netdev - Re: [PATCH v2] bonding: rejoin multicast groups on VLANs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 5 Oct 2010 19:07:57 -0300
From:	Flavio Leitner <fbl@...hat.com>
To:	Andy Gospodarek <andy@...yhouse.net>
Cc:	netdev@...r.kernel.org, bonding-devel@...ts.sourceforge.net,
	Jay Vosburgh <fubar@...ibm.com>
Subject: Re: [PATCH v2] bonding: rejoin multicast groups on VLANs

On Mon, Oct 04, 2010 at 09:24:10AM -0400, Andy Gospodarek wrote:
> On Thu, Sep 30, 2010 at 05:45:24PM -0300, Flavio Leitner wrote:
> > It fixes bonding to rejoin multicast groups added
> > to VLAN devices on top of bonding when a failover
> > happens.
> > 
> > Signed-off-by: Flavio Leitner <fleitner@...hat.com>
> > ---
> >  drivers/net/bonding/bond_main.c |   61 +++++++++++++++++++++++++++++++++-----
> >  drivers/net/bonding/bonding.h   |    1 +
> >  2 files changed, 54 insertions(+), 8 deletions(-)
> > 
> [...]
> > @@ -944,7 +979,9 @@ static void bond_mc_swap(struct bonding *bond, struct slave *new_active,
> >  
> >  		netdev_for_each_mc_addr(ha, bond->dev)
> >  			dev_mc_add(new_active->dev, ha->addr);
> > -		bond_resend_igmp_join_requests(bond);
> > +
> > +		/* rejoin multicast groups */
> > +		queue_delayed_work(bond->wq, &bond->mcast_work, 0);
> >  	}
> >  }
> >  
> 
> I was hoping that one patch would just make the changes so
> retransmission was supported on VLANs and the second patch would queue
> the work and add the tunable for multiple retransmissions, but I guess
> it wasn't clear enough.
> 
> I felt like that would divide the patches up into the bug-fix (VLANs +
> multicast not working) and the new feature (multiple retransmissions
> from the workqueue).

I'm not seeing how to that because the first patch changes to send
the membership directly and not using timers anymore, so the call
trace usually is:

 write_lock_bh(&bond->curr_slave_lock); <-- hold lock
 bond_select_active_slave() 
  ->bond_change_active_slave()
     ->bond_mc_swap()
        ->tx membership reports
 

then trying to send the IGMP packet directly, it will end up at:
  bond_start_xmit()
   ->bond_xmit_activebackup() (mode=1, for example)
      ->read_lock(&bond->lock);
      ->read_lock(&bond->curr_slave_lock); <-- deadlock


> I will test these out this morning and make sure things look good.

I have a better patch sequence which I'm planning post replying to
this thread as soon as I finish up testing them. It is still using
workqueue though.

--
Flavio
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html