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. system_wq is a per-CPU worqueue, yet nothing in its name tells about that CPU affinity constraint, which is very often not required by users. Make it clear by adding a system_percpu_wq. system_unbound_wq should be the default workqueue so as not to enforce locality constraints for random work whenever it's not required. Adding system_dfl_wq to encourage its use when unbound work should be used.
Showing the changes to kernel/workqueue.c (1 of 2 files). Show all
Unified · Split
| @@ -505,12 +505,16 @@ static struct kthread_worker *pwq_release_worker __ro_after_init; · 504 unmodified lines | ||
| 505 | 505 | |
| 506 | 506 | struct workqueue_struct *system_wq __ro_after_init; |
| 507 | 507 | EXPORT_SYMBOL(system_wq); |
| 508 | +struct workqueue_struct *system_percpu_wq __ro_after_init; | |
| 509 | +EXPORT_SYMBOL(system_percpu_wq); | |
| 508 | 510 | struct workqueue_struct *system_highpri_wq __ro_after_init; |
| 509 | 511 | EXPORT_SYMBOL_GPL(system_highpri_wq); |
| 510 | 512 | struct workqueue_struct *system_long_wq __ro_after_init; |
| 511 | 513 | EXPORT_SYMBOL_GPL(system_long_wq); |
| 512 | 514 | struct workqueue_struct *system_unbound_wq __ro_after_init; |
| 513 | 515 | EXPORT_SYMBOL_GPL(system_unbound_wq); |
| 516 | +struct workqueue_struct *system_dfl_wq __ro_after_init; | |
| 517 | +EXPORT_SYMBOL_GPL(system_dfl_wq); | |
| 514 | 518 | struct workqueue_struct *system_freezable_wq __ro_after_init; |
| 515 | 519 | EXPORT_SYMBOL_GPL(system_freezable_wq); |
| 516 | 520 | struct workqueue_struct *system_power_efficient_wq __ro_after_init; |
| @@ -7816,10 +7820,11 @@ void __init workqueue_init_early(void) · 7299 unmodified lines | ||
| 7816 | 7820 | } |
| 7817 | 7821 | |
| 7818 | 7822 | system_wq = alloc_workqueue("events", 0, 0); |
| 7823 | + system_percpu_wq = alloc_workqueue("events", 0, 0); | |
| 7819 | 7824 | system_highpri_wq = alloc_workqueue("events_highpri", WQ_HIGHPRI, 0); |
| 7820 | 7825 | system_long_wq = alloc_workqueue("events_long", 0, 0); |
| 7821 | − system_unbound_wq = alloc_workqueue("events_unbound", WQ_UNBOUND, | |
| 7822 | − WQ_MAX_ACTIVE); | |
| 7826 | + system_unbound_wq = alloc_workqueue("events_unbound", WQ_UNBOUND, WQ_MAX_ACTIVE); | |
| 7827 | + system_dfl_wq = alloc_workqueue("events_unbound", WQ_UNBOUND, WQ_MAX_ACTIVE); | |
| 7823 | 7828 | system_freezable_wq = alloc_workqueue("events_freezable", |
| 7824 | 7829 | WQ_FREEZABLE, 0); |
| 7825 | 7830 | system_power_efficient_wq = alloc_workqueue("events_power_efficient", |
| @@ -7830,8 +7835,8 @@ void __init workqueue_init_early(void) · 4 unmodified lines | ||
| 7830 | 7835 | system_bh_wq = alloc_workqueue("events_bh", WQ_BH, 0); |
| 7831 | 7836 | system_bh_highpri_wq = alloc_workqueue("events_bh_highpri", |
| 7832 | 7837 | WQ_BH | WQ_HIGHPRI, 0); |
| 7833 | − BUG_ON(!system_wq || !system_highpri_wq || !system_long_wq || | |
| 7834 | − !system_unbound_wq || !system_freezable_wq || | |
| 7838 | + BUG_ON(!system_wq || !system_percpu_wq|| !system_highpri_wq || !system_long_wq || | |
| 7839 | + !system_unbound_wq || !system_freezable_wq || !system_dfl_wq || | |
| 7835 | 7840 | !system_power_efficient_wq || |
| 7836 | 7841 | !system_freezable_power_efficient_wq || |
| 7837 | 7842 | !system_bh_wq || !system_bh_highpri_wq); |