linux-kernel - Re: [PATCH 3/3] PM: runtime: Update kerneldoc return codes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aLaaLc0TPhtm-H5j@kekkonen.localdomain>
Date: Tue, 2 Sep 2025 10:18:05 +0300
From: Sakari Ailus <sakari.ailus@...ux.intel.com>
To: Brian Norris <briannorris@...omium.org>
Cc: "Rafael J. Wysocki" <rafael@...nel.org>,
	Pavel Machek <pavel@...nel.org>,
	"Rafael J . Wysocki" <rafael.j.wysocki@...el.com>,
	kunit-dev@...glegroups.com, Len Brown <lenb@...nel.org>,
	linux-pm@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/3] PM: runtime: Update kerneldoc return codes

Hi Brian,

Thanks for posting this. A few comments below.

On Thu, Aug 28, 2025 at 05:28:28PM -0700, Brian Norris wrote:
> APIs based on __pm_runtime_idle() (pm_runtime_idle(), pm_request_idle())
> do not return 1 when already suspended. They return -EAGAIN. This is
> already covered in the docs, so the entry for "1" is redundant and
> conflicting.
> 
> (pm_runtime_put() and pm_runtime_put_sync() were previously incorrect,
> but that's fixed in "PM: runtime: pm_runtime_put{,_sync}() returns 1
> when already suspended", to ensure consistency with APIs like
> pm_runtime_put_autosuspend().)
> 
> RPM_GET_PUT APIs based on __pm_runtime_suspend() do return 1 when
> already suspended, but the language is a little unclear -- it's not
> really an "error", so it seems better to list as a clarification before
> the 0/success case. Additionally, they only actually return 1 when the
> refcount makes it to 0; if the usage_count is still non-zero, we return
> 0.
> 
> pm_runtime_put(), etc., also don't appear at first like they can ever
> see "-EAGAIN: Runtime PM usage_count non-zero", because in non-racy
> conditions, pm_runtime_put() would drop its reference count, see it's
> non-zero, and return early (in __pm_runtime_idle()). However, it's
> possible to race with another actor that increments the usage_count
> afterward, since rpm_idle() is protected by a separate lock; in such a
> case, we may see -EAGAIN.
> 
> Because this case is only seen in the presence of concurrent actors, it
> makes sense to clarify that this is when "usage_count **became**
> non-zero", by way of some racing actor.
> 
> Lastly, pm_runtime_put_sync_suspend() duplicated some -EAGAIN language.
> Fix that.
> 
> Fixes: 271ff96d6066 ("PM: runtime: Document return values of suspend-related API functions")
> Link: https://lore.kernel.org/linux-pm/aJ5pkEJuixTaybV4@google.com/
> Signed-off-by: Brian Norris <briannorris@...omium.org>
> ---
> 
>  include/linux/pm_runtime.h | 33 ++++++++++++++++++---------------
>  1 file changed, 18 insertions(+), 15 deletions(-)
> 
> diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
> index d88d6b6ccf5b..fd17ffe1bc79 100644
> --- a/include/linux/pm_runtime.h
> +++ b/include/linux/pm_runtime.h
> @@ -356,7 +356,6 @@ static inline int pm_runtime_force_resume(struct device *dev) { return -ENXIO; }
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -EINPROGRESS: Suspend already in progress.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   * Other values and conditions for the above values are possible as returned by
>   * Runtime PM idle and suspend callbacks.
>   */
> @@ -439,7 +438,6 @@ static inline int pm_runtime_resume(struct device *dev)
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -EINPROGRESS: Suspend already in progress.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   */
>  static inline int pm_request_idle(struct device *dev)
>  {
> @@ -540,15 +538,16 @@ static inline int pm_runtime_resume_and_get(struct device *dev)
>   * equal to 0, queue up a work item for @dev like in pm_request_idle().
>   *
>   * Return:
> + * * 1: Usage counts dropped to zero, but device was already suspended.

Does this actually happen? pm_runtime_put() calls __pm_runtime_idle() that
doesn't appear to return 1 in any case.

>   * * 0: Success.
>   * * -EINVAL: Runtime PM error.
>   * * -EACCES: Runtime PM disabled.
> - * * -EAGAIN: Runtime PM usage_count non-zero or Runtime PM status change ongoing.
> + * * -EAGAIN: Runtime PM usage_count became non-zero or Runtime PM status
> + *            change ongoing.
>   * * -EBUSY: Runtime PM child_count non-zero.
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -EINPROGRESS: Suspend already in progress.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   */
>  static inline int pm_runtime_put(struct device *dev)
>  {
> @@ -565,15 +564,16 @@ DEFINE_FREE(pm_runtime_put, struct device *, if (_T) pm_runtime_put(_T))
>   * equal to 0, queue up a work item for @dev like in pm_request_autosuspend().
>   *
>   * Return:
> + * * 1: Usage counts dropped to zero, but device was already suspended.
>   * * 0: Success.

"usage_count" and "usage counter" is being used in kernel-doc already, I'd
use either of the two. "usage_count" refers directly to the field in struct
dev_pm_info (and is used a few lines below). Same elsewhere.

>   * * -EINVAL: Runtime PM error.
>   * * -EACCES: Runtime PM disabled.
> - * * -EAGAIN: Runtime PM usage_count non-zero or Runtime PM status change ongoing.
> + * * -EAGAIN: Runtime PM usage_count became non-zero or Runtime PM status
> + *            change ongoing.
>   * * -EBUSY: Runtime PM child_count non-zero.
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -EINPROGRESS: Suspend already in progress.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   */
>  static inline int __pm_runtime_put_autosuspend(struct device *dev)
>  {
> @@ -590,15 +590,16 @@ static inline int __pm_runtime_put_autosuspend(struct device *dev)
>   * in pm_request_autosuspend().
>   *
>   * Return:
> + * * 1: Usage counts dropped to zero, but device was already suspended.
>   * * 0: Success.
>   * * -EINVAL: Runtime PM error.
>   * * -EACCES: Runtime PM disabled.
> - * * -EAGAIN: Runtime PM usage_count non-zero or Runtime PM status change ongoing.
> + * * -EAGAIN: Runtime PM usage_count became non-zero or Runtime PM status
> + *            change ongoing.
>   * * -EBUSY: Runtime PM child_count non-zero.
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -EINPROGRESS: Suspend already in progress.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   */
>  static inline int pm_runtime_put_autosuspend(struct device *dev)
>  {
> @@ -619,14 +620,15 @@ static inline int pm_runtime_put_autosuspend(struct device *dev)
>   * if it returns an error code.
>   *
>   * Return:
> + * * 1: Usage counts dropped to zero, but device was already suspended.
>   * * 0: Success.

Does this happen (pm_runtime_put_sync() calls __pm_runtime_idle())?

>   * * -EINVAL: Runtime PM error.
>   * * -EACCES: Runtime PM disabled.
> - * * -EAGAIN: Runtime PM usage_count non-zero or Runtime PM status change ongoing.
> + * * -EAGAIN: Runtime PM usage_count became non-zero or Runtime PM status
> + *            change ongoing.
>   * * -EBUSY: Runtime PM child_count non-zero.
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   * Other values and conditions for the above values are possible as returned by
>   * Runtime PM suspend callbacks.
>   */
> @@ -646,15 +648,15 @@ static inline int pm_runtime_put_sync(struct device *dev)
>   * if it returns an error code.
>   *
>   * Return:
> + * * 1: Usage counts dropped to zero, but device was already suspended.
>   * * 0: Success.
>   * * -EINVAL: Runtime PM error.
>   * * -EACCES: Runtime PM disabled.
> - * * -EAGAIN: Runtime PM usage_count non-zero or Runtime PM status change ongoing.
> - * * -EAGAIN: usage_count non-zero or Runtime PM status change ongoing.
> + * * -EAGAIN: Runtime PM usage_count became non-zero or Runtime PM status
> + *            change ongoing.
>   * * -EBUSY: Runtime PM child_count non-zero.
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   * Other values and conditions for the above values are possible as returned by
>   * Runtime PM suspend callbacks.
>   */
> @@ -677,15 +679,16 @@ static inline int pm_runtime_put_sync_suspend(struct device *dev)
>   * if it returns an error code.
>   *
>   * Return:
> + * * 1: Usage counts dropped to zero, but device was already suspended.
>   * * 0: Success.
>   * * -EINVAL: Runtime PM error.
>   * * -EACCES: Runtime PM disabled.
> - * * -EAGAIN: Runtime PM usage_count non-zero or Runtime PM status change ongoing.
> + * * -EAGAIN: Runtime PM usage_count became non-zero or Runtime PM status
> + *            change ongoing.
>   * * -EBUSY: Runtime PM child_count non-zero.
>   * * -EPERM: Device PM QoS resume latency 0.
>   * * -EINPROGRESS: Suspend already in progress.
>   * * -ENOSYS: CONFIG_PM not enabled.
> - * * 1: Device already suspended.
>   * Other values and conditions for the above values are possible as returned by
>   * Runtime PM suspend callbacks.
>   */

-- 
Kind regards,

Sakari Ailus