This patch series introduces wrapper functions for locking
system_transition_mutex used in system sleep (suspend/hibernate) code,
and refactors the existing usage in kernel/power to improve maintainability,
readability, and future extensibility.

Currently, mutex_lock/unlock(&system_transition_mutex) is used directly
in multiple places in the suspend and hibernation paths. This results in
boilerplate repetition and makes it harder to change the locking mechanism
later if needed (e.g., for debugging, tracing, or replacing the mutex
with another primitive).

Summary:

- Patch 1 replaces mutex_lock with lock_system_sleep.

- Patch 2 adds a non-blocking `try_lock_system_sleep()` for code paths that
  should proceed only if the lock is immediately available.

- Patch 3 replaces remaining uses of `system_transition_mutex` in
  `kernel/power/` with the new helper functions.

This change brings all system sleep transition locking behind well-named
abstractions, preparing the code for future evolution while simplifying
reasoning and avoiding repeated direct mutex usage.

change log:
    V2:
    Fix PF_NOFREEZE leakage on try_lock_system_sleep() failure.

Zihuan Zhang (3):
  PM / Sleep: Replace mutex_[un]lock(&system_transition_mutex) with
    [un]lock_system_sleep()
  PM / Sleep:  Introduce try_lock_system_sleep()
  PM / Sleep: Replace mutex_trylock(&system_transition_mutex) with
    try_lock_system_sleep()

 include/linux/suspend.h  |  2 ++
 kernel/power/hibernate.c | 11 +++++++----
 kernel/power/main.c      | 14 ++++++++++++++
 kernel/power/suspend.c   |  7 +++++--
 kernel/power/user.c      |  6 ++++--
 kernel/reboot.c          |  5 +++--
 6 files changed, 35 insertions(+), 10 deletions(-)

-- 
2.25.1

On Thu, May 22, 2025 at 4:35 AM Zihuan Zhang <zhangzihuan@kylinos.cn> wrote:
>
> This patch series introduces wrapper functions for locking
> system_transition_mutex used in system sleep (suspend/hibernate) code,
> and refactors the existing usage in kernel/power to improve maintainability,
> readability, and future extensibility.
>
> Currently, mutex_lock/unlock(&system_transition_mutex) is used directly
> in multiple places in the suspend and hibernation paths. This results in
> boilerplate repetition and makes it harder to change the locking mechanism
> later if needed (e.g., for debugging, tracing, or replacing the mutex
> with another primitive).
>
> Summary:
>
> - Patch 1 replaces mutex_lock with lock_system_sleep.
>
> - Patch 2 adds a non-blocking `try_lock_system_sleep()` for code paths that
>   should proceed only if the lock is immediately available.
>
> - Patch 3 replaces remaining uses of `system_transition_mutex` in
>   `kernel/power/` with the new helper functions.
>
> This change brings all system sleep transition locking behind well-named
> abstractions, preparing the code for future evolution while simplifying
> reasoning and avoiding repeated direct mutex usage.
>
> change log:
>     V2:
>     Fix PF_NOFREEZE leakage on try_lock_system_sleep() failure.
>
> Zihuan Zhang (3):
>   PM / Sleep: Replace mutex_[un]lock(&system_transition_mutex) with
>     [un]lock_system_sleep()
>   PM / Sleep:  Introduce try_lock_system_sleep()
>   PM / Sleep: Replace mutex_trylock(&system_transition_mutex) with
>     try_lock_system_sleep()
>
>  include/linux/suspend.h  |  2 ++
>  kernel/power/hibernate.c | 11 +++++++----
>  kernel/power/main.c      | 14 ++++++++++++++
>  kernel/power/suspend.c   |  7 +++++--
>  kernel/power/user.c      |  6 ++++--
>  kernel/reboot.c          |  5 +++--
>  6 files changed, 35 insertions(+), 10 deletions(-)
>
> --

This is a bit too late for 6.16, please resubmit when 6.16-rc1 is out.

Thanks!