Hello Marcel Holtmann,
The patch cdc52faac5f3: "Bluetooth: Fix memory leaking when
hdev->send returns an error" from Jul 6, 2014, leads to the following
static checker warning:
net/bluetooth/hci_core.c:3385 hci_send_frame()
warn: 'skb' was already freed.
net/bluetooth/hci_core.c
3377 if (!test_bit(HCI_RUNNING, &hdev->flags)) {
3378 kfree_skb(skb);
3379 return;
3380 }
3381
3382 err = hdev->send(hdev, skb);
3383 if (err < 0) {
3384 BT_ERR("%s sending frame failed (%d)", hdev->name, err);
3385 kfree_skb(skb);
3386 }
The ti_st_send_frame() frees skb on error. I'm surprised this bug
wasn't found by KAsan when we found acf91ec384dd ("Bluetooth: btwilink:
Save the packet type before sending").
I don't totally understand how skb is freed on the success path either.
bfusb_send_frame(), dtl1_hci_send_frame() and btqcomsmd_send() have
calls to kfree_skb() but I can't find the calls in bpa10x_send_frame()
or the other ->send functions.
regards,
dan carpenter
Hi Dan,
On 23/05/2017 07:34, Dan Carpenter wrote:
> The ti_st_send_frame() frees skb on error. I'm surprised this bug
> wasn't found by KAsan when we found acf91ec384dd ("Bluetooth: btwilink:
> Save the packet type before sending").
Maybe an unreachable condition in the current ti_st implementation.
But yes this needs to be fixed.
> I don't totally understand how skb is freed on the success path either.
> bfusb_send_frame(), dtl1_hci_send_frame() and btqcomsmd_send() have
> calls to kfree_skb() but I can't find the calls in bpa10x_send_frame()
> or the other ->send functions.
bpa10x_send_frame stores a skb reference when filling urb, skb is then freed
asynchronously in the bpa10x_tx_complete callback.
What are the other send functions ? bt skb is often enqueued and not
released
in the send function itself.
Regards,
Loic