[RFC PATCH v4 2/4] dpll: Add DPLL framework base functions

[Jiri Pirko]

Tue, Dec 06, 2022 at 01:19:33AM CET, kuba at kernel.org wrote:
>On Mon, 5 Dec 2022 11:32:04 +0100 Jiri Pirko wrote:
>> >> I believe we should do it only the other way around. Assign
>> >> dpll_pin pointer to struct net_device and expose this over new attr
>> >> IFLA_DPLL_PIN over RT netlink.  
>> >
>> >The ID table is global, what's the relationship between DPLLs
>> >and net namespaces? We tie DPLLs to a devlink instance which
>> >has a namespace? We pretend namespaces don't exist? :S  
>> 
>> Well, if would be odd to put dpll itself into a namespace. It might not
>> have anything to do with networking, for example in case of ptp_ocp.
>> What would mean for a dpll to be in a net namespace?
>
>Yeah, that's a slightly tricky one. We'd probably need some form 
>of second order association. Easiest if we link it to a devlink
>instance, I reckon. The OCP clock card does not have netdevs so we
>can't follow the namespace of netdevs (which would be the second
>option).

Why do we need this association at all?