Hi Andre, Thanks for reviewing, and apologies for reposting so quickly. It's my first patch submitted, and I leaned more than I should on the LLM without doing my due diligence.
I will hold off on submitting any more versions to this one until the DTS makes it in the kernel. On Thu, Oct 22, 2026 at 4:48 AM Andre Przywara <[email protected]> wrote: > > Hi Cristian, > > please try to avoid unsolicited new version posts, in particular shortly > after the previous version. > It's not like maintainers and reviewers twiddling their thumbs waiting > for new patches, so please give everybody enough time to have a look. > I'd say a few days at least. > If you find issues in your own patches, please reply on the list, > pointing this out. And then still wait for proper reviews before reposting. > > On 10/22/26 07:55, Cristian Filipescu via B4 Relay wrote: > > Add a device tree and a separate defconfig for the Orange Pi 3 LTS. The > > LTS has an H6, 2 GiB LPDDR3, 8 GiB eMMC and a Motorcomm YT8531C Ethernet > > PHY, with power and reset wiring different from the original Orange Pi 3. > > > > The device tree is based on Jernej Skrabec's Linux submission: > > https://lore.kernel.org/linux-sunxi/[email protected]/ > > Which had comments, so have those been addressed? The Linux series is also posted. Several April review fixes are included; however, the LED drive strength and faster eMMC modesI haven't investigated yet. Looking at it now. this is the high-level of the fixes: - use hyphens in oscillator node/output names: ADDRESSED - use CLK_OSC32K_FANOUT: ADDRESSED - add input supplies for GMAC/Wi-Fi regulators: ADDRESSED - Keep BLDO3 always on: ADDRESSED - document DCDCA/DCDCB polyphase arrangement: ADDRESSED - correct Ethernet delay configuration: ADDRESSED; Andrew’s suggested settings passed the DHCP/ping tests - explicit LED drive strength: STILL OPEN - enable faster eMMC modes, including HS200: STILL OPEN Following your defconfig reply, I’ll remove the custom LED handoff code and return to the existing SPL power LED configuration. > > It lives under arch/arm/dts until the board description is merged into > > Please don't do this. If there is a mainline DTS pending, move that into > dts/upstream anyway, that way we don't have two versions of patches to > deal with. We don't take non-mainline DTs anyway, so the support has to > wait until the .dts ended up in a tagged mainline Linux release, then we > can cherry-pick that DT patch. Compare U-Boot commit 4112b26537cf04 for > an example. > For the sake of getting the (defconfig) patch out, and allow people to > play with that, it's fine to post, but in a setup that matches the final > submission. I’ll use dts/upstream and make the Linux dependency explicit. > > > Linux and reaches dts/upstream. The current dts/upstream H6 SoC and OPP > > includes are reused. The original authorship and sign-off are retained. > > > > Clock naming, regulator supply chaining and BLDO3 always-on corrections > > from the earlier review are included. Ethernet uses the Motorcomm driver > > with rgmii-id, zero MAC delays and explicit 1950 ps RX/TX PHY delays, as > > proposed during the Linux review. This configuration has been tested on > > the board with DHCP and ICMP. > > That level of detail here and below smells like an LLM having been > unleashed. I don't think it's necessary wasting people's time with > details that are also shown in the diff. > If there is something surprising, it's worth mentioning here, but in > general I would expect a simple defconfig and .dts post to not cause > much problems. > > Cheers, > Andre > > > Validation (including the earlier board build): > > - Cross-build the revised orangepi_3_lts_defconfig with GCC 14.2.0 and > > TF-A v2.13.0 (sun50i_h6, DEBUG=1). Cross-build the original > > orangepi_3_defconfig too; its handoff hook remains an empty return. > > - Verify the LTS DT model, Ethernet configuration and final combined FIT. > > - Match the new defconfig to savedefconfig output. > > - Check both patches with checkpatch.pl --strict, excluding only the > > FILE_PATH_CHANGES warning: board/sunxi/MAINTAINERS is updated, but > > checkpatch recognises only a top-level MAINTAINERS change. > > > > The GPIO LEDs are enabled in main U-Boot, with green PL7 on and red > > PL4 off. The early SPL LED is disabled. At the OS handoff immediately > > after "Starting kernel ...", the LTS-specific board_quiesce_devices() > > hook switches green off and red on. Linux's LED defaults retain those > > states; diagnostic userspace clears both before starting the shell. > > Cristian confirmed this sequence on the corrected 60d788d7056c build: > > green in main U-Boot, red at the kernel handoff, both off in the shell. > > > > On-board testing of the pre-LED-change U-Boot 2026.10-gc8f95275fc8f: > > - Reaches the console and reports 2 GiB RAM. > > - Enumerates SD at 50 MHz/4-bit and eMMC at 52 MHz/8-bit. > > - Obtains a DHCP lease and pings two LAN hosts successfully. The control > > device tree confirms phy-mode = "rgmii-id". > > - Enumerates EHCI, OHCI and XHCI root hubs. External USB peripherals have > > not been tested. > > > > The submitted v1 (11d60b7149e6) also loaded a mainline Linux Image, DTB > > and initramfs from SD and reached a RAM shell with four CPUs and 2 GiB > > RAM. Its test-kernel configuration caused a PMIC I2C failure; Linux > > peripheral validation remained pending in that first test. > > > > The v2 build (f5a667df5ada) now boots Linux with CONFIG_MODULES=y into > > the RAM shell. Linux initializes the AXP806, enumerates SD and eMMC, > > registers eth0 and all USB host root hubs. The YT8531 PHY links at > > 1 Gbps/full duplex; DHCP, DNS and an IPv6 Internet ping passed on that > > build. Two direct 8 MiB reads from each of SD and eMMC compared > > identical under Linux. Two cold starts with LED, shell and Ethernet > > checks were confirmed on the final build. Both LEDs were confirmed off > > in the diagnostic shell. Sustained transfer, external USB peripherals > > and HDMI remain untested. > > The local test builds omit optional Crust firmware; suspend is not tested. > > > > Signed-off-by: Cristian Filipescu <[email protected]> > > --- > > Changes in v2: > > - Enable GPIO LEDs in main U-Boot; drop early SPL status activation. > > - Add board-specific U-Boot DT overrides: green on, red off. > > - Set shared Linux defaults to retain green off and red on. > > - Switch green off and red on in the LTS OS handoff hook at Starting kernel. > > - Retain the original two-patch DTS/defconfig series structure. > > - Link to v1: > > https://patch.msgid.link/[email protected] > > > > To: [email protected] > > Cc: Tom Rini <[email protected]> > > Cc: Ilias Apalodimas <[email protected]> > > Cc: Cristian Filipescu <[email protected]> > > Cc: Andre Przywara <[email protected]> > > Cc: Hans de Goede <[email protected]> > > Cc: Jernej Skrabec <[email protected]> > > > > --- > > Cristian Filipescu (1): > > sunxi: Add Orange Pi 3 LTS defconfig > > > > Jernej Skrabec (1): > > arm: dts: sunxi: Add Orange Pi 3 LTS device tree > > > > arch/arm/dts/Makefile | 1 + > > arch/arm/dts/sun50i-h6-orangepi-3-lts-u-boot.dtsi | 11 + > > arch/arm/dts/sun50i-h6-orangepi-3-lts.dts | 362 > > ++++++++++++++++++++++ > > board/sunxi/MAINTAINERS | 7 + > > board/sunxi/board.c | 19 ++ > > configs/orangepi_3_lts_defconfig | 19 ++ > > 6 files changed, 419 insertions(+) > > --- > > base-commit: 62c9805d8d57cad407f0295879cc9a6182da464a > > change-id: 20261021-sunxi-orangepi-3-lts-2fa76de6f361 > > > > Best regards, > > -- > > Cristian Filipescu <[email protected]> >
