Re: [PATCH 3/3] regulator: core: make regulator_register() EPROBE_DEFER aware

From: Marco Felsch
Date: Wed Sep 18 2019 - 12:06:53 EST


On 19-09-18 08:53, Dmitry Torokhov wrote:
> On Wed, Sep 18, 2019 at 1:18 AM Marco Felsch <m.felsch@xxxxxxxxxxxxxx> wrote:
> >
> > On 19-09-17 17:57, Dmitry Torokhov wrote:
> > > On Tue, Sep 17, 2019 at 4:42 PM Marco Felsch <m.felsch@xxxxxxxxxxxxxx> wrote:
> > > >
> > > > Sometimes it can happen that the regulator_of_get_init_data() can't
> > > > retrieve the config due to a not probed device the regulator depends on.
> > > > Fix that by checking the return value of of_parse_cb() and return
> > > > EPROBE_DEFER in such cases.
> > >
> > > Treating EPROBE_DEFER in a special way is usually wrong.
> > > regulator_of_get_init_data() may fail for multiple reasons (no memory,
> > > invalid DT, etc, etc). All of them should abort instantiating
> > > regulator.
> >
> > Those errors are handled but the behaviour of this funciton is to return
> > NULL in such errors which is fine for the caller of this function. I
> > only want to handle EPROBE_DEFER special..
>
> And I am saying it is wrong to handle only EPROBE_DEFER.
> regulator_of_get_init_data() should always return ERR_PTR()-encoded
> error code when parsing callback returns error, so that regulator core
> does not mistakenly believe that there is no configuration/init data
> when in fact there is, but we failed to handle it properly.
>
> IOW I'm advocating for extending you patch so that it reads:
>
> + ret = desc->of_parse_cb(child, desc, config);
> + if (ret) {
> + of_node_put(child);
> + return ERR_PTR(ret);
> + }

I know what you mean but I wanted to keep the core changes minimal and I
tought that it was intentional by the core.

Regards,
Marco

> Thanks.
>
> --
> Dmitry
>
>

--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |