Re: trunk(4) shouldn't need to play with a port's if_type

2019-06-11 Thread sven falempin
Some interfaces are marked as busy ( promiscuitous ?) so they can’t be added in two bridges for example, this could be extended and make things more consistent? Just my two cents On Tue, Jun 11, 2019 at 8:27 PM David Gwynne wrote: > I get that trunk ports should not be able to be added to bridg

Re: trunk(4) shouldn't need to play with a port's if_type

2019-06-11 Thread David Gwynne
I get that trunk ports should not be able to be added to bridges or have carp interfaces hanging off them, but I think the value of making the if_type immutable outweighs this usability feature. Especially when you consider that you can have an interface that is already a member of a bridge (or

Re: trunk(4) shouldn't need to play with a port's if_type

2019-06-11 Thread Reyk Floeter
Hi, the initial intention was to differentiate a trunk port from a regular Ethernet interface. As long as an interface is a member of a trunk, it is not a fully featured Ethernet interface. The changed type prevented from using it elsewhere. I‘m not so familiar with the current network stack a

trunk(4) shouldn't need to play with a port's if_type

2019-06-10 Thread David Gwynne
i think trunk(4) is the only thing left in the kernel that modifies an interfaces if_type at runtime. this diff removes that fiddling, so hopefully we can say that if_type is immutable after this. however, while this diff reads well to me, i don't actually know if it works. could someone kick the