Queueing operations use WORK_STRUCT_PENDING_BIT to synchronize access
to the target work item. They first try to claim the bit and proceed
with queueing only after that succeeds and there's a window between
PENDING being set and the actual queueing where the task can be
interrupted or preempted.
There's also a similar window in process_one_work() when clearing
PENDING. A work item is dequeued, gcwq->lock is released and then
PENDING is cleared and the worker might get interrupted or preempted
between releasing gcwq->lock and clearing PENDING.
cancel[_delayed]_work_sync() tries to claim or steal PENDING. The
function assumes that a work item with PENDING is either queued or in
the process of being [de]queued. In the latter case, it busy-loops
until either the work item loses PENDING or is queued. If canceling
coincides with the above described interrupts or preemptions, the
canceling task will busy-loop while the queueing or executing task is
preempted.
This patch keeps irq disabled across claiming PENDING and actual
queueing and moves PENDING clearing in process_one_work() inside
gcwq->lock so that busy looping from PENDING && !queued doesn't wait
for interrupted/preempted tasks. Note that, in process_one_work(),
setting last CPU and clearing PENDING got merged into single
operation.
This removes possible long busy-loops and will allow using
try_to_grab_pending() from bh and irq contexts.
v2: __queue_work() was testing preempt_count() to ensure that the
caller has disabled preemption. This triggers spuriously if
!CONFIG_PREEMPT_COUNT. Use preemptible() instead. Reported by
Fengguang Wu.
v3: Disable irq instead of preemption. IRQ will be disabled while
grabbing gcwq->lock later anyway and this allows using
try_to_grab_pending() from bh and irq contexts.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
WORK_STRUCT_PENDING is used to claim ownership of a work item and
process_one_work() releases it before starting execution. When
someone else grabs PENDING, all pre-release updates to the work item
should be visible and all updates made by the new owner should happen
afterwards.
Grabbing PENDING uses test_and_set_bit() and thus has a full barrier;
however, clearing doesn't have a matching wmb. Given the preceding
spin_unlock and use of clear_bit, I don't believe this can be a
problem on an actual machine and there hasn't been any related report
but it still is theretically possible for clear_pending to permeate
upwards and happen before work->entry update.
Add an explicit smp_wmb() before work_clear_pending().
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: stable@vger.kernel.org
All queueing functions return 1 on success, 0 if the work item was
already pending. Update them to return bool instead. This signifies
better that they don't return 0 / -errno.
This is cleanup and doesn't cause any functional difference.
While at it, fix comment opening for schedule_work_on().
Signed-off-by: Tejun Heo <tj@kernel.org>
Currently, queue/schedule[_delayed]_work_on() are located below the
counterpart without the _on postifx even though the latter is usually
implemented using the former. Swap them.
This is cleanup and doesn't cause any functional difference.
Signed-off-by: Tejun Heo <tj@kernel.org>
25511a4776 "workqueue: reimplement CPU online rebinding to handle idle
workers" added CPU locality sanity check in process_one_work(). It
triggers if a worker is executing on a different CPU without UNBOUND
or REBIND set.
This works for all normal workers but rescuers can trigger this
spuriously when they're serving the unbound or a disassociated
global_cwq - rescuers don't have either flag set and thus its
gcwq->cpu can be a different value including %WORK_CPU_UNBOUND.
Fix it by additionally testing %GCWQ_DISASSOCIATED.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
LKML-Refence: <20120721213656.GA7783@linux.vnet.ibm.com>
With trustee gone, CPU hotplug code can be simplified.
* gcwq_claim/release_management() now grab and release gcwq lock too
respectively and gained _and_lock and _and_unlock postfixes.
* All CPU hotplug logic was implemented in workqueue_cpu_callback()
which was called by workqueue_cpu_up/down_callback() for the correct
priority. This was because up and down paths shared a lot of logic,
which is no longer true. Remove workqueue_cpu_callback() and move
all hotplug logic into the two actual callbacks.
This patch doesn't make any functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
With the previous changes, a disassociated global_cwq now can run as
an unbound one on its own - it can create workers as necessary to
drain remaining works after the CPU has been brought down and manage
the number of workers using the usual idle timer mechanism making
trustee completely redundant except for the actual unbinding
operation.
This patch removes the trustee and let a disassociated global_cwq
manage itself. Unbinding is moved to a work item (for CPU affinity)
which is scheduled and flushed from CPU_DONW_PREPARE.
This patch moves nr_running clearing outside gcwq and manager locks to
simplify the code. As nr_running is unused at the point, this is
safe.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Currently, during CPU offlining, after all pending work items are
drained, the trustee butchers all workers. Also, on CPU onlining
failure, workqueue_cpu_callback() ensures that the first idle worker
is destroyed. Combined, these guarantee that an offline CPU doesn't
have any worker for it once all the lingering work items are finished.
This guarantee isn't really necessary and makes CPU on/offlining more
expensive than needs to be, especially for platforms which use CPU
hotplug for powersaving.
This patch lets offline CPUs removes idle worker butchering from the
trustee and let a CPU which failed onlining keep the created first
worker. The first worker is created if the CPU doesn't have any
during CPU_DOWN_PREPARE and started right away. If onlining succeeds,
the rebind_workers() call in CPU_ONLINE will rebind it like any other
workers. If onlining fails, the worker is left alone till the next
try.
This makes CPU hotplugs cheaper by allowing global_cwqs to keep
workers across them and simplifies code.
Note that trustee doesn't re-arm idle timer when it's done and thus
the disassociated global_cwq will keep all workers until it comes back
online. This will be improved by further patches.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Currently, if there are left workers when a CPU is being brough back
online, the trustee kills all idle workers and scheduled rebind_work
so that they re-bind to the CPU after the currently executing work is
finished. This works for busy workers because concurrency management
doesn't try to wake up them from scheduler callbacks, which require
the target task to be on the local run queue. The busy worker bumps
concurrency counter appropriately as it clears WORKER_UNBOUND from the
rebind work item and it's bound to the CPU before returning to the
idle state.
To reduce CPU on/offlining overhead (as many embedded systems use it
for powersaving) and simplify the code path, workqueue is planned to
be modified to retain idle workers across CPU on/offlining. This
patch reimplements CPU online rebinding such that it can also handle
idle workers.
As noted earlier, due to the local wakeup requirement, rebinding idle
workers is tricky. All idle workers must be re-bound before scheduler
callbacks are enabled. This is achieved by interlocking idle
re-binding. Idle workers are requested to re-bind and then hold until
all idle re-binding is complete so that no bound worker starts
executing work item. Only after all idle workers are re-bound and
parked, CPU_ONLINE proceeds to release them and queue rebind work item
to busy workers thus guaranteeing scheduler callbacks aren't invoked
until all idle workers are ready.
worker_rebind_fn() is renamed to busy_worker_rebind_fn() and
idle_worker_rebind() for idle workers is added. Rebinding logic is
moved to rebind_workers() and now called from CPU_ONLINE after
flushing trustee. While at it, add CPU sanity check in
worker_thread().
Note that now a worker may become idle or the manager between trustee
release and rebinding during CPU_ONLINE. As the previous patch
updated create_worker() so that it can be used by regular manager
while unbound and this patch implements idle re-binding, this is safe.
This prepares for removal of trustee and keeping idle workers across
CPU hotplugs.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Currently, create_worker()'s callers are responsible for deciding
whether the newly created worker should be bound to the associated CPU
and create_worker() sets WORKER_UNBOUND only for the workers for the
unbound global_cwq. Creation during normal operation is always via
maybe_create_worker() and @bind is true. For workers created during
hotplug, @bind is false.
Normal operation path is planned to be used even while the CPU is
going through hotplug operations or offline and this static decision
won't work.
Drop @bind from create_worker() and decide whether to bind by looking
at GCWQ_DISASSOCIATED. create_worker() will also set WORKER_UNBOUND
autmatically if disassociated. To avoid flipping GCWQ_DISASSOCIATED
while create_worker() is in progress, the flag is now allowed to be
changed only while holding all manager_mutexes on the global_cwq.
This requires that GCWQ_DISASSOCIATED is not cleared behind trustee's
back. CPU_ONLINE no longer clears DISASSOCIATED before flushing
trustee, which clears DISASSOCIATED before rebinding remaining workers
if asked to release. For cases where trustee isn't around, CPU_ONLINE
clears DISASSOCIATED after flushing trustee. Also, now, first_idle
has UNBOUND set on creation which is explicitly cleared by CPU_ONLINE
while binding it. These convolutions will soon be removed by further
simplification of CPU hotplug path.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
POOL_MANAGING_WORKERS is used to ensure that at most one worker takes
the manager role at any given time on a given global_cwq. Trustee
later hitched on it to assume manager adding blocking wait for the
bit. As trustee already needed a custom wait mechanism, waiting for
MANAGING_WORKERS was rolled into the same mechanism.
Trustee is scheduled to be removed. This patch separates out
MANAGING_WORKERS wait into per-pool mutex. Workers use
mutex_trylock() to test for manager role and trustee uses mutex_lock()
to claim manager roles.
gcwq_claim/release_management() helpers are added to grab and release
manager roles of all pools on a global_cwq. gcwq_claim_management()
always grabs pool manager mutexes in ascending pool index order and
uses pool index as lockdep subclass.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Currently, WORKER_UNBOUND is used to mark workers for the unbound
global_cwq and WORKER_ROGUE is used to mark workers for disassociated
per-cpu global_cwqs. Both are used to make the marked worker skip
concurrency management and the only place they make any difference is
in worker_enter_idle() where WORKER_ROGUE is used to skip scheduling
idle timer, which can easily be replaced with trustee state testing.
This patch replaces WORKER_ROGUE with WORKER_UNBOUND and drops
WORKER_ROGUE. This is to prepare for removing trustee and handling
disassociated global_cwqs as unbound.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Workqueue used CPU_DYING notification to mark GCWQ_DISASSOCIATED.
This was necessary because workqueue's CPU_DOWN_PREPARE happened
before other DOWN_PREPARE notifiers and workqueue needed to stay
associated across the rest of DOWN_PREPARE.
After the previous patch, workqueue's DOWN_PREPARE happens after
others and can set GCWQ_DISASSOCIATED directly. Drop CPU_DYING and
let the trustee set GCWQ_DISASSOCIATED after disabling concurrency
management.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Currently, all workqueue cpu hotplug operations run off
CPU_PRI_WORKQUEUE which is higher than normal notifiers. This is to
ensure that workqueue is up and running while bringing up a CPU before
other notifiers try to use workqueue on the CPU.
Per-cpu workqueues are supposed to remain working and bound to the CPU
for normal CPU_DOWN_PREPARE notifiers. This holds mostly true even
with workqueue offlining running with higher priority because
workqueue CPU_DOWN_PREPARE only creates a bound trustee thread which
runs the per-cpu workqueue without concurrency management without
explicitly detaching the existing workers.
However, if the trustee needs to create new workers, it creates
unbound workers which may wander off to other CPUs while
CPU_DOWN_PREPARE notifiers are in progress. Furthermore, if the CPU
down is cancelled, the per-CPU workqueue may end up with workers which
aren't bound to the CPU.
While reliably reproducible with a convoluted artificial test-case
involving scheduling and flushing CPU burning work items from CPU down
notifiers, this isn't very likely to happen in the wild, and, even
when it happens, the effects are likely to be hidden by the following
successful CPU down.
Fix it by using different priorities for up and down notifiers - high
priority for up operations and low priority for down operations.
Workqueue cpu hotplug operations will soon go through further cleanup.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: stable@vger.kernel.org
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
WQ_HIGHPRI was implemented by queueing highpri work items at the head
of the global worklist. Other than queueing at the head, they weren't
handled differently; unfortunately, this could lead to execution
latency of a few seconds on heavily loaded systems.
Now that workqueue code has been updated to deal with multiple
worker_pools per global_cwq, this patch reimplements WQ_HIGHPRI using
a separate worker_pool. NR_WORKER_POOLS is bumped to two and
gcwq->pools[0] is used for normal pri work items and ->pools[1] for
highpri. Highpri workers get -20 nice level and has 'H' suffix in
their names. Note that this change increases the number of kworkers
per cpu.
POOL_HIGHPRI_PENDING, pool_determine_ins_pos() and highpri chain
wakeup code in process_one_work() are no longer used and removed.
This allows proper prioritization of highpri work items and removes
high execution latency of highpri work items.
v2: nr_running indexing bug in get_pool_nr_running() fixed.
v3: Refreshed for the get_pool_nr_running() update in the previous
patch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Josh Hunt <joshhunt00@gmail.com>
LKML-Reference: <CAKA=qzaHqwZ8eqpLNFjxnO2fX-tgAOjmpvxgBFjv6dJeQaOW1w@mail.gmail.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Introduce NR_WORKER_POOLS and for_each_worker_pool() and convert code
paths which need to manipulate all pools in a gcwq to use them.
NR_WORKER_POOLS is currently one and for_each_worker_pool() iterates
over only @gcwq->pool.
Note that nr_running is per-pool property and converted to an array
with NR_WORKER_POOLS elements and renamed to pool_nr_running. Note
that get_pool_nr_running() currently assumes 0 index. The next patch
will make use of non-zero index.
The changes in this patch are mechanical and don't caues any
functional difference. This is to prepare for multiple pools per
gcwq.
v2: nr_running indexing bug in get_pool_nr_running() fixed.
v3: Pointer to array is stupid. Don't use it in get_pool_nr_running()
as suggested by Linus.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
GCWQ_MANAGE_WORKERS, GCWQ_MANAGING_WORKERS and GCWQ_HIGHPRI_PENDING
are per-pool properties. Add worker_pool->flags and make the above
three flags per-pool flags.
The changes in this patch are mechanical and don't caues any
functional difference. This is to prepare for multiple pools per
gcwq.
Signed-off-by: Tejun Heo <tj@kernel.org>
Modify all functions which deal with per-pool properties to pass
around @pool instead of @gcwq or @cpu.
The changes in this patch are mechanical and don't caues any
functional difference. This is to prepare for multiple pools per
gcwq.
Signed-off-by: Tejun Heo <tj@kernel.org>
Move worklist and all worker management fields from global_cwq into
the new struct worker_pool. worker_pool points back to the containing
gcwq. worker and cpu_workqueue_struct are updated to point to
worker_pool instead of gcwq too.
This change is mechanical and doesn't introduce any functional
difference other than rearranging of fields and an added level of
indirection in some places. This is to prepare for multiple pools per
gcwq.
v2: Comment typo fixes as suggested by Namhyung.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Unbound wqs aren't concurrency-managed and try to execute work items
as soon as possible. This is currently achieved by implicitly setting
%WQ_HIGHPRI on all unbound workqueues; however, WQ_HIGHPRI
implementation is about to be restructured and this usage won't be
valid anymore.
Add an explicit chain-wakeup path for unbound workqueues in
process_one_work() instead of piggy backing on %WQ_HIGHPRI.
Signed-off-by: Tejun Heo <tj@kernel.org>
Under memory load, on x86_64, with lockdep enabled, the workqueue's
process_one_work() has been seen to oops in __lock_acquire(), barfing
on a 0xffffffff00000000 pointer in the lockdep_map's class_cache[].
Because it's permissible to free a work_struct from its callout function,
the map used is an onstack copy of the map given in the work_struct: and
that copy is made without any locking.
Surprisingly, gcc (4.5.1 in Hugh's case) uses "rep movsl" rather than
"rep movsq" for that structure copy: which might race with a workqueue
user's wait_on_work() doing lock_map_acquire() on the source of the
copy, putting a pointer into the class_cache[], but only in time for
the top half of that pointer to be copied to the destination map.
Boom when process_one_work() subsequently does lock_map_acquire()
on its onstack copy of the lockdep_map.
Fix this, and a similar instance in call_timer_fn(), with a
lockdep_copy_map() function which additionally NULLs the class_cache[].
Note: this oops was actually seen on 3.4-next, where flush_work() newly
does the racing lock_map_acquire(); but Tejun points out that 3.4 and
earlier are already vulnerable to the same through wait_on_work().
* Patch orginally from Peter. Hugh modified it a bit and wrote the
description.
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Reported-by: Hugh Dickins <hughd@google.com>
LKML-Reference: <alpine.LSU.2.00.1205070951170.1544@eggly.anvils>
Signed-off-by: Tejun Heo <tj@kernel.org>
worker_enter_idle() has WARN_ON_ONCE() which triggers if nr_running
isn't zero when every worker is idle. This can trigger spuriously
while a cpu is going down due to the way trustee sets %WORKER_ROGUE
and zaps nr_running.
It first sets %WORKER_ROGUE on all workers without updating
nr_running, releases gcwq->lock, schedules, regrabs gcwq->lock and
then zaps nr_running. If the last running worker enters idle
inbetween, it would see stale nr_running which hasn't been zapped yet
and trigger the WARN_ON_ONCE().
Fix it by performing the sanity check iff the trustee is idle.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org
If a workqueue is flushed with flush_work() lockdep checking can
be circumvented. For example:
static DEFINE_MUTEX(mutex);
static void my_work(struct work_struct *w)
{
mutex_lock(&mutex);
mutex_unlock(&mutex);
}
static DECLARE_WORK(work, my_work);
static int __init start_test_module(void)
{
schedule_work(&work);
return 0;
}
module_init(start_test_module);
static void __exit stop_test_module(void)
{
mutex_lock(&mutex);
flush_work(&work);
mutex_unlock(&mutex);
}
module_exit(stop_test_module);
would not always print a warning when flush_work() was called.
In this trivial example nothing could go wrong since we are
guaranteed module_init() and module_exit() don't run concurrently,
but if the work item is schedule asynchronously we could have a
scenario where the work item is running just at the time flush_work()
is called resulting in a classic ABBA locking problem.
Add a lockdep hint by acquiring and releasing the work item
lockdep_map in flush_work() so that we always catch this
potential deadlock scenario.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Yong Zhang <yong.zhang0@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This BUG_ON() can be triggered if you call schedule_work() before
calling INIT_WORK(). It is a bug definitely, but it's nicer to just
print a stack trace and return.
Reported-by: Matt Renzelmann <mjr@cs.wisc.edu>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Pull workqueue changes from Tejun Heo:
"This contains only one commit which cleans up UP allocation path a
bit."
* 'for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: use percpu allocator for cwq on UP
I notice that the commit bbddff makes percpu allocator can work on UP,
So we don't need the magic way for UP.
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
This patch (as1519) fixes a bug in the block layer's disk-events
polling. The polling is done by a work routine queued on the
system_nrt_wq workqueue. Since that workqueue isn't freezable, the
polling continues even in the middle of a system sleep transition.
Obviously, polling a suspended drive for media changes and such isn't
a good thing to do; in the case of USB mass-storage devices it can
lead to real problems requiring device resets and even re-enumeration.
The patch fixes things by creating a new system-wide, non-reentrant,
freezable workqueue and using it for disk-events polling.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
CC: <stable@kernel.org>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
alloc_workqueue() currently expects the passed in @name pointer to remain
accessible. This is inconvenient and a bit silly given that the whole wq
is being dynamically allocated. This patch updates alloc_workqueue() and
friends to take printf format string instead of opaque string and matching
varargs at the end. The name is allocated together with the wq and
formatted.
alloc_ordered_workqueue() is converted to a macro to unify varargs
handling with alloc_workqueue(), and, while at it, add comment to
alloc_workqueue().
None of the current in-kernel users pass in string with '%' as constant
name and this change shouldn't cause any problem.
[akpm@linux-foundation.org: use __printf]
Signed-off-by: Tejun Heo <tj@kernel.org>
Suggested-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The changed files were only including linux/module.h for the
EXPORT_SYMBOL infrastructure, and nothing else. Revector them
onto the isolated export header for faster compile times.
Nothing to see here but a whole lot of instances of:
-#include <linux/module.h>
+#include <linux/export.h>
This commit is only changing the kernel dir; next targets
will probably be mm, fs, the arch dirs, etc.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Take cwq->gcwq->lock to avoid racing between drain_workqueue checking to
make sure the workqueues are empty and cwq_dec_nr_in_flight decrementing
and then incrementing nr_active when it activates a delayed work.
We discovered this when a corner case in one of our drivers resulted in
us trying to destroy a workqueue in which the remaining work would
always requeue itself again in the same workqueue. We would hit this
race condition and trip the BUG_ON on workqueue.c:3080.
Signed-off-by: Thomas Tuttle <ttuttle@chromium.org>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-3.1' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: separate out drain_workqueue() from destroy_workqueue()
workqueue: remove cancel_rearming_delayed_work[queue]()
* 'for-2.6.40' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu:
percpu: Unify input section names
percpu: Avoid extra NOP in percpu_cmpxchg16b_double
percpu: Cast away printk format warning
percpu: Always align percpu output section to PAGE_SIZE
Fix up fairly trivial conflict in arch/x86/include/asm/percpu.h as per Tejun
There are users which want to drain workqueues without destroying it.
Separate out drain functionality from destroy_workqueue() into
drain_workqueue() and make it accessible to workqueue users.
To guarantee forward-progress, only chain queueing is allowed while
drain is in progress. If a new work item which isn't chained from the
running or pending work items is queued while draining is in progress,
WARN_ON_ONCE() is triggered.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: James Bottomley <James.Bottomley@hansenpartnership.com>
If a rescuer and stop_machine() bringing down a CPU race with each
other, they may deadlock on non-preemptive kernel. The CPU won't
accept a new task, so the rescuer can't migrate to the target CPU,
while stop_machine() can't proceed because the rescuer is holding one
of the CPU retrying migration. GCWQ_DISASSOCIATED is never cleared
and worker_maybe_bind_and_lock() retries indefinitely.
This problem can be reproduced semi reliably while the system is
entering suspend.
http://thread.gmane.org/gmane.linux.kernel/1122051
A lot of kudos to Thilo-Alexander for reporting this tricky issue and
painstaking testing.
stable: This affects all kernels with cmwq, so all kernels since and
including v2.6.36 need this fix.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Thilo-Alexander Ginkel <thilo@ginkel.com>
Tested-by: Thilo-Alexander Ginkel <thilo@ginkel.com>
Cc: stable@kernel.org
Percpu allocator honors alignment request upto PAGE_SIZE and both the
percpu addresses in the percpu address space and the translated kernel
addresses should be aligned accordingly. The calculation of the
former depends on the alignment of percpu output section in the kernel
image.
The linker script macros PERCPU_VADDR() and PERCPU() are used to
define this output section and the latter takes @align parameter.
Several architectures are using @align smaller than PAGE_SIZE breaking
percpu memory alignment.
This patch removes @align parameter from PERCPU(), renames it to
PERCPU_SECTION() and makes it always align to PAGE_SIZE. While at it,
add PCPU_SETUP_BUG_ON() checks such that alignment problems are
reliably detected and remove percpu alignment comment recently added
in workqueue.c as the condition would trigger BUG way before reaching
there.
For um, this patch raises the alignment of percpu area. As the area
is in .init, there shouldn't be any noticeable difference.
This problem was discovered by David Howells while debugging boot
failure on mn10300.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Cc: uclinux-dist-devel@blackfin.uclinux.org
Cc: David Howells <dhowells@redhat.com>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: user-mode-linux-devel@lists.sourceforge.net
ksoftirqd, kworker, migration, and pktgend kthreads can be created with
kthread_create_on_node(), to get proper NUMA affinities for their stack and
task_struct.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: David Howells <dhowells@redhat.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-2.6.39' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
workqueue: fix build failure introduced by s/freezeable/freezable/
workqueue: add system_freezeable_wq
rds/ib: use system_wq instead of rds_ib_fmr_wq
net/9p: replace p9_poll_task with a work
net/9p: use system_wq instead of p9_mux_wq
xfs: convert to alloc_workqueue()
reiserfs: make commit_wq use the default concurrency level
ocfs2: use system_wq instead of ocfs2_quota_wq
ext4: convert to alloc_workqueue()
scsi/scsi_tgt_lib: scsi_tgtd isn't used in memory reclaim path
scsi/be2iscsi,qla2xxx: convert to alloc_workqueue()
misc/iwmc3200top: use system_wq instead of dedicated workqueues
i2o: use alloc_workqueue() instead of create_workqueue()
acpi: kacpi*_wq don't need WQ_MEM_RECLAIM
fs/aio: aio_wq isn't used in memory reclaim path
input/tps6507x-ts: use system_wq instead of dedicated workqueue
cpufreq: use system_wq instead of dedicated workqueues
wireless/ipw2x00: use system_wq instead of dedicated workqueues
arm/omap: use system_wq in mailbox
workqueue: use WQ_MEM_RECLAIM instead of WQ_RESCUER
In complex subsystems like mac80211 structures can contain several
timers and work structs, so identifying a specific instance from the
call trace and object type output of debugobjects can be hard.
Allow the subsystems which support debugobjects to provide a hint
function. This function returns a pointer to a kernel address
(preferrably the objects callback function) which is printed along
with the debugobjects type.
Add hint methods for timer_list, work_struct and hrtimer.
[ tglx: Massaged changelog, made it compile ]
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
LKML-Reference: <20110307085809.GA9334@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
wq:fixes-2.6.38 does s/WQ_FREEZEABLE/WQ_FREEZABLE and wq:for-2.6.39
adds new usage of the flag. The combination of the two creates a
build failure after merge. Fix it by renaming all freezeables to
freezables.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
MAYDAY_INITIAL_TIMEOUT is defined as HZ / 100 and depending on
configuration may end up 0 or 1. Even when it's 1, depending on when
the mayday timer is added in the current jiffy interval, it may expire
way before a jiffy has passed.
Make sure MAYDAY_INITIAL_TIMEOUT is at least two to guarantee that at
least a full jiffy has passed before calling rescuers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Ray Jui <rjui@broadcom.com>
Cc: stable@kernel.org
There are two spellings in use for 'freeze' + 'able' - 'freezable' and
'freezeable'. The former is the more prominent one. The latter is
mostly used by workqueue and in a few other odd places. Unify the
spelling to 'freezable'.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Alan Stern <stern@rowland.harvard.edu>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Alex Dubov <oakad@yahoo.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Steven Whitehouse <swhiteho@redhat.com>
After executing the matching works, a rescuer leaves the gcwq whether
there are more pending works or not. This may decrease the
concurrency level to zero and stall execution until a new work item is
queued on the gcwq.
Make rescuer wake up a regular worker when it leaves a gcwq if there
are more works to execute, so that execution isn't stalled.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Ray Jui <rjui@broadcom.com>
Cc: stable@kernel.org
The nested NOT_RUNNING test in worker_clr_flags() is slightly
misleading in that if NOT_RUNNING were a single flag the nested test
would be always %true and thus noop. Add a comment noting that the
test isn't a noop.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Currently, the lockdep annotation in flush_work() requires exclusive
access on the workqueue the target work is queued on and triggers
warning if a work is trying to flush another work on the same
workqueue; however, this is no longer true as workqueues can now
execute multiple works concurrently.
This patch adds lock_map_acquire_read() and make process_one_work()
hold read access to the workqueue while executing a work and
start_flush_work() check for write access if concurrnecy level is one
or the workqueue has a rescuer (as only one execution resource - the
rescuer - is guaranteed to be available under memory pressure), and
read access if higher.
This better represents what's going on and removes spurious lockdep
warnings which are triggered by fake dependency chain created through
flush_work().
* Peter pointed out that flushing another work from a WQ_MEM_RECLAIM
wq breaks forward progress guarantee under memory pressure.
Condition check accordingly updated.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Tested-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: stable@kernel.org
Currently, destroy_workqueue() makes the workqueue deny all new
queueing by setting WQ_DYING and flushes the workqueue once before
proceeding with destruction; however, there are cases where work items
queue more related work items. Currently, such users need to
explicitly flush the workqueue multiple times depending on the
possible depth of such chained queueing.
This patch updates the queueing path such that a work item can queue
further work items on the same workqueue even when WQ_DYING is set.
The flush on destruction is automatically retried until the workqueue
is empty. This guarantees that the workqueue is empty on destruction
while allowing chained queueing.
The flush retry logic whines if it takes too many retries to drain the
workqueue.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Running the annotate branch profiler on three boxes, including my
main box that runs firefox, evolution, xchat, and is part of the distcc farm,
showed this with the likelys in the workqueue code:
correct incorrect % Function File Line
------- --------- - -------- ---- ----
96 996253 99 wq_worker_sleeping workqueue.c 703
96 996247 99 wq_worker_waking_up workqueue.c 677
The likely()s in this case were assuming that WORKER_NOT_RUNNING will
most likely be false. But this is not the case. The reason is
(and shown by adding trace_printks and testing it) that most of the time
WORKER_PREP is set.
In worker_thread() we have:
worker_clr_flags(worker, WORKER_PREP);
[ do work stuff ]
worker_set_flags(worker, WORKER_PREP, false);
(that 'false' means not to wake up an idle worker)
The wq_worker_sleeping() is called from schedule when a worker thread
is putting itself to sleep. Which happens most of the time outside
of that [ do work stuff ].
The wq_worker_waking_up is called by the wakeup worker code, which
is also callod outside that [ do work stuff ].
Thus, the likely and unlikely used by those two functions are actually
backwards.
Remove the annotation and let gcc figure it out.
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Tejun Heo <tj@kernel.org>
I found a trivial bug on initialization of workqueue.
Current init_workqueues doesn't check the result of
allocation of system_unbound_wq, this should be checked
like other queues.
Signed-off-by: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: David Howells <dhowells@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Silly though it is, completions and wait_queue_heads use foo_ONSTACK
(COMPLETION_INITIALIZER_ONSTACK, DECLARE_COMPLETION_ONSTACK,
__WAIT_QUEUE_HEAD_INIT_ONSTACK and DECLARE_WAIT_QUEUE_HEAD_ONSTACK) so I
guess workqueues should do the same thing.
s/INIT_WORK_ON_STACK/INIT_WORK_ONSTACK/
s/INIT_DELAYED_WORK_ON_STACK/INIT_DELAYED_WORK_ONSTACK/
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In the MN10300 arch, we occasionally see an assertion being tripped in
alloc_cwqs() at the following line:
/* just in case, make sure it's actually aligned */
---> BUG_ON(!IS_ALIGNED(wq->cpu_wq.v, align));
return wq->cpu_wq.v ? 0 : -ENOMEM;
The values are:
wa->cpu_wq.v => 0x902776e0
align => 0x100
and align is calculated by the following:
const size_t align = max_t(size_t, 1 << WORK_STRUCT_FLAG_BITS,
__alignof__(unsigned long long));
This is because the pointer in question (wq->cpu_wq.v) loses some of its
lower bits to control flags, and so the object it points to must be
sufficiently aligned to avoid the need to use those bits for pointing to
things.
Currently, 4 control bits and 4 colour bits are used in normal
circumstances, plus a debugging bit if debugging is set. This requires
the cpu_workqueue_struct struct to be at least 256 bytes aligned (or 512
bytes aligned with debugging).
PERCPU() alignment on MN13000, however, is only 32 bytes as set in
vmlinux.lds.S. So we set this to PAGE_SIZE (4096) to match most other
arches and stick a comment in alloc_cwqs() for anyone else who triggers
the assertion.
Reported-by: Akira Takeuchi <takeuchi.akr@jp.panasonic.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Mark Salter <msalter@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit a25909a4 (lockdep: Add an in_workqueue_context() lockdep-based
test function) added in_workqueue_context() but there hasn't been any
in-kernel user and the lockdep annotation in workqueue is scheduled to
change. Remove the unused function.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The documentation for schedule_on_each_cpu() states that it calls a
function on each online CPU from keventd. This can easily be
interpreted as an asyncronous call because the description does not
mention that flush_work is called. Clarify that it is synchronous.
tj: rephrased a bit
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Add WQ_MEM_RECLAIM flag which currently maps to WQ_RESCUER, mark
WQ_RESCUER as internal and replace all external WQ_RESCUER usages to
WQ_MEM_RECLAIM.
This makes the API users express the intent of the workqueue instead
of indicating the internal mechanism used to guarantee forward
progress. This is also to make it cleaner to add more semantics to
WQ_MEM_RECLAIM. For example, if deemed necessary, memory reclaim
workqueues can be made highpri.
This patch doesn't introduce any functional change.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jeff Garzik <jgarzik@pobox.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
The policy function keep_working() didn't check GCWQ_HIGHPRI_PENDING
and could return %false with highpri work pending. This could lead to
late execution of a highpri work which was delayed due to @max_active
throttling if other works are actively consuming CPU cycles.
For example, the following could happen.
1. Work W0 which burns CPU cycles.
2. Two works W1 and W2 are queued to a highpri wq w/ @max_active of 1.
3. W1 starts executing and W2 is put to delayed queue. W0 and W1 are
both runnable.
4. W1 finishes which puts W2 to pending queue but keep_working()
incorrectly returns %false and the worker goes to sleep.
5. W0 finishes and W2 starts execution.
With this patch applied, W2 starts execution as soon as W1 finishes.
Signed-off-by: Tejun Heo <tj@kernel.org>
These two tracepoints allow tracking when and how a work is queued and
activated. This patch is based on Frederic's patch to add queue_work
trace point.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Define workqueue_work event class and use it for workqueue_execute_end
trace point. Also, move trace/events/workqueue.h include downwards
such that all struct definitions are visible to it. This is to
prepare for more tracepoints and doesn't cause any functional change.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Implement flush[_delayed]_work_sync(). These are flush functions
which also make sure no CPU is still executing the target work from
earlier queueing instances. These are similar to
cancel[_delayed]_work_sync() except that the target work item is
flushed instead of cancelled.
Signed-off-by: Tejun Heo <tj@kernel.org>
Factor out start_flush_work() from flush_work(). start_flush_work()
has @wait_executing argument which controls whether the barrier is
queued only if the work is pending or also if executing. As
flush_work() needs to wait for execution too, it uses %true.
This commit doesn't cause any behavior difference. start_flush_work()
will be used to implement flush_work_sync().
Signed-off-by: Tejun Heo <tj@kernel.org>
Make the following cleanup changes.
* Relocate flush/cancel function prototypes and definitions.
* Relocate wait_on_cpu_work() and wait_on_work() before
try_to_grab_pending(). These will be used to implement
flush_work_sync().
* Make all flush/cancel functions return bool instead of int.
* Update wait_on_cpu_work() and wait_on_work() to return %true if they
actually waited.
* Add / update comments.
This patch doesn't cause any functional changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
alloc_mayday_mask() was using alloc_cpumask_var() making
gcwq->mayday_mask contain garbage after initialization on
CONFIG_CPUMASK_OFFSTACK=y configurations. This combined with the
previously fixed GCWQ_DISASSOCIATED initialization bug could make
rescuers fall into infinite loop trying to bind to an offline cpu.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: CAI Qian <caiqian@redhat.com>
init_workqueues() incorrectly marks workqueues for all possible CPUs
associated. Combined with mayday_mask initialization bug, this can
make rescuers keep trying to bind to an offline gcwq indefinitely.
Fix init_workqueues() such that only online CPUs have their gcwqs have
GCWQ_DISASSOCIATED cleared.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: CAI Qian <caiqian@redhat.com>
cwq->nr_active is used to keep track of how many work items are active
for the cpu workqueue, where 'active' is defined as either pending on
global worklist or executing. This is used to implement the
max_active limit and workqueue freezing. If a work item is queued
after nr_active has already reached max_active, the work item doesn't
increment nr_active and is put on the delayed queue and gets activated
later as previous active work items retire.
try_to_grab_pending() which is used in the cancellation path
unconditionally decremented nr_active whether the work item being
cancelled is currently active or delayed, so cancelling a delayed work
item makes nr_active underflow. This breaks max_active enforcement
and triggers BUG_ON() in destroy_workqueue() later on.
This patch fixes this bug by adding a flag WORK_STRUCT_DELAYED, which
is set while a work item in on the delayed list and making
try_to_grab_pending() decrement nr_active iff the work item is
currently active.
The addition of the flag enlarges cwq alignment to 256 bytes which is
getting a bit too large. It's scheduled to be reduced back to 128
bytes by merging WORK_STRUCT_PENDING and WORK_STRUCT_CWQ in the next
devel cycle.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Johannes Berg <johannes@sipsolutions.net>
Now that the worklist is global, having works pending after wq
destruction can easily lead to oops and destroy_workqueue() have
several BUG_ON()s to catch these cases. Unfortunately, BUG_ON()
doesn't tell much about how the work became pending after the final
flush_workqueue().
This patch adds WQ_DYING which is set before the final flush begins.
If a work is requested to be queued on a dying workqueue,
WARN_ON_ONCE() is triggered and the request is ignored. This clearly
indicates which caller is trying to queue a work on a dying workqueue
and keeps the system working in most cases.
Locking rule comment is updated such that the 'I' rule includes
modifying the field from destruction path.
Signed-off-by: Tejun Heo <tj@kernel.org>
worker_maybe_bind_and_lock() actually grabs gcwq->lock but was missing proper
annotation. Add it. So this patch will remove following sparse warnings:
kernel/workqueue.c:1214:13: warning: context imbalance in 'worker_maybe_bind_and_lock' - wrong count at exit
arch/x86/include/asm/irqflags.h:44:9: warning: context imbalance in 'worker_rebind_fn' - unexpected unlock
kernel/workqueue.c:1991:17: warning: context imbalance in 'rescuer_thread' - unexpected unlock
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Some of internal functions called within gcwq->lock context releases and
regrabs the lock but were missing proper annotations. Add it.
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
With the introduction of the new unified work queue thread pools,
we lost one feature: It's no longer possible to know which worker
is causing the CPU to wake out of idle. The result is that PowerTOP
now reports a lot of "kworker/a:b" instead of more readable results.
This patch adds a pair of tracepoints to the new workqueue code,
similar in style to the timer/hrtimer tracepoints.
With this pair of tracepoints, the next PowerTOP can correctly
report which work item caused the wakeup (and how long it took):
Interrupt (43) i915 time 3.51ms wakeups 141
Work ieee80211_iface_work time 0.81ms wakeups 29
Work do_dbs_timer time 0.55ms wakeups 24
Process Xorg time 21.36ms wakeups 4
Timer sched_rt_period_timer time 0.01ms wakeups 1
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
wq->rescuer is not freed when wq is destroyed, leads a memory leak
then. This patch also remove a redundant line.
Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Commit 6ee0578b (workqueue: mark init_workqueues as early_initcall)
made workqueue SMP initialization depend on workqueue_cpu_callback(),
which however was registered as hotcpu_notifier() and didn't get
called if CONFIG_HOTPLUG_CPU is not set. This made gcwqs on non-boot
CPUs not create their initial workers leading to boot failures. Fix
it by making it a cpu_notifier.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-bisected-by: walt <w41ter@gmail.com>
Tested-by: Markus Trippelsdorf <markus@trippelsdorf.de>
works in schecule_on_each_cpu() is a percpu pointer but was missing
__percpu markup. Add it.
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq: (55 commits)
workqueue: mark init_workqueues() as early_initcall()
workqueue: explain for_each_*cwq_cpu() iterators
fscache: fix build on !CONFIG_SYSCTL
slow-work: kill it
gfs2: use workqueue instead of slow-work
drm: use workqueue instead of slow-work
cifs: use workqueue instead of slow-work
fscache: drop references to slow-work
fscache: convert operation to use workqueue instead of slow-work
fscache: convert object to use workqueue instead of slow-work
workqueue: fix how cpu number is stored in work->data
workqueue: fix mayday_mask handling on UP
workqueue: fix build problem on !CONFIG_SMP
workqueue: fix locking in retry path of maybe_create_worker()
async: use workqueue for worker pool
workqueue: remove WQ_SINGLE_CPU and use WQ_UNBOUND instead
workqueue: implement unbound workqueue
workqueue: prepare for WQ_UNBOUND implementation
libata: take advantage of cmwq and remove concurrency limitations
workqueue: fix worker management invocation without pending works
...
Fixed up conflicts in fs/cifs/* as per Tejun. Other trivial conflicts in
include/linux/workqueue.h, kernel/trace/Kconfig and kernel/workqueue.c
Mark init_workqueues() as early_initcall() and thus it will be initialized
before smp bringup. init_workqueues() registers for the hotcpu notifier
and thus it should cope with the processors that are brought online after
the workqueues are initialized.
x86 smp bringup code uses workqueues and uses a workaround for the
cold boot process (as the workqueues are initialized post smp_init()).
Marking init_workqueues() as early_initcall() will pave the way for
cleaning up this code.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
for_each_*cwq_cpu() are similar to regular CPU iterators except that
it also considers the pseudo CPU number used for unbound workqueues.
Explain them.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Once a work starts execution, its data contains the cpu number it was
on instead of pointing to cwq. This is added by commit 7a22ad75
(workqueue: carry cpu number in work data once execution starts) to
reliably determine the work was last on even if the workqueue itself
was destroyed inbetween.
Whether data points to a cwq or contains a cpu number was
distinguished by comparing the value against PAGE_OFFSET. The
assumption was that a cpu number should be below PAGE_OFFSET while a
pointer to cwq should be above it. However, on architectures which
use separate address spaces for user and kernel spaces, this doesn't
hold as PAGE_OFFSET is zero.
Fix it by using an explicit flag, WORK_STRUCT_CWQ, to mark what the
data field contains. If the flag is set, it's pointing to a cwq;
otherwise, it contains a cpu number.
Reported on s390 and microblaze during linux-next testing.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Sachin Sant <sachinp@in.ibm.com>
Reported-by: Michal Simek <michal.simek@petalogix.com>
Reported-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Tested-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Tested-by: Michal Simek <monstr@monstr.eu>
All cpumasks are assumed to have cpu 0 permanently set on UP, so it
can't be used to signify whether there's something to be done for the
CPU. workqueue was using cpumask to track which CPU requested rescuer
assistance and this led rescuer thread to think there always are
pending mayday requests on UP, which resulted in infinite busy loops.
This patch fixes the problem by introducing mayday_mask_t and
associated helpers which wrap cpumask on SMP and emulates its behavior
using bitops and unsigned long on UP.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Commit f3421797 (workqueue: implement unbound workqueue) incorrectly
tested CONFIG_SMP as part of a C expression in alloc/free_cwqs(). As
CONFIG_SMP is not defined in UP, this breaks build. Fix it by using
Found during linux-next build test.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
maybe_create_worker() mismanaged locking when worker creation fails
and it has to retry. Fix locking and simplify lock manipulation.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Yong Zhang <yong.zhang@windriver.com>
WQ_SINGLE_CPU combined with @max_active of 1 is used to achieve full
ordering among works queued to a workqueue. The same can be achieved
using WQ_UNBOUND as unbound workqueues always use the gcwq for
WORK_CPU_UNBOUND. As @max_active is always one and benefits from cpu
locality isn't accessible anyway, serving them with unbound workqueues
should be fine.
Drop WQ_SINGLE_CPU support and use WQ_UNBOUND instead. Note that most
single thread workqueue users will be converted to use multithread or
non-reentrant instead and only the ones which require strict ordering
will keep using WQ_UNBOUND + @max_active of 1.
Signed-off-by: Tejun Heo <tj@kernel.org>
This patch implements unbound workqueue which can be specified with
WQ_UNBOUND flag on creation. An unbound workqueue has the following
properties.
* It uses a dedicated gcwq with a pseudo CPU number WORK_CPU_UNBOUND.
This gcwq is always online and disassociated.
* Workers are not bound to any CPU and not concurrency managed. Works
are dispatched to workers as soon as possible and the only applied
limitation is @max_active. IOW, all unbound workqeueues are
implicitly high priority.
Unbound workqueues can be used as simple execution context provider.
Contexts unbound to any cpu are served as soon as possible.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: David Howells <dhowells@redhat.com>
In preparation of WQ_UNBOUND addition, make the following changes.
* Add WORK_CPU_* constants for pseudo cpu id numbers used (currently
only WORK_CPU_NONE) and use them instead of NR_CPUS. This is to
allow another pseudo cpu id for unbound cpu.
* Reorder WQ_* flags.
* Make workqueue_struct->cpu_wq a union which contains a percpu
pointer, regular pointer and an unsigned long value and use
kzalloc/kfree() in UP allocation path. This will be used to
implement unbound workqueues which will use only one cwq on SMPs.
* Move alloc_cwqs() allocation after initialization of wq fields, so
that alloc_cwqs() has access to wq->flags.
* Trivial relocation of wq local variables in freeze functions.
These changes don't cause any functional change.
Signed-off-by: Tejun Heo <tj@kernel.org>
When there's no pending work to do, worker_thread() goes back to sleep
after waking up without checking whether worker management is
necessary. This means that idle worker exit requests can be ignored
if the gcwq stays empty.
Fix it by making worker_thread() always check whether worker
management is necessary before going to sleep.
Signed-off-by: Tejun Heo <tj@kernel.org>
get_work_gcwq() was incorrectly triggering BUG_ON() if cpu number is
equal to or higher than num_possible_cpus() instead of nr_cpu_ids.
Fix it.
Signed-off-by: Tejun Heo <tj@kernel.org>
When one flusher is cascading to the next flusher, it first sets
wq->first_flusher to the next one and sets up the next flush cycle.
If there's nothing to do for the next cycle, it clears
wq->flush_flusher and proceeds to the one after that.
If the woken up flusher checks wq->first_flusher before it gets
cleared, it will incorrectly assume the role of the first flusher,
which triggers BUG_ON() sanity check.
Fix it by checking wq->first_flusher again after grabbing the mutex.
Signed-off-by: Tejun Heo <tj@kernel.org>
worker_set/clr_flags() assume that if none of NOT_RUNNING flags is set
the worker must be contributing to nr_running which is only true if
the worker is actually running.
As when called from self, it is guaranteed that the worker is running,
those functions can be safely used from the worker itself and they
aren't necessary from other places anyway. Make the following changes
to fix the bug.
* Make worker_set/clr_flags() whine if not called from self.
* Convert all places which called those functions from other tasks to
manipulate flags directly.
* Make trustee_thread() directly clear nr_running after setting
WORKER_ROGUE on all workers. This is the only place where
nr_running manipulation is necessary outside of workers themselves.
* While at it, add sanity check for nr_running in worker_enter_idle().
Signed-off-by: Tejun Heo <tj@kernel.org>
This patch implements cpu intensive workqueue which can be specified
with WQ_CPU_INTENSIVE flag on creation. Works queued to a cpu
intensive workqueue don't participate in concurrency management. IOW,
it doesn't contribute to gcwq->nr_running and thus doesn't delay
excution of other works.
Note that although cpu intensive works won't delay other works, they
can be delayed by other works. Combine with WQ_HIGHPRI to avoid being
delayed by other works too.
As the name suggests this is useful when using workqueue for cpu
intensive works. Workers executing cpu intensive works are not
considered for workqueue concurrency management and left for the
scheduler to manage.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
This patch implements high priority workqueue which can be specified
with WQ_HIGHPRI flag on creation. A high priority workqueue has the
following properties.
* A work queued to it is queued at the head of the worklist of the
respective gcwq after other highpri works, while normal works are
always appended at the end.
* As long as there are highpri works on gcwq->worklist,
[__]need_more_worker() remains %true and process_one_work() wakes up
another worker before it start executing a work.
The above two properties guarantee that works queued to high priority
workqueues are dispatched to workers and start execution as soon as
possible regardless of the state of other works.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Implement the following utility APIs.
workqueue_set_max_active() : adjust max_active of a wq
workqueue_congested() : test whether a wq is contested
work_cpu() : determine the last / current cpu of a work
work_busy() : query whether a work is busy
* Anton Blanchard fixed missing ret initialization in work_busy().
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Anton Blanchard <anton@samba.org>
This patch makes changes to make new workqueue features available to
its users.
* Now that workqueue is more featureful, there should be a public
workqueue creation function which takes paramters to control them.
Rename __create_workqueue() to alloc_workqueue() and make 0
max_active mean WQ_DFL_ACTIVE. In the long run, all
create_workqueue_*() will be converted over to alloc_workqueue().
* To further unify access interface, rename keventd_wq to system_wq
and export it.
* Add system_long_wq and system_nrt_wq. The former is to host long
running works separately (so that flush_scheduled_work() dosen't
take so long) and the latter guarantees any queued work item is
never executed in parallel by multiple CPUs. These will be used by
future patches to update workqueue users.
Signed-off-by: Tejun Heo <tj@kernel.org>
Define WQ_MAX_ACTIVE and create keventd with max_active set to half of
it which means that keventd now can process upto WQ_MAX_ACTIVE / 2 - 1
works concurrently. Unless some combination can result in dependency
loop longer than max_active, deadlock won't happen and thus it's
unnecessary to check whether current_is_keventd() before trying to
schedule a work. Kill current_is_keventd().
(Lockdep annotations are broken. We need lock_map_acquire_read_norecurse())
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Instead of creating a worker for each cwq and putting it into the
shared pool, manage per-cpu workers dynamically.
Works aren't supposed to be cpu cycle hogs and maintaining just enough
concurrency to prevent work processing from stalling due to lack of
processing context is optimal. gcwq keeps the number of concurrent
active workers to minimum but no less. As long as there's one or more
running workers on the cpu, no new worker is scheduled so that works
can be processed in batch as much as possible but when the last
running worker blocks, gcwq immediately schedules new worker so that
the cpu doesn't sit idle while there are works to be processed.
gcwq always keeps at least single idle worker around. When a new
worker is necessary and the worker is the last idle one, the worker
assumes the role of "manager" and manages the worker pool -
ie. creates another worker. Forward-progress is guaranteed by having
dedicated rescue workers for workqueues which may be necessary while
creating a new worker. When the manager is having problem creating a
new worker, mayday timer activates and rescue workers are summoned to
the cpu and execute works which might be necessary to create new
workers.
Trustee is expanded to serve the role of manager while a CPU is being
taken down and stays down. As no new works are supposed to be queued
on a dead cpu, it just needs to drain all the existing ones. Trustee
continues to try to create new workers and summon rescuers as long as
there are pending works. If the CPU is brought back up while the
trustee is still trying to drain the gcwq from the previous offlining,
the trustee will kill all idles ones and tell workers which are still
busy to rebind to the cpu, and pass control over to gcwq which assumes
the manager role as necessary.
Concurrency managed worker pool reduces the number of workers
drastically. Only workers which are necessary to keep the processing
going are created and kept. Also, it reduces cache footprint by
avoiding unnecessarily switching contexts between different workers.
Please note that this patch does not increase max_active of any
workqueue. All workqueues can still only process one work per cpu.
Signed-off-by: Tejun Heo <tj@kernel.org>
Implement worker_{set|clr}_flags() to manipulate worker flags. These
are currently simple wrappers but logics to track the current worker
state and the current level of concurrency will be added.
Signed-off-by: Tejun Heo <tj@kernel.org>
Use gcwq->worklist instead of cwq->worklist and break the strict
association between a cwq and its worker. All works queued on a cpu
are queued on gcwq->worklist and processed by any available worker on
the gcwq.
As there no longer is strict association between a cwq and its worker,
whether a work is executing can now only be determined by calling
[__]find_worker_executing_work().
After this change, the only association between a cwq and its worker
is that a cwq puts a worker into shared worker pool on creation and
kills it on destruction. As all workqueues are still limited to
max_active of one, this means that there are always at least as many
workers as active works and thus there's no danger for deadlock.
The break of strong association between cwqs and workers requires
somewhat clumsy changes to current_is_keventd() and
destroy_workqueue(). Dynamic worker pool management will remove both
clumsy changes. current_is_keventd() won't be necessary at all as the
only reason it exists is to avoid queueing a work from a work which
will be allowed just fine. The clumsy part of destroy_workqueue() is
added because a worker can only be destroyed while idle and there's no
guarantee a worker is idle when its wq is going down. With dynamic
pool management, workers are not associated with workqueues at all and
only idle ones will be submitted to destroy_workqueue() so the code
won't be necessary anymore.
Signed-off-by: Tejun Heo <tj@kernel.org>
With gcwq managing all the workers and work->data pointing to the last
gcwq it was on, non-reentrance can be easily implemented by checking
whether the work is still running on the previous gcwq on queueing.
Implement it.
Signed-off-by: Tejun Heo <tj@kernel.org>
To implement non-reentrant workqueue, the last gcwq a work was
executed on must be reliably obtainable as long as the work structure
is valid even if the previous workqueue has been destroyed.
To achieve this, work->data will be overloaded to carry the last cpu
number once execution starts so that the previous gcwq can be located
reliably. This means that cwq can't be obtained from work after
execution starts but only gcwq.
Implement set_work_{cwq|cpu}(), get_work_[g]cwq() and
clear_work_data() to set work data to the cpu number when starting
execution, access the overloaded work data and clear it after
cancellation.
queue_delayed_work_on() is updated to preserve the last cpu while
in-flight in timer and other callers which depended on getting cwq
from work after execution starts are converted to depend on gcwq
instead.
* Anton Blanchard fixed compile error on powerpc due to missing
linux/threads.h include.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Anton Blanchard <anton@samba.org>
Now that all the workers are tracked by gcwq, we can find which worker
is executing a work from gcwq. Implement find_worker_executing_work()
and make worker track its current_cwq so that we can find things the
other way around. This will be used to implement non-reentrant wqs.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reimplement st (single thread) workqueue so that it's friendly to
shared worker pool. It was originally implemented by confining st
workqueues to use cwq of a fixed cpu and always having a worker for
the cpu. This implementation isn't very friendly to shared worker
pool and suboptimal in that it ends up crossing cpu boundaries often.
Reimplement st workqueue using dynamic single cpu binding and
cwq->limit. WQ_SINGLE_THREAD is replaced with WQ_SINGLE_CPU. In a
single cpu workqueue, at most single cwq is bound to the wq at any
given time. Arbitration is done using atomic accesses to
wq->single_cpu when queueing a work. Once bound, the binding stays
till the workqueue is drained.
Note that the binding is never broken while a workqueue is frozen.
This is because idle cwqs may have works waiting in delayed_works
queue while frozen. On thaw, the cwq is restarted if there are any
delayed works or unbound otherwise.
When combined with max_active limit of 1, single cpu workqueue has
exactly the same execution properties as the original single thread
workqueue while allowing sharing of per-cpu workers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reimplement CPU hotplugging support using trustee thread. On CPU
down, a trustee thread is created and each step of CPU down is
executed by the trustee and workqueue_cpu_callback() simply drives and
waits for trustee state transitions.
CPU down operation no longer waits for works to be drained but trustee
sticks around till all pending works have been completed. If CPU is
brought back up while works are still draining,
workqueue_cpu_callback() tells trustee to step down and tell workers
to rebind to the cpu.
As it's difficult to tell whether cwqs are empty if it's freezing or
frozen, trustee doesn't consider draining to be complete while a gcwq
is freezing or frozen (tracked by new GCWQ_FREEZING flag). Also,
workers which get unbound from their cpu are marked with WORKER_ROGUE.
Trustee based implementation doesn't bring any new feature at this
point but it will be used to manage worker pool when dynamic shared
worker pool is implemented.
Signed-off-by: Tejun Heo <tj@kernel.org>
Implement worker states. After created, a worker is STARTED. While a
worker isn't processing a work, it's IDLE and chained on
gcwq->idle_list. While processing a work, a worker is BUSY and
chained on gcwq->busy_hash. Also, gcwq now counts the number of all
workers and idle ones.
worker_thread() is restructured to reflect state transitions.
cwq->more_work is removed and waking up a worker makes it check for
events. A worker is killed by setting DIE flag while it's IDLE and
waking it up.
This gives gcwq better visibility of what's going on and allows it to
find out whether a work is executing quickly which is necessary to
have multiple workers processing the same cwq.
Signed-off-by: Tejun Heo <tj@kernel.org>
Tejun Heo