Re: [RESEND 1/3] Bluetooth: Use cancel_work instead of cancel_work_sync

Hi Ulisses,

On Tue, Jan 31, 2012 at 03:49:19PM -0200, Ulisses Furquim wrote:

Hi Marcel,

On Tue, Jan 31, 2012 at 1:58 PM, Marcel Holtmann [off-list ref] wrote:

quoted

Hi Ulisses,

quoted

quoted

quoted

quoted

quoted

-       cancel_delayed_work_sync(&conn->disc_work);
+       cancel_delayed_work(&conn->disc_work);

I'm afraid we must use _sync variant here. disc_work is not supposed to
be running after hci_conn is deleted.

BTW, I believe we already addressed this issue in patches [PATCH 1/2]
Bluetooth: Fix potential deadlock and [PATCH 2/2] Bluetooth: Remove
unneeded locking. These patches are now pushed upstream. Could you

I will check those patches from upstream and let you know.

crap, I just acked these.

Johan, forget about my acks and just ignore them. Lets wait until we get
a clean new series.

This change is really wrong because we're on the delete path and Andre
sent other patches which I'm almost sure will address this problem.

lets do it this way, I only look at final patches that you and Andrei
signed off / acked.

Sure, we can do that this way.

I've checked recent upstream code and it works fine so far, please forget
about this particular patch.

The other 2 patches are still valid.

Best regards 
Andrei Emeltchenko