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>
Cc: stable@vger.kernel.org
When there is serious memory pressure, all workers in a pool could be
blocked, and a new thread cannot be created because it requires memory
allocation.
In this situation a WQ_MEM_RECLAIM workqueue will wake up the
rescuer thread to do some work.
The rescuer will only handle requests that are already on ->worklist.
If max_requests is 1, that means it will handle a single request.
The rescuer will be woken again in 100ms to handle another max_requests
requests.
I've seen a machine (running a 3.0 based "enterprise" kernel) with
thousands of requests queued for xfslogd, which has a max_requests of
1, and is needed for retiring all 'xfs' write requests. When one of
the worker pools gets into this state, it progresses extremely slowly
and possibly never recovers (only waited an hour or two).
With this patch we leave a pool_workqueue on mayday list
until it is clearly no longer in need of assistance. This allows
all requests to be handled in a timely fashion.
We keep each pool_workqueue on the mayday list until
need_to_create_worker() is false, and no work for this workqueue is
found in the pool.
I have tested this in combination with a (hackish) patch which forces
all work items to be handled by the rescuer thread. In that context
it significantly improves performance. A similar patch for a 3.0
kernel significantly improved performance on a heavy work load.
Thanks to Jan Kara for some design ideas, and to Dongsu Park for
some comments and testing.
tj: Inverted the lock order between wq_mayday_lock and pool->lock with
a preceding patch and simplified this patch. Added comment and
updated changelog accordingly. Dongsu spotted missing get_pwq()
in the simplified code.
Cc: Dongsu Park <dongsu.park@profitbricks.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
Currently, pool->lock nests inside pool->lock. There's no inherent
reason for this order. The only place where the two locks are held
together is pool_mayday_timeout() and it just got decided that way.
This nesting order turns out to complicate things with the planned
rescuer_thread() update. Let's invert them. This doesn't cause any
behavior differences.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: NeilBrown <neilb@suse.de>
Cc: Dongsu Park <dongsu.park@profitbricks.com>
rescuer_thread() caches &rescuer->scheduled in a local variable
scheduled for convenience. There's one WARN_ON_ONCE() which was using
&rescuer->scheduled directly. Replace it with the local variable.
This patch causes no functional difference.
Signed-off-by: Tejun Heo <tj@kernel.org>
Tidy up and use cond_resched_rcu_qs when calling cond_resched and
reporting potential quiescent state to RCU. Splitting this change in
this way allows easy backporting to -stable for kernel versions not
having cond_resched_rcu_qs().
Signed-off-by: Joe Lawrence <joe.lawrence@stratus.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Pull percpu updates from Tejun Heo:
- Major reorganization of percpu header files which I think makes
things a lot more readable and logical than before.
- percpu-refcount is updated so that it requires explicit destruction
and can be reinitialized if necessary. This was pulled into the
block tree to replace the custom percpu refcnting implemented in
blk-mq.
- In the process, percpu and percpu-refcount got cleaned up a bit
* 'for-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu: (21 commits)
percpu-refcount: implement percpu_ref_reinit() and percpu_ref_is_zero()
percpu-refcount: require percpu_ref to be exited explicitly
percpu-refcount: use unsigned long for pcpu_count pointer
percpu-refcount: add helpers for ->percpu_count accesses
percpu-refcount: one bit is enough for REF_STATUS
percpu-refcount, aio: use percpu_ref_cancel_init() in ioctx_alloc()
workqueue: stronger test in process_one_work()
workqueue: clear POOL_DISASSOCIATED in rebind_workers()
percpu: Use ALIGN macro instead of hand coding alignment calculation
percpu: invoke __verify_pcpu_ptr() from the generic part of accessors and operations
percpu: preffity percpu header files
percpu: use raw_cpu_*() to define __this_cpu_*()
percpu: reorder macros in percpu header files
percpu: move {raw|this}_cpu_*() definitions to include/linux/percpu-defs.h
percpu: move generic {raw|this}_cpu_*_N() definitions to include/asm-generic/percpu.h
percpu: only allow sized arch overrides for {raw|this}_cpu_*() ops
percpu: reorganize include/linux/percpu-defs.h
percpu: move accessors from include/linux/percpu.h to percpu-defs.h
percpu: include/asm-generic/percpu.h should contain only arch-overridable parts
percpu: introduce arch_raw_cpu_ptr()
...
Pull workqueue updates from Tejun Heo:
"Lai has been doing a lot of cleanups of workqueue and kthread_work.
No significant behavior change. Just a lot of cleanups all over the
place. Some are a bit invasive but overall nothing too dangerous"
* 'for-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
kthread_work: remove the unused wait_queue_head
kthread_work: wake up worker only when the worker is idle
workqueue: use nr_node_ids instead of wq_numa_tbl_len
workqueue: remove the misnamed out_unlock label in get_unbound_pool()
workqueue: remove the stale comment in pwq_unbound_release_workfn()
workqueue: move rescuer pool detachment to the end
workqueue: unfold start_worker() into create_worker()
workqueue: remove @wakeup from worker_set_flags()
workqueue: remove an unneeded UNBOUND test before waking up the next worker
workqueue: wake regular worker if need_more_worker() when rescuer leave the pool
workqueue: alloc struct worker on its local node
workqueue: reuse the already calculated pwq in try_to_grab_pending()
workqueue: stronger test in process_one_work()
workqueue: clear POOL_DISASSOCIATED in rebind_workers()
workqueue: sanity check pool->cpu in wq_worker_sleeping()
workqueue: clear leftover flags when detached
workqueue: remove useless WARN_ON_ONCE()
workqueue: use schedule_timeout_interruptible() instead of open code
workqueue: remove the empty check in too_many_workers()
workqueue: use "pool->cpu < 0" to stand for an unbound pool
They are the same and nr_node_ids is provided by the memory subsystem.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
After the locking was moved up to the caller of the get_unbound_pool(),
out_unlock label doesn't need to do any unlock operation and the name
became bad, so we just remove this label, and the only usage-site
"goto out_unlock" is subsituted to "return pool".
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
In 75ccf5950f ("workqueue: prepare flush_workqueue() for dynamic
creation and destrucion of unbound pool_workqueues"), a comment
about the synchronization for the pwq in pwq_unbound_release_workfn()
was added. The comment claimed the flush_mutex wasn't strictly
necessary, it was correct in that time, due to the pwq was protected
by workqueue_lock.
But it is incorrect now since the wq->flush_mutex was renamed to
wq->mutex and workqueue_lock was removed, the wq->mutex is strictly
needed. But the comment was miss-updated when the synchronization
was changed.
This patch removes the incorrect comments and doesn't add any new
comment to explain why wq->mutex is needed here, which is definitely
obvious and wq->pwqs_node has "WQ" notation in its definition which is
better comment.
The old commit mentioned above also introduced a comment in link_pwq()
about the synchronization. This comment is also removed in this patch
since the whole link_pwq() is proteced by wq->mutex.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
In 51697d3939 ("workqueue: use generic attach/detach routine for
rescuers"), The rescuer detaches itself from the pool before put_pwq()
so that the put_unbound_pool() will not destroy the rescuer-attached
pool.
It is unnecessary. worker_detach_from_pool() can be used as the last
statement to access to the pool just like the regular workers,
put_unbound_pool() will wait for it to detach and then free the pool.
So we move the worker_detach_from_pool() down, make it coincide with
the regular workers.
tj: Minor description update.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Simply unfold the code of start_worker() into create_worker() and
remove the original start_worker() and create_and_start_worker().
The only trade-off is the introduced overhead that the pool->lock
is released and regrabbed after the newly worker is started.
The overhead is acceptible since the manager is slow path.
And because this new locking behavior, the newly created worker
may grab the lock earlier than the manager and go to process
work items. In this case, the recheck need_to_create_worker() may be
true as expected and the manager goes to restart which is the
correct behavior.
tj: Minor updates to description and comments.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
worker_set_flags() has only two callers, each specifying %true and
%false for @wakeup. Let's push the wake up to the caller and remove
@wakeup from worker_set_flags(). The caller can use the following
instead if wakeup is necessary:
worker_set_flags();
if (need_more_worker(pool))
wake_up_worker(pool);
This makes the code simpler. This patch doesn't introduce behavior
changes.
tj: Updated description and comments.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
In process_one_work():
if ((worker->flags & WORKER_UNBOUND) && need_more_worker(pool))
wake_up_worker(pool);
the first test is unneeded. Even if the first test is removed, it
doesn't affect the wake-up logic for WORKER_UNBOUND, and it will not
introduce any useless wake-ups for normal per-cpu workers since
nr_running is always >= 1. It will introduce useless/redundant
wake-ups for CPU_INTENSIVE, but this case is rare and the next patch
will also remove this redundant wake-up.
tj: Minor updates to the description and comment.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
We don't need to wake up regular worker when nr_running==1,
so need_more_worker() is sufficient here.
And need_more_worker() gives us better readability due to the name of
"keep_working()" implies the rescuer should keep working now but
the rescuer is actually leaving.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
When the create_worker() is called from non-manager, the struct worker
is allocated from the node of the caller which may be different from the
node of pool->node.
So we add a node ID argument for the alloc_worker() to ensure the
struct worker is allocated from the preferable node.
tj: @nid renamed to @node for consistency.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
try_to_grab_pending() was re-calculating the associated pwq using
get_work_pwq() when it already has it cached in a local varible and
the association can't change. Reuse the local variable instead.
This doesn't introduce any functional changes.
tj: Updated description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
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>
Cc: stable@vger.kernel.org
Fixes: bce903809a ("workqueue: add wq_numa_tbl_len and wq_numa_possible_cpumask[]")
When POOL_DISASSOCIATED is cleared, the running worker's local CPU should
be the same as pool->cpu without any exception even during cpu-hotplug.
This patch changes "(proposition_A && proposition_B && proposition_C)"
to "(proposition_B && proposition_C)", so if the old compound
proposition is true, the new one must be true too. so this won't hide
any possible bug which can be hit by old test.
tj: Minor description update and dropped the obvious comment.
CC: Jason J. Herne <jjherne@linux.vnet.ibm.com>
CC: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
a9ab775bca ("workqueue: directly restore CPU affinity of workers
from CPU_ONLINE") moved pool locking into rebind_workers() but left
"pool->flags &= ~POOL_DISASSOCIATED" in workqueue_cpu_up_callback().
There is nothing necessarily wrong with it, but there is no benefit
either. Let's move it into rebind_workers() and achieve the following
benefits:
1) better readability, POOL_DISASSOCIATED is cleared in rebind_workers()
as expected.
2) we can guarantee that, when POOL_DISASSOCIATED is clear, the
running workers of the pool are on the local CPU (pool->cpu).
tj: Minor description update.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
After the recent changes, when POOL_DISASSOCIATED is cleared, the
running worker's local CPU should be the same as pool->cpu without any
exception even during cpu-hotplug. Update the sanity check in
process_one_work() accordingly.
This patch changes "(proposition_A && proposition_B && proposition_C)"
to "(proposition_B && proposition_C)", so if the old compound
proposition is true, the new one must be true too. so this will not
hide any possible bug which can be caught by the old test.
tj: Minor updates to the description.
CC: Jason J. Herne <jjherne@linux.vnet.ibm.com>
CC: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
The commit a9ab775bca ("workqueue: directly restore CPU affinity of
workers from CPU_ONLINE") moved the pool->lock into rebind_workers()
without also moving "pool->flags &= ~POOL_DISASSOCIATED".
There is nothing wrong with "pool->flags &= ~POOL_DISASSOCIATED" not
being moved together, but there isn't any benefit either. We move it
into rebind_workers() and achieve these benefits:
1) Better readability. POOL_DISASSOCIATED is cleared in
rebind_workers() as expected.
2) When POOL_DISASSOCIATED is cleared, we can ensure that all the
running workers of the pool are on the local CPU (pool->cpu).
tj: Cosmetic updates to the code and description.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
In theory, pool->cpu is equals to @cpu in wq_worker_sleeping() after
worker->flags is checked.
And "pool->cpu != cpu" sanity check will help us if something wrong.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
