Re: [PATCH v5 2/2] ACPI: amba bus probing support

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Mon, Jan 11, 2016 at 06:26:00PM +0200, Andy Shevchenko wrote:
> On Mon, Jan 11, 2016 at 5:27 PM, Russell King - ARM Linux
> <linux@arm.linux.org.uk> wrote:
> > On Mon, Jan 11, 2016 at 05:13:20PM +0200, Andy Shevchenko wrote:
> >> On Mon, Jan 11, 2016 at 3:26 PM, Aleksey Makarov
> >> <aleksey.makarov@linaro.org> wrote:
> >> dev->res is 0 before this one, right? Could you use this fact instead
> >> of address_found flag?
> >
> > amba_device_alloc() zero-initialises everything.  However, dev->res is
> > a struct resource, and I'd prefer _this_ method that the OT is using
> > to testing some random part of struct resource.
> 
> So, you mean resource->start = 0 is not enough reliable?

I'd rather not make assumptions about what in a resource is valid
or not valid.

> >> > +               default:
> >> > +                       dev_warn(&adev->dev, "Invalid resource\n");
> >>
> >> Why? Isn't possible to have other resources for the devices?
> >
> > AMBA primecell devices have one memory region, and a number of
> > interrupts.  Other resource types don't make sense.
> 
> But isn't warning on the other side too noisy?

Why would it be "too noisy" ?  Isn't it saying that the ACPI is in
error to include more resource types that aren't part of specifying
the AMBA primecell device?  Maybe it should be dev_err(), because
it's technically an error...

Are you expecting people to create ACPI tables with a lot of rubbish
resources attached to these devices?

> Yes, and in this case ret > 0 should be converted to an appropriate
> error code, otherwise ACPI core will consider this as a normal
> execution, right?

You are assuming that it does return a positive non-zero value in the
first place.

-- 
RMK's Patch system: http://www.arm.linux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.