Re: [PATCH net-next v10 04/14] net: hold netdev instance lock during qdisc ndo_setup_tc

[Jakub Kicinski <kuba@kernel.org>]

On Mon, 5 May 2025 15:12:39 +0000 Cosmin Ratiu wrote:
> On Mon, 2025-05-05 at 08:07 -0700, Stanislav Fomichev wrote:
> > On 05/05, Cosmin Ratiu wrote:  
> > > diff --git a/net/core/dev.c b/net/core/dev.c
> > > index d1a8cad0c99c..134ceddf7fa5 100644
> > > --- a/net/core/dev.c
> > > +++ b/net/core/dev.c
> > > @@ -12020,9 +12020,9 @@ void
> > > unregister_netdevice_many_notify(struct
> > > list_head *head,
> > >                 struct sk_buff *skb = NULL;
> > >  
> > >                 /* Shutdown queueing discipline. */
> > > +               netdev_lock_ops(dev);
> > >                 dev_shutdown(dev);
> > >                 dev_tcx_uninstall(dev);  
> > 
> > There is a synchronize_net hidden inside of dev_tcx_uninstall, so
> > let's ops-lock only dev_shutdown here? Other than that, don't see
> > anything wrong. Can you send this separately and target net tree?  

Avoiding synchronize_net() under the instance lock as an optimization? 
We're under rtnl_lock here, probably a premature optimization?
But no strong preference..

BTW isn't the naming scheme now that dev_* takes the lock? So we should
probably add netif_ versions or rename these existing calls?