Re: [PATCH] soc: mediatek: SCPSYS: use builtin_platform_driver

[Paul Gortmaker <paul.gortmaker-CWA4WttNNZF54TAoqtyWWQ@public.gmane.org>]

[Re: [PATCH] soc: mediatek: SCPSYS: use builtin_platform_driver] On 22/12/2015 (Tue 21:36) Daniel Kurtz wrote:

> Hi Matthias,
> 
> One thing below is missing from this version of the patch...
> 
> > On Fri, Dec 18, 2015 at 7:19 PM, Matthias Brugger
> > <matthias.bgg-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> >> SCPSYS can't be built as module. Use builtin_platform_driver instead.
> >> For this probe must not be __init and the data accessed can't be
> >> __initconst. Remove this macros. To make the impact as small as possible,
> >> fold scp_domain_data into scp_domain via a pointer.
> 
> [snip]
> 
> >> @@ -542,10 +533,11 @@ static const struct of_device_id of_scpsys_match_tbl[] = {
> >>  };
> >>
> >>  static struct platform_driver scpsys_drv = {
> >> +       .probe = scpsys_probe,
> >>         .driver = {
> >>                 .name = "mtk-scpsys",
> 
> This was in Sascha's version of the patch:
> 
>   .suppress_bind_attrs = true,

Unless you've got a very specific use case where an unbind makes sense
(like a quad port ethernet card, where you unbind one port and then do
PCI pass through of it to a kvm guest for example) then yes, you should
suppress unbind.  So for most drivers aimed at a single device, or where
being built in but unbinding just makes no sense for any possible use
case, then yes you should set the above.

Paul.
--

> 
> Do we still need it now that we use "builtin_platform_driver"?
> 
> >>                 .owner = THIS_MODULE,
> >>                 .of_match_table = of_match_ptr(of_scpsys_match_tbl),
> >>         },
> >>  };
> >> -builtin_platform_driver_probe(scpsys_drv, scpsys_probe);
> >> +builtin_platform_driver(scpsys_drv);
> >> --
> >> 2.6.2
> >>