All of lore.kernel.org
 help / color / mirror / Atom feed
From: Lee Jones <lee@kernel.org>
To: lee@kernel.org, Pavel Machek <pavel@kernel.org>,
	Ben Whitten <ben.whitten@gmail.com>,
	Oliver Jowett <oliver@opencloud.com>,
	Richard Purdie <rpurdie@openedhand.com>,
	Christian Marangi <ansuelsmth@gmail.com>,
	"David S. Miller" <davem@davemloft.net>,
	Andrew Lunn <andrew@lunn.ch>,
	linux-leds@vger.kernel.org, linux-kernel@vger.kernel.org
Cc: stable@vger.kernel.org
Subject: [PATCH 03/21] leds: trigger: netdev: Hold RTNL lock across hw_control_get_device() in activate
Date: Wed, 21 Oct 2026 12:32:12 +0000	[thread overview]
Message-ID: <20261021123406.3163547-4-lee@kernel.org> (raw)
In-Reply-To: <20261021123406.3163547-1-lee@kernel.org>

In netdev_trig_activate(), hw_control_get_device() is invoked without
holding rtnl_lock, and the returned struct device pointer is then
dereferenced via dev_name() and passed into set_device_name().

Because implementations such as phy_led_hw_control_get_device() and DSA
port LED callbacks return the embedded struct device of an attached
net_device without incrementing its reference count, a concurrent
network device unregistration or rename can detach and free the
net_device (or its kobject name) before set_device_name() copies the
name and acquires rtnl_lock, leading to a use-after-free.

Move rtnl_lock acquisition to the callers of set_device_name() so that
netdev_trig_activate() holds rtnl_lock across hw_control_get_device(),
dev_name(), and set_device_name().

Fixes: 0316cc5629d1 ("leds: trigger: netdev: init mode if hw control already active")
Closes: https://sashiko.dev/#/bug/linux-ef006cc6-9a69-46b2-ae9f-8de934d37aff
Cc: stable@vger.kernel.org
Signed-off-by: Lee Jones <lee@kernel.org>
---
 drivers/leds/trigger/ledtrig-netdev.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/leds/trigger/ledtrig-netdev.c b/drivers/leds/trigger/ledtrig-netdev.c
index 5b0132484594..d69031bbd2b8 100644
--- a/drivers/leds/trigger/ledtrig-netdev.c
+++ b/drivers/leds/trigger/ledtrig-netdev.c
@@ -279,6 +279,8 @@ static int set_device_name(struct led_netdev_data *trigger_data,
 	struct net_device *new_dev = NULL;
 	char device_name[IFNAMSIZ];
 
+	ASSERT_RTNL();
+
 	if (size >= IFNAMSIZ)
 		return -EINVAL;
 
@@ -292,7 +294,6 @@ static int set_device_name(struct led_netdev_data *trigger_data,
 	/*
 	 * Lock order: rtnl_lock -> netdev instance lock -> trigger_data lock.
 	 */
-	rtnl_lock();
 	if (device_name[0]) {
 		new_dev = dev_get_by_name(&init_net, device_name);
 		if (new_dev)
@@ -319,7 +320,6 @@ static int set_device_name(struct led_netdev_data *trigger_data,
 	mutex_unlock(&trigger_data->lock);
 	if (new_dev)
 		netdev_unlock_ops(new_dev);
-	rtnl_unlock();
 
 	return 0;
 }
@@ -331,7 +331,9 @@ static ssize_t device_name_store(struct device *dev,
 	struct led_netdev_data *trigger_data = led_trigger_get_drvdata(dev);
 	int ret;
 
+	rtnl_lock();
 	ret = set_device_name(trigger_data, buf, size);
+	rtnl_unlock();
 
 	if (ret < 0)
 		return ret;
@@ -760,6 +762,7 @@ static int netdev_trig_activate(struct led_classdev *led_cdev)
 	 * Init already enabled mode in hw control.
 	 */
 	if (supports_hw_control(led_cdev)) {
+		rtnl_lock();
 		dev = led_cdev->hw_control_get_device(led_cdev);
 		if (dev) {
 			const char *name = dev_name(dev);
@@ -771,6 +774,7 @@ static int netdev_trig_activate(struct led_classdev *led_cdev)
 			if (!rc)
 				trigger_data->mode = mode;
 		}
+		rtnl_unlock();
 	}
 
 	led_set_trigger_data(led_cdev, trigger_data);
-- 
2.56.0.218.g1ec904b173-goog


  parent reply	other threads:[~2026-10-21 12:34 UTC|newest]

Thread overview: 28+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-21 12:32 [PATCH 00/21] leds: Assorted bug fixes across core, flash, trigger, and drivers Lee Jones
2026-10-21 12:32 ` [PATCH 01/21] leds: lm3532: Propagate error from devm_regulator_get() Lee Jones
2026-10-21 12:32 ` [PATCH 02/21] leds: lm3532: Use dev_err_probe() when lm3532_parse_node() fails Lee Jones
2026-10-21 12:32 ` Lee Jones [this message]
2026-10-21 12:32 ` [PATCH 04/21] leds: aw2013: Guard LED array index in aw2013_probe_dt() Lee Jones
2026-10-27 14:27   ` Nikita Travkin
2026-10-21 12:32 ` [PATCH 05/21] leds: pca955x: Take a reference on the child firmware node Lee Jones
2026-10-21 12:32 ` [PATCH 06/21] leds: lp50xx: Guard sub-LED count in lp50xx_probe_dt() Lee Jones
2026-10-21 12:32 ` [PATCH 07/21] leds: trigger: netdev: Serialise work cancellation and state updates Lee Jones
2026-10-21 12:32 ` [PATCH 08/21] leds: sgm3140: Replace powerdown timer with delayed work Lee Jones
2026-10-21 12:32 ` [PATCH 09/21] leds: sgm3140: Cancel powerdown work via devres after LED unregistration Lee Jones
2026-10-21 12:32 ` [PATCH 10/21] leds: class: Initialise trigger_lock before exposing sysfs attributes Lee Jones
2026-10-21 12:32 ` [PATCH 11/21] leds: class: Unlock led_access before unregistering device on error Lee Jones
2026-10-21 12:32 ` [PATCH 12/21] leds: aat1290: Retain child device_node reference for device lifetime Lee Jones
2026-10-21 12:32 ` [PATCH 13/21] leds: aat1290: Prevent out-of-bounds read in aat1290_brightness_to_intensity() Lee Jones
2026-10-21 12:32 ` [PATCH 14/21] leds: flash: Avoid divide-by-zero in led_clamp_align() Lee Jones
2026-10-22 19:20   ` Jacek Anaszewski
2026-10-22 21:00   ` David Laight
2026-10-26 17:31     ` Lee Jones
2026-10-26 18:37       ` David Laight
2026-10-21 12:32 ` [PATCH 15/21] leds: lm3532: Validate LED string indices in lm3532_parse_node() Lee Jones
2026-10-21 12:32 ` [PATCH 16/21] leds: lm3532: Clamp full-scale current to minimum bound Lee Jones
2026-10-21 12:32 ` [PATCH 17/21] leds: aat1290: Avoid leaking devres pinctrl handles on strobe source changes Lee Jones
2026-10-21 12:32 ` [PATCH 18/21] leds: aw2013: Unlock chip mutex before registering LEDs Lee Jones
2026-10-27 14:28   ` Nikita Travkin
2026-10-21 12:32 ` [PATCH 19/21] leds: is31fl32xx: Clamp channel current limit in is31fl3293_reset() Lee Jones
2026-10-21 12:32 ` [PATCH 20/21] leds: flash: Terminate sysfs_groups array with NULL Lee Jones
2026-10-21 12:32 ` [PATCH 21/21] leds: lm3692x: Use devm_mutex_init() to avoid premature mutex destruction Lee Jones

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=20261021123406.3163547-4-lee@kernel.org \
    --to=lee@kernel.org \
    --cc=andrew@lunn.ch \
    --cc=ansuelsmth@gmail.com \
    --cc=ben.whitten@gmail.com \
    --cc=davem@davemloft.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-leds@vger.kernel.org \
    --cc=oliver@opencloud.com \
    --cc=pavel@kernel.org \
    --cc=rpurdie@openedhand.com \
    --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 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.