Compatibility issues with other TLS libraries

[SmallTown123]

Version

5.5.1 and below

Description

I think there are two compatibility issues in the TLS 1.2 implementation:

(1) When processing the SessionID field in the ClientHello message, the wolfssl server side seems to respond normally only to SessionID of length 32 at present.

However, RFC5246 specifies the SessionID length as opaque SessionID<0..32>.

There may be compatibility issues with other TLS library clients.

(2) It seems that wolfssl server has support upper limit 150 for Cipher Suites in ClientHello.

In general, this upper number is sufficient, but is there a potential compatibility issue with this implementation?

For example, a client with an individual TLS implementation library will send all the cipher suites it supports, and the number happens to be greater than 150.

[dgarske]

@julek-wolfssl can you look into this report?

[julek-wolfssl]

Hi @SmallTown123,

thank you for your report.

1. I agree that a session ID that is shorter than ID_LEN should not result in an error. In this case we should simply ignore the session ID and continue as if it was not present. I will post a pull request that fixes this.
2. The maximum amount of ciphersuites that wolfSSL is willing to process is governed by the WOLFSSL_MAX_SUITE_SZ compile time definition. By default it is 300 bytes (or 150 ciphersuites) but this can be modified to fit your use case. We feel that 150 ciphersuites covers most use cases but also doesn't use too much memory.

Sincerely
Juliusz

[SmallTown123]

Ok, I understand. Thank you for your reply! @julek-wolfssl

[julek-wolfssl]

@SmallTown123 I have opened #6415 which fixes the session ID incompatibility.

[dgarske]

Hi @SmallTown123 ,

Have you had a chance to try the fixes in PR #6415 to resolve the shorter session ID issue? You feedback would be appreciated.

Thanks,
David Garske, wolfSSL

[julek-wolfssl]

Hi @SmallTown123,

the PR fixing this has been merged.

Juliusz