Re: [PATCH 05/11] net: core: provide devm_register_netdev()

[Bartosz Golaszewski <bgolaszewski@baylibre.com>]

czw., 7 maj 2020 o 18:53 Jakub Kicinski <kuba@kernel.org> napisał(a):
>
> On Thu, 7 May 2020 11:25:01 +0200 Bartosz Golaszewski wrote:
> > śr., 6 maj 2020 o 19:12 Jakub Kicinski <kuba@kernel.org> napisał(a):
> > >
> > > On Wed, 6 May 2020 08:39:47 +0200 Bartosz Golaszewski wrote:
> > > > wt., 5 maj 2020 o 19:31 Jakub Kicinski <kuba@kernel.org> napisał(a):
> > > > >
> > > > > On Tue,  5 May 2020 16:02:25 +0200 Bartosz Golaszewski wrote:
> > > > > > From: Bartosz Golaszewski <bgolaszewski@baylibre.com>
> > > > > >
> > > > > > Provide devm_register_netdev() - a device resource managed variant
> > > > > > of register_netdev(). This new helper will only work for net_device
> > > > > > structs that have a parent device assigned and are devres managed too.
> > > > > >
> > > > > > Signed-off-by: Bartosz Golaszewski <bgolaszewski@baylibre.com>
> > > > >
> > > > > > diff --git a/net/core/dev.c b/net/core/dev.c
> > > > > > index 522288177bbd..99db537c9468 100644
> > > > > > --- a/net/core/dev.c
> > > > > > +++ b/net/core/dev.c
> > > > > > @@ -9519,6 +9519,54 @@ int register_netdev(struct net_device *dev)
> > > > > >  }
> > > > > >  EXPORT_SYMBOL(register_netdev);
> > > > > >
> > > > > > +struct netdevice_devres {
> > > > > > +     struct net_device *ndev;
> > > > > > +};
> > > > >
> > > > > Is there really a need to define a structure if we only need a pointer?
> > > > >
> > > >
> > > > There is no need for that, but it really is more readable this way.
> > > > Also: using a pointer directly doesn't save us any memory nor code
> > > > here.
> > >
> > > I don't care either way but devm_alloc_etherdev_mqs() and co. are using
> > > the double pointer directly. Please make things consistent. Either do
> > > the same, or define the structure in some header and convert other
> > > helpers to also make use of it.
> >
> > In order to use devres_find() to check if struct net_device is managed
> > in devm_register_netdev() I need to know the address of the release
> > function used by devm_alloc_etherdev_mqs(). Do you mind if I move all
> > networking devres routines (currently only devm_alloc_etherdev_mqs())
> > into a separate .c file (e.g. under net/devres.c)?
>
> To implement Edwin's suggestion? Makes sense, but I'm no expert, let's
> also CC Heiner since he was asking about it last time.

Yes, because taking the last bit of priv_flags from net_device seems
to be more controversial but if net maintainers are fine with that I
can simply go with the current approach.

Bart