Your placement is the only possible location as the code currently stands, but I would like to suggest that of_set_phy_eee_broken() should _not_ be calling linkmode_zero(modes), and we should be able to set phydev->eee_disabled_modes in the .get_features() method.
Andrew, would you agree?
I dug back through the git history. Originally, the code would read a value from DT which was literally used as a mask against the value in MDIO_MMD_AN. If the mask was not zero, it was applied. That later got converted to a collection of Boolean DT properties, one per link speed, and the mask was created as a collection of |= statements, with the initial value being 0, and the result assigned to phydev->eee_broken_modes. It would of been possible at that stage to do phydev->eee_broken_modes |=, but it guess a local variable was used to keep the lines shorter. The u32 then got converted to a linux bitmap, with the initial = 0; replaced with a linkmode_zero().
I don't see anything in any of the commit messages to indicate there was a reason to initialise phydev->eee_broken_modes to 0 before parsing the DT properties.
So i think it should be O.K. to remove the linkmode_zero(). For broken PHYs like this, the earlier we mask out the broken behaviour the better.
Andrew