On 2025-03-04 07:44, Dragan Simic wrote:
On 2025-03-03 23:36, Heiko Stuebner wrote:
On Sun, 02 Mar 2025 19:48:02 +0100, Dragan Simic wrote:
This is a small series that introduces small improvements to the way
Pine64 RockPro64 [1] single-board-computer is described in the DT
files.
This applies to both production-run revisions of the RockPro64.
The introduced improvements boil down to eliminating some warnings
from
the kernel log, by adding a previously undefined regulator and by
adding
some previously missing references to the regulators.
[...]
Applied, thanks!
[1/2] arm64: dts: rockchip: Add avdd HDMI supplies to RockPro64 board
dtsi
commit: bd1c959f37f384b477f51572331b0dc828bd009a
[2/2] arm64: dts: rockchip: Add missing PCIe supplies to RockPro64
board dtsi
commit: 64ef4a4320e7aa3f0f267e01f170f52b90bf0b1b
I've moved the pcie12v supply up one line.
While in a mathematical sense it's true 12 > 3.3, we're sorting
alphabetical, so it's 1?? < 3?? .
And yes I sympathize with 3.3 < 12, but also have come to appreciate
not
having overly many special cases :-)
Great, thanks! :)
I'm fine with the alphabetical ordering, albeit with some caveats
described below, but the following part of the patch description
should also be removed, if possible, so the patch description fully
matches the introduced changes:
Shuffle and reorder the "vpcie*-supply" properties a bit, so they're
sorted
alphanumerically, which is a bit more logical and more useful than
having
these properties listed in their strict alphabetical order.
Admittedly, the patch itself indeed looks nicer [1] with the
alphabetical ordering, :) (which is quasi-alphabetical, actually,
please see below) but it might be better to have the DT guidelines
adjusted to specify alphanumerical ordering as well. Even ls(1)
can sort filenames alphanumerically, with its "-v" option that
performs "natural sort of (version) numbers within text". [2]
The real trouble is that, if we were to apply the alphabetical
ordering strictly, we'd end up with the following order of node
properties in which 12 is before 1.8 and 3.3, which IMHO doesn't
make much sense:
vpcie0v9-supply = <&vcca_0v9>;
vpcie12v-supply = <&vcc12v_dcin>;
vpcie1v8-supply = <&vcca_1v8>;
vpcie3v3-supply = <&vcc3v3_pcie>;
I'm hoping you'll agree that specifying alphanumerical ordering
for the properties in the DTS coding style is the way to go, just
like it's already specified for the ordering of the nodes. I'll
go ahead and submit an appropriate patch for the DT guidelines.
[1]
https://web.git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip.git/commit/?id=64ef4a4320e7aa3f0f267e01f170f52b90bf0b1b
[2] https://man.archlinux.org/man/ls.1.en#v
For future reference and for anyone interested, here's a link to
the above-mentioned DTS coding style patch:
https://lore.kernel.org/linux-kernel/09d6f2fc111b3d6e58987336944f93ec36b65118.1741071107.git.dsimic@xxxxxxxxxxx/T/#u
