Re: [PATCH v2] Input: cyapa - fix runtime PM leak in cyapa_update_rt_suspend_scanrate()
From: Dmitry Torokhov
Date: Mon Nov 09 2026 - 02:25:55 EST
On Sun, Nov 08, 2026 at 06:45:47AM +0000, Wentao Liang wrote:
> Hi Dmitry,
>
> Sorry for the very late reply.
>
> It does not have to be the "sync" variant. I used
> pm_runtime_put_sync_autosuspend() only to mirror the normal flow of the
> same function and the other runtime PM call sites in cyapa.c, but there
> is nothing on this path that needs a synchronous transition: no device
> state has been changed yet, both put variants mark the device busy
> first, and cyapa_start_runtime() sets a 2 s autosuspend delay, so either
> of them would only arm the autosuspend timer before the store returns.
> v2 uses pm_runtime_put_autosuspend(), the counterpart of
> pm_runtime_get_sync() for devices that use autosuspend.
>
> For the record on how the patch was produced: the candidate came from
> static analysis tooling, and DeepSeek and GPT-6.1 Sol were used to audit
> it and to help write the change. I reviewed the result and I am
> responsible for it. v2 is build-tested only: it compiles cleanly on
> x86-64 with CONFIG_MOUSE_CYAPA=y, including W=1, and passes checkpatch.
> I have no Cypress trackpad to test it on.
>
> Thanks,
> Wentao
>
> From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
> From: Wentao Liang <vulab@xxxxxxxxxxx>
> Date: Sun, 8 Nov 2026 06:45:47 +0000
> Subject: [PATCH v2] Input: cyapa - fix runtime PM leak in cyapa_update_rt_suspend_scanrate()
>
> pm_runtime_get_sync() unconditionally increments the runtime PM usage
> counter. If mutex_lock_interruptible() is interrupted, the function
> returns without dropping that reference, so the usage count stays
> elevated for as long as the driver is bound and the device can no longer
> be autosuspended. Drop the reference on the error path.
>
> Use pm_runtime_put_autosuspend() rather than the synchronous variant
> there: nothing has been done to the device on this path, so there is
> nothing to wait for, and with the 2 s autosuspend delay the driver
> configures the put only arms the autosuspend timer instead of suspending
> the device.
>
> Fixes: 672865080a8f ("Input: cyapa - add runtime power management support")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Wentao Liang <vulab@xxxxxxxxxxx>
Applied, thank you.
--
Dmitry