Currently if a user enqueue a work item using schedule_delayed_work() the used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to schedule_work() that is using system_wq and queue_work(), that makes use again of WORK_CPU_UNBOUND. This lack of consistentcy cannot be addressed without refactoring the API. alloc_workqueue() treats all queues as per-CPU by default, while unbound workqueues must opt-in via WQ_UNBOUND. This default is suboptimal: most workloads benefit from unbound queues, allowing the scheduler to place worker threads where they’re needed and reducing noise when CPUs are isolated. This patch adds a new WQ_PERCPU flag to explicitly request the use of the per-CPU behavior. Both flags coexist for one release cycle to allow callers to transition their calls. Once migration is complete, WQ_UNBOUND can be removed and unbound will become the implicit default. With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND), any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND must now use WQ_PERCPU. All existing users have been updated accordingly.
Showing the changes to kernel/workqueue.c (1 of 2 files). Show all
Unified · Split
| @@ -7828,22 +7828,22 @@ void __init workqueue_init_early(void) · 7827 unmodified lines | ||
| 7828 | 7828 | ordered_wq_attrs[i] = attrs; |
| 7829 | 7829 | } |
| 7830 | 7830 | |
| 7831 | − system_wq = alloc_workqueue("events", 0, 0); | |
| 7832 | − system_percpu_wq = alloc_workqueue("events", 0, 0); | |
| 7833 | − system_highpri_wq = alloc_workqueue("events_highpri", WQ_HIGHPRI, 0); | |
| 7834 | − system_long_wq = alloc_workqueue("events_long", 0, 0); | |
| 7831 | + system_wq = alloc_workqueue("events", WQ_PERCPU, 0); | |
| 7832 | + system_percpu_wq = alloc_workqueue("events", WQ_PERCPU, 0); | |
| 7833 | + system_highpri_wq = alloc_workqueue("events_highpri", | |
| 7834 | + WQ_HIGHPRI | WQ_PERCPU, 0); | |
| 7835 | + system_long_wq = alloc_workqueue("events_long", WQ_PERCPU, 0); | |
| 7835 | 7836 | system_unbound_wq = alloc_workqueue("events_unbound", WQ_UNBOUND, WQ_MAX_ACTIVE); |
| 7836 | 7837 | system_dfl_wq = alloc_workqueue("events_unbound", WQ_UNBOUND, WQ_MAX_ACTIVE); |
| 7837 | 7838 | system_freezable_wq = alloc_workqueue("events_freezable", |
| 7838 | − WQ_FREEZABLE, 0); | |
| 7839 | + WQ_FREEZABLE | WQ_PERCPU, 0); | |
| 7839 | 7840 | system_power_efficient_wq = alloc_workqueue("events_power_efficient", |
| 7840 | − WQ_POWER_EFFICIENT, 0); | |
| 7841 | + WQ_POWER_EFFICIENT | WQ_PERCPU, 0); | |
| 7841 | 7842 | system_freezable_power_efficient_wq = alloc_workqueue("events_freezable_pwr_efficient", |
| 7842 | − WQ_FREEZABLE | WQ_POWER_EFFICIENT, | |
| 7843 | − 0); | |
| 7844 | − system_bh_wq = alloc_workqueue("events_bh", WQ_BH, 0); | |
| 7843 | + WQ_FREEZABLE | WQ_POWER_EFFICIENT | WQ_PERCPU, 0); | |
| 7844 | + system_bh_wq = alloc_workqueue("events_bh", WQ_BH | WQ_PERCPU, 0); | |
| 7845 | 7845 | system_bh_highpri_wq = alloc_workqueue("events_bh_highpri", |
| 7846 | − WQ_BH | WQ_HIGHPRI, 0); | |
| 7846 | + WQ_BH | WQ_HIGHPRI | WQ_PERCPU, 0); | |
| 7847 | 7847 | BUG_ON(!system_wq || !system_percpu_wq|| !system_highpri_wq || !system_long_wq || |
| 7848 | 7848 | !system_unbound_wq || !system_freezable_wq || !system_dfl_wq || |
| 7849 | 7849 | !system_power_efficient_wq || |