Re: [PATCH v6 bluetooth-next] 6lowpan: Use skb_cow in IPHC decompression.

[Alexander Aring <alex.aring@gmail.com>]

Hi Jukka,

On Tue, Oct 14, 2014 at 11:35:05AM +0300, Jukka Rissanen wrote:
> Hi Alex,
> 
> On ma, 2014-10-13 at 19:22 +0200, Alexander Aring wrote:
> > Hi Jukka,
> > 
> > sorry.
> > 
> > I was a little too fast here, because I am sure now this should solve your
> > lockdep issue.
> 
> Unfortunately this patch did not help with the issue.
> 

mhh, ok.

> > 
> > On Mon, Oct 13, 2014 at 07:11:11PM +0200, Alexander Aring wrote:
> > > Hi Jukka,
> > > 
> > > On Mon, Oct 13, 2014 at 06:09:14PM +0300, Jukka Rissanen wrote:
> > > > Hi Martin,
> > > > 
> > > > On ma, 2014-10-13 at 15:56 +0100, Martin Townsend wrote:
> > > > > Hi Jukka,
> > > > > 
> > > > > Does this patch help?
> > > > 
> > > > Unfortunately no, I still see inconsistent lock state. It would probably
> > > > have been too easy :)
> > > > 
> > > 
> > > I remeber something, I think 802.15.4 had similar issues long time ago.
> > > 
> > s/remeber/remember/
> > 
> > > The fix was 20e7c4e80dcd01dad5e6c8b32455228b8fe9c619 ("6lowpan: fix
> > > lockdep splats"). Please check that, you need something like that!
> > > 
> > 
> > Something like that:
> > 
> > diff --git a/net/bluetooth/6lowpan.c b/net/bluetooth/6lowpan.c
> > index 4ebc806..02fd21a 100644
> > --- a/net/bluetooth/6lowpan.c
> > +++ b/net/bluetooth/6lowpan.c
> > @@ -642,7 +642,27 @@ static netdev_tx_t bt_xmit(struct sk_buff *skb, struct net_device *netdev)
> >         return err < 0 ? NET_XMIT_DROP : err;
> >  }
> >  
> > +static struct lock_class_key bt_tx_busylock;
> > +static struct lock_class_key bt_netdev_xmit_lock_key;
> > +
> > +static void bt_set_lockdep_class_one(struct net_device *dev,
> > +                                    struct netdev_queue *txq,
> > +                                    void *_unused)
> > +{
> > +       lockdep_set_class(&txq->_xmit_lock,
> > +                         &bt_netdev_xmit_lock_key);
> > +}
> > +
> > +
> > +static int bt_dev_init(struct net_device *dev)
> > +{
> > +       netdev_for_each_tx_queue(dev, bt_set_lockdep_class_one, NULL);
> > +       dev->qdisc_tx_busylock = &bt_tx_busylock;
> > +       return 0;
> > +}
> > +
> >  static const struct net_device_ops netdev_ops = {
> > +       .ndo_init               = bt_dev_init,
> >         .ndo_start_xmit         = bt_xmit,
> >  };
> > 
> > 
> > - Alex
> > 
> > [0] https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=20e7c4e80dcd01dad5e6c8b32455228b8fe9c619
> 
> 
> Martin had an idea about the issue being related to work queues, I will
> investigate that further.
> 

ok. The above one should fix if some dev_queue_xmit call another
dev_queue_xmit. Don't know if you have some architecture like this in
bluetooth.

- Alex