Re: [RESEND PATCHv4 1/1] drivers/amba: add reset control to amba bus probe

[Philipp Zabel <p.zabel@pengutronix.de>]

Hi Dinh, Linus,

On Fri, 2019-08-23 at 10:42 -0500, Dinh Nguyen wrote:
> 
> On 8/23/19 4:19 AM, Linus Walleij wrote:
> > On Tue, Aug 20, 2019 at 4:58 PM Dinh Nguyen <dinguyen@kernel.org> wrote:
> > 
> > > @@ -401,6 +402,26 @@ static int amba_device_try_add(struct amba_device *dev, struct resource *parent)
> > >         ret = amba_get_enable_pclk(dev);
> > >         if (ret == 0) {
> > >                 u32 pid, cid;
> > > +               int count;
> > > +               struct reset_control *rstc;
> > > +
> > > +               /*
> > > +                * Find reset control(s) of the amba bus and de-assert them.
> > > +                */
> > > +               count = reset_control_get_count(&dev->dev);

The reset_control_get_count() inline stub returns -ENOENT, so the
compiler can throw away the complete loop.

> > > +               while (count > 0) {
> > > +                       rstc = of_reset_control_get_shared_by_index(dev->dev.of_node, count - 1);

Since resets are looked up with of_reset_control_get, I'd use
of_reset_control_get_count() above for consistency. But see below:

> > > +                       if (IS_ERR(rstc)) {
> > > +                               if (PTR_ERR(rstc) == -EPROBE_DEFER)
> > > +                                       ret = -EPROBE_DEFER;
> > > +                               else
> > > +                                       dev_err(&dev->dev, "Can't get amba reset!\n");
> > > +                               break;
> > > +                       }
> > > +                       reset_control_deassert(rstc);
> > > +                       reset_control_put(rstc);
> > > +                       count--;
> > > +               }

It looks like the order of deassertions is irrelevant. If so,
You can use of_reset_control_array_get() to simplify this:

+		rstc = of_reset_control_array_get_optional_shared(dev->dev.of_node);
+		if (IS_ERR(rstc)) {
+			if (PTR_ERR(rstc) != -EPROBE_DEFER)
+				dev_err(&dev->dev, "Can't get amba reset!\n");
+			return PTR_ERR(rstc);
+		}
+		reset_control_deassert(rstc);
+		reset_control_put(rstc);

> > I'm not normally a footprint person, but the looks of the stubs in
> > <linux/reset.h> makes me suspicious whether this will have zero impact
> > in size on platforms without reset controllers.
> > 
> > Can you just ls -al on the kernel without CONFIG_RESET_CONTROLLER
> > before and after this patch and ascertain that it has zero footprint effect?
> 
> Thanks for the review. I checked it, and indeed, it does have a zero
> footprint effect.
> 
> > 
> > If it doesn't I'd sure like to break this into its own function and
> > stick a if (!IS_ENABLED(CONFIG_RESET_CONTROLLER)) return 0;
> > in there to make sure the compiler drops it.
> > 
> > Also it'd be nice to get Philipp's ACK on the semantics, though they
> > look correct to me.

regards
Philipp

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel