[v2] Input: powermate - fix use-after-free in powermate_config_complete

Message ID 20230916-topic-powermate_use_after_free-v2-1-505f49019f2f@gmail.com
State New
Headers
Series [v2] Input: powermate - fix use-after-free in powermate_config_complete |

Commit Message

Javier Carrasco Oct. 1, 2023, 3:35 p.m. UTC
  syzbot has found a use-after-free bug [1] in the powermate driver. This
happens when the device is disconnected, which leads to a memory free
from the powermate_device struct.
When an asynchronous control message completes after the kfree and its
callback is invoked, the lock does not exist anymore and hence the bug.

Use usb_kill_urb() on pm->config to cancel any in-progress requests upon
device disconnection. Given that this action is already done on pm->irq,
reorder the code to have both calls after the call to
input_unregister_device(), which is the most common approach.

[1] https://syzkaller.appspot.com/bug?extid=0434ac83f907a1dbdd1e

Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Reported-by: syzbot+0434ac83f907a1dbdd1e@syzkaller.appspotmail.com
---
Changes in v2:
- Use usb_kill_urb() on pm->config upon device disconnection.
- Link to v1: https://lore.kernel.org/r/20230916-topic-powermate_use_after_free-v1-1-2ffa46652869@gmail.com
---
 drivers/input/misc/powermate.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)


---
base-commit: cefc06e4de1477dbdc3cb2a91d4b1873b7797a5c
change-id: 20230916-topic-powermate_use_after_free-c703c7969c91

Best regards,
  

Comments

Dmitry Torokhov Oct. 4, 2023, 2:30 p.m. UTC | #1
Hi Javier,

On Sun, Oct 01, 2023 at 05:35:54PM +0200, Javier Carrasco wrote:
> syzbot has found a use-after-free bug [1] in the powermate driver. This
> happens when the device is disconnected, which leads to a memory free
> from the powermate_device struct.
> When an asynchronous control message completes after the kfree and its
> callback is invoked, the lock does not exist anymore and hence the bug.
> 
> Use usb_kill_urb() on pm->config to cancel any in-progress requests upon
> device disconnection. Given that this action is already done on pm->irq,
> reorder the code to have both calls after the call to
> input_unregister_device(), which is the most common approach.
> 
> [1] https://syzkaller.appspot.com/bug?extid=0434ac83f907a1dbdd1e
> 
> Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
> Reported-by: syzbot+0434ac83f907a1dbdd1e@syzkaller.appspotmail.com
> ---
> Changes in v2:
> - Use usb_kill_urb() on pm->config upon device disconnection.
> - Link to v1: https://lore.kernel.org/r/20230916-topic-powermate_use_after_free-v1-1-2ffa46652869@gmail.com
> ---
>  drivers/input/misc/powermate.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/input/misc/powermate.c b/drivers/input/misc/powermate.c
> index c1c733a9cb89..c0aea75eb087 100644
> --- a/drivers/input/misc/powermate.c
> +++ b/drivers/input/misc/powermate.c
> @@ -423,8 +423,9 @@ static void powermate_disconnect(struct usb_interface *intf)
>  	usb_set_intfdata(intf, NULL);
>  	if (pm) {
>  		pm->requires_update = 0;
> -		usb_kill_urb(pm->irq);

No, you do not want interrupts coming in while input device is being
destroyed, this one needs to stay where it is.

>  		input_unregister_device(pm->input);
> +		usb_kill_urb(pm->irq);
> +		usb_kill_urb(pm->config);
>  		usb_free_urb(pm->irq);
>  		usb_free_urb(pm->config);
>  		powermate_free_buffers(interface_to_usbdev(intf), pm);

Thanks.
  

Patch

diff --git a/drivers/input/misc/powermate.c b/drivers/input/misc/powermate.c
index c1c733a9cb89..c0aea75eb087 100644
--- a/drivers/input/misc/powermate.c
+++ b/drivers/input/misc/powermate.c
@@ -423,8 +423,9 @@  static void powermate_disconnect(struct usb_interface *intf)
 	usb_set_intfdata(intf, NULL);
 	if (pm) {
 		pm->requires_update = 0;
-		usb_kill_urb(pm->irq);
 		input_unregister_device(pm->input);
+		usb_kill_urb(pm->irq);
+		usb_kill_urb(pm->config);
 		usb_free_urb(pm->irq);
 		usb_free_urb(pm->config);
 		powermate_free_buffers(interface_to_usbdev(intf), pm);