[PATCH v2 0/2] efi: Expose the runtime-services workqueue via sysfs

[PATCH v2 0/2] efi: Expose the runtime-services workqueue via sysfs

[Sebastian Andrzej Siewior]

EFI runtime services are disabled on PREEMPT_RT by default which can be
overwritten on the boot command line. For native EFI, an invocation
requires to disable preemption an x86 while a call is made into EFI.
Even if it does not (as it is the case on arm64) the EFI function can
disable interrupts. This makes the time spent in EFI is not
deterministic.
While accessing the efi-rtc device can be avoided by using a native
driver, accessing the "variables" is important and there is no second
path.

The "runtime-wrappers" is wrapping access to the EFI callback via a
workqueue. On a SMP system one CPU could be declared as housekeeping/
not-realtime-capable and force all EFI invocation to be performed on
this CPU. This could be achieved by setting workqueue.unbound_cpus or
	/sys/devices/virtual/workqueue/cpumask

at runtime. This however will affect all workqueues and might not be
desired. With an explicit setting such as
	/sys/devices/virtual/workqueue/efi_runtime/cpumask

it looks like an official way to limit the CPUs involved here.

With this in place I was wondering if EFI_DISABLE_RUNTIME could be
lifted at runtime on SMP systems. But given the unbound_cpus option
and the auto-config based on NOHZ-full it might not be wise to add yet
another smart option here. Also it needs to be a subset of root cpumask
or it won't be effective.

v1…v2: https://lore.kernel.org/all/20260205115559.1625236-1-bigeasy@linutronix.de/
  - For BH and WQ_ORDERED worker make the max_active RO.

Sebastian Andrzej Siewior (2):
  workqueue: Allow to expose ordered workqueues via sysfs
  efi: Allow to expose the workqueue via sysfs

 drivers/firmware/efi/efi.c |  2 +-
 kernel/workqueue.c         | 28 ++++++++++++++++++++--------
 2 files changed, 21 insertions(+), 9 deletions(-)

-- 
2.51.0

[PATCH v2 1/2] workqueue: Allow to expose ordered workqueues via sysfs

[Sebastian Andrzej Siewior]

Ordered workqueues are not exposed via sysfs because the 'max_active'
attribute changes the number actives worker. More than one active worker
can break ordering guarantees.

This can be avoided by forbidding writes the file for ordered
workqueues. Exposing it via sysfs allows to alter other attributes such
as the cpumask on which CPU the worker can run.

The 'max_active' value shouldn't be changed for BH worker because the
core never spawns additional worker and the worker itself can not be
preempted. So this make no sense.

Allow to expose ordered workqueues via sysfs if requested and forbid
changing 'max_active' value for ordered and BH worker.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 kernel/workqueue.c | 28 ++++++++++++++++++++--------
 1 file changed, 20 insertions(+), 8 deletions(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index aeaec79bc09c4..2f95cb0d2f1b8 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -7176,7 +7176,26 @@ static struct attribute *wq_sysfs_attrs[] = {
 	&dev_attr_max_active.attr,
 	NULL,
 };
-ATTRIBUTE_GROUPS(wq_sysfs);
+
+static umode_t wq_sysfs_is_visible(struct kobject *kobj, struct attribute *a, int n)
+{
+	struct device *dev = kobj_to_dev(kobj);
+	struct workqueue_struct *wq = dev_to_wq(dev);
+
+	/*
+	 * Adjusting max_active breaks ordering guarantee. Changing it has no
+	 * effect on BH worker. Limit max_active to RO in such case.
+	 */
+	if (wq->flags & (WQ_BH | __WQ_ORDERED))
+		return 0444;
+	return a->mode;
+}
+
+static const struct attribute_group wq_sysfs_group = {
+	.is_visible = wq_sysfs_is_visible,
+	.attrs = wq_sysfs_attrs,
+};
+__ATTRIBUTE_GROUPS(wq_sysfs);
 
 static ssize_t wq_nice_show(struct device *dev, struct device_attribute *attr,
 			    char *buf)
@@ -7479,13 +7498,6 @@ int workqueue_sysfs_register(struct workqueue_struct *wq)
 	struct wq_device *wq_dev;
 	int ret;
 
-	/*
-	 * Adjusting max_active breaks ordering guarantee.  Disallow exposing
-	 * ordered workqueues.
-	 */
-	if (WARN_ON(wq->flags & __WQ_ORDERED))
-		return -EINVAL;
-
 	wq->wq_dev = wq_dev = kzalloc_obj(*wq_dev);
 	if (!wq_dev)
 		return -ENOMEM;
-- 
2.51.0

Re: [PATCH v2 1/2] workqueue: Allow to expose ordered workqueues via sysfs

[Tejun Heo]

On Fri, Feb 27, 2026 at 06:01:02PM +0100, Sebastian Andrzej Siewior wrote:
> Ordered workqueues are not exposed via sysfs because the 'max_active'
> attribute changes the number actives worker. More than one active worker
> can break ordering guarantees.
> 
> This can be avoided by forbidding writes the file for ordered
> workqueues. Exposing it via sysfs allows to alter other attributes such
> as the cpumask on which CPU the worker can run.
> 
> The 'max_active' value shouldn't be changed for BH worker because the
> core never spawns additional worker and the worker itself can not be
> preempted. So this make no sense.
> 
> Allow to expose ordered workqueues via sysfs if requested and forbid
> changing 'max_active' value for ordered and BH worker.
> 
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>

Acked-by: Tejun Heo <tj@kernel.org>

I can apply this to wq/for-7.1. Would that work?

Thanks.

-- 
tejun

Re: [PATCH v2 1/2] workqueue: Allow to expose ordered workqueues via sysfs

[Sebastian Andrzej Siewior]

On 2026-02-27 07:12:46 [-1000], Tejun Heo wrote:
> Acked-by: Tejun Heo <tj@kernel.org>
Thank.

> I can apply this to wq/for-7.1. Would that work?

I think you need to take both since 2/2 can't be applied independently.
Maybe Ard is going to give his blessing ;)

> Thanks.

Sebastian

Re: [PATCH v2 1/2] workqueue: Allow to expose ordered workqueues via sysfs

[Ard Biesheuvel]

On Fri, 27 Feb 2026, at 18:14, Sebastian Andrzej Siewior wrote:
> On 2026-02-27 07:12:46 [-1000], Tejun Heo wrote:
>> Acked-by: Tejun Heo <tj@kernel.org>
> Thank.
>
>> I can apply this to wq/for-7.1. Would that work?
>
> I think you need to take both since 2/2 can't be applied independently.
> Maybe Ard is going to give his blessing ;)
>

iIRC I already acked v1.

Re: [PATCH v2 1/2] workqueue: Allow to expose ordered workqueues via sysfs

[Tejun Heo]

On Fri, Feb 27, 2026 at 06:20:47PM +0100, Ard Biesheuvel wrote:
> 
> 
> On Fri, 27 Feb 2026, at 18:14, Sebastian Andrzej Siewior wrote:
> > On 2026-02-27 07:12:46 [-1000], Tejun Heo wrote:
> >> Acked-by: Tejun Heo <tj@kernel.org>
> > Thank.
> >
> >> I can apply this to wq/for-7.1. Would that work?
> >
> > I think you need to take both since 2/2 can't be applied independently.
> > Maybe Ard is going to give his blessing ;)
> >
> 
> iIRC I already acked v1.

Ard, do you want me to take both patches or wann pull from wq/for-7.1?

Thanks.

-- 
tejun

Re: [PATCH v2 1/2] workqueue: Allow to expose ordered workqueues via sysfs

[Ard Biesheuvel]

On Fri, 27 Feb 2026, at 18:27, Tejun Heo wrote:
> On Fri, Feb 27, 2026 at 06:20:47PM +0100, Ard Biesheuvel wrote:
>> 
>> 
>> On Fri, 27 Feb 2026, at 18:14, Sebastian Andrzej Siewior wrote:
>> > On 2026-02-27 07:12:46 [-1000], Tejun Heo wrote:
>> >> Acked-by: Tejun Heo <tj@kernel.org>
>> > Thank.
>> >
>> >> I can apply this to wq/for-7.1. Would that work?
>> >
>> > I think you need to take both since 2/2 can't be applied independently.
>> > Maybe Ard is going to give his blessing ;)
>> >
>> 
>> iIRC I already acked v1.
>
> Ard, do you want me to take both patches or wann pull from wq/for-7.1?
>

Please take both - thanks.

[PATCH v2 2/2] efi: Allow to expose the workqueue via sysfs

[Sebastian Andrzej Siewior]

Exposing the efi_rts_wq workqueue via sysfs provides an easy mechanism
to restrict EFI firmware invocation to certain CPU(s).
This can be used to restrict EFI invocations to specific CPUs while
allowing other workqueue to use the remaning CPUs.

Expose the workqueue via sysfs. Change the name to efi_runtime which is
what will be visible under sysfs.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 drivers/firmware/efi/efi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/firmware/efi/efi.c b/drivers/firmware/efi/efi.c
index b2fb92a4bbd11..3dab284a7754d 100644
--- a/drivers/firmware/efi/efi.c
+++ b/drivers/firmware/efi/efi.c
@@ -423,7 +423,7 @@ static int __init efisubsys_init(void)
 		 * ordered workqueue (which creates only one execution context)
 		 * should suffice for all our needs.
 		 */
-		efi_rts_wq = alloc_ordered_workqueue("efi_rts_wq", 0);
+		efi_rts_wq = alloc_ordered_workqueue("efi_runtime", WQ_SYSFS);
 		if (!efi_rts_wq) {
 			pr_err("Creating efi_rts_wq failed, EFI runtime services disabled.\n");
 			clear_bit(EFI_RUNTIME_SERVICES, &efi.flags);
-- 
2.51.0

Re: [PATCH v2 0/2] efi: Expose the runtime-services workqueue via sysfs

[Tejun Heo]

Applied to wq/for-7.1.

Thanks.

--
tejun