On Tue, Apr 12, 2016 at 9:39 AM, Thierry Reding
<thierry.reding@xxxxxxxxx> wrote:
> On Mon, Apr 11, 2016 at 12:47:23PM -0600, Stephen Warren wrote:
>> On 04/11/2016 12:32 PM, Rob Herring wrote:
>> > On Mon, Apr 11, 2016 at 11:05 AM, Stephen Warren <swarren@xxxxxxxxxxxxx> wrote:
>> > > On 04/11/2016 08:01 AM, Thierry Reding wrote:
>> > > >
>> > > > From: Thierry Reding <treding@xxxxxxxxxx>
>> > > >
>> > > > When Tegra124 support was first merged the unit-addresses of all devices
>> > > > were listed with a "0," prefix to encode the reg property's second cell.
>> > > > It turns out that this notation is not correct, and the "," separator is
>> > > > only used to separate fields in the unit address (such as the device and
>> > > > function number in PCI devices), not individual cells for addresses with
>> > > > more than one cell.
>> > >
>> > >
>> > > > diff --git a/arch/arm/boot/dts/tegra124-jetson-tk1.dts
>> > > > b/arch/arm/boot/dts/tegra124-jetson-tk1.dts
>> > >
>> > >
>> > > > - gpu@0,57000000 {
>> > > > + gpu@57000000 {
>> > > > /*
>> > > > * Node left disabled on purpose - the bootloader will
>> > > > enable
>> > > > * it after having set the VPR up
>> > >
>> > >
>> > > So the bootloader doesn't actually do that for the new node name at present.
>> > > I have written a patch to make it do so, but haven't sent it yet since I
>> > > wrote it in the middle of a large cleanup of U-Boot. I expect I can shuffle
>> > > it to the front of the series and send it soon though. Without a new
>> > > bootloader that contains this change, IIUC all graphics will be
>> > > non-operative if this change is applied.
>> >
>> > Then you should leave this one alone for a while.
>> >
>> > I also found this looking at u-boot:
>> >
>> > arch/arm/dts/tegra124.dtsi: gpu@57000000 {
>>
>> FWIW, that's because the U-Boot DTs use #address-cells=<1> on this chip so
>> there was no question of using commas in the unit address or not. That may
>> have been because U-Boot imported the DTs (or parts of them) from Linux
>> before Linux switched to #address-cells=<2> on this chip.
>
> I thought the reason had been that U-Boot didn't support any more than
> 32-bits for addresses on 32-bit ARM anyway, hence we never bothered to
> sync with the kernel DTS files regarding #address-cells.
Ah yes, there are lots of issues with u-boot around that. That's a
wonderful work-around though.
> Technically it would be possible for someone to write a DTS file with a
> GPU node with yet another name (the binding after all doesn't specify
> what the node name should be) so I think better matching would be safer
> in any case.
Strictly speaking, the node name should be part of the binding. We're
not too good at specifying it.
I've not looked, but I would think the compatible string would be
unique in this case.
Rob
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
