If a PHY has no driver, the genphy driver is probed/removed directly in
phy_attach/detach. If the PHY's ofnode has an "leds" subnode, then the
LEDs will be (un)registered when probing/removing the genphy driver.
This could occur if the leds are for a non-generic driver that isn't
loaded for whatever reason. Synchronously removing the PHY device in
phy_detach leads to the following deadlock:
rtnl_lock()
ndo_close()
    ...
    phy_detach()
        phy_remove()
            phy_leds_unregister()
                led_classdev_unregister()
                    led_trigger_set()
                        netdev_trigger_deactivate()
                            unregister_netdevice_notifier()
                                rtnl_lock()
There is a corresponding deadlock on the open/register side of things
(and that one is reported by lockdep), but it requires a race while this
one is deterministic.
Generic PHYs do not support LEDs anyway, so don't bother registering
them.
Fixes: 01e5b728e9e4 ("net: phy: Add a binding for PHY LEDs")
Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
Link: https://patch.msgid.link/20250707195803.666097-1-sean.anderson@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
        /* Get the LEDs from the device tree, and instantiate standard
         * LEDs for them.
         */
-       if (IS_ENABLED(CONFIG_PHYLIB_LEDS))
+       if (IS_ENABLED(CONFIG_PHYLIB_LEDS) && !phy_driver_is_genphy(phydev) &&
+           !phy_driver_is_genphy_10g(phydev))
                err = of_phy_leds(phydev);
 
 out:
 
        cancel_delayed_work_sync(&phydev->state_queue);
 
-       if (IS_ENABLED(CONFIG_PHYLIB_LEDS))
+       if (IS_ENABLED(CONFIG_PHYLIB_LEDS) && !phy_driver_is_genphy(phydev) &&
+           !phy_driver_is_genphy_10g(phydev))
                phy_leds_unregister(phydev);
 
        phydev->state = PHY_DOWN;