Hi Andrew, On Mon, 17 Feb 2025 14:43:00 +0100 Andrew Lunn <andrew@xxxxxxx> wrote: > > One way to avoid that would be to extract out of phylink/phylib all the > > functions for linkmode handling that aren't tied to phylink/phylib > > directly, but are about managing the capabilities of each interface, > > linkmode, speed, duplex, etc. For phylink, that would be : > > > > phylink_merge_link_mode > > phylink_get_capabilities > > phylink_cap_from_speed_duplex > > phylink_limit_mac_speed > > phylink_caps_to_linkmodes > > phylink_interface_max_speed > > phylink_interface_signal_rate > > phylink_is_empty_linkmode > > phylink_an_mode_str > > phylink_set_port_modes > > ... > > > These would go into linkmode.c/h for example, and we'd have a shared set > > of helpers that we can use in phylink, phylib and phy_port. > > Please be careful with the scope of these. Heiner is going through > phylib and trying to reduce the scope of some of the functions we > exporting in include/linux/phy.h to just being available in > drivers/net/phy. That will help stop MAC drivers abuse them. We should > do the same here, limit what can actually use these helpers to stop > abuse. Can we consider having an header file sitting in drivers/net/phy directly for this kind of things ? Maxime
