Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [mosquitto-dev] Help with TLS and X25519 EC for key agreement


The easy answer is that I don't know.

More usefully, I would suggest trying to use the "openssl s_client"
and "openssl s_server" tools to reproduce your setup and see if that
works or gives you any more information on why it doesn't work.



On Tue, 19 Apr 2022 at 16:26, aweatherguy <wsdl@xxxxxxxxxx> wrote:
> I hope this is the proper mailing list for this issue. I looked for a users list but couldn't find one. Let me know if this belongs elsewhere.
> Having a problem trying to use X25519 with mosquitto broker. A client device (Tasmota) has been configured with a certificate that contains an X25519 public key, and an RSA signature. The broker is configured with "cafile" matching the signature on the device certificate.
> A Wireshark capture of the initial TLS connection looks okay with Client and Server Hello packets, but I don't have the expertise to do a deep dive into these packets to say for certain that all is okay. However, I do see the device's Client Hello packet advertising the expected cipher suite (ECDHE_RSA_WITH_AES_128_GCM_SHA256), and the expected public key type (X25519).
> The Mosquitto broker logs a message "New connection from <ip address> on port 8883".
> For some reason the Client-Hello/Server-Hello exchange happens twice, and then Mosquitto prints this error from the OpenSSL library:
>     tls_process_cert_verify: signature for non signing certificate
> My only guess here is that the error happens because X25519 cannot be used for signing -- only key exchange...but the actual signature on the certificate is an RSA signature from a self-signed root CA that is configured into the broker via "cafile".
> Can anyone help me with this? I can provide the broker log text and Wireshark pcap files it that will help any...
> _______________________________________________
> mosquitto-dev mailing list
> mosquitto-dev@xxxxxxxxxxx
> To unsubscribe from this list, visit

Back to the top