[PATCH net] net: phy: skip EEE advertisement write when autoneg is disabled

[PATCH net] net: phy: skip EEE advertisement write when autoneg is disabled

[Nerijus Bendžiūnas]

genphy_c45_an_config_eee_aneg() writes the EEE advertisement to the
auto-negotiation device's MMD register space (MDIO_MMD_AN, register
MDIO_AN_EEE_ADV).  These registers are only consumed by the link
partner during auto-negotiation, so writing them while autoneg is
disabled is semantically a no-op.

On at least the Broadcom BCM54213PE PHY, however, the indirect MMD
write triggered from this path while the link is currently forced
(autoneg off) disturbs the receive datapath: the kernel still
reports the link as UP at the configured speed/duplex, but no
frames are received until the link is renegotiated.  Concretely,
running

    ethtool -s eth0 speed 100 duplex full autoneg off
    ethtool --set-eee eth0 eee off

leaves eth0 with TX working and RX completely silent on a
Raspberry Pi 4 / CM4 board (bcmgenet + BCM54213PE in rgmii-rxid).
Switching back to autoneg recovers the link.

Prior to commit f26a29a038ee ("net: phy: ensure that genphy_c45_an_config_eee_aneg() sees new value of phydev->eee_cfg.eee_enabled"),
the disable path was effectively a no-op because the helper read
the stale eee_cfg.eee_enabled, so the underlying PHY behavior never
surfaced.

Bisected on rpi-6.12.y between commits 83943264 (good) and
effcbc88 (bad) to f26a29a038ee.

Fixes: f26a29a038ee ("net: phy: ensure that genphy_c45_an_config_eee_aneg() sees new value of phydev->eee_cfg.eee_enabled")
Cc: stable@vger.kernel.org
Signed-off-by: Nerijus Bendžiūnas <nerijus.bendziunas@gmail.com>
---
 drivers/net/phy/phy-c45.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/phy/phy-c45.c b/drivers/net/phy/phy-c45.c
index d48aa7231b37..65de6a5fa204 100644
--- a/drivers/net/phy/phy-c45.c
+++ b/drivers/net/phy/phy-c45.c
@@ -940,6 +940,10 @@ EXPORT_SYMBOL_GPL(genphy_c45_read_eee_abilities);
  */
 int genphy_c45_an_config_eee_aneg(struct phy_device *phydev)
 {
+	/* MMD AN advertisements are only consumed during autoneg. */
+	if (phydev->autoneg == AUTONEG_DISABLE)
+		return 0;
+
 	if (!phydev->eee_cfg.eee_enabled) {
 		__ETHTOOL_DECLARE_LINK_MODE_MASK(adv) = {};
 
-- 
2.54.0

Re: [PATCH net] net: phy: skip EEE advertisement write when autoneg is disabled

[Andrew Lunn]

> On at least the Broadcom BCM54213PE PHY, however, the indirect MMD
> write triggered from this path while the link is currently forced
> (autoneg off) disturbs the receive datapath


> +	/* MMD AN advertisements are only consumed during autoneg. */
> +	if (phydev->autoneg == AUTONEG_DISABLE)
> +		return 0;

You are not adding this condition because it is not consumed, but
because it is consumed and then bad things happen. Please could you
make the comment accurate.

    Andrew

---
pw-bot: cr

[PATCH net v2] net: phy: skip EEE advertisement write when autoneg is disabled

[Nerijus Bendžiūnas]

genphy_c45_an_config_eee_aneg() writes the EEE advertisement to the
auto-negotiation device's MMD register space (MDIO_MMD_AN, register
MDIO_AN_EEE_ADV).  These registers are read by the link partner only
during auto-negotiation, so writing them while autoneg is disabled
cannot influence the link.  On some PHYs (e.g. Broadcom BCM54213PE)
the write nevertheless reaches the chip and disturbs the receive
datapath.

Concretely, running

    ethtool -s eth0 speed 100 duplex full autoneg off
    ethtool --set-eee eth0 eee off

leaves eth0 with TX working and RX completely silent on a
Raspberry Pi 4 / CM4 board (bcmgenet + BCM54213PE in rgmii-rxid).
Switching back to autoneg recovers the link.

Prior to commit f26a29a038ee ("net: phy: ensure that genphy_c45_an_config_eee_aneg() sees new value of phydev->eee_cfg.eee_enabled"),
the disable path was effectively a no-op because the helper read
the stale eee_cfg.eee_enabled, so the underlying PHY behavior never
surfaced.

Bisected on rpi-6.12.y between commits 83943264 (good) and
effcbc88 (bad) to f26a29a038ee.

Fixes: f26a29a038ee ("net: phy: ensure that genphy_c45_an_config_eee_aneg() sees new value of phydev->eee_cfg.eee_enabled")
Cc: stable@vger.kernel.org
Signed-off-by: Nerijus Bendžiūnas <nerijus.bendziunas@gmail.com>
---
Changes since v1:
 - Reworded the inline comment and commit message: the MMD AN
   write is read by the chip with side-effects, not silently
   ignored. Suggested by Andrew Lunn.
 - Tightened the reproducer paragraph (was repeating paragraph 1).

v1: https://lore.kernel.org/netdev/20260516114334.812828-1-nerijus.bendziunas@gmail.com/
 drivers/net/phy/phy-c45.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/drivers/net/phy/phy-c45.c b/drivers/net/phy/phy-c45.c
index d48aa7231b37..126951741428 100644
--- a/drivers/net/phy/phy-c45.c
+++ b/drivers/net/phy/phy-c45.c
@@ -940,6 +940,14 @@ EXPORT_SYMBOL_GPL(genphy_c45_read_eee_abilities);
  */
 int genphy_c45_an_config_eee_aneg(struct phy_device *phydev)
 {
+	/* Writing MMD AN advertisements while autoneg is disabled has no
+	 * effect on link-partner negotiation, but on some PHYs (e.g. the
+	 * Broadcom BCM54213PE) the write itself disturbs the receive
+	 * datapath. Skip it.
+	 */
+	if (phydev->autoneg == AUTONEG_DISABLE)
+		return 0;
+
 	if (!phydev->eee_cfg.eee_enabled) {
 		__ETHTOOL_DECLARE_LINK_MODE_MASK(adv) = {};
 
-- 
2.54.0