syzbot is reporting that calling hci_release_dev() from hci_error_reset()
due to hci_dev_put() from hci_error_reset() can cause deadlock at
destroy_workqueue(), for hci_error_reset() is called from
hdev->req_workqueue which destroy_workqueue() needs to flush.
We need to make sure that hdev->{rx_work,cmd_work,tx_work} which are
queued into hdev->workqueue and hdev->{power_on,error_reset} which are
queued into hdev->req_workqueue are no longer running by the moment
destroy_workqueue(hdev->workqueue);
destroy_workqueue(hdev->req_workqueue);
are called from hci_release_dev().
Call cancel_work_sync() on these work items from hci_unregister_dev()
as soon as hdev->list is removed from hci_dev_list.
Reported-by: syzbot <syzbot+da0a9c9721e36db712e8@xxxxxxxxxxxxxxxxxxxxxxxxx>
Closes: https://syzkaller.appspot.com/bug?extid=da0a9c9721e36db712e8
Signed-off-by: Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx>
---
Completely untested. Please do tests with lockdep enabled before committing.
Maybe it is too early to cancel hdev->{rx_work,cmd_work,tx_work}.
Maybe there are more work items which should be canceled before
hci_unregister_dev() completes. I don't know...
net/bluetooth/hci_core.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/net/bluetooth/hci_core.c b/net/bluetooth/hci_core.c
index dd3b0f501018..dbbe5e2da210 100644
--- a/net/bluetooth/hci_core.c
+++ b/net/bluetooth/hci_core.c
@@ -2751,7 +2751,11 @@ void hci_unregister_dev(struct hci_dev *hdev)
list_del(&hdev->list);
write_unlock(&hci_dev_list_lock);
+ cancel_work_sync(&hdev->rx_work);
+ cancel_work_sync(&hdev->cmd_work);
+ cancel_work_sync(&hdev->tx_work);
cancel_work_sync(&hdev->power_on);
+ cancel_work_sync(&hdev->error_reset);
hci_cmd_sync_clear(hdev);
--
2.18.4
