[PATCH net-next v15 0/9] net: stmmac: qcom-ethqos: add support for SCMI power domains

[PATCH net-next v15 0/9] net: stmmac: qcom-ethqos: add support for SCMI power domains

[Bartosz Golaszewski]

This iteration addresses several issues raised by Sashiko. Some are not
addressed as they were false positives (nothing in the PHY driver ever
calls phy_runtime_get() on the parent device) or I decided they make no
sense - like the suggestion to remove the power-domain-names property
because there's only one power domain.

--

Add support for the firmware-managed variant of the DesignWare MAC on
the sa8255p platform. This series contains new DT bindings and driver
changes required to support the MAC in the STMMAC driver.

It also reorganizes the ethqos code quite a bit to make the introduction
of power domains into the driver a bit easier on the eye.

The DTS changes will go in separately.

Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
---
Changes in v15:
- Squash patches 6 and 7 to avoid a potential hang-on-resume regression
  (Sashiko)
- Fix the AUTONEG_DISABLE early return path in forced mode (Sashiko)
- Fix Kconfig dependencies on generic power domains (Sashiko)
- Set the performance level back to 0 on unbind in the PHY driver
  (Sashiko)
- Use devm_pm_runtime_enable() where applicable
- Link to v14: https://patch.msgid.link/20260908-qcom-sa8255p-emac-v14-0-ec212b747a7d@oss.qualcomm.com

Changes in v14:
- let genpd's own runtime suspend/resume callbacks own the perf state entirely
  by calling dev_pm_genpd_set_performance_state() in set_mode()
  unconditionally in the sgmii PHY driver
- drop usleep_range() from the PHY driver when setting the state as the level
  setting is already synchronous
- add missing power-domains limits for ma35d1/k3-dwmac
- fix ordering of clearing the sgmii loopback
- configure the wrapper clocks synchronously in scmi setup
- Link to v13: https://patch.msgid.link/20260713-qcom-sa8255p-emac-v13-0-119f8699ef8e@oss.qualcomm.com

Changes in v13:
- Reapply OPP on speed change in the PHY driver
- Use linux/device-id/of.h instead of mod_devicetable.h
- Fix the "invalid PTP clock rate" warning on sa8255p
- Add commits necessary to allow correct link speed change with SCMI
- Link to v12: https://patch.msgid.link/20260706-qcom-sa8255p-emac-v12-0-e3ab1ecf2901@oss.qualcomm.com

Changes in v12:
- Drop the power-domain-names from the SGMII PHY DT bindings
- Add missing Kconfig dependency on PM for the new SGMII PHY driver
- Link to v11: https://patch.msgid.link/20260629-qcom-sa8255p-emac-v11-0-1b7fb95b51f9@oss.qualcomm.com

Changes in v11:
- Take a new approach: add a dedicated driver for the firmware-managed
  SGMII PHY and simplify changes made to the MAC driver
- Link to v10: https://patch.msgid.link/20260323-qcom-sa8255p-emac-v10-0-79302b238a16@oss.qualcomm.com

Changes in v10:
- Fix unit address in DT example
- Link to v9: https://patch.msgid.link/20260316-qcom-sa8255p-emac-v9-0-c58934e76ff2@oss.qualcomm.com

Changes in v9:
- Rebase on top of current linux-next again
- Link to v8: https://patch.msgid.link/20260311-qcom-sa8255p-emac-v8-0-58227bcf1018@oss.qualcomm.com

Changes in v8:
- Rebase on top of recent changes in linux-next which required an
  extensive rework
- Drop partial R-b tags
- Link to v7: https://patch.msgid.link/20260306-qcom-sa8255p-emac-v7-0-d6a3013094b7@oss.qualcomm.com

Changes in v7:
- Restored the correct authorship after learning git uses .mailmap for
  the --author switch
- Rebased on top of changes from Russell
- Fixed resource management issues in error paths
- Link to v6: https://lore.kernel.org/r/20260112-qcom-sa8255p-emac-v6-0-86a3d4b2ad83@oss.qualcomm.com

Changes in v6:
- Fix $id value in the bindings
- Drop patch 3/8 from the series
- Update init/exit callback signatures
- Link to v5: https://lore.kernel.org/r/20251107-qcom-sa8255p-emac-v5-0-01d3e3aaf388@linaro.org
- Link to v6: https://lore.kernel.org/r/20251219-qcom-sa8255p-emac-v6-0-487f1082461e@oss.qualcomm.com

Changes in v5:
- Name the DT binding document after the new compatbile
- Add missing space
- Make the power-domains limits stricter
- Link to v4: https://lore.kernel.org/r/20251104-qcom-sa8255p-emac-v4-0-f76660087cea@linaro.org

Changes in v4:
- Remove the phys property from the SCMI bindings
- Mark the power-domain-names property as required
- Set maxItems for power-domains to 1 for all existing bindings to
  maintain the current requirements after modifying the value in the
  top-level document
- Link to v3: https://lore.kernel.org/r/20251027-qcom-sa8255p-emac-v3-0-75767b9230ab@linaro.org

Changes in v3:
- Drop 'power' and 'perf' prefixes from power domain names
- Rebase on top of Russell's changes to dwmac
- Rebase on top of even more changes from Russell that are not yet
  in next (E1vB6ld-0000000BIPy-2Qi4@rmk-PC.armlinux.org.uk)
- Link to v2: https://lore.kernel.org/all/20251008-qcom-sa8255p-emac-v2-0-92bc29309fce@linaro.org/

Changes in v2:
- Fix the power-domains property in DT bindings
- Rework the DT bindings example
- Drop the DTS patch, it will go upstream separately
- Link to v1: https://lore.kernel.org/r/20250910-qcom-sa8255p-emac-v1-0-32a79cf1e668@linaro.org

---
Bartosz Golaszewski (9):
      net: phy: aquantia: fix system interface type not updated in forced mode
      dt-bindings: phy: document the serdes PHY on sa8255p
      phy: qcom: add the SGMII SerDes PHY driver for SCMI systems
      dt-bindings: net: qcom: document the ethqos device for SCMI-based systems
      net: stmmac: qcom-ethqos: set serdes mode before powerup
      net: stmmac: qcom-ethqos: update phy_mode to the resolved interface in mac_finish()
      net: stmmac: qcom-ethqos: reuse the address of ethqos_emac_driver_data
      net: stmmac: qcom-ethqos: factor out linux-level setup into a separate function
      net: stmmac: qcom-ethqos: add support for sa8255p

 .../bindings/net/allwinner,sun7i-a20-gmac.yaml     |   3 +
 .../bindings/net/altr,socfpga-stmmac.yaml          |   3 +
 .../bindings/net/amlogic,meson-dwmac.yaml          |   3 +
 .../devicetree/bindings/net/eswin,eic7700-eth.yaml |   3 +
 .../devicetree/bindings/net/intel,dwmac-plat.yaml  |   3 +
 .../bindings/net/loongson,ls1b-gmac.yaml           |   3 +
 .../bindings/net/loongson,ls1c-emac.yaml           |   3 +
 .../bindings/net/nuvoton,ma35d1-dwmac.yaml         |   3 +
 .../devicetree/bindings/net/nxp,dwmac-imx.yaml     |   3 +
 .../devicetree/bindings/net/nxp,lpc1850-dwmac.yaml |   3 +
 .../devicetree/bindings/net/nxp,s32-dwmac.yaml     |   3 +
 .../devicetree/bindings/net/qcom,ethqos.yaml       |   3 +
 .../bindings/net/qcom,sa8255p-ethqos.yaml          | 107 +++++++++
 .../devicetree/bindings/net/renesas,rzn1-gmac.yaml |   3 +
 .../bindings/net/renesas,rzv2h-gbeth.yaml          |   3 +
 .../devicetree/bindings/net/rockchip-dwmac.yaml    |   3 +
 .../devicetree/bindings/net/snps,dwmac.yaml        |   5 +-
 .../bindings/net/sophgo,cv1800b-dwmac.yaml         |   3 +
 .../bindings/net/sophgo,sg2044-dwmac.yaml          |   3 +
 .../devicetree/bindings/net/spacemit,k3-dwmac.yaml |   3 +
 .../bindings/net/starfive,jh7110-dwmac.yaml        |   3 +
 .../devicetree/bindings/net/stm32-dwmac.yaml       |   3 +
 .../devicetree/bindings/net/tesla,fsd-ethqos.yaml  |   3 +
 .../devicetree/bindings/net/thead,th1520-gmac.yaml |   3 +
 .../bindings/net/toshiba,visconti-dwmac.yaml       |   3 +
 .../bindings/phy/qcom,sa8255p-dwmac-sgmii-phy.yaml |  50 ++++
 MAINTAINERS                                        |   1 +
 .../ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c    | 263 ++++++++++++++++-----
 drivers/net/phy/aquantia/aquantia_main.c           |   5 +-
 drivers/phy/qualcomm/Kconfig                       |  10 +
 drivers/phy/qualcomm/Makefile                      |   1 +
 drivers/phy/qualcomm/phy-qcom-sgmii-eth-scmi.c     | 117 +++++++++
 32 files changed, 570 insertions(+), 58 deletions(-)
---
base-commit: ab8dad6c8262729415164de729ba533466fc3649
change-id: 20250704-qcom-sa8255p-emac-8460235ac512

Best regards,
-- 
Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>

[PATCH net-next v15 1/9] net: phy: aquantia: fix system interface type not updated in forced mode

[Bartosz Golaszewski]

aqr_gen1_read_status() decodes the MDIO_PHYXS_VEND_IF_STATUS register
to determine which SerDes interface the PHY is currently using on its
system side and stores the result in phydev->interface. phylink relies
on this value to configure the MAC.

The autoneg == AUTONEG_DISABLE check is not correct:
MDIO_PHYXS_VEND_IF_STATUS is set by the PHY firmware based on the
negotiated link speed, not based on whether autoneg was used to reach
it. When the link comes up at 1G in forced mode, the register correctly
reads SGMII, but the early return prevents phydev->interface from being
updated. It stays at whatever value it held before (typically 2500BASE-X
from the initial autoneg run), so phylink configures the MAC for the
wrong interface and the link cannot come up.

Remove the autoneg guard so that the system interface type is always
decoded when the link is up.

Cc: stable@vger.kernel.org
Fixes: 110a2432c520 ("net: phy: aquantia: add downshift support")
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
---
 drivers/net/phy/aquantia/aquantia_main.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drivers/net/phy/aquantia/aquantia_main.c b/drivers/net/phy/aquantia/aquantia_main.c
index e95d391a6dc233879ae1fb2a97758b3df9600ce5..096407bebd52a8e644084c3e286fb5dbcfa3f598 100644
--- a/drivers/net/phy/aquantia/aquantia_main.c
+++ b/drivers/net/phy/aquantia/aquantia_main.c
@@ -544,7 +544,7 @@ static int aqr_gen1_read_status(struct phy_device *phydev)
 	if (ret)
 		return ret;
 
-	if (!phydev->link || phydev->autoneg == AUTONEG_DISABLE)
+	if (!phydev->link)
 		return 0;
 
 	/* The status register is not immediately correct on line side link up.
@@ -594,6 +594,9 @@ static int aqr_gen1_read_status(struct phy_device *phydev)
 
 	phydev->interface = aqr_translate_interface(phydev, interface);
 
+	if (phydev->autoneg == AUTONEG_DISABLE)
+		return 0;
+
 	/* Read rate from vendor register */
 	return aqr_gen1_read_rate(phydev);
 }

-- 
2.47.3

Re: [PATCH net-next v15 1/9] net: phy: aquantia: fix system interface type not updated in forced mode

[Jakub Kicinski]

On Wed, 23 Sep 2026 17:14:35 +0200 Bartosz Golaszewski wrote:
> aqr_gen1_read_status() decodes the MDIO_PHYXS_VEND_IF_STATUS register
> to determine which SerDes interface the PHY is currently using on its
> system side and stores the result in phydev->interface. phylink relies
> on this value to configure the MAC.
> 
> The autoneg == AUTONEG_DISABLE check is not correct:
> MDIO_PHYXS_VEND_IF_STATUS is set by the PHY firmware based on the
> negotiated link speed, not based on whether autoneg was used to reach
> it. When the link comes up at 1G in forced mode, the register correctly
> reads SGMII, but the early return prevents phydev->interface from being
> updated. It stays at whatever value it held before (typically 2500BASE-X
> from the initial autoneg run), so phylink configures the MAC for the
> wrong interface and the link cannot come up.
> 
> Remove the autoneg guard so that the system interface type is always
> decoded when the link is up.
> 
> Cc: stable@vger.kernel.org
> Fixes: 110a2432c520 ("net: phy: aquantia: add downshift support")
> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>

Sounds like a real regression, is it? Should we ship it to Linus
tomorrow so that it makes 7.3? stable + next does not compute for me

Re: [PATCH net-next v15 1/9] net: phy: aquantia: fix system interface type not updated in forced mode

[Bartosz Golaszewski]

On Thu, 1 Oct 2026 01:33:34 +0200, Jakub Kicinski <kuba@kernel.org> said:
> On Wed, 23 Sep 2026 17:14:35 +0200 Bartosz Golaszewski wrote:
>> aqr_gen1_read_status() decodes the MDIO_PHYXS_VEND_IF_STATUS register
>> to determine which SerDes interface the PHY is currently using on its
>> system side and stores the result in phydev->interface. phylink relies
>> on this value to configure the MAC.
>>
>> The autoneg == AUTONEG_DISABLE check is not correct:
>> MDIO_PHYXS_VEND_IF_STATUS is set by the PHY firmware based on the
>> negotiated link speed, not based on whether autoneg was used to reach
>> it. When the link comes up at 1G in forced mode, the register correctly
>> reads SGMII, but the early return prevents phydev->interface from being
>> updated. It stays at whatever value it held before (typically 2500BASE-X
>> from the initial autoneg run), so phylink configures the MAC for the
>> wrong interface and the link cannot come up.
>>
>> Remove the autoneg guard so that the system interface type is always
>> decoded when the link is up.
>>
>> Cc: stable@vger.kernel.org
>> Fixes: 110a2432c520 ("net: phy: aquantia: add downshift support")
>> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
>
> Sounds like a real regression, is it? Should we ship it to Linus
> tomorrow so that it makes 7.3? stable + next does not compute for me
>

Yes, this is independent from the rest of the series and a small fix. It sounds
good to send it upstream ASAP.

Bart

Re: [PATCH net-next v15 0/9] net: stmmac: qcom-ethqos: add support for SCMI power domains

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (main)
by Paolo Abeni <pabeni@redhat.com>:

On Wed, 23 Sep 2026 17:14:34 +0200 you wrote:
> This iteration addresses several issues raised by Sashiko. Some are not
> addressed as they were false positives (nothing in the PHY driver ever
> calls phy_runtime_get() on the parent device) or I decided they make no
> sense - like the suggestion to remove the power-domain-names property
> because there's only one power domain.
> 
> --
> 
> [...]

Here is the summary with links:
  - [net-next,v15,1/9] net: phy: aquantia: fix system interface type not updated in forced mode
    https://git.kernel.org/netdev/net/c/6e0022b5ae3d
  - [net-next,v15,2/9] dt-bindings: phy: document the serdes PHY on sa8255p
    (no matching commit)
  - [net-next,v15,3/9] phy: qcom: add the SGMII SerDes PHY driver for SCMI systems
    (no matching commit)
  - [net-next,v15,4/9] dt-bindings: net: qcom: document the ethqos device for SCMI-based systems
    (no matching commit)
  - [net-next,v15,5/9] net: stmmac: qcom-ethqos: set serdes mode before powerup
    (no matching commit)
  - [net-next,v15,6/9] net: stmmac: qcom-ethqos: update phy_mode to the resolved interface in mac_finish()
    (no matching commit)
  - [net-next,v15,7/9] net: stmmac: qcom-ethqos: reuse the address of ethqos_emac_driver_data
    (no matching commit)
  - [net-next,v15,8/9] net: stmmac: qcom-ethqos: factor out linux-level setup into a separate function
    (no matching commit)
  - [net-next,v15,9/9] net: stmmac: qcom-ethqos: add support for sa8255p
    (no matching commit)

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html