support of IPv6 by NFS
Subsystems (2)
I'm working on the support of IPv6 by NFS and the RPC on Linux.
As now preconized for the developing of new networking applications, I have developed a prototype implementation on which I have migrated all the NFS / RPC kernel stack and the user commands to use IPv6 addresses. The IPv4-mapping mechanism is used to assume the backward compatibility for IPv4 addresses which are still the most used. This works but this needs that the kernel has been compiled with IPv6, which is not mandotary. A lot of people in the Linux community do not have experience with IPv6 yet and are not ready to use it. So making it mandatory for NFS, even in a pure IPv4 network, is not easy.
It seems that the most of the major distributions already provide default kernel built with IPv6, but the reference on kernel.org is still providing with the IPv6 support not set; and there are some unwillingness to make mandatory the compilation of the kernel with IPv6 to support NFS.
The problem is not specific to NFS, any networking application written using IPv6 mechanisms for both IPv4 and IPv6 addresses (AF_INET6 socket opened, IPv4 addresses mapped) couldn't work without a kernel built with IPv6.
Are the final users really against the use of kernels built with IPv6 ?
What is preconized on Linux for the support of IPv6 ? The solution described above or the cohabitation of the two modes (struct sockaddr or sockaddr_storage used to contain either struct sockaddr_in or struct sockaddr_in6) with specific processing according to the family of the addresses ?
Regards,
Gilles
Quoted 4 lines · Gilles Quillard
This works but this needs that the kernel has been compiled with IPv6, which is not mandotary. A lot of people in the Linux community do not have experience with IPv6 yet and are not ready to use it. So making it mandatory for NFS, even in a pure IPv4 network, is not easy.
My experience is that IPV6 is extremely difficult to figure out how to set up securely, for the time being, due to lack of connection-sharing. This little fact goes completely unmentioned in ALL of the HowTos. Thank goodness for the USAGI project.
Also one must become an ip6tables expert in order to have a reasonably secure firewall, because ip6tables and 6tables are dead, and Shorewall does not support IPV6 security for some reason. Another deterrant.
And 80% of potential users are behind a cable/DSL 4 NATting router. There is no clarity that it is possible overcome this by either setting to DMZ, or hoping your cablemodem passes protos 41, 50 & 51. Even some tunnel operators do not know this, so I had to figure it out myself. There is no Linux 6to4 UDP tunnelling app, but there should be, because this is such a common problem. (As I understand, Teredo is Winduhs-only, and is not supported by most tunnel operators)
And frankly, most Linux users' only contact with IPV6 has been the DNS AAAA browser delay seemingly inherent in some distros. Although I realize that all of us who run Linux are ostensibly uber-gurus, fact is this is a negative first experience for most, stemming from attempts by distros to encourage ppl to use it with an inoperative function, and without an obvious way to troubleshoot/repair.
These issues contradict assertions that IPV6 is beneficial and easy. If I didn't have a strong motivation and lots of time, I would have chucked early-on. Speaking the actual truth, not propaganda or spin, leads to understanding of the *real* world.
Carl Cook
Re: (usagi-users 03222) Re: support of IPv6 by NFS
Quoted 8 lines · Quantum Scientific
Quoted 4 lines · Gilles Quillard
This works but this needs that the kernel has been compiled with IPv6, which is not mandotary. A lot of people in the Linux community do not have experience with IPv6 yet and are not ready to use it. So making it mandatory for NFS, even in a pure IPv4 network, is not easy.
My experience is that IPV6 is extremely difficult to figure out how to set up securely, for the time being, due to lack of connection-sharing.
NAT is not a firewall. Get that into your brain.
And indeed there is no Linux firewalling code yet, in the mainstream that can do connection tracking. There is no non-EFT Cisco PIX code for this either. The only OS that can do it is the various BSD's.
Quoted 4 lines
And 80% of potential users are behind a cable/DSL 4 NATting router. There is no clarity that it is possible overcome this by either setting to DMZ, or hoping your cablemodem passes protos 41, 50 & 51. Even some tunnel operators do not know this, so I had to figure it out myself.
Freenet6/Hexago have a UDP protocol and SixXS has AYIYA. Works perfectly fine.
In most cases, I know from quite a bit of experience, proto-41 forwarding works very well in most of these DSL boxes.
Quoted 4 lines
There is no Linux 6to4
UDP tunnelling app, but there should be, because this is such a common problem. (As I understand, Teredo is Winduhs-only, and is not supported by most tunnel operators)
The protocol for Teredo is open and can be implemented at will:
http://www-rp.lip6.fr/teredo/
http://www.simphalempin.com/dev/miredo
http://people.via.ecp.fr/~rem/miredo/?C=N;O=D
First couple of hits when doing a google on "Teredo BSD", or for you to click as that might be difficult:
http://www.google.com/search?q=teredo+bsd
Quoted 6 lines
And frankly, most Linux users' only contact with IPV6 has been the DNS AAAA browser delay seemingly inherent in some distros. Although I realize that all of us who run Linux are ostensibly uber-gurus, fact is this is a negative first experience for most, stemming from attempts by distros to encourage ppl to use it with an inoperative function, and without an obvious way to troubleshoot/repair.
I can clearly assume that you are not part of the 'ostensibly uber-gurus' you try to mention.
Quoted 2 lines
These issues contradict assertions that IPV6 is beneficial and easy.
That you don't understand it is your problem ;)
Quoted 4 lines
If I
didn't have a strong motivation and lots of time, I would have chucked early-on. Speaking the actual truth, not propaganda or spin, leads to understanding of the *real* world.
Loads of people seem to have no problem at all with IPv6, getting it up and running and actually using it for a lot of traffic.
That fact that you are only complaining, without doing any actual research, typing two words in google, says enough. You are not even capable of configuring your mailer properly to include your own name, the field is not called "Realname" for nothing...
Greets,
Jeroen
Re: (usagi-users 03222) Re: support of IPv6 by NFS
Quoted 6 lines · Jeroen Massar
Quoted 4 lines
There is no Linux 6to4
UDP tunnelling app, but there should be, because this is such a common problem. (As I understand, Teredo is Winduhs-only, and is not supported by most tunnel operators)The protocol for Teredo is open and can be implemented at will:
Except that it's quite horrible, and it requires a centralized broker, and IIRC it also makes assumptions about the way your NAT implementation assigns ports.
Olaf
Signature · 2 lines
Olaf Kirch | --- o --- Nous sommes du soleil we love when we play [email protected] | / | \ sol.dhoop.naytheet.ah kin.ir.samse.qurax
Re: (usagi-users 03222) Re: support of IPv6 by NFS
Quoted 9 lines · Olaf Kirch
Quoted 6 lines · Jeroen Massar
Quoted 4 lines
There is no Linux 6to4
UDP tunnelling app, but there should be, because this is such a common problem. (As I understand, Teredo is Winduhs-only, and is not supported by most tunnel operators)The protocol for Teredo is open and can be implemented at will:
Except that it's quite horrible,
It needs to be horrible as it needs to cross horrible NAT's.
Quoted 1 line
and it requires a centralized broker,
Doesn't every tunneling method require this? Or is 6to4 anycasted and thus not central?
Do note that you can setup your own Teredo relay, see the docs at the Miredo site for more information.
Quoted 2 lines
and IIRC it also makes assumptions about the way your NAT implementation assigns ports.
It expects a Cone NAT (or was it the other thing?). The functionality for the others where taken out because of 'security' concerns from some people.
Greets,
Jeroen
Re: (usagi-users 03224) Re: support of IPv6 by NFS
In reply to Olaf KirchQuoted 3 lines · Olaf Kirch
Quoted 1 line
The protocol for Teredo is open and can be implemented at will:
Except that it's quite horrible,
Yes, it is, and that's its biggest weakness. NAT traversal is horrible by design. So either you use a point-to-point tunnel over UDP (or TCP, but it is slow), either you end up with something horrible.
Quoted 1 line
and it requires a centralized broker,
Actually, Teredo is much more decentralised than, say, TSP. There could be several Teredo relays among the IPv6 world, from different entities, much like there are currently 6to4 relays.
The only centralized thing is the server whose only purpose is autoconf and NAT traversal.
Quoted 2 lines
and IIRC it also makes assumptions about the way your NAT implementation assigns ports.
Yes, indeed. Unfortunately, the only way to avoid such assumptions is to use point-to-point IPv6 tunnels (or not try to use IPv6 from behind a NAT at all).
Point-to-point tunneling might be fine, but, as far as I know, there is no automatic and registration-less IPv6 point-to-point tunneling solution at the moment :-(
Signature · 1 line
Rémi Denis-Courmont
Re: (usagi-users 03222) Re: support of IPv6 by NFS
In reply to Jeroen MassarQuoted 8 lines · Jeroen Massar
Quoted 7 lines · Quantum Scientific
Quoted 4 lines · Gilles Quillard
This works but this needs that the kernel has been compiled with IPv6, which is not mandotary. A lot of people in the Linux community do not have experience with IPv6 yet and are not ready to use it. So making it mandatory for NFS, even in a pure IPv4 network, is not easy.
My experience is that IPV6 is extremely difficult to figure out how to set
up
Quoted 3 lines
Quoted 1 line
securely, for the time being, due to lack of connection-sharing.
NAT is not a firewall. Get that into your brain.
Jeroen, was this addressed to me, or to Giles? Never mind, it doesn't matter; your words show that you are an uneducated man.
Quoted 3 lines · Jeroen Massar
First couple of hits when doing a google on "Teredo BSD", or for you to click as that might be difficult:
http://www.google.com/search?q=teredo+bsd
...
Quoted 8 lines · Quantum Scientific
Quoted 4 lines
Although I realize that all of us who run Linux are ostensibly uber-gurus, fact is this is a negative first experience for most, stemming from attempts by distros to encourage ppl to use it with an inoperative function, and without an obvious way to troubleshoot/repair.
I can clearly assume that you are not part of the 'ostensibly uber-gurus' you try to mention.
And we can clearly assume that you are petty, and just an asshole. No, I am not a Linux uber-guru. I am a commercial real estate developer, using Linux as a hobby. You may not want my input, but others seem to, judging from emails I've gotten in back-channel about you.
Quoted 6 lines
Loads of people seem to have no problem at all with IPv6, getting it up and running and actually using it for a lot of traffic.
That fact that you are only complaining, without doing any actual research, typing two words in google, says enough. You are not even capable of configuring your mailer properly to include your own name, the field is not called "Realname" for nothing...
Obviously you have not been following my emails, and have simply written your response to carp and ignorantly pretend you are superior in some way. This is no different than noise.
As most here have ascertained, I said the things I have said, as reflective of the experiences of the majority when trying to set up IPV6. If you have a problem with that, you are unable to understand the true issues, and show it with every word.
You will have no more responses from me.
Carl Cook
Re: (usagi-users 03222) Re: support of IPv6 by NFS
Quoted 8 lines · Quantum Scientific
Quoted 7 lines · Jeroen Massar
Quoted 6 lines · Quantum Scientific
Quoted 4 lines · Gilles Quillard
This works but this needs that the kernel has been compiled with IPv6, which is not mandotary. A lot of people in the Linux community do not have experience with IPv6 yet and are not ready to use it. So making it mandatory for NFS, even in a pure IPv4 network, is not easy.
Quoted 8 lines
Quoted 1 line
Quoted 1 line
My experience is that IPV6 is extremely difficult to figure out how to set
up
Quoted 3 lines
Quoted 1 line
securely, for the time being, due to lack of connection-sharing.
NAT is not a firewall. Get that into your brain.
Jeroen, was this addressed to me, or to Giles? Never mind, it doesn't matter; your words show that you are an uneducated man.
As you have read correctly, and how the indentation of the message shows it was a reply to your post. Btw, I am 'educated' enough ;)
Quoted 11 lines · Jeroen Massar
Quoted 3 lines
First couple of hits when doing a google on "Teredo BSD", or for you to click as that might be difficult:
http://www.google.com/search?q=teredo+bsd...
Quoted 6 lines · Quantum Scientific
Quoted 4 lines
Although I realize that all of us who run Linux are ostensibly uber-gurus, fact is this is a negative first experience for most, stemming from attempts by distros to encourage ppl to use it with an inoperative function, and without an obvious way to troubleshoot/repair.
Quoted 4 lines
Quoted 2 lines
I can clearly assume that you are not part of the 'ostensibly uber-gurus' you try to mention.
And we can clearly assume that you are petty, and just an asshole.
Pretty depends on who you ask of course, most ladies will say so fortunately and I don't care about a guys opinion ;)
Quoted 4 lines
No, I am
not a Linux uber-guru. I am a commercial real estate developer, using Linux as a hobby. You may not want my input, but others seem to, judging from emails I've gotten in back-channel about you.
Could you please publish these 'back-channel' communications? I would love to hear comments about me. They are apparently about me, and reading from your sentence you are implying that they are accusing me of a lot of bad things. I don't need names, but please publish them, then everybody knows what it is so bad about me, and even better, then I might learn from these 'issues' that so 'others' might be having.
But I'll just assume you've misjudged me. The fact that you need faul words tells a lot about your reasoning.
Quoted 10 lines
Quoted 6 lines
Loads of people seem to have no problem at all with IPv6, getting it up and running and actually using it for a lot of traffic.
That fact that you are only complaining, without doing any actual research, typing two words in google, says enough. You are not even capable of configuring your mailer properly to include your own name, the field is not called "Realname" for nothing...Obviously you have not been following my emails, and have simply written your response to carp and ignorantly pretend you are superior in some way. This is no different than noise.
Where is your actual technical arguments then? The only few items you named are wellknown and are being addressed already, but things like that take time, especially in an environment where people are doing it on a free basis.
As for the 'superiority', let your 'back-channel' decide on that.
Quoted 2 lines
As most here have ascertained, I said the things I have said, as reflective of the experiences of the majority when trying to set up IPV6.
"most" of the participants of these mailinglists, both of them to which you where at first unable to subscribe, contain people who simply lurk and listen and try to learn from the content that is brought forth here. Claiming 'most' is simply silly.
Quoted 3 lines
If you have a
problem with that, you are unable to understand the true issues, and show it with every word.
The problems are known, but you are trying to misleadingly shove them under the wrong header. Check http://www.v6fix.net as others have also pointed out to you. You might have also wanted to read my mails in where I noted that even Cisco PIX's don't support it yet, unless you get an EFT or the brand sprankling new 7.0 image.
Quoted 1 line
You will have no more responses from me.
Thank you very much, that saves me quite some valuable time trying to reply to posts which are misleading in various ways.
Greets,
Jeroen
Re: (usagi-users 03226) Re: support of IPv6 by NFS
In reply to Quantum ScientificQuoted 9 lines
From: Quantum Scientific <[email protected]>
Quoted 7 lines · Jeroen Massar
Quoted 6 lines · Quantum Scientific
Quoted 4 lines · Gilles Quillard
This works but this needs that the kernel has been compiled with IPv6, which is not mandotary. A lot of people in the Linux community do not have experience with IPv6 yet and are not ready to use it. So making it mandatory for NFS, even in a pure IPv4 network, is not easy.
Quoted 8 lines
Quoted 1 line
Quoted 1 line
My experience is that IPV6 is extremely difficult to figure out how to set
up
Quoted 3 lines
Quoted 1 line
securely, for the time being, due to lack of connection-sharing.
NAT is not a firewall. Get that into your brain.
Jeroen, was this addressed to me, or to Giles? Never mind, it doesn't matter; your words show that you are an uneducated man.
Though I was planning to be more polite, I was going to write a similar message. If you're depending on a firewall as a main defense, you're already dead. If you wish your hosts to be secure, they MUST be secure even if they didn't have a firewall!
The already mentioned approach works quite well. Filter packets with only the SYN bit set, no incoming connections will work, outgoing connections will be unaffected. No state needed. Though important for a firewall, stateful filtering isn't a critical feature to state the IPv6 stack is working.
Signature · 4 lines
(\___(\___(\______ --=> 8-) EHM <=-- ______/)___/)___/) \ ( | [email protected] PGP 8881EF59 | ) / \_ \ | _____ -O #include <stddisclaimer.h> O- _____ | / _/ \___\_|_/82 04 A1 3C C7 B1 37 2A*E3 6E 84 DA 97 4C 40 E6\_|_/___/
Re: (usagi-users 03222) Re: support of IPv6 by NFS
In reply to Jeroen MassarQuoted 5 lines · Jeroen Massar
Quoted 3 lines
My experience is that IPV6 is extremely difficult to figure out how to set up securely, for the time being, due to lack of connection-sharing.
NAT is not a firewall. Get that into your brain.
oh, that was what he meant. I wasn't familiar with the term 'connection sharing'.
I've stated numerous time that IPv6<->IPv6 NAT will only end up in netfilter/iptables over my dead body. IPv4<->IPv6 NAT-PT is a different issue, obviously.
btw, the IETF BEHAVE group is actually demanding that a NAT device does not NAT ipv6 traffic!!
Quoted 2 lines
And indeed there is no Linux firewalling code yet, in the mainstream that can do connection tracking.
still, ip6_conntrack is shipped by commercial distributions like SuSE...
Signature · 4 lines
- Harald Welte <[email protected]> http://gnumonks.org/ ============================================================================ "Privacy in residential applications is a desirable marketing option." (ETSI EN 300 175-7 Ch. A6)
Re: (usagi-users 03249) Re: support of IPv6 by NFS
Quoted 9 lines · Harald Welte
Quoted 5 lines · Jeroen Massar
Quoted 3 lines
My experience is that IPV6 is extremely difficult to figure out how to set up securely, for the time being, due to lack of connection-sharing.
NAT is not a firewall. Get that into your brain.
oh, that was what he meant. I wasn't familiar with the term 'connection sharing'.
That is the Windows term for it ;)
Quoted 2 lines
I've stated numerous time that IPv6<->IPv6 NAT will only end up in netfilter/iptables over my dead body.
Hmmm..... then I guess that I'll have to kill you at some point ;) But I'll leave it to printing out a kernel source and throwing it on your casket in a year or 100 or so.
Quoted 4 lines
IPv4<->IPv6 NAT-PT is a different issue, obviously.
btw, the IETF BEHAVE group is actually demanding that a NAT device does not NAT ipv6 traffic!!
Doing the NAT as in the 'connection sharing', or better said, "rewriting source/dest addresses and packet contents" is evil. But the other method for which we are going to use a "translation of addresses", but on both sides will be very interesting and will cost you your dead body <grin>.
Quoted 4 lines
Quoted 2 lines
And indeed there is no Linux firewalling code yet, in the mainstream that can do connection tracking.
still,
ip6_conntrackis shipped by commercial distributions like SuSE...
There is nothing wrong with connection tracking as that can be used for checking if a certain packet is allowed to come back into the firewall or not, one of the basic principles of stateful firewalling
Greets,
Jeroen
Re: (usagi-users 03222) Re: support of IPv6 by NFS
In reply to Quantum ScientificIn article <[email protected]> (at Tue, 1 Mar 2005 07:44:37 -0600), Quantum Scientific <[email protected]> says:
Quoted 6 lines
And frankly, most Linux users' only contact with IPV6 has been the DNS AAAA browser delay seemingly inherent in some distros. Although I realize that all of us who run Linux are ostensibly uber-gurus, fact is this is a negative first experience for most, stemming from attempts by distros to encourage ppl to use it with an inoperative function, and without an obvious way to troubleshoot/repair.
:
Quoted 4 lines
These issues contradict assertions that IPV6 is beneficial and easy. If I didn't have a strong motivation and lots of time, I would have chucked early-on. Speaking the actual truth, not propaganda or spin, leads to understanding of the *real* world.
Well, we really need to analyse and solve "negative experiences" and berries against IPv6, and the "IPv6-Fix" Project started:
http://v6fix.net
Please report any incidents to <contact AT v6fix DOT net>. We might need to list up pitwalls the people may have and tips to solve those issues.
Thank you.
Signature · 2 lines
Hideaki YOSHIFUJI @ USAGI Project <[email protected]> GPG FP: 9022 65EB 1ECF 3AD1 0BDF 80D8 4807 F894 E062 0EEA
Re: (usagi-users 03222) Re: support of IPv6 by NFS
In reply to Quantum ScientificQuoted 8 lines · Quantum Scientific
And 80% of potential users are behind a cable/DSL 4 NATting router. There is no clarity that it is possible overcome this by either setting to DMZ, or hoping your cablemodem passes protos 41, 50 & 51. Even some tunnel operators do not know this, so I had to figure it out myself. There is no Linux 6to4 UDP tunnelling app, but there should be, because this is such a common problem. (As I understand, Teredo is Winduhs-only, and is not supported by most tunnel operators)
There is at least one Teredo client for Linux :
http://www.simphalempin.com/dev/miredo/
Alternatively, TSP tunneling might also work through NAT devices.
Signature · 1 line
Rémi Denis-Courmont
Quoted 3 lines · Quantum Scientific
Also one must become an ip6tables expert in order to have a reasonably secure firewall, because ip6tables and 6tables are dead, and Shorewall does not support IPV6 security for some reason. Another deterrant.
I have to oppose that statement. ip6tables is not dead, it's alive. We're even at the brink of submitting nf_conntrack, the new connection tracking engine that covers ipv4 and ipv6, to the mainline kernel.
I'm running a number of ipv6 packet filters, and as of now we are not aware of any known issues or bugs in the current ip6tables code base.
Signature · 4 lines
- Harald Welte <[email protected]> http://gnumonks.org/ ============================================================================ "Privacy in residential applications is a desirable marketing option." (ETSI EN 300 175-7 Ch. A6)
In article <[email protected]> (at Tue, 01 Mar 2005 11:10:21 +0100), Gilles Quillard <[email protected]> says:
Quoted 4 lines
The problem is not specific to NFS, any networking application written using IPv6 mechanisms for both IPv4 and IPv6 addresses (
AF_INET6socket opened, IPv4 addresses mapped) couldn't work without a kernel built with IPv6.
:
Quoted 6 lines
Are the final users really against the use of kernels built with IPv6 ? What is preconized on Linux for the support of IPv6 ? The solution described above or the cohabitation of the two modes (struct sockaddr or
sockaddr_storageused to contain either structsockaddr_inor structsockaddr_in6) with specific processing according to the family of the addresses ?
You cannot assume whether the user enables IPv6 or not, and you cannot assume s/he has connectivity to global Internet, in most cases.
So, you likely need to try both IPv6 and IPv4. Getaddrinfo() / getnameinfo(), or "protocol independent progarmming," are your friend.
--yoshfuji