Skip to contentSkip to conversation

Input: goodix - Fix double free on managed resource

PATCH In review José Expósito 2 participants 7 messages Apr 22, 2022 – Apr 23, 2022

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH 1/3 Input: goodix - Fix double free on managed resource

As described in the documentation for devm_input_allocate_device():

Managed input devices do not need to be explicitly unregistered or freed as it will be done automatically when owner device unbinds from its driver (or binding fails).

However this driver was explicitly freeing the input device.

Remove the calls to input_free_device() to avoid a possible double free error.

Fixes 5ede7f0cfb93f ("Input: goodix - add pen support")
---
 drivers/input/touchscreen/goodix.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
drivers/input/touchscreen/goodix.c +2 -6
index 752e8ba4fecb..61eb69f3a259 100644
@@ -308,10 +308,8 @@ static struct input_dev *goodix_create_pen_input(struct goodix_ts_data *ts)
 		return NULL;
 
 	input_alloc_absinfo(input);
-	if (!input->absinfo) {
-		input_free_device(input);
+	if (!input->absinfo)
 		return NULL;
-	}
 
 	input->absinfo[ABS_X] = ts->input_dev->absinfo[ABS_MT_POSITION_X];
 	input->absinfo[ABS_Y] = ts->input_dev->absinfo[ABS_MT_POSITION_Y];
@@ -340,10 +338,8 @@ static struct input_dev *goodix_create_pen_input(struct goodix_ts_data *ts)
 		input->id.product = 0x1001;
 	input->id.version = ts->version;
 
-	if (input_register_device(input) != 0) {
-		input_free_device(input);
+	if (input_register_device(input) != 0)
 		return NULL;
-	}
 
 	return input;
 }
Signature · 1 line
2.25.1
#1 lore raw

PATCH 2/3 HID: logitech-hidpp: Fix double free on managed resource

As described in the documentation for devm_input_allocate_device():

Managed input devices do not need to be explicitly unregistered or freed as it will be done automatically when owner device unbinds from its driver (or binding fails).

However this driver was explicitly freeing the input device, allocated using devm_input_allocate_device() through hidpp_allocate_input().

Remove the call to input_free_device() to avoid a possible double free error.

Fixes c39e3d5fc9dd3 ("HID: logitech-hidpp: late bind the input device on wireless connection")
---
 drivers/hid/hid-logitech-hidpp.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)
drivers/hid/hid-logitech-hidpp.c +1 -5
index 81de88ab2ecc..9c00a781ab57 100644
@@ -3957,11 +3957,7 @@ static void hidpp_connect_event(struct hidpp_device *hidpp)
 	}
 
 	hidpp_populate_input(hidpp, input);
-
-	ret = input_register_device(input);
-	if (ret)
-		input_free_device(input);
-
+	input_register_device(input);
 	hidpp->delayed_input = input;
 }
 
Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH 2/3] HID: logitech-hidpp: Fix double free on managed resource

Hi,

Quoted 11 lines · José Expósito

As described in the documentation for devm_input_allocate_device():

Managed input devices do not need to be explicitly unregistered or freed as it will be done automatically when owner device unbinds from its driver (or binding fails).

However this driver was explicitly freeing the input device, allocated using devm_input_allocate_device() through hidpp_allocate_input().

Remove the call to input_free_device() to avoid a possible double free error.

Actually calling input_free_device() on a devm allocated input device is fine. The input subsystem has chosen to not have a separate devm_input_free_device(), instead input_free_device() knows if a device is allocated through devm and then also frees the devres tied to it:

void input_free_device(struct input_dev *dev)

{
        if (dev) {
                if (dev->devres_managed)
                        WARN_ON(devres_destroy(dev->dev.parent,
                                                devm_input_device_release,
                                                devm_input_device_match,
                                                dev));
                input_put_device(dev);
        }
}
Quoted 12 lines
Fixes c39e3d5fc9dd3 ("HID: logitech-hidpp: late bind the input device on wireless connection")
---
 drivers/hid/hid-logitech-hidpp.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)
drivers/hid/hid-logitech-hidpp.c +0 -0
index 81de88ab2ecc..9c00a781ab57 100644

drivers/hid/hid-logitech-hidpp.c static void hidpp_connect_event(struct hidpp_device *hidpp) · line 3964

Quoted 8 lines
drivers/hid/hid-logitech-hidpp.c +0 -5
@@ -3957,11 +3957,7 @@ static void hidpp_connect_event(struct hidpp_device *hidpp)
 	}
 
 	hidpp_populate_input(hidpp, input);
-
-	ret = input_register_device(input);
-	if (ret)
-		input_free_device(input);
-

The original code does look wrong there though, since the input device is free-ed it should not be stored in hidpp->delayed_input, so this should be comes:

	ret = input_register_device(input);
	if (ret) {
		input_free_device(input);
		return;
	}

Regards,

Hans

Quoted 3 lines
drivers/hid/hid-logitech-hidpp.c +0 -0
 	hidpp->delayed_input = input;
 }
 
#3 lore raw

Re: [PATCH 2/3] HID: logitech-hidpp: Fix double free on managed resource

Quoted 22 lines · Hans de Goede

Hi,

Quoted 11 lines · José Expósito

As described in the documentation for devm_input_allocate_device():

Managed input devices do not need to be explicitly unregistered or freed as it will be done automatically when owner device unbinds from its driver (or binding fails).

However this driver was explicitly freeing the input device, allocated using devm_input_allocate_device() through hidpp_allocate_input().

Remove the call to input_free_device() to avoid a possible double free error.

Actually calling input_free_device() on a devm allocated input device is fine. The input subsystem has chosen to not have a separate devm_input_free_device(), instead input_free_device() knows if a device is allocated through devm and then also frees the devres tied to it:

void input_free_device(struct input_dev *dev)

Quoted 10 lines
{
        if (dev) {
                if (dev->devres_managed)
                        WARN_ON(devres_destroy(dev->dev.parent,
                                                devm_input_device_release,
                                                devm_input_device_match,
                                                dev));
                input_put_device(dev);
        }
}

Hi Hans,

Thanks for the code review.

Obviously, I completely misunderstood these functions, my bad. Thanks for the explanation.

Please ignore the patchset.

Jose

Quoted 41 lines
Quoted 12 lines
Fixes c39e3d5fc9dd3 ("HID: logitech-hidpp: late bind the input device on wireless connection")
---
 drivers/hid/hid-logitech-hidpp.c | 6 +-----
 1 file changed, 1 insertion(+), 5 deletions(-)
drivers/hid/hid-logitech-hidpp.c +0 -0
index 81de88ab2ecc..9c00a781ab57 100644

drivers/hid/hid-logitech-hidpp.c static void hidpp_connect_event(struct hidpp_device *hidpp) · line 3964

Quoted 8 lines
drivers/hid/hid-logitech-hidpp.c +0 -5
@@ -3957,11 +3957,7 @@ static void hidpp_connect_event(struct hidpp_device *hidpp)
 	}
 
 	hidpp_populate_input(hidpp, input);
-
-	ret = input_register_device(input);
-	if (ret)
-		input_free_device(input);
-

The original code does look wrong there though, since the input device is free-ed it should not be stored in hidpp->delayed_input, so this should be comes:

	ret = input_register_device(input);
	if (ret) {
		input_free_device(input);
		return;
	}

Regards,

Hans

Quoted 3 lines
drivers/hid/hid-logitech-hidpp.c +0 -0
 	hidpp->delayed_input = input;
 }
 
#4 lore raw

PATCH 3/3 HID: wacom: Fix double free on managed resource

In reply to José Expósito

As described in the documentation for devm_input_allocate_device():

Managed input devices do not need to be explicitly unregistered or freed as it will be done automatically when owner device unbinds from its driver (or binding fails).

However this driver was explicitly freeing the input device, allocated using devm_input_allocate_device() through wacom_allocate_input().

Remove the calls to input_free_device() to avoid a possible double free error.

Fixes d2d13f18aaa51 ("Input: wacom - create a separate input device for pads")
---
 drivers/hid/wacom_sys.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/hid/wacom_sys.c +0 -3
index 066c567dbaa2..164c0f7cb796 100644
@@ -2098,7 +2098,6 @@ static int wacom_register_inputs(struct wacom *wacom)
 	error = wacom_setup_pen_input_capabilities(pen_input_dev, wacom_wac);
 	if (error) {
 		/* no pen in use on this interface */
-		input_free_device(pen_input_dev);
 		wacom_wac->pen_input = NULL;
 		pen_input_dev = NULL;
 	} else {
@@ -2110,7 +2109,6 @@ static int wacom_register_inputs(struct wacom *wacom)
 	error = wacom_setup_touch_input_capabilities(touch_input_dev, wacom_wac);
 	if (error) {
 		/* no touch in use on this interface */
-		input_free_device(touch_input_dev);
 		wacom_wac->touch_input = NULL;
 		touch_input_dev = NULL;
 	} else {
@@ -2122,7 +2120,6 @@ static int wacom_register_inputs(struct wacom *wacom)
 	error = wacom_setup_pad_input_capabilities(pad_input_dev, wacom_wac);
 	if (error) {
 		/* no pad in use on this interface */
-		input_free_device(pad_input_dev);
 		wacom_wac->pad_input = NULL;
 		pad_input_dev = NULL;
 	} else {
Signature · 1 line
2.25.1
#5 lore raw

Re: [PATCH 3/3] HID: wacom: Fix double free on managed resource

Hi,

Quoted 11 lines · José Expósito

As described in the documentation for devm_input_allocate_device():

Managed input devices do not need to be explicitly unregistered or freed as it will be done automatically when owner device unbinds from its driver (or binding fails).

However this driver was explicitly freeing the input device, allocated using devm_input_allocate_device() through wacom_allocate_input().

Remove the calls to input_free_device() to avoid a possible double free error.

Actually calling input_free_device() on a devm allocated input device is fine. The input subsystem has chosen to not have a separate devm_input_free_device(), instead input_free_device() knows if a device is allocated through devm and then also frees the devres tied to it:

void input_free_device(struct input_dev *dev)

{
        if (dev) {
                if (dev->devres_managed)
                        WARN_ON(devres_destroy(dev->dev.parent,
                                                devm_input_device_release,
                                                devm_input_device_match,
                                                dev));
                input_put_device(dev);
        }
}

So there is no need for this patch.

Regards,

Hans

Quoted 35 lines
Fixes d2d13f18aaa51 ("Input: wacom - create a separate input device for pads")
---
 drivers/hid/wacom_sys.c | 3 ---
 1 file changed, 3 deletions(-)
drivers/hid/wacom_sys.c +0 -3
index 066c567dbaa2..164c0f7cb796 100644
@@ -2098,7 +2098,6 @@ static int wacom_register_inputs(struct wacom *wacom)
 	error = wacom_setup_pen_input_capabilities(pen_input_dev, wacom_wac);
 	if (error) {
 		/* no pen in use on this interface */
-		input_free_device(pen_input_dev);
 		wacom_wac->pen_input = NULL;
 		pen_input_dev = NULL;
 	} else {
@@ -2110,7 +2109,6 @@ static int wacom_register_inputs(struct wacom *wacom)
 	error = wacom_setup_touch_input_capabilities(touch_input_dev, wacom_wac);
 	if (error) {
 		/* no touch in use on this interface */
-		input_free_device(touch_input_dev);
 		wacom_wac->touch_input = NULL;
 		touch_input_dev = NULL;
 	} else {
@@ -2122,7 +2120,6 @@ static int wacom_register_inputs(struct wacom *wacom)
 	error = wacom_setup_pad_input_capabilities(pad_input_dev, wacom_wac);
 	if (error) {
 		/* no pad in use on this interface */
-		input_free_device(pad_input_dev);
 		wacom_wac->pad_input = NULL;
 		pad_input_dev = NULL;
 	} else {
#6 lore raw
In reply to José Expósito

Hi José,

Quoted 23 lines · José Expósito

As described in the documentation for devm_input_allocate_device():

Managed input devices do not need to be explicitly unregistered or freed as it will be done automatically when owner device unbinds from its driver (or binding fails).

However this driver was explicitly freeing the input device.

Remove the calls to input_free_device() to avoid a possible double free error.

Fixes 5ede7f0cfb93f ("Input: goodix - add pen support")
---
 drivers/input/touchscreen/goodix.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
drivers/input/touchscreen/goodix.c +0 -0
index 752e8ba4fecb..61eb69f3a259 100644
@@ -308,10 +308,8 @@ static struct input_dev *goodix_create_pen_input(struct goodix_ts_data *ts)
 		return NULL;

drivers/input/touchscreen/goodix.c static struct input_dev *goodix_create_pen_input(struct goodix_ts_data *ts) · line 315

Quoted 10 lines
drivers/input/touchscreen/goodix.c +1 -3
 
 	input_alloc_absinfo(input);
-	if (!input->absinfo) {
-		input_free_device(input);
+	if (!input->absinfo)
 		return NULL;
-	}
 
 	input->absinfo[ABS_X] = ts->input_dev->absinfo[ABS_MT_POSITION_X];
 	input->absinfo[ABS_Y] = ts->input_dev->absinfo[ABS_MT_POSITION_Y];

I don't know what tree you've based this on, but the above code has been replaced with the new input_copy_abs helper in Linus' current master branch:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/input/touchscreen/goodix.c#n310

drivers/input/touchscreen/goodix.c static struct input_dev *goodix_create_pen_input(struct goodix_ts_data *ts) · line 345

Quoted 12 lines
drivers/input/touchscreen/goodix.c +1 -3
@@ -340,10 +338,8 @@ static struct input_dev *goodix_create_pen_input(struct goodix_ts_data *ts)
 		input->id.product = 0x1001;
 	input->id.version = ts->version;
 
-	if (input_register_device(input) != 0) {
-		input_free_device(input);
+	if (input_register_device(input) != 0)
 		return NULL;
-	}
 
 	return input;
 }

And this code has also already been fixed, so this patch can be dropped.

Regards,

Hans