workqueue: Add new WQ_PERCPU flag
authorMarco Crivellari <marco.crivellari@suse.com>
Sat, 14 Jun 2025 13:35:30 +0000 (15:35 +0200)
committerTejun Heo <tj@kernel.org>
Tue, 17 Jun 2025 18:52:09 +0000 (08:52 -1000)
commit930c2ea566aff59e962c50b2421d5fcc3b98b8be
treebc7851f564351e810bf3103ccd91876af6c4d5d6
parent128ea9f6ccfb6960293ae4212f4f97165e42222d
workqueue: Add new WQ_PERCPU flag

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.

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.

tj: Merged doc patch.

Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Documentation/core-api/workqueue.rst
include/linux/workqueue.h