On Tuesday 30 Dec 2014 10:10:04 Zesen Qian wrote: > Mick <michaelkintz...@gmail.com> writes: > > On Tuesday 30 Dec 2014 03:27:34 Zesen Qian wrote: > >> Hello list, > >> Sorry for hijacking Rich's thread, resend here. > >> I want to disable IPv6 on an certain interface, a simple google > >> tell me to add one line to /etc/sysctl.d/local.conf > >> net.ipv6.conf.enp4s0.disable_ipv6=1 > >> Simple enough, but the problem I have is that both ipv6 and the > >> dirver of the network card(tg3) is loaded by modules. They 're not > >> loaded when the service sysctl is started, so there's no entry > >> named "net.ipv6.conf.enp4s0", which make sysctl not working. > >> So my question is that, is there any way to disable IPv6 on > >> specific interface, as early as I can? I want to make it early > >> because I don't want to receive any RA to mess up my route table. > >> I guess just adding sysctl to preup() in /etc/conf.d/net should do > >> the trick, but may I make it earlier? > >> Any comment is appreciated. > > > > Unless the kernel knows of the enp4s0 interface and therefore lists it > > under sysctl, I can't see how it can be disabled. You could try > > disabling IPv6 altogether in the kernel, but this may not be what you > > want to achieve. > > > > Alternatively, have a look with modinfo in the module options in the > > unlikely chance that the module has some option which disables IPv6 > > functionality. > > Hello Mick, > The module ipv6 has only three parameters, each of which is "disable > ipv6", "disable ipv6 on all interfaces", "disable autoconf on all > interfaces", none of these is what I want to achieve.
Right, I didn't think that you would want to disable IPv6 completely, but mentioned it for completeness. > Module tg3 has > only on parameters which control the debug level. Yes, I thought it unlikely that a hardware driver would be used to configure IPv6, but it was worth checking. > Currently I just call "sysctl net.ipv6.conf.enp4s0.disable_ipv6=1" in > preup(), which works just fine. I 'm just wondering if there 's some > more "proper" way to do it. Since in my opinion, this should be done at > the first moment when the interface is present, not when the interface > is up. I'm afraid I don't know of any other way to do it. -- Regards, Mick
signature.asc
Description: This is a digitally signed message part.