Skip to content

server.listen's bind argument does not accept [::] for ipv6 #54441

Description

@ThisIsMissEm

Version

22, 20

Platform

Darwin the-beast-kitten.local 23.6.0 Darwin Kernel Version 23.6.0: Mon Jul 29 21:14:46 PDT 2024; root:xnu-10063.141.2~1/RELEASE_ARM64_T6031 arm64

Subsystem

net or dns

What steps will reproduce the bug?

require('http').createServer((req, res) => {
  res.statusCode = 200;
  res.end("OK");
}).listen(4001, '[::]');

Outputs:

node:events:497
      throw er; // Unhandled 'error' event
      ^

Error: getaddrinfo ENOTFOUND [::]
    at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:109:26)
Emitted 'error' event on Server instance at:
    at GetAddrInfoReqWrap.doListen [as callback] (node:net:2132:12)
    at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:109:17) {
  errno: -3008,
  code: 'ENOTFOUND',
  syscall: 'getaddrinfo',
  hostname: '[::]'
}

Node.js v20.16.0

(also reproducible on 22.2.0, same error)

How often does it reproduce? Is there a required condition?

Always

What is the expected behavior? Why is that the expected behavior?

Should listen on all interfaces for ipv6.

What do you see instead?

Server fails to start with an error.

Additional information

Using bind of :: works, so the following succeeds:

require('http').createServer((req, res) => {
  res.statusCode = 200;
  res.end("OK");
}).listen(4001, '::');

This causes compatibility issues if you've configuration that has a BIND parameter that needs to be passed to Node.js and another process, such as the Ruby on Rails built-in server, which doesn't accept bind of :: but does accept [::]

Activity

  1. ThisIsMissEm commented on Aug 18, 2024

    @ThisIsMissEm
    Author

    This was discovered via mastodon/mastodon#31395

  2. added
    httpIssues and PRs related to the http subsystem.
    on Aug 18, 2024
  3. ThisIsMissEm commented on Aug 18, 2024

    @ThisIsMissEm
    Author

    Interestingly, net.isIPv6('[::]') also returns false, when I'd expect it to return true, I think?

  4. ThisIsMissEm commented on Aug 18, 2024

    @ThisIsMissEm
    Author

    @redyetidev I believe this affects all net#listen calls, not just http#listen

  5. added
    netIssues and PRs related to the net subsystem.
    on Aug 18, 2024
  6. avivkeller commented on Aug 18, 2024

    @avivkeller
    Member

    @nodejs/net + @nodejs/dns

  7. avivkeller commented on Aug 18, 2024

    @avivkeller
    Member
    require('http').createServer((req, res) => {
      res.statusCode = 200;
      res.end("OK");
    }).listen(4001, '[::]');
    $ node repro.js                  
    node:events:498
          throw er; // Unhandled 'error' event
          ^
    
    Error: getaddrinfo ENOTFOUND [::]
        at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:109:26)
    Emitted 'error' event on Server instance at:
        at GetAddrInfoReqWrap.doListen [as callback] (node:net:2130:12)
        at GetAddrInfoReqWrap.onlookup [as oncomplete] (node:dns:109:17) {
      errno: -3008,
      code: 'ENOTFOUND',
      syscall: 'getaddrinfo',
      hostname: '[::]'
    }
    
    Node.js v22.6.0
  8. added
    dnsIssues and PRs related to the dns subsystem.
    and removed
    httpIssues and PRs related to the http subsystem.
    on Aug 18, 2024
  9. bnoordhuis commented on Aug 19, 2024

    @bnoordhuis
    Member

    This has been reported and discussed before but [::] is a phrase GitHub's search function doesn't do well on...

    [::] is URL syntax, whereas node's dns, net and http modules expect plain addresses and host names.

    I think it should be possible to accept and ignore the brackets without introducing ambiguities. It is however definitely a change in behavior and therefore may have backwards compatibility and/or security implications for downstream users.

  10. ThisIsMissEm commented on Aug 19, 2024

    @ThisIsMissEm
    Author

    Yeah, it could also be ruby/rails which is wrong in using url syntax in their BIND environment variable.

    i think automatically stripping brackets with a warning might be okay?

  11. mcollina commented on Aug 20, 2024

    @mcollina
    SponsorMember

    I would not add any specific stripping to this; these kinds of things often end up as attack surfaces for vulnerability hunters. The loopback interface is ::, not [::], use the right one.

  12. ThisIsMissEm commented on Aug 20, 2024

    @ThisIsMissEm
    Author

    Would it be possible to throw a better error here? e.g., an invalid argument error that explains :: vs [::] if the input argument for bind starts with [ ?

  13. mcollina commented on Aug 20, 2024

    @mcollina
    SponsorMember

    I think that would be a very good idea.

  14. 2 remaining items

  15. jazelly commented on Aug 28, 2024

    @jazelly
    Member

    My attempt to add a general regex check at a higher level is not feasible. The server.listen method doesn't perform validation on the host parameter; instead, it simply passes it down to the uv layer for lookup, which is why we're encountering this error. The method relies on the lookup process to reject invalid host values. Currently, it accepts almost any character, and there isn't a straightforward way to implement a high-level guard to cover what is valid domain name and what is not.

    For this specific issue, I think it's kind of like a matter of deciding whether/how we want to enforce host validation within server.listen. I'm hesitant to introduce a check that only targets specific cases, such as [], but also don't have better idea.

    Edit: for reference, ada-url parses [::], which comes back as is, and leave it to uv to lookup

    std::string ascii_hostname = ada::idna::to_ascii(hostname.ToStringView());

  16. github-actions commented on Apr 28, 2026

    @github-actions
    Contributor

    This issue has been marked as stale due to 210 days of inactivity.
    It will be automatically closed in 30 days if no further activity occurs. If this is still relevant, please leave a comment or update it to keep it open.

  17. added
    staleIssues and PRs marked stale due to inactivity and scheduled for automatic closure.
    on Apr 28, 2026
  18. github-actions commented on May 29, 2026

    @github-actions
    Contributor

    This issue has been automatically closed after 30 days of inactivity following its stale status (no activity for a total of 240 days).
    If this is still relevant, feel free to reopen it or leave a comment with additional details so we can continue the discussion.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    dnsIssues and PRs related to the dns subsystem.netIssues and PRs related to the net subsystem.staleIssues and PRs marked stale due to inactivity and scheduled for automatic closure.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions