Re: [PATCH 2/2] net: phy: Ensure the MDIO bus module is held

[Ezequiel Garcia <ezequiel.garcia@free-electrons.com>]

On 23 Jul 03:40 PM, Ezequiel Garcia wrote:
> On 23 Jul 11:22 AM, Florian Fainelli wrote:
> > 2014-07-22 17:16 GMT-07:00 Ezequiel Garcia <ezequiel.garcia@free-electrons.com>:
> > > @@ -614,10 +623,14 @@ int phy_attach_direct(struct net_device *dev, struct phy_device *phydev,
> > >          */
> > >         err = phy_init_hw(phydev);
> > >         if (err)
> > > -               phy_detach(phydev);
> > > -       else
> > > -               phy_resume(phydev);
> > > +               goto err_module_put;
> > >
> > > +       phy_resume(phydev);
> > > +       return err;
> > > +
> > > +err_module_put:
> > > +       module_put(bus_module);
> > > +       phy_detach(phydev);
> > 
> > Since we are calling phy_detach() which is also attempting to do a
> > module_put(), does not that result in one too many calls to
> > module_put() on error path?
> > 
> 
> As far as I can see all that module_get() and module_put() do
> is increment and decrement refcounters. Therefore, I think you can
> have as many module_{get,put}() calls as you want on a path.

After reading this again, I know see what your pointing. The call to
phy_detach already puts the module, so we don't need to do it on the
error path above.

I'll cook up v2.
-- 
Ezequiel García, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com