Re: bluetooth: use-after-free in vhci_send_frame

[Jiri Slaby <jslaby@suse.cz>]

On 03/08/2016, 07:32 PM, Marcel Holtmann wrote:
> that means very little to me actually. So is the real issue caused by opening /dev/vhci or is that theoretical one via some internal kernel compile time feature.

Hi, what do you think about this one?

--- a/drivers/bluetooth/hci_vhci.c
+++ b/drivers/bluetooth/hci_vhci.c
@@ -196,6 +196,11 @@ static inline ssize_t vhci_get_user(stru

                cancel_delayed_work_sync(&data->open_timeout);

+               if (data->hdev) {
+                       kfree_skb(skb);
+                       return -EBADFD;
+               }
+
                opcode = *((__u8 *) skb->data);
                skb_pull(skb, 1);


open_timeout could be in progress (raced with us) and _sync cancel
waited for vhci_create_device to actually finish and create the device
the second time.

thanks,
-- 
js
suse labs