mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users
@ 2025-09-05  9:06 Marco Crivellari
  2025-09-05  9:06 ` [PATCH 1/2] power: supply: replace use of system_wq with system_percpu_wq Marco Crivellari
                   ` (2 more replies)
  0 siblings, 3 replies; 5+ messages in thread
From: Marco Crivellari @ 2025-09-05  9:06 UTC (permalink / raw)
  To: linux-kernel, linux-pm
  Cc: Tejun Heo, Lai Jiangshan, Frederic Weisbecker,
	Sebastian Andrzej Siewior, Marco Crivellari, Michal Hocko,
	Sebastian Reichel

Hi!

Below is a summary of a discussion about the Workqueue API and cpu isolation
considerations. Details and more information are available here:

        "workqueue: Always use wq_select_unbound_cpu() for WORK_CPU_UNBOUND."
        https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/

=== Current situation: problems ===

Let's consider a nohz_full system with isolated CPUs: wq_unbound_cpumask is
set to the housekeeping CPUs, for !WQ_UNBOUND the local CPU is selected.

This leads to different scenarios if a work item is scheduled on an isolated
CPU where "delay" value is 0 or greater then 0:
        schedule_delayed_work(, 0);

This will be handled by __queue_work() that will queue the work item on the
current local (isolated) CPU, while:

        schedule_delayed_work(, 1);

Will move the timer on an housekeeping CPU, and schedule the work there.

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.

=== Plan and future plans ===

This patchset is the first stone on a refactoring needed in order to
address the points aforementioned; it will have a positive impact also
on the cpu isolation, in the long term, moving away percpu workqueue in
favor to an unbound model.

These are the main steps:
1)  API refactoring (that this patch is introducing)
    -   Make more clear and uniform the system wq names, both per-cpu and
        unbound. This to avoid any possible confusion on what should be
        used.

    -   Introduction of WQ_PERCPU: this flag is the complement of WQ_UNBOUND,
        introduced in this patchset and used on all the callers that are not
        currently using WQ_UNBOUND.

        WQ_UNBOUND will be removed in a future release cycle.

        Most users don't need to be per-cpu, because they don't have
        locality requirements, because of that, a next future step will be
        make "unbound" the default behavior.

2)  Check who really needs to be per-cpu
    -   Remove the WQ_PERCPU flag when is not strictly required.

3)  Add a new API (prefer local cpu)
    -   There are users that don't require a local execution, like mentioned
        above; despite that, local execution yeld to performance gain.

        This new API will prefer the local execution, without requiring it.

=== Introduced Changes by this series ===

1) [P 1] Replace use of system_wq and system_unbound_wq

        system_wq is a per-CPU workqueue, but his name is not clear.

        Because of that, system_wq has been renamed in system_percpu_wq.

2) [P 2] add WQ_PERCPU to remaining alloc_workqueue() users

        Every alloc_workqueue() caller should use one among WQ_PERCPU or
        WQ_UNBOUND. This is actually enforced warning if both or none of them
        are present at the same time.

        WQ_UNBOUND will be removed in a next release cycle.

=== For Maintainers ===

There are prerequisites for this series, already merged in the master branch.
The commits are:

128ea9f6ccfb6960293ae4212f4f97165e42222d ("workqueue: Add system_percpu_wq and
system_dfl_wq")

930c2ea566aff59e962c50b2421d5fcc3b98b8be ("workqueue: Add new WQ_PERCPU flag")


Thanks!

Marco Crivellari (2):
  power: supply: replace use of system_wq with system_percpu_wq
  power: supply: WQ_PERCPU added to alloc_workqueue users

 drivers/power/supply/ab8500_btemp.c       | 3 ++-
 drivers/power/supply/bq2415x_charger.c    | 2 +-
 drivers/power/supply/bq24190_charger.c    | 2 +-
 drivers/power/supply/bq27xxx_battery.c    | 6 +++---
 drivers/power/supply/ipaq_micro_battery.c | 3 ++-
 drivers/power/supply/rk817_charger.c      | 6 +++---
 drivers/power/supply/ucs1002_power.c      | 2 +-
 drivers/power/supply/ug3105_battery.c     | 6 +++---
 8 files changed, 16 insertions(+), 14 deletions(-)

-- 
2.51.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH 1/2] power: supply: replace use of system_wq with system_percpu_wq
  2025-09-05  9:06 [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Marco Crivellari
@ 2025-09-05  9:06 ` Marco Crivellari
  2025-09-05  9:06 ` [PATCH 2/2] power: supply: WQ_PERCPU added to alloc_workqueue users Marco Crivellari
  2025-09-09  0:19 ` [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Sebastian Reichel
  2 siblings, 0 replies; 5+ messages in thread
From: Marco Crivellari @ 2025-09-05  9:06 UTC (permalink / raw)
  To: linux-kernel, linux-pm
  Cc: Tejun Heo, Lai Jiangshan, Frederic Weisbecker,
	Sebastian Andrzej Siewior, Marco Crivellari, Michal Hocko,
	Sebastian Reichel

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_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.

queue_work() / queue_delayed_work() / mod_delayed_work() will now use the
new unbound wq: whether the user still use the old wq a warn will be
printed along with a wq redirect to the new one.

The old system_unbound_wq will be kept for a few release cycles.

Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/power/supply/bq2415x_charger.c | 2 +-
 drivers/power/supply/bq24190_charger.c | 2 +-
 drivers/power/supply/bq27xxx_battery.c | 6 +++---
 drivers/power/supply/rk817_charger.c   | 6 +++---
 drivers/power/supply/ucs1002_power.c   | 2 +-
 drivers/power/supply/ug3105_battery.c  | 6 +++---
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/drivers/power/supply/bq2415x_charger.c b/drivers/power/supply/bq2415x_charger.c
index 9e3b9181ee76..03837c831643 100644
--- a/drivers/power/supply/bq2415x_charger.c
+++ b/drivers/power/supply/bq2415x_charger.c
@@ -842,7 +842,7 @@ static int bq2415x_notifier_call(struct notifier_block *nb,
 	if (bq->automode < 1)
 		return NOTIFY_OK;
 
-	mod_delayed_work(system_wq, &bq->work, 0);
+	mod_delayed_work(system_percpu_wq, &bq->work, 0);
 
 	return NOTIFY_OK;
 }
diff --git a/drivers/power/supply/bq24190_charger.c b/drivers/power/supply/bq24190_charger.c
index f0d97ab45bd8..a19fca6d0a29 100644
--- a/drivers/power/supply/bq24190_charger.c
+++ b/drivers/power/supply/bq24190_charger.c
@@ -1474,7 +1474,7 @@ static void bq24190_charger_external_power_changed(struct power_supply *psy)
 	 * too low default 500mA iinlim. Delay setting the input-current-limit
 	 * for 300ms to avoid this.
 	 */
-	queue_delayed_work(system_wq, &bdi->input_current_limit_work,
+	queue_delayed_work(system_percpu_wq, &bdi->input_current_limit_work,
 			   msecs_to_jiffies(300));
 }
 
diff --git a/drivers/power/supply/bq27xxx_battery.c b/drivers/power/supply/bq27xxx_battery.c
index 2f31d750a4c1..d670ccf9661b 100644
--- a/drivers/power/supply/bq27xxx_battery.c
+++ b/drivers/power/supply/bq27xxx_battery.c
@@ -1127,7 +1127,7 @@ static int poll_interval_param_set(const char *val, const struct kernel_param *k
 
 	mutex_lock(&bq27xxx_list_lock);
 	list_for_each_entry(di, &bq27xxx_battery_devices, list)
-		mod_delayed_work(system_wq, &di->work, 0);
+		mod_delayed_work(system_percpu_wq, &di->work, 0);
 	mutex_unlock(&bq27xxx_list_lock);
 
 	return ret;
@@ -1945,7 +1945,7 @@ static void bq27xxx_battery_update_unlocked(struct bq27xxx_device_info *di)
 	di->last_update = jiffies;
 
 	if (!di->removed && poll_interval > 0)
-		mod_delayed_work(system_wq, &di->work, poll_interval * HZ);
+		mod_delayed_work(system_percpu_wq, &di->work, poll_interval * HZ);
 }
 
 void bq27xxx_battery_update(struct bq27xxx_device_info *di)
@@ -2221,7 +2221,7 @@ static void bq27xxx_external_power_changed(struct power_supply *psy)
 	struct bq27xxx_device_info *di = power_supply_get_drvdata(psy);
 
 	/* After charger plug in/out wait 0.5s for things to stabilize */
-	mod_delayed_work(system_wq, &di->work, HZ / 2);
+	mod_delayed_work(system_percpu_wq, &di->work, HZ / 2);
 }
 
 static void bq27xxx_battery_mutex_destroy(void *data)
diff --git a/drivers/power/supply/rk817_charger.c b/drivers/power/supply/rk817_charger.c
index 945c7720c4ae..032b191ddbf5 100644
--- a/drivers/power/supply/rk817_charger.c
+++ b/drivers/power/supply/rk817_charger.c
@@ -1046,7 +1046,7 @@ static void rk817_charging_monitor(struct work_struct *work)
 	rk817_read_props(charger);
 
 	/* Run every 8 seconds like the BSP driver did. */
-	queue_delayed_work(system_wq, &charger->work, msecs_to_jiffies(8000));
+	queue_delayed_work(system_percpu_wq, &charger->work, msecs_to_jiffies(8000));
 }
 
 static void rk817_cleanup_node(void *data)
@@ -1206,7 +1206,7 @@ static int rk817_charger_probe(struct platform_device *pdev)
 		return ret;
 
 	/* Force the first update immediately. */
-	mod_delayed_work(system_wq, &charger->work, 0);
+	mod_delayed_work(system_percpu_wq, &charger->work, 0);
 
 	return 0;
 }
@@ -1226,7 +1226,7 @@ static int __maybe_unused rk817_resume(struct device *dev)
 	struct rk817_charger *charger = dev_get_drvdata(dev);
 
 	/* force an immediate update */
-	mod_delayed_work(system_wq, &charger->work, 0);
+	mod_delayed_work(system_percpu_wq, &charger->work, 0);
 
 	return 0;
 }
diff --git a/drivers/power/supply/ucs1002_power.c b/drivers/power/supply/ucs1002_power.c
index d32a7633f9e7..fe94435340de 100644
--- a/drivers/power/supply/ucs1002_power.c
+++ b/drivers/power/supply/ucs1002_power.c
@@ -493,7 +493,7 @@ static irqreturn_t ucs1002_alert_irq(int irq, void *data)
 {
 	struct ucs1002_info *info = data;
 
-	mod_delayed_work(system_wq, &info->health_poll, 0);
+	mod_delayed_work(system_percpu_wq, &info->health_poll, 0);
 
 	return IRQ_HANDLED;
 }
diff --git a/drivers/power/supply/ug3105_battery.c b/drivers/power/supply/ug3105_battery.c
index 38e23bdd4603..15b62952f953 100644
--- a/drivers/power/supply/ug3105_battery.c
+++ b/drivers/power/supply/ug3105_battery.c
@@ -276,7 +276,7 @@ static void ug3105_work(struct work_struct *work)
 out:
 	mutex_unlock(&chip->lock);
 
-	queue_delayed_work(system_wq, &chip->work,
+	queue_delayed_work(system_percpu_wq, &chip->work,
 			   (chip->poll_count <= UG3105_INIT_POLL_COUNT) ?
 					UG3105_INIT_POLL_TIME : UG3105_POLL_TIME);
 
@@ -352,7 +352,7 @@ static void ug3105_external_power_changed(struct power_supply *psy)
 	struct ug3105_chip *chip = power_supply_get_drvdata(psy);
 
 	dev_dbg(&chip->client->dev, "external power changed\n");
-	mod_delayed_work(system_wq, &chip->work, UG3105_SETTLE_TIME);
+	mod_delayed_work(system_percpu_wq, &chip->work, UG3105_SETTLE_TIME);
 }
 
 static const struct power_supply_desc ug3105_psy_desc = {
@@ -373,7 +373,7 @@ static void ug3105_init(struct ug3105_chip *chip)
 				  UG3105_MODE_RUN);
 	i2c_smbus_write_byte_data(chip->client, UG3105_REG_CTRL1,
 				  UG3105_CTRL1_RESET_COULOMB_CNT);
-	queue_delayed_work(system_wq, &chip->work, 0);
+	queue_delayed_work(system_percpu_wq, &chip->work, 0);
 	flush_delayed_work(&chip->work);
 }
 
-- 
2.51.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* [PATCH 2/2] power: supply: WQ_PERCPU added to alloc_workqueue users
  2025-09-05  9:06 [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Marco Crivellari
  2025-09-05  9:06 ` [PATCH 1/2] power: supply: replace use of system_wq with system_percpu_wq Marco Crivellari
@ 2025-09-05  9:06 ` Marco Crivellari
  2025-09-09  0:19 ` [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Sebastian Reichel
  2 siblings, 0 replies; 5+ messages in thread
From: Marco Crivellari @ 2025-09-05  9:06 UTC (permalink / raw)
  To: linux-kernel, linux-pm
  Cc: Tejun Heo, Lai Jiangshan, Frederic Weisbecker,
	Sebastian Andrzej Siewior, Marco Crivellari, Michal Hocko,
	Sebastian Reichel

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 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.

Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/power/supply/ab8500_btemp.c       | 3 ++-
 drivers/power/supply/ipaq_micro_battery.c | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/power/supply/ab8500_btemp.c b/drivers/power/supply/ab8500_btemp.c
index b00c84fbc33c..e5202a7b6209 100644
--- a/drivers/power/supply/ab8500_btemp.c
+++ b/drivers/power/supply/ab8500_btemp.c
@@ -667,7 +667,8 @@ static int ab8500_btemp_bind(struct device *dev, struct device *master,
 
 	/* Create a work queue for the btemp */
 	di->btemp_wq =
-		alloc_workqueue("ab8500_btemp_wq", WQ_MEM_RECLAIM, 0);
+		alloc_workqueue("ab8500_btemp_wq", WQ_MEM_RECLAIM | WQ_PERCPU,
+				0);
 	if (di->btemp_wq == NULL) {
 		dev_err(dev, "failed to create work queue\n");
 		return -ENOMEM;
diff --git a/drivers/power/supply/ipaq_micro_battery.c b/drivers/power/supply/ipaq_micro_battery.c
index 7e0568a5353f..ff8573a5ca6d 100644
--- a/drivers/power/supply/ipaq_micro_battery.c
+++ b/drivers/power/supply/ipaq_micro_battery.c
@@ -232,7 +232,8 @@ static int micro_batt_probe(struct platform_device *pdev)
 		return -ENOMEM;
 
 	mb->micro = dev_get_drvdata(pdev->dev.parent);
-	mb->wq = alloc_workqueue("ipaq-battery-wq", WQ_MEM_RECLAIM, 0);
+	mb->wq = alloc_workqueue("ipaq-battery-wq",
+				 WQ_MEM_RECLAIM | WQ_PERCPU, 0);
 	if (!mb->wq)
 		return -ENOMEM;
 
-- 
2.51.0


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users
  2025-09-05  9:06 [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Marco Crivellari
  2025-09-05  9:06 ` [PATCH 1/2] power: supply: replace use of system_wq with system_percpu_wq Marco Crivellari
  2025-09-05  9:06 ` [PATCH 2/2] power: supply: WQ_PERCPU added to alloc_workqueue users Marco Crivellari
@ 2025-09-09  0:19 ` Sebastian Reichel
  2025-09-09  7:28   ` Marco Crivellari
  2 siblings, 1 reply; 5+ messages in thread
From: Sebastian Reichel @ 2025-09-09  0:19 UTC (permalink / raw)
  To: linux-kernel, linux-pm, Marco Crivellari
  Cc: Tejun Heo, Lai Jiangshan, Frederic Weisbecker,
	Sebastian Andrzej Siewior, Michal Hocko, Sebastian Reichel


On Fri, 05 Sep 2025 11:06:39 +0200, Marco Crivellari wrote:
> Below is a summary of a discussion about the Workqueue API and cpu isolation
> considerations. Details and more information are available here:
> 
>         "workqueue: Always use wq_select_unbound_cpu() for WORK_CPU_UNBOUND."
>         https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
> 
> === Current situation: problems ===
> 
> [...]

Applied, thanks!

[1/2] power: supply: replace use of system_wq with system_percpu_wq
      commit: c4a7748b551e5a06fe9a3862001192b1b5cfe195
[2/2] power: supply: WQ_PERCPU added to alloc_workqueue users
      commit: cc2ec444e461b6ca2bc73cd7cbd06aaf15bdfa1a

Best regards,
-- 
Sebastian Reichel <sebastian.reichel@collabora.com>


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users
  2025-09-09  0:19 ` [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Sebastian Reichel
@ 2025-09-09  7:28   ` Marco Crivellari
  0 siblings, 0 replies; 5+ messages in thread
From: Marco Crivellari @ 2025-09-09  7:28 UTC (permalink / raw)
  To: Sebastian Reichel
  Cc: linux-kernel, linux-pm, Tejun Heo, Lai Jiangshan,
	Frederic Weisbecker, Sebastian Andrzej Siewior, Michal Hocko,
	Sebastian Reichel

On Tue, Sep 9, 2025 at 2:19 AM Sebastian Reichel
<sebastian.reichel@collabora.com> wrote:
> Applied, thanks!
>
> [1/2] power: supply: replace use of system_wq with system_percpu_wq
>       commit: c4a7748b551e5a06fe9a3862001192b1b5cfe195
> [2/2] power: supply: WQ_PERCPU added to alloc_workqueue users
>       commit: cc2ec444e461b6ca2bc73cd7cbd06aaf15bdfa1a

Thank you!
-- 

Marco Crivellari

L3 Support Engineer, Technology & Product

marco.crivellari@suse.com

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2025-09-09  7:29 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-09-05  9:06 [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Marco Crivellari
2025-09-05  9:06 ` [PATCH 1/2] power: supply: replace use of system_wq with system_percpu_wq Marco Crivellari
2025-09-05  9:06 ` [PATCH 2/2] power: supply: WQ_PERCPU added to alloc_workqueue users Marco Crivellari
2025-09-09  0:19 ` [PATCH 0/2] power: replace wq users and add WQ_PERCPU to alloc_workqueue() users Sebastian Reichel
2025-09-09  7:28   ` Marco Crivellari

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®