From mboxrd@z Thu Jan  1 00:00:00 1970
From: ezequiel@vanguardiasur.com.ar (Ezequiel Garcia)
Date: Wed, 16 Jul 2014 10:30:01 -0300
Subject: [PATCH 1/5] [RFC-HACK] pinctrl-single: init by postcore_initcall()
In-Reply-To: <20140716123641.GB18374@atomide.com>
References: <1400799986-20043-1-git-send-email-chf.fritz@googlemail.com>
 <1400799986-20043-2-git-send-email-chf.fritz@googlemail.com>
 <20140522233154.GE29577@atomide.com>
 <20140716114558.GA1842@arch.cereza>
 <20140716123641.GB18374@atomide.com>
Message-ID: <20140716133001.GA4033@arch.cereza>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On 16 Jul 05:36 AM, Tony Lindgren wrote:
> * Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> [140716 04:48]:
> > On 22 May 04:31 PM, Tony Lindgren wrote:
> > > * Christoph Fritz <chf.fritz@googlemail.com> [140522 16:07]:
> > > > --- a/drivers/pinctrl/pinctrl-single.c
> > > > +++ b/drivers/pinctrl/pinctrl-single.c
> > > > @@ -2069,7 +2069,22 @@ static struct platform_driver pcs_driver = {
> > > >  #endif
> > > >  };
> > > >  
> > > > +#ifdef CONFIG_USE_OF
> > > > +static int __init pcs_driver_drv_init(void)
> > > > +{
> > > > +       return platform_driver_register(&pcs_driver);
> > > > +}
> > > > +postcore_initcall(pcs_driver_drv_init);
> > > > +
> > > > +static void __exit pcs_driver_drv_exit(void)
> > > > +{
> > > > +       platform_driver_unregister(&pcs_driver);
> > > > +}
> > > > +module_exit(pcs_driver_drv_exit);
> > > > +#else
> > > >  module_platform_driver(pcs_driver);
> > > > +#endif
> > > > +
> > > 
> > > I would prefer to initialize all the drivers with module_init
> > > and and if necessary, selected frameworks with subsys_initcall. 
> > > 
> > > In drivers/Makefile we do have painctrl before gpio, so this
> > > too could be just module_init once the other drivers are fixed.
> > > 
> > 
> > I've been using patches that change gpio and i2c initialization to module_init
> > for some time now. However, I still see some non-DT boards using the GPIO API
> > in early functions such as .init_machine.
> > 
> > Only when these boards get converted and the legacy support is dropped, we'll
> > be able to move all the drivers to module_init.
> 
> Hmm yeah tons of dependencies to dropping the legacy booting.. But to not
> have that dependency maybe we can convert those .init_machine functions
> easily to late_initcall.
> 

Could be. I can drop a few patches and then wait for Acks from the maintainers.
Although I'm not really sure this is worth the pain.
-- 
Ezequiel Garcia, VanguardiaSur
www.vanguardiasur.com.ar