[PATCH] PM: runtime: call pm_runtime_dont_use_autosuspend() on reinit()

[PATCH] PM: runtime: call pm_runtime_dont_use_autosuspend() on reinit()

[Joshua Crofts]

Add a pm_runtime_dont_use_autosuspend() call to pm_runtime_reinit() to
ensure autosuspend is disabled on driver teardown.

Suggested-by: Frank Li <Frank.Li@nxp.com>
Signed-off-by: Joshua Crofts <joshua.crofts1@gmail.com>
---
If a driver doesn't use devm_* functions for automatic cleanup on
teardown and autosuspend is enabled during probe, it should also be
disabled on teardown in a _remove() function or equivalent. However,
due to bad documentation, driver authors usually omit the
pm_runtime_dont_use_autosuspend() call on teardown.

Reason for investigating: this week, several patches popped up in
the linux-iio mailing list, adding these missing calls to several
IIO drivers. For some reason I ended up writing a cocci script to
find other instances of this and found ~195 offending drivers.

After a discussion with Frank Li on the matter [1] when sending a similar
patch adding these missing calls to the clk subsystem, we arrived to the
idea of actually adding the dont_use_autosuspend() call to
pm_runtime_reinit() and have driver core handle this instead of
submitting 195+ patches to amend the issue.

[1] https://lore.kernel.org/all/aq1N863eJpD_ZrLU@SMW015318/
---
 drivers/base/power/runtime.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
index fab38bc98113..9a37f6162fe0 100644
--- a/drivers/base/power/runtime.c
+++ b/drivers/base/power/runtime.c
@@ -1874,6 +1874,9 @@ void pm_runtime_reinit(struct device *dev)
 	if (dev->power.runtime_status == RPM_ACTIVE)
 		pm_runtime_set_suspended(dev);
 
+	if (dev->power.use_autosuspend)
+		pm_runtime_dont_use_autosuspend(dev);
+
 	if (dev->power.irq_safe) {
 		spin_lock_irq(&dev->power.lock);
 		dev->power.irq_safe = 0;

---
base-commit: 40288c9206c17eb66a603262e06a58d300d0f279
change-id: 20260919-move-dont-use-autosuspend-0203336ca270

Best regards,
-- 
Kind regards,
Joshua Crofts

Re: [PATCH] PM: runtime: call pm_runtime_dont_use_autosuspend() on reinit()

[Rafael J. Wysocki (Intel)]

On Sat, Sep 19, 2026 at 10:25 AM Joshua Crofts <joshua.crofts1@gmail.com> wrote:
>
> Add a pm_runtime_dont_use_autosuspend() call to pm_runtime_reinit() to
> ensure autosuspend is disabled on driver teardown.
>
> Suggested-by: Frank Li <Frank.Li@nxp.com>
> Signed-off-by: Joshua Crofts <joshua.crofts1@gmail.com>
> ---
> If a driver doesn't use devm_* functions for automatic cleanup on
> teardown and autosuspend is enabled during probe, it should also be
> disabled on teardown in a _remove() function or equivalent. However,
> due to bad documentation, driver authors usually omit the
> pm_runtime_dont_use_autosuspend() call on teardown.
>
> Reason for investigating: this week, several patches popped up in
> the linux-iio mailing list, adding these missing calls to several
> IIO drivers. For some reason I ended up writing a cocci script to
> find other instances of this and found ~195 offending drivers.
>
> After a discussion with Frank Li on the matter [1] when sending a similar
> patch adding these missing calls to the clk subsystem, we arrived to the
> idea of actually adding the dont_use_autosuspend() call to
> pm_runtime_reinit() and have driver core handle this instead of
> submitting 195+ patches to amend the issue.
>
> [1] https://lore.kernel.org/all/aq1N863eJpD_ZrLU@SMW015318/
> ---
>  drivers/base/power/runtime.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
> index fab38bc98113..9a37f6162fe0 100644
> --- a/drivers/base/power/runtime.c
> +++ b/drivers/base/power/runtime.c
> @@ -1874,6 +1874,9 @@ void pm_runtime_reinit(struct device *dev)
>         if (dev->power.runtime_status == RPM_ACTIVE)
>                 pm_runtime_set_suspended(dev);
>
> +       if (dev->power.use_autosuspend)
> +               pm_runtime_dont_use_autosuspend(dev);
> +
>         if (dev->power.irq_safe) {
>                 spin_lock_irq(&dev->power.lock);
>                 dev->power.irq_safe = 0;
>
> ---

Applied as 7.4 material, thanks!