> On Mon, Mar 11, 2013 at 08:22:05PM +0400, Alexander Shiyan wrote:
> > > On Mon, Mar 11, 2013 at 12:01:58PM +0100, Jean-Christophe PLAGNIOL-VILLARD wrote:
> > > > > +static int clps711x_cs_probe(struct device_d *dev)
> > > > > +{
> > > > > + u32 rate;
> > > > > + struct clk *timer_clk;
> > > > > +
> > > > > + timer_clk = clk_get(dev, NULL);
> > > > > + if (IS_ERR(timer_clk))
> > > > > + return PTR_ERR(timer_clk);
> > > > > +
> > > > > + rate = clk_get_rate(timer_clk);
> > > > > + clps711x_timer_base = dev_request_mem_region(dev, 0);
> > > > > + if (!clps711x_timer_base) {
> > > > > + clk_put(timer_clk);
> > > > > + return -ENOENT;
> > > > > + }
> > > > this deserve a nice crash
> > >
> > > No, it doesn't. First of all we are very early here, so we might not even
> > > see the crash. Then, with devicetree probing we may often end up with
> > > the same devices registered from the devicetree and from the
> > > platform/soc. While this should find a way to avoid duplicate device
> > > registration, it is not nice having barebox crash in this case.
> >
> > So what is the solution you propose in this case?
>
> For the device duplication I don't have a solution yet. Basically I
> wanted to say that your patch looks good the way you did it.
Later I can move check for clps711x_timer_base at start of probe function.
So at least we may avoid to add another clocksource.
---
_______________________________________________
barebox mailing list
barebox@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/barebox
