Re: [PATCH 1/2] xen-netfront: don't nest queue locks in xennet_connect()

[David Vrabel <david.vrabel@citrix.com>]

On 02/07/14 18:36, Boris Ostrovsky wrote:
> On 07/02/2014 11:09 AM, David Vrabel wrote:
>> The nesting of the per-queue rx_lock and tx_lock in xennet_connect()
>> is confusing to both humans and lockdep.  The locking is safe because
>> this is the only place where the locks are nested in this way but
>> lockdep still warns.
>>
>> Instead of adding the missing lockdep annotations, refactor the
>> locking to avoid the confusing nesting.  This is still safe, because
>> the xenbus connection state changes are all serialized by the xenwatch
>> thread.
> 
> Tulin (GSoC student) is working on adding some concurrency to the
> xenwatch thread (possibly eliminating it and replacing with a
> workqueue). Which means that we will need to add some sort of
> synchronization in watch callbacks, whenever they are required.

The back/frontend changed callbacks should be serialized with the device
mutex in the same way that the device core does for probe and remove etc.

David