Re: [PATCH] server: Use SO_BINDTODEVICE in bind_to_index() if possible.

[Paul Gofman]

On 10/12/21 21:50, Erich E. Hoover wrote:
> On 10/12/21 21:06, Paul Gofman wrote:
> > Bug 50499 is not related, that one is
> > about SO_REUSEADDR explicitly used by an app (and not set by us to
> > workaround broadcast issue) and behaving differently on Windows and
> > Linux WRT which socket receives packets.
>
> This is correct, though the solution is "the same".  The SO_REUSEADDR
> behavior is easier to implement though, it just involves applying a
> SO_ATTACH_REUSEPORT_CBPF rule that always returns zero (zero is
> oldest).

I am afraid it might be not that simple. If you close the oldest socket 
the last one will take its place which is probably not the one which 
should now start receiving all the packets.