Hi,
On Wed, Oct 19, 2022 at 12:31 AM Yang Yingliang
<yangyingliang@xxxxxxxxxx> wrote:
>
> If hci_register_suspend_notifier() returns error, the hdev
> need to be put and deleted and rfkill need be unregistered
> to avoid leaks.
>
> Fixes: 9952d90ea288 ("Bluetooth: Handle PM_SUSPEND_PREPARE and PM_POST_SUSPEND")
> Signed-off-by: Yang Yingliang <yangyingliang@xxxxxxxxxx>
> ---
> net/bluetooth/hci_core.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
> index 0540555b3704..03b9374f3a97 100644
> --- a/net/bluetooth/hci_core.c
> +++ b/net/bluetooth/hci_core.c
> @@ -2660,7 +2660,7 @@ int hci_register_dev(struct hci_dev *hdev)
>
> error = hci_register_suspend_notifier(hdev);
> if (error)
> - goto err_wqueue;
> + goto err_hdev;
Perhaps we could disregard the error here, or print a warning, as it
just means we won't be handing suspend requests which I guess is
better than having to fail to register which would render the
controller useless.
>
> queue_work(hdev->req_workqueue, &hdev->power_on);
>
> @@ -2669,6 +2669,11 @@ int hci_register_dev(struct hci_dev *hdev)
>
> return id;
>
> +err_hdev:
> + hci_dev_put(hdev);
> + if (hdev->rfkill)
> + rfkill_unregister(hdev->rfkill);
> + device_del(&hdev->dev);
> err_wqueue:
> debugfs_remove_recursive(hdev->debugfs);
> destroy_workqueue(hdev->workqueue);
> --
> 2.25.1
>
--
Luiz Augusto von Dentz
