Re: [PATCH] Bluetooth: postpone hci_dev unregistration

[Marcel Holtmann <marcel@holtmann.org>]

Hi Jiri,

> (I sent this a week ago but it seems to have got lost in other noise, 
> resending)
> 
> From: Jiri Kosina <jkosina@suse.cz>
> 
> Bluetooth: postpone hci_dev unregistration
>     
> Commit b40df57 substituted bh_lock_sock() in hci_sock_dev_event() for 
> lock_sock() when unregistering HCI device, in order to prevent deadlock 
> against locking in l2cap_connect_cfm() from softirq context.
>     
> This however introduces another problem - hci_sock_dev_event() for 
> HCI_DEV_UNREG can also be triggered in atomic context, in which calling 
> lock_sock() is not safe as it could sleep. Reported by Jeremy Fitzhardinge 
> at http://lkml.org/lkml/2007/4/23/271
> 
> This patch moves the detaching of sockets from hci_device into workqueue, 
> so that lock_sock() can be used safely. This requires movement of 
> deallocation of hci_dev - deallocating device just after 
> hci_unregister_dev() would be too soon, as it could happen before the 
> workqueue has been run.

I saw the report on LKML, but I am not really comfortable with this
approach. It feels like an ugly hack. This needs more thinking and I
think that simplifying the looking between HCI and L2CAP should be the
goal.

Regards

Marcel