All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>
To: hyunjungg <hj351016@gmail.com>
Cc: Hans de Goede <hansg@kernel.org>,
	Mark Pearson <mpearson-lenovo@squebb.ca>,
	 "Derek J. Clark" <derekjohn.clark@gmail.com>,
	Rong Zhang <i@rong.moe>,
	 platform-driver-x86@vger.kernel.org,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2] platform/x86: lenovo-yoga-tab2-pro-1380-fastcharger: cancel work on teardown
Date: Mon, 16 Nov 2026 13:30:10 +0200 (EET)	[thread overview]
Message-ID: <6857ee54-81b3-4c8e-106f-d5c871f44375@linux.intel.com> (raw)
In-Reply-To: <20261106074316.2204392-1-hj351016@gmail.com>

On Fri, 6 Nov 2026, hyunjungg wrote:

> From: Hyunjung Ko <hj351016@gmail.com>
> 
> fc->work is scheduled from the extcon notifier callback
> yt2_1380_fc_extcon_evt() (and once from probe), but it is never cancelled.
> The containing struct yt2_1380_fc and the extcon notifier are both
> device-managed on the serdev device, and the serdev driver has no
> ->remove callback, so teardown happens entirely via devres unwinding
> when yt2_1380_fc_pdev_remove() calls serdev_device_remove().
> 
> devm_extcon_register_notifier_all() unregistering on teardown only stops
> future notifications; it does not drain a work item that the notifier has
> already queued. A fast-charger extcon event that races device unbind can
> therefore queue fc->work just before teardown; the work then runs after
> devres has torn the device down. yt2_1380_fc_worker() dereferences not
> only fc itself but also fc->pinctrl (pinctrl_select_state()) and the
> serdev (serdev_device_write_buf()), so it must be drained before those
> device-managed resources are released, not merely before fc is freed.
> 
> Cancel the work via devm_add_action_or_reset(), registered after the
> serdev is opened and pinctrl is acquired but before
> devm_extcon_register_notifier_all(). Because devres runs cleanup in
> reverse order, teardown then unregisters the notifier first (so it can no
> longer queue the work), next cancels the work while fc->pinctrl and the
> serdev are still valid, and only then closes the serdev, releases pinctrl
> and frees fc. This closes both the re-arm window and the use-after-free
> of fc and of its sub-resources in the worker.
> 
> Found by static analysis / code review; compile-tested only, not
> reproduced on hardware (requires the Lenovo Yoga Tablet 2 1380 charger).
> 
> Fixes: b2ed33e8d486 ("platform/x86: Add lenovo-yoga-tab2-pro-1380-fastcharger driver")
> Signed-off-by: Hyunjung Ko <hj351016@gmail.com>
> ---
> v2:
>  - Move the devm_add_action_or_reset(cancel_work) registration from right
>    after INIT_WORK to after devm_serdev_device_open()/devm_pinctrl_get()
>    and before devm_extcon_register_notifier_all(). The v1 placement
>    registered the cancel action first, so devres (LIFO) drained the work
>    last - after the serdev was closed and pinctrl released - leaving the
>    worker's fc->pinctrl / serdev dereferences racing freed resources.
>    The new placement drains the work while those resources are still live.
> 
>  drivers/platform/x86/lenovo/yoga-tab2-pro-1380-fastcharger.c | 17 +++++++++++++++++
>  1 file changed, 17 insertions(+)
> 
> diff --git a/drivers/platform/x86/lenovo/yoga-tab2-pro-1380-fastcharger.c b/drivers/platform/x86/lenovo/yoga-tab2-pro-1380-fastcharger.c
> --- a/drivers/platform/x86/lenovo/yoga-tab2-pro-1380-fastcharger.c
> +++ b/drivers/platform/x86/lenovo/yoga-tab2-pro-1380-fastcharger.c
> @@ -148,6 +148,11 @@
>  	.write_wakeup = serdev_device_write_wakeup,
>  };
> 
> +static void yt2_1380_fc_cancel_work(void *data)
> +{
> +	cancel_work_sync(data);
> +}
> +
>  static int yt2_1380_fc_serdev_probe(struct serdev_device *serdev)
>  {
>  	struct device *dev = &serdev->dev;
> @@ -209,6 +214,18 @@
>  	serdev_device_set_baudrate(serdev, 600);
>  	serdev_device_set_flow_control(serdev, false);
> 
> +	/*
> +	 * Register work cancellation after the serdev is opened and pinctrl
> +	 * is acquired, but before the extcon notifier.  devres unwinds in
> +	 * reverse, so teardown unregisters the notifier first (no new work
> +	 * can be queued), then cancels the work while fc->pinctrl and the
> +	 * serdev are still valid, and only then closes the serdev and
> +	 * releases pinctrl.
> +	 */
> +	ret = devm_add_action_or_reset(dev, yt2_1380_fc_cancel_work, &fc->work);
> +	if (ret)
> +		return ret;

Why not devm_work_autocancel() ?

> +
>  	ret = devm_extcon_register_notifier_all(dev, fc->extcon, &fc->nb);
>  	if (ret)
>  		return dev_err_probe(dev, ret, "registering extcon notifier\n");
> 

-- 
 i.


  reply	other threads:[~2026-11-16 11:30 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-06  4:36 [PATCH] platform/x86: lenovo-yoga-tab2-pro-1380-fastcharger: cancel work on teardown hyunjungg
2026-11-06  7:43 ` [PATCH v2] " hyunjungg
2026-11-16 11:30   ` Ilpo Järvinen [this message]
2026-11-16 12:19     ` khj

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=6857ee54-81b3-4c8e-106f-d5c871f44375@linux.intel.com \
    --to=ilpo.jarvinen@linux.intel.com \
    --cc=derekjohn.clark@gmail.com \
    --cc=hansg@kernel.org \
    --cc=hj351016@gmail.com \
    --cc=i@rong.moe \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mpearson-lenovo@squebb.ca \
    --cc=platform-driver-x86@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.