Re: [PATCH] crypto: omap-sham: remove redundant pm_runtime_dont_use_autosuspend() calls

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [High] crypto: omap-sham: missing runtime suspend before
  pm_runtime_disable()
--

commit 3e6145ea4d9da519423f99c72ef5ab8b8512ee57
Author: Joshua Crofts <crofts@kernel.org>

crypto: omap-sham: remove redundant pm_runtime_dont_use_autosuspend() calls

This commit removes the pm_runtime_dont_use_autosuspend() calls from the
probe error path and driver remove callback in omap-sham.

> diff --git a/drivers/crypto/omap-sham.c b/drivers/crypto/omap-sham.c
> index 5e739415a0102..72182f1900849 100644
> --- a/drivers/crypto/omap-sham.c
> +++ b/drivers/crypto/omap-sham.c
> @@ -2109,7 +2109,6 @@ static int omap_sham_probe(struct platform_device *pdev)
>  	list_del(&dd->list);
>  	spin_unlock_bh(&sham.lock);
>  err_pm:
> -	pm_runtime_dont_use_autosuspend(dev);
>  	pm_runtime_disable(dev);
>  	if (!dd->polling_mode)
>  		dma_release_channel(dd->dma_lch);
> @@ -2131,7 +2130,6 @@ static void omap_sham_remove(struct platform_device *pdev)
>  	omap_sham_unregister_algs(dd->pdata);
>  	crypto_engine_exit(dd->engine);
>  	cancel_work_sync(&dd->done_task);
> -	pm_runtime_dont_use_autosuspend(&pdev->dev);
>  	pm_runtime_disable(&pdev->dev);
>  
>  	if (!dd->polling_mode)

[Severity: High]
Does removing pm_runtime_dont_use_autosuspend() before pm_runtime_disable()
prevent .runtime_suspend from running on probe failure and driver removal?

Because autosuspend is enabled with a 1000 ms delay, pm_runtime_put_sync()
in omap_sham_probe() and in __device_release_driver() before
omap_sham_remove() only arms the autosuspend timer and leaves the device in
RPM_ACTIVE.  Previously, calling pm_runtime_dont_use_autosuspend() while
runtime PM was still enabled triggered a synchronous rpm_idle() ->
rpm_suspend() transition before pm_runtime_disable().

Without it, pm_runtime_disable() cancels the pending autosuspend timer and
disables runtime PM while the device is still active.  When
pm_runtime_reinit() later runs in device_unbind_cleanup(), runtime PM is
already disabled, so it only resets the software status to RPM_SUSPENDED via
pm_runtime_set_suspended() without invoking .runtime_suspend.  This leaves
hardware clocks and the power domain enabled, and on OMAP2/OMAP3 platforms
using omap_device_pm_domain, probe failure leaves od->_state at
OMAP_DEVICE_STATE_ENABLED while runtime_status is RPM_SUSPENDED, causing any
subsequent probe to fail with -EINVAL in omap_device_enable().

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261107-crypto-dont-use-autosuspend-removal-v1-1-62696ad3ff0e@kernel.org?part=1