Re: [PATCH 00/10] phy/drm: amlogic: Extract Meson HDMI TX PHY into a dedicated driver

From: Neil Armstrong

Date: Thu Nov 19 2026 - 02:58:30 EST


On 11/18/26 20:35, linux-kernel-dev@xxxxxxxx wrote:
On 11/18/26 13:38, Neil Armstrong wrote:
Hi,

On 11/17/26 20:33, Ronald Claveau via B4 Relay wrote:
This series moves the Amlogic Meson HDMI TX PHY handling (covering GXBB,
GXL, GXM, and G12A SoCs) out of the meson_dw_hdmi DRM driver and into a
dedicated generic PHY driver.

Until now, meson_dw_hdmi directly manipulated the HHI system controller
registers (HHI_HDMI_PHY_CNTL0..5) to configure analog parameters, toggle
resets, and manage power states, as well as manually ungating HDMI
clocks and power domains already described in the DT.

By extracting this logic into a standalone PHY driver under
drivers/phy/amlogic/:
- Hardware representation in the DT is cleaner, modeling the HDMI TX PHY
   as a child node of the HHI syscon referenced via standard "phys" and
   "phy-names" properties.
- The DRM bridge driver relies on the standard generic PHY API
   (phy_init, phy_configure, phy_power_on/off, phy_exit) rather than
   touching syscon registers directly.
- Legacy DTs lacking the "phys" property remain supported via a
   temporary fallback helper (meson_hdmi_phy_create_legacy) to preserve
   backward compatibility to be removed on next phase of component
   removal.

Overview of the series:
- Patches 1-3: Fix compatible matching in meson-gx-hhi-sysctrl bindings,
   add the new amlogic,meson-hdmi-tx-phy DT schema, and allow the
   "hdmi-phy" child node under the HHI syscon.
- Patch 4: Add the new Meson HDMI TX PHY driver implementing the HDMI
   PHY configuration ops, along with the temporary legacy creation
   helper.
- Patch 5: Document the optional "phys" and "phy-names" properties in
   the amlogic,meson-dw-hdmi binding.
- Patches 6-7: Clean up manual HHI clock/power-domain writes in
   meson_dw_hdmi (relying on runtime PM instead) and convert the driver
   to use the generic PHY framework.
- Patches 8-10: Add the "hdmi-phy" child node and wire up the "phys"
   phandle for G12, GXBB, and GXL Device Trees.

Tested on:
- Khadas VIM1 (GXL)
- dtbs_check and dt_binding_check run cleanly with no new warnings.

Signed-off-by: Ronald Claveau <linux-kernel-dev@xxxxxxxx>
---
Ronald Claveau (10):
       dt-bindings: soc: amlogic: meson-gx-hhi-sysctrl: Fix compatible
matching in conditionals
       dt-bindings: phy: amlogic: Add Meson HDMI TX PHY binding
       dt-bindings: soc: amlogic: meson-gx-hhi-sysctrl: Add hdmi-phy
child node
       phy: amlogic: Add Meson HDMI TX PHY driver
       dt-bindings: display: amlogic,meson-dw-hdmi: Add phys and phy-
names
       drm/meson: dw-hdmi: Drop manual HHI clock and memory PD setup
       drm/meson: dw-hdmi: Use the generic PHY framework for the HDMI
TX PHY
       arm64: dts: amlogic: Add HDMI TX PHY node for G12
       arm64: dts: amlogic: Add HDMI TX PHY node for GXBB
       arm64: dts: amlogic: Add HDMI TX PHY node for GXL

  .../bindings/display/amlogic,meson-dw-hdmi.yaml    |   8 +
  .../bindings/phy/amlogic,meson-hdmi-tx-phy.yaml    |  33 +++
  .../soc/amlogic/amlogic,meson-gx-hhi-sysctrl.yaml  |  43 ++-
  arch/arm64/boot/dts/amlogic/meson-g12-common.dtsi  |   7 +
  arch/arm64/boot/dts/amlogic/meson-gxbb.dtsi        |   7 +
  arch/arm64/boot/dts/amlogic/meson-gxl.dtsi         |   7 +
  drivers/gpu/drm/meson/Kconfig                      |   1 +
  drivers/gpu/drm/meson/meson_dw_hdmi.c              | 203 ++++++--------
  drivers/phy/amlogic/Kconfig                        |  11 +
  drivers/phy/amlogic/Makefile                       |   1 +
  drivers/phy/amlogic/phy-meson-hdmi-tx.c            | 309 +++++++++++
++++++++++
  include/linux/phy/phy-meson-hdmi-tx.h              |  19 ++
  12 files changed, 513 insertions(+), 136 deletions(-)
---
base-commit: 6d582d78fad9d828e575ca0cdcff8ea47a77220b
change-id: 20261115-meson-drm-create-hdmitx-phy-4ac40a654d09

Best regards,
--
Ronald Claveau <linux-kernel-dev@xxxxxxxx>

Overall it's very good, but I would've prefered a more sequencial move
to the PHY driver, with an exact copy of the code and simplification
towards the current PHY code.

Anyway, the sashiko review comments are good, especially the one on
patch 6 to keep the driver working on old DTs, sorry my bad...

Neil

Thanks for your feedback, I am analyzing/testing sashiko feedback.

About the suspend/resume unbind/bind, I do not have any issue on my
side, but I can change the orders.

About patch 6 and the old DTs, I will reintroduce them with the legacy
check. Do you know what would be a good timeline to add a require on
those elements and drop the fallback code ?


The problem is not necessarily the kernel DT but the U-Boot DT which
could be used by the kernel.

The U-Boot DT was using v6.2 before v2024.07 and after it uses the synchronized
DT from Linux, which coincides with the v6.11 DT which contains the DT clocks
changes.

It seems most of the SBCs distro uses recent U-Boot versions, and some nice distro
ships older U-Boot version, so it could break on those if a new kernel was deployed.

But, U-Boot already set HHI_HDMI_CLK_CNTL and HHI_MEM_PD_REG0 when probing dw_hdmi
since 2019.04 so considering that I thing the risk is super low.

Add this to your commit message and send a v2.

Neil