Re: [PATCH] accel/amdxdna: fix usage_count leak when autosuspend_delay is negative

On Fri, Aug 07, 2026 at 10:32:00PM +0800, Guangshuo Li wrote:
> amdxdna_pm_init() calls pm_runtime_use_autosuspend(), but
> amdxdna_pm_fini() does not call the matching
> pm_runtime_dont_use_autosuspend() when tearing down runtime PM.
> 
> If autosuspend_delay is set to a negative value while autosuspend is
> enabled, the runtime PM core increments usage_count to prevent runtime
> suspend. Without calling pm_runtime_dont_use_autosuspend() during
> driver teardown, this reference is not dropped and usage_count remains
> unbalanced.

As I've explained elsewhere, this is just misleading. There is no usage
count leak here as the count is balanced whenever the user re-enables
autosuspend through sysfs (by writing a non-negative timeout).

Drivers should clean up after themselves and disable autosuspend, but
this is more of a clean up than a fix and should not be backported.

You've sent upwards of 60 of these in the matter of just a few days,
some of which have even been picked up. Please send follow-ups (replies
or v2s) as soon as possible to prevent further of these from getting
merged.

> Add the missing pm_runtime_dont_use_autosuspend() call in
> amdxdna_pm_fini() before restoring the runtime PM usage reference and
> forbidding runtime PM.
> 
> This issue was found by manual code inspection.
> 
> Fixes: 063db451832b ("accel/amdxdna: Enhance runtime power management")
> Signed-off-by: Guangshuo Li <[email protected]>

Johan