netdev - Re: [PATCH net V4 1/3] bonding: implement xdo_dev_state

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]

Message-ID: <07bad330d9259f851a5b6354c1c6a72587048c0e.camel@nvidia.com>
Date: Fri, 16 Aug 2024 09:10:25 +0000
From: Jianbo Liu <jianbol@...dia.com>
To: Tariq Toukan <tariqt@...dia.com>, "liuhangbin@...il.com"
	<liuhangbin@...il.com>
CC: "davem@...emloft.net" <davem@...emloft.net>, Leon Romanovsky
	<leonro@...dia.com>, "andy@...yhouse.net" <andy@...yhouse.net>, Gal Pressman
	<gal@...dia.com>, "jv@...sburgh.net" <jv@...sburgh.net>, "kuba@...nel.org"
	<kuba@...nel.org>, "pabeni@...hat.com" <pabeni@...hat.com>,
	"edumazet@...gle.com" <edumazet@...gle.com>, Saeed Mahameed
	<saeedm@...dia.com>, "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH net V4 1/3] bonding: implement xdo_dev_state_free and call
 it after deletion

On Fri, 2024-08-16 at 11:07 +0800, Hangbin Liu wrote:
> On Thu, Aug 15, 2024 at 05:21:01PM +0300, Tariq Toukan wrote:
> > From: Jianbo Liu <jianbol@...dia.com>
> > 
> > Add this implementation for bonding, so hardware resources can be
> > freed from the active slave after xfrm state is deleted. The netdev
> > used to invoke xdo_dev_state_free callback, is saved in the xfrm
> > state
> > (xs->xso.real_dev), which is also the bond's active slave.
> > 
> > And call it when deleting all SAs from old active real interface
> > while
> > switching current active slave.
> > 
> > Fixes: 9a5605505d9c ("bonding: Add struct bond_ipesc to manage SA")
> > Signed-off-by: Jianbo Liu <jianbol@...dia.com>
> > Signed-off-by: Tariq Toukan <tariqt@...dia.com>
> > Reviewed-by: Hangbin Liu <liuhangbin@...il.com>
> > ---
> >  drivers/net/bonding/bond_main.c | 32
> > ++++++++++++++++++++++++++++++++
> >  1 file changed, 32 insertions(+)
> > 
> > diff --git a/drivers/net/bonding/bond_main.c
> > b/drivers/net/bonding/bond_main.c
> > index 1cd92c12e782..eb5e43860670 100644
> > --- a/drivers/net/bonding/bond_main.c
> > +++ b/drivers/net/bonding/bond_main.c
> > @@ -581,6 +581,8 @@ static void bond_ipsec_del_sa_all(struct
> > bonding *bond)
> >                                    __func__);
> >                 } else {
> >                         slave->dev->xfrmdev_ops-
> > >xdo_dev_state_delete(ipsec->xs);
> > +                       if (slave->dev->xfrmdev_ops-
> > >xdo_dev_state_free)
> > +                               slave->dev->xfrmdev_ops-
> > >xdo_dev_state_free(ipsec->xs);
> >                 }
> >                 ipsec->xs->xso.real_dev = NULL;
> >         }
> > @@ -588,6 +590,35 @@ static void bond_ipsec_del_sa_all(struct
> > bonding *bond)
> >         rcu_read_unlock();
> >  }
> >  
> > +static void bond_ipsec_free_sa(struct xfrm_state *xs)
> > +{
> > +       struct net_device *bond_dev = xs->xso.dev;
> > +       struct net_device *real_dev;
> > +       struct bonding *bond;
> > +       struct slave *slave;
> > +
> > +       if (!bond_dev)
> > +               return;
> > +
> > +       rcu_read_lock();
> > +       bond = netdev_priv(bond_dev);
> > +       slave = rcu_dereference(bond->curr_active_slave);
> > +       real_dev = slave ? slave->dev : NULL;
> > +       rcu_read_unlock();
> 
> As I replied in   
> https://lore.kernel.org/netdev/ZrwgRaDc1Vo0Jhcj@Laptop-X1/,
> 

As I replied, the RCU lock is to protect the reading of the content
pointed by curr_active_slave, not the slave->dev itself.

> > +
> > +       if (!slave)
> > +               return;
> > +
> > +       if (!xs->xso.real_dev)
> > +               return;
> > +
> > +       WARN_ON(xs->xso.real_dev != real_dev);
> > +
> > +       if (real_dev && real_dev->xfrmdev_ops &&
> > +           real_dev->xfrmdev_ops->xdo_dev_state_free)
> > +               real_dev->xfrmdev_ops->xdo_dev_state_free(xs);
> 
> How do you make sure the slave not freed after rcu_read_unlock()?
> 

So do you want to move rcu_read_unlock after xdo_dev_state_free, just
like what you did in your patches? 
 

Thanks!
Jianbo