2005-04-17 02:20:36 +04:00
/*
* linux / kernel / workqueue . c
*
* Generic mechanism for defining kernel helper threads for running
* arbitrary tasks in process context .
*
* Started by Ingo Molnar , Copyright ( C ) 2002
*
* Derived from the taskqueue / keventd code by :
*
* David Woodhouse < dwmw2 @ infradead . org >
* Andrew Morton < andrewm @ uow . edu . au >
* Kai Petzke < wpp @ marie . physik . tu - berlin . de >
* Theodore Ts ' o < tytso @ mit . edu >
2005-10-31 02:01:59 +03:00
*
* Made to use alloc_percpu by Christoph Lameter < clameter @ sgi . com > .
2005-04-17 02:20:36 +04:00
*/
# include <linux/module.h>
# 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 21:43:43 +03:00
# include <linux/hardirq.h>
2006-10-11 12:21:26 +04:00
# include <linux/mempolicy.h>
2006-12-07 07:34:49 +03:00
# include <linux/freezer.h>
2006-12-07 07:37:26 +03:00
# include <linux/kallsyms.h>
# include <linux/debug_locks.h>
2005-04-17 02:20:36 +04:00
/*
2006-01-08 12:05:12 +03:00
* The per - CPU workqueue ( if single thread , we always use the first
* possible cpu ) .
2005-04-17 02:20:36 +04:00
*/
struct cpu_workqueue_struct {
spinlock_t lock ;
struct list_head worklist ;
wait_queue_head_t more_work ;
2007-05-09 13:34:09 +04:00
struct work_struct * current_work ;
2005-04-17 02:20:36 +04:00
struct workqueue_struct * wq ;
2006-07-03 11:25:41 +04:00
struct task_struct * thread ;
2007-05-09 13:34:09 +04:00
int should_stop ;
2005-04-17 02:20:36 +04:00
int run_depth ; /* Detect run_workqueue() recursion depth */
} ____cacheline_aligned ;
/*
* The externally visible workqueue abstraction is an array of
* per - CPU workqueues :
*/
struct workqueue_struct {
2005-10-31 02:01:59 +03:00
struct cpu_workqueue_struct * cpu_wq ;
2007-05-09 13:34:13 +04:00
struct list_head list ;
2005-04-17 02:20:36 +04:00
const char * name ;
2007-05-09 13:34:13 +04:00
int singlethread ;
2007-05-09 13:34:06 +04:00
int freezeable ; /* Freeze threads during suspend */
2005-04-17 02:20:36 +04:00
} ;
/* All the per-cpu workqueues on the system, for hotplug cpu to add/remove
threads to each one as cpus come / go . */
2006-08-14 10:24:26 +04:00
static DEFINE_MUTEX ( workqueue_mutex ) ;
2005-04-17 02:20:36 +04:00
static LIST_HEAD ( workqueues ) ;
2007-05-09 13:34:09 +04:00
static int singlethread_cpu __read_mostly ;
2007-05-09 13:34:12 +04:00
static cpumask_t cpu_singlethread_map __read_mostly ;
2007-05-09 13:34:09 +04:00
/* optimization, we could use cpu_possible_map */
static cpumask_t cpu_populated_map __read_mostly ;
2006-01-08 12:05:12 +03:00
2005-04-17 02:20:36 +04:00
/* If it's single threaded, it isn't in the list of workqueues. */
static inline int is_single_threaded ( struct workqueue_struct * wq )
{
2007-05-09 13:34:13 +04:00
return wq - > singlethread ;
2005-04-17 02:20:36 +04:00
}
2007-05-09 13:34:12 +04:00
static const cpumask_t * wq_cpu_map ( struct workqueue_struct * wq )
{
return is_single_threaded ( wq )
? & cpu_singlethread_map : & cpu_populated_map ;
}
[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 14:33:26 +03:00
/*
* Set the workqueue on which a work item is to be run
* - Must * only * be called if the pending flag is set
*/
2006-11-22 17:54:49 +03:00
static inline void set_wq_data ( struct work_struct * work , void * wq )
{
[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 14:33:26 +03:00
unsigned long new ;
BUG_ON ( ! work_pending ( work ) ) ;
2006-11-22 17:54:49 +03:00
new = ( unsigned long ) wq | ( 1UL < < WORK_STRUCT_PENDING ) ;
2006-12-16 20:53:50 +03:00
new | = WORK_STRUCT_FLAG_MASK & * work_data_bits ( work ) ;
atomic_long_set ( & work - > data , new ) ;
2006-11-22 17:54:49 +03:00
}
static inline void * get_wq_data ( struct work_struct * work )
{
2006-12-16 20:53:50 +03:00
return ( void * ) ( atomic_long_read ( & work - > data ) & WORK_STRUCT_WQ_DATA_MASK ) ;
2006-11-22 17:54:49 +03:00
}
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 13:33:52 +04:00
static void insert_work ( struct cpu_workqueue_struct * cwq ,
struct work_struct * work , int tail )
{
set_wq_data ( work , cwq ) ;
if ( tail )
list_add_tail ( & work - > entry , & cwq - > worklist ) ;
else
list_add ( & work - > entry , & cwq - > worklist ) ;
wake_up ( & cwq - > more_work ) ;
}
2005-04-17 02:20:36 +04:00
/* Preempt must be disabled. */
static void __queue_work ( struct cpu_workqueue_struct * cwq ,
struct work_struct * work )
{
unsigned long flags ;
spin_lock_irqsave ( & cwq - > lock , 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 13:33:52 +04:00
insert_work ( cwq , work , 1 ) ;
2005-04-17 02:20:36 +04:00
spin_unlock_irqrestore ( & cwq - > lock , flags ) ;
}
2006-07-30 14:03:42 +04:00
/**
* queue_work - queue work on a workqueue
* @ wq : workqueue to use
* @ work : work to queue
*
2006-10-28 21:38:58 +04:00
* Returns 0 if @ work was already on a queue , non - zero otherwise .
2005-04-17 02:20:36 +04:00
*
* We queue the work to the CPU it was submitted , but there is no
* guarantee that it will be processed by that CPU .
*/
int fastcall queue_work ( struct workqueue_struct * wq , struct work_struct * work )
{
int ret = 0 , cpu = get_cpu ( ) ;
2006-12-16 20:53:50 +03:00
if ( ! test_and_set_bit ( WORK_STRUCT_PENDING , work_data_bits ( work ) ) ) {
2005-04-17 02:20:36 +04:00
if ( unlikely ( is_single_threaded ( wq ) ) )
2006-01-08 12:05:12 +03:00
cpu = singlethread_cpu ;
2005-04-17 02:20:36 +04:00
BUG_ON ( ! list_empty ( & work - > entry ) ) ;
2005-10-31 02:01:59 +03:00
__queue_work ( per_cpu_ptr ( wq - > cpu_wq , cpu ) , work ) ;
2005-04-17 02:20:36 +04:00
ret = 1 ;
}
put_cpu ( ) ;
return ret ;
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL_GPL ( queue_work ) ;
2005-04-17 02:20:36 +04:00
[PATCH] Add debugging feature /proc/timer_stat
Add /proc/timer_stats support: debugging feature to profile timer expiration.
Both the starting site, process/PID and the expiration function is captured.
This allows the quick identification of timer event sources in a system.
Sample output:
# echo 1 > /proc/timer_stats
# cat /proc/timer_stats
Timer Stats Version: v0.1
Sample period: 4.010 s
24, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
11, 0 swapper sk_reset_timer (tcp_delack_timer)
6, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
17, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
4, 2050 pcscd do_nanosleep (hrtimer_wakeup)
5, 4179 sshd sk_reset_timer (tcp_write_timer)
4, 2248 yum-updatesd schedule_timeout (process_timeout)
18, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
3, 0 swapper sk_reset_timer (tcp_delack_timer)
1, 1 swapper neigh_table_init_no_netlink (neigh_periodic_timer)
2, 1 swapper e1000_up (e1000_watchdog)
1, 1 init schedule_timeout (process_timeout)
100 total events, 25.24 events/sec
[ cleanups and hrtimers support from Thomas Gleixner <tglx@linutronix.de> ]
[bunk@stusta.de: nr_entries can become static]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: john stultz <johnstul@us.ibm.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-16 12:28:13 +03:00
void delayed_work_timer_fn ( unsigned long __data )
2005-04-17 02:20:36 +04:00
{
2006-11-22 17:54:01 +03:00
struct delayed_work * dwork = ( struct delayed_work * ) __data ;
2006-11-22 17:54:49 +03:00
struct workqueue_struct * wq = get_wq_data ( & dwork - > work ) ;
2005-04-17 02:20:36 +04:00
int cpu = smp_processor_id ( ) ;
if ( unlikely ( is_single_threaded ( wq ) ) )
2006-01-08 12:05:12 +03:00
cpu = singlethread_cpu ;
2005-04-17 02:20:36 +04:00
2006-11-22 17:54:01 +03:00
__queue_work ( per_cpu_ptr ( wq - > cpu_wq , cpu ) , & dwork - > work ) ;
2005-04-17 02:20:36 +04:00
}
2006-07-30 14:03:42 +04:00
/**
* queue_delayed_work - queue work on a workqueue after delay
* @ wq : workqueue to use
2006-12-22 12:06:52 +03:00
* @ dwork : delayable work to queue
2006-07-30 14:03:42 +04:00
* @ delay : number of jiffies to wait before queueing
*
2006-10-28 21:38:58 +04:00
* Returns 0 if @ work was already on a queue , non - zero otherwise .
2006-07-30 14:03:42 +04:00
*/
2005-04-17 02:20:36 +04:00
int fastcall queue_delayed_work ( struct workqueue_struct * wq ,
2006-11-22 17:54:01 +03:00
struct delayed_work * dwork , unsigned long delay )
2005-04-17 02:20:36 +04:00
{
int ret = 0 ;
2006-11-22 17:54:01 +03:00
struct timer_list * timer = & dwork - > timer ;
struct work_struct * work = & dwork - > work ;
[PATCH] Add debugging feature /proc/timer_stat
Add /proc/timer_stats support: debugging feature to profile timer expiration.
Both the starting site, process/PID and the expiration function is captured.
This allows the quick identification of timer event sources in a system.
Sample output:
# echo 1 > /proc/timer_stats
# cat /proc/timer_stats
Timer Stats Version: v0.1
Sample period: 4.010 s
24, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
11, 0 swapper sk_reset_timer (tcp_delack_timer)
6, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
17, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
4, 2050 pcscd do_nanosleep (hrtimer_wakeup)
5, 4179 sshd sk_reset_timer (tcp_write_timer)
4, 2248 yum-updatesd schedule_timeout (process_timeout)
18, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
3, 0 swapper sk_reset_timer (tcp_delack_timer)
1, 1 swapper neigh_table_init_no_netlink (neigh_periodic_timer)
2, 1 swapper e1000_up (e1000_watchdog)
1, 1 init schedule_timeout (process_timeout)
100 total events, 25.24 events/sec
[ cleanups and hrtimers support from Thomas Gleixner <tglx@linutronix.de> ]
[bunk@stusta.de: nr_entries can become static]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: john stultz <johnstul@us.ibm.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-16 12:28:13 +03:00
timer_stats_timer_set_start_info ( timer ) ;
2006-11-22 17:54:01 +03:00
if ( delay = = 0 )
return queue_work ( wq , work ) ;
2005-04-17 02:20:36 +04:00
2006-12-16 20:53:50 +03:00
if ( ! test_and_set_bit ( WORK_STRUCT_PENDING , work_data_bits ( work ) ) ) {
2005-04-17 02:20:36 +04:00
BUG_ON ( timer_pending ( timer ) ) ;
BUG_ON ( ! list_empty ( & work - > entry ) ) ;
/* This stores wq for the moment, for the timer_fn */
2006-11-22 17:54:49 +03:00
set_wq_data ( work , wq ) ;
2005-04-17 02:20:36 +04:00
timer - > expires = jiffies + delay ;
2006-11-22 17:54:01 +03:00
timer - > data = ( unsigned long ) dwork ;
2005-04-17 02:20:36 +04:00
timer - > function = delayed_work_timer_fn ;
add_timer ( timer ) ;
ret = 1 ;
}
return ret ;
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL_GPL ( queue_delayed_work ) ;
2005-04-17 02:20:36 +04:00
2006-07-30 14:03:42 +04:00
/**
* 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 12:06:52 +03:00
* @ dwork : work to queue
2006-07-30 14:03:42 +04:00
* @ delay : number of jiffies to wait before queueing
*
2006-10-28 21:38:58 +04:00
* Returns 0 if @ work was already on a queue , non - zero otherwise .
2006-07-30 14:03:42 +04:00
*/
2006-06-29 00:50:33 +04:00
int queue_delayed_work_on ( int cpu , struct workqueue_struct * wq ,
2006-11-22 17:54:01 +03:00
struct delayed_work * dwork , unsigned long delay )
2006-06-29 00:50:33 +04:00
{
int ret = 0 ;
2006-11-22 17:54:01 +03:00
struct timer_list * timer = & dwork - > timer ;
struct work_struct * work = & dwork - > work ;
2006-06-29 00:50:33 +04:00
2006-12-16 20:53:50 +03:00
if ( ! test_and_set_bit ( WORK_STRUCT_PENDING , work_data_bits ( work ) ) ) {
2006-06-29 00:50:33 +04:00
BUG_ON ( timer_pending ( timer ) ) ;
BUG_ON ( ! list_empty ( & work - > entry ) ) ;
/* This stores wq for the moment, for the timer_fn */
2006-11-22 17:54:49 +03:00
set_wq_data ( work , wq ) ;
2006-06-29 00:50:33 +04:00
timer - > expires = jiffies + delay ;
2006-11-22 17:54:01 +03:00
timer - > data = ( unsigned long ) dwork ;
2006-06-29 00:50:33 +04:00
timer - > function = delayed_work_timer_fn ;
add_timer_on ( timer , cpu ) ;
ret = 1 ;
}
return ret ;
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL_GPL ( queue_delayed_work_on ) ;
2005-04-17 02:20:36 +04:00
2006-01-15 00:20:43 +03:00
static void run_workqueue ( struct cpu_workqueue_struct * cwq )
2005-04-17 02:20:36 +04:00
{
2007-05-09 13:34:10 +04:00
spin_lock_irq ( & cwq - > lock ) ;
2005-04-17 02:20:36 +04:00
cwq - > run_depth + + ;
if ( cwq - > run_depth > 3 ) {
/* morton gets to eat his hat */
printk ( " %s: recursion depth exceeded: %d \n " ,
__FUNCTION__ , cwq - > run_depth ) ;
dump_stack ( ) ;
}
while ( ! list_empty ( & cwq - > worklist ) ) {
struct work_struct * work = list_entry ( cwq - > worklist . next ,
struct work_struct , entry ) ;
2006-11-22 17:54:45 +03:00
work_func_t f = work - > func ;
2005-04-17 02:20:36 +04:00
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 13:33:52 +04:00
cwq - > current_work = work ;
2005-04-17 02:20:36 +04:00
list_del_init ( cwq - > worklist . next ) ;
2007-05-09 13:34:10 +04:00
spin_unlock_irq ( & cwq - > lock ) ;
2005-04-17 02:20:36 +04:00
2006-11-22 17:54:49 +03:00
BUG_ON ( get_wq_data ( work ) ! = cwq ) ;
2006-12-16 20:53:50 +03:00
if ( ! test_bit ( WORK_STRUCT_NOAUTOREL , work_data_bits ( work ) ) )
2006-11-22 17:55:48 +03:00
work_release ( work ) ;
f ( work ) ;
2005-04-17 02:20:36 +04:00
2006-12-07 07:37:26 +03:00
if ( unlikely ( in_atomic ( ) | | lockdep_depth ( current ) > 0 ) ) {
printk ( KERN_ERR " BUG: workqueue leaked lock or atomic: "
" %s/0x%08x/%d \n " ,
current - > comm , preempt_count ( ) ,
current - > pid ) ;
printk ( KERN_ERR " last function: " ) ;
print_symbol ( " %s \n " , ( unsigned long ) f ) ;
debug_show_held_locks ( current ) ;
dump_stack ( ) ;
}
2007-05-09 13:34:10 +04:00
spin_lock_irq ( & cwq - > lock ) ;
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 13:33:52 +04:00
cwq - > current_work = NULL ;
2005-04-17 02:20:36 +04:00
}
cwq - > run_depth - - ;
2007-05-09 13:34:10 +04:00
spin_unlock_irq ( & cwq - > lock ) ;
2005-04-17 02:20:36 +04:00
}
2007-05-09 13:34:09 +04:00
/*
* NOTE : the caller must not touch * cwq if this func returns true
*/
static int cwq_should_stop ( struct cpu_workqueue_struct * cwq )
{
int should_stop = cwq - > should_stop ;
if ( unlikely ( should_stop ) ) {
spin_lock_irq ( & cwq - > lock ) ;
should_stop = cwq - > should_stop & & list_empty ( & cwq - > worklist ) ;
if ( should_stop )
cwq - > thread = NULL ;
spin_unlock_irq ( & cwq - > lock ) ;
}
return should_stop ;
}
2005-04-17 02:20:36 +04:00
static int worker_thread ( void * __cwq )
{
struct cpu_workqueue_struct * cwq = __cwq ;
2007-05-09 13:34:09 +04:00
DEFINE_WAIT ( wait ) ;
2005-04-17 02:20:36 +04:00
struct k_sigaction sa ;
sigset_t blocked ;
2007-05-09 13:34:06 +04:00
if ( ! cwq - > wq - > freezeable )
2006-12-07 07:34:49 +03:00
current - > flags | = PF_NOFREEZE ;
2005-04-17 02:20:36 +04:00
set_user_nice ( current , - 5 ) ;
/* Block and flush all signals */
sigfillset ( & blocked ) ;
sigprocmask ( SIG_BLOCK , & blocked , NULL ) ;
flush_signals ( current ) ;
2006-10-11 12:21:26 +04:00
/*
* We inherited MPOL_INTERLEAVE from the booting kernel .
* Set MPOL_DEFAULT to insure node local allocations .
*/
numa_default_policy ( ) ;
2005-04-17 02:20:36 +04:00
/* SIG_IGN makes children autoreap: see do_notify_parent(). */
sa . sa . sa_handler = SIG_IGN ;
sa . sa . sa_flags = 0 ;
siginitset ( & sa . sa . sa_mask , sigmask ( SIGCHLD ) ) ;
do_sigaction ( SIGCHLD , & sa , ( struct k_sigaction * ) 0 ) ;
2007-05-09 13:34:09 +04:00
for ( ; ; ) {
2007-05-09 13:34:06 +04:00
if ( cwq - > wq - > freezeable )
2006-12-07 07:34:49 +03:00
try_to_freeze ( ) ;
2007-05-09 13:34:09 +04:00
prepare_to_wait ( & cwq - > more_work , & wait , TASK_INTERRUPTIBLE ) ;
if ( ! cwq - > should_stop & & list_empty ( & cwq - > worklist ) )
2005-04-17 02:20:36 +04:00
schedule ( ) ;
2007-05-09 13:34:09 +04:00
finish_wait ( & cwq - > more_work , & wait ) ;
if ( cwq_should_stop ( cwq ) )
break ;
2005-04-17 02:20:36 +04:00
2007-05-09 13:34:09 +04:00
run_workqueue ( cwq ) ;
2005-04-17 02:20:36 +04:00
}
2007-05-09 13:34:09 +04:00
2005-04-17 02:20:36 +04:00
return 0 ;
}
2007-05-09 13:33:51 +04:00
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 ) ;
}
2007-05-09 13:33:54 +04:00
static void insert_wq_barrier ( struct cpu_workqueue_struct * cwq ,
struct wq_barrier * barr , int tail )
2007-05-09 13:33:51 +04:00
{
INIT_WORK ( & barr - > work , wq_barrier_func ) ;
__set_bit ( WORK_STRUCT_PENDING , work_data_bits ( & barr - > work ) ) ;
init_completion ( & barr - > done ) ;
2007-05-09 13:33:54 +04:00
insert_work ( cwq , & barr - > work , tail ) ;
2007-05-09 13:33:51 +04:00
}
2005-04-17 02:20:36 +04:00
static void flush_cpu_workqueue ( struct cpu_workqueue_struct * cwq )
{
if ( cwq - > thread = = current ) {
/*
* Probably keventd trying to flush its own queue . So simply run
* it by hand rather than deadlocking .
*/
run_workqueue ( cwq ) ;
} else {
2007-05-09 13:33:51 +04:00
struct wq_barrier barr ;
2007-05-09 13:33:54 +04:00
int active = 0 ;
2005-04-17 02:20:36 +04:00
2007-05-09 13:33:54 +04:00
spin_lock_irq ( & cwq - > lock ) ;
if ( ! list_empty ( & cwq - > worklist ) | | cwq - > current_work ! = NULL ) {
insert_wq_barrier ( cwq , & barr , 1 ) ;
active = 1 ;
}
spin_unlock_irq ( & cwq - > lock ) ;
2005-04-17 02:20:36 +04:00
workqueue: fix flush_workqueue() vs CPU_DEAD race
Many thanks to Srivatsa Vaddagiri for the helpful discussion and for spotting
the bug in my previous attempt.
work->func() (and thus flush_workqueue()) must not use workqueue_mutex,
this leads to deadlock when CPU_DEAD does kthread_stop(). However without
this mutex held we can't detect CPU_DEAD in progress, which can move pending
works to another CPU while the dead one is not on cpu_online_map.
Change flush_workqueue() to use for_each_possible_cpu(). This means that
flush_cpu_workqueue() may hit CPU which is already dead. However in that
case
!list_empty(&cwq->worklist) || cwq->current_work != NULL
means that CPU_DEAD in progress, it will do kthread_stop() + take_over_work()
so we can proceed and insert a barrier. We hold cwq->lock, so we are safe.
Also, add migrate_sequence incremented by take_over_work() under cwq->lock.
If take_over_work() happened before we checked this CPU, we should see the
new value after spin_unlock().
Further possible changes:
remove CPU_DEAD handling (along with take_over_work, migrate_sequence)
from workqueue.c. CPU_DEAD just sets cwq->please_exit_after_flush flag.
CPU_UP_PREPARE->create_workqueue_thread() clears this flag, and creates
the new thread if cwq->thread == NULL.
This way the workqueue/cpu-hotplug interaction is almost zero, workqueue_mutex
just protects "workqueues" list, CPU_LOCK_ACQUIRE/CPU_LOCK_RELEASE go away.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: "Pallipadi, Venkatesh" <venkatesh.pallipadi@intel.com>
Cc: Gautham shenoy <ego@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-05-09 13:34:07 +04:00
if ( active )
2007-05-09 13:33:54 +04:00
wait_for_completion ( & barr . done ) ;
2005-04-17 02:20:36 +04:00
}
}
2006-07-30 14:03:42 +04:00
/**
2005-04-17 02:20:36 +04:00
* flush_workqueue - ensure that any scheduled work has run to completion .
2006-07-30 14:03:42 +04:00
* @ wq : workqueue to flush
2005-04-17 02:20:36 +04:00
*
* Forces execution of the workqueue and blocks until its completion .
* This is typically used in driver shutdown handlers .
*
2007-05-09 13:33:51 +04:00
* We sleep until all works which were queued on entry have been handled ,
* but we are not livelocked by new incoming ones .
2005-04-17 02:20:36 +04:00
*
* This function used to run the workqueues itself . Now we just wait for the
* helper threads to do it .
*/
void fastcall flush_workqueue ( struct workqueue_struct * wq )
{
2007-05-09 13:34:12 +04:00
const cpumask_t * cpu_map = wq_cpu_map ( wq ) ;
2007-05-09 13:34:13 +04:00
int cpu ;
2005-04-17 02:20:36 +04:00
2007-05-09 13:34:12 +04:00
might_sleep ( ) ;
for_each_cpu_mask ( cpu , * cpu_map )
flush_cpu_workqueue ( per_cpu_ptr ( wq - > cpu_wq , cpu ) ) ;
2005-04-17 02:20:36 +04:00
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL_GPL ( flush_workqueue ) ;
2005-04-17 02:20:36 +04:00
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 13:33:52 +04:00
static void wait_on_work ( struct cpu_workqueue_struct * cwq ,
struct work_struct * work )
{
struct wq_barrier barr ;
int running = 0 ;
spin_lock_irq ( & cwq - > lock ) ;
if ( unlikely ( cwq - > current_work = = work ) ) {
2007-05-09 13:33:54 +04:00
insert_wq_barrier ( cwq , & barr , 0 ) ;
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 13:33:52 +04:00
running = 1 ;
}
spin_unlock_irq ( & cwq - > lock ) ;
2007-05-09 13:34:09 +04:00
if ( unlikely ( running ) )
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 13:33:52 +04:00
wait_for_completion ( & barr . done ) ;
}
/**
* flush_work - block until a work_struct ' s callback has terminated
* @ wq : the workqueue on which the work is queued
* @ work : the work which is to be flushed
*
* flush_work ( ) will attempt to cancel the work if it is queued . If the work ' s
* callback appears to be running , flush_work ( ) will block until it has
* completed .
*
* flush_work ( ) is designed to be used when the caller is tearing down data
* structures which the callback function operates upon . It is expected that ,
* prior to calling flush_work ( ) , the caller has arranged for the work to not
* be requeued .
*/
void flush_work ( struct workqueue_struct * wq , struct work_struct * work )
{
2007-05-09 13:34:12 +04:00
const cpumask_t * cpu_map = wq_cpu_map ( wq ) ;
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 13:33:52 +04:00
struct cpu_workqueue_struct * cwq ;
2007-05-09 13:34:12 +04:00
int cpu ;
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 13:33:52 +04:00
2007-05-09 13:34:10 +04:00
might_sleep ( ) ;
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 13:33:52 +04:00
cwq = get_wq_data ( work ) ;
/* Was it ever queued ? */
if ( ! cwq )
2007-05-09 13:34:09 +04:00
return ;
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 13:33:52 +04:00
/*
2007-05-09 13:34:09 +04:00
* This work can ' t be re - queued , no need to re - check that
* get_wq_data ( ) is still the same when we take cwq - > lock .
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 13:33:52 +04:00
*/
spin_lock_irq ( & cwq - > lock ) ;
list_del_init ( & work - > entry ) ;
work_release ( work ) ;
spin_unlock_irq ( & cwq - > lock ) ;
2007-05-09 13:34:12 +04:00
for_each_cpu_mask ( cpu , * cpu_map )
wait_on_work ( per_cpu_ptr ( wq - > cpu_wq , cpu ) , work ) ;
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 13:33:52 +04:00
}
EXPORT_SYMBOL_GPL ( flush_work ) ;
2005-04-17 02:20:36 +04:00
static struct workqueue_struct * keventd_wq ;
2006-07-30 14:03:42 +04:00
/**
* schedule_work - put work task in global workqueue
* @ work : job to be done
*
* This puts a job in the kernel - global workqueue .
*/
2005-04-17 02:20:36 +04:00
int fastcall schedule_work ( struct work_struct * work )
{
return queue_work ( keventd_wq , work ) ;
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL ( schedule_work ) ;
2005-04-17 02:20:36 +04:00
2006-07-30 14:03:42 +04:00
/**
* schedule_delayed_work - put work task in global workqueue after delay
2006-11-22 17:54:01 +03:00
* @ dwork : job to be done
* @ delay : number of jiffies to wait or 0 for immediate execution
2006-07-30 14:03:42 +04:00
*
* After waiting for a given time this puts a job in the kernel - global
* workqueue .
*/
[PATCH] Add debugging feature /proc/timer_stat
Add /proc/timer_stats support: debugging feature to profile timer expiration.
Both the starting site, process/PID and the expiration function is captured.
This allows the quick identification of timer event sources in a system.
Sample output:
# echo 1 > /proc/timer_stats
# cat /proc/timer_stats
Timer Stats Version: v0.1
Sample period: 4.010 s
24, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
11, 0 swapper sk_reset_timer (tcp_delack_timer)
6, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
17, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
4, 2050 pcscd do_nanosleep (hrtimer_wakeup)
5, 4179 sshd sk_reset_timer (tcp_write_timer)
4, 2248 yum-updatesd schedule_timeout (process_timeout)
18, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
3, 0 swapper sk_reset_timer (tcp_delack_timer)
1, 1 swapper neigh_table_init_no_netlink (neigh_periodic_timer)
2, 1 swapper e1000_up (e1000_watchdog)
1, 1 init schedule_timeout (process_timeout)
100 total events, 25.24 events/sec
[ cleanups and hrtimers support from Thomas Gleixner <tglx@linutronix.de> ]
[bunk@stusta.de: nr_entries can become static]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: john stultz <johnstul@us.ibm.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-16 12:28:13 +03:00
int fastcall schedule_delayed_work ( struct delayed_work * dwork ,
unsigned long delay )
2005-04-17 02:20:36 +04:00
{
[PATCH] Add debugging feature /proc/timer_stat
Add /proc/timer_stats support: debugging feature to profile timer expiration.
Both the starting site, process/PID and the expiration function is captured.
This allows the quick identification of timer event sources in a system.
Sample output:
# echo 1 > /proc/timer_stats
# cat /proc/timer_stats
Timer Stats Version: v0.1
Sample period: 4.010 s
24, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
11, 0 swapper sk_reset_timer (tcp_delack_timer)
6, 0 swapper hrtimer_stop_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
17, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
2, 1 swapper queue_delayed_work_on (delayed_work_timer_fn)
4, 2050 pcscd do_nanosleep (hrtimer_wakeup)
5, 4179 sshd sk_reset_timer (tcp_write_timer)
4, 2248 yum-updatesd schedule_timeout (process_timeout)
18, 0 swapper hrtimer_restart_sched_tick (hrtimer_sched_tick)
3, 0 swapper sk_reset_timer (tcp_delack_timer)
1, 1 swapper neigh_table_init_no_netlink (neigh_periodic_timer)
2, 1 swapper e1000_up (e1000_watchdog)
1, 1 init schedule_timeout (process_timeout)
100 total events, 25.24 events/sec
[ cleanups and hrtimers support from Thomas Gleixner <tglx@linutronix.de> ]
[bunk@stusta.de: nr_entries can become static]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: john stultz <johnstul@us.ibm.com>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-02-16 12:28:13 +03:00
timer_stats_timer_set_start_info ( & dwork - > timer ) ;
2006-11-22 17:54:01 +03:00
return queue_delayed_work ( keventd_wq , dwork , delay ) ;
2005-04-17 02:20:36 +04:00
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL ( schedule_delayed_work ) ;
2005-04-17 02:20:36 +04:00
2006-07-30 14:03:42 +04:00
/**
* schedule_delayed_work_on - queue work in global workqueue on CPU after delay
* @ cpu : cpu to use
2006-11-22 17:54:01 +03:00
* @ dwork : job to be done
2006-07-30 14:03:42 +04:00
* @ delay : number of jiffies to wait
*
* After waiting for a given time this puts a job in the kernel - global
* workqueue on the specified CPU .
*/
2005-04-17 02:20:36 +04:00
int schedule_delayed_work_on ( int cpu ,
2006-11-22 17:54:01 +03:00
struct delayed_work * dwork , unsigned long delay )
2005-04-17 02:20:36 +04:00
{
2006-11-22 17:54:01 +03:00
return queue_delayed_work_on ( cpu , keventd_wq , dwork , delay ) ;
2005-04-17 02:20:36 +04:00
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL ( schedule_delayed_work_on ) ;
2005-04-17 02:20:36 +04:00
2006-06-25 16:47:49 +04:00
/**
* schedule_on_each_cpu - call a function on each online CPU from keventd
* @ func : the function to call
*
* Returns zero on success .
* Returns - ve errno on failure .
*
* Appears to be racy against CPU hotplug .
*
* schedule_on_each_cpu ( ) is very slow .
*/
2006-11-22 17:55:48 +03:00
int schedule_on_each_cpu ( work_func_t func )
2006-01-08 12:00:43 +03:00
{
int cpu ;
2006-06-25 16:47:49 +04:00
struct work_struct * works ;
2006-01-08 12:00:43 +03:00
2006-06-25 16:47:49 +04:00
works = alloc_percpu ( struct work_struct ) ;
if ( ! works )
2006-01-08 12:00:43 +03:00
return - ENOMEM ;
2006-06-25 16:47:49 +04:00
2007-05-09 13:33:50 +04:00
preempt_disable ( ) ; /* CPU hotplug */
2006-01-08 12:00:43 +03:00
for_each_online_cpu ( cpu ) {
2006-12-18 22:05:09 +03:00
struct work_struct * work = per_cpu_ptr ( works , cpu ) ;
INIT_WORK ( work , func ) ;
set_bit ( WORK_STRUCT_PENDING , work_data_bits ( work ) ) ;
__queue_work ( per_cpu_ptr ( keventd_wq - > cpu_wq , cpu ) , work ) ;
2006-01-08 12:00:43 +03:00
}
2007-05-09 13:33:50 +04:00
preempt_enable ( ) ;
2006-01-08 12:00:43 +03:00
flush_workqueue ( keventd_wq ) ;
2006-06-25 16:47:49 +04:00
free_percpu ( works ) ;
2006-01-08 12:00:43 +03:00
return 0 ;
}
2005-04-17 02:20:36 +04:00
void flush_scheduled_work ( void )
{
flush_workqueue ( keventd_wq ) ;
}
2006-06-30 09:40:45 +04:00
EXPORT_SYMBOL ( flush_scheduled_work ) ;
2005-04-17 02:20:36 +04:00
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 13:33:52 +04:00
void flush_work_keventd ( struct work_struct * work )
{
flush_work ( keventd_wq , work ) ;
}
EXPORT_SYMBOL ( flush_work_keventd ) ;
2005-04-17 02:20:36 +04:00
/**
2007-02-10 12:45:59 +03:00
* cancel_rearming_delayed_workqueue - reliably kill off a delayed work whose handler rearms the delayed work .
2005-04-17 02:20:36 +04:00
* @ wq : the controlling workqueue structure
2006-11-22 17:54:01 +03:00
* @ dwork : the delayed work struct
2005-04-17 02:20:36 +04:00
*/
2005-04-17 02:23:59 +04:00
void cancel_rearming_delayed_workqueue ( struct workqueue_struct * wq ,
2006-11-22 17:54:01 +03:00
struct delayed_work * dwork )
2005-04-17 02:20:36 +04:00
{
2007-05-09 13:34:11 +04:00
/* Was it ever queued ? */
if ( ! get_wq_data ( & dwork - > work ) )
return ;
2006-11-22 17:54:01 +03:00
while ( ! cancel_delayed_work ( dwork ) )
2005-04-17 02:20:36 +04:00
flush_workqueue ( wq ) ;
}
2005-04-17 02:23:59 +04:00
EXPORT_SYMBOL ( cancel_rearming_delayed_workqueue ) ;
2005-04-17 02:20:36 +04:00
/**
2007-02-10 12:45:59 +03:00
* cancel_rearming_delayed_work - reliably kill off a delayed keventd work whose handler rearms the delayed work .
2006-11-22 17:54:01 +03:00
* @ dwork : the delayed work struct
2005-04-17 02:20:36 +04:00
*/
2006-11-22 17:54:01 +03:00
void cancel_rearming_delayed_work ( struct delayed_work * dwork )
2005-04-17 02:20:36 +04:00
{
2006-11-22 17:54:01 +03:00
cancel_rearming_delayed_workqueue ( keventd_wq , dwork ) ;
2005-04-17 02:20:36 +04:00
}
EXPORT_SYMBOL ( cancel_rearming_delayed_work ) ;
2006-02-23 21:43:43 +03:00
/**
* 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 .
*
* Returns : 0 - function was executed
* 1 - function was scheduled for execution
*/
2006-11-22 17:55:48 +03:00
int execute_in_process_context ( work_func_t fn , struct execute_work * ew )
2006-02-23 21:43:43 +03:00
{
if ( ! in_interrupt ( ) ) {
2006-11-22 17:55:48 +03:00
fn ( & ew - > work ) ;
2006-02-23 21:43:43 +03:00
return 0 ;
}
2006-11-22 17:55:48 +03:00
INIT_WORK ( & ew - > work , fn ) ;
2006-02-23 21:43:43 +03:00
schedule_work ( & ew - > work ) ;
return 1 ;
}
EXPORT_SYMBOL_GPL ( execute_in_process_context ) ;
2005-04-17 02:20:36 +04:00
int keventd_up ( void )
{
return keventd_wq ! = NULL ;
}
int current_is_keventd ( void )
{
struct cpu_workqueue_struct * cwq ;
int cpu = smp_processor_id ( ) ; /* preempt-safe: keventd is per-cpu */
int ret = 0 ;
BUG_ON ( ! keventd_wq ) ;
2005-10-31 02:01:59 +03:00
cwq = per_cpu_ptr ( keventd_wq - > cpu_wq , cpu ) ;
2005-04-17 02:20:36 +04:00
if ( current = = cwq - > thread )
ret = 1 ;
return ret ;
}
2007-05-09 13:34:09 +04:00
static struct cpu_workqueue_struct *
init_cpu_workqueue ( struct workqueue_struct * wq , int cpu )
2005-04-17 02:20:36 +04:00
{
2005-10-31 02:01:59 +03:00
struct cpu_workqueue_struct * cwq = per_cpu_ptr ( wq - > cpu_wq , cpu ) ;
2005-04-17 02:20:36 +04:00
2007-05-09 13:34:09 +04:00
cwq - > wq = wq ;
spin_lock_init ( & cwq - > lock ) ;
INIT_LIST_HEAD ( & cwq - > worklist ) ;
init_waitqueue_head ( & cwq - > more_work ) ;
return cwq ;
2005-04-17 02:20:36 +04:00
}
2007-05-09 13:34:09 +04:00
static int create_workqueue_thread ( struct cpu_workqueue_struct * cwq , int cpu )
{
struct workqueue_struct * wq = cwq - > wq ;
const char * fmt = is_single_threaded ( wq ) ? " %s " : " %s/%d " ;
struct task_struct * p ;
p = kthread_create ( worker_thread , cwq , fmt , wq - > name , cpu ) ;
/*
* Nobody can add the work_struct to this cwq ,
* if ( caller is __create_workqueue )
* nobody should see this wq
* else // caller is CPU_UP_PREPARE
* cpu is not on cpu_online_map
* so we can abort safely .
*/
if ( IS_ERR ( p ) )
return PTR_ERR ( p ) ;
cwq - > thread = p ;
cwq - > should_stop = 0 ;
if ( ! is_single_threaded ( wq ) )
kthread_bind ( p , cpu ) ;
if ( is_single_threaded ( wq ) | | cpu_online ( cpu ) )
wake_up_process ( p ) ;
return 0 ;
}
struct workqueue_struct * __create_workqueue ( const char * name ,
int singlethread , int freezeable )
2005-04-17 02:20:36 +04:00
{
struct workqueue_struct * wq ;
2007-05-09 13:34:09 +04:00
struct cpu_workqueue_struct * cwq ;
int err = 0 , cpu ;
2005-04-17 02:20:36 +04:00
2007-05-09 13:34:09 +04:00
wq = kzalloc ( sizeof ( * wq ) , GFP_KERNEL ) ;
if ( ! wq )
return NULL ;
wq - > cpu_wq = alloc_percpu ( struct cpu_workqueue_struct ) ;
if ( ! wq - > cpu_wq ) {
kfree ( wq ) ;
return NULL ;
}
wq - > name = name ;
2007-05-09 13:34:13 +04:00
wq - > singlethread = singlethread ;
2007-05-09 13:34:09 +04:00
wq - > freezeable = freezeable ;
2007-05-09 13:34:13 +04:00
INIT_LIST_HEAD ( & wq - > list ) ;
2007-05-09 13:34:09 +04:00
if ( singlethread ) {
cwq = init_cpu_workqueue ( wq , singlethread_cpu ) ;
err = create_workqueue_thread ( cwq , singlethread_cpu ) ;
} else {
2006-08-14 10:24:26 +04:00
mutex_lock ( & workqueue_mutex ) ;
2007-05-09 13:34:09 +04:00
list_add ( & wq - > list , & workqueues ) ;
for_each_possible_cpu ( cpu ) {
cwq = init_cpu_workqueue ( wq , cpu ) ;
if ( err | | ! cpu_online ( cpu ) )
continue ;
err = create_workqueue_thread ( cwq , cpu ) ;
2005-04-17 02:20:36 +04:00
}
2007-05-09 13:34:09 +04:00
mutex_unlock ( & workqueue_mutex ) ;
}
if ( err ) {
destroy_workqueue ( wq ) ;
wq = NULL ;
}
return wq ;
}
EXPORT_SYMBOL_GPL ( __create_workqueue ) ;
2005-04-17 02:20:36 +04:00
2007-05-09 13:34:09 +04:00
static void cleanup_workqueue_thread ( struct cpu_workqueue_struct * cwq , int cpu )
{
struct wq_barrier barr ;
int alive = 0 ;
2005-10-31 02:01:59 +03:00
2007-05-09 13:34:09 +04:00
spin_lock_irq ( & cwq - > lock ) ;
if ( cwq - > thread ! = NULL ) {
insert_wq_barrier ( cwq , & barr , 1 ) ;
cwq - > should_stop = 1 ;
alive = 1 ;
}
spin_unlock_irq ( & cwq - > lock ) ;
if ( alive ) {
wait_for_completion ( & barr . done ) ;
while ( unlikely ( cwq - > thread ! = NULL ) )
cpu_relax ( ) ;
/*
* Wait until cwq - > thread unlocks cwq - > lock ,
* it won ' t touch * cwq after that .
*/
smp_rmb ( ) ;
spin_unlock_wait ( & cwq - > lock ) ;
}
}
/**
* 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 )
{
2007-05-09 13:34:12 +04:00
const cpumask_t * cpu_map = wq_cpu_map ( wq ) ;
2007-05-09 13:34:09 +04:00
struct cpu_workqueue_struct * cwq ;
2007-05-09 13:34:12 +04:00
int cpu ;
2007-05-09 13:34:09 +04:00
2007-05-09 13:34:12 +04:00
mutex_lock ( & workqueue_mutex ) ;
list_del ( & wq - > list ) ;
mutex_unlock ( & workqueue_mutex ) ;
2007-05-09 13:34:09 +04:00
2007-05-09 13:34:12 +04:00
for_each_cpu_mask ( cpu , * cpu_map ) {
cwq = per_cpu_ptr ( wq - > cpu_wq , cpu ) ;
cleanup_workqueue_thread ( cwq , cpu ) ;
2007-05-09 13:34:09 +04:00
}
2006-08-14 10:24:26 +04:00
2007-05-09 13:34:09 +04:00
free_percpu ( wq - > cpu_wq ) ;
kfree ( wq ) ;
}
EXPORT_SYMBOL_GPL ( destroy_workqueue ) ;
static int __devinit workqueue_cpu_callback ( struct notifier_block * nfb ,
unsigned long action ,
void * hcpu )
{
unsigned int cpu = ( unsigned long ) hcpu ;
struct cpu_workqueue_struct * cwq ;
struct workqueue_struct * wq ;
switch ( action ) {
case CPU_LOCK_ACQUIRE :
2006-08-14 10:24:26 +04:00
mutex_lock ( & workqueue_mutex ) ;
2007-05-09 13:34:09 +04:00
return NOTIFY_OK ;
2006-08-14 10:24:26 +04:00
2007-05-09 13:34:09 +04:00
case CPU_LOCK_RELEASE :
2006-08-14 10:24:26 +04:00
mutex_unlock ( & workqueue_mutex ) ;
2007-05-09 13:34:09 +04:00
return NOTIFY_OK ;
2005-04-17 02:20:36 +04:00
2007-05-09 13:34:09 +04:00
case CPU_UP_PREPARE :
cpu_set ( cpu , cpu_populated_map ) ;
}
list_for_each_entry ( wq , & workqueues , list ) {
cwq = per_cpu_ptr ( wq - > cpu_wq , cpu ) ;
switch ( action ) {
case CPU_UP_PREPARE :
if ( ! create_workqueue_thread ( cwq , cpu ) )
break ;
printk ( KERN_ERR " workqueue for %i failed \n " , cpu ) ;
return NOTIFY_BAD ;
case CPU_ONLINE :
wake_up_process ( cwq - > thread ) ;
break ;
case CPU_UP_CANCELED :
if ( cwq - > thread )
wake_up_process ( cwq - > thread ) ;
case CPU_DEAD :
cleanup_workqueue_thread ( cwq , cpu ) ;
break ;
}
2005-04-17 02:20:36 +04:00
}
return NOTIFY_OK ;
}
void init_workqueues ( void )
{
2007-05-09 13:34:09 +04:00
cpu_populated_map = cpu_online_map ;
2006-01-08 12:05:12 +03:00
singlethread_cpu = first_cpu ( cpu_possible_map ) ;
2007-05-09 13:34:12 +04:00
cpu_singlethread_map = cpumask_of_cpu ( singlethread_cpu ) ;
2005-04-17 02:20:36 +04:00
hotcpu_notifier ( workqueue_cpu_callback , 0 ) ;
keventd_wq = create_workqueue ( " events " ) ;
BUG_ON ( ! keventd_wq ) ;
}