mayday is
|
2012-07-12 15:46:37 -06:00
|
|
|
* sent to all rescuers with works scheduled on @pool to resolve
|
2010-06-29 02:07:14 -06:00
|
|
|
* possible allocation deadlock.
|
|
|
|
*
|
2013-03-13 17:51:36 -06:00
|
|
|
* On return, need_to_create_worker() is guaranteed to be %false and
|
|
|
|
* may_start_working() %true.
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
|
|
|
* LOCKING:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock) which may be released and regrabbed
|
2010-06-29 02:07:14 -06:00
|
|
|
* multiple times. Does GFP_KERNEL allocations. Called only from
|
|
|
|
* manager.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2013-03-13 17:51:36 -06:00
|
|
|
* %false if no action was taken and pool->lock stayed locked, %true
|
2010-06-29 02:07:14 -06:00
|
|
|
* otherwise.
|
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool maybe_create_worker(struct worker_pool *pool)
|
2013-01-24 12:01:33 -07:00
|
|
|
__releases(&pool->lock)
|
|
|
|
__acquires(&pool->lock)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
if (!need_to_create_worker(pool))
|
2010-06-29 02:07:14 -06:00
|
|
|
return false;
|
|
|
|
restart:
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-07-14 03:31:20 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/* if we don't make progress in MAYDAY_INITIAL_TIMEOUT, call for help */
|
2012-07-12 15:46:37 -06:00
|
|
|
mod_timer(&pool->mayday_timer, jiffies + MAYDAY_INITIAL_TIMEOUT);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
while (true) {
|
|
|
|
struct worker *worker;
|
|
|
|
|
2012-07-17 13:39:27 -06:00
|
|
|
worker = create_worker(pool);
|
2010-06-29 02:07:14 -06:00
|
|
|
if (worker) {
|
2012-07-12 15:46:37 -06:00
|
|
|
del_timer_sync(&pool->mayday_timer);
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
start_worker(worker);
|
2013-03-12 12:29:57 -06:00
|
|
|
if (WARN_ON_ONCE(need_to_create_worker(pool)))
|
|
|
|
goto restart;
|
2010-06-29 02:07:14 -06:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
if (!need_to_create_worker(pool))
|
2010-06-29 02:07:14 -06:00
|
|
|
break;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
__set_current_state(TASK_INTERRUPTIBLE);
|
|
|
|
schedule_timeout(CREATE_COOLDOWN);
|
2010-07-14 03:31:20 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
if (!need_to_create_worker(pool))
|
2010-06-29 02:07:14 -06:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
del_timer_sync(&pool->mayday_timer);
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2012-07-12 15:46:37 -06:00
|
|
|
if (need_to_create_worker(pool))
|
2010-06-29 02:07:14 -06:00
|
|
|
goto restart;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* maybe_destroy_worker - destroy workers which have been idle for a while
|
2012-07-12 15:46:37 -06:00
|
|
|
* @pool: pool to destroy workers for
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
2012-07-12 15:46:37 -06:00
|
|
|
* Destroy @pool workers which have been idle for longer than
|
2010-06-29 02:07:14 -06:00
|
|
|
* IDLE_WORKER_TIMEOUT.
|
|
|
|
*
|
|
|
|
* LOCKING:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock) which may be released and regrabbed
|
2010-06-29 02:07:14 -06:00
|
|
|
* multiple times. Called only from manager.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2013-03-13 17:51:36 -06:00
|
|
|
* %false if no action was taken and pool->lock stayed locked, %true
|
2010-06-29 02:07:14 -06:00
|
|
|
* otherwise.
|
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
static bool maybe_destroy_workers(struct worker_pool *pool)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
|
|
|
bool ret = false;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
while (too_many_workers(pool)) {
|
2010-06-29 02:07:14 -06:00
|
|
|
struct worker *worker;
|
|
|
|
unsigned long expires;
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
worker = list_entry(pool->idle_list.prev, struct worker, entry);
|
2010-06-29 02:07:14 -06:00
|
|
|
expires = worker->last_active + IDLE_WORKER_TIMEOUT;
|
2007-05-09 03:34:20 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
if (time_before(jiffies, expires)) {
|
2012-07-12 15:46:37 -06:00
|
|
|
mod_timer(&pool->idle_timer, expires);
|
2007-05-09 03:34:09 -06:00
|
|
|
break;
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
destroy_worker(worker);
|
|
|
|
ret = true;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
return ret;
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
/**
|
2010-06-29 02:07:14 -06:00
|
|
|
* manage_workers - manage worker pool
|
|
|
|
* @worker: self
|
2010-06-29 02:07:11 -06:00
|
|
|
*
|
2013-01-24 12:01:34 -07:00
|
|
|
* Assume the manager role and manage the worker pool @worker belongs
|
2010-06-29 02:07:14 -06:00
|
|
|
* to. At any given time, there can be only zero or one manager per
|
2013-01-24 12:01:34 -07:00
|
|
|
* pool. The exclusion is handled automatically by this function.
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
|
|
|
* The caller can safely start processing works on false return. On
|
|
|
|
* true return, it's guaranteed that need_to_create_worker() is false
|
|
|
|
* and may_start_working() is true.
|
2010-06-29 02:07:11 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock) which may be released and regrabbed
|
2010-06-29 02:07:14 -06:00
|
|
|
* multiple times. Does GFP_KERNEL allocations.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2013-08-20 18:50:40 -06:00
|
|
|
* %false if the pool don't need management and the caller can safely start
|
|
|
|
* processing works, %true indicates that the function released pool->lock
|
|
|
|
* and reacquired it to perform some management function and that the
|
|
|
|
* conditions that the caller verified while holding the lock before
|
|
|
|
* calling the function might no longer be true.
|
2010-06-29 02:07:11 -06:00
|
|
|
*/
|
2010-06-29 02:07:14 -06:00
|
|
|
static bool manage_workers(struct worker *worker)
|
2010-06-29 02:07:11 -06:00
|
|
|
{
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
bool ret = false;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
/*
|
|
|
|
* Managership is governed by two mutexes - manager_arb and
|
|
|
|
* manager_mutex. manager_arb handles arbitration of manager role.
|
|
|
|
* Anyone who successfully grabs manager_arb wins the arbitration
|
|
|
|
* and becomes the manager. mutex_trylock() on pool->manager_arb
|
|
|
|
* failure while holding pool->lock reliably indicates that someone
|
|
|
|
* else is managing the pool and the worker which failed trylock
|
|
|
|
* can proceed to executing work items. This means that anyone
|
|
|
|
* grabbing manager_arb is responsible for actually performing
|
|
|
|
* manager duties. If manager_arb is grabbed and released without
|
|
|
|
* actual management, the pool may stall indefinitely.
|
|
|
|
*
|
|
|
|
* manager_mutex is used for exclusion of actual management
|
|
|
|
* operations. The holder of manager_mutex can be sure that none
|
|
|
|
* of management operations, including creation and destruction of
|
|
|
|
* workers, won't take place until the mutex is released. Because
|
|
|
|
* manager_mutex doesn't interfere with manager role arbitration,
|
|
|
|
* it is guaranteed that the pool's management, while may be
|
|
|
|
* delayed, won't be disturbed by someone else grabbing
|
|
|
|
* manager_mutex.
|
|
|
|
*/
|
2013-03-12 12:30:00 -06:00
|
|
|
if (!mutex_trylock(&pool->manager_arb))
|
2010-06-29 02:07:14 -06:00
|
|
|
return ret;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2012-09-10 11:03:44 -06:00
|
|
|
/*
|
2013-03-13 20:47:39 -06:00
|
|
|
* With manager arbitration won, manager_mutex would be free in
|
|
|
|
* most cases. trylock first without dropping @pool->lock.
|
2012-09-10 11:03:44 -06:00
|
|
|
*/
|
2013-03-13 20:47:39 -06:00
|
|
|
if (unlikely(!mutex_trylock(&pool->manager_mutex))) {
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_lock(&pool->manager_mutex);
|
2013-04-30 09:07:00 -06:00
|
|
|
spin_lock_irq(&pool->lock);
|
2012-09-10 11:03:44 -06:00
|
|
|
ret = true;
|
|
|
|
}
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
pool->flags &= ~POOL_MANAGE_WORKERS;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
/*
|
2010-06-29 02:07:14 -06:00
|
|
|
* Destroy and then create so that may_start_working() is true
|
|
|
|
* on return.
|
2010-06-29 02:07:11 -06:00
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
ret |= maybe_destroy_workers(pool);
|
|
|
|
ret |= maybe_create_worker(pool);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_unlock(&pool->manager_mutex);
|
2013-03-12 12:30:00 -06:00
|
|
|
mutex_unlock(&pool->manager_arb);
|
2010-06-29 02:07:14 -06:00
|
|
|
return ret;
|
2010-06-29 02:07:11 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
/**
|
|
|
|
* process_one_work - process single work
|
2010-06-29 02:07:11 -06:00
|
|
|
* @worker: self
|
2010-06-29 02:07:10 -06:00
|
|
|
* @work: work to process
|
|
|
|
*
|
|
|
|
* Process @work. This function contains all the logics necessary to
|
|
|
|
* process a single work including synchronization against and
|
|
|
|
* interaction with other workers on the same cpu, queueing and
|
|
|
|
* flushing. As long as context requirement is met, any worker can
|
|
|
|
* call this function to process a work.
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock) which is released and regrabbed.
|
2010-06-29 02:07:10 -06:00
|
|
|
*/
|
2010-06-29 02:07:11 -06:00
|
|
|
static void process_one_work(struct worker *worker, struct work_struct *work)
|
2013-01-24 12:01:33 -07:00
|
|
|
__releases(&pool->lock)
|
|
|
|
__acquires(&pool->lock)
|
2010-06-29 02:07:10 -06:00
|
|
|
{
|
2013-02-13 20:29:12 -07:00
|
|
|
struct pool_workqueue *pwq = get_work_pwq(work);
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2013-02-13 20:29:12 -07:00
|
|
|
bool cpu_intensive = pwq->wq->flags & WQ_CPU_INTENSIVE;
|
2010-06-29 02:07:11 -06:00
|
|
|
int work_color;
|
2010-06-29 02:07:13 -06:00
|
|
|
struct worker *collision;
|
2010-06-29 02:07:10 -06:00
|
|
|
#ifdef CONFIG_LOCKDEP
|
|
|
|
/*
|
|
|
|
* It is permissible to free the struct work_struct from
|
|
|
|
* inside the function that is called from it, this we need to
|
|
|
|
* take into account for lockdep too. To avoid bogus "held
|
|
|
|
* lock freed" warnings as well as problems when looking into
|
|
|
|
* work->lockdep_map, make a copy and use that here.
|
|
|
|
*/
|
lockdep: fix oops in processing workqueue
Under memory load, on x86_64, with lockdep enabled, the workqueue's
process_one_work() has been seen to oops in __lock_acquire(), barfing
on a 0xffffffff00000000 pointer in the lockdep_map's class_cache[].
Because it's permissible to free a work_struct from its callout function,
the map used is an onstack copy of the map given in the work_struct: and
that copy is made without any locking.
Surprisingly, gcc (4.5.1 in Hugh's case) uses "rep movsl" rather than
"rep movsq" for that structure copy: which might race with a workqueue
user's wait_on_work() doing lock_map_acquire() on the source of the
copy, putting a pointer into the class_cache[], but only in time for
the top half of that pointer to be copied to the destination map.
Boom when process_one_work() subsequently does lock_map_acquire()
on its onstack copy of the lockdep_map.
Fix this, and a similar instance in call_timer_fn(), with a
lockdep_copy_map() function which additionally NULLs the class_cache[].
Note: this oops was actually seen on 3.4-next, where flush_work() newly
does the racing lock_map_acquire(); but Tejun points out that 3.4 and
earlier are already vulnerable to the same through wait_on_work().
* Patch orginally from Peter. Hugh modified it a bit and wrote the
description.
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Reported-by: Hugh Dickins <hughd@google.com>
LKML-Reference: <alpine.LSU.2.00.1205070951170.1544@eggly.anvils>
Signed-off-by: Tejun Heo <tj@kernel.org>
2012-05-15 09:06:19 -06:00
|
|
|
struct lockdep_map lockdep_map;
|
|
|
|
|
|
|
|
lockdep_copy_map(&lockdep_map, &work->lockdep_map);
|
2010-06-29 02:07:10 -06:00
|
|
|
#endif
|
2012-07-22 11:16:34 -06:00
|
|
|
/*
|
|
|
|
* Ensure we're on the correct CPU. DISASSOCIATED test is
|
|
|
|
* necessary to avoid spurious warnings from rescuers servicing the
|
2013-01-24 12:01:33 -07:00
|
|
|
* unbound or a disassociated pool.
|
2012-07-22 11:16:34 -06:00
|
|
|
*/
|
2012-09-18 10:59:23 -06:00
|
|
|
WARN_ON_ONCE(!(worker->flags & WORKER_UNBOUND) &&
|
2013-01-24 12:01:33 -07:00
|
|
|
!(pool->flags & POOL_DISASSOCIATED) &&
|
2013-01-24 12:01:33 -07:00
|
|
|
raw_smp_processor_id() != pool->cpu);
|
2012-07-17 13:39:27 -06:00
|
|
|
|
2010-06-29 02:07:13 -06:00
|
|
|
/*
|
|
|
|
* A single work shouldn't be executed concurrently by
|
|
|
|
* multiple workers on a single cpu. Check whether anyone is
|
|
|
|
* already processing the work. If so, defer the work to the
|
|
|
|
* currently executing one.
|
|
|
|
*/
|
2013-01-24 12:01:33 -07:00
|
|
|
collision = find_worker_executing_work(pool, work);
|
2010-06-29 02:07:13 -06:00
|
|
|
if (unlikely(collision)) {
|
|
|
|
move_linked_works(work, &collision->scheduled, NULL);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-08-03 11:30:45 -06:00
|
|
|
/* claim and dequeue */
|
2010-06-29 02:07:10 -06:00
|
|
|
debug_work_deactivate(work);
|
2013-01-24 12:01:33 -07:00
|
|
|
hash_add(pool->busy_hash, &worker->hentry, (unsigned long)work);
|
2010-06-29 02:07:11 -06:00
|
|
|
worker->current_work = work;
|
2012-12-18 11:35:02 -07:00
|
|
|
worker->current_func = work->func;
|
2013-02-13 20:29:12 -07:00
|
|
|
worker->current_pwq = pwq;
|
2010-06-29 02:07:11 -06:00
|
|
|
work_color = get_work_color(work);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
list_del_init(&work->entry);
|
|
|
|
|
2010-06-29 02:07:15 -06:00
|
|
|
/*
|
|
|
|
* CPU intensive works don't participate in concurrency
|
|
|
|
* management. They're the scheduler's responsibility.
|
|
|
|
*/
|
|
|
|
if (unlikely(cpu_intensive))
|
|
|
|
worker_set_flags(worker, WORKER_CPU_INTENSIVE, true);
|
|
|
|
|
2012-07-12 15:46:37 -06:00
|
|
|
/*
|
2013-01-24 12:01:33 -07:00
|
|
|
* Unbound pool isn't concurrency managed and work items should be
|
2012-07-12 15:46:37 -06:00
|
|
|
* executed ASAP. Wake up another worker if necessary.
|
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
if ((worker->flags & WORKER_UNBOUND) && need_more_worker(pool))
|
|
|
|
wake_up_worker(pool);
|
2012-07-12 15:46:37 -06:00
|
|
|
|
2012-08-03 11:30:45 -06:00
|
|
|
/*
|
2013-01-24 12:01:33 -07:00
|
|
|
* Record the last pool and clear PENDING which should be the last
|
2013-01-24 12:01:33 -07:00
|
|
|
* update to @work. Also, do this inside @pool->lock so that
|
2012-08-13 18:08:19 -06:00
|
|
|
* PENDING and queued state changes happen together while IRQ is
|
|
|
|
* disabled.
|
2012-08-03 11:30:45 -06:00
|
|
|
*/
|
2013-01-24 12:01:33 -07:00
|
|
|
set_work_pool_and_clear_pending(work, pool->id);
|
2010-06-29 02:07:10 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-06-29 02:07:10 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
lock_map_acquire_read(&pwq->wq->lockdep_map);
|
2010-06-29 02:07:10 -06:00
|
|
|
lock_map_acquire(&lockdep_map);
|
2010-08-21 14:07:26 -06:00
|
|
|
trace_workqueue_execute_start(work);
|
2012-12-18 11:35:02 -07:00
|
|
|
worker->current_func(work);
|
2010-08-21 14:07:26 -06:00
|
|
|
/*
|
|
|
|
* While we must be careful to not use "work" after this, the trace
|
|
|
|
* point will only record its address.
|
|
|
|
*/
|
|
|
|
trace_workqueue_execute_end(work);
|
2010-06-29 02:07:10 -06:00
|
|
|
lock_map_release(&lockdep_map);
|
2013-02-13 20:29:12 -07:00
|
|
|
lock_map_release(&pwq->wq->lockdep_map);
|
2010-06-29 02:07:10 -06:00
|
|
|
|
|
|
|
if (unlikely(in_atomic() || lockdep_depth(current) > 0)) {
|
2012-08-18 15:52:42 -06:00
|
|
|
pr_err("BUG: workqueue leaked lock or atomic: %s/0x%08x/%d\n"
|
|
|
|
" last function: %pf\n",
|
2012-12-18 11:35:02 -07:00
|
|
|
current->comm, preempt_count(), task_pid_nr(current),
|
|
|
|
worker->current_func);
|
2010-06-29 02:07:10 -06:00
|
|
|
debug_show_held_locks(current);
|
|
|
|
dump_stack();
|
|
|
|
}
|
|
|
|
|
2013-08-28 15:33:37 -06:00
|
|
|
/*
|
|
|
|
* The following prevents a kworker from hogging CPU on !PREEMPT
|
|
|
|
* kernels, where a requeueing work item waiting for something to
|
|
|
|
* happen could deadlock with stop_machine as such work item could
|
|
|
|
* indefinitely requeue itself while all other CPUs are trapped in
|
|
|
|
* stop_machine.
|
|
|
|
*/
|
|
|
|
cond_resched();
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2010-06-29 02:07:10 -06:00
|
|
|
|
2010-06-29 02:07:15 -06:00
|
|
|
/* clear cpu intensive status */
|
|
|
|
if (unlikely(cpu_intensive))
|
|
|
|
worker_clr_flags(worker, WORKER_CPU_INTENSIVE);
|
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
/* we're done with it, release */
|
2012-12-17 08:01:23 -07:00
|
|
|
hash_del(&worker->hentry);
|
2010-06-29 02:07:11 -06:00
|
|
|
worker->current_work = NULL;
|
2012-12-18 11:35:02 -07:00
|
|
|
worker->current_func = NULL;
|
2013-02-13 20:29:12 -07:00
|
|
|
worker->current_pwq = NULL;
|
2013-04-30 16:27:22 -06:00
|
|
|
worker->desc_valid = false;
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq_dec_nr_in_flight(pwq, work_color);
|
2010-06-29 02:07:10 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/**
|
|
|
|
* process_scheduled_works - process scheduled works
|
|
|
|
* @worker: self
|
|
|
|
*
|
|
|
|
* Process all scheduled works. Please note that the scheduled list
|
|
|
|
* may change while processing a work, so this function repeatedly
|
|
|
|
* fetches a work from the top and executes it.
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock) which may be released and regrabbed
|
2010-06-29 02:07:12 -06:00
|
|
|
* multiple times.
|
|
|
|
*/
|
|
|
|
static void process_scheduled_works(struct worker *worker)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2010-06-29 02:07:12 -06:00
|
|
|
while (!list_empty(&worker->scheduled)) {
|
|
|
|
struct work_struct *work = list_first_entry(&worker->scheduled,
|
2005-04-16 16:20:36 -06:00
|
|
|
struct work_struct, entry);
|
2010-06-29 02:07:11 -06:00
|
|
|
process_one_work(worker, work);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
/**
|
|
|
|
* worker_thread - the worker thread function
|
2010-06-29 02:07:11 -06:00
|
|
|
* @__worker: self
|
2010-06-29 02:07:10 -06:00
|
|
|
*
|
2013-03-13 17:51:36 -06:00
|
|
|
* The worker thread function. All workers belong to a worker_pool -
|
|
|
|
* either a per-cpu one or dynamic unbound one. These workers process all
|
|
|
|
* work items regardless of their specific target workqueue. The only
|
|
|
|
* exception is work items which belong to workqueues with a rescuer which
|
|
|
|
* will be explained in rescuer_thread().
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: 0
|
2010-06-29 02:07:10 -06:00
|
|
|
*/
|
2010-06-29 02:07:11 -06:00
|
|
|
static int worker_thread(void *__worker)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2010-06-29 02:07:11 -06:00
|
|
|
struct worker *worker = __worker;
|
2012-07-12 15:46:37 -06:00
|
|
|
struct worker_pool *pool = worker->pool;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/* tell the scheduler that this is a workqueue worker */
|
|
|
|
worker->task->flags |= PF_WQ_WORKER;
|
2010-06-29 02:07:12 -06:00
|
|
|
woke_up:
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
/* am I supposed to die? */
|
|
|
|
if (unlikely(worker->flags & WORKER_DIE)) {
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
WARN_ON_ONCE(!list_empty(&worker->entry));
|
|
|
|
worker->task->flags &= ~PF_WQ_WORKER;
|
|
|
|
return 0;
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
worker_leave_idle(worker);
|
2010-06-29 02:07:12 -06:00
|
|
|
recheck:
|
2010-06-29 02:07:14 -06:00
|
|
|
/* no more worker necessary? */
|
2012-07-12 15:46:37 -06:00
|
|
|
if (!need_more_worker(pool))
|
2010-06-29 02:07:14 -06:00
|
|
|
goto sleep;
|
|
|
|
|
|
|
|
/* do we need to manage? */
|
2012-07-12 15:46:37 -06:00
|
|
|
if (unlikely(!may_start_working(pool)) && manage_workers(worker))
|
2010-06-29 02:07:14 -06:00
|
|
|
goto recheck;
|
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/*
|
|
|
|
* ->scheduled list can only be filled while a worker is
|
|
|
|
* preparing to process a work or actually processing it.
|
|
|
|
* Make sure nobody diddled with it while I was sleeping.
|
|
|
|
*/
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(!list_empty(&worker->scheduled));
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/*
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
* Finish PREP stage. We're guaranteed to have at least one idle
|
|
|
|
* worker or that someone else has already assumed the manager
|
|
|
|
* role. This is where @worker starts participating in concurrency
|
|
|
|
* management if applicable and concurrency management is restored
|
|
|
|
* after being rebound. See rebind_workers() for details.
|
2010-06-29 02:07:14 -06:00
|
|
|
*/
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
worker_clr_flags(worker, WORKER_PREP | WORKER_REBOUND);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
do {
|
2010-06-29 02:07:12 -06:00
|
|
|
struct work_struct *work =
|
2012-07-12 15:46:37 -06:00
|
|
|
list_first_entry(&pool->worklist,
|
2010-06-29 02:07:12 -06:00
|
|
|
struct work_struct, entry);
|
|
|
|
|
|
|
|
if (likely(!(*work_data_bits(work) & WORK_STRUCT_LINKED))) {
|
|
|
|
/* optimization path, not strictly necessary */
|
|
|
|
process_one_work(worker, work);
|
|
|
|
if (unlikely(!list_empty(&worker->scheduled)))
|
2010-06-29 02:07:12 -06:00
|
|
|
process_scheduled_works(worker);
|
2010-06-29 02:07:12 -06:00
|
|
|
} else {
|
|
|
|
move_linked_works(work, &worker->scheduled, NULL);
|
|
|
|
process_scheduled_works(worker);
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
2012-07-12 15:46:37 -06:00
|
|
|
} while (keep_working(pool));
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
worker_set_flags(worker, WORKER_PREP, false);
|
2010-07-02 02:03:51 -06:00
|
|
|
sleep:
|
2012-07-12 15:46:37 -06:00
|
|
|
if (unlikely(need_to_manage_workers(pool)) && manage_workers(worker))
|
2010-06-29 02:07:14 -06:00
|
|
|
goto recheck;
|
2010-07-02 02:03:51 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/*
|
2013-01-24 12:01:33 -07:00
|
|
|
* pool->lock is held and there's no work to process and no need to
|
|
|
|
* manage, sleep. Workers are woken up only while holding
|
|
|
|
* pool->lock or from local cpu, so setting the current state
|
|
|
|
* before releasing pool->lock is enough to prevent losing any
|
|
|
|
* event.
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
|
|
|
worker_enter_idle(worker);
|
|
|
|
__set_current_state(TASK_INTERRUPTIBLE);
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-06-29 02:07:12 -06:00
|
|
|
schedule();
|
|
|
|
goto woke_up;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/**
|
|
|
|
* rescuer_thread - the rescuer thread function
|
2013-01-17 18:16:24 -07:00
|
|
|
* @__rescuer: self
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
|
|
|
* Workqueue rescuer thread function. There's one rescuer for each
|
2013-03-12 12:30:03 -06:00
|
|
|
* workqueue which has WQ_MEM_RECLAIM set.
|
2010-06-29 02:07:14 -06:00
|
|
|
*
|
2013-01-24 12:01:34 -07:00
|
|
|
* Regular work processing on a pool may block trying to create a new
|
2010-06-29 02:07:14 -06:00
|
|
|
* worker which uses GFP_KERNEL allocation which has slight chance of
|
|
|
|
* developing into deadlock if some works currently on the same queue
|
|
|
|
* need to be processed to satisfy the GFP_KERNEL allocation. This is
|
|
|
|
* the problem rescuer solves.
|
|
|
|
*
|
2013-01-24 12:01:34 -07:00
|
|
|
* When such condition is possible, the pool summons rescuers of all
|
|
|
|
* workqueues which have works queued on the pool and let them process
|
2010-06-29 02:07:14 -06:00
|
|
|
* those works so that forward progress can be guaranteed.
|
|
|
|
*
|
|
|
|
* This should happen rarely.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: 0
|
2010-06-29 02:07:14 -06:00
|
|
|
*/
|
2013-01-17 18:16:24 -07:00
|
|
|
static int rescuer_thread(void *__rescuer)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2013-01-17 18:16:24 -07:00
|
|
|
struct worker *rescuer = __rescuer;
|
|
|
|
struct workqueue_struct *wq = rescuer->rescue_wq;
|
2010-06-29 02:07:14 -06:00
|
|
|
struct list_head *scheduled = &rescuer->scheduled;
|
2014-04-18 09:04:16 -06:00
|
|
|
bool should_stop;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
set_user_nice(current, RESCUER_NICE_LEVEL);
|
2013-01-17 18:16:24 -07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Mark rescuer as worker too. As WORKER_PREP is never cleared, it
|
|
|
|
* doesn't participate in concurrency management.
|
|
|
|
*/
|
|
|
|
rescuer->task->flags |= PF_WQ_WORKER;
|
2010-06-29 02:07:14 -06:00
|
|
|
repeat:
|
|
|
|
set_current_state(TASK_INTERRUPTIBLE);
|
|
|
|
|
2014-04-18 09:04:16 -06:00
|
|
|
/*
|
|
|
|
* By the time the rescuer is requested to stop, the workqueue
|
|
|
|
* shouldn't have any work pending, but @wq->maydays may still have
|
|
|
|
* pwq(s) queued. This can happen by non-rescuer workers consuming
|
|
|
|
* all the work items before the rescuer got to them. Go through
|
|
|
|
* @wq->maydays processing before acting on should_stop so that the
|
|
|
|
* list is always empty on exit.
|
|
|
|
*/
|
|
|
|
should_stop = kthread_should_stop();
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-12 12:29:59 -06:00
|
|
|
/* see whether any pwq is asking for help */
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_lock_irq(&wq_mayday_lock);
|
2013-03-12 12:29:59 -06:00
|
|
|
|
|
|
|
while (!list_empty(&wq->maydays)) {
|
|
|
|
struct pool_workqueue *pwq = list_first_entry(&wq->maydays,
|
|
|
|
struct pool_workqueue, mayday_node);
|
2013-02-13 20:29:12 -07:00
|
|
|
struct worker_pool *pool = pwq->pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
struct work_struct *work, *n;
|
|
|
|
|
|
|
|
__set_current_state(TASK_RUNNING);
|
2013-03-12 12:29:59 -06:00
|
|
|
list_del_init(&pwq->mayday_node);
|
|
|
|
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_unlock_irq(&wq_mayday_lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
/* migrate to the target cpu if possible */
|
2013-02-19 13:17:02 -07:00
|
|
|
worker_maybe_bind_and_lock(pool);
|
2013-02-19 13:17:02 -07:00
|
|
|
rescuer->pool = pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Slurp in all works issued via this workqueue and
|
|
|
|
* process'em.
|
|
|
|
*/
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(!list_empty(&rescuer->scheduled));
|
2012-07-12 15:46:37 -06:00
|
|
|
list_for_each_entry_safe(work, n, &pool->worklist, entry)
|
2013-02-13 20:29:12 -07:00
|
|
|
if (get_work_pwq(work) == pwq)
|
2010-06-29 02:07:14 -06:00
|
|
|
move_linked_works(work, scheduled, &n);
|
|
|
|
|
|
|
|
process_scheduled_works(rescuer);
|
2011-02-14 06:04:46 -07:00
|
|
|
|
2014-04-18 09:04:16 -06:00
|
|
|
/*
|
|
|
|
* Put the reference grabbed by send_mayday(). @pool won't
|
|
|
|
* go away while we're holding its lock.
|
|
|
|
*/
|
|
|
|
put_pwq(pwq);
|
|
|
|
|
2011-02-14 06:04:46 -07:00
|
|
|
/*
|
2013-01-24 12:01:33 -07:00
|
|
|
* Leave this pool. If keep_working() is %true, notify a
|
2011-02-14 06:04:46 -07:00
|
|
|
* regular worker; otherwise, we end up with 0 concurrency
|
|
|
|
* and stalling the execution.
|
|
|
|
*/
|
2012-07-12 15:46:37 -06:00
|
|
|
if (keep_working(pool))
|
|
|
|
wake_up_worker(pool);
|
2011-02-14 06:04:46 -07:00
|
|
|
|
2013-02-19 13:17:02 -07:00
|
|
|
rescuer->pool = NULL;
|
2013-03-12 12:29:59 -06:00
|
|
|
spin_unlock(&pool->lock);
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_lock(&wq_mayday_lock);
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_unlock_irq(&wq_mayday_lock);
|
2013-03-12 12:29:59 -06:00
|
|
|
|
2014-04-18 09:04:16 -06:00
|
|
|
if (should_stop) {
|
|
|
|
__set_current_state(TASK_RUNNING);
|
|
|
|
rescuer->task->flags &= ~PF_WQ_WORKER;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-01-17 18:16:24 -07:00
|
|
|
/* rescuers should never participate in concurrency management */
|
|
|
|
WARN_ON_ONCE(!(rescuer->flags & WORKER_NOT_RUNNING));
|
2010-06-29 02:07:14 -06:00
|
|
|
schedule();
|
|
|
|
goto repeat;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2007-05-09 03:33:51 -06:00
|
|
|
struct wq_barrier {
|
|
|
|
struct work_struct work;
|
|
|
|
struct completion done;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void wq_barrier_func(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct wq_barrier *barr = container_of(work, struct wq_barrier, work);
|
|
|
|
complete(&barr->done);
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:10 -06:00
|
|
|
/**
|
|
|
|
* insert_wq_barrier - insert a barrier work
|
2013-02-13 20:29:12 -07:00
|
|
|
* @pwq: pwq to insert barrier into
|
2010-06-29 02:07:10 -06:00
|
|
|
* @barr: wq_barrier to insert
|
2010-06-29 02:07:12 -06:00
|
|
|
* @target: target work to attach @barr to
|
|
|
|
* @worker: worker currently executing @target, NULL if @target is not executing
|
2010-06-29 02:07:10 -06:00
|
|
|
*
|
2010-06-29 02:07:12 -06:00
|
|
|
* @barr is linked to @target such that @barr is completed only after
|
|
|
|
* @target finishes execution. Please note that the ordering
|
|
|
|
* guarantee is observed only with respect to @target and on the local
|
|
|
|
* cpu.
|
|
|
|
*
|
|
|
|
* Currently, a queued barrier can't be canceled. This is because
|
|
|
|
* try_to_grab_pending() can't determine whether the work to be
|
|
|
|
* grabbed is at the head of the queue and thus can't clear LINKED
|
|
|
|
* flag of the previous work while there must be a valid next work
|
|
|
|
* after a work with LINKED flag set.
|
|
|
|
*
|
|
|
|
* Note that when @worker is non-NULL, @target may be modified
|
2013-02-13 20:29:12 -07:00
|
|
|
* underneath us, so we can't reliably determine pwq from @target.
|
2010-06-29 02:07:10 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-01-24 12:01:33 -07:00
|
|
|
* spin_lock_irq(pool->lock).
|
2010-06-29 02:07:10 -06:00
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
static void insert_wq_barrier(struct pool_workqueue *pwq,
|
2010-06-29 02:07:12 -06:00
|
|
|
struct wq_barrier *barr,
|
|
|
|
struct work_struct *target, struct worker *worker)
|
2007-05-09 03:33:51 -06:00
|
|
|
{
|
2010-06-29 02:07:12 -06:00
|
|
|
struct list_head *head;
|
|
|
|
unsigned int linked = 0;
|
|
|
|
|
2009-11-15 09:09:48 -07:00
|
|
|
/*
|
2013-01-24 12:01:33 -07:00
|
|
|
* debugobject calls are safe here even with pool->lock locked
|
2009-11-15 09:09:48 -07:00
|
|
|
* as we know for sure that this will not trigger any of the
|
|
|
|
* checks and call back into the fixup functions where we
|
|
|
|
* might deadlock.
|
|
|
|
*/
|
2010-10-26 15:22:34 -06:00
|
|
|
INIT_WORK_ONSTACK(&barr->work, wq_barrier_func);
|
2010-06-29 02:07:10 -06:00
|
|
|
__set_bit(WORK_STRUCT_PENDING_BIT, work_data_bits(&barr->work));
|
2007-05-09 03:33:51 -06:00
|
|
|
init_completion(&barr->done);
|
2007-05-09 03:33:54 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/*
|
|
|
|
* If @target is currently being executed, schedule the
|
|
|
|
* barrier to the worker; otherwise, put it after @target.
|
|
|
|
*/
|
|
|
|
if (worker)
|
|
|
|
head = worker->scheduled.next;
|
|
|
|
else {
|
|
|
|
unsigned long *bits = work_data_bits(target);
|
|
|
|
|
|
|
|
head = target->entry.next;
|
|
|
|
/* there can already be other linked works, inherit and set */
|
|
|
|
linked = *bits & WORK_STRUCT_LINKED;
|
|
|
|
__set_bit(WORK_STRUCT_LINKED_BIT, bits);
|
|
|
|
}
|
|
|
|
|
2009-11-15 09:09:48 -07:00
|
|
|
debug_work_activate(&barr->work);
|
2013-02-13 20:29:12 -07:00
|
|
|
insert_work(pwq, &barr->work, head,
|
2010-06-29 02:07:12 -06:00
|
|
|
work_color_to_flags(WORK_NO_COLOR) | linked);
|
2007-05-09 03:33:51 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
/**
|
2013-02-13 20:29:12 -07:00
|
|
|
* flush_workqueue_prep_pwqs - prepare pwqs for workqueue flushing
|
2010-06-29 02:07:11 -06:00
|
|
|
* @wq: workqueue being flushed
|
|
|
|
* @flush_color: new flush color, < 0 for no-op
|
|
|
|
* @work_color: new work color, < 0 for no-op
|
|
|
|
*
|
2013-02-13 20:29:12 -07:00
|
|
|
* Prepare pwqs for workqueue flushing.
|
2010-06-29 02:07:11 -06:00
|
|
|
*
|
2013-02-13 20:29:12 -07:00
|
|
|
* If @flush_color is non-negative, flush_color on all pwqs should be
|
|
|
|
* -1. If no pwq has in-flight commands at the specified color, all
|
|
|
|
* pwq->flush_color's stay at -1 and %false is returned. If any pwq
|
|
|
|
* has in flight commands, its pwq->flush_color is set to
|
|
|
|
* @flush_color, @wq->nr_pwqs_to_flush is updated accordingly, pwq
|
2010-06-29 02:07:11 -06:00
|
|
|
* wakeup logic is armed and %true is returned.
|
|
|
|
*
|
|
|
|
* The caller should have initialized @wq->first_flusher prior to
|
|
|
|
* calling this function with non-negative @flush_color. If
|
|
|
|
* @flush_color is negative, no flush color update is done and %false
|
|
|
|
* is returned.
|
|
|
|
*
|
2013-02-13 20:29:12 -07:00
|
|
|
* If @work_color is non-negative, all pwqs should have the same
|
2010-06-29 02:07:11 -06:00
|
|
|
* work_color which is previous to @work_color and all will be
|
|
|
|
* advanced to @work_color.
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-03-25 17:57:17 -06:00
|
|
|
* mutex_lock(wq->mutex).
|
2010-06-29 02:07:11 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-06-29 02:07:11 -06:00
|
|
|
* %true if @flush_color >= 0 and there's something to flush. %false
|
|
|
|
* otherwise.
|
|
|
|
*/
|
2013-02-13 20:29:12 -07:00
|
|
|
static bool flush_workqueue_prep_pwqs(struct workqueue_struct *wq,
|
2010-06-29 02:07:11 -06:00
|
|
|
int flush_color, int work_color)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2010-06-29 02:07:11 -06:00
|
|
|
bool wait = false;
|
2013-03-12 12:29:58 -06:00
|
|
|
struct pool_workqueue *pwq;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
if (flush_color >= 0) {
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(atomic_read(&wq->nr_pwqs_to_flush));
|
2013-02-13 20:29:12 -07:00
|
|
|
atomic_set(&wq->nr_pwqs_to_flush, 1);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2009-04-02 17:58:24 -06:00
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
for_each_pwq(pwq, wq) {
|
2013-02-13 20:29:12 -07:00
|
|
|
struct worker_pool *pool = pwq->pool;
|
2007-05-09 03:33:51 -06:00
|
|
|
|
2013-03-25 17:57:18 -06:00
|
|
|
spin_lock_irq(&pool->lock);
|
2007-05-09 03:33:54 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
if (flush_color >= 0) {
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(pwq->flush_color != -1);
|
2007-05-09 03:33:51 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (pwq->nr_in_flight[flush_color]) {
|
|
|
|
pwq->flush_color = flush_color;
|
|
|
|
atomic_inc(&wq->nr_pwqs_to_flush);
|
2010-06-29 02:07:11 -06:00
|
|
|
wait = true;
|
|
|
|
}
|
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
if (work_color >= 0) {
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(work_color != work_next_color(pwq->work_color));
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq->work_color = work_color;
|
2010-06-29 02:07:11 -06:00
|
|
|
}
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-03-25 17:57:18 -06:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2009-04-02 17:58:24 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (flush_color >= 0 && atomic_dec_and_test(&wq->nr_pwqs_to_flush))
|
2010-06-29 02:07:11 -06:00
|
|
|
complete(&wq->first_flusher->done);
|
2007-05-23 14:57:57 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
return wait;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
|
|
|
|
2006-07-30 04:03:42 -06:00
|
|
|
/**
|
2005-04-16 16:20:36 -06:00
|
|
|
* flush_workqueue - ensure that any scheduled work has run to completion.
|
2006-07-30 04:03:42 -06:00
|
|
|
* @wq: workqueue to flush
|
2005-04-16 16:20:36 -06:00
|
|
|
*
|
2013-03-13 17:51:36 -06:00
|
|
|
* This function sleeps until all work items which were queued on entry
|
|
|
|
* have finished execution, but it is not livelocked by new incoming ones.
|
2005-04-16 16:20:36 -06:00
|
|
|
*/
|
2008-02-08 05:19:53 -07:00
|
|
|
void flush_workqueue(struct workqueue_struct *wq)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2010-06-29 02:07:11 -06:00
|
|
|
struct wq_flusher this_flusher = {
|
|
|
|
.list = LIST_HEAD_INIT(this_flusher.list),
|
|
|
|
.flush_color = -1,
|
|
|
|
.done = COMPLETION_INITIALIZER_ONSTACK(this_flusher.done),
|
|
|
|
};
|
|
|
|
int next_color;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2008-08-11 02:30:30 -06:00
|
|
|
lock_map_acquire(&wq->lockdep_map);
|
|
|
|
lock_map_release(&wq->lockdep_map);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Start-to-wait phase
|
|
|
|
*/
|
|
|
|
next_color = work_next_color(wq->work_color);
|
|
|
|
|
|
|
|
if (next_color != wq->flush_color) {
|
|
|
|
/*
|
|
|
|
* Color space is not full. The current work_color
|
|
|
|
* becomes our flush_color and work_color is advanced
|
|
|
|
* by one.
|
|
|
|
*/
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(!list_empty(&wq->flusher_overflow));
|
2010-06-29 02:07:11 -06:00
|
|
|
this_flusher.flush_color = wq->work_color;
|
|
|
|
wq->work_color = next_color;
|
|
|
|
|
|
|
|
if (!wq->first_flusher) {
|
|
|
|
/* no flush in progress, become the first flusher */
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(wq->flush_color != this_flusher.flush_color);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
wq->first_flusher = &this_flusher;
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (!flush_workqueue_prep_pwqs(wq, wq->flush_color,
|
2010-06-29 02:07:11 -06:00
|
|
|
wq->work_color)) {
|
|
|
|
/* nothing to flush, done */
|
|
|
|
wq->flush_color = next_color;
|
|
|
|
wq->first_flusher = NULL;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/* wait in queue */
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(wq->flush_color == this_flusher.flush_color);
|
2010-06-29 02:07:11 -06:00
|
|
|
list_add_tail(&this_flusher.list, &wq->flusher_queue);
|
2013-02-13 20:29:12 -07:00
|
|
|
flush_workqueue_prep_pwqs(wq, -1, wq->work_color);
|
2010-06-29 02:07:11 -06:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Oops, color space is full, wait on overflow queue.
|
|
|
|
* The next flush completion will assign us
|
|
|
|
* flush_color and transfer to flusher_queue.
|
|
|
|
*/
|
|
|
|
list_add_tail(&this_flusher.list, &wq->flusher_overflow);
|
|
|
|
}
|
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
wait_for_completion(&this_flusher.done);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Wake-up-and-cascade phase
|
|
|
|
*
|
|
|
|
* First flushers are responsible for cascading flushes and
|
|
|
|
* handling overflow. Non-first flushers can simply return.
|
|
|
|
*/
|
|
|
|
if (wq->first_flusher != &this_flusher)
|
|
|
|
return;
|
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2010-07-02 02:03:51 -06:00
|
|
|
/* we might have raced, check again with mutex held */
|
|
|
|
if (wq->first_flusher != &this_flusher)
|
|
|
|
goto out_unlock;
|
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
wq->first_flusher = NULL;
|
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(!list_empty(&this_flusher.list));
|
|
|
|
WARN_ON_ONCE(wq->flush_color != this_flusher.flush_color);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
while (true) {
|
|
|
|
struct wq_flusher *next, *tmp;
|
|
|
|
|
|
|
|
/* complete all the flushers sharing the current flush color */
|
|
|
|
list_for_each_entry_safe(next, tmp, &wq->flusher_queue, list) {
|
|
|
|
if (next->flush_color != wq->flush_color)
|
|
|
|
break;
|
|
|
|
list_del_init(&next->list);
|
|
|
|
complete(&next->done);
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(!list_empty(&wq->flusher_overflow) &&
|
|
|
|
wq->flush_color != work_next_color(wq->work_color));
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
/* this flush_color is finished, advance by one */
|
|
|
|
wq->flush_color = work_next_color(wq->flush_color);
|
|
|
|
|
|
|
|
/* one color has been freed, handle overflow queue */
|
|
|
|
if (!list_empty(&wq->flusher_overflow)) {
|
|
|
|
/*
|
|
|
|
* Assign the same color to all overflowed
|
|
|
|
* flushers, advance work_color and append to
|
|
|
|
* flusher_queue. This is the start-to-wait
|
|
|
|
* phase for these overflowed flushers.
|
|
|
|
*/
|
|
|
|
list_for_each_entry(tmp, &wq->flusher_overflow, list)
|
|
|
|
tmp->flush_color = wq->work_color;
|
|
|
|
|
|
|
|
wq->work_color = work_next_color(wq->work_color);
|
|
|
|
|
|
|
|
list_splice_tail_init(&wq->flusher_overflow,
|
|
|
|
&wq->flusher_queue);
|
2013-02-13 20:29:12 -07:00
|
|
|
flush_workqueue_prep_pwqs(wq, -1, wq->work_color);
|
2010-06-29 02:07:11 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
if (list_empty(&wq->flusher_queue)) {
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(wq->flush_color != wq->work_color);
|
2010-06-29 02:07:11 -06:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Need to flush more colors. Make the next flusher
|
2013-02-13 20:29:12 -07:00
|
|
|
* the new first flusher and arm pwqs.
|
2010-06-29 02:07:11 -06:00
|
|
|
*/
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(wq->flush_color == wq->work_color);
|
|
|
|
WARN_ON_ONCE(wq->flush_color != next->flush_color);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
|
|
|
list_del_init(&next->list);
|
|
|
|
wq->first_flusher = next;
|
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
if (flush_workqueue_prep_pwqs(wq, wq->flush_color, -1))
|
2010-06-29 02:07:11 -06:00
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Meh... this color is already done, clear first
|
|
|
|
* flusher and repeat cascading.
|
|
|
|
*/
|
|
|
|
wq->first_flusher = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
out_unlock:
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2006-06-29 23:40:45 -06:00
|
|
|
EXPORT_SYMBOL_GPL(flush_workqueue);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2011-04-05 10:01:44 -06:00
|
|
|
/**
|
|
|
|
* drain_workqueue - drain a workqueue
|
|
|
|
* @wq: workqueue to drain
|
|
|
|
*
|
|
|
|
* Wait until the workqueue becomes empty. While draining is in progress,
|
|
|
|
* only chain queueing is allowed. IOW, only currently pending or running
|
|
|
|
* work items on @wq can queue further work items on it. @wq is flushed
|
|
|
|
* repeatedly until it becomes empty. The number of flushing is detemined
|
|
|
|
* by the depth of chaining and should be relatively short. Whine if it
|
|
|
|
* takes too long.
|
|
|
|
*/
|
|
|
|
void drain_workqueue(struct workqueue_struct *wq)
|
|
|
|
{
|
|
|
|
unsigned int flush_cnt = 0;
|
2013-03-12 12:29:58 -06:00
|
|
|
struct pool_workqueue *pwq;
|
2011-04-05 10:01:44 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* __queue_work() needs to test whether there are drainers, is much
|
|
|
|
* hotter than drain_workqueue() and already looks at @wq->flags.
|
2013-03-12 12:30:04 -06:00
|
|
|
* Use __WQ_DRAINING so that queue doesn't have to check nr_drainers.
|
2011-04-05 10:01:44 -06:00
|
|
|
*/
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2011-04-05 10:01:44 -06:00
|
|
|
if (!wq->nr_drainers++)
|
2013-03-12 12:30:04 -06:00
|
|
|
wq->flags |= __WQ_DRAINING;
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2011-04-05 10:01:44 -06:00
|
|
|
reflush:
|
|
|
|
flush_workqueue(wq);
|
|
|
|
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
for_each_pwq(pwq, wq) {
|
2011-09-14 17:22:28 -06:00
|
|
|
bool drained;
|
2011-04-05 10:01:44 -06:00
|
|
|
|
2013-03-25 17:57:18 -06:00
|
|
|
spin_lock_irq(&pwq->pool->lock);
|
2013-02-13 20:29:12 -07:00
|
|
|
drained = !pwq->nr_active && list_empty(&pwq->delayed_works);
|
2013-03-25 17:57:18 -06:00
|
|
|
spin_unlock_irq(&pwq->pool->lock);
|
2011-09-14 17:22:28 -06:00
|
|
|
|
|
|
|
if (drained)
|
2011-04-05 10:01:44 -06:00
|
|
|
continue;
|
|
|
|
|
|
|
|
if (++flush_cnt == 10 ||
|
|
|
|
(flush_cnt % 100 == 0 && flush_cnt <= 1000))
|
2013-03-13 17:51:36 -06:00
|
|
|
pr_warn("workqueue %s: drain_workqueue() isn't complete after %u tries\n",
|
2012-08-18 15:52:42 -06:00
|
|
|
wq->name, flush_cnt);
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2011-04-05 10:01:44 -06:00
|
|
|
goto reflush;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!--wq->nr_drainers)
|
2013-03-12 12:30:04 -06:00
|
|
|
wq->flags &= ~__WQ_DRAINING;
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2011-04-05 10:01:44 -06:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(drain_workqueue);
|
|
|
|
|
2012-08-20 15:51:23 -06:00
|
|
|
static bool start_flush_work(struct work_struct *work, struct wq_barrier *barr)
|
workqueues: implement flush_work()
Most of users of flush_workqueue() can be changed to use cancel_work_sync(),
but sometimes we really need to wait for the completion and cancelling is not
an option. schedule_on_each_cpu() is good example.
Add the new helper, flush_work(work), which waits for the completion of the
specific work_struct. More precisely, it "flushes" the result of of the last
queue_work() which is visible to the caller.
For example, this code
queue_work(wq, work);
/* WINDOW */
queue_work(wq, work);
flush_work(work);
doesn't necessary work "as expected". What can happen in the WINDOW above is
- wq starts the execution of work->func()
- the caller migrates to another CPU
now, after the 2nd queue_work() this work is active on the previous CPU, and
at the same time it is queued on another. In this case flush_work(work) may
return before the first work->func() completes.
It is trivial to add another helper
int flush_work_sync(struct work_struct *work)
{
return flush_work(work) || wait_on_work(work);
}
which works "more correctly", but it has to iterate over all CPUs and thus
it much slower than flush_work().
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Max Krasnyansky <maxk@qualcomm.com>
Acked-by: Jarek Poplawski <jarkao2@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 02:47:49 -06:00
|
|
|
{
|
2010-06-29 02:07:12 -06:00
|
|
|
struct worker *worker = NULL;
|
2013-01-24 12:01:33 -07:00
|
|
|
struct worker_pool *pool;
|
2013-02-13 20:29:12 -07:00
|
|
|
struct pool_workqueue *pwq;
|
workqueues: implement flush_work()
Most of users of flush_workqueue() can be changed to use cancel_work_sync(),
but sometimes we really need to wait for the completion and cancelling is not
an option. schedule_on_each_cpu() is good example.
Add the new helper, flush_work(work), which waits for the completion of the
specific work_struct. More precisely, it "flushes" the result of of the last
queue_work() which is visible to the caller.
For example, this code
queue_work(wq, work);
/* WINDOW */
queue_work(wq, work);
flush_work(work);
doesn't necessary work "as expected". What can happen in the WINDOW above is
- wq starts the execution of work->func()
- the caller migrates to another CPU
now, after the 2nd queue_work() this work is active on the previous CPU, and
at the same time it is queued on another. In this case flush_work(work) may
return before the first work->func() completes.
It is trivial to add another helper
int flush_work_sync(struct work_struct *work)
{
return flush_work(work) || wait_on_work(work);
}
which works "more correctly", but it has to iterate over all CPUs and thus
it much slower than flush_work().
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Max Krasnyansky <maxk@qualcomm.com>
Acked-by: Jarek Poplawski <jarkao2@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 02:47:49 -06:00
|
|
|
|
|
|
|
might_sleep();
|
2013-03-12 12:30:00 -06:00
|
|
|
|
|
|
|
local_irq_disable();
|
2013-01-24 12:01:33 -07:00
|
|
|
pool = get_work_pool(work);
|
2013-03-12 12:30:00 -06:00
|
|
|
if (!pool) {
|
|
|
|
local_irq_enable();
|
2010-09-16 02:42:16 -06:00
|
|
|
return false;
|
2013-03-12 12:30:00 -06:00
|
|
|
}
|
workqueues: implement flush_work()
Most of users of flush_workqueue() can be changed to use cancel_work_sync(),
but sometimes we really need to wait for the completion and cancelling is not
an option. schedule_on_each_cpu() is good example.
Add the new helper, flush_work(work), which waits for the completion of the
specific work_struct. More precisely, it "flushes" the result of of the last
queue_work() which is visible to the caller.
For example, this code
queue_work(wq, work);
/* WINDOW */
queue_work(wq, work);
flush_work(work);
doesn't necessary work "as expected". What can happen in the WINDOW above is
- wq starts the execution of work->func()
- the caller migrates to another CPU
now, after the 2nd queue_work() this work is active on the previous CPU, and
at the same time it is queued on another. In this case flush_work(work) may
return before the first work->func() completes.
It is trivial to add another helper
int flush_work_sync(struct work_struct *work)
{
return flush_work(work) || wait_on_work(work);
}
which works "more correctly", but it has to iterate over all CPUs and thus
it much slower than flush_work().
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Max Krasnyansky <maxk@qualcomm.com>
Acked-by: Jarek Poplawski <jarkao2@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 02:47:49 -06:00
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
spin_lock(&pool->lock);
|
workqueue: simplify is-work-item-queued-here test
Currently, determining whether a work item is queued on a locked pool
involves somewhat convoluted memory barrier dancing. It goes like the
following.
* When a work item is queued on a pool, work->data is updated before
work->entry is linked to the pending list with a wmb() inbetween.
* When trying to determine whether a work item is currently queued on
a pool pointed to by work->data, it locks the pool and looks at
work->entry. If work->entry is linked, we then do rmb() and then
check whether work->data points to the current pool.
This works because, work->data can only point to a pool if it
currently is or were on the pool and,
* If it currently is on the pool, the tests would obviously succeed.
* It it left the pool, its work->entry was cleared under pool->lock,
so if we're seeing non-empty work->entry, it has to be from the work
item being linked on another pool. Because work->data is updated
before work->entry is linked with wmb() inbetween, work->data update
from another pool is guaranteed to be visible if we do rmb() after
seeing non-empty work->entry. So, we either see empty work->entry
or we see updated work->data pointin to another pool.
While this works, it's convoluted, to put it mildly. With recent
updates, it's now guaranteed that work->data points to cwq only while
the work item is queued and that updating work->data to point to cwq
or back to pool is done under pool->lock, so we can simply test
whether work->data points to cwq which is associated with the
currently locked pool instead of the convoluted memory barrier
dancing.
This patch replaces the memory barrier based "are you still here,
really?" test with much simpler "does work->data points to me?" test -
if work->data points to a cwq which is associated with the currently
locked pool, the work item is guaranteed to be queued on the pool as
work->data can start and stop pointing to such cwq only under
pool->lock and the start and stop coincide with queue and dequeue.
tj: Rewrote the comments and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
2013-02-06 19:04:53 -07:00
|
|
|
/* see the comment in try_to_grab_pending() with the same code */
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq = get_work_pwq(work);
|
|
|
|
if (pwq) {
|
|
|
|
if (unlikely(pwq->pool != pool))
|
2010-06-29 02:07:10 -06:00
|
|
|
goto already_gone;
|
2012-08-20 15:51:23 -06:00
|
|
|
} else {
|
2013-01-24 12:01:33 -07:00
|
|
|
worker = find_worker_executing_work(pool, work);
|
2010-06-29 02:07:12 -06:00
|
|
|
if (!worker)
|
2010-06-29 02:07:10 -06:00
|
|
|
goto already_gone;
|
2013-02-13 20:29:12 -07:00
|
|
|
pwq = worker->current_pwq;
|
2012-08-20 15:51:23 -06:00
|
|
|
}
|
workqueues: implement flush_work()
Most of users of flush_workqueue() can be changed to use cancel_work_sync(),
but sometimes we really need to wait for the completion and cancelling is not
an option. schedule_on_each_cpu() is good example.
Add the new helper, flush_work(work), which waits for the completion of the
specific work_struct. More precisely, it "flushes" the result of of the last
queue_work() which is visible to the caller.
For example, this code
queue_work(wq, work);
/* WINDOW */
queue_work(wq, work);
flush_work(work);
doesn't necessary work "as expected". What can happen in the WINDOW above is
- wq starts the execution of work->func()
- the caller migrates to another CPU
now, after the 2nd queue_work() this work is active on the previous CPU, and
at the same time it is queued on another. In this case flush_work(work) may
return before the first work->func() completes.
It is trivial to add another helper
int flush_work_sync(struct work_struct *work)
{
return flush_work(work) || wait_on_work(work);
}
which works "more correctly", but it has to iterate over all CPUs and thus
it much slower than flush_work().
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Max Krasnyansky <maxk@qualcomm.com>
Acked-by: Jarek Poplawski <jarkao2@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 02:47:49 -06:00
|
|
|
|
2013-02-13 20:29:12 -07:00
|
|
|
insert_wq_barrier(pwq, barr, work, worker);
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2011-01-09 15:32:15 -07:00
|
|
|
/*
|
|
|
|
* If @max_active is 1 or rescuer is in use, flushing another work
|
|
|
|
* item on the same workqueue may lead to deadlock. Make sure the
|
|
|
|
* flusher is not running on the same workqueue by verifying write
|
|
|
|
* access.
|
|
|
|
*/
|
2013-03-12 12:30:03 -06:00
|
|
|
if (pwq->wq->saved_max_active == 1 || pwq->wq->rescuer)
|
2013-02-13 20:29:12 -07:00
|
|
|
lock_map_acquire(&pwq->wq->lockdep_map);
|
2011-01-09 15:32:15 -07:00
|
|
|
else
|
2013-02-13 20:29:12 -07:00
|
|
|
lock_map_acquire_read(&pwq->wq->lockdep_map);
|
|
|
|
lock_map_release(&pwq->wq->lockdep_map);
|
2011-01-09 15:32:15 -07:00
|
|
|
|
2010-09-16 02:36:00 -06:00
|
|
|
return true;
|
2010-06-29 02:07:10 -06:00
|
|
|
already_gone:
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2010-09-16 02:36:00 -06:00
|
|
|
return false;
|
workqueues: implement flush_work()
Most of users of flush_workqueue() can be changed to use cancel_work_sync(),
but sometimes we really need to wait for the completion and cancelling is not
an option. schedule_on_each_cpu() is good example.
Add the new helper, flush_work(work), which waits for the completion of the
specific work_struct. More precisely, it "flushes" the result of of the last
queue_work() which is visible to the caller.
For example, this code
queue_work(wq, work);
/* WINDOW */
queue_work(wq, work);
flush_work(work);
doesn't necessary work "as expected". What can happen in the WINDOW above is
- wq starts the execution of work->func()
- the caller migrates to another CPU
now, after the 2nd queue_work() this work is active on the previous CPU, and
at the same time it is queued on another. In this case flush_work(work) may
return before the first work->func() completes.
It is trivial to add another helper
int flush_work_sync(struct work_struct *work)
{
return flush_work(work) || wait_on_work(work);
}
which works "more correctly", but it has to iterate over all CPUs and thus
it much slower than flush_work().
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Max Krasnyansky <maxk@qualcomm.com>
Acked-by: Jarek Poplawski <jarkao2@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 02:47:49 -06:00
|
|
|
}
|
2010-09-16 02:42:16 -06:00
|
|
|
|
|
|
|
/**
|
|
|
|
* flush_work - wait for a work to finish executing the last queueing instance
|
|
|
|
* @work: the work to flush
|
|
|
|
*
|
2012-08-20 15:51:23 -06:00
|
|
|
* Wait until @work has finished execution. @work is guaranteed to be idle
|
|
|
|
* on return if it hasn't been requeued since flush started.
|
2010-09-16 02:42:16 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-09-16 02:42:16 -06:00
|
|
|
* %true if flush_work() waited for the work to finish execution,
|
|
|
|
* %false if it was already idle.
|
|
|
|
*/
|
|
|
|
bool flush_work(struct work_struct *work)
|
|
|
|
{
|
2013-11-18 11:00:29 -07:00
|
|
|
struct wq_barrier barr;
|
|
|
|
|
2012-04-20 18:28:50 -06:00
|
|
|
lock_map_acquire(&work->lockdep_map);
|
|
|
|
lock_map_release(&work->lockdep_map);
|
|
|
|
|
2013-11-18 11:00:29 -07:00
|
|
|
if (start_flush_work(work, &barr)) {
|
|
|
|
wait_for_completion(&barr.done);
|
|
|
|
destroy_work_on_stack(&barr.work);
|
|
|
|
return true;
|
|
|
|
} else {
|
|
|
|
return false;
|
|
|
|
}
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
}
|
2012-08-20 15:51:23 -06:00
|
|
|
EXPORT_SYMBOL_GPL(flush_work);
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
static bool __cancel_work_timer(struct work_struct *work, bool is_dwork)
|
2007-07-16 00:41:44 -06:00
|
|
|
{
|
2012-08-03 11:30:46 -06:00
|
|
|
unsigned long flags;
|
2007-07-16 00:41:44 -06:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
do {
|
2012-08-03 11:30:46 -06:00
|
|
|
ret = try_to_grab_pending(work, is_dwork, &flags);
|
|
|
|
/*
|
|
|
|
* If someone else is canceling, wait for the same event it
|
|
|
|
* would be waiting for before retrying.
|
|
|
|
*/
|
|
|
|
if (unlikely(ret == -ENOENT))
|
2012-08-20 15:51:23 -06:00
|
|
|
flush_work(work);
|
2007-07-16 00:41:44 -06:00
|
|
|
} while (unlikely(ret < 0));
|
|
|
|
|
2012-08-03 11:30:46 -06:00
|
|
|
/* tell other tasks trying to grab @work to back off */
|
|
|
|
mark_work_canceling(work);
|
|
|
|
local_irq_restore(flags);
|
|
|
|
|
2012-08-20 15:51:23 -06:00
|
|
|
flush_work(work);
|
2010-06-29 02:07:13 -06:00
|
|
|
clear_work_data(work);
|
2007-07-16 00:41:44 -06:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
/**
|
2010-09-16 02:36:00 -06:00
|
|
|
* cancel_work_sync - cancel a work and wait for it to finish
|
|
|
|
* @work: the work to cancel
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
*
|
2010-09-16 02:36:00 -06:00
|
|
|
* Cancel @work and wait for its execution to finish. This function
|
|
|
|
* can be used even if the work re-queues itself or migrates to
|
|
|
|
* another workqueue. On return from this function, @work is
|
|
|
|
* guaranteed to be not pending or executing on any CPU.
|
2007-07-16 00:41:44 -06:00
|
|
|
*
|
2010-09-16 02:36:00 -06:00
|
|
|
* cancel_work_sync(&delayed_work->work) must not be used for
|
|
|
|
* delayed_work's. Use cancel_delayed_work_sync() instead.
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
*
|
2010-09-16 02:36:00 -06:00
|
|
|
* The caller must ensure that the workqueue on which @work was last
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
* queued can't be destroyed before this function returns.
|
2010-09-16 02:36:00 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-09-16 02:36:00 -06:00
|
|
|
* %true if @work was pending, %false otherwise.
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
*/
|
2010-09-16 02:36:00 -06:00
|
|
|
bool cancel_work_sync(struct work_struct *work)
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
{
|
2012-08-03 11:30:46 -06:00
|
|
|
return __cancel_work_timer(work, false);
|
implement flush_work()
A basic problem with flush_scheduled_work() is that it blocks behind _all_
presently-queued works, rather than just the work whcih the caller wants to
flush. If the caller holds some lock, and if one of the queued work happens
to want that lock as well then accidental deadlocks can occur.
One example of this is the phy layer: it wants to flush work while holding
rtnl_lock(). But if a linkwatch event happens to be queued, the phy code will
deadlock because the linkwatch callback function takes rtnl_lock.
So we implement a new function which will flush a *single* work - just the one
which the caller wants to free up. Thus we avoid the accidental deadlocks
which can arise from unrelated subsystems' callbacks taking shared locks.
flush_work() non-blockingly dequeues the work_struct which we want to kill,
then it waits for its handler to complete on all CPUs.
Add ->current_work to the "struct cpu_workqueue_struct", it points to
currently running "struct work_struct". When flush_work(work) detects
->current_work == work, it inserts a barrier at the _head_ of ->worklist
(and thus right _after_ that work) and waits for completition. This means
that the next work fired on that CPU will be this barrier, or another
barrier queued by concurrent flush_work(), so the caller of flush_work()
will be woken before any "regular" work has a chance to run.
When wait_on_work() unlocks workqueue_mutex (or whatever we choose to protect
against CPU hotplug), CPU may go away. But in that case take_over_work() will
move a barrier we queued to another CPU, it will be fired sometime, and
wait_on_work() will be woken.
Actually, we are doing cleanup_workqueue_thread()->kthread_stop() before
take_over_work(), so cwq->thread should complete its ->worklist (and thus
the barrier), because currently we don't check kthread_should_stop() in
run_workqueue(). But even if we did, everything should be ok.
[akpm@osdl.org: cleanup]
[akpm@osdl.org: add flush_work_keventd() wrapper]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:33:52 -06:00
|
|
|
}
|
2007-05-09 03:34:22 -06:00
|
|
|
EXPORT_SYMBOL_GPL(cancel_work_sync);
|
implement flush_work()
A basic problem with flush_scheduled_work() is that it blocks behind _all_
presently-queued works, rather than just the work whcih the caller wants to
flush. If the caller holds some lock, and if one of the queued work happens
to want that lock as well then accidental deadlocks can occur.
One example of this is the phy layer: it wants to flush work while holding
rtnl_lock(). But if a linkwatch event happens to be queued, the phy code will
deadlock because the linkwatch callback function takes rtnl_lock.
So we implement a new function which will flush a *single* work - just the one
which the caller wants to free up. Thus we avoid the accidental deadlocks
which can arise from unrelated subsystems' callbacks taking shared locks.
flush_work() non-blockingly dequeues the work_struct which we want to kill,
then it waits for its handler to complete on all CPUs.
Add ->current_work to the "struct cpu_workqueue_struct", it points to
currently running "struct work_struct". When flush_work(work) detects
->current_work == work, it inserts a barrier at the _head_ of ->worklist
(and thus right _after_ that work) and waits for completition. This means
that the next work fired on that CPU will be this barrier, or another
barrier queued by concurrent flush_work(), so the caller of flush_work()
will be woken before any "regular" work has a chance to run.
When wait_on_work() unlocks workqueue_mutex (or whatever we choose to protect
against CPU hotplug), CPU may go away. But in that case take_over_work() will
move a barrier we queued to another CPU, it will be fired sometime, and
wait_on_work() will be woken.
Actually, we are doing cleanup_workqueue_thread()->kthread_stop() before
take_over_work(), so cwq->thread should complete its ->worklist (and thus
the barrier), because currently we don't check kthread_should_stop() in
run_workqueue(). But even if we did, everything should be ok.
[akpm@osdl.org: cleanup]
[akpm@osdl.org: add flush_work_keventd() wrapper]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:33:52 -06:00
|
|
|
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
/**
|
2010-09-16 02:36:00 -06:00
|
|
|
* flush_delayed_work - wait for a dwork to finish executing the last queueing
|
|
|
|
* @dwork: the delayed work to flush
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
*
|
2010-09-16 02:36:00 -06:00
|
|
|
* Delayed timer is cancelled and the pending work is queued for
|
|
|
|
* immediate execution. Like flush_work(), this function only
|
|
|
|
* considers the last queueing instance of @dwork.
|
2007-07-16 00:41:44 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-09-16 02:36:00 -06:00
|
|
|
* %true if flush_work() waited for the work to finish execution,
|
|
|
|
* %false if it was already idle.
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
*/
|
2010-09-16 02:36:00 -06:00
|
|
|
bool flush_delayed_work(struct delayed_work *dwork)
|
|
|
|
{
|
2012-08-03 11:30:45 -06:00
|
|
|
local_irq_disable();
|
2010-09-16 02:36:00 -06:00
|
|
|
if (del_timer_sync(&dwork->timer))
|
2013-02-06 19:04:53 -07:00
|
|
|
__queue_work(dwork->cpu, dwork->wq, &dwork->work);
|
2012-08-03 11:30:45 -06:00
|
|
|
local_irq_enable();
|
2010-09-16 02:36:00 -06:00
|
|
|
return flush_work(&dwork->work);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(flush_delayed_work);
|
|
|
|
|
2010-09-16 02:48:29 -06:00
|
|
|
/**
|
2012-08-21 14:18:24 -06:00
|
|
|
* cancel_delayed_work - cancel a delayed work
|
|
|
|
* @dwork: delayed_work to cancel
|
2010-09-16 02:48:29 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Kill off a pending delayed_work.
|
|
|
|
*
|
|
|
|
* Return: %true if @dwork was pending and canceled; %false if it wasn't
|
|
|
|
* pending.
|
|
|
|
*
|
|
|
|
* Note:
|
|
|
|
* The work callback function may still be running on return, unless
|
|
|
|
* it returns %true and the work doesn't re-arm itself. Explicitly flush or
|
|
|
|
* use cancel_delayed_work_sync() to wait on it.
|
2010-09-16 02:48:29 -06:00
|
|
|
*
|
2012-08-21 14:18:24 -06:00
|
|
|
* This function is safe to call from any context including IRQ handler.
|
2010-09-16 02:48:29 -06:00
|
|
|
*/
|
2012-08-21 14:18:24 -06:00
|
|
|
bool cancel_delayed_work(struct delayed_work *dwork)
|
2010-09-16 02:48:29 -06:00
|
|
|
{
|
2012-08-21 14:18:24 -06:00
|
|
|
unsigned long flags;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
do {
|
|
|
|
ret = try_to_grab_pending(&dwork->work, true, &flags);
|
|
|
|
} while (unlikely(ret == -EAGAIN));
|
|
|
|
|
|
|
|
if (unlikely(ret < 0))
|
|
|
|
return false;
|
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
set_work_pool_and_clear_pending(&dwork->work,
|
|
|
|
get_work_pool_id(&dwork->work));
|
2012-08-21 14:18:24 -06:00
|
|
|
local_irq_restore(flags);
|
2012-10-18 17:31:37 -06:00
|
|
|
return ret;
|
2010-09-16 02:48:29 -06:00
|
|
|
}
|
2012-08-21 14:18:24 -06:00
|
|
|
EXPORT_SYMBOL(cancel_delayed_work);
|
2010-09-16 02:48:29 -06:00
|
|
|
|
2010-09-16 02:36:00 -06:00
|
|
|
/**
|
|
|
|
* cancel_delayed_work_sync - cancel a delayed work and wait for it to finish
|
|
|
|
* @dwork: the delayed work cancel
|
|
|
|
*
|
|
|
|
* This is cancel_work_sync() for delayed works.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-09-16 02:36:00 -06:00
|
|
|
* %true if @dwork was pending, %false otherwise.
|
|
|
|
*/
|
|
|
|
bool cancel_delayed_work_sync(struct delayed_work *dwork)
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
{
|
2012-08-03 11:30:46 -06:00
|
|
|
return __cancel_work_timer(&dwork->work, true);
|
make cancel_rearming_delayed_work() reliable
Thanks to Jarek Poplawski for the ideas and for spotting the bug in the
initial draft patch.
cancel_rearming_delayed_work() currently has many limitations, because it
requires that dwork always re-arms itself via queue_delayed_work(). So it
hangs forever if dwork doesn't do this, or cancel_rearming_delayed_work/
cancel_delayed_work was already called. It uses flush_workqueue() in a
loop, so it can't be used if workqueue was freezed, and it is potentially
live- lockable on busy system if delay is small.
With this patch cancel_rearming_delayed_work() doesn't make any assumptions
about dwork, it can re-arm itself via queue_delayed_work(), or
queue_work(), or do nothing.
As a "side effect", cancel_work_sync() was changed to handle re-arming works
as well.
Disadvantages:
- this patch adds wmb() to insert_work().
- slowdowns the fast path (when del_timer() succeeds on entry) of
cancel_rearming_delayed_work(), because wait_on_work() is called
unconditionally. In that case, compared to the old version, we are
doing "unneeded" lock/unlock for each online CPU.
On the other hand, this means we don't need to use cancel_work_sync()
after cancel_rearming_delayed_work().
- complicates the code (.text grows by 130 bytes).
[akpm@linux-foundation.org: fix speling]
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Chinner <dgc@sgi.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Gautham Shenoy <ego@in.ibm.com>
Acked-by: Jarek Poplawski <jarkao2@o2.pl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 03:34:46 -06:00
|
|
|
}
|
2007-07-16 00:41:44 -06:00
|
|
|
EXPORT_SYMBOL(cancel_delayed_work_sync);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2006-06-25 06:47:49 -06:00
|
|
|
/**
|
2010-10-19 03:14:49 -06:00
|
|
|
* schedule_on_each_cpu - execute a function synchronously on each online CPU
|
2006-06-25 06:47:49 -06:00
|
|
|
* @func: the function to call
|
|
|
|
*
|
2010-10-19 03:14:49 -06:00
|
|
|
* schedule_on_each_cpu() executes @func on each online CPU using the
|
|
|
|
* system workqueue and blocks until all CPUs have completed.
|
2006-06-25 06:47:49 -06:00
|
|
|
* schedule_on_each_cpu() is very slow.
|
2010-10-19 03:14:49 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-10-19 03:14:49 -06:00
|
|
|
* 0 on success, -errno on failure.
|
2006-06-25 06:47:49 -06:00
|
|
|
*/
|
2006-11-22 07:55:48 -07:00
|
|
|
int schedule_on_each_cpu(work_func_t func)
|
2006-01-08 02:00:43 -07:00
|
|
|
{
|
|
|
|
int cpu;
|
2010-08-08 06:24:09 -06:00
|
|
|
struct work_struct __percpu *works;
|
2006-01-08 02:00:43 -07:00
|
|
|
|
2006-06-25 06:47:49 -06:00
|
|
|
works = alloc_percpu(struct work_struct);
|
|
|
|
if (!works)
|
2006-01-08 02:00:43 -07:00
|
|
|
return -ENOMEM;
|
2006-06-25 06:47:49 -06:00
|
|
|
|
2009-11-17 15:06:20 -07:00
|
|
|
get_online_cpus();
|
|
|
|
|
2006-01-08 02:00:43 -07:00
|
|
|
for_each_online_cpu(cpu) {
|
2006-12-18 12:05:09 -07:00
|
|
|
struct work_struct *work = per_cpu_ptr(works, cpu);
|
|
|
|
|
|
|
|
INIT_WORK(work, func);
|
2010-06-29 02:07:14 -06:00
|
|
|
schedule_work_on(cpu, work);
|
2009-10-13 22:22:47 -06:00
|
|
|
}
|
2009-11-17 15:06:20 -07:00
|
|
|
|
|
|
|
for_each_online_cpu(cpu)
|
|
|
|
flush_work(per_cpu_ptr(works, cpu));
|
|
|
|
|
2008-01-25 13:08:02 -07:00
|
|
|
put_online_cpus();
|
2006-06-25 06:47:49 -06:00
|
|
|
free_percpu(works);
|
2006-01-08 02:00:43 -07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-02-12 01:39:21 -07:00
|
|
|
/**
|
|
|
|
* flush_scheduled_work - ensure that any scheduled work has run to completion.
|
|
|
|
*
|
|
|
|
* Forces execution of the kernel-global workqueue and blocks until its
|
|
|
|
* completion.
|
|
|
|
*
|
|
|
|
* Think twice before calling this function! It's very easy to get into
|
|
|
|
* trouble if you don't take great care. Either of the following situations
|
|
|
|
* will lead to deadlock:
|
|
|
|
*
|
|
|
|
* One of the work items currently on the workqueue needs to acquire
|
|
|
|
* a lock held by your code or its caller.
|
|
|
|
*
|
|
|
|
* Your code is running in the context of a work routine.
|
|
|
|
*
|
|
|
|
* They will be detected by lockdep when they occur, but the first might not
|
|
|
|
* occur very often. It depends on what work items are on the workqueue and
|
|
|
|
* what locks they need, which you have no control over.
|
|
|
|
*
|
|
|
|
* In most situations flushing the entire workqueue is overkill; you merely
|
|
|
|
* need to know that a particular work item isn't queued and isn't running.
|
|
|
|
* In such cases you should use cancel_delayed_work_sync() or
|
|
|
|
* cancel_work_sync() instead.
|
|
|
|
*/
|
2005-04-16 16:20:36 -06:00
|
|
|
void flush_scheduled_work(void)
|
|
|
|
{
|
2010-06-29 02:07:14 -06:00
|
|
|
flush_workqueue(system_wq);
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2006-06-29 23:40:45 -06:00
|
|
|
EXPORT_SYMBOL(flush_scheduled_work);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2006-02-23 11:43:43 -07:00
|
|
|
/**
|
|
|
|
* execute_in_process_context - reliably execute the routine with user context
|
|
|
|
* @fn: the function to execute
|
|
|
|
* @ew: guaranteed storage for the execute work structure (must
|
|
|
|
* be available when the work executes)
|
|
|
|
*
|
|
|
|
* Executes the function immediately if process context is available,
|
|
|
|
* otherwise schedules the function for delayed execution.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return: 0 - function was executed
|
2006-02-23 11:43:43 -07:00
|
|
|
* 1 - function was scheduled for execution
|
|
|
|
*/
|
2006-11-22 07:55:48 -07:00
|
|
|
int execute_in_process_context(work_func_t fn, struct execute_work *ew)
|
2006-02-23 11:43:43 -07:00
|
|
|
{
|
|
|
|
if (!in_interrupt()) {
|
2006-11-22 07:55:48 -07:00
|
|
|
fn(&ew->work);
|
2006-02-23 11:43:43 -07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-11-22 07:55:48 -07:00
|
|
|
INIT_WORK(&ew->work, fn);
|
2006-02-23 11:43:43 -07:00
|
|
|
schedule_work(&ew->work);
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(execute_in_process_context);
|
|
|
|
|
2013-03-12 12:30:05 -06:00
|
|
|
#ifdef CONFIG_SYSFS
|
|
|
|
/*
|
|
|
|
* Workqueues with WQ_SYSFS flag set is visible to userland via
|
|
|
|
* /sys/bus/workqueue/devices/WQ_NAME. All visible workqueues have the
|
|
|
|
* following attributes.
|
|
|
|
*
|
|
|
|
* per_cpu RO bool : whether the workqueue is per-cpu or unbound
|
|
|
|
* max_active RW int : maximum number of in-flight work items
|
|
|
|
*
|
|
|
|
* Unbound workqueues have the following extra attributes.
|
|
|
|
*
|
|
|
|
* id RO int : the associated pool ID
|
|
|
|
* nice RW int : nice value of the workers
|
|
|
|
* cpumask RW mask : bitmask of allowed CPUs for the workers
|
|
|
|
*/
|
|
|
|
struct wq_device {
|
|
|
|
struct workqueue_struct *wq;
|
|
|
|
struct device dev;
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct workqueue_struct *dev_to_wq(struct device *dev)
|
|
|
|
{
|
|
|
|
struct wq_device *wq_dev = container_of(dev, struct wq_device, dev);
|
|
|
|
|
|
|
|
return wq_dev->wq;
|
|
|
|
}
|
|
|
|
|
2013-08-23 15:24:41 -06:00
|
|
|
static ssize_t per_cpu_show(struct device *dev, struct device_attribute *attr,
|
|
|
|
char *buf)
|
2013-03-12 12:30:05 -06:00
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
|
|
|
|
return scnprintf(buf, PAGE_SIZE, "%d\n", (bool)!(wq->flags & WQ_UNBOUND));
|
|
|
|
}
|
2013-08-23 15:24:41 -06:00
|
|
|
static DEVICE_ATTR_RO(per_cpu);
|
2013-03-12 12:30:05 -06:00
|
|
|
|
2013-08-23 15:24:41 -06:00
|
|
|
static ssize_t max_active_show(struct device *dev,
|
|
|
|
struct device_attribute *attr, char *buf)
|
2013-03-12 12:30:05 -06:00
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
|
|
|
|
return scnprintf(buf, PAGE_SIZE, "%d\n", wq->saved_max_active);
|
|
|
|
}
|
|
|
|
|
2013-08-23 15:24:41 -06:00
|
|
|
static ssize_t max_active_store(struct device *dev,
|
|
|
|
struct device_attribute *attr, const char *buf,
|
|
|
|
size_t count)
|
2013-03-12 12:30:05 -06:00
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
int val;
|
|
|
|
|
|
|
|
if (sscanf(buf, "%d", &val) != 1 || val <= 0)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
workqueue_set_max_active(wq, val);
|
|
|
|
return count;
|
|
|
|
}
|
2013-08-23 15:24:41 -06:00
|
|
|
static DEVICE_ATTR_RW(max_active);
|
2013-03-12 12:30:05 -06:00
|
|
|
|
2013-08-23 15:24:41 -06:00
|
|
|
static struct attribute *wq_sysfs_attrs[] = {
|
|
|
|
&dev_attr_per_cpu.attr,
|
|
|
|
&dev_attr_max_active.attr,
|
|
|
|
NULL,
|
2013-03-12 12:30:05 -06:00
|
|
|
};
|
2013-08-23 15:24:41 -06:00
|
|
|
ATTRIBUTE_GROUPS(wq_sysfs);
|
2013-03-12 12:30:05 -06:00
|
|
|
|
2013-04-01 12:23:38 -06:00
|
|
|
static ssize_t wq_pool_ids_show(struct device *dev,
|
|
|
|
struct device_attribute *attr, char *buf)
|
2013-03-12 12:30:05 -06:00
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
2013-04-01 12:23:38 -06:00
|
|
|
const char *delim = "";
|
|
|
|
int node, written = 0;
|
2013-03-12 12:30:05 -06:00
|
|
|
|
|
|
|
rcu_read_lock_sched();
|
2013-04-01 12:23:38 -06:00
|
|
|
for_each_node(node) {
|
|
|
|
written += scnprintf(buf + written, PAGE_SIZE - written,
|
|
|
|
"%s%d:%d", delim, node,
|
|
|
|
unbound_pwq_by_node(wq, node)->pool->id);
|
|
|
|
delim = " ";
|
|
|
|
}
|
|
|
|
written += scnprintf(buf + written, PAGE_SIZE - written, "\n");
|
2013-03-12 12:30:05 -06:00
|
|
|
rcu_read_unlock_sched();
|
|
|
|
|
|
|
|
return written;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t wq_nice_show(struct device *dev, struct device_attribute *attr,
|
|
|
|
char *buf)
|
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
int written;
|
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
|
|
|
written = scnprintf(buf, PAGE_SIZE, "%d\n", wq->unbound_attrs->nice);
|
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:30:05 -06:00
|
|
|
|
|
|
|
return written;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* prepare workqueue_attrs for sysfs store operations */
|
|
|
|
static struct workqueue_attrs *wq_sysfs_prep_attrs(struct workqueue_struct *wq)
|
|
|
|
{
|
|
|
|
struct workqueue_attrs *attrs;
|
|
|
|
|
|
|
|
attrs = alloc_workqueue_attrs(GFP_KERNEL);
|
|
|
|
if (!attrs)
|
|
|
|
return NULL;
|
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
|
|
|
copy_workqueue_attrs(attrs, wq->unbound_attrs);
|
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:30:05 -06:00
|
|
|
return attrs;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t wq_nice_store(struct device *dev, struct device_attribute *attr,
|
|
|
|
const char *buf, size_t count)
|
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
struct workqueue_attrs *attrs;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
attrs = wq_sysfs_prep_attrs(wq);
|
|
|
|
if (!attrs)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
if (sscanf(buf, "%d", &attrs->nice) == 1 &&
|
2014-02-11 00:34:52 -07:00
|
|
|
attrs->nice >= MIN_NICE && attrs->nice <= MAX_NICE)
|
2013-03-12 12:30:05 -06:00
|
|
|
ret = apply_workqueue_attrs(wq, attrs);
|
|
|
|
else
|
|
|
|
ret = -EINVAL;
|
|
|
|
|
|
|
|
free_workqueue_attrs(attrs);
|
|
|
|
return ret ?: count;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t wq_cpumask_show(struct device *dev,
|
|
|
|
struct device_attribute *attr, char *buf)
|
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
int written;
|
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
|
|
|
written = cpumask_scnprintf(buf, PAGE_SIZE, wq->unbound_attrs->cpumask);
|
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:30:05 -06:00
|
|
|
|
|
|
|
written += scnprintf(buf + written, PAGE_SIZE - written, "\n");
|
|
|
|
return written;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t wq_cpumask_store(struct device *dev,
|
|
|
|
struct device_attribute *attr,
|
|
|
|
const char *buf, size_t count)
|
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
struct workqueue_attrs *attrs;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
attrs = wq_sysfs_prep_attrs(wq);
|
|
|
|
if (!attrs)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
ret = cpumask_parse(buf, attrs->cpumask);
|
|
|
|
if (!ret)
|
|
|
|
ret = apply_workqueue_attrs(wq, attrs);
|
|
|
|
|
|
|
|
free_workqueue_attrs(attrs);
|
|
|
|
return ret ?: count;
|
|
|
|
}
|
|
|
|
|
2013-04-01 12:23:38 -06:00
|
|
|
static ssize_t wq_numa_show(struct device *dev, struct device_attribute *attr,
|
|
|
|
char *buf)
|
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
int written;
|
|
|
|
|
|
|
|
mutex_lock(&wq->mutex);
|
|
|
|
written = scnprintf(buf, PAGE_SIZE, "%d\n",
|
|
|
|
!wq->unbound_attrs->no_numa);
|
|
|
|
mutex_unlock(&wq->mutex);
|
|
|
|
|
|
|
|
return written;
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t wq_numa_store(struct device *dev, struct device_attribute *attr,
|
|
|
|
const char *buf, size_t count)
|
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = dev_to_wq(dev);
|
|
|
|
struct workqueue_attrs *attrs;
|
|
|
|
int v, ret;
|
|
|
|
|
|
|
|
attrs = wq_sysfs_prep_attrs(wq);
|
|
|
|
if (!attrs)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
if (sscanf(buf, "%d", &v) == 1) {
|
|
|
|
attrs->no_numa = !v;
|
|
|
|
ret = apply_workqueue_attrs(wq, attrs);
|
|
|
|
}
|
|
|
|
|
|
|
|
free_workqueue_attrs(attrs);
|
|
|
|
return ret ?: count;
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:05 -06:00
|
|
|
static struct device_attribute wq_sysfs_unbound_attrs[] = {
|
2013-04-01 12:23:38 -06:00
|
|
|
__ATTR(pool_ids, 0444, wq_pool_ids_show, NULL),
|
2013-03-12 12:30:05 -06:00
|
|
|
__ATTR(nice, 0644, wq_nice_show, wq_nice_store),
|
|
|
|
__ATTR(cpumask, 0644, wq_cpumask_show, wq_cpumask_store),
|
2013-04-01 12:23:38 -06:00
|
|
|
__ATTR(numa, 0644, wq_numa_show, wq_numa_store),
|
2013-03-12 12:30:05 -06:00
|
|
|
__ATTR_NULL,
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct bus_type wq_subsys = {
|
|
|
|
.name = "workqueue",
|
2013-08-23 15:24:41 -06:00
|
|
|
.dev_groups = wq_sysfs_groups,
|
2013-03-12 12:30:05 -06:00
|
|
|
};
|
|
|
|
|
|
|
|
static int __init wq_sysfs_init(void)
|
|
|
|
{
|
|
|
|
return subsys_virtual_register(&wq_subsys, NULL);
|
|
|
|
}
|
|
|
|
core_initcall(wq_sysfs_init);
|
|
|
|
|
|
|
|
static void wq_device_release(struct device *dev)
|
|
|
|
{
|
|
|
|
struct wq_device *wq_dev = container_of(dev, struct wq_device, dev);
|
|
|
|
|
|
|
|
kfree(wq_dev);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* workqueue_sysfs_register - make a workqueue visible in sysfs
|
|
|
|
* @wq: the workqueue to register
|
|
|
|
*
|
|
|
|
* Expose @wq in sysfs under /sys/bus/workqueue/devices.
|
|
|
|
* alloc_workqueue*() automatically calls this function if WQ_SYSFS is set
|
|
|
|
* which is the preferred method.
|
|
|
|
*
|
|
|
|
* Workqueue user should use this function directly iff it wants to apply
|
|
|
|
* workqueue_attrs before making the workqueue visible in sysfs; otherwise,
|
|
|
|
* apply_workqueue_attrs() may race against userland updating the
|
|
|
|
* attributes.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return: 0 on success, -errno on failure.
|
2013-03-12 12:30:05 -06:00
|
|
|
*/
|
|
|
|
int workqueue_sysfs_register(struct workqueue_struct *wq)
|
|
|
|
{
|
|
|
|
struct wq_device *wq_dev;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Adjusting max_active or creating new pwqs by applyting
|
|
|
|
* attributes breaks ordering guarantee. Disallow exposing ordered
|
|
|
|
* workqueues.
|
|
|
|
*/
|
|
|
|
if (WARN_ON(wq->flags & __WQ_ORDERED))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
wq->wq_dev = wq_dev = kzalloc(sizeof(*wq_dev), GFP_KERNEL);
|
|
|
|
if (!wq_dev)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
wq_dev->wq = wq;
|
|
|
|
wq_dev->dev.bus = &wq_subsys;
|
|
|
|
wq_dev->dev.init_name = wq->name;
|
|
|
|
wq_dev->dev.release = wq_device_release;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* unbound_attrs are created separately. Suppress uevent until
|
|
|
|
* everything is ready.
|
|
|
|
*/
|
|
|
|
dev_set_uevent_suppress(&wq_dev->dev, true);
|
|
|
|
|
|
|
|
ret = device_register(&wq_dev->dev);
|
|
|
|
if (ret) {
|
|
|
|
kfree(wq_dev);
|
|
|
|
wq->wq_dev = NULL;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (wq->flags & WQ_UNBOUND) {
|
|
|
|
struct device_attribute *attr;
|
|
|
|
|
|
|
|
for (attr = wq_sysfs_unbound_attrs; attr->attr.name; attr++) {
|
|
|
|
ret = device_create_file(&wq_dev->dev, attr);
|
|
|
|
if (ret) {
|
|
|
|
device_unregister(&wq_dev->dev);
|
|
|
|
wq->wq_dev = NULL;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
kobject_uevent(&wq_dev->dev.kobj, KOBJ_ADD);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* workqueue_sysfs_unregister - undo workqueue_sysfs_register()
|
|
|
|
* @wq: the workqueue to unregister
|
|
|
|
*
|
|
|
|
* If @wq is registered to sysfs by workqueue_sysfs_register(), unregister.
|
|
|
|
*/
|
|
|
|
static void workqueue_sysfs_unregister(struct workqueue_struct *wq)
|
|
|
|
{
|
|
|
|
struct wq_device *wq_dev = wq->wq_dev;
|
|
|
|
|
|
|
|
if (!wq->wq_dev)
|
|
|
|
return;
|
|
|
|
|
|
|
|
wq->wq_dev = NULL;
|
|
|
|
device_unregister(&wq_dev->dev);
|
|
|
|
}
|
|
|
|
#else /* CONFIG_SYSFS */
|
|
|
|
static void workqueue_sysfs_unregister(struct workqueue_struct *wq) { }
|
|
|
|
#endif /* CONFIG_SYSFS */
|
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
/**
|
|
|
|
* free_workqueue_attrs - free a workqueue_attrs
|
|
|
|
* @attrs: workqueue_attrs to free
|
|
|
|
*
|
|
|
|
* Undo alloc_workqueue_attrs().
|
|
|
|
*/
|
|
|
|
void free_workqueue_attrs(struct workqueue_attrs *attrs)
|
|
|
|
{
|
|
|
|
if (attrs) {
|
|
|
|
free_cpumask_var(attrs->cpumask);
|
|
|
|
kfree(attrs);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* alloc_workqueue_attrs - allocate a workqueue_attrs
|
|
|
|
* @gfp_mask: allocation mask to use
|
|
|
|
*
|
|
|
|
* Allocate a new workqueue_attrs, initialize with default settings and
|
2013-07-31 15:59:24 -06:00
|
|
|
* return it.
|
|
|
|
*
|
|
|
|
* Return: The allocated new workqueue_attr on success. %NULL on failure.
|
2013-03-12 12:30:00 -06:00
|
|
|
*/
|
|
|
|
struct workqueue_attrs *alloc_workqueue_attrs(gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
struct workqueue_attrs *attrs;
|
|
|
|
|
|
|
|
attrs = kzalloc(sizeof(*attrs), gfp_mask);
|
|
|
|
if (!attrs)
|
|
|
|
goto fail;
|
|
|
|
if (!alloc_cpumask_var(&attrs->cpumask, gfp_mask))
|
|
|
|
goto fail;
|
|
|
|
|
2013-04-01 12:23:31 -06:00
|
|
|
cpumask_copy(attrs->cpumask, cpu_possible_mask);
|
2013-03-12 12:30:00 -06:00
|
|
|
return attrs;
|
|
|
|
fail:
|
|
|
|
free_workqueue_attrs(attrs);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
static void copy_workqueue_attrs(struct workqueue_attrs *to,
|
|
|
|
const struct workqueue_attrs *from)
|
|
|
|
{
|
|
|
|
to->nice = from->nice;
|
|
|
|
cpumask_copy(to->cpumask, from->cpumask);
|
2013-07-31 19:56:36 -06:00
|
|
|
/*
|
|
|
|
* Unlike hash and equality test, this function doesn't ignore
|
|
|
|
* ->no_numa as it is used for both pool and wq attrs. Instead,
|
|
|
|
* get_unbound_pool() explicitly clears ->no_numa after copying.
|
|
|
|
*/
|
|
|
|
to->no_numa = from->no_numa;
|
2013-03-12 12:30:03 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/* hash value of the content of @attr */
|
|
|
|
static u32 wqattrs_hash(const struct workqueue_attrs *attrs)
|
|
|
|
{
|
|
|
|
u32 hash = 0;
|
|
|
|
|
|
|
|
hash = jhash_1word(attrs->nice, hash);
|
2013-04-01 12:23:31 -06:00
|
|
|
hash = jhash(cpumask_bits(attrs->cpumask),
|
|
|
|
BITS_TO_LONGS(nr_cpumask_bits) * sizeof(long), hash);
|
2013-03-12 12:30:03 -06:00
|
|
|
return hash;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* content equality test */
|
|
|
|
static bool wqattrs_equal(const struct workqueue_attrs *a,
|
|
|
|
const struct workqueue_attrs *b)
|
|
|
|
{
|
|
|
|
if (a->nice != b->nice)
|
|
|
|
return false;
|
|
|
|
if (!cpumask_equal(a->cpumask, b->cpumask))
|
|
|
|
return false;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
/**
|
|
|
|
* init_worker_pool - initialize a newly zalloc'd worker_pool
|
|
|
|
* @pool: worker_pool to initialize
|
|
|
|
*
|
|
|
|
* Initiailize a newly zalloc'd @pool. It also allocates @pool->attrs.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: 0 on success, -errno on failure. Even on failure, all fields
|
2013-03-12 12:30:03 -06:00
|
|
|
* inside @pool proper are initialized and put_unbound_pool() can be called
|
|
|
|
* on @pool safely to release it.
|
2013-03-12 12:30:00 -06:00
|
|
|
*/
|
|
|
|
static int init_worker_pool(struct worker_pool *pool)
|
2013-03-12 12:30:00 -06:00
|
|
|
{
|
|
|
|
spin_lock_init(&pool->lock);
|
2013-03-12 12:30:03 -06:00
|
|
|
pool->id = -1;
|
|
|
|
pool->cpu = -1;
|
2013-04-01 12:23:34 -06:00
|
|
|
pool->node = NUMA_NO_NODE;
|
2013-03-12 12:30:00 -06:00
|
|
|
pool->flags |= POOL_DISASSOCIATED;
|
|
|
|
INIT_LIST_HEAD(&pool->worklist);
|
|
|
|
INIT_LIST_HEAD(&pool->idle_list);
|
|
|
|
hash_init(pool->busy_hash);
|
|
|
|
|
|
|
|
init_timer_deferrable(&pool->idle_timer);
|
|
|
|
pool->idle_timer.function = idle_worker_timeout;
|
|
|
|
pool->idle_timer.data = (unsigned long)pool;
|
|
|
|
|
|
|
|
setup_timer(&pool->mayday_timer, pool_mayday_timeout,
|
|
|
|
(unsigned long)pool);
|
|
|
|
|
|
|
|
mutex_init(&pool->manager_arb);
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_init(&pool->manager_mutex);
|
2013-03-19 14:45:21 -06:00
|
|
|
idr_init(&pool->worker_idr);
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
INIT_HLIST_NODE(&pool->hash_node);
|
|
|
|
pool->refcnt = 1;
|
|
|
|
|
|
|
|
/* shouldn't fail above this point */
|
2013-03-12 12:30:00 -06:00
|
|
|
pool->attrs = alloc_workqueue_attrs(GFP_KERNEL);
|
|
|
|
if (!pool->attrs)
|
|
|
|
return -ENOMEM;
|
|
|
|
return 0;
|
2013-03-12 12:30:00 -06:00
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
static void rcu_free_pool(struct rcu_head *rcu)
|
|
|
|
{
|
|
|
|
struct worker_pool *pool = container_of(rcu, struct worker_pool, rcu);
|
|
|
|
|
2013-03-19 14:45:21 -06:00
|
|
|
idr_destroy(&pool->worker_idr);
|
2013-03-12 12:30:03 -06:00
|
|
|
free_workqueue_attrs(pool->attrs);
|
|
|
|
kfree(pool);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* put_unbound_pool - put a worker_pool
|
|
|
|
* @pool: worker_pool to put
|
|
|
|
*
|
|
|
|
* Put @pool. If its refcnt reaches zero, it gets destroyed in sched-RCU
|
2013-03-13 17:51:36 -06:00
|
|
|
* safe manner. get_unbound_pool() calls this function on its failure path
|
|
|
|
* and this function should be able to release pools which went through,
|
|
|
|
* successfully or not, init_worker_pool().
|
2013-04-01 12:23:32 -06:00
|
|
|
*
|
|
|
|
* Should be called with wq_pool_mutex held.
|
2013-03-12 12:30:03 -06:00
|
|
|
*/
|
|
|
|
static void put_unbound_pool(struct worker_pool *pool)
|
|
|
|
{
|
|
|
|
struct worker *worker;
|
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
lockdep_assert_held(&wq_pool_mutex);
|
|
|
|
|
|
|
|
if (--pool->refcnt)
|
2013-03-12 12:30:03 -06:00
|
|
|
return;
|
|
|
|
|
|
|
|
/* sanity checks */
|
|
|
|
if (WARN_ON(!(pool->flags & POOL_DISASSOCIATED)) ||
|
2013-04-01 12:23:32 -06:00
|
|
|
WARN_ON(!list_empty(&pool->worklist)))
|
2013-03-12 12:30:03 -06:00
|
|
|
return;
|
|
|
|
|
|
|
|
/* release id and unhash */
|
|
|
|
if (pool->id >= 0)
|
|
|
|
idr_remove(&worker_pool_idr, pool->id);
|
|
|
|
hash_del(&pool->hash_node);
|
|
|
|
|
2013-03-13 17:51:36 -06:00
|
|
|
/*
|
|
|
|
* Become the manager and destroy all workers. Grabbing
|
|
|
|
* manager_arb prevents @pool's workers from blocking on
|
|
|
|
* manager_mutex.
|
|
|
|
*/
|
2013-03-12 12:30:03 -06:00
|
|
|
mutex_lock(&pool->manager_arb);
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_lock(&pool->manager_mutex);
|
2013-03-12 12:30:03 -06:00
|
|
|
spin_lock_irq(&pool->lock);
|
|
|
|
|
|
|
|
while ((worker = first_worker(pool)))
|
|
|
|
destroy_worker(worker);
|
|
|
|
WARN_ON(pool->nr_workers || pool->nr_idle);
|
|
|
|
|
|
|
|
spin_unlock_irq(&pool->lock);
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_unlock(&pool->manager_mutex);
|
2013-03-12 12:30:03 -06:00
|
|
|
mutex_unlock(&pool->manager_arb);
|
|
|
|
|
|
|
|
/* shut down the timers */
|
|
|
|
del_timer_sync(&pool->idle_timer);
|
|
|
|
del_timer_sync(&pool->mayday_timer);
|
|
|
|
|
|
|
|
/* sched-RCU protected to allow dereferences from get_work_pool() */
|
|
|
|
call_rcu_sched(&pool->rcu, rcu_free_pool);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* get_unbound_pool - get a worker_pool with the specified attributes
|
|
|
|
* @attrs: the attributes of the worker_pool to get
|
|
|
|
*
|
|
|
|
* Obtain a worker_pool which has the same attributes as @attrs, bump the
|
|
|
|
* reference count and return it. If there already is a matching
|
|
|
|
* worker_pool, it will be used; otherwise, this function attempts to
|
2013-07-31 15:59:24 -06:00
|
|
|
* create a new one.
|
2013-04-01 12:23:32 -06:00
|
|
|
*
|
|
|
|
* Should be called with wq_pool_mutex held.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: On success, a worker_pool with the same attributes as @attrs.
|
|
|
|
* On failure, %NULL.
|
2013-03-12 12:30:03 -06:00
|
|
|
*/
|
|
|
|
static struct worker_pool *get_unbound_pool(const struct workqueue_attrs *attrs)
|
|
|
|
{
|
|
|
|
u32 hash = wqattrs_hash(attrs);
|
|
|
|
struct worker_pool *pool;
|
2013-04-01 12:23:34 -06:00
|
|
|
int node;
|
2013-03-12 12:30:03 -06:00
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
lockdep_assert_held(&wq_pool_mutex);
|
2013-03-12 12:30:03 -06:00
|
|
|
|
|
|
|
/* do we already have a matching pool? */
|
|
|
|
hash_for_each_possible(unbound_pool_hash, pool, hash_node, hash) {
|
|
|
|
if (wqattrs_equal(pool->attrs, attrs)) {
|
|
|
|
pool->refcnt++;
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* nope, create a new one */
|
|
|
|
pool = kzalloc(sizeof(*pool), GFP_KERNEL);
|
|
|
|
if (!pool || init_worker_pool(pool) < 0)
|
|
|
|
goto fail;
|
|
|
|
|
2013-03-19 13:28:01 -06:00
|
|
|
if (workqueue_freezing)
|
|
|
|
pool->flags |= POOL_FREEZING;
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
lockdep_set_subclass(&pool->lock, 1); /* see put_pwq() */
|
2013-03-12 12:30:03 -06:00
|
|
|
copy_workqueue_attrs(pool->attrs, attrs);
|
|
|
|
|
2013-07-31 19:56:36 -06:00
|
|
|
/*
|
|
|
|
* no_numa isn't a worker_pool attribute, always clear it. See
|
|
|
|
* 'struct workqueue_attrs' comments for detail.
|
|
|
|
*/
|
|
|
|
pool->attrs->no_numa = false;
|
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
/* if cpumask is contained inside a NUMA node, we belong to that node */
|
|
|
|
if (wq_numa_enabled) {
|
|
|
|
for_each_node(node) {
|
|
|
|
if (cpumask_subset(pool->attrs->cpumask,
|
|
|
|
wq_numa_possible_cpumask[node])) {
|
|
|
|
pool->node = node;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
if (worker_pool_assign_id(pool) < 0)
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
/* create and start the initial worker */
|
2013-03-13 20:47:39 -06:00
|
|
|
if (create_and_start_worker(pool) < 0)
|
2013-03-12 12:30:03 -06:00
|
|
|
goto fail;
|
|
|
|
|
|
|
|
/* install */
|
|
|
|
hash_add(unbound_pool_hash, &pool->hash_node, hash);
|
|
|
|
out_unlock:
|
|
|
|
return pool;
|
|
|
|
fail:
|
|
|
|
if (pool)
|
|
|
|
put_unbound_pool(pool);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
static void rcu_free_pwq(struct rcu_head *rcu)
|
|
|
|
{
|
|
|
|
kmem_cache_free(pwq_cache,
|
|
|
|
container_of(rcu, struct pool_workqueue, rcu));
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Scheduled on system_wq by put_pwq() when an unbound pwq hits zero refcnt
|
|
|
|
* and needs to be destroyed.
|
|
|
|
*/
|
|
|
|
static void pwq_unbound_release_workfn(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct pool_workqueue *pwq = container_of(work, struct pool_workqueue,
|
|
|
|
unbound_release_work);
|
|
|
|
struct workqueue_struct *wq = pwq->wq;
|
|
|
|
struct worker_pool *pool = pwq->pool;
|
2013-04-01 12:23:31 -06:00
|
|
|
bool is_last;
|
2013-03-12 12:30:04 -06:00
|
|
|
|
|
|
|
if (WARN_ON_ONCE(!(wq->flags & WQ_UNBOUND)))
|
|
|
|
return;
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/*
|
2013-03-25 17:57:17 -06:00
|
|
|
* Unlink @pwq. Synchronization against wq->mutex isn't strictly
|
2013-03-12 12:30:04 -06:00
|
|
|
* necessary on release but do it anyway. It's easier to verify
|
|
|
|
* and consistent with the linking path.
|
|
|
|
*/
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2013-03-12 12:30:04 -06:00
|
|
|
list_del_rcu(&pwq->pwqs_node);
|
2013-04-01 12:23:31 -06:00
|
|
|
is_last = list_empty(&wq->pwqs);
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:30:04 -06:00
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
2013-03-12 12:30:04 -06:00
|
|
|
put_unbound_pool(pool);
|
2013-04-01 12:23:32 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
call_rcu_sched(&pwq->rcu, rcu_free_pwq);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If we're the last pwq going away, @wq is already dead and no one
|
|
|
|
* is gonna access it anymore. Free it.
|
|
|
|
*/
|
2013-04-01 12:23:34 -06:00
|
|
|
if (is_last) {
|
|
|
|
free_workqueue_attrs(wq->unbound_attrs);
|
2013-03-12 12:30:04 -06:00
|
|
|
kfree(wq);
|
2013-04-01 12:23:34 -06:00
|
|
|
}
|
2013-03-12 12:30:04 -06:00
|
|
|
}
|
|
|
|
|
2013-03-13 17:51:35 -06:00
|
|
|
/**
|
2013-03-13 17:51:35 -06:00
|
|
|
* pwq_adjust_max_active - update a pwq's max_active to the current setting
|
2013-03-13 17:51:35 -06:00
|
|
|
* @pwq: target pool_workqueue
|
|
|
|
*
|
2013-03-13 17:51:35 -06:00
|
|
|
* If @pwq isn't freezing, set @pwq->max_active to the associated
|
|
|
|
* workqueue's saved_max_active and activate delayed work items
|
|
|
|
* accordingly. If @pwq is freezing, clear @pwq->max_active to zero.
|
2013-03-13 17:51:35 -06:00
|
|
|
*/
|
2013-03-13 17:51:35 -06:00
|
|
|
static void pwq_adjust_max_active(struct pool_workqueue *pwq)
|
2013-03-13 17:51:35 -06:00
|
|
|
{
|
2013-03-13 17:51:35 -06:00
|
|
|
struct workqueue_struct *wq = pwq->wq;
|
|
|
|
bool freezable = wq->flags & WQ_FREEZABLE;
|
|
|
|
|
|
|
|
/* for @wq->saved_max_active */
|
2013-03-25 17:57:19 -06:00
|
|
|
lockdep_assert_held(&wq->mutex);
|
2013-03-13 17:51:35 -06:00
|
|
|
|
|
|
|
/* fast exit for non-freezable wqs */
|
|
|
|
if (!freezable && pwq->max_active == wq->saved_max_active)
|
|
|
|
return;
|
|
|
|
|
2013-03-25 17:57:19 -06:00
|
|
|
spin_lock_irq(&pwq->pool->lock);
|
2013-03-13 17:51:35 -06:00
|
|
|
|
|
|
|
if (!freezable || !(pwq->pool->flags & POOL_FREEZING)) {
|
|
|
|
pwq->max_active = wq->saved_max_active;
|
2013-03-13 17:51:35 -06:00
|
|
|
|
2013-03-13 17:51:35 -06:00
|
|
|
while (!list_empty(&pwq->delayed_works) &&
|
|
|
|
pwq->nr_active < pwq->max_active)
|
|
|
|
pwq_activate_first_delayed(pwq);
|
2013-03-20 11:52:30 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Need to kick a worker after thawed or an unbound wq's
|
|
|
|
* max_active is bumped. It's a slow path. Do it always.
|
|
|
|
*/
|
|
|
|
wake_up_worker(pwq->pool);
|
2013-03-13 17:51:35 -06:00
|
|
|
} else {
|
|
|
|
pwq->max_active = 0;
|
|
|
|
}
|
|
|
|
|
2013-03-25 17:57:19 -06:00
|
|
|
spin_unlock_irq(&pwq->pool->lock);
|
2013-03-13 17:51:35 -06:00
|
|
|
}
|
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
/* initialize newly alloced @pwq which is associated with @wq and @pool */
|
2013-04-01 12:23:35 -06:00
|
|
|
static void init_pwq(struct pool_workqueue *pwq, struct workqueue_struct *wq,
|
|
|
|
struct worker_pool *pool)
|
2013-03-12 12:30:04 -06:00
|
|
|
{
|
|
|
|
BUG_ON((unsigned long)pwq & WORK_STRUCT_FLAG_MASK);
|
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
memset(pwq, 0, sizeof(*pwq));
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
pwq->pool = pool;
|
|
|
|
pwq->wq = wq;
|
|
|
|
pwq->flush_color = -1;
|
2013-03-12 12:30:04 -06:00
|
|
|
pwq->refcnt = 1;
|
2013-03-12 12:30:04 -06:00
|
|
|
INIT_LIST_HEAD(&pwq->delayed_works);
|
2013-04-01 12:23:35 -06:00
|
|
|
INIT_LIST_HEAD(&pwq->pwqs_node);
|
2013-03-12 12:30:04 -06:00
|
|
|
INIT_LIST_HEAD(&pwq->mayday_node);
|
2013-03-12 12:30:04 -06:00
|
|
|
INIT_WORK(&pwq->unbound_release_work, pwq_unbound_release_workfn);
|
2013-04-01 12:23:35 -06:00
|
|
|
}
|
2013-03-12 12:30:04 -06:00
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
/* sync @pwq with the current state of its associated wq and link it */
|
2013-04-01 12:23:35 -06:00
|
|
|
static void link_pwq(struct pool_workqueue *pwq)
|
2013-04-01 12:23:35 -06:00
|
|
|
{
|
|
|
|
struct workqueue_struct *wq = pwq->wq;
|
|
|
|
|
|
|
|
lockdep_assert_held(&wq->mutex);
|
2013-03-12 12:30:04 -06:00
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
/* may be called multiple times, ignore if already linked */
|
|
|
|
if (!list_empty(&pwq->pwqs_node))
|
|
|
|
return;
|
|
|
|
|
2013-03-13 17:51:35 -06:00
|
|
|
/*
|
|
|
|
* Set the matching work_color. This is synchronized with
|
2013-03-25 17:57:17 -06:00
|
|
|
* wq->mutex to avoid confusing flush_workqueue().
|
2013-03-13 17:51:35 -06:00
|
|
|
*/
|
2013-03-12 12:30:04 -06:00
|
|
|
pwq->work_color = wq->work_color;
|
2013-03-13 17:51:35 -06:00
|
|
|
|
|
|
|
/* sync max_active to the current setting */
|
|
|
|
pwq_adjust_max_active(pwq);
|
|
|
|
|
|
|
|
/* link in @pwq */
|
2013-03-12 12:30:04 -06:00
|
|
|
list_add_rcu(&pwq->pwqs_node, &wq->pwqs);
|
2013-04-01 12:23:35 -06:00
|
|
|
}
|
2013-03-25 17:57:19 -06:00
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
/* obtain a pool matching @attr and create a pwq associating the pool and @wq */
|
|
|
|
static struct pool_workqueue *alloc_unbound_pwq(struct workqueue_struct *wq,
|
|
|
|
const struct workqueue_attrs *attrs)
|
|
|
|
{
|
|
|
|
struct worker_pool *pool;
|
|
|
|
struct pool_workqueue *pwq;
|
|
|
|
|
|
|
|
lockdep_assert_held(&wq_pool_mutex);
|
|
|
|
|
|
|
|
pool = get_unbound_pool(attrs);
|
|
|
|
if (!pool)
|
|
|
|
return NULL;
|
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
pwq = kmem_cache_alloc_node(pwq_cache, GFP_KERNEL, pool->node);
|
2013-04-01 12:23:35 -06:00
|
|
|
if (!pwq) {
|
|
|
|
put_unbound_pool(pool);
|
|
|
|
return NULL;
|
2013-04-01 12:23:35 -06:00
|
|
|
}
|
2013-04-01 12:23:34 -06:00
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
init_pwq(pwq, wq, pool);
|
|
|
|
return pwq;
|
2013-03-12 12:30:04 -06:00
|
|
|
}
|
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/* undo alloc_unbound_pwq(), used only in the error path */
|
|
|
|
static void free_unbound_pwq(struct pool_workqueue *pwq)
|
|
|
|
{
|
|
|
|
lockdep_assert_held(&wq_pool_mutex);
|
|
|
|
|
|
|
|
if (pwq) {
|
|
|
|
put_unbound_pool(pwq->pool);
|
2013-04-09 00:29:11 -06:00
|
|
|
kmem_cache_free(pwq_cache, pwq);
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* wq_calc_node_mask - calculate a wq_attrs' cpumask for the specified node
|
|
|
|
* @attrs: the wq_attrs of interest
|
|
|
|
* @node: the target NUMA node
|
|
|
|
* @cpu_going_down: if >= 0, the CPU to consider as offline
|
|
|
|
* @cpumask: outarg, the resulting cpumask
|
|
|
|
*
|
|
|
|
* Calculate the cpumask a workqueue with @attrs should use on @node. If
|
|
|
|
* @cpu_going_down is >= 0, that cpu is considered offline during
|
2013-07-31 15:59:24 -06:00
|
|
|
* calculation. The result is stored in @cpumask.
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
*
|
|
|
|
* If NUMA affinity is not enabled, @attrs->cpumask is always used. If
|
|
|
|
* enabled and @node has online CPUs requested by @attrs, the returned
|
|
|
|
* cpumask is the intersection of the possible CPUs of @node and
|
|
|
|
* @attrs->cpumask.
|
|
|
|
*
|
|
|
|
* The caller is responsible for ensuring that the cpumask of @node stays
|
|
|
|
* stable.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: %true if the resulting @cpumask is different from @attrs->cpumask,
|
|
|
|
* %false if equal.
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
*/
|
|
|
|
static bool wq_calc_node_cpumask(const struct workqueue_attrs *attrs, int node,
|
|
|
|
int cpu_going_down, cpumask_t *cpumask)
|
|
|
|
{
|
2013-04-01 12:23:38 -06:00
|
|
|
if (!wq_numa_enabled || attrs->no_numa)
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
goto use_dfl;
|
|
|
|
|
|
|
|
/* does @node have any online CPUs @attrs wants? */
|
|
|
|
cpumask_and(cpumask, cpumask_of_node(node), attrs->cpumask);
|
|
|
|
if (cpu_going_down >= 0)
|
|
|
|
cpumask_clear_cpu(cpu_going_down, cpumask);
|
|
|
|
|
|
|
|
if (cpumask_empty(cpumask))
|
|
|
|
goto use_dfl;
|
|
|
|
|
|
|
|
/* yeap, return possible CPUs in @node that @attrs wants */
|
|
|
|
cpumask_and(cpumask, attrs->cpumask, wq_numa_possible_cpumask[node]);
|
|
|
|
return !cpumask_equal(cpumask, attrs->cpumask);
|
|
|
|
|
|
|
|
use_dfl:
|
|
|
|
cpumask_copy(cpumask, attrs->cpumask);
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
/* install @pwq into @wq's numa_pwq_tbl[] for @node and return the old pwq */
|
|
|
|
static struct pool_workqueue *numa_pwq_tbl_install(struct workqueue_struct *wq,
|
|
|
|
int node,
|
|
|
|
struct pool_workqueue *pwq)
|
|
|
|
{
|
|
|
|
struct pool_workqueue *old_pwq;
|
|
|
|
|
|
|
|
lockdep_assert_held(&wq->mutex);
|
|
|
|
|
|
|
|
/* link_pwq() can handle duplicate calls */
|
|
|
|
link_pwq(pwq);
|
|
|
|
|
|
|
|
old_pwq = rcu_access_pointer(wq->numa_pwq_tbl[node]);
|
|
|
|
rcu_assign_pointer(wq->numa_pwq_tbl[node], pwq);
|
|
|
|
return old_pwq;
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/**
|
|
|
|
* apply_workqueue_attrs - apply new workqueue_attrs to an unbound workqueue
|
|
|
|
* @wq: the target workqueue
|
|
|
|
* @attrs: the workqueue_attrs to apply, allocated with alloc_workqueue_attrs()
|
|
|
|
*
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
* Apply @attrs to an unbound workqueue @wq. Unless disabled, on NUMA
|
|
|
|
* machines, this function maps a separate pwq to each NUMA node with
|
|
|
|
* possibles CPUs in @attrs->cpumask so that work items are affine to the
|
|
|
|
* NUMA node it was issued on. Older pwqs are released as in-flight work
|
|
|
|
* items finish. Note that a work item which repeatedly requeues itself
|
|
|
|
* back-to-back will stay on its current pwq.
|
2013-03-12 12:30:04 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Performs GFP_KERNEL allocations.
|
|
|
|
*
|
|
|
|
* Return: 0 on success and -errno on failure.
|
2013-03-12 12:30:04 -06:00
|
|
|
*/
|
|
|
|
int apply_workqueue_attrs(struct workqueue_struct *wq,
|
|
|
|
const struct workqueue_attrs *attrs)
|
|
|
|
{
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
struct workqueue_attrs *new_attrs, *tmp_attrs;
|
|
|
|
struct pool_workqueue **pwq_tbl, *dfl_pwq;
|
2013-04-01 12:23:35 -06:00
|
|
|
int node, ret;
|
2013-03-12 12:30:04 -06:00
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/* only unbound workqueues can change attributes */
|
2013-03-12 12:30:04 -06:00
|
|
|
if (WARN_ON(!(wq->flags & WQ_UNBOUND)))
|
|
|
|
return -EINVAL;
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/* creating multiple pwqs breaks ordering guarantee */
|
|
|
|
if (WARN_ON((wq->flags & __WQ_ORDERED) && !list_empty(&wq->pwqs)))
|
|
|
|
return -EINVAL;
|
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
pwq_tbl = kzalloc(wq_numa_tbl_len * sizeof(pwq_tbl[0]), GFP_KERNEL);
|
2013-04-01 12:23:31 -06:00
|
|
|
new_attrs = alloc_workqueue_attrs(GFP_KERNEL);
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
tmp_attrs = alloc_workqueue_attrs(GFP_KERNEL);
|
|
|
|
if (!pwq_tbl || !new_attrs || !tmp_attrs)
|
2013-04-01 12:23:31 -06:00
|
|
|
goto enomem;
|
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/* make a copy of @attrs and sanitize it */
|
2013-04-01 12:23:31 -06:00
|
|
|
copy_workqueue_attrs(new_attrs, attrs);
|
|
|
|
cpumask_and(new_attrs->cpumask, new_attrs->cpumask, cpu_possible_mask);
|
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/*
|
|
|
|
* We may create multiple pwqs with differing cpumasks. Make a
|
|
|
|
* copy of @new_attrs which will be modified and used to obtain
|
|
|
|
* pools.
|
|
|
|
*/
|
|
|
|
copy_workqueue_attrs(tmp_attrs, new_attrs);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* CPUs should stay stable across pwq creations and installations.
|
|
|
|
* Pin CPUs, determine the target cpumask for each node and create
|
|
|
|
* pwqs accordingly.
|
|
|
|
*/
|
|
|
|
get_online_cpus();
|
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If something goes wrong during CPU up/down, we'll fall back to
|
|
|
|
* the default pwq covering whole @attrs->cpumask. Always create
|
|
|
|
* it even if we don't use it immediately.
|
|
|
|
*/
|
|
|
|
dfl_pwq = alloc_unbound_pwq(wq, new_attrs);
|
|
|
|
if (!dfl_pwq)
|
|
|
|
goto enomem_pwq;
|
|
|
|
|
|
|
|
for_each_node(node) {
|
|
|
|
if (wq_calc_node_cpumask(attrs, node, -1, tmp_attrs->cpumask)) {
|
|
|
|
pwq_tbl[node] = alloc_unbound_pwq(wq, tmp_attrs);
|
|
|
|
if (!pwq_tbl[node])
|
|
|
|
goto enomem_pwq;
|
|
|
|
} else {
|
|
|
|
dfl_pwq->refcnt++;
|
|
|
|
pwq_tbl[node] = dfl_pwq;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2013-03-12 12:30:04 -06:00
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/* all pwqs have been created successfully, let's install'em */
|
2013-04-01 12:23:35 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2013-04-01 12:23:32 -06:00
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
copy_workqueue_attrs(wq->unbound_attrs, new_attrs);
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
|
|
|
|
/* save the previous pwq and install the new one */
|
2013-04-01 12:23:35 -06:00
|
|
|
for_each_node(node)
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
pwq_tbl[node] = numa_pwq_tbl_install(wq, node, pwq_tbl[node]);
|
|
|
|
|
|
|
|
/* @dfl_pwq might not have been used, ensure it's linked */
|
|
|
|
link_pwq(dfl_pwq);
|
|
|
|
swap(wq->dfl_pwq, dfl_pwq);
|
2013-04-01 12:23:35 -06:00
|
|
|
|
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:30:04 -06:00
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/* put the old pwqs */
|
|
|
|
for_each_node(node)
|
|
|
|
put_pwq_unlocked(pwq_tbl[node]);
|
|
|
|
put_pwq_unlocked(dfl_pwq);
|
|
|
|
|
|
|
|
put_online_cpus();
|
2013-04-01 12:23:31 -06:00
|
|
|
ret = 0;
|
|
|
|
/* fall through */
|
|
|
|
out_free:
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
free_workqueue_attrs(tmp_attrs);
|
2013-04-01 12:23:31 -06:00
|
|
|
free_workqueue_attrs(new_attrs);
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
kfree(pwq_tbl);
|
2013-04-01 12:23:31 -06:00
|
|
|
return ret;
|
2013-04-01 12:23:31 -06:00
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
enomem_pwq:
|
|
|
|
free_unbound_pwq(dfl_pwq);
|
|
|
|
for_each_node(node)
|
|
|
|
if (pwq_tbl && pwq_tbl[node] != dfl_pwq)
|
|
|
|
free_unbound_pwq(pwq_tbl[node]);
|
|
|
|
mutex_unlock(&wq_pool_mutex);
|
|
|
|
put_online_cpus();
|
2013-04-01 12:23:31 -06:00
|
|
|
enomem:
|
2013-04-01 12:23:31 -06:00
|
|
|
ret = -ENOMEM;
|
|
|
|
goto out_free;
|
2013-03-12 12:30:04 -06:00
|
|
|
}
|
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/**
|
|
|
|
* wq_update_unbound_numa - update NUMA affinity of a wq for CPU hot[un]plug
|
|
|
|
* @wq: the target workqueue
|
|
|
|
* @cpu: the CPU coming up or going down
|
|
|
|
* @online: whether @cpu is coming up or going down
|
|
|
|
*
|
|
|
|
* This function is to be called from %CPU_DOWN_PREPARE, %CPU_ONLINE and
|
|
|
|
* %CPU_DOWN_FAILED. @cpu is being hot[un]plugged, update NUMA affinity of
|
|
|
|
* @wq accordingly.
|
|
|
|
*
|
|
|
|
* If NUMA affinity can't be adjusted due to memory allocation failure, it
|
|
|
|
* falls back to @wq->dfl_pwq which may not be optimal but is always
|
|
|
|
* correct.
|
|
|
|
*
|
|
|
|
* Note that when the last allowed CPU of a NUMA node goes offline for a
|
|
|
|
* workqueue with a cpumask spanning multiple nodes, the workers which were
|
|
|
|
* already executing the work items for the workqueue will lose their CPU
|
|
|
|
* affinity and may execute on any CPU. This is similar to how per-cpu
|
|
|
|
* workqueues behave on CPU_DOWN. If a workqueue user wants strict
|
|
|
|
* affinity, it's the user's responsibility to flush the work item from
|
|
|
|
* CPU_DOWN_PREPARE.
|
|
|
|
*/
|
|
|
|
static void wq_update_unbound_numa(struct workqueue_struct *wq, int cpu,
|
|
|
|
bool online)
|
|
|
|
{
|
|
|
|
int node = cpu_to_node(cpu);
|
|
|
|
int cpu_off = online ? -1 : cpu;
|
|
|
|
struct pool_workqueue *old_pwq = NULL, *pwq;
|
|
|
|
struct workqueue_attrs *target_attrs;
|
|
|
|
cpumask_t *cpumask;
|
|
|
|
|
|
|
|
lockdep_assert_held(&wq_pool_mutex);
|
|
|
|
|
|
|
|
if (!wq_numa_enabled || !(wq->flags & WQ_UNBOUND))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We don't wanna alloc/free wq_attrs for each wq for each CPU.
|
|
|
|
* Let's use a preallocated one. The following buf is protected by
|
|
|
|
* CPU hotplug exclusion.
|
|
|
|
*/
|
|
|
|
target_attrs = wq_update_unbound_numa_attrs_buf;
|
|
|
|
cpumask = target_attrs->cpumask;
|
|
|
|
|
|
|
|
mutex_lock(&wq->mutex);
|
2013-04-01 12:23:38 -06:00
|
|
|
if (wq->unbound_attrs->no_numa)
|
|
|
|
goto out_unlock;
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
|
|
|
|
copy_workqueue_attrs(target_attrs, wq->unbound_attrs);
|
|
|
|
pwq = unbound_pwq_by_node(wq, node);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Let's determine what needs to be done. If the target cpumask is
|
|
|
|
* different from wq's, we need to compare it to @pwq's and create
|
|
|
|
* a new one if they don't match. If the target cpumask equals
|
|
|
|
* wq's, the default pwq should be used. If @pwq is already the
|
|
|
|
* default one, nothing to do; otherwise, install the default one.
|
|
|
|
*/
|
|
|
|
if (wq_calc_node_cpumask(wq->unbound_attrs, node, cpu_off, cpumask)) {
|
|
|
|
if (cpumask_equal(cpumask, pwq->pool->attrs->cpumask))
|
|
|
|
goto out_unlock;
|
|
|
|
} else {
|
|
|
|
if (pwq == wq->dfl_pwq)
|
|
|
|
goto out_unlock;
|
|
|
|
else
|
|
|
|
goto use_dfl_pwq;
|
|
|
|
}
|
|
|
|
|
|
|
|
mutex_unlock(&wq->mutex);
|
|
|
|
|
|
|
|
/* create a new pwq */
|
|
|
|
pwq = alloc_unbound_pwq(wq, target_attrs);
|
|
|
|
if (!pwq) {
|
|
|
|
pr_warning("workqueue: allocation failed while updating NUMA affinity of \"%s\"\n",
|
|
|
|
wq->name);
|
2014-04-15 23:32:29 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
|
|
|
goto use_dfl_pwq;
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Install the new pwq. As this function is called only from CPU
|
|
|
|
* hotplug callbacks and applying a new attrs is wrapped with
|
|
|
|
* get/put_online_cpus(), @wq->unbound_attrs couldn't have changed
|
|
|
|
* inbetween.
|
|
|
|
*/
|
|
|
|
mutex_lock(&wq->mutex);
|
|
|
|
old_pwq = numa_pwq_tbl_install(wq, node, pwq);
|
|
|
|
goto out_unlock;
|
|
|
|
|
|
|
|
use_dfl_pwq:
|
|
|
|
spin_lock_irq(&wq->dfl_pwq->pool->lock);
|
|
|
|
get_pwq(wq->dfl_pwq);
|
|
|
|
spin_unlock_irq(&wq->dfl_pwq->pool->lock);
|
|
|
|
old_pwq = numa_pwq_tbl_install(wq, node, wq->dfl_pwq);
|
|
|
|
out_unlock:
|
|
|
|
mutex_unlock(&wq->mutex);
|
|
|
|
put_pwq_unlocked(old_pwq);
|
|
|
|
}
|
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
static int alloc_and_link_pwqs(struct workqueue_struct *wq)
|
2010-06-29 02:07:11 -06:00
|
|
|
{
|
2013-03-12 12:29:58 -06:00
|
|
|
bool highpri = wq->flags & WQ_HIGHPRI;
|
2013-09-05 10:30:04 -06:00
|
|
|
int cpu, ret;
|
2013-03-12 12:29:57 -06:00
|
|
|
|
|
|
|
if (!(wq->flags & WQ_UNBOUND)) {
|
2013-03-12 12:29:59 -06:00
|
|
|
wq->cpu_pwqs = alloc_percpu(struct pool_workqueue);
|
|
|
|
if (!wq->cpu_pwqs)
|
2013-03-12 12:29:57 -06:00
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
for_each_possible_cpu(cpu) {
|
2013-03-12 12:30:00 -06:00
|
|
|
struct pool_workqueue *pwq =
|
|
|
|
per_cpu_ptr(wq->cpu_pwqs, cpu);
|
2013-03-12 12:30:03 -06:00
|
|
|
struct worker_pool *cpu_pools =
|
2013-03-12 12:30:03 -06:00
|
|
|
per_cpu(cpu_worker_pools, cpu);
|
2010-07-02 02:03:51 -06:00
|
|
|
|
2013-04-01 12:23:35 -06:00
|
|
|
init_pwq(pwq, wq, &cpu_pools[highpri]);
|
|
|
|
|
|
|
|
mutex_lock(&wq->mutex);
|
2013-04-01 12:23:35 -06:00
|
|
|
link_pwq(pwq);
|
2013-04-01 12:23:35 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:29:57 -06:00
|
|
|
}
|
2013-03-12 12:30:04 -06:00
|
|
|
return 0;
|
2013-09-05 10:30:04 -06:00
|
|
|
} else if (wq->flags & __WQ_ORDERED) {
|
|
|
|
ret = apply_workqueue_attrs(wq, ordered_wq_attrs[highpri]);
|
|
|
|
/* there should only be single pwq for ordering guarantee */
|
|
|
|
WARN(!ret && (wq->pwqs.next != &wq->dfl_pwq->pwqs_node ||
|
|
|
|
wq->pwqs.prev != &wq->dfl_pwq->pwqs_node),
|
|
|
|
"ordering guarantee broken for workqueue %s\n", wq->name);
|
|
|
|
return ret;
|
2013-03-12 12:29:57 -06:00
|
|
|
} else {
|
2013-03-12 12:30:04 -06:00
|
|
|
return apply_workqueue_attrs(wq, unbound_std_wq_attrs[highpri]);
|
2013-03-12 12:29:57 -06:00
|
|
|
}
|
2010-06-29 02:07:11 -06:00
|
|
|
}
|
|
|
|
|
2010-07-02 02:03:51 -06:00
|
|
|
static int wq_clamp_max_active(int max_active, unsigned int flags,
|
|
|
|
const char *name)
|
2010-06-29 02:07:14 -06:00
|
|
|
{
|
2010-07-02 02:03:51 -06:00
|
|
|
int lim = flags & WQ_UNBOUND ? WQ_UNBOUND_MAX_ACTIVE : WQ_MAX_ACTIVE;
|
|
|
|
|
|
|
|
if (max_active < 1 || max_active > lim)
|
2012-08-18 15:52:42 -06:00
|
|
|
pr_warn("workqueue: max_active %d requested for %s is out of range, clamping between %d and %d\n",
|
|
|
|
max_active, name, 1, lim);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2010-07-02 02:03:51 -06:00
|
|
|
return clamp_val(max_active, 1, lim);
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
2012-01-10 16:11:35 -07:00
|
|
|
struct workqueue_struct *__alloc_workqueue_key(const char *fmt,
|
2010-06-29 02:07:14 -06:00
|
|
|
unsigned int flags,
|
|
|
|
int max_active,
|
|
|
|
struct lock_class_key *key,
|
2012-01-10 16:11:35 -07:00
|
|
|
const char *lock_name, ...)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-04-01 12:23:35 -06:00
|
|
|
size_t tbl_size = 0;
|
2013-04-01 12:23:34 -06:00
|
|
|
va_list args;
|
2005-04-16 16:20:36 -06:00
|
|
|
struct workqueue_struct *wq;
|
2013-03-12 12:29:58 -06:00
|
|
|
struct pool_workqueue *pwq;
|
2012-01-10 16:11:35 -07:00
|
|
|
|
2013-04-08 05:15:40 -06:00
|
|
|
/* see the comment above the definition of WQ_POWER_EFFICIENT */
|
|
|
|
if ((flags & WQ_POWER_EFFICIENT) && wq_power_efficient)
|
|
|
|
flags |= WQ_UNBOUND;
|
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
/* allocate wq and format name */
|
2013-04-01 12:23:35 -06:00
|
|
|
if (flags & WQ_UNBOUND)
|
|
|
|
tbl_size = wq_numa_tbl_len * sizeof(wq->numa_pwq_tbl[0]);
|
|
|
|
|
|
|
|
wq = kzalloc(sizeof(*wq) + tbl_size, GFP_KERNEL);
|
2012-01-10 16:11:35 -07:00
|
|
|
if (!wq)
|
2013-03-12 12:30:04 -06:00
|
|
|
return NULL;
|
2012-01-10 16:11:35 -07:00
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
if (flags & WQ_UNBOUND) {
|
|
|
|
wq->unbound_attrs = alloc_workqueue_attrs(GFP_KERNEL);
|
|
|
|
if (!wq->unbound_attrs)
|
|
|
|
goto err_free_wq;
|
|
|
|
}
|
|
|
|
|
2013-04-01 12:23:34 -06:00
|
|
|
va_start(args, lock_name);
|
|
|
|
vsnprintf(wq->name, sizeof(wq->name), fmt, args);
|
2012-01-10 16:11:35 -07:00
|
|
|
va_end(args);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
max_active = max_active ?: WQ_DFL_ACTIVE;
|
2012-01-10 16:11:35 -07:00
|
|
|
max_active = wq_clamp_max_active(max_active, flags, wq->name);
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2012-01-10 16:11:35 -07:00
|
|
|
/* init wq */
|
2010-06-29 02:07:10 -06:00
|
|
|
wq->flags = flags;
|
2010-06-29 02:07:12 -06:00
|
|
|
wq->saved_max_active = max_active;
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_init(&wq->mutex);
|
2013-02-13 20:29:12 -07:00
|
|
|
atomic_set(&wq->nr_pwqs_to_flush, 0);
|
2013-03-12 12:29:57 -06:00
|
|
|
INIT_LIST_HEAD(&wq->pwqs);
|
2010-06-29 02:07:11 -06:00
|
|
|
INIT_LIST_HEAD(&wq->flusher_queue);
|
|
|
|
INIT_LIST_HEAD(&wq->flusher_overflow);
|
2013-03-12 12:29:59 -06:00
|
|
|
INIT_LIST_HEAD(&wq->maydays);
|
2010-06-29 02:07:13 -06:00
|
|
|
|
2008-01-16 01:51:58 -07:00
|
|
|
lockdep_init_map(&wq->lockdep_map, lock_name, key, 0);
|
2007-05-09 03:34:13 -06:00
|
|
|
INIT_LIST_HEAD(&wq->list);
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
if (alloc_and_link_pwqs(wq) < 0)
|
2013-03-12 12:30:04 -06:00
|
|
|
goto err_free_wq;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
/*
|
|
|
|
* Workqueues which may be used during memory reclaim should
|
|
|
|
* have a rescuer to guarantee forward progress.
|
|
|
|
*/
|
|
|
|
if (flags & WQ_MEM_RECLAIM) {
|
2010-06-29 02:07:14 -06:00
|
|
|
struct worker *rescuer;
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
rescuer = alloc_worker();
|
2010-06-29 02:07:14 -06:00
|
|
|
if (!rescuer)
|
2013-03-12 12:30:04 -06:00
|
|
|
goto err_destroy;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-01-17 18:16:24 -07:00
|
|
|
rescuer->rescue_wq = wq;
|
|
|
|
rescuer->task = kthread_create(rescuer_thread, rescuer, "%s",
|
2012-01-10 16:11:35 -07:00
|
|
|
wq->name);
|
2013-03-12 12:30:04 -06:00
|
|
|
if (IS_ERR(rescuer->task)) {
|
|
|
|
kfree(rescuer);
|
|
|
|
goto err_destroy;
|
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
wq->rescuer = rescuer;
|
2013-03-19 14:45:20 -06:00
|
|
|
rescuer->task->flags |= PF_NO_SETAFFINITY;
|
2010-06-29 02:07:14 -06:00
|
|
|
wake_up_process(rescuer->task);
|
2007-05-09 03:34:09 -06:00
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:05 -06:00
|
|
|
if ((wq->flags & WQ_SYSFS) && workqueue_sysfs_register(wq))
|
|
|
|
goto err_destroy;
|
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/*
|
2013-03-25 17:57:17 -06:00
|
|
|
* wq_pool_mutex protects global freeze state and workqueues list.
|
|
|
|
* Grab it, adjust max_active and add the new @wq to workqueues
|
|
|
|
* list.
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2013-03-13 17:51:35 -06:00
|
|
|
for_each_pwq(pwq, wq)
|
|
|
|
pwq_adjust_max_active(pwq);
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2010-06-29 02:07:11 -06:00
|
|
|
list_add(&wq->list, &workqueues);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2007-05-09 03:34:09 -06:00
|
|
|
return wq;
|
2013-03-12 12:30:04 -06:00
|
|
|
|
|
|
|
err_free_wq:
|
2013-04-01 12:23:34 -06:00
|
|
|
free_workqueue_attrs(wq->unbound_attrs);
|
2013-03-12 12:30:04 -06:00
|
|
|
kfree(wq);
|
|
|
|
return NULL;
|
|
|
|
err_destroy:
|
|
|
|
destroy_workqueue(wq);
|
2010-06-29 02:07:10 -06:00
|
|
|
return NULL;
|
2007-05-09 03:34:09 -06:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
EXPORT_SYMBOL_GPL(__alloc_workqueue_key);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2007-05-09 03:34:09 -06:00
|
|
|
/**
|
|
|
|
* destroy_workqueue - safely terminate a workqueue
|
|
|
|
* @wq: target workqueue
|
|
|
|
*
|
|
|
|
* Safely destroy a workqueue. All work currently pending will be done first.
|
|
|
|
*/
|
|
|
|
void destroy_workqueue(struct workqueue_struct *wq)
|
|
|
|
{
|
2013-03-12 12:29:58 -06:00
|
|
|
struct pool_workqueue *pwq;
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
int node;
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2011-04-05 10:01:44 -06:00
|
|
|
/* drain it before proceeding with destruction */
|
|
|
|
drain_workqueue(wq);
|
2010-12-20 11:32:04 -07:00
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
/* sanity checks */
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2013-03-12 12:29:58 -06:00
|
|
|
for_each_pwq(pwq, wq) {
|
2013-03-12 12:29:57 -06:00
|
|
|
int i;
|
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
for (i = 0; i < WORK_NR_COLORS; i++) {
|
|
|
|
if (WARN_ON(pwq->nr_in_flight[i])) {
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:29:57 -06:00
|
|
|
return;
|
2013-03-12 12:30:00 -06:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-04-03 20:05:38 -06:00
|
|
|
if (WARN_ON((pwq != wq->dfl_pwq) && (pwq->refcnt > 1)) ||
|
2013-03-12 12:30:04 -06:00
|
|
|
WARN_ON(pwq->nr_active) ||
|
2013-03-12 12:30:00 -06:00
|
|
|
WARN_ON(!list_empty(&pwq->delayed_works))) {
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:29:57 -06:00
|
|
|
return;
|
2013-03-12 12:30:00 -06:00
|
|
|
}
|
2013-03-12 12:29:57 -06:00
|
|
|
}
|
2013-03-25 17:57:18 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2013-03-12 12:29:57 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/*
|
|
|
|
* wq list is used to freeze wq, remove from list after
|
|
|
|
* flushing is complete in case freeze races us.
|
|
|
|
*/
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
2013-03-12 12:30:04 -06:00
|
|
|
list_del_init(&wq->list);
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2013-03-12 12:30:05 -06:00
|
|
|
workqueue_sysfs_unregister(wq);
|
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
if (wq->rescuer) {
|
2010-06-29 02:07:14 -06:00
|
|
|
kthread_stop(wq->rescuer->task);
|
2010-08-16 01:54:28 -06:00
|
|
|
kfree(wq->rescuer);
|
2013-03-12 12:30:03 -06:00
|
|
|
wq->rescuer = NULL;
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
if (!(wq->flags & WQ_UNBOUND)) {
|
|
|
|
/*
|
|
|
|
* The base ref is never dropped on per-cpu pwqs. Directly
|
|
|
|
* free the pwqs and wq.
|
|
|
|
*/
|
|
|
|
free_percpu(wq->cpu_pwqs);
|
|
|
|
kfree(wq);
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* We're the sole accessor of @wq at this point. Directly
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
* access numa_pwq_tbl[] and dfl_pwq to put the base refs.
|
|
|
|
* @wq will be freed when the last pwq is released.
|
2013-03-12 12:30:04 -06:00
|
|
|
*/
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
for_each_node(node) {
|
|
|
|
pwq = rcu_access_pointer(wq->numa_pwq_tbl[node]);
|
|
|
|
RCU_INIT_POINTER(wq->numa_pwq_tbl[node], NULL);
|
|
|
|
put_pwq_unlocked(pwq);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Put dfl_pwq. @wq may be freed any time after dfl_pwq is
|
|
|
|
* put. Don't access it afterwards.
|
|
|
|
*/
|
|
|
|
pwq = wq->dfl_pwq;
|
|
|
|
wq->dfl_pwq = NULL;
|
2013-04-01 12:23:35 -06:00
|
|
|
put_pwq_unlocked(pwq);
|
2013-03-12 12:30:03 -06:00
|
|
|
}
|
2007-05-09 03:34:09 -06:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(destroy_workqueue);
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/**
|
|
|
|
* workqueue_set_max_active - adjust max_active of a workqueue
|
|
|
|
* @wq: target workqueue
|
|
|
|
* @max_active: new max_active value.
|
|
|
|
*
|
|
|
|
* Set max_active of @wq to @max_active.
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
|
|
|
* Don't call from IRQ context.
|
|
|
|
*/
|
|
|
|
void workqueue_set_max_active(struct workqueue_struct *wq, int max_active)
|
|
|
|
{
|
2013-03-12 12:29:58 -06:00
|
|
|
struct pool_workqueue *pwq;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-12 12:30:04 -06:00
|
|
|
/* disallow meddling with max_active for ordered workqueues */
|
|
|
|
if (WARN_ON(wq->flags & __WQ_ORDERED))
|
|
|
|
return;
|
|
|
|
|
2010-07-02 02:03:51 -06:00
|
|
|
max_active = wq_clamp_max_active(max_active, wq->flags, wq->name);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2010-06-29 02:07:14 -06:00
|
|
|
|
|
|
|
wq->saved_max_active = max_active;
|
|
|
|
|
2013-03-13 17:51:35 -06:00
|
|
|
for_each_pwq(pwq, wq)
|
|
|
|
pwq_adjust_max_active(pwq);
|
2009-11-17 15:06:20 -07:00
|
|
|
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2006-01-08 02:00:43 -07:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
EXPORT_SYMBOL_GPL(workqueue_set_max_active);
|
2006-01-08 02:00:43 -07:00
|
|
|
|
2013-03-12 18:41:37 -06:00
|
|
|
/**
|
|
|
|
* current_is_workqueue_rescuer - is %current workqueue rescuer?
|
|
|
|
*
|
|
|
|
* Determine whether %current is a workqueue rescuer. Can be used from
|
|
|
|
* work functions to determine whether it's being run off the rescuer task.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: %true if %current is a workqueue rescuer. %false otherwise.
|
2013-03-12 18:41:37 -06:00
|
|
|
*/
|
|
|
|
bool current_is_workqueue_rescuer(void)
|
|
|
|
{
|
|
|
|
struct worker *worker = current_wq_worker();
|
|
|
|
|
2013-03-19 13:28:03 -06:00
|
|
|
return worker && worker->rescue_wq;
|
2013-03-12 18:41:37 -06:00
|
|
|
}
|
|
|
|
|
2010-02-12 01:39:21 -07:00
|
|
|
/**
|
2010-06-29 02:07:14 -06:00
|
|
|
* workqueue_congested - test whether a workqueue is congested
|
|
|
|
* @cpu: CPU in question
|
|
|
|
* @wq: target workqueue
|
2010-02-12 01:39:21 -07:00
|
|
|
*
|
2010-06-29 02:07:14 -06:00
|
|
|
* Test whether @wq's cpu workqueue for @cpu is congested. There is
|
|
|
|
* no synchronization around this function and the test result is
|
|
|
|
* unreliable and only useful as advisory hints or for debugging.
|
2010-02-12 01:39:21 -07:00
|
|
|
*
|
2013-05-10 12:10:17 -06:00
|
|
|
* If @cpu is WORK_CPU_UNBOUND, the test is performed on the local CPU.
|
|
|
|
* Note that both per-cpu and unbound workqueues may be associated with
|
|
|
|
* multiple pool_workqueues which have separate congested states. A
|
|
|
|
* workqueue being congested on one CPU doesn't mean the workqueue is also
|
|
|
|
* contested on other CPUs / NUMA nodes.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-06-29 02:07:14 -06:00
|
|
|
* %true if congested, %false otherwise.
|
2010-02-12 01:39:21 -07:00
|
|
|
*/
|
2013-03-12 12:29:59 -06:00
|
|
|
bool workqueue_congested(int cpu, struct workqueue_struct *wq)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-03-12 12:30:00 -06:00
|
|
|
struct pool_workqueue *pwq;
|
2013-03-12 12:30:00 -06:00
|
|
|
bool ret;
|
|
|
|
|
2013-03-19 13:28:10 -06:00
|
|
|
rcu_read_lock_sched();
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-05-10 12:10:17 -06:00
|
|
|
if (cpu == WORK_CPU_UNBOUND)
|
|
|
|
cpu = smp_processor_id();
|
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
if (!(wq->flags & WQ_UNBOUND))
|
|
|
|
pwq = per_cpu_ptr(wq->cpu_pwqs, cpu);
|
|
|
|
else
|
2013-04-01 12:23:35 -06:00
|
|
|
pwq = unbound_pwq_by_node(wq, cpu_to_node(cpu));
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
ret = !list_empty(&pwq->delayed_works);
|
2013-03-19 13:28:10 -06:00
|
|
|
rcu_read_unlock_sched();
|
2013-03-12 12:30:00 -06:00
|
|
|
|
|
|
|
return ret;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
EXPORT_SYMBOL_GPL(workqueue_congested);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/**
|
|
|
|
* work_busy - test whether a work is currently pending or running
|
|
|
|
* @work: the work to be tested
|
|
|
|
*
|
|
|
|
* Test whether @work is currently pending or running. There is no
|
|
|
|
* synchronization around this function and the test result is
|
|
|
|
* unreliable and only useful as advisory hints or for debugging.
|
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2010-06-29 02:07:14 -06:00
|
|
|
* OR'd bitmask of WORK_BUSY_* bits.
|
|
|
|
*/
|
|
|
|
unsigned int work_busy(struct work_struct *work)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-03-12 12:30:00 -06:00
|
|
|
struct worker_pool *pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
unsigned long flags;
|
|
|
|
unsigned int ret = 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
if (work_pending(work))
|
|
|
|
ret |= WORK_BUSY_PENDING;
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
local_irq_save(flags);
|
|
|
|
pool = get_work_pool(work);
|
2013-02-06 19:04:53 -07:00
|
|
|
if (pool) {
|
2013-03-12 12:30:00 -06:00
|
|
|
spin_lock(&pool->lock);
|
2013-02-06 19:04:53 -07:00
|
|
|
if (find_worker_executing_work(pool, work))
|
|
|
|
ret |= WORK_BUSY_RUNNING;
|
2013-03-12 12:30:00 -06:00
|
|
|
spin_unlock(&pool->lock);
|
2013-02-06 19:04:53 -07:00
|
|
|
}
|
2013-03-12 12:30:00 -06:00
|
|
|
local_irq_restore(flags);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
return ret;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
EXPORT_SYMBOL_GPL(work_busy);
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-04-30 16:27:22 -06:00
|
|
|
/**
|
|
|
|
* set_worker_desc - set description for the current work item
|
|
|
|
* @fmt: printf-style format string
|
|
|
|
* @...: arguments for the format string
|
|
|
|
*
|
|
|
|
* This function can be called by a running work function to describe what
|
|
|
|
* the work item is about. If the worker task gets dumped, this
|
|
|
|
* information will be printed out together to help debugging. The
|
|
|
|
* description can be at most WORKER_DESC_LEN including the trailing '\0'.
|
|
|
|
*/
|
|
|
|
void set_worker_desc(const char *fmt, ...)
|
|
|
|
{
|
|
|
|
struct worker *worker = current_wq_worker();
|
|
|
|
va_list args;
|
|
|
|
|
|
|
|
if (worker) {
|
|
|
|
va_start(args, fmt);
|
|
|
|
vsnprintf(worker->desc, sizeof(worker->desc), fmt, args);
|
|
|
|
va_end(args);
|
|
|
|
worker->desc_valid = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* print_worker_info - print out worker information and description
|
|
|
|
* @log_lvl: the log level to use when printing
|
|
|
|
* @task: target task
|
|
|
|
*
|
|
|
|
* If @task is a worker and currently executing a work item, print out the
|
|
|
|
* name of the workqueue being serviced and worker description set with
|
|
|
|
* set_worker_desc() by the currently executing work item.
|
|
|
|
*
|
|
|
|
* This function can be safely called on any task as long as the
|
|
|
|
* task_struct itself is accessible. While safe, this function isn't
|
|
|
|
* synchronized and may print out mixups or garbages of limited length.
|
|
|
|
*/
|
|
|
|
void print_worker_info(const char *log_lvl, struct task_struct *task)
|
|
|
|
{
|
|
|
|
work_func_t *fn = NULL;
|
|
|
|
char name[WQ_NAME_LEN] = { };
|
|
|
|
char desc[WORKER_DESC_LEN] = { };
|
|
|
|
struct pool_workqueue *pwq = NULL;
|
|
|
|
struct workqueue_struct *wq = NULL;
|
|
|
|
bool desc_valid = false;
|
|
|
|
struct worker *worker;
|
|
|
|
|
|
|
|
if (!(task->flags & PF_WQ_WORKER))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This function is called without any synchronization and @task
|
|
|
|
* could be in any state. Be careful with dereferences.
|
|
|
|
*/
|
|
|
|
worker = probe_kthread_data(task);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Carefully copy the associated workqueue's workfn and name. Keep
|
|
|
|
* the original last '\0' in case the original contains garbage.
|
|
|
|
*/
|
|
|
|
probe_kernel_read(&fn, &worker->current_func, sizeof(fn));
|
|
|
|
probe_kernel_read(&pwq, &worker->current_pwq, sizeof(pwq));
|
|
|
|
probe_kernel_read(&wq, &pwq->wq, sizeof(wq));
|
|
|
|
probe_kernel_read(name, wq->name, sizeof(name) - 1);
|
|
|
|
|
|
|
|
/* copy worker description */
|
|
|
|
probe_kernel_read(&desc_valid, &worker->desc_valid, sizeof(desc_valid));
|
|
|
|
if (desc_valid)
|
|
|
|
probe_kernel_read(desc, worker->desc, sizeof(desc) - 1);
|
|
|
|
|
|
|
|
if (fn || name[0] || desc[0]) {
|
|
|
|
printk("%sWorkqueue: %s %pf", log_lvl, name, fn);
|
|
|
|
if (desc[0])
|
|
|
|
pr_cont(" (%s)", desc);
|
|
|
|
pr_cont("\n");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
/*
|
|
|
|
* CPU hotplug.
|
|
|
|
*
|
2010-06-29 02:07:14 -06:00
|
|
|
* There are two challenges in supporting CPU hotplug. Firstly, there
|
2013-02-13 20:29:12 -07:00
|
|
|
* are a lot of assumptions on strong associations among work, pwq and
|
2013-01-24 12:01:34 -07:00
|
|
|
* pool which make migrating pending and scheduled works very
|
2010-06-29 02:07:14 -06:00
|
|
|
* difficult to implement without impacting hot paths. Secondly,
|
2013-01-24 12:01:33 -07:00
|
|
|
* worker pools serve mix of short, long and very long running works making
|
2010-06-29 02:07:14 -06:00
|
|
|
* blocked draining impractical.
|
|
|
|
*
|
2013-01-24 12:01:33 -07:00
|
|
|
* This is solved by allowing the pools to be disassociated from the CPU
|
2012-07-17 13:39:27 -06:00
|
|
|
* running as an unbound one and allowing it to be reattached later if the
|
|
|
|
* cpu comes back online.
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
2005-04-16 16:20:36 -06:00
|
|
|
|
2013-01-24 12:01:34 -07:00
|
|
|
static void wq_unbind_fn(struct work_struct *work)
|
2007-05-09 03:34:09 -06:00
|
|
|
{
|
2013-01-24 12:01:34 -07:00
|
|
|
int cpu = smp_processor_id();
|
2012-07-13 23:16:44 -06:00
|
|
|
struct worker_pool *pool;
|
2010-06-29 02:07:12 -06:00
|
|
|
struct worker *worker;
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
int wi;
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
for_each_cpu_worker_pool(pool, cpu) {
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(cpu != smp_processor_id());
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_lock(&pool->manager_mutex);
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_lock_irq(&pool->lock);
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
/*
|
2013-03-13 20:47:39 -06:00
|
|
|
* We've blocked all manager operations. Make all workers
|
2013-01-24 12:01:33 -07:00
|
|
|
* unbound and set DISASSOCIATED. Before this, all workers
|
|
|
|
* except for the ones which are still executing works from
|
|
|
|
* before the last CPU down must be on the cpu. After
|
|
|
|
* this, they may become diasporas.
|
|
|
|
*/
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
for_each_pool_worker(worker, wi, pool)
|
2013-01-24 12:01:33 -07:00
|
|
|
worker->flags |= WORKER_UNBOUND;
|
2007-05-09 03:34:15 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
pool->flags |= POOL_DISASSOCIATED;
|
2012-07-17 13:39:26 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_unlock(&pool->manager_mutex);
|
2012-07-17 13:39:27 -06:00
|
|
|
|
2013-03-08 16:18:28 -07:00
|
|
|
/*
|
|
|
|
* Call schedule() so that we cross rq->lock and thus can
|
|
|
|
* guarantee sched callbacks see the %WORKER_UNBOUND flag.
|
|
|
|
* This is necessary as scheduler callbacks may be invoked
|
|
|
|
* from other cpus.
|
|
|
|
*/
|
|
|
|
schedule();
|
2007-05-09 03:34:15 -06:00
|
|
|
|
2013-03-08 16:18:28 -07:00
|
|
|
/*
|
|
|
|
* 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. This pool now behaves as an
|
|
|
|
* unbound (in terms of concurrency management) pool which
|
|
|
|
* are served by workers tied to the pool.
|
|
|
|
*/
|
2013-01-24 12:39:44 -07:00
|
|
|
atomic_set(&pool->nr_running, 0);
|
2013-03-08 16:18:28 -07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* With concurrency management just turned off, a busy
|
|
|
|
* worker blocking could lead to lengthy stalls. Kick off
|
|
|
|
* unbound chain execution of currently pending work items.
|
|
|
|
*/
|
|
|
|
spin_lock_irq(&pool->lock);
|
|
|
|
wake_up_worker(pool);
|
|
|
|
spin_unlock_irq(&pool->lock);
|
|
|
|
}
|
2007-05-09 03:34:09 -06:00
|
|
|
}
|
|
|
|
|
2013-03-19 14:45:21 -06:00
|
|
|
/**
|
|
|
|
* rebind_workers - rebind all workers of a pool to the associated CPU
|
|
|
|
* @pool: pool of interest
|
|
|
|
*
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
* @pool->cpu is coming online. Rebind all workers to the CPU.
|
2013-03-19 14:45:21 -06:00
|
|
|
*/
|
|
|
|
static void rebind_workers(struct worker_pool *pool)
|
|
|
|
{
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
struct worker *worker;
|
|
|
|
int wi;
|
2013-03-19 14:45:21 -06:00
|
|
|
|
|
|
|
lockdep_assert_held(&pool->manager_mutex);
|
|
|
|
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
/*
|
|
|
|
* Restore CPU affinity of all workers. As all idle workers should
|
|
|
|
* be on the run-queue of the associated CPU before any local
|
|
|
|
* wake-ups for concurrency management happen, restore CPU affinty
|
|
|
|
* of all workers first and then clear UNBOUND. As we're called
|
|
|
|
* from CPU_ONLINE, the following shouldn't fail.
|
|
|
|
*/
|
|
|
|
for_each_pool_worker(worker, wi, pool)
|
|
|
|
WARN_ON_ONCE(set_cpus_allowed_ptr(worker->task,
|
|
|
|
pool->attrs->cpumask) < 0);
|
2013-03-19 14:45:21 -06:00
|
|
|
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
spin_lock_irq(&pool->lock);
|
2013-03-19 14:45:21 -06:00
|
|
|
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
for_each_pool_worker(worker, wi, pool) {
|
|
|
|
unsigned int worker_flags = worker->flags;
|
2013-03-19 14:45:21 -06:00
|
|
|
|
|
|
|
/*
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
* A bound idle worker should actually be on the runqueue
|
|
|
|
* of the associated CPU for local wake-ups targeting it to
|
|
|
|
* work. Kick all idle workers so that they migrate to the
|
|
|
|
* associated CPU. Doing this in the same loop as
|
|
|
|
* replacing UNBOUND with REBOUND is safe as no worker will
|
|
|
|
* be bound before @pool->lock is released.
|
2013-03-19 14:45:21 -06:00
|
|
|
*/
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
if (worker_flags & WORKER_IDLE)
|
|
|
|
wake_up_process(worker->task);
|
2013-03-19 14:45:21 -06:00
|
|
|
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
/*
|
|
|
|
* We want to clear UNBOUND but can't directly call
|
|
|
|
* worker_clr_flags() or adjust nr_running. Atomically
|
|
|
|
* replace UNBOUND with another NOT_RUNNING flag REBOUND.
|
|
|
|
* @worker will clear REBOUND using worker_clr_flags() when
|
|
|
|
* it initiates the next execution cycle thus restoring
|
|
|
|
* concurrency management. Note that when or whether
|
|
|
|
* @worker clears REBOUND doesn't affect correctness.
|
|
|
|
*
|
|
|
|
* ACCESS_ONCE() is necessary because @worker->flags may be
|
|
|
|
* tested without holding any lock in
|
|
|
|
* wq_worker_waking_up(). Without it, NOT_RUNNING test may
|
|
|
|
* fail incorrectly leading to premature concurrency
|
|
|
|
* management operations.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(!(worker_flags & WORKER_UNBOUND));
|
|
|
|
worker_flags |= WORKER_REBOUND;
|
|
|
|
worker_flags &= ~WORKER_UNBOUND;
|
|
|
|
ACCESS_ONCE(worker->flags) = worker_flags;
|
2013-03-19 14:45:21 -06:00
|
|
|
}
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
|
|
|
|
spin_unlock_irq(&pool->lock);
|
2013-03-19 14:45:21 -06:00
|
|
|
}
|
|
|
|
|
workqueue: restore CPU affinity of unbound workers on CPU_ONLINE
With the recent addition of the custom attributes support, unbound
pools may have allowed cpumask which isn't full. As long as some of
CPUs in the cpumask are online, its workers will maintain cpus_allowed
as set on worker creation; however, once no online CPU is left in
cpus_allowed, the scheduler will reset cpus_allowed of any workers
which get scheduled so that they can execute.
To remain compliant to the user-specified configuration, CPU affinity
needs to be restored when a CPU becomes online for an unbound pool
which doesn't currently have any online CPUs before.
This patch implement restore_unbound_workers_cpumask(), which is
called from CPU_ONLINE for all unbound pools, checks whether the
coming up CPU is the first allowed online one, and, if so, invokes
set_cpus_allowed_ptr() with the configured cpumask on all workers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
/**
|
|
|
|
* restore_unbound_workers_cpumask - restore cpumask of unbound workers
|
|
|
|
* @pool: unbound pool of interest
|
|
|
|
* @cpu: the CPU which is coming up
|
|
|
|
*
|
|
|
|
* An unbound pool may end up with a cpumask which doesn't have any online
|
|
|
|
* CPUs. When a worker of such pool get scheduled, the scheduler resets
|
|
|
|
* its cpus_allowed. If @cpu is in @pool's cpumask which didn't have any
|
|
|
|
* online CPU before, cpus_allowed of all its workers should be restored.
|
|
|
|
*/
|
|
|
|
static void restore_unbound_workers_cpumask(struct worker_pool *pool, int cpu)
|
|
|
|
{
|
|
|
|
static cpumask_t cpumask;
|
|
|
|
struct worker *worker;
|
|
|
|
int wi;
|
|
|
|
|
|
|
|
lockdep_assert_held(&pool->manager_mutex);
|
|
|
|
|
|
|
|
/* is @cpu allowed for @pool? */
|
|
|
|
if (!cpumask_test_cpu(cpu, pool->attrs->cpumask))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* is @cpu the only online CPU? */
|
|
|
|
cpumask_and(&cpumask, pool->attrs->cpumask, cpu_online_mask);
|
|
|
|
if (cpumask_weight(&cpumask) != 1)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* as we're called from CPU_ONLINE, the following shouldn't fail */
|
|
|
|
for_each_pool_worker(worker, wi, pool)
|
|
|
|
WARN_ON_ONCE(set_cpus_allowed_ptr(worker->task,
|
|
|
|
pool->attrs->cpumask) < 0);
|
|
|
|
}
|
|
|
|
|
2012-07-17 13:39:28 -06:00
|
|
|
/*
|
|
|
|
* Workqueues should be brought up before normal priority CPU notifiers.
|
|
|
|
* This will be registered high priority CPU notifier.
|
|
|
|
*/
|
2013-06-19 12:53:51 -06:00
|
|
|
static int workqueue_cpu_up_callback(struct notifier_block *nfb,
|
2012-07-17 13:39:28 -06:00
|
|
|
unsigned long action,
|
|
|
|
void *hcpu)
|
2007-05-09 03:34:09 -06:00
|
|
|
{
|
2013-03-12 12:29:59 -06:00
|
|
|
int cpu = (unsigned long)hcpu;
|
2012-07-13 23:16:44 -06:00
|
|
|
struct worker_pool *pool;
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
struct workqueue_struct *wq;
|
workqueue: restore CPU affinity of unbound workers on CPU_ONLINE
With the recent addition of the custom attributes support, unbound
pools may have allowed cpumask which isn't full. As long as some of
CPUs in the cpumask are online, its workers will maintain cpus_allowed
as set on worker creation; however, once no online CPU is left in
cpus_allowed, the scheduler will reset cpus_allowed of any workers
which get scheduled so that they can execute.
To remain compliant to the user-specified configuration, CPU affinity
needs to be restored when a CPU becomes online for an unbound pool
which doesn't currently have any online CPUs before.
This patch implement restore_unbound_workers_cpumask(), which is
called from CPU_ONLINE for all unbound pools, checks whether the
coming up CPU is the first allowed online one, and, if so, invokes
set_cpus_allowed_ptr() with the configured cpumask on all workers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
int pi;
|
2012-07-17 13:39:27 -06:00
|
|
|
|
2012-07-17 13:39:28 -06:00
|
|
|
switch (action & ~CPU_TASKS_FROZEN) {
|
2007-05-09 03:34:09 -06:00
|
|
|
case CPU_UP_PREPARE:
|
2013-03-12 12:30:03 -06:00
|
|
|
for_each_cpu_worker_pool(pool, cpu) {
|
2012-07-17 13:39:27 -06:00
|
|
|
if (pool->nr_workers)
|
|
|
|
continue;
|
2013-03-13 20:47:39 -06:00
|
|
|
if (create_and_start_worker(pool) < 0)
|
2012-07-17 13:39:27 -06:00
|
|
|
return NOTIFY_BAD;
|
2007-05-09 03:34:09 -06:00
|
|
|
}
|
2012-07-17 13:39:28 -06:00
|
|
|
break;
|
2007-05-09 03:34:09 -06:00
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
case CPU_DOWN_FAILED:
|
|
|
|
case CPU_ONLINE:
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
workqueue: restore CPU affinity of unbound workers on CPU_ONLINE
With the recent addition of the custom attributes support, unbound
pools may have allowed cpumask which isn't full. As long as some of
CPUs in the cpumask are online, its workers will maintain cpus_allowed
as set on worker creation; however, once no online CPU is left in
cpus_allowed, the scheduler will reset cpus_allowed of any workers
which get scheduled so that they can execute.
To remain compliant to the user-specified configuration, CPU affinity
needs to be restored when a CPU becomes online for an unbound pool
which doesn't currently have any online CPUs before.
This patch implement restore_unbound_workers_cpumask(), which is
called from CPU_ONLINE for all unbound pools, checks whether the
coming up CPU is the first allowed online one, and, if so, invokes
set_cpus_allowed_ptr() with the configured cpumask on all workers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
|
|
|
|
for_each_pool(pool, pi) {
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_lock(&pool->manager_mutex);
|
2013-01-24 12:01:33 -07:00
|
|
|
|
workqueue: restore CPU affinity of unbound workers on CPU_ONLINE
With the recent addition of the custom attributes support, unbound
pools may have allowed cpumask which isn't full. As long as some of
CPUs in the cpumask are online, its workers will maintain cpus_allowed
as set on worker creation; however, once no online CPU is left in
cpus_allowed, the scheduler will reset cpus_allowed of any workers
which get scheduled so that they can execute.
To remain compliant to the user-specified configuration, CPU affinity
needs to be restored when a CPU becomes online for an unbound pool
which doesn't currently have any online CPUs before.
This patch implement restore_unbound_workers_cpumask(), which is
called from CPU_ONLINE for all unbound pools, checks whether the
coming up CPU is the first allowed online one, and, if so, invokes
set_cpus_allowed_ptr() with the configured cpumask on all workers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
if (pool->cpu == cpu) {
|
|
|
|
spin_lock_irq(&pool->lock);
|
|
|
|
pool->flags &= ~POOL_DISASSOCIATED;
|
|
|
|
spin_unlock_irq(&pool->lock);
|
workqueue: directly restore CPU affinity of workers from CPU_ONLINE
Rebinding workers of a per-cpu pool after a CPU comes online involves
a lot of back-and-forth mostly because only the task itself could
adjust CPU affinity if PF_THREAD_BOUND was set.
As CPU_ONLINE itself couldn't adjust affinity, it had to somehow
coerce the workers themselves to perform set_cpus_allowed_ptr(). Due
to the various states a worker can be in, this led to three different
paths a worker may be rebound. worker->rebind_work is queued to busy
workers. Idle ones are signaled by unlinking worker->entry and call
idle_worker_rebind(). The manager isn't covered by either and
implements its own mechanism.
PF_THREAD_BOUND has been relaced with PF_NO_SETAFFINITY and CPU_ONLINE
itself now can manipulate CPU affinity of workers. This patch
replaces the existing rebind mechanism with direct one where
CPU_ONLINE iterates over all workers using for_each_pool_worker(),
restores CPU affinity, and clears WORKER_UNBOUND.
There are a couple subtleties. All bound idle workers should have
their runqueues set to that of the bound CPU; however, if the target
task isn't running, set_cpus_allowed_ptr() just updates the
cpus_allowed mask deferring the actual migration to when the task
wakes up. This is worked around by waking up idle workers after
restoring CPU affinity before any workers can become bound.
Another subtlety is stems from matching @pool->nr_running with the
number of running unbound workers. While DISASSOCIATED, all workers
are unbound and nr_running is zero. As workers become bound again,
nr_running needs to be adjusted accordingly; however, there is no good
way to tell whether a given worker is running without poking into
scheduler internals. Instead of clearing UNBOUND directly,
rebind_workers() replaces UNBOUND with another new NOT_RUNNING flag -
REBOUND, which will later be cleared by the workers themselves while
preparing for the next round of work item execution. The only change
needed for the workers is clearing REBOUND along with PREP.
* This patch leaves for_each_busy_worker() without any user. Removed.
* idle_worker_rebind(), busy_worker_rebind_fn(), worker->rebind_work
and rebind logic in manager_workers() removed.
* worker_thread() now looks at WORKER_DIE instead of testing whether
@worker->entry is empty to determine whether it needs to do
something special as dying is the only special thing now.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
|
workqueue: restore CPU affinity of unbound workers on CPU_ONLINE
With the recent addition of the custom attributes support, unbound
pools may have allowed cpumask which isn't full. As long as some of
CPUs in the cpumask are online, its workers will maintain cpus_allowed
as set on worker creation; however, once no online CPU is left in
cpus_allowed, the scheduler will reset cpus_allowed of any workers
which get scheduled so that they can execute.
To remain compliant to the user-specified configuration, CPU affinity
needs to be restored when a CPU becomes online for an unbound pool
which doesn't currently have any online CPUs before.
This patch implement restore_unbound_workers_cpumask(), which is
called from CPU_ONLINE for all unbound pools, checks whether the
coming up CPU is the first allowed online one, and, if so, invokes
set_cpus_allowed_ptr() with the configured cpumask on all workers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
rebind_workers(pool);
|
|
|
|
} else if (pool->cpu < 0) {
|
|
|
|
restore_unbound_workers_cpumask(pool, cpu);
|
|
|
|
}
|
2013-01-24 12:01:33 -07:00
|
|
|
|
2013-03-13 20:47:39 -06:00
|
|
|
mutex_unlock(&pool->manager_mutex);
|
2013-01-24 12:01:33 -07:00
|
|
|
}
|
workqueue: restore CPU affinity of unbound workers on CPU_ONLINE
With the recent addition of the custom attributes support, unbound
pools may have allowed cpumask which isn't full. As long as some of
CPUs in the cpumask are online, its workers will maintain cpus_allowed
as set on worker creation; however, once no online CPU is left in
cpus_allowed, the scheduler will reset cpus_allowed of any workers
which get scheduled so that they can execute.
To remain compliant to the user-specified configuration, CPU affinity
needs to be restored when a CPU becomes online for an unbound pool
which doesn't currently have any online CPUs before.
This patch implement restore_unbound_workers_cpumask(), which is
called from CPU_ONLINE for all unbound pools, checks whether the
coming up CPU is the first allowed online one, and, if so, invokes
set_cpus_allowed_ptr() with the configured cpumask on all workers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-03-19 14:45:21 -06:00
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/* update NUMA affinity of unbound workqueues */
|
|
|
|
list_for_each_entry(wq, &workqueues, list)
|
|
|
|
wq_update_unbound_numa(wq, cpu, true);
|
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
break;
|
2008-04-29 02:00:27 -06:00
|
|
|
}
|
2012-07-17 13:39:26 -06:00
|
|
|
return NOTIFY_OK;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Workqueues should be brought down after normal priority CPU notifiers.
|
|
|
|
* This will be registered as low priority CPU notifier.
|
|
|
|
*/
|
2013-06-19 12:53:51 -06:00
|
|
|
static int workqueue_cpu_down_callback(struct notifier_block *nfb,
|
2012-07-17 13:39:26 -06:00
|
|
|
unsigned long action,
|
|
|
|
void *hcpu)
|
|
|
|
{
|
2013-03-12 12:29:59 -06:00
|
|
|
int cpu = (unsigned long)hcpu;
|
2012-07-17 13:39:28 -06:00
|
|
|
struct work_struct unbind_work;
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
struct workqueue_struct *wq;
|
2012-07-17 13:39:28 -06:00
|
|
|
|
2012-07-17 13:39:26 -06:00
|
|
|
switch (action & ~CPU_TASKS_FROZEN) {
|
|
|
|
case CPU_DOWN_PREPARE:
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
/* unbinding per-cpu workers should happen on the local CPU */
|
2013-01-24 12:01:34 -07:00
|
|
|
INIT_WORK_ONSTACK(&unbind_work, wq_unbind_fn);
|
2012-08-15 08:25:41 -06:00
|
|
|
queue_work_on(cpu, system_highpri_wq, &unbind_work);
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
|
|
|
|
/* update NUMA affinity of unbound workqueues */
|
|
|
|
mutex_lock(&wq_pool_mutex);
|
|
|
|
list_for_each_entry(wq, &workqueues, list)
|
|
|
|
wq_update_unbound_numa(wq, cpu, false);
|
|
|
|
mutex_unlock(&wq_pool_mutex);
|
|
|
|
|
|
|
|
/* wait for per-cpu unbinding to finish */
|
2012-07-17 13:39:28 -06:00
|
|
|
flush_work(&unbind_work);
|
2014-01-11 20:26:33 -07:00
|
|
|
destroy_work_on_stack(&unbind_work);
|
2012-07-17 13:39:28 -06:00
|
|
|
break;
|
2012-07-17 13:39:26 -06:00
|
|
|
}
|
|
|
|
return NOTIFY_OK;
|
|
|
|
}
|
|
|
|
|
2008-11-04 19:39:10 -07:00
|
|
|
#ifdef CONFIG_SMP
|
2009-01-16 16:31:15 -07:00
|
|
|
|
2008-11-04 19:39:10 -07:00
|
|
|
struct work_for_cpu {
|
2012-09-18 13:48:43 -06:00
|
|
|
struct work_struct work;
|
2008-11-04 19:39:10 -07:00
|
|
|
long (*fn)(void *);
|
|
|
|
void *arg;
|
|
|
|
long ret;
|
|
|
|
};
|
|
|
|
|
2012-09-18 13:48:43 -06:00
|
|
|
static void work_for_cpu_fn(struct work_struct *work)
|
2008-11-04 19:39:10 -07:00
|
|
|
{
|
2012-09-18 13:48:43 -06:00
|
|
|
struct work_for_cpu *wfc = container_of(work, struct work_for_cpu, work);
|
|
|
|
|
2008-11-04 19:39:10 -07:00
|
|
|
wfc->ret = wfc->fn(wfc->arg);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* work_on_cpu - run a function in user context on a particular cpu
|
|
|
|
* @cpu: the cpu to run on
|
|
|
|
* @fn: the function to run
|
|
|
|
* @arg: the function arg
|
|
|
|
*
|
2009-01-16 16:31:15 -07:00
|
|
|
* It is up to the caller to ensure that the cpu doesn't go offline.
|
2009-04-09 09:50:37 -06:00
|
|
|
* The caller must not hold any locks which would prevent @fn from completing.
|
2013-07-31 15:59:24 -06:00
|
|
|
*
|
|
|
|
* Return: The value @fn returns.
|
2008-11-04 19:39:10 -07:00
|
|
|
*/
|
2013-03-12 12:29:59 -06:00
|
|
|
long work_on_cpu(int cpu, long (*fn)(void *), void *arg)
|
2008-11-04 19:39:10 -07:00
|
|
|
{
|
2012-09-18 13:48:43 -06:00
|
|
|
struct work_for_cpu wfc = { .fn = fn, .arg = arg };
|
2009-04-09 09:50:37 -06:00
|
|
|
|
2012-09-18 13:48:43 -06:00
|
|
|
INIT_WORK_ONSTACK(&wfc.work, work_for_cpu_fn);
|
|
|
|
schedule_work_on(cpu, &wfc.work);
|
2013-11-18 11:00:29 -07:00
|
|
|
flush_work(&wfc.work);
|
2014-01-11 20:26:33 -07:00
|
|
|
destroy_work_on_stack(&wfc.work);
|
2008-11-04 19:39:10 -07:00
|
|
|
return wfc.ret;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(work_on_cpu);
|
|
|
|
#endif /* CONFIG_SMP */
|
|
|
|
|
2010-06-29 02:07:12 -06:00
|
|
|
#ifdef CONFIG_FREEZER
|
|
|
|
|
|
|
|
/**
|
|
|
|
* freeze_workqueues_begin - begin freezing workqueues
|
|
|
|
*
|
2011-02-16 01:25:31 -07:00
|
|
|
* Start freezing workqueues. After this function returns, all freezable
|
2013-03-13 17:51:36 -06:00
|
|
|
* workqueues will queue new works to their delayed_works list instead of
|
2013-01-24 12:01:34 -07:00
|
|
|
* pool->worklist.
|
2010-06-29 02:07:12 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-03-25 17:57:19 -06:00
|
|
|
* Grabs and releases wq_pool_mutex, wq->mutex and pool->lock's.
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
|
|
|
void freeze_workqueues_begin(void)
|
|
|
|
{
|
2013-03-12 12:29:58 -06:00
|
|
|
struct worker_pool *pool;
|
2013-03-12 12:29:58 -06:00
|
|
|
struct workqueue_struct *wq;
|
|
|
|
struct pool_workqueue *pwq;
|
2013-03-13 17:51:36 -06:00
|
|
|
int pi;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(workqueue_freezing);
|
2010-06-29 02:07:12 -06:00
|
|
|
workqueue_freezing = true;
|
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
/* set FREEZING */
|
2013-03-13 17:51:36 -06:00
|
|
|
for_each_pool(pool, pi) {
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_lock_irq(&pool->lock);
|
2013-03-12 12:29:58 -06:00
|
|
|
WARN_ON_ONCE(pool->flags & POOL_FREEZING);
|
|
|
|
pool->flags |= POOL_FREEZING;
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2013-03-12 12:29:58 -06:00
|
|
|
}
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
list_for_each_entry(wq, &workqueues, list) {
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2013-03-13 17:51:35 -06:00
|
|
|
for_each_pwq(pwq, wq)
|
|
|
|
pwq_adjust_max_active(pwq);
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
2013-03-13 20:47:40 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2011-02-16 01:25:31 -07:00
|
|
|
* freeze_workqueues_busy - are freezable workqueues still busy?
|
2010-06-29 02:07:12 -06:00
|
|
|
*
|
|
|
|
* Check whether freezing is complete. This function must be called
|
|
|
|
* between freeze_workqueues_begin() and thaw_workqueues().
|
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-03-25 17:57:17 -06:00
|
|
|
* Grabs and releases wq_pool_mutex.
|
2010-06-29 02:07:12 -06:00
|
|
|
*
|
2013-07-31 15:59:24 -06:00
|
|
|
* Return:
|
2011-02-16 01:25:31 -07:00
|
|
|
* %true if some freezable workqueues are still busy. %false if freezing
|
|
|
|
* is complete.
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
|
|
|
bool freeze_workqueues_busy(void)
|
|
|
|
{
|
|
|
|
bool busy = false;
|
2013-03-12 12:29:58 -06:00
|
|
|
struct workqueue_struct *wq;
|
|
|
|
struct pool_workqueue *pwq;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(!workqueue_freezing);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
list_for_each_entry(wq, &workqueues, list) {
|
|
|
|
if (!(wq->flags & WQ_FREEZABLE))
|
|
|
|
continue;
|
2010-06-29 02:07:12 -06:00
|
|
|
/*
|
|
|
|
* nr_active is monotonically decreasing. It's safe
|
|
|
|
* to peek without lock.
|
|
|
|
*/
|
2013-03-19 13:28:10 -06:00
|
|
|
rcu_read_lock_sched();
|
2013-03-12 12:29:58 -06:00
|
|
|
for_each_pwq(pwq, wq) {
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON_ONCE(pwq->nr_active < 0);
|
2013-02-13 20:29:12 -07:00
|
|
|
if (pwq->nr_active) {
|
2010-06-29 02:07:12 -06:00
|
|
|
busy = true;
|
2013-03-19 13:28:10 -06:00
|
|
|
rcu_read_unlock_sched();
|
2010-06-29 02:07:12 -06:00
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
}
|
2013-03-19 13:28:10 -06:00
|
|
|
rcu_read_unlock_sched();
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
|
|
|
out_unlock:
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
return busy;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* thaw_workqueues - thaw workqueues
|
|
|
|
*
|
|
|
|
* Thaw workqueues. Normal queueing is restored and all collected
|
2013-01-24 12:01:34 -07:00
|
|
|
* frozen works are transferred to their respective pool worklists.
|
2010-06-29 02:07:12 -06:00
|
|
|
*
|
|
|
|
* CONTEXT:
|
2013-03-25 17:57:19 -06:00
|
|
|
* Grabs and releases wq_pool_mutex, wq->mutex and pool->lock's.
|
2010-06-29 02:07:12 -06:00
|
|
|
*/
|
|
|
|
void thaw_workqueues(void)
|
|
|
|
{
|
2013-03-12 12:29:58 -06:00
|
|
|
struct workqueue_struct *wq;
|
|
|
|
struct pool_workqueue *pwq;
|
|
|
|
struct worker_pool *pool;
|
2013-03-13 17:51:36 -06:00
|
|
|
int pi;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
|
|
|
if (!workqueue_freezing)
|
|
|
|
goto out_unlock;
|
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
/* clear FREEZING */
|
2013-03-13 17:51:36 -06:00
|
|
|
for_each_pool(pool, pi) {
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_lock_irq(&pool->lock);
|
2013-03-12 12:29:58 -06:00
|
|
|
WARN_ON_ONCE(!(pool->flags & POOL_FREEZING));
|
|
|
|
pool->flags &= ~POOL_FREEZING;
|
2013-03-13 20:47:40 -06:00
|
|
|
spin_unlock_irq(&pool->lock);
|
2013-03-12 12:29:58 -06:00
|
|
|
}
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:29:58 -06:00
|
|
|
/* restore max_active and repopulate worklist */
|
|
|
|
list_for_each_entry(wq, &workqueues, list) {
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_lock(&wq->mutex);
|
2013-03-13 17:51:35 -06:00
|
|
|
for_each_pwq(pwq, wq)
|
|
|
|
pwq_adjust_max_active(pwq);
|
2013-03-25 17:57:19 -06:00
|
|
|
mutex_unlock(&wq->mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
workqueue_freezing = false;
|
|
|
|
out_unlock:
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
|
|
|
#endif /* CONFIG_FREEZER */
|
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
static void __init wq_numa_init(void)
|
|
|
|
{
|
|
|
|
cpumask_var_t *tbl;
|
|
|
|
int node, cpu;
|
|
|
|
|
|
|
|
/* determine NUMA pwq table len - highest node id + 1 */
|
|
|
|
for_each_node(node)
|
|
|
|
wq_numa_tbl_len = max(wq_numa_tbl_len, node + 1);
|
|
|
|
|
|
|
|
if (num_possible_nodes() <= 1)
|
|
|
|
return;
|
|
|
|
|
2013-04-01 12:23:38 -06:00
|
|
|
if (wq_disable_numa) {
|
|
|
|
pr_info("workqueue: NUMA affinity support disabled\n");
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
workqueue: implement NUMA affinity for unbound workqueues
Currently, an unbound workqueue has single current, or first, pwq
(pool_workqueue) to which all new work items are queued. This often
isn't optimal on NUMA machines as workers may jump around across node
boundaries and work items get assigned to workers without any regard
to NUMA affinity.
This patch implements NUMA affinity for unbound workqueues. Instead
of mapping all entries of numa_pwq_tbl[] to the same pwq,
apply_workqueue_attrs() now creates a separate pwq covering the
intersecting CPUs for each NUMA node which has online CPUs in
@attrs->cpumask. Nodes which don't have intersecting possible CPUs
are mapped to pwqs covering whole @attrs->cpumask.
As CPUs come up and go down, the pool association is changed
accordingly. Changing pool association may involve allocating new
pools which may fail. To avoid failing CPU_DOWN, each workqueue
always keeps a default pwq which covers whole attrs->cpumask which is
used as fallback if pool creation fails during a CPU hotplug
operation.
This ensures that all work items issued on a NUMA node is executed on
the same node as long as the workqueue allows execution on the CPUs of
the node.
As this maps a workqueue to multiple pwqs and max_active is per-pwq,
this change the behavior of max_active. The limit is now per NUMA
node instead of global. While this is an actual change, max_active is
already per-cpu for per-cpu workqueues and primarily used as safety
mechanism rather than for active concurrency control. Concurrency is
usually limited from workqueue users by the number of concurrently
active work items and this change shouldn't matter much.
v2: Fixed pwq freeing in apply_workqueue_attrs() error path. Spotted
by Lai.
v3: The previous version incorrectly made a workqueue spanning
multiple nodes spread work items over all online CPUs when some of
its nodes don't have any desired cpus. Reimplemented so that NUMA
affinity is properly updated as CPUs go up and down. This problem
was spotted by Lai Jiangshan.
v4: destroy_workqueue() was putting wq->dfl_pwq and then clearing it;
however, wq may be freed at any time after dfl_pwq is put making
the clearing use-after-free. Clear wq->dfl_pwq before putting it.
v5: apply_workqueue_attrs() was leaking @tmp_attrs, @new_attrs and
@pwq_tbl after success. Fixed.
Retry loop in wq_update_unbound_numa_attrs() isn't necessary as
application of new attrs is excluded via CPU hotplug. Removed.
Documentation on CPU affinity guarantee on CPU_DOWN added.
All changes are suggested by Lai Jiangshan.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
2013-04-01 12:23:36 -06:00
|
|
|
wq_update_unbound_numa_attrs_buf = alloc_workqueue_attrs(GFP_KERNEL);
|
|
|
|
BUG_ON(!wq_update_unbound_numa_attrs_buf);
|
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
/*
|
|
|
|
* We want masks of possible CPUs of each node which isn't readily
|
|
|
|
* available. Build one from cpu_to_node() which should have been
|
|
|
|
* fully initialized by now.
|
|
|
|
*/
|
|
|
|
tbl = kzalloc(wq_numa_tbl_len * sizeof(tbl[0]), GFP_KERNEL);
|
|
|
|
BUG_ON(!tbl);
|
|
|
|
|
|
|
|
for_each_node(node)
|
2013-05-15 15:24:24 -06:00
|
|
|
BUG_ON(!alloc_cpumask_var_node(&tbl[node], GFP_KERNEL,
|
|
|
|
node_online(node) ? node : NUMA_NO_NODE));
|
2013-04-01 12:23:32 -06:00
|
|
|
|
|
|
|
for_each_possible_cpu(cpu) {
|
|
|
|
node = cpu_to_node(cpu);
|
|
|
|
if (WARN_ON(node == NUMA_NO_NODE)) {
|
|
|
|
pr_warn("workqueue: NUMA node mapping not available for cpu%d, disabling NUMA support\n", cpu);
|
|
|
|
/* happens iff arch is bonkers, let's just proceed */
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
cpumask_set_cpu(cpu, tbl[node]);
|
|
|
|
}
|
|
|
|
|
|
|
|
wq_numa_possible_cpumask = tbl;
|
|
|
|
wq_numa_enabled = true;
|
|
|
|
}
|
|
|
|
|
2010-07-30 15:57:37 -06:00
|
|
|
static int __init init_workqueues(void)
|
2005-04-16 16:20:36 -06:00
|
|
|
{
|
2013-03-12 12:30:00 -06:00
|
|
|
int std_nice[NR_STD_WORKER_POOLS] = { 0, HIGHPRI_NICE_LEVEL };
|
|
|
|
int i, cpu;
|
2010-06-29 02:07:11 -06:00
|
|
|
|
2013-03-12 12:29:57 -06:00
|
|
|
WARN_ON(__alignof__(struct pool_workqueue) < __alignof__(long long));
|
|
|
|
|
|
|
|
pwq_cache = KMEM_CACHE(pool_workqueue, SLAB_PANIC);
|
|
|
|
|
2012-07-17 13:39:26 -06:00
|
|
|
cpu_notifier(workqueue_cpu_up_callback, CPU_PRI_WORKQUEUE_UP);
|
2012-09-18 10:59:23 -06:00
|
|
|
hotcpu_notifier(workqueue_cpu_down_callback, CPU_PRI_WORKQUEUE_DOWN);
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-04-01 12:23:32 -06:00
|
|
|
wq_numa_init();
|
|
|
|
|
2013-01-24 12:01:34 -07:00
|
|
|
/* initialize CPU pools */
|
2013-03-12 12:30:03 -06:00
|
|
|
for_each_possible_cpu(cpu) {
|
2012-07-13 23:16:44 -06:00
|
|
|
struct worker_pool *pool;
|
2010-06-29 02:07:12 -06:00
|
|
|
|
2013-03-12 12:30:00 -06:00
|
|
|
i = 0;
|
2013-03-12 12:30:03 -06:00
|
|
|
for_each_cpu_worker_pool(pool, cpu) {
|
2013-03-12 12:30:00 -06:00
|
|
|
BUG_ON(init_worker_pool(pool));
|
2013-01-24 12:01:33 -07:00
|
|
|
pool->cpu = cpu;
|
2013-03-12 12:30:03 -06:00
|
|
|
cpumask_copy(pool->attrs->cpumask, cpumask_of(cpu));
|
2013-03-12 12:30:00 -06:00
|
|
|
pool->attrs->nice = std_nice[i++];
|
2013-04-01 12:23:34 -06:00
|
|
|
pool->node = cpu_to_node(cpu);
|
2013-03-12 12:30:00 -06:00
|
|
|
|
2013-01-24 12:01:33 -07:00
|
|
|
/* alloc pool ID */
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_lock(&wq_pool_mutex);
|
2013-01-24 12:01:33 -07:00
|
|
|
BUG_ON(worker_pool_assign_id(pool));
|
2013-03-25 17:57:17 -06:00
|
|
|
mutex_unlock(&wq_pool_mutex);
|
2012-07-13 23:16:44 -06:00
|
|
|
}
|
2010-06-29 02:07:12 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
/* create the initial worker */
|
2013-03-12 12:30:03 -06:00
|
|
|
for_each_online_cpu(cpu) {
|
2012-07-13 23:16:44 -06:00
|
|
|
struct worker_pool *pool;
|
2010-06-29 02:07:14 -06:00
|
|
|
|
2013-03-12 12:30:03 -06:00
|
|
|
for_each_cpu_worker_pool(pool, cpu) {
|
2013-03-12 12:30:03 -06:00
|
|
|
pool->flags &= ~POOL_DISASSOCIATED;
|
2013-03-13 20:47:39 -06:00
|
|
|
BUG_ON(create_and_start_worker(pool) < 0);
|
2012-07-13 23:16:44 -06:00
|
|
|
}
|
2010-06-29 02:07:14 -06:00
|
|
|
}
|
|
|
|
|
2013-09-05 10:30:04 -06:00
|
|
|
/* create default unbound and ordered wq attrs */
|
2013-03-12 12:30:03 -06:00
|
|
|
for (i = 0; i < NR_STD_WORKER_POOLS; i++) {
|
|
|
|
struct workqueue_attrs *attrs;
|
|
|
|
|
|
|
|
BUG_ON(!(attrs = alloc_workqueue_attrs(GFP_KERNEL)));
|
|
|
|
attrs->nice = std_nice[i];
|
|
|
|
unbound_std_wq_attrs[i] = attrs;
|
2013-09-05 10:30:04 -06:00
|
|
|
|
|
|
|
/*
|
|
|
|
* An ordered wq should have only one pwq as ordering is
|
|
|
|
* guaranteed by max_active which is enforced by pwqs.
|
|
|
|
* Turn off NUMA so that dfl_pwq is used for all nodes.
|
|
|
|
*/
|
|
|
|
BUG_ON(!(attrs = alloc_workqueue_attrs(GFP_KERNEL)));
|
|
|
|
attrs->nice = std_nice[i];
|
|
|
|
attrs->no_numa = true;
|
|
|
|
ordered_wq_attrs[i] = attrs;
|
2013-03-12 12:30:03 -06:00
|
|
|
}
|
|
|
|
|
2010-06-29 02:07:14 -06:00
|
|
|
system_wq = alloc_workqueue("events", 0, 0);
|
2012-08-15 08:25:39 -06:00
|
|
|
system_highpri_wq = alloc_workqueue("events_highpri", WQ_HIGHPRI, 0);
|
2010-06-29 02:07:14 -06:00
|
|
|
system_long_wq = alloc_workqueue("events_long", 0, 0);
|
2010-07-02 02:03:51 -06:00
|
|
|
system_unbound_wq = alloc_workqueue("events_unbound", WQ_UNBOUND,
|
|
|
|
WQ_UNBOUND_MAX_ACTIVE);
|
2011-02-21 01:52:50 -07:00
|
|
|
system_freezable_wq = alloc_workqueue("events_freezable",
|
|
|
|
WQ_FREEZABLE, 0);
|
2013-04-24 05:42:54 -06:00
|
|
|
system_power_efficient_wq = alloc_workqueue("events_power_efficient",
|
|
|
|
WQ_POWER_EFFICIENT, 0);
|
|
|
|
system_freezable_power_efficient_wq = alloc_workqueue("events_freezable_power_efficient",
|
|
|
|
WQ_FREEZABLE | WQ_POWER_EFFICIENT,
|
|
|
|
0);
|
2012-08-15 08:25:39 -06:00
|
|
|
BUG_ON(!system_wq || !system_highpri_wq || !system_long_wq ||
|
2013-04-24 05:42:54 -06:00
|
|
|
!system_unbound_wq || !system_freezable_wq ||
|
|
|
|
!system_power_efficient_wq ||
|
|
|
|
!system_freezable_power_efficient_wq);
|
2010-07-30 15:57:37 -06:00
|
|
|
return 0;
|
2005-04-16 16:20:36 -06:00
|
|
|
}
|
2010-07-30 15:57:37 -06:00
|
|
|
early_initcall(init_workqueues);
|