commit 8603e1b300 upstream.
cancel[_delayed]_work_sync() are implemented using
__cancel_work_timer() which grabs the PENDING bit using
try_to_grab_pending() and then flushes the work item with PENDING set
to prevent the on-going execution of the work item from requeueing
itself.
try_to_grab_pending() can always grab PENDING bit without blocking
except when someone else is doing the above flushing during
cancelation. In that case, try_to_grab_pending() returns -ENOENT. In
this case, __cancel_work_timer() currently invokes flush_work(). The
assumption is that the completion of the work item is what the other
canceling task would be waiting for too and thus waiting for the same
condition and retrying should allow forward progress without excessive
busy looping
Unfortunately, this doesn't work if preemption is disabled or the
latter task has real time priority. Let's say task A just got woken
up from flush_work() by the completion of the target work item. If,
before task A starts executing, task B gets scheduled and invokes
__cancel_work_timer() on the same work item, its try_to_grab_pending()
will return -ENOENT as the work item is still being canceled by task A
and flush_work() will also immediately return false as the work item
is no longer executing. This puts task B in a busy loop possibly
preventing task A from executing and clearing the canceling state on
the work item leading to a hang.
task A task B worker
executing work
__cancel_work_timer()
try_to_grab_pending()
set work CANCELING
flush_work()
block for work completion
completion, wakes up A
__cancel_work_timer()
while (forever) {
try_to_grab_pending()
-ENOENT as work is being canceled
flush_work()
false as work is no longer executing
}
This patch removes the possible hang by updating __cancel_work_timer()
to explicitly wait for clearing of CANCELING rather than invoking
flush_work() after try_to_grab_pending() fails with -ENOENT.
Link: http://lkml.kernel.org/g/20150206171156.GA8942@axis.com
v3: bit_waitqueue() can't be used for work items defined in vmalloc
area. Switched to custom wake function which matches the target
work item and exclusive wait and wakeup.
v2: v1 used wake_up() on bit_waitqueue() which leads to NULL deref if
the target bit waitqueue has wait_bit_queue's on it. Use
DEFINE_WAIT_BIT() and __wake_up_bit() instead. Reported by Tomeu
Vizoso.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Rabin Vincent <rabin.vincent@axis.com>
Cc: Tomeu Vizoso <tomeu.vizoso@gmail.com>
Tested-by: Jesper Nilsson <jesper.nilsson@axis.com>
Tested-by: Rabin Vincent <rabin.vincent@axis.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 29187a9eea upstream.
A worker_pool's forward progress is guaranteed by the fact that the
last idle worker assumes the manager role to create more workers and
summon the rescuers if creating workers doesn't succeed in timely
manner before proceeding to execute work items.
This manager role is implemented in manage_workers(), which indicates
whether the worker may proceed to work item execution with its return
value. This is necessary because multiple workers may contend for the
manager role, and, if there already is a manager, others should
proceed to work item execution.
Unfortunately, the function also indicates that the worker may proceed
to work item execution if need_to_create_worker() is false at the head
of the function. need_to_create_worker() tests the following
conditions.
pending work items && !nr_running && !nr_idle
The first and third conditions are protected by pool->lock and thus
won't change while holding pool->lock; however, nr_running can change
asynchronously as other workers block and resume and while it's likely
to be zero, as someone woke this worker up in the first place, some
other workers could have become runnable inbetween making it non-zero.
If this happens, manage_worker() could return false even with zero
nr_idle making the worker, the last idle one, proceed to execute work
items. If then all workers of the pool end up blocking on a resource
which can only be released by a work item which is pending on that
pool, the whole pool can deadlock as there's no one to create more
workers or summon the rescuers.
This patch fixes the problem by removing the early exit condition from
maybe_create_worker() and making manage_workers() return false iff
there's already another manager, which ensures that the last worker
doesn't start executing work items.
We can leave the early exit condition alone and just ignore the return
value but the only reason it was put there is because the
manage_workers() used to perform both creations and destructions of
workers and thus the function may be invoked while the pool is trying
to reduce the number of workers. Now that manage_workers() is called
only when more workers are needed, the only case this early exit
condition is triggered is rare race conditions rendering it pointless.
Tested with simulated workload and modified workqueue code which
trigger the pool deadlock reliably without this patch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Eric Sandeen <sandeen@sandeen.net>
Link: http://lkml.kernel.org/g/54B019F4.8030009@sandeen.net
Cc: Dave Chinner <david@fromorbit.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 5a6024f160 upstream.
When hot-adding and onlining CPU, kernel panic occurs, showing following
call trace.
BUG: unable to handle kernel paging request at 0000000000001d08
IP: [<ffffffff8114acfd>] __alloc_pages_nodemask+0x9d/0xb10
PGD 0
Oops: 0000 [#1] SMP
...
Call Trace:
[<ffffffff812b8745>] ? cpumask_next_and+0x35/0x50
[<ffffffff810a3283>] ? find_busiest_group+0x113/0x8f0
[<ffffffff81193bc9>] ? deactivate_slab+0x349/0x3c0
[<ffffffff811926f1>] new_slab+0x91/0x300
[<ffffffff815de95a>] __slab_alloc+0x2bb/0x482
[<ffffffff8105bc1c>] ? copy_process.part.25+0xfc/0x14c0
[<ffffffff810a3c78>] ? load_balance+0x218/0x890
[<ffffffff8101a679>] ? sched_clock+0x9/0x10
[<ffffffff81105ba9>] ? trace_clock_local+0x9/0x10
[<ffffffff81193d1c>] kmem_cache_alloc_node+0x8c/0x200
[<ffffffff8105bc1c>] copy_process.part.25+0xfc/0x14c0
[<ffffffff81114d0d>] ? trace_buffer_unlock_commit+0x4d/0x60
[<ffffffff81085a80>] ? kthread_create_on_node+0x140/0x140
[<ffffffff8105d0ec>] do_fork+0xbc/0x360
[<ffffffff8105d3b6>] kernel_thread+0x26/0x30
[<ffffffff81086652>] kthreadd+0x2c2/0x300
[<ffffffff81086390>] ? kthread_create_on_cpu+0x60/0x60
[<ffffffff815f20ec>] ret_from_fork+0x7c/0xb0
[<ffffffff81086390>] ? kthread_create_on_cpu+0x60/0x60
In my investigation, I found the root cause is wq_numa_possible_cpumask.
All entries of wq_numa_possible_cpumask is allocated by
alloc_cpumask_var_node(). And these entries are used without initializing.
So these entries have wrong value.
When hot-adding and onlining CPU, wq_update_unbound_numa() is called.
wq_update_unbound_numa() calls alloc_unbound_pwq(). And alloc_unbound_pwq()
calls get_unbound_pool(). In get_unbound_pool(), worker_pool->node is set
as follow:
3592 /* if cpumask is contained inside a NUMA node, we belong to that node */
3593 if (wq_numa_enabled) {
3594 for_each_node(node) {
3595 if (cpumask_subset(pool->attrs->cpumask,
3596 wq_numa_possible_cpumask[node])) {
3597 pool->node = node;
3598 break;
3599 }
3600 }
3601 }
But wq_numa_possible_cpumask[node] does not have correct cpumask. So, wrong
node is selected. As a result, kernel panic occurs.
By this patch, all entries of wq_numa_possible_cpumask are allocated by
zalloc_cpumask_var_node to initialize them. And the panic disappeared.
Signed-off-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: bce903809a ("workqueue: add wq_numa_tbl_len and wq_numa_possible_cpumask[]")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit bddbceb688 upstream.
Uevents are suppressed during attributes registration, but never
restored, so kobject_uevent() does nothing.
Signed-off-by: Maxime Bizon <mbizon@freebox.fr>
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: 226223ab3c
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 4d595b866d upstream.
After a @pwq is scheduled for emergency execution, other workers may
consume the affectd work items before the rescuer gets to them. This
means that a workqueue many have pwqs queued on @wq->maydays list
while not having any work item pending or in-flight. If
destroy_workqueue() executes in such condition, the rescuer may exit
without emptying @wq->maydays.
This currently doesn't cause any actual harm. destroy_workqueue() can
safely destroy all the involved data structures whether @wq->maydays
is populated or not as nobody access the list once the rescuer exits.
However, this is nasty and makes future development difficult. Let's
update rescuer_thread() so that it empties @wq->maydays after seeing
should_stop to guarantee that the list is empty on rescuer exit.
tj: Updated comment and patch description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 77668c8b55 upstream.
There is a race condition between rescuer_thread() and
pwq_unbound_release_workfn().
Even after a pwq is scheduled for rescue, the associated work items
may be consumed by any worker. If all of them are consumed before the
rescuer gets to them and the pwq's base ref was put due to attribute
change, the pwq may be released while still being linked on
@wq->maydays list making the rescuer dereference already freed pwq
later.
Make send_mayday() pin the target pwq until the rescuer is done with
it.
tj: Updated comment and patch description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 77f300b198 upstream.
wq_update_unbound_numa() failure path has the following two bugs.
- alloc_unbound_pwq() is called without holding wq->mutex; however, if
the allocation fails, it jumps to out_unlock which tries to unlock
wq->mutex.
- The function should switch to dfl_pwq on failure but didn't do so
after alloc_unbound_pwq() failure.
Fix it by regrabbing wq->mutex and jumping to use_dfl_pwq on
alloc_unbound_pwq() failure.
Signed-off-by: Daeseok Youn <daeseok.youn@gmail.com>
Acked-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: 4c16bd327c ("workqueue: implement NUMA affinity for unbound workqueues")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 5bdfff96c6 upstream.
When a kworker should die, the kworkre is notified through WORKER_DIE
flag instead of kthread_should_stop(). This, IIRC, is primarily to
keep the test synchronized inside worker_pool lock. WORKER_DIE is
first set while holding pool->lock, the lock is dropped and
kthread_stop() is called.
Unfortunately, this means that there's a slight chance that the target
kworker may see WORKER_DIE before kthread_stop() finishes and exits
and frees the target task before or during kthread_stop().
Fix it by pinning the target task before setting WORKER_DIE and
putting it after kthread_stop() is done.
tj: Improved patch description and comment. Moved pinning above
WORKER_DIE for better signify what it's protecting.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 8a2b753844 upstream.
An ordered workqueue implements execution ordering by using single
pool_workqueue with max_active == 1. On a given pool_workqueue, work
items are processed in FIFO order and limiting max_active to 1
enforces the queued work items to be processed one by one.
Unfortunately, 4c16bd327c ("workqueue: implement NUMA affinity for
unbound workqueues") accidentally broke this guarantee by applying
NUMA affinity to ordered workqueues too. On NUMA setups, an ordered
workqueue would end up with separate pool_workqueues for different
nodes. Each pool_workqueue still limits max_active to 1 but multiple
work items may be executed concurrently and out of order depending on
which node they are queued to.
Fix it by using dedicated ordered_wq_attrs[] when creating ordered
workqueues. The new attrs match the unbound ones except that no_numa
is always set thus forcing all NUMA nodes to share the default
pool_workqueue.
While at it, add sanity check in workqueue creation path which
verifies that an ordered workqueues has only the default
pool_workqueue.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Libin <huawei.libin@huawei.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b22ce2785d upstream.
If !PREEMPT, a kworker running work items back to back can hog CPU.
This becomes dangerous when a self-requeueing work item which is
waiting for something to happen races against stop_machine. Such
self-requeueing work item would requeue itself indefinitely hogging
the kworker and CPU it's running on while stop_machine would wait for
that CPU to enter stop_machine while preventing anything else from
happening on all other CPUs. The two would deadlock.
Jamie Liu reports that this deadlock scenario exists around
scsi_requeue_run_queue() and libata port multiplier support, where one
port may exclude command processing from other ports. With the right
timing, scsi_requeue_run_queue() can end up requeueing itself trying
to execute an IO which is asked to be retried while another device has
an exclusive access, which in turn can't make forward progress due to
stop_machine.
Fix it by invoking cond_resched() after executing each work item.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Jamie Liu <jamieliu@google.com>
References: http://thread.gmane.org/gmane.linux.kernel/1552567
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 2865a8fb44 upstream.
$echo '0' > /sys/bus/workqueue/devices/xxx/numa
$cat /sys/bus/workqueue/devices/xxx/numa
I got 1. It should be 0, the reason is copy_workqueue_attrs() called
in apply_workqueue_attrs() doesn't copy no_numa field.
Fix it by making copy_workqueue_attrs() copy ->no_numa too. This
would also make get_unbound_pool() set a pool's ->no_numa attribute
according to the workqueue attributes used when the pool was created.
While harmelss, as ->no_numa isn't a pool attribute, this is a bit
confusing. Clear it explicitly.
tj: Updated description and comments a bit.
Signed-off-by: Shaohua Li <shli@fusionio.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch adds system wide workqueues aligned towards power saving. This is
done by allocating them with WQ_UNBOUND flag if 'wq_power_efficient' is set to
'true'.
tj: updated comments a bit.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
(cherry picked from commit 0668106ca3)
Signed-off-by: Mark Brown <broonie@linaro.org>
Workqueues can be performance or power-oriented. Currently, most workqueues are
bound to the CPU they were created on. This gives good performance (due to cache
effects) at the cost of potentially waking up otherwise idle cores (Idle from
scheduler's perspective. Which may or may not be physically idle) just to
process some work. To save power, we can allow the work to be rescheduled on a
core that is already awake.
Workqueues created with the WQ_UNBOUND flag will allow some power savings.
However, we don't change the default behaviour of the system. To enable
power-saving behaviour, a new config option CONFIG_WQ_POWER_EFFICIENT needs to
be turned on. This option can also be overridden by the
workqueue.power_efficient boot parameter.
tj: Updated config description and comments. Renamed
CONFIG_WQ_POWER_EFFICIENT to CONFIG_WQ_POWER_EFFICIENT_DEFAULT.
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Reviewed-by: Amit Kucheria <amit.kucheria@linaro.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
(cherry picked from commit cee22a1505)
Signed-off-by: Mark Brown <broonie@linaro.org>
Commit 8425e3d5bd ("workqueue: inline trivial wrappers") changed
schedule_work() and schedule_delayed_work() to inline wrappers,
but these rely on some symbols that are EXPORT_SYMBOL_GPL, while
the original functions were EXPORT_SYMBOL. This has the effect of
changing the licensing requirement for these functions and making
them unavailable to non GPL modules.
Make them available again by removing the restriction on the
required symbols.
Signed-off-by: Marc Dionne <marc.dionne@your-file-system.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
When we fail to mutex_trylock(), we release the pool spin_lock and do
mutex_lock(). After that, we should regrab the pool spin_lock, but,
regrabbing is missed in current code. So correct it.
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
