[PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[Bartosz Golaszewski]

From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

Linus Torvalds pointed out that using trylock here is wrong. This iteration
drops it in favor of unconditional locks but keeps all the fixes that came
later.

I will also not send it for this release but make it part of the updates PR
for v6.2 to give it some time in next.

v7 -> v8:
- don't use down_read_trylock(), just go straight for a full lock

v6 -> v7:
- fix a build issue with CDEV_V1 code disabled (giving credit to Nick Hainke)
- protect the gdev->chip also in gpio_chrdev_open()

v5 -> v6:
- signal an error in poll callbacks instead of returning 0 which would make
  the user-space assume a timeout occurred (which could lead to user-space
  spinning a timeout loop forever)

v4 -> v5:
- try to acquire the semaphore for reading and bail out of syscall callbacks
  immediately in case of lock contention

v3 -> v4:
- use function typedefs to make code cleaner
- add a blank line after down_write()

v2 -> v3:
- drop the helper variable in patch 1/2 as we won't be using it in 2/2
- refactor patch 2/2 to use locking wrappers around the syscall callbacks

v1 -> v2:
- add missing gdev->chip checks in patch 1/2
- add a second patch that protects the structures that can be accessed
  by user-space calls against concurrent removal

Bartosz Golaszewski (2):
  gpiolib: cdev: fix NULL-pointer dereferences
  gpiolib: protect the GPIO device against being dropped while in use by
    user-space

 drivers/gpio/gpiolib-cdev.c | 204 +++++++++++++++++++++++++++++++-----
 drivers/gpio/gpiolib.c      |   4 +
 drivers/gpio/gpiolib.h      |   5 +
 3 files changed, 188 insertions(+), 25 deletions(-)

-- 
2.37.2

[PATCH v8 1/2] gpiolib: cdev: fix NULL-pointer dereferences

[Bartosz Golaszewski]

From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

There are several places where we can crash the kernel by requesting
lines, unbinding the GPIO device, then calling any of the system calls
relevant to the GPIO character device's annonymous file descriptors:
ioctl(), read(), poll().

While I observed it with the GPIO simulator, it will also happen for any
of the GPIO devices that can be hot-unplugged - for instance any HID GPIO
expander (e.g. CP2112).

This affects both v1 and v2 uAPI.

This fixes it partially by checking if gdev->chip is not NULL but it
doesn't entirely remedy the situation as we still have a race condition
in which another thread can remove the device after the check.

Fixes: d7c51b47ac11 ("gpio: userspace ABI for reading/writing GPIO lines")
Fixes: 3c0d9c635ae2 ("gpiolib: cdev: support GPIO_V2_GET_LINE_IOCTL and GPIO_V2_LINE_GET_VALUES_IOCTL")
Fixes: aad955842d1c ("gpiolib: cdev: support GPIO_V2_GET_LINEINFO_IOCTL and GPIO_V2_GET_LINEINFO_WATCH_IOCTL")
Fixes: a54756cb24ea ("gpiolib: cdev: support GPIO_V2_LINE_SET_CONFIG_IOCTL")
Fixes: 7b8e00d98168 ("gpiolib: cdev: support GPIO_V2_LINE_SET_VALUES_IOCTL")
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
---
 drivers/gpio/gpiolib-cdev.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)

diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
index 0cb6b468f364..6fa5c2169985 100644
--- a/drivers/gpio/gpiolib-cdev.c
+++ b/drivers/gpio/gpiolib-cdev.c
@@ -201,6 +201,9 @@ static long linehandle_ioctl(struct file *file, unsigned int cmd,
 	unsigned int i;
 	int ret;
 
+	if (!lh->gdev->chip)
+		return -ENODEV;
+
 	switch (cmd) {
 	case GPIOHANDLE_GET_LINE_VALUES_IOCTL:
 		/* NOTE: It's okay to read values of output lines */
@@ -1384,6 +1387,9 @@ static long linereq_ioctl(struct file *file, unsigned int cmd,
 	struct linereq *lr = file->private_data;
 	void __user *ip = (void __user *)arg;
 
+	if (!lr->gdev->chip)
+		return -ENODEV;
+
 	switch (cmd) {
 	case GPIO_V2_LINE_GET_VALUES_IOCTL:
 		return linereq_get_values(lr, ip);
@@ -1410,6 +1416,9 @@ static __poll_t linereq_poll(struct file *file,
 	struct linereq *lr = file->private_data;
 	__poll_t events = 0;
 
+	if (!lr->gdev->chip)
+		return EPOLLHUP | EPOLLERR;
+
 	poll_wait(file, &lr->wait, wait);
 
 	if (!kfifo_is_empty_spinlocked_noirqsave(&lr->events,
@@ -1429,6 +1438,9 @@ static ssize_t linereq_read(struct file *file,
 	ssize_t bytes_read = 0;
 	int ret;
 
+	if (!lr->gdev->chip)
+		return -ENODEV;
+
 	if (count < sizeof(le))
 		return -EINVAL;
 
@@ -1716,6 +1728,9 @@ static __poll_t lineevent_poll(struct file *file,
 	struct lineevent_state *le = file->private_data;
 	__poll_t events = 0;
 
+	if (!le->gdev->chip)
+		return EPOLLHUP | EPOLLERR;
+
 	poll_wait(file, &le->wait, wait);
 
 	if (!kfifo_is_empty_spinlocked_noirqsave(&le->events, &le->wait.lock))
@@ -1740,6 +1755,9 @@ static ssize_t lineevent_read(struct file *file,
 	ssize_t ge_size;
 	int ret;
 
+	if (!le->gdev->chip)
+		return -ENODEV;
+
 	/*
 	 * When compatible system call is being used the struct gpioevent_data,
 	 * in case of at least ia32, has different size due to the alignment
@@ -1821,6 +1839,9 @@ static long lineevent_ioctl(struct file *file, unsigned int cmd,
 	void __user *ip = (void __user *)arg;
 	struct gpiohandle_data ghd;
 
+	if (!le->gdev->chip)
+		return -ENODEV;
+
 	/*
 	 * We can get the value for an event line but not set it,
 	 * because it is input by definition.
@@ -2407,6 +2428,9 @@ static __poll_t lineinfo_watch_poll(struct file *file,
 	struct gpio_chardev_data *cdev = file->private_data;
 	__poll_t events = 0;
 
+	if (!cdev->gdev->chip)
+		return EPOLLHUP | EPOLLERR;
+
 	poll_wait(file, &cdev->wait, pollt);
 
 	if (!kfifo_is_empty_spinlocked_noirqsave(&cdev->events,
@@ -2425,6 +2449,9 @@ static ssize_t lineinfo_watch_read(struct file *file, char __user *buf,
 	int ret;
 	size_t event_size;
 
+	if (!cdev->gdev->chip)
+		return -ENODEV;
+
 #ifndef CONFIG_GPIO_CDEV_V1
 	event_size = sizeof(struct gpio_v2_line_info_changed);
 	if (count < event_size)
-- 
2.37.2

[PATCH v8 2/2] gpiolib: protect the GPIO device against being dropped while in use by user-space

[Bartosz Golaszewski]

From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

While any of the GPIO cdev syscalls is in progress, the kernel can call
gpiochip_remove() (for instance, when a USB GPIO expander is disconnected)
which will set gdev->chip to NULL after which any subsequent access will
cause a crash.

To avoid that: use an RW-semaphore in which the syscalls take it for
reading (so that we don't needlessly prohibit the user-space from calling
syscalls simultaneously) while gpiochip_remove() takes it for writing so
that it can only happen once all syscalls return.

Fixes: d7c51b47ac11 ("gpio: userspace ABI for reading/writing GPIO lines")
Fixes: 3c0d9c635ae2 ("gpiolib: cdev: support GPIO_V2_GET_LINE_IOCTL and GPIO_V2_LINE_GET_VALUES_IOCTL")
Fixes: aad955842d1c ("gpiolib: cdev: support GPIO_V2_GET_LINEINFO_IOCTL and GPIO_V2_GET_LINEINFO_WATCH_IOCTL")
Fixes: a54756cb24ea ("gpiolib: cdev: support GPIO_V2_LINE_SET_CONFIG_IOCTL")
Fixes: 7b8e00d98168 ("gpiolib: cdev: support GPIO_V2_LINE_SET_VALUES_IOCTL")
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
[Nick: fixed a build failure with CDEV_V1 disabled]
Co-authored-by: Nick Hainke <vincent@systemli.org>
Reviewed-by: Kent Gibson <warthog618@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
---
 drivers/gpio/gpiolib-cdev.c | 177 +++++++++++++++++++++++++++++++-----
 drivers/gpio/gpiolib.c      |   4 +
 drivers/gpio/gpiolib.h      |   5 +
 3 files changed, 161 insertions(+), 25 deletions(-)

diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
index 6fa5c2169985..6ab1cf489d03 100644
--- a/drivers/gpio/gpiolib-cdev.c
+++ b/drivers/gpio/gpiolib-cdev.c
@@ -55,6 +55,50 @@ static_assert(IS_ALIGNED(sizeof(struct gpio_v2_line_values), 8));
  * interface to gpiolib GPIOs via ioctl()s.
  */
 
+typedef __poll_t (*poll_fn)(struct file *, struct poll_table_struct *);
+typedef long (*ioctl_fn)(struct file *, unsigned int, unsigned long);
+typedef ssize_t (*read_fn)(struct file *, char __user *,
+			   size_t count, loff_t *);
+
+static __poll_t call_poll_locked(struct file *file,
+				 struct poll_table_struct *wait,
+				 struct gpio_device *gdev, poll_fn func)
+{
+	__poll_t ret;
+
+	down_read(&gdev->sem);
+	ret = func(file, wait);
+	up_read(&gdev->sem);
+
+	return ret;
+}
+
+static long call_ioctl_locked(struct file *file, unsigned int cmd,
+			      unsigned long arg, struct gpio_device *gdev,
+			      ioctl_fn func)
+{
+	long ret;
+
+	down_read(&gdev->sem);
+	ret = func(file, cmd, arg);
+	up_read(&gdev->sem);
+
+	return ret;
+}
+
+static ssize_t call_read_locked(struct file *file, char __user *buf,
+				size_t count, loff_t *f_ps,
+				struct gpio_device *gdev, read_fn func)
+{
+	ssize_t ret;
+
+	down_read(&gdev->sem);
+	ret = func(file, buf, count, f_ps);
+	up_read(&gdev->sem);
+
+	return ret;
+}
+
 /*
  * GPIO line handle management
  */
@@ -191,8 +235,8 @@ static long linehandle_set_config(struct linehandle_state *lh,
 	return 0;
 }
 
-static long linehandle_ioctl(struct file *file, unsigned int cmd,
-			     unsigned long arg)
+static long linehandle_ioctl_unlocked(struct file *file, unsigned int cmd,
+				      unsigned long arg)
 {
 	struct linehandle_state *lh = file->private_data;
 	void __user *ip = (void __user *)arg;
@@ -250,6 +294,15 @@ static long linehandle_ioctl(struct file *file, unsigned int cmd,
 	}
 }
 
+static long linehandle_ioctl(struct file *file, unsigned int cmd,
+			     unsigned long arg)
+{
+	struct linehandle_state *lh = file->private_data;
+
+	return call_ioctl_locked(file, cmd, arg, lh->gdev,
+				 linehandle_ioctl_unlocked);
+}
+
 #ifdef CONFIG_COMPAT
 static long linehandle_ioctl_compat(struct file *file, unsigned int cmd,
 				    unsigned long arg)
@@ -1381,8 +1434,8 @@ static long linereq_set_config(struct linereq *lr, void __user *ip)
 	return ret;
 }
 
-static long linereq_ioctl(struct file *file, unsigned int cmd,
-			  unsigned long arg)
+static long linereq_ioctl_unlocked(struct file *file, unsigned int cmd,
+				   unsigned long arg)
 {
 	struct linereq *lr = file->private_data;
 	void __user *ip = (void __user *)arg;
@@ -1402,6 +1455,15 @@ static long linereq_ioctl(struct file *file, unsigned int cmd,
 	}
 }
 
+static long linereq_ioctl(struct file *file, unsigned int cmd,
+			  unsigned long arg)
+{
+	struct linereq *lr = file->private_data;
+
+	return call_ioctl_locked(file, cmd, arg, lr->gdev,
+				 linereq_ioctl_unlocked);
+}
+
 #ifdef CONFIG_COMPAT
 static long linereq_ioctl_compat(struct file *file, unsigned int cmd,
 				 unsigned long arg)
@@ -1410,8 +1472,8 @@ static long linereq_ioctl_compat(struct file *file, unsigned int cmd,
 }
 #endif
 
-static __poll_t linereq_poll(struct file *file,
-			    struct poll_table_struct *wait)
+static __poll_t linereq_poll_unlocked(struct file *file,
+				      struct poll_table_struct *wait)
 {
 	struct linereq *lr = file->private_data;
 	__poll_t events = 0;
@@ -1428,10 +1490,16 @@ static __poll_t linereq_poll(struct file *file,
 	return events;
 }
 
-static ssize_t linereq_read(struct file *file,
-			    char __user *buf,
-			    size_t count,
-			    loff_t *f_ps)
+static __poll_t linereq_poll(struct file *file,
+			     struct poll_table_struct *wait)
+{
+	struct linereq *lr = file->private_data;
+
+	return call_poll_locked(file, wait, lr->gdev, linereq_poll_unlocked);
+}
+
+static ssize_t linereq_read_unlocked(struct file *file, char __user *buf,
+				     size_t count, loff_t *f_ps)
 {
 	struct linereq *lr = file->private_data;
 	struct gpio_v2_line_event le;
@@ -1485,6 +1553,15 @@ static ssize_t linereq_read(struct file *file,
 	return bytes_read;
 }
 
+static ssize_t linereq_read(struct file *file, char __user *buf,
+			    size_t count, loff_t *f_ps)
+{
+	struct linereq *lr = file->private_data;
+
+	return call_read_locked(file, buf, count, f_ps, lr->gdev,
+				linereq_read_unlocked);
+}
+
 static void linereq_free(struct linereq *lr)
 {
 	unsigned int i;
@@ -1722,8 +1799,8 @@ struct lineevent_state {
 	(GPIOEVENT_REQUEST_RISING_EDGE | \
 	GPIOEVENT_REQUEST_FALLING_EDGE)
 
-static __poll_t lineevent_poll(struct file *file,
-			       struct poll_table_struct *wait)
+static __poll_t lineevent_poll_unlocked(struct file *file,
+					struct poll_table_struct *wait)
 {
 	struct lineevent_state *le = file->private_data;
 	__poll_t events = 0;
@@ -1739,15 +1816,21 @@ static __poll_t lineevent_poll(struct file *file,
 	return events;
 }
 
+static __poll_t lineevent_poll(struct file *file,
+			       struct poll_table_struct *wait)
+{
+	struct lineevent_state *le = file->private_data;
+
+	return call_poll_locked(file, wait, le->gdev, lineevent_poll_unlocked);
+}
+
 struct compat_gpioeevent_data {
 	compat_u64	timestamp;
 	u32		id;
 };
 
-static ssize_t lineevent_read(struct file *file,
-			      char __user *buf,
-			      size_t count,
-			      loff_t *f_ps)
+static ssize_t lineevent_read_unlocked(struct file *file, char __user *buf,
+				       size_t count, loff_t *f_ps)
 {
 	struct lineevent_state *le = file->private_data;
 	struct gpioevent_data ge;
@@ -1815,6 +1898,15 @@ static ssize_t lineevent_read(struct file *file,
 	return bytes_read;
 }
 
+static ssize_t lineevent_read(struct file *file, char __user *buf,
+			      size_t count, loff_t *f_ps)
+{
+	struct lineevent_state *le = file->private_data;
+
+	return call_read_locked(file, buf, count, f_ps, le->gdev,
+				lineevent_read_unlocked);
+}
+
 static void lineevent_free(struct lineevent_state *le)
 {
 	if (le->irq)
@@ -1832,8 +1924,8 @@ static int lineevent_release(struct inode *inode, struct file *file)
 	return 0;
 }
 
-static long lineevent_ioctl(struct file *file, unsigned int cmd,
-			    unsigned long arg)
+static long lineevent_ioctl_unlocked(struct file *file, unsigned int cmd,
+				     unsigned long arg)
 {
 	struct lineevent_state *le = file->private_data;
 	void __user *ip = (void __user *)arg;
@@ -1864,6 +1956,15 @@ static long lineevent_ioctl(struct file *file, unsigned int cmd,
 	return -EINVAL;
 }
 
+static long lineevent_ioctl(struct file *file, unsigned int cmd,
+			    unsigned long arg)
+{
+	struct lineevent_state *le = file->private_data;
+
+	return call_ioctl_locked(file, cmd, arg, le->gdev,
+				 lineevent_ioctl_unlocked);
+}
+
 #ifdef CONFIG_COMPAT
 static long lineevent_ioctl_compat(struct file *file, unsigned int cmd,
 				   unsigned long arg)
@@ -2422,8 +2523,8 @@ static int lineinfo_changed_notify(struct notifier_block *nb,
 	return NOTIFY_OK;
 }
 
-static __poll_t lineinfo_watch_poll(struct file *file,
-				    struct poll_table_struct *pollt)
+static __poll_t lineinfo_watch_poll_unlocked(struct file *file,
+					     struct poll_table_struct *pollt)
 {
 	struct gpio_chardev_data *cdev = file->private_data;
 	__poll_t events = 0;
@@ -2440,8 +2541,17 @@ static __poll_t lineinfo_watch_poll(struct file *file,
 	return events;
 }
 
-static ssize_t lineinfo_watch_read(struct file *file, char __user *buf,
-				   size_t count, loff_t *off)
+static __poll_t lineinfo_watch_poll(struct file *file,
+				    struct poll_table_struct *pollt)
+{
+	struct gpio_chardev_data *cdev = file->private_data;
+
+	return call_poll_locked(file, pollt, cdev->gdev,
+				lineinfo_watch_poll_unlocked);
+}
+
+static ssize_t lineinfo_watch_read_unlocked(struct file *file, char __user *buf,
+					    size_t count, loff_t *off)
 {
 	struct gpio_chardev_data *cdev = file->private_data;
 	struct gpio_v2_line_info_changed event;
@@ -2519,6 +2629,15 @@ static ssize_t lineinfo_watch_read(struct file *file, char __user *buf,
 	return bytes_read;
 }
 
+static ssize_t lineinfo_watch_read(struct file *file, char __user *buf,
+				   size_t count, loff_t *off)
+{
+	struct gpio_chardev_data *cdev = file->private_data;
+
+	return call_read_locked(file, buf, count, off, cdev->gdev,
+				lineinfo_watch_read_unlocked);
+}
+
 /**
  * gpio_chrdev_open() - open the chardev for ioctl operations
  * @inode: inode for this chardev
@@ -2532,13 +2651,17 @@ static int gpio_chrdev_open(struct inode *inode, struct file *file)
 	struct gpio_chardev_data *cdev;
 	int ret = -ENOMEM;
 
+	down_read(&gdev->sem);
+
 	/* Fail on open if the backing gpiochip is gone */
-	if (!gdev->chip)
-		return -ENODEV;
+	if (!gdev->chip) {
+		ret = -ENODEV;
+		goto out_unlock;
+	}
 
 	cdev = kzalloc(sizeof(*cdev), GFP_KERNEL);
 	if (!cdev)
-		return -ENOMEM;
+		goto out_unlock;
 
 	cdev->watched_lines = bitmap_zalloc(gdev->chip->ngpio, GFP_KERNEL);
 	if (!cdev->watched_lines)
@@ -2561,6 +2684,8 @@ static int gpio_chrdev_open(struct inode *inode, struct file *file)
 	if (ret)
 		goto out_unregister_notifier;
 
+	up_read(&gdev->sem);
+
 	return ret;
 
 out_unregister_notifier:
@@ -2570,6 +2695,8 @@ static int gpio_chrdev_open(struct inode *inode, struct file *file)
 	bitmap_free(cdev->watched_lines);
 out_free_cdev:
 	kfree(cdev);
+out_unlock:
+	up_read(&gdev->sem);
 	return ret;
 }
 
diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index 4756ea08894f..e0e73bd756ca 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -731,6 +731,7 @@ int gpiochip_add_data_with_key(struct gpio_chip *gc, void *data,
 	spin_unlock_irqrestore(&gpio_lock, flags);
 
 	BLOCKING_INIT_NOTIFIER_HEAD(&gdev->notifier);
+	init_rwsem(&gdev->sem);
 
 #ifdef CONFIG_PINCTRL
 	INIT_LIST_HEAD(&gdev->pin_ranges);
@@ -865,6 +866,8 @@ void gpiochip_remove(struct gpio_chip *gc)
 	unsigned long	flags;
 	unsigned int	i;
 
+	down_write(&gdev->sem);
+
 	/* FIXME: should the legacy sysfs handling be moved to gpio_device? */
 	gpiochip_sysfs_unregister(gdev);
 	gpiochip_free_hogs(gc);
@@ -899,6 +902,7 @@ void gpiochip_remove(struct gpio_chip *gc)
 	 * gone.
 	 */
 	gcdev_unregister(gdev);
+	up_write(&gdev->sem);
 	put_device(&gdev->dev);
 }
 EXPORT_SYMBOL_GPL(gpiochip_remove);
diff --git a/drivers/gpio/gpiolib.h b/drivers/gpio/gpiolib.h
index d900ecdbac46..9ad68a0adf4a 100644
--- a/drivers/gpio/gpiolib.h
+++ b/drivers/gpio/gpiolib.h
@@ -15,6 +15,7 @@
 #include <linux/device.h>
 #include <linux/module.h>
 #include <linux/cdev.h>
+#include <linux/rwsem.h>
 
 #define GPIOCHIP_NAME	"gpiochip"
 
@@ -39,6 +40,9 @@
  * @list: links gpio_device:s together for traversal
  * @notifier: used to notify subscribers about lines being requested, released
  *            or reconfigured
+ * @sem: protects the structure from a NULL-pointer dereference of @chip by
+ *       user-space operations when the device gets unregistered during
+ *       a hot-unplug event
  * @pin_ranges: range of pins served by the GPIO driver
  *
  * This state container holds most of the runtime variable data
@@ -60,6 +64,7 @@ struct gpio_device {
 	void			*data;
 	struct list_head        list;
 	struct blocking_notifier_head notifier;
+	struct rw_semaphore	sem;
 
 #ifdef CONFIG_PINCTRL
 	/*
-- 
2.37.2

Re: [PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[Andy Shevchenko]

On Mon, Dec 05, 2022 at 01:39:01PM +0100, Bartosz Golaszewski wrote:
> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> 
> Linus Torvalds pointed out that using trylock here is wrong. This iteration
> drops it in favor of unconditional locks but keeps all the fixes that came
> later.
> 
> I will also not send it for this release but make it part of the updates PR
> for v6.2 to give it some time in next.
> 
> v7 -> v8:
> - don't use down_read_trylock(), just go straight for a full lock

Yep, it was a good wrap-up explanation.

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[Andy Shevchenko]

On Mon, Dec 05, 2022 at 02:59:42PM +0200, Andy Shevchenko wrote:
> On Mon, Dec 05, 2022 at 01:39:01PM +0100, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> > 
> > Linus Torvalds pointed out that using trylock here is wrong. This iteration
> > drops it in favor of unconditional locks but keeps all the fixes that came
> > later.
> > 
> > I will also not send it for this release but make it part of the updates PR
> > for v6.2 to give it some time in next.
> > 
> > v7 -> v8:
> > - don't use down_read_trylock(), just go straight for a full lock
> 
> Yep, it was a good wrap-up explanation.

But he also suggested to fold NULL check into call_*_locked(). Any points why
you decided not to go that way?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[Bartosz Golaszewski]

On Mon, Dec 5, 2022 at 2:01 PM Andy Shevchenko
<andriy.shevchenko@linux.intel.com> wrote:
>
> On Mon, Dec 05, 2022 at 02:59:42PM +0200, Andy Shevchenko wrote:
> > On Mon, Dec 05, 2022 at 01:39:01PM +0100, Bartosz Golaszewski wrote:
> > > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> > >
> > > Linus Torvalds pointed out that using trylock here is wrong. This iteration
> > > drops it in favor of unconditional locks but keeps all the fixes that came
> > > later.
> > >
> > > I will also not send it for this release but make it part of the updates PR
> > > for v6.2 to give it some time in next.
> > >
> > > v7 -> v8:
> > > - don't use down_read_trylock(), just go straight for a full lock
> >
> > Yep, it was a good wrap-up explanation.
>
> But he also suggested to fold NULL check into call_*_locked(). Any points why
> you decided not to go that way?
>

Yes! Every read(), ioctl() and poll() variant extract a different
structure from file->private_data. I'm afraid we need to keep those in
the callbacks down the stack. I too thought it was a good idea but
that won't work.

Bart

Re: [PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[Kent Gibson]

On Mon, Dec 05, 2022 at 03:01:13PM +0200, Andy Shevchenko wrote:
> On Mon, Dec 05, 2022 at 02:59:42PM +0200, Andy Shevchenko wrote:
> > On Mon, Dec 05, 2022 at 01:39:01PM +0100, Bartosz Golaszewski wrote:
> > > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> > > 
> > > Linus Torvalds pointed out that using trylock here is wrong. This iteration
> > > drops it in favor of unconditional locks but keeps all the fixes that came
> > > later.
> > > 
> > > I will also not send it for this release but make it part of the updates PR
> > > for v6.2 to give it some time in next.
> > > 
> > > v7 -> v8:
> > > - don't use down_read_trylock(), just go straight for a full lock
> > 
> > Yep, it was a good wrap-up explanation.
> 
> But he also suggested to fold NULL check into call_*_locked(). Any points why
> you decided not to go that way?
> 

He also mentioned making it more back-portable.
Does that mean splitting out the patches for uAPI v1 and v2, if not for
each of the Fixes? Or does he mean back-porting it to 6.1?

Cheers,
Kent.

Re: [PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[Bartosz Golaszewski]

On Mon, Dec 5, 2022 at 2:17 PM Kent Gibson <warthog618@gmail.com> wrote:
>
> On Mon, Dec 05, 2022 at 03:01:13PM +0200, Andy Shevchenko wrote:
> > On Mon, Dec 05, 2022 at 02:59:42PM +0200, Andy Shevchenko wrote:
> > > On Mon, Dec 05, 2022 at 01:39:01PM +0100, Bartosz Golaszewski wrote:
> > > > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> > > >
> > > > Linus Torvalds pointed out that using trylock here is wrong. This iteration
> > > > drops it in favor of unconditional locks but keeps all the fixes that came
> > > > later.
> > > >
> > > > I will also not send it for this release but make it part of the updates PR
> > > > for v6.2 to give it some time in next.
> > > >
> > > > v7 -> v8:
> > > > - don't use down_read_trylock(), just go straight for a full lock
> > >
> > > Yep, it was a good wrap-up explanation.
> >
> > But he also suggested to fold NULL check into call_*_locked(). Any points why
> > you decided not to go that way?
> >
>
> He also mentioned making it more back-portable.
> Does that mean splitting out the patches for uAPI v1 and v2, if not for
> each of the Fixes? Or does he mean back-porting it to 6.1?

I think he mentioned not sending it for v6.1 that late in the cycle
and instead sending it for v6.2 and backporting it to v6.1. I'm afraid
that backporting of this fix will require some manual labor anyway.
We'll want to backport it to branches that don't have v2 yet after
all, where that code resided elsewhere.

Bartosz

Re: [PATCH v8 0/2] gpiolib: don't allow user-space to crash the kernel with hot-unplugs

[Bartosz Golaszewski]

On Mon, Dec 5, 2022 at 1:39 PM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
>
> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
>
> Linus Torvalds pointed out that using trylock here is wrong. This iteration
> drops it in favor of unconditional locks but keeps all the fixes that came
> later.
>
> I will also not send it for this release but make it part of the updates PR
> for v6.2 to give it some time in next.
>
> v7 -> v8:
> - don't use down_read_trylock(), just go straight for a full lock
>
> v6 -> v7:
> - fix a build issue with CDEV_V1 code disabled (giving credit to Nick Hainke)
> - protect the gdev->chip also in gpio_chrdev_open()
>
> v5 -> v6:
> - signal an error in poll callbacks instead of returning 0 which would make
>   the user-space assume a timeout occurred (which could lead to user-space
>   spinning a timeout loop forever)
>
> v4 -> v5:
> - try to acquire the semaphore for reading and bail out of syscall callbacks
>   immediately in case of lock contention
>
> v3 -> v4:
> - use function typedefs to make code cleaner
> - add a blank line after down_write()
>
> v2 -> v3:
> - drop the helper variable in patch 1/2 as we won't be using it in 2/2
> - refactor patch 2/2 to use locking wrappers around the syscall callbacks
>
> v1 -> v2:
> - add missing gdev->chip checks in patch 1/2
> - add a second patch that protects the structures that can be accessed
>   by user-space calls against concurrent removal
>
> Bartosz Golaszewski (2):
>   gpiolib: cdev: fix NULL-pointer dereferences
>   gpiolib: protect the GPIO device against being dropped while in use by
>     user-space
>
>  drivers/gpio/gpiolib-cdev.c | 204 +++++++++++++++++++++++++++++++-----
>  drivers/gpio/gpiolib.c      |   4 +
>  drivers/gpio/gpiolib.h      |   5 +
>  3 files changed, 188 insertions(+), 25 deletions(-)
>
> --
> 2.37.2
>

Reapplied again to get it back into next for testing.

Bart