On Sun, 2006-06-11 at 16:11 +0200, nigel henry wrote: > On Sunday 11 June 2006 13:31, Tim wrote: > > On Sun, 2006-06-11 at 12:11 +0100, Leon wrote: > > > What's the clean way of preventing the kernel from loading ipv6 > > > module? I don't need ipv6 for the moment. > > > > One way, on FC4, is to add the following alias line into > > your /etc/modprobe.conf file: > > > > alias net-pf-10 off > Hi Tim. Just out of interest, how does "net-pf-10" reference just the ipv6 > part of the "S10network" shellscript in /etc/rc.d/rc5.d ? No connection between the two. The alias for net-pf-10 is referring to the network protocol family (pf) 10. If you look in /usr/include/linux/socket.h you will find that AF_INET6 (Address Family INET6) is defined as "10" and PF_INET6 is defined as AF_INET6. When an application (in this case, one of the configuration utilities, either ifconfig or ip or both) references PF_INET6 or AF_INET6, in a system call with a PF_ or AF_ parameter, the kernel uses net-pf-10 to load the appropriate protocol (IPv6) to support that address/protocol family (10). So it then kmod loads it for you if it's not already there. No scripts involved. That's why you can't just turn it off by saying "NETWORKING_IPV6=no" (which happens to be the default anyways) in /etc/sysconfig/network. It's not being done by the network scripts. It's being done by the kernel in response to a protocol/address family reference in one of the configuration utilities. This is at the heart of why the default is broken in many ways. IPv6 is NOT enabled by default! It should be! It would seriously work better if it were. It's enabled by accident and the proper configuration utilities to set it up properly (setting up the routing tables and defining the appropriate link-local black hole routes and behavior) are not being run. It's loaded and functioning but it hasn't been properly configured. If it's in an IPv6 native environment it will autoconfigure the interface addresses and it will work well enough as an end node, but it still should be set up properly to avoid some of the timeouts and delays that people have reported. I never experienced any of those problems, but I have NETWORK_IPV6=yes and it's set up correctly. BTW... I just noticed that it seems in the last few weeks (or, maybe the last couple of months) the default got changed for the network preference. Before, if IPv6 was available and a site had both an IPv6 address and an IPv4 address, IPv6 would be used preferentially over IPv4. This was good. I often see BETTER latency between the US and Europe on IPv6 than on IPv4 (thanks to some v6 backbones out there plus cleaner routing). Lately though, I see that the preference has been switched and it will choose the IPv4 address over the IPv6 address if a site has both. :-( So, when I browse to www.kame.net it says I'm connecting over IPv4 now, instead of IPv6, and all I see is the static Kame turtle. If I connect to www.sixxs.net, or IPv6style.jp, or DeepSpac6, they all report an IPv4 connection. All those sites have both v4 and v6 addresses and I was connecting to them on v6 for ages (last 5 years or so) until just a few weeks ago (by my best guess). I only noticed it was no longer happening last week and I haven't found any changelog announcing the change in preference. It's definitely been since the beginning of the year. I've tried the resolver option in /etc/resolve.conf but that didn't fix the problem. Anyone recall where the v6/v4 preference option is stored? I know it's handled down in glibc somewhere, and I recall reading about an option to define that preference, I just don't recall where. I would like to set it back the way it was. Oh... For those who missed it... Comcast gave a presentation, http://www.nanog.org/mtg-0606/pdf/alain-durand.pdf, a week ago at Nanog (the North American Network Operators Group) about managing 100+ million IP addresses (their current need for management addresses, which can not be NAT'ed, for managing their broadband facility with all the cable modems and set-top boxes and VoIP devices). They exhausted the entire 10.* net way back in 2005 and are now deploying IPv6 in their infrastructure to manage new devices and new segments (and only over IPv6, no dual stack - if it can be managed over IPv6 they won't even give the device a v4 address). New cable modems are being managed over IPv6 even if they are delivering IPv4 to the end customer. Plans for delivering IPv6 to the end customer is still down the road yet but is on their charts. So, if you want to keep playing with that cable modem, you better bone up on IPv6 and if you don't think you've got IPv6, you could be surprised. > For earlier versions (FC1,2,and3) I normally just added modules I did not want > loaded at bootup to hotplugs blacklist, stuff like the OSS modules "audio" > and "usb_midi" when using Alsa. But hotplug is no longer being used on FC5. > > I have made changes to /etc/modprobe.conf before, but only for the load > ordering of Alsa soundcard modules, snd-emu10k1, snd-usb-audio, etc, and > these are written just as they appear in /sbin/lsmod. /sbib/lsmod shows ipv6 > as a module, which is why I ask about the specific syntax you use to have it > set to "off" at bootup. > > Nigel. > > > > > > -- > > (Currently running FC4, occasionally trying FC5.) > > > > Don't send private replies to my address, the mailbox is ignored. > > I read messages from the public lists. Regards, Mike -- Michael H. Warfield (AI4NB) | (770) 985-6132 | mhw@xxxxxxxxxxxx /\/\|=mhw=|\/\/ | (678) 463-0932 | http://www.wittsend.com/mhw/ NIC whois: MHW9 | An optimist believes we live in the best of all PGP Key: 0xDF1DD471 | possible worlds. A pessimist is sure of it!
Attachment:
signature.asc
Description: This is a digitally signed message part
