From: Chengfeng Ye <nicoyip.dev@gmail.com>
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
Jiri Slaby <jirislaby@kernel.org>,
Richard Watts <rrw@kynesim.co.uk>
Cc: linux-kernel@vger.kernel.org, linux-serial@vger.kernel.org,
Chengfeng Ye <nicoyip.dev@gmail.com>,
stable@vger.kernel.org
Subject: [PATCH] tty: serialize device registration and removal
Date: Fri, 31 Jul 2026 22:22:20 +0800 [thread overview]
Message-ID: <20260731142220.3000214-1-nicoyip.dev@gmail.com> (raw)
tty_cdev_add() stores a newly allocated cdev in driver->cdevs[index]
before initializing and registering it. tty_unregister_device() reads and
deletes the same pointer without serialization.
The race can proceed as follows:
registration removal
------------ -------
driver->cdevs[index] = cdev
cdev_del(driver->cdevs[index])
kobject_put() frees cdev
driver->cdevs[index]->ops = &tty_fops
The final write accesses freed memory. Concurrent registration can also
replace the slot while another caller is still using its cdev.
KASAN reported:
BUG: KASAN: slab-use-after-free in tty_cdev_add+0x5c7/0x670
Write of size 8 at addr ffff88810b298c48
Call Trace:
tty_cdev_add+0x5c7/0x670
tty_register_device_attr+0x458/0x810
gsm_activate_mux+0x105/0x2e0
gsmld_ioctl+0x92f/0x14d0
Allocated by task 111:
cdev_alloc+0x99/0x130
tty_cdev_add+0x93/0x670
Freed by task 110:
cdev_dynamic_release+0x38/0x50
kobject_put+0x14d/0x280
tty_unregister_device+0x1b7/0x3c0
Add a per-driver mutex and hold it across complete device registration and
removal. This prevents either transition from replacing or freeing a cdev
until the other has finished.
Fixes: a3a10ce3429e ("Avoid usb reset crashes by making tty_io cdevs truly dynamic")
Cc: stable@vger.kernel.org
Signed-off-by: Chengfeng Ye <nicoyip.dev@gmail.com>
---
drivers/tty/tty_io.c | 13 ++++++++++---
include/linux/tty_driver.h | 3 +++
2 files changed, 13 insertions(+), 3 deletions(-)
diff --git a/drivers/tty/tty_io.c b/drivers/tty/tty_io.c
index 6b283fd03ff8..926a9b586c7d 100644
--- a/drivers/tty/tty_io.c
+++ b/drivers/tty/tty_io.c
@@ -3185,7 +3185,8 @@ static int tty_cdev_add(struct tty_driver *driver, dev_t dev,
* that bit is not set, this function should not be called by a tty
* driver.
*
- * Locking: ??
+ * Locking: device registration and removal are serialized by
+ * @driver->device_mutex.
*
* Return: A pointer to the struct device for this tty device (or
* ERR_PTR(-EFOO) on error).
@@ -3217,7 +3218,8 @@ static void tty_device_create_release(struct device *dev)
* tty driver's flags have the %TTY_DRIVER_DYNAMIC_DEV bit set. If that bit is
* not set, this function should not be called by a tty driver.
*
- * Locking: ??
+ * Locking: device registration and removal are serialized by
+ * @driver->device_mutex.
*
* Return: A pointer to the struct device for this tty device (or
* ERR_PTR(-EFOO) on error).
@@ -3257,6 +3259,7 @@ struct device *tty_register_device_attr(struct tty_driver *driver,
dev_set_drvdata(dev, drvdata);
dev_set_uevent_suppress(dev, 1);
+ guard(mutex)(&driver->device_mutex);
retval = device_register(dev);
if (retval)
@@ -3300,10 +3303,13 @@ EXPORT_SYMBOL_GPL(tty_register_device_attr);
* If a tty device is registered with a call to tty_register_device() then
* this function must be called when the tty device is gone.
*
- * Locking: ??
+ * Locking: device registration and removal are serialized by
+ * @driver->device_mutex.
*/
void tty_unregister_device(struct tty_driver *driver, unsigned index)
{
+ guard(mutex)(&driver->device_mutex);
+
device_destroy(&tty_class, MKDEV(driver->major, driver->minor_start) + index);
if (!(driver->flags & TTY_DRIVER_DYNAMIC_ALLOC)) {
cdev_del(driver->cdevs[index]);
@@ -3338,6 +3344,7 @@ struct tty_driver *__tty_alloc_driver(unsigned int lines, struct module *owner,
return ERR_PTR(-ENOMEM);
kref_init(&driver->kref);
+ mutex_init(&driver->device_mutex);
driver->num = lines;
driver->owner = owner;
driver->flags = flags;
diff --git a/include/linux/tty_driver.h b/include/linux/tty_driver.h
index 1f2896e56e77..bf8c9d8ec698 100644
--- a/include/linux/tty_driver.h
+++ b/include/linux/tty_driver.h
@@ -7,6 +7,7 @@
#include <linux/kref.h>
#include <linux/list.h>
#include <linux/cdev.h>
+#include <linux/mutex.h>
#include <linux/uaccess.h>
#include <linux/termios.h>
#include <linux/seq_file.h>
@@ -497,6 +498,7 @@ struct tty_operations {
* @kref: reference counting. Reaching zero frees all the internals and the
* driver.
* @cdevs: allocated/registered character /dev devices
+ * @device_mutex: serializes device registration and removal
* @owner: modules owning this driver. Used drivers cannot be rmmod'ed.
* Automatically set by tty_alloc_driver().
* @driver_name: name of the driver used in /proc/tty
@@ -532,6 +534,7 @@ struct tty_operations {
struct tty_driver {
struct kref kref;
struct cdev **cdevs;
+ struct mutex device_mutex;
struct module *owner;
const char *driver_name;
const char *name;
--
2.43.0
reply other threads:[~2026-07-31 14:22 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
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=20260731142220.3000214-1-nicoyip.dev@gmail.com \
--to=nicoyip.dev@gmail.com \
--cc=gregkh@linuxfoundation.org \
--cc=jirislaby@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-serial@vger.kernel.org \
--cc=rrw@kynesim.co.uk \
--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;
as well as URLs for NNTP newsgroup(s).