Currently if a user enqueues 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 consistency cannot be addressed without refactoring the API.

This continues the effort to refactor workqueue APIs, which began with
the introduction of new workqueues and a new alloc_workqueue flag in:

commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

Switch to using system_percpu_wq because system_wq is going away as part of
a workqueue restructuring.

Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/usb/host/xhci-dbgcap.c | 8 ++++----
 drivers/usb/host/xhci-ring.c   | 2 +-
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/usb/host/xhci-dbgcap.c b/drivers/usb/host/xhci-dbgcap.c
index ecda964e018a..9da4f3b452cb 100644
--- a/drivers/usb/host/xhci-dbgcap.c
+++ b/drivers/usb/host/xhci-dbgcap.c
@@ -374,7 +374,7 @@ int dbc_ep_queue(struct dbc_request *req)
 		ret = dbc_ep_do_queue(req);
 	spin_unlock_irqrestore(&dbc->lock, flags);
 
-	mod_delayed_work(system_wq, &dbc->event_work, 0);
+	mod_delayed_work(system_percpu_wq, &dbc->event_work, 0);
 
 	trace_xhci_dbc_queue_request(req);
 
@@ -677,7 +677,7 @@ static int xhci_dbc_start(struct xhci_dbc *dbc)
 		return ret;
 	}
 
-	return mod_delayed_work(system_wq, &dbc->event_work,
+	return mod_delayed_work(system_percpu_wq, &dbc->event_work,
 				msecs_to_jiffies(dbc->poll_interval));
 }
 
@@ -1023,7 +1023,7 @@ static void xhci_dbc_handle_events(struct work_struct *work)
 		return;
 	}
 
-	mod_delayed_work(system_wq, &dbc->event_work,
+	mod_delayed_work(system_percpu_wq, &dbc->event_work,
 			 msecs_to_jiffies(poll_interval));
 }
 
@@ -1274,7 +1274,7 @@ static ssize_t dbc_poll_interval_ms_store(struct device *dev,
 
 	dbc->poll_interval = value;
 
-	mod_delayed_work(system_wq, &dbc->event_work, 0);
+	mod_delayed_work(system_percpu_wq, &dbc->event_work, 0);
 
 	return size;
 }
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 8e209aa33ea7..3d05734c8377 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -434,7 +434,7 @@ void xhci_ring_cmd_db(struct xhci_hcd *xhci)
 
 static bool xhci_mod_cmd_timer(struct xhci_hcd *xhci)
 {
-	return mod_delayed_work(system_wq, &xhci->cmd_timer,
+	return mod_delayed_work(system_percpu_wq, &xhci->cmd_timer,
 			msecs_to_jiffies(xhci->current_cmd->timeout_ms));
 }
 
-- 
2.51.1

On 11/6/25 16:54, Marco Crivellari wrote:
> Currently if a user enqueues 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 consistency cannot be addressed without refactoring the API.
> 
> This continues the effort to refactor workqueue APIs, which began with
> the introduction of new workqueues and a new alloc_workqueue flag in:
> 
> commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
> commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")
> 
> Switch to using system_percpu_wq because system_wq is going away as part of
> a workqueue restructuring.
> 
> Suggested-by: Tejun Heo <tj@kernel.org>
> Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
> ---

Thanks, added to queue
-Mathias

On Wed, Nov 12, 2025 at 1:00 PM Mathias Nyman
<mathias.nyman@linux.intel.com> wrote:
> [...]
> Thanks, added to queue
> -Mathias
>

Many thanks!


-- 

Marco Crivellari

L3 Support Engineer, Technology & Product