[U-Boot] [PATCH 2/2] arm: mvebu: armada-xp/37x/38x: Enable PCIe DT node

[Chris Packham]

On Fri, 18 Jan 2019, 7:15 PM Stefan Roese <sr at denx.de wrote:

> Hi Chris,
>
> On 17.01.19 20:57, Chris Packham wrote:
> > Hi Stefan,
> >
> > On Fri, Jan 18, 2019 at 2:35 AM Stefan Roese <sr at denx.de> wrote:
> >>
> >> This patch enables the DT PCIe nodes for the Armada XP/37x/38x boards.
> >> This is needed for the new DM_PCI support in the MVEBU PCIe driver.
> >>
> >> Signed-off-by: Stefan Roese <sr at denx.de>
> >> Cc: Dirk Eibach <dirk.eibach at gdsys.cc>
> >> Cc: Mario Six <mario.six at gdsys.cc>
> >> Cc: Chris Packham <chris.packham at alliedtelesis.co.nz>
> >> Cc: Phil Sutter <phil at nwl.cc>
> >> Cc: Marek Behún <marek.behun at nic.cz>
> >> Cc: VlaoMao <vlaomao at gmail.com>
> >> ---
> >>   arch/arm/dts/armada-375.dtsi        | 2 +-
> >>   arch/arm/dts/armada-380.dtsi        | 2 +-
> >>   arch/arm/dts/armada-385.dtsi        | 2 +-
> >>   arch/arm/dts/armada-xp-mv78230.dtsi | 2 +-
> >>   arch/arm/dts/armada-xp-mv78260.dtsi | 2 +-
> >>   arch/arm/dts/armada-xp-mv78460.dtsi | 2 +-
> >>   6 files changed, 6 insertions(+), 6 deletions(-)
> >
> > I think this should be at the board level instead of the SoC.
> I agree (please see below).
>
> > There
> > are boards that don't use the pcie controller even though it's in the
> > SoC.
>
> This use case is handled by enabling the PCI_MVEBU driver on a board
> per board case in U-Boot. But nevertheless I agree in general.
>
> > Another good reason is that this deviates from the dtsi files in
> > Linux
> >
> >
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm/boot/dts/armada-375.dtsi#n548
>
> Yes, my first version did not change the dtsi file but the board dts
> file (theadorable in this case). My motivation was to not change anything
> in the configuration for these Armada XP/38x boards and to enable the
> PCIe DT node in general for all boards (frankly I was too lazy to do these
> DT changes for all those boards). But I agree that this approach is
> cumbersome and I'll follow-up with a v2 patch moving this "okay" to the
> board DT files.
>

You could do theodorable an leave the others for the board maintainers.
Even the defconfig changes should be innocuous without the dts change.

Anyways, it would be great if you (and others) could test the new
> DM PCI driver on your board(s).
>

Yes I  was planning to do that this weekend.


> Thanks,
> Stefan
>