Stories
Slash Boxes
Comments

News for nerds, stuff that matters

Slashdot Log In

Log In

Create Account  |  Retrieve Password

Examining ICMP Flaws

Posted by samzenpus on Wed Jul 06, 2005 08:00 PM
from the under-the-microscope dept.
An anonymous reader writes "A recent internet-draft pointed out a number of security flaws in the design of the ICMP protocol. Most open source projects and vendors have addressed the flaws to some level, but this interesting article on KernelTrap examines the true extent of the problem, and how so far only OpenBSD has implemented all possible counter-measures. Theo de Raadt is quoted saying, "here we have a 20 year old protocol, a part of the Internet infrastructure that hasn't been touched in 10 years and we were all sure was right, and now is cast in doubt.""
+ -
story
This discussion has been archived. No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
 Full
 Abbreviated
 Hidden
More
Loading... please wait.
  • ICMP... (Score:5, Funny)

    by Jugalator (259273) on Wednesday July 06 2005, @08:04PM (#12999358) Journal
    In that case... I See More Patches. :-(
  • by DanielMarkham (765899) on Wednesday July 06 2005, @08:07PM (#12999372) Homepage
    The spec calls for a sequence number in the block. Vendors aren't checking it. There are a lot of technical details about how TCP connections can be slowed down by a ICMP attack, but if the vendors checked the sequence number it would make it almost impossible to implement these attacks.
    Researcher found the bugs, tried to work with major vendors. Lawyers got involved, turns out Cisco had been working on a fix for years (so they say). Seems like vendors are more concerned about getting credit than fixing the bugs.
    Reading between the lines, I take it the major vendors have patched their stacks and life is good. Linux implemented all the fixes for all the errors, but addressing the sequence number should be enough for now.
    Makes me wonder: what did the guys writing the code back in the 80s think about the sequence number, anyway? It was obviously there for some reason. I guess because it wasn't part of the "official" spec it was ignored? Shame, that. That was back in the day when people probably didn't think of ICMP being used as a cyber attack vector.

    Smart Identification Of Cost Savings, One Key to Program Management [whattofix.com]
    • by AndrewStephens (815287) on Wednesday July 06 2005, @08:18PM (#12999430) Homepage
      I cannot speak for the orginal implementors of the network stacks, but it was probably for performance (if they thought about it all). Back in the day, when the summers were warm and music was still good, it was not uncommon for the network stack to take a hefty portion of the processor load just processing packets. Shaving off a few cycles per packet by ignoring part of the spec probably looked like a good idea, especially since people were not as concerned with deliberate attacks back then.
    • Help me out here, I seriously want to understand you comment. How can ICMP outlined in RFC792 http://www.ietf.org/rfc/rfc0792.txt?number=792/ [ietf.org] allow for a sequence number when it is designed as an unreliable protocol without a port number or sequence number in the header? I mean unless we rewrite ICMP as a layer four protocol I don't see how we can prevent this?
      • by ph0enix (87965) on Wednesday July 06 2005, @08:58PM (#12999616)
        The ICMP error message that's returned contains a chunk of the packet that the message refers to. The recipient of the error message uses this to demultiplex the ICMP error and apply the error to the correct connection. To do this for TCP, it needs the source and destination IP addresss and ports.

        The TCP sequence number is also there, but the protocol doesn't require it to be checked.

        By an large most protocols are designed to work, not to be secure. This needs to change. The TCPM working group [massiveshill.com] needs to admit that this is a flaw, and change the standard rather than sticking their heads in the sand.
        • I just ran my sniffer and inspected a icmp echo request. I see the ICMP seq#, there is no TCP header on the packet. While I can see this being a potential tool to defend against a barage of ICMP echo replies (or any reply protocol under ICMP) but I still cannot see how it would be a usefull tool in other scenarios, e.g. if I were sending Echo's or Information Requests.

          Regardless I see greater challenges for ICMP with much more complicated answers needed. Until then I still see old school acl's as the bes

          • by ph0enix (87965) on Wednesday July 06 2005, @09:30PM (#12999750)
            I just ran my sniffer and inspected a icmp echo request. I see the ICMP seq#, there is no TCP header on the packet.


            Well duh, ICMP echo request/reply messages are not the ICMP error messages being discussed. From TFA:

            There are three ICMP type 3 'destination unreachable' errors that are defined in RFC 1122 as hard errors. Code 2, 'protocol unreachable', code 3, 'port unreachable', and possibly code 4, 'fragmentation needed and don't fragment bit set' are all hard errors that if received can cause a TCP stack to tear down an existing connection.
    • The spec calls for a sequence number in the block. Vendors aren't checking it.There are a lot of technical details about how TCP connections can be slowed down by a ICMP attack, but if the vendors checked the sequence number it would make it almost impossible to implement these attacks.

      The spec does not require the sequence number be checked (or even mention it), but some OSs check it anyways. Even with the sequence number check, the attack is still not impossible, now just on par with the TCP reset [kerneltrap.org]

      • by JohnQPublic (158027) on Thursday July 07 2005, @07:36AM (#13001805)
        The whole idea behind the RFC system was that the documents, once published, are unchanging. Just like every standard ANSI and ISO publish. There can be and often are documents that revise the protocol, but that's the nature of the standards game and even the ISO does it. Despite common references to ISO standards by number, their proper names are ISO nnnnn:yyyy. For example, SGML is ISO 8879:1986, not ISO 8879, and it has been updated three times since it was issued, by ISO 8879:1986/Cor 1:1996, ISO 8879:1986/Cor 2:1999 and ISO 8879:1986/Amd 1:1988.

        And "back in the day", if you didn't implement part of an RFC for a protocol you implemented, you got lambasted for it. Search around the net for the early 1908s discussions of the TCP Bakeoffs if you want to see how serious we were about it.
  • This is ridiculous! (Score:5, Interesting)

    by garcia (6573) * on Wednesday July 06 2005, @08:10PM (#12999387) Homepage
    While the patent issue was happening with Cisco, CERT/CC created a mailing list to allow vendors to communicate amongst themselves about the newly discovered vulnerability. "They blamed me for submitting my work," Fernando said in exasperation. "One of Cisco's managers of PSIRT said I was cooperating with terrorists, because a terrorist could have gotten the information in the paper I wrote!"

    Of course. We know of problems but we are going to go the Security by Obscurity route and then when the cover is blown we'll claim they are supporting terrorism instead of admitting that we were wrong!

    If the terrorism route doesn't work we always have the patent on the issue to sue him with!

    Way to fucking go, thanks Cisco!
    • by jrockway (229604) <jon-nospam@jrock.us> on Wednesday July 06 2005, @09:02PM (#12999639) Homepage Journal
      Wow. It is about time to stop buying Cisco products. Their idea of security is calling people who help them make it better "terrorists". No fucking thanks.
      • by dcam (615646) <.moc.tpecnocrebu. .ta. .divad.> on Thursday July 07 2005, @02:37AM (#13000900) Homepage
        That is only half of it, read the full article for the way cisco behaved:

        He continued to reply thoroughly to all their questions, until two months later when he received an email from Cisco's lawyer claiming that Cisco held a patent on his work. He asked their lawyer for specifics, but they refused to reveal any details. ...

        Fernando went on to point out that from his experience vendors seem to be more concerned about who gets credit for finding a flaw, rather than about actually fixing it. Fernando explained, "Cisco was worried about not giving me credit because they claimed to have been working on the problem for four years. They offered to set up a meeting with some people of Cisco Argentina to show me documentation that would prove they had been working on the Path MTU Discovery attack for more than a year. It didn't happen. ...

        One week prior to the eventual discloser, Fernando received a call from the CTO of Cisco Argentina who asked him for a copy of his resume. "He said he wanted to have a meeting with me, telling me they might have a job for me," Fernando shrugged. "The meeting was delayed a few times, then I never heard from him again. I wouldn't have thought much of it, but I mentioned it to other people and it turns out they'd had similar experiences. It seems this is a common practice for Cisco to offer someone work in the hopes you'll not talk to the media when the security issues are disclosed."


        Way to go Cisco!
  • by mveloso (325617) on Wednesday July 06 2005, @08:41PM (#12999548)
    I think you could send a redirect via ICMP too, to generate your own man-in-the-middle attack. It's been too long since I've read the RFC.

    DHCP has a whole bunch of issues too. For example, what if a DHCP client gets a DHCPACK from some machine? I'll bet that it would just reconfigure itself using the information in the packet. Bam, you've got a man-in-the-middle attack.

  • by OverCode@work (196386) <overcode AT gmail DOT com> on Wednesday July 06 2005, @08:51PM (#12999593) Homepage
    Often when Internet providers disable your cable/DSL/LAN connection for security or billing reasons, they just block TCP and UDP but leave ICMP available. I've observed Georgia Tech's ResNet to do this, and reportedly Adelphia's cable ISP does the same. You can ping to your heart's content, but can't send data.

    Except that you can.

    A ping packet (ICMP echo request) can have a completely arbitrary payload. You can put any data you want there. You could even tunnel IP inside it. You would have to have to have a friendly server on the outside to receive these packets and forward the contents, but that's easily done.

    This trick might also be useful for tunnelling past content filters. I don't think any of them scan ICMP packets.

    I'm writing a simple userspace IP stack (gets packets from the tun/tap interface), and I intend to try this out once it's a bit more mature.

    -John
    • by isometrick (817436) on Wednesday July 06 2005, @09:16PM (#12999690)
      Save yourself some trouble, check out PingTunnel [cs.uit.no]. :)
    • DON'T DO IT! (Score:4, Informative)

      by DigiShaman (671371) on Thursday July 07 2005, @01:38AM (#13000743) Homepage
      FYI, I'm a Time Warner employee in Austin, TX.

      When we disable a modem for non-payment or virus/spam abuse, we do it through rebooting the modem with a new BIN file. Once done, you will not get an IP address. The modem will still have a 10.net address attached to our network to configure. However, it's not accessible so don't bother wasting your time.

      Regardless if you could get online through a disabled modem, don't do it. Theft of cable service (including internet service through our cable) is federal crime. So don't even THINK about getting crafty with your connection that has been explicitly disabled for non-payment.
      • However,.... (Score:3, Insightful)

        by WindBourne (631190)
        it is not illegal to write the code. It would be useful to see if this hack does work. If so, then it will point out weaknesses in the networks as well as an issue with the protocol.

        Now, with that said, Yeah, it is theft, and it is federal. So, I am sure that s?he will not be using illegally. Or at least, I hope not.
  • by possible (123857) on Wednesday July 06 2005, @08:52PM (#12999595)
    Here are some facts about these vulnerabilities in no particular order.
    1. These are blind exploits, meaning you do NOT have to be a man-in-the-middle.
    2. Sequence number checking is not enough. Therefore Linux has not fully fixed these issues yet. Only OpenBSD has fixed them all, and it must be considered the reference implementation for these fixes. TCP window sizes are fairly large these days. You can EASILY exploit this in a few seconds simply by brute forcing into the window.
    3. This is much worse than the TCP reset attacks we read about. Why? Because using these ICMP exploits, you can stall a connection without the application layer ever receiving notification that something is amiss.
    4. Why does this matter? BGP. How do people secure BGP these days? They filter TCP packets with a firewall. Or they use tunnels. Guess what? That doesn't protect you from these vulnerabilities, because they use ICMP. Guess what? Home users with firewalls aside, most ISPs do not (and cannot, if they expect the Internet to work) filter ICMP.
    5. "Responsible disclosure" is incredibly broken these days and it's getting worse. The vendors have hijacked the process. This is at least the 3rd time Cisco has tried to patent somebody else's security work. NISCC and CERT totally blew it. The IETF blew it AGAIN (remember VRRP?) Gont was asked during his presentation "Knowing what you know, how would you handle the disclosure of these issues if you had to do it over." His answer was, he would just write things up and publish them to Bugtraq without notifying anyone ahead of time. And he's not alone. More and more researchers are anonymously publishing things without notifying the vendors, because they don't want to go through this stuff every time they discover an issue.
    • by graf0z (464763) on Thursday July 07 2005, @04:32AM (#13001161)
      • These are blind exploits, meaning you do NOT have to be a man-in-the-middle.

        If the error receiving system is checking the header of the error generating tcp or udp packet (at least 8 byte have to be contained in the icmp error), the attacker has to guess the source port and - in case if tcp - the tcp sequence number to work blindly.

      • Sequence number checking is not enough. Therefore Linux has not fully fixed these issues yet. Only OpenBSD has fixed them all, and it must be considered the reference implementation for these fixes. TCP window sizes are fairly large these days. You can EASILY exploit this in a few seconds simply by brute forcing into the window.

        Again: you have to guess the source port, too. There are very few tcp protocols with predictable source ports nowadays. So it's not 2^32/windowsize but probably (2^16-1024)*2^32/windowsize. Have fun brute forcing that.

      • This is much worse than the TCP reset attacks we read about. Why? Because using these ICMP exploits, you can stall a connection without the application layer ever receiving notification that something is amiss.

        True: such an attack would feel more like a network problem than like an attack.

      • Why does this matter? BGP. How do people secure BGP these days? They filter TCP packets with a firewall. Or they use tunnels.

        And they secure them by no longer using predictable source ports (many BGP implementations used dest port = source port (179) before).

      This issue has to be considered, but as D. Adams said: Don't panic!

      /graf0z.

  • Typical science (Score:3, Insightful)

    by pstreck (558593) * on Wednesday July 06 2005, @09:21PM (#12999716)
    here we have a 20 year old protocol, a part of the Internet infrastructure that hasn't been touched in 10 years and we were all sure was right, and now is cast in doubt.

    This seems very typical of science in all fields. An unproven theory goes unrebutted for some time untill someone realizes we made a big mistake. The world was once flat remember guys! One thing this should point us to is that no matter how solid something appears, it will always be broken whether it be a theory, a protocol or yes even the fortress known as open bsd *duck* Jokes a side, remember nothing is secure.

    • Actually, first the world was dish shaped, then it was cilindrical, but the God Atlas has been carrying a globe on his shoulder for about 2500 years already. The American view of the world as a flat Pizza, is very modern...
  • by NotWulfen (219204) on Wednesday July 06 2005, @09:27PM (#12999737) Homepage
    IRC networks have been plagued with ICMP unreachables for years

    http://www.rs-labs.com/papers/tacticas/ircutils/pu ke.html [rs-labs.com]

    nothing new to see here, move along.
    • ``nothing new to see here, move along.''

      Well, no. The point is not that there is something new here, but that there isn't. These vulnerabilities are older than the www, but they still haven't been addressed properly. Now, that isn't exactly a new thing either, but it's good to give some attention to it, so that, maybe, they will be addressed soon.

      And there are interesting philosophical implications, too. Fundamentally, these messages can be spoofed to cause problems. You get the bad with the good. We prob
  • by matman (71405) on Wednesday July 06 2005, @09:33PM (#12999770)
    Using spoofed unreach packets to drop TCP sessions has been around for a LONG time - it used to be called a "nuke" (before the Windows OOB attack, "WinNuke", became more widely known). I know that I've heard of the quench spoof attack, but hadn't heard of the path MTU attack, yet. Using ICMP redirect messages to arrange MITM attacks was also an old one, but I don't think that most stacks pay attention to redirect any more.

    Here's a post from 1993, for example:
    http://groups.google.ca/group/comp.protocols.tcp-i p/browse_thread/thread/439b09e36f4738eb/2eacbab1d4 9e966d?q=icmp+unreach+nuke&rnum=3&hl=en#2eacbab1d4 9e966d [google.ca]
    One from 2000:
    http://groups.google.ca/group/sol.lists.freebsd.se curity/browse_thread/thread/37d9a0a870080133/711f4 cc20af1a450?q=icmp+quench+spoof&rnum=1&hl=en#711f4 cc20af1a450 [google.ca]
    One from 2003:
    http://groups.google.ca/group/linux.kernel/browse_ thread/thread/e96bd4e594c808d5/3f66eac2a5aa8665?q= icmp+path+mtu+spoof&rnum=2&hl=en#3f66eac2a5aa8665 [google.ca]

    While these kinds of risks have been known for a long time, there hasn't really been much attempt to mitigate them. Fernando seems to be a little green, initially thinking that he discovered new vulnerabilities, but he's doing the right thing in pressuring for methods of mitigation. It's a hard fight against complacency. Some of the ideas are clever, but it'll take a lot of convincing to change something so low level as ICMP. For how simple ICMP is, it has lots of security issues; it has got to be made more complicated very carefully.
      • Re:Blind attacks (Score:4, Informative)

        by matman (71405) on Thursday July 07 2005, @10:23AM (#13003799)
        First, while source quench is pretty blind, it isn't much of an issue - it's ignored for TCP and I'm not sure that it's used for UDP either (if it is, few important services use UDP over the internet).

        Path MTU spoofing is really just a variation of the ICMP Unreach spoof attack (same ICMP type). Unreach packets need to "quote" the header of the packet that couldn't be delivered - including source (random 1024-65535) and target port numbers - this allows the sending host to know what connection is being affected. In order for the attacked host to accept a spoofed unreach, the unreach needs to quote the right source IP/port and target IP/port. Most of the time, the source IP, and target IP/port are known but the source port could be one in a few thousand. It used to be that, on modem connections, sending thousands of unreach packets took a few minutes, but now it can be done in seconds or less. Now you can even guess the source IP (drop all connections from a network to a server). Thus, now, the attack is essentially (if not technically) blind since you don't have to find the right combo - you just send all combos.
  • Theo (Score:5, Insightful)

    by Anonymous Coward on Wednesday July 06 2005, @10:15PM (#12999993)
    Theo may be a belligerent asshole, no question. But he is a belligerent asshole working for my side.

    I run OpenBSD stable, and some belligerent asshole stays up all night worrying about the best possible response to the latest threats. Sure, I will buy a CD http://openbsd.org/items.html#37 [openbsd.org].

    And Theo, thank you for being a belligerent asshole for the good guys.

  • Eyeroll (Score:3, Interesting)

    by Spazmania (174582) on Wednesday July 06 2005, @11:07PM (#13000238) Homepage
    Thing is, none of those "vulnerabilities" matter.

    Yes, they're real. Yes, you really can use them to bring non-OpenBSD servers to a halt -- for as long as you keep sending packets.

    But think it through: to use those vulnerabilities without getting very busted very fast, you have to have control of a botnet -- a significant anonymous source of packets. If you have control of a botnet, you can DDOS the server to death regardless of whether it has these vulnerabilies -- simply fill the pipe with normal packets.

    And guess what? Getting a hold of a botnet is a lot easier than exploiting these vulnerabilities.

    So, on a practical level, whats the difference between fixing these particular denial-of-service vulnerabilities and ignoring them? Damned if you do, damned if you don't. Better to spend your time worrying about problems whose solution might actually make a difference.
    • Re:Eyeroll (Score:4, Informative)

      by dmiller (581) <{djm} {at} {mindrot.org}> on Thursday July 07 2005, @12:05AM (#13000490) Homepage
      I'm sorry, but you are completely wrong on every one of your points.

      First, you don't need to keep sending packets to continue the DOS. You can degrade connections with the MTU attack with exactly one packet per connection. I.e. you don't need to flood.

      Second, these vulnerabilities don't require a botnet to execute or to hide your identity. By their nature you must spoof the source address of your packets so you would be difficult to find anyway.

      Next, getting hold and maintaining anonymous control of a botnet is a lot harder than running a script-kiddie tool that implements these attacks.

      Finally, fixing these problems makes them go away. They exist independantly of susceptibility to flooding. I still lock my doors despite the fact that a thief can smash my windows.
  • by SpaceLifeForm (228190) on Wednesday July 06 2005, @11:26PM (#13000337)
    With the BS traffic that comes down the pipe when visiting certain websites, could this actually be used to reduce the unwanted traffic?

    I'm thinking, they are attacking me, so I'll attack them back! (Normally, I drop all garbage packets).

  • by grumling (94709) on Wednesday July 06 2005, @11:33PM (#13000368) Homepage
    From the article: On the other hand, if it were true, then it would mean that Cisco takes about two years to address these issues. I would be concerned about this if I were one of their customers.

    This is the biggest problem with large companies. Sure, it is has been pointed out adnausium over the years in various sources -The Mythical Man Month and The Innovators Dilemma being two very good ones. It is too bad that our network is now being ruled by bandits because of it. MS has become everything that it hated about IBM. Cisco has so much hardware out there that IOS has to be tested on everything before a new release. How can it be possible that when FOSS gets updated and corrected quicker? Of course, I work for a large company, and I see how long it takes to get a simple task completed. I'm guessing it has a lot to do with modivation. The open source folks really do believe in their product. For the people working in big companies, it is just a paycheck.

    Of course, there's always this possibility [phule.net]

    • by Anonymous Coward on Wednesday July 06 2005, @08:16PM (#12999423)
      Where's the -1 misguided when you need it?


      ICMP is in the kernel because it's part of TCP/IP, which wouldn't be hard to remove from a Linux kernel.

      Kinda makes all your other services pointless thought, don't it.

      http://en.wikipedia.org/wiki/TCP/IP [wikipedia.org]

        • by burns210 (572621) <maburns@gmail.com> on Thursday July 07 2005, @12:43AM (#13000614) Homepage Journal
          Probably just a typo, but I wanted to clarify a mistype in your post.

          ICMP IS a subset of the Internet Protocol(IP). IP, part of TCP/IP, has an error reporting mechanism for when things get screwed up, it is called ICMP. It doesn't sit on top nor beside IP, it sits inside of it, logically speaking.

          Both ICMP, which is consider its own entity at times, but is a subset of IP as a whole, and IP are at layer 3. The Networking layer.

          TCP and UDP are layer 4.

          microkernels(as mentioned in another post) do exactly this: move as much OUT of the kernel as possible, including the networking (TCP/IP) stack. This isn't a bad idea, necesarily, it gives some advantages that microkernels are all about. If your networking stack gets completely destroyed, it doesn't take down your kernel, etc, etc.

          monolithic kernels, like Linux(and most OSes, since they are 'easier' and more commonly accepted design) put more things inside the kernel like the networking stack. Not everything, but more things than a microkernel.

          All that being said, even in linux, you could still write an userspace TCP/IP stack and use it, AFAIK. Though things like performance would be an issue.
    • by Anonymous Coward
      This should not have been modded up. ICMP is not a service like http or smb, though if your only experience with icmp is ping, it might seem that way. ICMP is a low-level part of the network stack, used internally by TCP/IP, and as such it needs to reside wherever the network stack resides. The entire network stack could be moved to userland, but that is a separate question.
    • We don't put HTTP servers in the kernel.

      I'm sure this will end up redundant but I'm going to comment anyway.

      I give you TUX [wikipedia.org]
    • You know what's ironic? You just described a microkernel...

      The whole reason for monolithic kernels like Linux is performance and simplicity. Having to do a context switch for every single packet gets expensive...

    • You have been modded up, but this is a pretty uninformed argument. ICMP is part of TCP/IP, which you definitely want in the kernel. I guess it would be possible to put it user space, but performance would be terrible. In any case, user space wouldn't protect against what the article is talking about.
      As a matter of fact, people DO put HTTP servers in the kernel. There is more than one Linux in-kernel HTTP server around, and parts of IIS run in kernel on later versions of Windows. If you really, really care a
    • by Anonymous Coward
      Please google TCP/IP, then quiz yourself against your grandmother. If she knows more than you about how networks operate, read more. Once you manage to outwit her you are free to state your apologies here.
    • by A beautiful mind (821714) on Wednesday July 06 2005, @08:32PM (#12999509)
      The scary thing is that the parent is talking about ICMP without actually knowing what it is.

      You see, this is one of the failures of the moderation system: when someone posts something like this, it seems intelligent because it mentions a lot of familiar things, but overally it's not even making sense. The problem is that moderators work like this:

      Argument: check
      Clear line of thinking: check
      Windows comparison: check

      The problem is that this checklist does not include VERIFYING THINGS like what ICMP is. This is how the parent got +5, insightful while it's one of the most misinformed posts i've seen in a while.
    • by Trick (3648) on Wednesday July 06 2005, @08:34PM (#12999522)
      How the heck did this get modded insightful?

      ICMP runs on a different layer than all of the services you mentioned. ICMP is a network layer protocol (like IP and IPv6, also called "layer 3"), and all the protocols you mentioned are application layer (layer 7) protocols. There's no direct comparison to be made to any of the protocols (HTTP, SMB, FTP and NFS) you mentioned.

      If you want to compare having ICMP in the kernel to other sinilar protocols, your best argument (if you can call it that) is that we should have *IP*, another layer 3 protocol, "running as an ordinary user process, not root, and especially not as a kernel process." Obviously, IP *is* included in the kernel, for plenty of good reasons. Comparing ICMP to application-layer protocols like HTTP holds no weight whatsoever, unless you're completely ignorant of network fundamentals.

      How it got modded to +5 Insightful baffles me. I'd have thought this crowd would have a better handle on the basics.

      • by Animats (122034) on Wednesday July 06 2005, @10:12PM (#12999981) Homepage
        Sigh.

        As someone who once implemented ICMP (in 1982, before BSD, even), I should say something.

        First, ICMP is a layer 3 protocol, like TCP and UDP. ICMP is IP protocol #1; TCP is #6 and UDP is #17.

        Second, it's quite feasible to put ICMP in user space. I'm writing this on a QNX system where it's in user space. My 1982 implementation was also in user space, as part of 3COM's UNET. Linux doesn't do it that way, but it's not fundamental that ICMP must be in the kernel. It needs to have a mechanism to pass messages to the other protocols, but that's a local message passing problem. But I'm not going to rehash the ever-growing monolithic kernel issue here.

        Third, we knew about many of those vulnerabilities back in the 1980s, but weren't as concerned about them because the Internet was a DoD/NSF operation. Destination Unreachable and Source Quench messages used to be taken more seriously than they are now. Destination Unreachable told you where the network was down, and Source Quench told you where it was congested, basic network management info back then. Today, nobody does network management that way and many TCP stacks don't do much, if anything, with ICMP information. I used to encourage the use of Source Quench for congestion management (see my RFC on this, from 1984) [ietf.org], but it's far less appropriate today. Back then, we were concerned about packet loss through transmission errors, a frequent occurence with leased-line synchronous modems. So, when a packet was lost, the question was whether you should retransmit rapidly (appropriate for an error) or slowly (appropriate for congestion). Source Quench could disambiguate that situation. Today, it's assumed that packets are lost almost entirely through congestion, since the lower levels are of much better quality than they used to be.

    • Putting things in the kernel rather than user space isn't purely a performance decision, which seems to be the logic behind your comment. If that was the case, there'd be a whole lot of things that should be moved to user space, such as the dumb terminal or low speed serial port handling.

      A lot of things should be put in user space if they can. When those things are put in the kernel, it is usually because the kernel is a better location than userspace. Sometimes it is for performance, sometimes for other

    • Yes... it could be done outside of the kernel. The problem is that it's basically a required component of IP (the two interact closely). IP uses ICMP to report errors, for instance. If ICMP were in a user process, it would have to jump the kernel/user boundary every time, which could become very expensive.

      ICMP really isn't a server. It's a protocol for performing odd tasks that don't quite belong inside IP itself but that are more or less essential for it to function. 'ping' is just one of about 16 message types ICMP supports. The others involve routing, destination unreachable messages, source quench ("slow down!"), etc.

      I think a better design would be to have the entire networking subsystem in userland.

      Incidentally, my current project is writing a tun/tap-based IP stack entirely in C#. It's mostly for fun, but when it's finished it'll be a complete userland networking subsystem. (Current status: decodes and defragments IP packets, starting to implement ICMP.)

      -John