EdDSA (ed25519) support

[pflanze]

Hi. I'm the author of "better-qmail-remote", a wrapper around qmail-remote that adds DKIM signatures. The project is currently using Mail::DKIM. I've been asked to support EdDSA (ed25519) keys, which I have done, but have realized that Mail::DKIM does not yet support such keys. What are your thoughts on adding such support?

Here is my project branch that supports EdDSA: https://github.com/pflanze/better-qmail-remote/tree/issue1

Here is the original issue on the matter: https://github.com/pflanze/better-qmail-remote/issues/1

Thanks a lot!

[marcbradshaw]

It's something on the roadmap, I don't have a timeline.

[pgnd]

@marcbradshaw is this still 'on the roadmap' ?

currently, for ref,

with DKIM dns zone data for a domain, example.com

    ;   SELECTOR (ed25519): 42c4465ce19e2af6
    dkim-42c4465ce19e2af6-ed25519._domainkey.example.com. 5 IN TXT "v=DKIM1; k=ed25519; p=8fD...=;"

    selector-ed25519._domainkey.example.com. 5 IN CNAME dkim-42c4465ce19e2af6-ed25519._domainkey.example.com.

    ;   SELECTOR (rsa):     42c4465ce19e2af6
    dkim-42c4465ce19e2af6-rsa._domainkey.example.com. 5 IN TXT (
      "v=DKIM1; k=rsa; h=sha256; s=email; t=s;"
      "p=MII..."
        ...
        "0wIDAQAB;"
    )

    selector-rsa._domainkey.example.com. 5 IN CNAME dkim-42c4465ce19e2af6-rsa._domainkey.example.com.

and dkimpy-milter config

    cat dkimpy-milter.conf
        ...
        SigningTable       refile:/opt/etc/dkimpy-milter/signing_table
        KeyTable           /opt/etc/dkimpy-milter/key_table_rsa
        KeyTableEd25519    /opt/etc/dkimpy-milter/key_table_ed25519
        Canonicalization   relaxed/relaxed
        Mode               s
        MacroList          daemon_name/DKIM_ORIGINATING
        SubDomains         No

    grep example.com  /opt/etc/dkimpy-milter/signing_table
        *@example.com    dkim-42c4465ce19e2af6-name.example.com

    grep example.com  /opt/etc/dkimpy-milter/key_table_rsa
        dkim-42c4465ce19e2af6-name.example.com  example.com:dkim-42c4465ce19e2af6-rsa:/opt/etc/dkim/dkim.1662932824.1661990400.1678233600.example.com.rsa.key.pem

    grep example.com  /opt/etc/dkimpy-milter/key_table_ed25519
        dkim-42c4465ce19e2af6-name.example.com  example.com:dkim-42c4465ce19e2af6-ed25519:/opt/etc/dkim/dkim.1662932824.1661990400.1678233600.example.com.ed25519.key

dkimpy-milter signs outbound correctly

    DKIM-Signature: v=1; a=ed25519-sha256; c=relaxed/relaxed;
     d=example.com; i=@example.com; q=dns/txt;
     s=dkim-42c4465ce19e2af6-ed25519; t=1662936130; h=message-id : date :
     from : subject : reply-to : to : content-type :
     content-transfer-encoding : from;
     bh=nos...=;
     b=w9u...==

    DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed;
     d=example.com; i=@example.com; q=dns/txt;
     s=dkim-42c4465ce19e2af6-rsa; t=1662936130; h=message-id : date : from
     : subject : reply-to : to : content-type : content-transfer-encoding :
     from; bh=nos...=;
     b=PMi...==

but inbound authentication check with Authentication-Milter @FastMailreports

    Authentication-Results: mx4.example.net;

        dkim=pass (2048-bit rsa key sha256) header.d=example.com
          header.i=@example.com header.b=PMirdrv4 header.a=rsa-sha256
          header.s=dkim-42c4465ce19e2af6-rsa x-bits=2048;

!!!!    dkim=invalid (unsupported algorithm ed25519-sha256, 0-bit  key)
          header.d=example.com header.i=@example.com
          header.b=w9uQT3c3 header.a=- header.s=dkim-42c4465ce19e2af6-ed25519
          x-bits=0;

[mnalis]

Any progress on this? I'm too getting more and more failures with amavis/spamassassin which trace down to Mail::DKIM leading to reason="invalid (unsupported algorithm ed25519-sha256)"

It would be really nice if it could be supported, especially as RFC 8463 has been mandating for years now that "verifiers MUST implement the Ed25519-SHA256 algorithm".

[pgnd]

@marcbradshaw

Can you please provide some comment/guidance on plans for Ed25519-SHA256 support?

[marcbradshaw]

Hi, I don't have any additional clarity on a timeline or priority for this work. It remains in the backlog but is not scheduled.

[pgnd]

@marcbradshaw Thx for the update. Works as expected so far in local installs. Do you have an estimate of when Ed25519 support will find its way into Fastmail itself?

[marcbradshaw]

@pgnd This is not an official timeline, however I expect verification will be added quite soon. For signing the majority of the work lies in updating the key rotation algorithms to rotate the 2 types of key, and I have no timeline to share for that.