No or distorted sound using SRTP and rtpengine

 

TL;DR;

SDES-pad 

Long version:

Recently had a chance of upgrading rtpengine from version 8 to version 12. 

For those, who don't know what it is, it's just one of the most popular RTP proxy in the opensource ecosystem. And usually coming with Kamailio and OpenSIPS as a default RTP Proxy option nowdays. 

And I must say, it's a really stable software with great backwards compatibility. I did not expected to get any issues on this upgrade, but quite fast, got into situation, when receiving a calls on Linphone (mobile, if it matters to someone), sound get distorted in a case of Android and completely absent in a case of iOS.

First I've blamed a new chiper suites, that had been added and disabled it. (SDES-no-<CHIPER> string in rtpengine_manage(), if someone interested). But that does not help.

In the process of searching I've found this issue on github. And yes, turning on SDES-pad resolved the issue, but I was really interested why this option was introduced in a first place.

So, according to the documentation,

pad

   RFC 4568 (section 6.1) is somewhat ambiguous regarding the base64 encoding format of a=crypto parameters added to an SDP body. The default interpretation is that trailing = characters used for padding should be omitted. With this flag set, these padding characters will be left in place.

Hm, let's read the RFC about it.

   When base64 decoding the key and salt, padding characters (i.e., one or two "=" at the end of the base64-encoded data) are discarded (see [RFC3548] for details).
   Base64 encoding assumes that the base64 encoding input is an integral number of octets.  If a given crypto-suite requires the use of a concatenated key and salt with a length that is not an integral number of octets, said crypto-suite MUST define a padding scheme that results in the base64 input being an integral number of octets.

What I can see here is reference to how base64 is made, just explicit remark, that padding characters are not carrying any data, just.. padding according to the corresponding RFC:

   In some circumstances, the use of padding ("=") in base encoded data is not required nor used.  In the general case, when assumptions on size of transported data cannot be made, padding is required to yield correct decoded data.
   Implementations MUST include appropriate pad characters at the end of encoded data unless the specification referring to this document explicitly states otherwise.

Really, I can't see any way, that padding is "optional" in base64 encoding by default, both RFC's expicitly saying, that padding is needed and this is the case for SDP option a=crypto.

As a takeway for me, "default" options even in poular software not meaning "right".