diff mbox series

Bluetooth: Fix use-after-free read in hci_cmd_timeout()

Message ID 20221105190446.1324053-1-syoshida@redhat.com
State New
Headers
Series Bluetooth: Fix use-after-free read in hci_cmd_timeout() |

Commit Message

Shigeru Yoshida Nov. 5, 2022, 7:04 p.m. UTC 
  syzbot reported use-after-free in hci_cmd_timeout() [1].  The scenario
for the issue is as follows:

Task                                    Workqueue
----------------------------------------------------------------------
hci_dev_open_sync
  ...
  hci_dev_init_sync  <- failed
  ...
  if (hdev->sent_cmd)
    kfree_skb(hdev->sent_cmd)
                                        hci_cmd_timeout
                                          ...
                                          if (hdev->sent_cmd)
                                            sent = hdev->sent_cmd->data
                                                   ^^ UAF occurred
    hdev->sent_cmd = NULL

When hci_dev_init_sync() failed, hci_dev_open_sync() frees
hdev->send_cmd and set it to NULL.  However, hci_cmd_timeout() can run
just after freeing hdev->sent_cmd because hdev->cmd_timer is not
canceled.

This patch fixes the issue by canceling hdev->cmd_timer before freeing
hdev->sent_cmd.

Link: https://syzkaller.appspot.com/bug?id=cb23ebfc8f304f510fb717cb783fe8b496c7ffb1 [1]
Reported-by: syzbot+19a9f729f05272857487@syzkaller.appspotmail.com
Signed-off-by: Shigeru Yoshida <syoshida@redhat.com>
---
 net/bluetooth/hci_sync.c | 1 +
 1 file changed, 1 insertion(+)

Comments

Luiz Augusto von Dentz Nov. 7, 2022, 6:49 p.m. UTC | #1 
Hi Shigeru,

On Sat, Nov 5, 2022 at 12:04 PM Shigeru Yoshida <syoshida@redhat.com> wrote:
>
> syzbot reported use-after-free in hci_cmd_timeout() [1].  The scenario
> for the issue is as follows:
>
> Task                                    Workqueue
> ----------------------------------------------------------------------
> hci_dev_open_sync
>   ...
>   hci_dev_init_sync  <- failed
>   ...
>   if (hdev->sent_cmd)
>     kfree_skb(hdev->sent_cmd)
>                                         hci_cmd_timeout
>                                           ...
>                                           if (hdev->sent_cmd)
>                                             sent = hdev->sent_cmd->data
>                                                    ^^ UAF occurred
>     hdev->sent_cmd = NULL
>
> When hci_dev_init_sync() failed, hci_dev_open_sync() frees
> hdev->send_cmd and set it to NULL.  However, hci_cmd_timeout() can run
> just after freeing hdev->sent_cmd because hdev->cmd_timer is not
> canceled.
>
> This patch fixes the issue by canceling hdev->cmd_timer before freeing
> hdev->sent_cmd.
>
> Link: https://syzkaller.appspot.com/bug?id=cb23ebfc8f304f510fb717cb783fe8b496c7ffb1 [1]
> Reported-by: syzbot+19a9f729f05272857487@syzkaller.appspotmail.com
> Signed-off-by: Shigeru Yoshida <syoshida@redhat.com>
> ---
>  net/bluetooth/hci_sync.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
> index 76c3107c9f91..a011065220e4 100644
> --- a/net/bluetooth/hci_sync.c
> +++ b/net/bluetooth/hci_sync.c
> @@ -4696,6 +4696,7 @@ int hci_dev_open_sync(struct hci_dev *hdev)
>                         hdev->flush(hdev);
>
>                 if (hdev->sent_cmd) {
> +                       cancel_delayed_work_sync(&hdev->cmd_timer);
>                         kfree_skb(hdev->sent_cmd);
>                         hdev->sent_cmd = NULL;
>                 }
> --
> 2.38.1

A similar fix has already been applied:

https://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth-next.git/commit/?id=64b5c4c8e79c131fe8f135bab5e5dfaa245c5776
diff mbox series

Patch

diff --git a/net/bluetooth/hci_sync.c b/net/bluetooth/hci_sync.c
index 76c3107c9f91..a011065220e4 100644
--- a/net/bluetooth/hci_sync.c
+++ b/net/bluetooth/hci_sync.c
@@ -4696,6 +4696,7 @@  int hci_dev_open_sync(struct hci_dev *hdev)
 			hdev->flush(hdev);
 
 		if (hdev->sent_cmd) {
+			cancel_delayed_work_sync(&hdev->cmd_timer);
 			kfree_skb(hdev->sent_cmd);
 			hdev->sent_cmd = NULL;
 		}