From: Bradley Morgan <bradley.morgan@linux.dev>
To: Daniel Lezcano <daniel.lezcano@oss.qualcomm.com>,
rafael@kernel.org, tglx@linutronix.de
Cc: Andrew Morton <akpm@linux-foundation.org>,
Peter Ziljstra <peterz@infradead.org>,
Heiko Stuebner <heiko@sntech.de>,
linux-arm-kernel@lists.infradead.org,
linux-rockchip@lists.infradead.org,
Thomas Gleixner <tglx@kernel.org>,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 1/2] smpboot: Allow restricting per-CPU threads to a CPU mask
Date: Wed, 18 Nov 2026 16:00:43 +0000 [thread overview]
Message-ID: <0453960B-FF17-4E49-8672-FC0FDA3EF5DE@linux.dev> (raw)
In-Reply-To: <7a1e55d24c39d70caa157e6e2c823d9be5db07e7.1794992905.git.daniel.lezcano@oss.qualcomm.com>
On 18 November 2026 10:33:48 GMT, Daniel Lezcano
<daniel.lezcano@oss.qualcomm.com> wrote:
>smpboot_register_percpu_thread() creates a thread for every online CPU
>and registers the descriptor for subsequent CPU hotplug operations.
>There is currently no way to restrict these threads to a subset of
>CPUs, even when the user only needs work performed on that subset.
>
>For example, an idle injection device targets a specific CPU mask.
>Creating threads outside that mask consumes resources unnecessarily.
>Supporting multiple such devices also requires each descriptor to
>manage only the threads associated with its own CPUs.
>
>Add an optional CPU mask to struct smp_hotplug_thread and apply it
>throughout the thread lifecycle. Registration creates threads only for
>online CPUs in the mask. CPU bring-up creates threads for selected
>CPUs that were offline at registration, while parking and unparking
>are restricted to the same mask.
>
>Apply the restriction to thread destruction as well, both during
>unregistration and registration failure cleanup. This is necessary
>when descriptors with disjoint masks share the same per-CPU task
>storage: tearing down one descriptor must not stop or clear threads
>owned by another descriptor.
>
>Reject registrations that share task storage with an existing
>descriptor and have overlapping CPU masks. Check for overlaps under
>smpboot_threads_lock before creating any threads, warn once and return
>-EBUSY on a conflict. Checking the masks also covers offline CPUs,
>whose task pointers may not yet be populated. This prevents descriptors
>from sharing a thread and interfering with each other's teardown.
>
>Default a NULL mask to cpu_possible_mask at registration, preserving
>the existing all-CPU behavior and allowing the lifecycle helpers to
>use the mask unconditionally. Use scoped guards for the registration
>locks so all error returns release them in the correct order.
>
>The caller must keep the mask valid and unchanged until the descriptor
>is unregistered. smpboot serializes the overlap check and descriptor
>registration against other registration and unregistration operations.
Cheers,
Reviewed-by: Bradley Morgan <bradley.morgan@linux.dev>
>
>Signed-off-by: Daniel Lezcano <daniel.lezcano@oss.qualcomm.com>
>---
> include/linux/smpboot.h | 6 ++++++
> kernel/smpboot.c | 45 ++++++++++++++++++++++++++++++++---------
> 2 files changed, 41 insertions(+), 10 deletions(-)
>
>diff --git a/include/linux/smpboot.h b/include/linux/smpboot.h
>index 9d1bc65d226c..57b945d51b19 100644
>--- a/include/linux/smpboot.h
>+++ b/include/linux/smpboot.h
>@@ -5,6 +5,7 @@
> #include <linux/types.h>
>
> struct task_struct;
>+struct cpumask;
> /* Cookie handed to the thread_fn*/
> struct smpboot_thread_data;
>
>@@ -27,6 +28,10 @@ struct smpboot_thread_data;
> * unparked (cpu online)
> * @selfparking: Thread is not parked by the park function.
> * @thread_comm: The base name of the thread
>+ * @cpumask: Optional mask of CPUs on which to create threads. Must remain
>+ * valid and unchanged until the descriptor is unregistered.
>+ * Registration defaults NULL to cpu_possible_mask. Descriptors
>+ * sharing @store must have disjoint masks.
> */
> struct smp_hotplug_thread {
> struct task_struct * __percpu *store;
>@@ -40,6 +45,7 @@ struct smp_hotplug_thread {
> void (*unpark)(unsigned int cpu);
> bool selfparking;
> const char *thread_comm;
>+ const struct cpumask *cpumask;
> };
>
> int smpboot_register_percpu_thread(struct smp_hotplug_thread
> *plug_thread);
>diff --git a/kernel/smpboot.c b/kernel/smpboot.c
>index 4503b60ce9bd..f997fe85f088 100644
>--- a/kernel/smpboot.c
>+++ b/kernel/smpboot.c
>@@ -168,6 +168,9 @@ __smpboot_create_thread(struct smp_hotplug_thread *ht, unsigned int cpu)
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
> struct smpboot_thread_data *td;
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ return 0;
>+
> if (tsk)
> return 0;
>
>@@ -225,6 +228,9 @@ static void smpboot_unpark_thread(struct smp_hotplug_thread *ht, unsigned int cp
> {
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ return;
>+
> if (!ht->selfparking)
> kthread_unpark(tsk);
> }
>@@ -244,6 +250,9 @@ static void smpboot_park_thread(struct smp_hotplug_thread *ht, unsigned int cpu)
> {
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ return;
>+
> if (tsk && !ht->selfparking)
> kthread_park(tsk);
> }
>@@ -267,6 +276,9 @@ static void smpboot_destroy_threads(struct smp_hotplug_thread *ht)
> for_each_possible_cpu(cpu) {
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ continue;
>+
> if (tsk) {
> kthread_stop_put(tsk);
> *per_cpu_ptr(ht->store, cpu) = NULL;
>@@ -279,28 +291,41 @@ static void smpboot_destroy_threads(struct smp_hotplug_thread *ht)
> * to hotplug
> * @plug_thread: Hotplug thread descriptor
> *
>- * Creates and starts the threads on all online cpus.
>+ * Creates and starts the threads on online CPUs selected by the descriptor's
>+ * cpumask, or all online CPUs if no mask is specified.
>+ *
>+ * Return: 0 on success, -EBUSY if a registered descriptor shares the same task
>+ * storage and has an overlapping CPU mask, or a thread creation error.
> */
> int smpboot_register_percpu_thread(struct smp_hotplug_thread
> *plug_thread)
> {
>+ struct smp_hotplug_thread *ht;
> unsigned int cpu;
>- int ret = 0;
>+ int ret;
>+
>+ if (!plug_thread->cpumask)
>+ plug_thread->cpumask = cpu_possible_mask;
>+
>+ guard(cpus_read_lock)();
>+ guard(mutex)(&smpboot_threads_lock);
>+
>+ list_for_each_entry(ht, &hotplug_threads, list) {
>+ if (ht->store != plug_thread->store)
>+ continue;
>+ if (WARN_ON_ONCE(cpumask_intersects(ht->cpumask, plug_thread->cpumask)))
>+ return -EBUSY;
>+ }
>
>- cpus_read_lock();
>- mutex_lock(&smpboot_threads_lock);
> for_each_online_cpu(cpu) {
> ret = __smpboot_create_thread(plug_thread, cpu);
> if (ret) {
> smpboot_destroy_threads(plug_thread);
>- goto out;
>+ return ret;
> }
> smpboot_unpark_thread(plug_thread, cpu);
> }
> list_add(&plug_thread->list, &hotplug_threads);
>-out:
>- mutex_unlock(&smpboot_threads_lock);
>- cpus_read_unlock();
>- return ret;
>+ return 0;
> }
> EXPORT_SYMBOL_GPL(smpboot_register_percpu_thread);
>
>@@ -308,7 +333,7 @@ EXPORT_SYMBOL_GPL(smpboot_register_percpu_thread);
> * smpboot_unregister_percpu_thread - Unregister a per_cpu thread related to hotplug
> * @plug_thread: Hotplug thread descriptor
> *
>- * Stops all threads on all possible cpus.
>+ * Stops all threads associated with the descriptor, including offline CPUs.
> */
> void smpboot_unregister_percpu_thread(struct smp_hotplug_thread
> *plug_thread)
> {
>
--- Thanks!
WARNING: multiple messages have this Message-ID (diff)
From: Bradley Morgan <bradley.morgan@linux.dev>
To: Daniel Lezcano <daniel.lezcano@oss.qualcomm.com>,
rafael@kernel.org, tglx@linutronix.de
Cc: Andrew Morton <akpm@linux-foundation.org>,
Peter Ziljstra <peterz@infradead.org>,
Heiko Stuebner <heiko@sntech.de>,
linux-arm-kernel@lists.infradead.org,
linux-rockchip@lists.infradead.org,
Thomas Gleixner <tglx@kernel.org>,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 1/2] smpboot: Allow restricting per-CPU threads to a CPU mask
Date: Wed, 18 Nov 2026 16:00:43 +0000 [thread overview]
Message-ID: <0453960B-FF17-4E49-8672-FC0FDA3EF5DE@linux.dev> (raw)
In-Reply-To: <7a1e55d24c39d70caa157e6e2c823d9be5db07e7.1794992905.git.daniel.lezcano@oss.qualcomm.com>
On 18 November 2026 10:33:48 GMT, Daniel Lezcano
<daniel.lezcano@oss.qualcomm.com> wrote:
>smpboot_register_percpu_thread() creates a thread for every online CPU
>and registers the descriptor for subsequent CPU hotplug operations.
>There is currently no way to restrict these threads to a subset of
>CPUs, even when the user only needs work performed on that subset.
>
>For example, an idle injection device targets a specific CPU mask.
>Creating threads outside that mask consumes resources unnecessarily.
>Supporting multiple such devices also requires each descriptor to
>manage only the threads associated with its own CPUs.
>
>Add an optional CPU mask to struct smp_hotplug_thread and apply it
>throughout the thread lifecycle. Registration creates threads only for
>online CPUs in the mask. CPU bring-up creates threads for selected
>CPUs that were offline at registration, while parking and unparking
>are restricted to the same mask.
>
>Apply the restriction to thread destruction as well, both during
>unregistration and registration failure cleanup. This is necessary
>when descriptors with disjoint masks share the same per-CPU task
>storage: tearing down one descriptor must not stop or clear threads
>owned by another descriptor.
>
>Reject registrations that share task storage with an existing
>descriptor and have overlapping CPU masks. Check for overlaps under
>smpboot_threads_lock before creating any threads, warn once and return
>-EBUSY on a conflict. Checking the masks also covers offline CPUs,
>whose task pointers may not yet be populated. This prevents descriptors
>from sharing a thread and interfering with each other's teardown.
>
>Default a NULL mask to cpu_possible_mask at registration, preserving
>the existing all-CPU behavior and allowing the lifecycle helpers to
>use the mask unconditionally. Use scoped guards for the registration
>locks so all error returns release them in the correct order.
>
>The caller must keep the mask valid and unchanged until the descriptor
>is unregistered. smpboot serializes the overlap check and descriptor
>registration against other registration and unregistration operations.
Cheers,
Reviewed-by: Bradley Morgan <bradley.morgan@linux.dev>
>
>Signed-off-by: Daniel Lezcano <daniel.lezcano@oss.qualcomm.com>
>---
> include/linux/smpboot.h | 6 ++++++
> kernel/smpboot.c | 45 ++++++++++++++++++++++++++++++++---------
> 2 files changed, 41 insertions(+), 10 deletions(-)
>
>diff --git a/include/linux/smpboot.h b/include/linux/smpboot.h
>index 9d1bc65d226c..57b945d51b19 100644
>--- a/include/linux/smpboot.h
>+++ b/include/linux/smpboot.h
>@@ -5,6 +5,7 @@
> #include <linux/types.h>
>
> struct task_struct;
>+struct cpumask;
> /* Cookie handed to the thread_fn*/
> struct smpboot_thread_data;
>
>@@ -27,6 +28,10 @@ struct smpboot_thread_data;
> * unparked (cpu online)
> * @selfparking: Thread is not parked by the park function.
> * @thread_comm: The base name of the thread
>+ * @cpumask: Optional mask of CPUs on which to create threads. Must remain
>+ * valid and unchanged until the descriptor is unregistered.
>+ * Registration defaults NULL to cpu_possible_mask. Descriptors
>+ * sharing @store must have disjoint masks.
> */
> struct smp_hotplug_thread {
> struct task_struct * __percpu *store;
>@@ -40,6 +45,7 @@ struct smp_hotplug_thread {
> void (*unpark)(unsigned int cpu);
> bool selfparking;
> const char *thread_comm;
>+ const struct cpumask *cpumask;
> };
>
> int smpboot_register_percpu_thread(struct smp_hotplug_thread
> *plug_thread);
>diff --git a/kernel/smpboot.c b/kernel/smpboot.c
>index 4503b60ce9bd..f997fe85f088 100644
>--- a/kernel/smpboot.c
>+++ b/kernel/smpboot.c
>@@ -168,6 +168,9 @@ __smpboot_create_thread(struct smp_hotplug_thread *ht, unsigned int cpu)
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
> struct smpboot_thread_data *td;
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ return 0;
>+
> if (tsk)
> return 0;
>
>@@ -225,6 +228,9 @@ static void smpboot_unpark_thread(struct smp_hotplug_thread *ht, unsigned int cp
> {
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ return;
>+
> if (!ht->selfparking)
> kthread_unpark(tsk);
> }
>@@ -244,6 +250,9 @@ static void smpboot_park_thread(struct smp_hotplug_thread *ht, unsigned int cpu)
> {
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ return;
>+
> if (tsk && !ht->selfparking)
> kthread_park(tsk);
> }
>@@ -267,6 +276,9 @@ static void smpboot_destroy_threads(struct smp_hotplug_thread *ht)
> for_each_possible_cpu(cpu) {
> struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
>
>+ if (!cpumask_test_cpu(cpu, ht->cpumask))
>+ continue;
>+
> if (tsk) {
> kthread_stop_put(tsk);
> *per_cpu_ptr(ht->store, cpu) = NULL;
>@@ -279,28 +291,41 @@ static void smpboot_destroy_threads(struct smp_hotplug_thread *ht)
> * to hotplug
> * @plug_thread: Hotplug thread descriptor
> *
>- * Creates and starts the threads on all online cpus.
>+ * Creates and starts the threads on online CPUs selected by the descriptor's
>+ * cpumask, or all online CPUs if no mask is specified.
>+ *
>+ * Return: 0 on success, -EBUSY if a registered descriptor shares the same task
>+ * storage and has an overlapping CPU mask, or a thread creation error.
> */
> int smpboot_register_percpu_thread(struct smp_hotplug_thread
> *plug_thread)
> {
>+ struct smp_hotplug_thread *ht;
> unsigned int cpu;
>- int ret = 0;
>+ int ret;
>+
>+ if (!plug_thread->cpumask)
>+ plug_thread->cpumask = cpu_possible_mask;
>+
>+ guard(cpus_read_lock)();
>+ guard(mutex)(&smpboot_threads_lock);
>+
>+ list_for_each_entry(ht, &hotplug_threads, list) {
>+ if (ht->store != plug_thread->store)
>+ continue;
>+ if (WARN_ON_ONCE(cpumask_intersects(ht->cpumask, plug_thread->cpumask)))
>+ return -EBUSY;
>+ }
>
>- cpus_read_lock();
>- mutex_lock(&smpboot_threads_lock);
> for_each_online_cpu(cpu) {
> ret = __smpboot_create_thread(plug_thread, cpu);
> if (ret) {
> smpboot_destroy_threads(plug_thread);
>- goto out;
>+ return ret;
> }
> smpboot_unpark_thread(plug_thread, cpu);
> }
> list_add(&plug_thread->list, &hotplug_threads);
>-out:
>- mutex_unlock(&smpboot_threads_lock);
>- cpus_read_unlock();
>- return ret;
>+ return 0;
> }
> EXPORT_SYMBOL_GPL(smpboot_register_percpu_thread);
>
>@@ -308,7 +333,7 @@ EXPORT_SYMBOL_GPL(smpboot_register_percpu_thread);
> * smpboot_unregister_percpu_thread - Unregister a per_cpu thread related to hotplug
> * @plug_thread: Hotplug thread descriptor
> *
>- * Stops all threads on all possible cpus.
>+ * Stops all threads associated with the descriptor, including offline CPUs.
> */
> void smpboot_unregister_percpu_thread(struct smp_hotplug_thread
> *plug_thread)
> {
>
--- Thanks!
_______________________________________________
Linux-rockchip mailing list
Linux-rockchip@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-rockchip
next prev parent reply other threads:[~2026-11-18 16:00 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-18 10:33 [PATCH v2 0/2] powercap: idle_inject: Create threads on demand Daniel Lezcano
2026-11-18 16:15 ` [PATCH RESEND " Daniel Lezcano
2026-11-18 16:15 ` Daniel Lezcano
2026-11-18 10:33 ` [PATCH " Daniel Lezcano
2026-11-18 10:33 ` [PATCH v2 1/2] smpboot: Allow restricting per-CPU threads to a CPU mask Daniel Lezcano
2026-11-18 16:15 ` [PATCH RESEND " Daniel Lezcano
2026-11-18 16:15 ` Daniel Lezcano
2026-11-18 10:33 ` [PATCH " Daniel Lezcano
2026-11-18 16:00 ` Bradley Morgan [this message]
2026-11-18 16:00 ` Bradley Morgan
2026-11-18 10:33 ` [PATCH v2 2/2] powercap: idle_inject: Create per-CPU threads on registration Daniel Lezcano
2026-11-18 16:15 ` [PATCH RESEND " Daniel Lezcano
2026-11-18 16:15 ` Daniel Lezcano
2026-11-18 10:33 ` [PATCH " Daniel Lezcano
2026-11-18 15:57 ` [PATCH v2 0/2] powercap: idle_inject: Create threads on demand Rafael J. Wysocki (Intel)
2026-11-18 15:57 ` Rafael J. Wysocki (Intel)
2026-11-20 19:17 ` [PATCH RESEND " Rafael J. Wysocki (Intel)
2026-11-20 19:17 ` Rafael J. Wysocki (Intel)
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=0453960B-FF17-4E49-8672-FC0FDA3EF5DE@linux.dev \
--to=bradley.morgan@linux.dev \
--cc=akpm@linux-foundation.org \
--cc=daniel.lezcano@oss.qualcomm.com \
--cc=heiko@sntech.de \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rockchip@lists.infradead.org \
--cc=peterz@infradead.org \
--cc=rafael@kernel.org \
--cc=tglx@kernel.org \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.