| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
An unbound workqueue can be served by multiple worker_pools to improve
locality. The segmentation is achieved by grouping CPUs into pods. By
default, the cache boundaries according to cpus_share_cache() define the
CPUs are grouped. Let's a workqueue is allowed to run on all CPUs and the
system has two L3 caches. The workqueue would be mapped to two worker_pools
each serving one L3 cache domains.
While this improves locality, because the pod boundaries are strict, it
limits the total bandwidth a given issuer can consume. For example, let's
say there is a thread pinned to a CPU issuing enough work items to saturate
the whole machine. With the machine segmented into two pods, no matter how
many work items it issues, it can only use half of the CPUs on the system.
While this limitation has existed for a very long time, it wasn't very
pronounced because the affinity grouping used to be always by NUMA nodes.
With cache boundaries as the default and support for even finer grained
scopes (smt and cpu), it is now an a lot more pressing problem.
This patch implements non-strict affinity scope where the pod boundaries
aren't enforced strictly. Going back to the previous example, the workqueue
would still be mapped to two worker_pools; however, the affinity enforcement
would be soft. The workers in both pools would have their cpus_allowed set
to the whole machine thus allowing the scheduler to migrate them anywhere on
the machine. However, whenever an idle worker is woken up, the workqueue
code asks the scheduler to bring back the task within the pod if the worker
is outside. ie. work items start executing within its affinity scope but can
be migrated outside as the scheduler sees fit. This removes the hard cap on
utilization while maintaining the benefits of affinity scopes.
After the earlier ->__pod_cpumask changes, the implementation is pretty
simple. When non-strict which is the new default:
* pool_allowed_cpus() returns @pool->attrs->cpumask instead of
->__pod_cpumask so that the workers are allowed to run on any CPU that
the associated workqueues allow.
* If the idle worker task's ->wake_cpu is outside the pod, kick_pool() sets
the field to a CPU within the pod.
This would be the first use of task_struct->wake_cpu outside scheduler
proper, so it isn't clear whether this would be acceptable. However, other
methods of migrating tasks are significantly more expensive and are likely
prohibitively so if we want to do this on every work item. This needs
discussion with scheduler folks.
There is also a race window where setting ->wake_cpu wouldn't be effective
as the target task is still on CPU. However, the window is pretty small and
this being a best-effort optimization, it doesn't seem to warrant more
complexity at the moment.
While the non-strict cache affinity scopes seem to be the best option, the
performance picture interacts with the affinity scope and is a bit
complicated to fully discuss in this patch, so the behavior is made easily
selectable through wqattrs and sysfs and the next patch will add
documentation to discuss performance implications.
v2: pool->attrs->affn_strict is set to true for per-cpu worker_pools.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add three more affinity scopes - WQ_AFFN_CPU, SMT and CACHE - and make CACHE
the default. The code changes to actually add the additional scopes are
trivial.
Also add module parameter "workqueue.default_affinity_scope" to override the
default scope and "affinity_scope" sysfs file to configure it per workqueue.
wq_dump.py and documentations are updated accordingly.
This enables significant flexibility in configuring how unbound workqueues
behave. If affinity scope is set to "cpu", it'll behave close to a per-cpu
workqueue. On the other hand, "system" removes all locality boundaries.
Many modern machines have multiple L3 caches often while being mostly
uniform in terms of memory access. Thus, workqueue's previous behavior of
spreading work items in each NUMA node had negative performance implications
from unncessarily crossing L3 boundaries between issue and execution.
However, picking a finer grained affinity scope also has a downside in that
an issuer in one group can't utilize CPUs in other groups.
While dependent on the specifics of workload, there's usually a noticeable
penalty in crossing L3 boundaries, so let's default to CACHE. This issue
will be further addressed and documented with examples in future patches.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
configuration
Lack of visibility has always been a pain point for workqueues. While the
recently added wq_monitor.py improved the situation, it's still difficult to
understand what worker pools are active in the system, how workqueues map to
them and why. The lack of visibility into how workqueues are configured is
going to become more noticeable as workqueue improves locality awareness and
provides more mechanisms to customize locality related behaviors.
Now that the basic framework for more flexible locality support is in place,
this is a good time to improve the situation. This patch adds
tools/workqueues/wq_dump.py which prints out the topology configuration,
worker pools and how workqueues are mapped to pools. Read the command's help
message for more details.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Now that wq_worker_tick() is there, we can easily track the rough CPU time
consumption of each workqueue by charging the whole tick whenever a tick
hits an active workqueue. While not super accurate, it provides reasonable
visibility into the workqueues that consume a lot of CPU cycles.
wq_monitor.py is updated to report the per-workqueue CPU times.
v2: wq_monitor.py was using "cputime" as the key when outputting in json
format. Use "cpu_time" instead for consistency with other fields.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If a per-cpu work item hogs the CPU, it can prevent other work items from
starting through concurrency management. A per-cpu workqueue which intends
to host such CPU-hogging work items can choose to not participate in
concurrency management by setting %WQ_CPU_INTENSIVE; however, this can be
error-prone and difficult to debug when missed.
This patch adds an automatic CPU usage based detection. If a
concurrency-managed work item consumes more CPU time than the threshold
(10ms by default) continuously without intervening sleeps, wq_worker_tick()
which is called from scheduler_tick() will detect the condition and
automatically mark it CPU_INTENSIVE.
The mechanism isn't foolproof:
* Detection depends on tick hitting the work item. Getting preempted at the
right timings may allow a violating work item to evade detection at least
temporarily.
* nohz_full CPUs may not be running ticks and thus can fail detection.
* Even when detection is working, the 10ms detection delays can add up if
many CPU-hogging work items are queued at the same time.
However, in vast majority of cases, this should be able to detect violations
reliably and provide reasonable protection with a small increase in code
complexity.
If some work items trigger this condition repeatedly, the bigger problem
likely is the CPU being saturated with such per-cpu work items and the
solution would be making them UNBOUND. The next patch will add a debug
mechanism to help spot such cases.
v4: Documentation for workqueue.cpu_intensive_thresh_us added to
kernel-parameters.txt.
v3: Switch to use wq_worker_tick() instead of hooking into preemptions as
suggested by Peter.
v2: Lai pointed out that wq_worker_stopping() also needs to be called from
preemption and rtlock paths and an earlier patch was updated
accordingly. This patch adds a comment describing the risk of infinte
recursions and how they're avoided.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
|
|
|
Currently, the only way to peer into workqueue operations is through
tracing. While possible, it isn't easy or convenient to monitor
per-workqueue behaviors over time this way. Let's add pwq->stats[] that
track relevant events and a drgn monitoring script -
tools/workqueue/wq_monitor.py.
It's arguable whether this needs to be configurable. However, it currently
only has several counters and the runtime overhead shouldn't be noticeable
given that they're on pwq's which are per-cpu on per-cpu workqueues and
per-numa-node on unbound ones. Let's keep it simple for the time being.
v2: Patch reordered to earlier with fewer fields. Field will be added back
gradually. Help message improved.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
|
