Linux kernel -stable discussions
 help / color / mirror / Atom feed
From: "Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>
To: Hongyan Xu <getshell@seu.edu.cn>
Cc: Hans de Goede <hansg@kernel.org>,
	MyungJoo Ham <myungjoo.ham@samsung.com>,
	 Chanwoo Choi <cw00.choi@samsung.com>,
	 "Derek J. Clark" <derekjohn.clark@gmail.com>,
	 Mark Pearson <mpearson-lenovo@squebb.ca>,
	Rong Zhang <i@rong.moe>,
	 platform-driver-x86@vger.kernel.org,
	LKML <linux-kernel@vger.kernel.org>,
	 stable@vger.kernel.org
Subject: Re: [PATCH v2 1/2] extcon: synchronize all-connector notifier removal
Date: Mon, 16 Nov 2026 13:28:00 +0200 (EET)	[thread overview]
Message-ID: <4f3601df-029b-0ee9-41ea-c0ec29c5f80d@linux.intel.com> (raw)
In-Reply-To: <179474326273.103840.4798359664806814358@seu.edu.cn>

On Sun, 15 Nov 2026, Hongyan Xu wrote:

> extcon_sync() invokes the all-connector raw notifier chain after
> dropping edev->lock. Unregister unlinks a block under that lock but
> does not wait for an in-flight callback, so managed consumers can
> release their state while a callback still publishes work.
> 
> Use an atomic notifier chain for the all-connector notifications.
> Its RCU read section and synchronized unregister provide the
> required callback quiescence. Use the notifier implementation's
> own writer lock and perform its grace-period wait outside edev's
> spinlock. Document callback and unregister context requirements.
> 
> The current all-connector consumers defer their operations to work
> items and their callbacks do not sleep. Per-ID raw chains retain
> their existing behavior: some OMAP callbacks perform regulator
> operations and need a separate context-compatible repair.
> 
> Found by the author's in-house static analysis tool.

I think sashiko found this too:

Reported-by: Sashiko <sashiko-bot@kernel.org>
Closes: https://sashiko.dev/#/patchset/20260806060526.1189-1-getshell%40seu.edu.cn

> Fixes: 815429b39d94 ("extcon: Add new extcon_register_notifier_all() to monitor all external connectors")
> Cc: stable@vger.kernel.org
> Link: https://lore.kernel.org/all/2ef4ad43-497f-c40f-b18a-f5b75799d95d@linux.intel.com/
> Signed-off-by: Hongyan Xu <getshell@seu.edu.cn>
> Signed-off-by: Zehao Gao <gaozehao@iie.ac.cn>

-- 
 i.


> ---
> 
> Discovery: found by the author's in-house static analysis tool.
> 
>  Documentation/driver-api/extcon.rst |  2 +-
>  drivers/extcon/extcon.c             | 28 +++++++++-------------------
>  drivers/extcon/extcon.h             |  2 +-
>  include/linux/extcon.h              |  4 ++++
>  4 files changed, 15 insertions(+), 21 deletions(-)
> 
> diff --git a/Documentation/driver-api/extcon.rst b/Documentation/driver-api/extcon.rst
> index d3217b9..384d743 100644
> --- a/Documentation/driver-api/extcon.rst
> +++ b/Documentation/driver-api/extcon.rst
> @@ -70,7 +70,7 @@ The core structure representing an Extcon device::
>          /* Internal data */
>          struct device dev;
>          unsigned int id;
> -        struct raw_notifier_head nh_all;
> +        struct atomic_notifier_head nh_all;
>          struct raw_notifier_head *nh;
>          struct list_head entry;
>          int max_supported;
> diff --git a/drivers/extcon/extcon.c b/drivers/extcon/extcon.c
> index 98d85cc..b8d8c2a 100644
> --- a/drivers/extcon/extcon.c
> +++ b/drivers/extcon/extcon.c
> @@ -456,10 +456,10 @@ int extcon_sync(struct extcon_dev *edev, unsigned int id)
>  	raw_notifier_call_chain(&edev->nh[index], state, edev);
>  
>  	/*
> -	 * Call functions in a raw notifier chain for the all supported
> +	 * Call functions in the atomic notifier chain for all supported
>  	 * external connectors.
>  	 */
> -	raw_notifier_call_chain(&edev->nh_all, state, edev);
> +	atomic_notifier_call_chain(&edev->nh_all, state, edev);
>  
>  	spin_lock_irqsave(&edev->lock, flags);
>  	/* This could be in interrupt handler */
> @@ -967,6 +967,7 @@ EXPORT_SYMBOL_GPL(extcon_unregister_notifier);
>   *
>   * Note that this function registers a notifier block in order to receive
>   * the state change of all supported external connectors from extcon device.
> + * Callbacks run under the RCU read-side critical section and must not sleep.
>   * And the second parameter given to the callback of nb (val) is
>   * the current state and the third pameter is the pointer of extcon device.
>   *
> @@ -975,17 +976,10 @@ EXPORT_SYMBOL_GPL(extcon_unregister_notifier);
>  int extcon_register_notifier_all(struct extcon_dev *edev,
>  				struct notifier_block *nb)
>  {
> -	unsigned long flags;
> -	int ret;
> -
>  	if (!edev || !nb)
>  		return -EINVAL;
>  
> -	spin_lock_irqsave(&edev->lock, flags);
> -	ret = raw_notifier_chain_register(&edev->nh_all, nb);
> -	spin_unlock_irqrestore(&edev->lock, flags);
> -
> -	return ret;
> +	return atomic_notifier_chain_register(&edev->nh_all, nb);
>  }
>  EXPORT_SYMBOL_GPL(extcon_register_notifier_all);
>  
> @@ -995,21 +989,17 @@ EXPORT_SYMBOL_GPL(extcon_register_notifier_all);
>   * @nb:		a notifier block to be registered
>   *
>   * Returns 0 if success or error number if fail.
> + *
> + * Waits for in-flight callbacks. Must be called from process context and
> + * outside a notifier callback.
>   */
>  int extcon_unregister_notifier_all(struct extcon_dev *edev,
>  				struct notifier_block *nb)
>  {
> -	unsigned long flags;
> -	int ret;
> -
>  	if (!edev || !nb)
>  		return -EINVAL;
>  
> -	spin_lock_irqsave(&edev->lock, flags);
> -	ret = raw_notifier_chain_unregister(&edev->nh_all, nb);
> -	spin_unlock_irqrestore(&edev->lock, flags);
> -
> -	return ret;
> +	return atomic_notifier_chain_unregister(&edev->nh_all, nb);
>  }
>  EXPORT_SYMBOL_GPL(extcon_unregister_notifier_all);
>  
> @@ -1300,7 +1290,7 @@ int extcon_dev_register(struct extcon_dev *edev)
>  	for (index = 0; index < edev->max_supported; index++)
>  		RAW_INIT_NOTIFIER_HEAD(&edev->nh[index]);
>  
> -	RAW_INIT_NOTIFIER_HEAD(&edev->nh_all);
> +	ATOMIC_INIT_NOTIFIER_HEAD(&edev->nh_all);
>  
>  	dev_set_drvdata(&edev->dev, edev);
>  	dev_set_name(&edev->dev, "extcon%d", edev->id);
> diff --git a/drivers/extcon/extcon.h b/drivers/extcon/extcon.h
> index 9461826..5a3f0c4 100644
> --- a/drivers/extcon/extcon.h
> +++ b/drivers/extcon/extcon.h
> @@ -48,7 +48,7 @@ struct extcon_dev {
>  	/* Internal data. Please do not set. */
>  	struct device dev;
>  	unsigned int id;
> -	struct raw_notifier_head nh_all;
> +	struct atomic_notifier_head nh_all;
>  	struct raw_notifier_head *nh;
>  	struct list_head entry;
>  	int max_supported;
> diff --git a/include/linux/extcon.h b/include/linux/extcon.h
> index e596a0a..869bac9 100644
> --- a/include/linux/extcon.h
> +++ b/include/linux/extcon.h
> @@ -209,6 +209,10 @@ void devm_extcon_unregister_notifier(struct device *dev,
>  				struct extcon_dev *edev, unsigned int id,
>  				struct notifier_block *nb);
>  
> +/*
> + * All-connector callbacks must not sleep. Unregister waits for callbacks and
> + * must be called from process context, outside a notifier callback.
> + */
>  int extcon_register_notifier_all(struct extcon_dev *edev,
>  				struct notifier_block *nb);
>  int extcon_unregister_notifier_all(struct extcon_dev *edev,
> 

  reply	other threads:[~2026-11-16 11:28 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-15 11:47 [PATCH v2 0/2] extcon: synchronize all-connector removal and fix Lenovo fastcharger teardown Hongyan Xu
2026-11-15 11:47 ` [PATCH v2 1/2] extcon: synchronize all-connector notifier removal Hongyan Xu
2026-11-16 11:28   ` Ilpo Järvinen [this message]
2026-11-15 11:47 ` [PATCH v2 2/2] platform/x86: lenovo: cancel fastcharger work on detach Hongyan Xu

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=4f3601df-029b-0ee9-41ea-c0ec29c5f80d@linux.intel.com \
    --to=ilpo.jarvinen@linux.intel.com \
    --cc=cw00.choi@samsung.com \
    --cc=derekjohn.clark@gmail.com \
    --cc=getshell@seu.edu.cn \
    --cc=hansg@kernel.org \
    --cc=i@rong.moe \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mpearson-lenovo@squebb.ca \
    --cc=myungjoo.ham@samsung.com \
    --cc=platform-driver-x86@vger.kernel.org \
    --cc=stable@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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox