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

[Bartosz Golaszewski]

wt., 5 maj 2020 o 21:25 Edwin Peer <edwin.peer@xxxxxxxxxxxx> napisaÅ(a):
> > +
> > +static void devm_netdev_release(struct device *dev, void *this)
> > +{
> > +       struct netdevice_devres *res = this;
> > +
> > +       unregister_netdev(res->ndev);
> > +}
> > +
> > +/**
> > + *     devm_register_netdev - resource managed variant of register_netdev()
> > + *     @ndev: device to register
> > + *
> > + *     This is a devres variant of register_netdev() for which the unregister
> > + *     function will be call automatically when the parent device of ndev
> > + *     is detached.
> > + */
> > +int devm_register_netdev(struct net_device *ndev)
> > +{
> > +       struct netdevice_devres *dr;
> > +       int ret;
> > +
> > +       /* struct net_device itself must be devres managed. */
> > +       BUG_ON(!(ndev->priv_flags & IFF_IS_DEVRES));
> > +       /* struct net_device must have a parent device - it will be the device
> > +        * managing this resource.
> > +        */
>
> Catching static programming errors seems like an expensive use of the
> last runtime flag in the enum. It would be weird to devres manage the
> unregister and not also choose to manage the underlying memory in the
> same fashion, so it wouldn't be an obvious mistake to make. If it must
> be enforced, one could also iterate over the registered release
> functions and check for the presence of devm_free_netdev without
> burning the flag.
>

Hi Edwin,

I've submitted this patch some time ago already and was told to check
if the underlying memory is managed too. I guess I could try to use
devres_find() here though.

Re the last bit in priv_flags: is this really a problem though? It's
not like struct net_device must remain stable - e.g. we can make
priv_flags a bitmap.

Bart