Re: [PATCH v2] tpm: use DEFINE_SIMPLE_DEV_PM_OPS and pm_sleep_ptr()

[Nathan Chancellor <nathan@kernel.org>]

On Wed, Sep 09, 2026 at 04:24:59PM +0800, Li Jun wrote:
> Replace the deprecated SIMPLE_DEV_PM_OPS with DEFINE_SIMPLE_DEV_PM_OPS
> and wrap the struct assignment with pm_sleep_ptr(). The old code keeps
> a completely empty dev_pm_ops structure -- all callbacks are NULL --
> as dead code in the object file when CONFIG_PM_SLEEP=n, because
> SET_SYSTEM_SLEEP_PM_OPS() expands to nothing while __maybe_unused on
> the struct only silences the warning without removing the struct.
> Switching to DEFINE_SIMPLE_DEV_PM_OPS() lets pm_sleep_ptr() eliminate
> the callback references at compile time, and wrapping the struct
> assignment with pm_sleep_ptr() drops the entire structure when
> CONFIG_PM_SLEEP is disabled, instead of retaining an empty one.
> 
> Signed-off-by: Li Jun <lijun01@kylinos.cn>
...
> diff --git a/drivers/char/tpm/tpm_infineon.c b/drivers/char/tpm/tpm_infineon.c
> index 7638b65b851b..84a0d3e5f679 100644
> --- a/drivers/char/tpm/tpm_infineon.c
> +++ b/drivers/char/tpm/tpm_infineon.c
> @@ -612,7 +612,7 @@ static int tpm_inf_resume(struct device *dev)
>  	return tpm_pm_resume(dev);
>  }
>  #endif
> -static SIMPLE_DEV_PM_OPS(tpm_inf_pm, tpm_pm_suspend, tpm_inf_resume);
> +static DEFINE_SIMPLE_DEV_PM_OPS(tpm_inf_pm, tpm_pm_suspend, tpm_inf_resume);
>  
>  static struct pnp_driver tpm_inf_pnp_driver = {
>  	.name = "tpm_inf_pnp",
> @@ -620,7 +620,7 @@ static struct pnp_driver tpm_inf_pnp_driver = {
>  	.probe = tpm_inf_pnp_probe,
>  	.remove = tpm_inf_pnp_remove,
>  	.driver = {
> -		.pm = &tpm_inf_pm,
> +		.pm = pm_sleep_ptr(&tpm_inf_pm),
>  	}
>  };
>  

This patch is in -next as commit 2249200ee420 ("tpm: use
DEFINE_SIMPLE_DEV_PM_OPS and pm_sleep_ptr()"), where it breaks the build
for CONFIG_PM_SLEEP=n kernels because some of the functions are not
visible in these configurations.

  https://lore.kernel.org/202609121207.tM2OfBmR-lkp@intel.com/
  https://lore.kernel.org/202609121338.qo2mQWRl-lkp@intel.com/

It should be fairly easy enough to expose these functions regardless of
CONFIG_PM_SLEEP (or add some stubs) but I think this patch should be
dropped until it can be done in a new version.

-- 
Cheers,
Nathan