make olm_pickle_* return the lengths of the base64-encoded pickles, rather than
the raw pickle. (From the application's POV, the format of the pickle is
opaque: it doesn't even know that it is base64-encoded. So returning the length
of the raw pickle is particularly unhelpful.)
All the other methods clear their random inputs. This one needs to do the same,
to reduce the risk of the randomness being used elsewhere and leaking key info.
Fixes a segfault when a group message had exactly the length of the mac +
signature.
Also tweak skipping of unknown tags to avoid an extra trip around the loop.
Some clients expect the session id to be globally unique,
so allowing the end devices to pick the session id will cause
problems.
Include the current ratchet index with the initial keys, this decreases
the risk that the client will supply the wrong index causing problems.
Sign the initial keys with the ratchet ed25519 key, this reduces the
risk of a client claiming a session that they didn't create.
Add ed25519 keys to the inbound and outbound sessions, and use them to sign and
verify megolm messages.
We just stuff the ed25519 public key in alongside the megolm session key (and
add a version byte), to save adding more boilerplate to the JS/python/etc
layers.
We were using olm::KEY_LENGTH for everything under the sun which happened to be
32 bytes long, and making a bunch of assumptions in the process. Create a bunch
of new constants (as C #defines rather than C++ consts so that I can use them
in another forthcoming refactor).
Ed25519 private keys, it turns out, have 64 bytes, not 32.
We were previously generating only 32 bytes (which is all that is required to
generate the public key), and then using the public key as the upper 32 bytes
when generating the per-message session key. This meant that everything
appeared to work, but the security of the private key was severely compromised.
By way of fixes:
* Use the correct algorithm for generating the Ed25519 private key, and store
all 512 bits of it.
* Update the account pickle format and refuse to load the old format (since we
should consider it compromised).
* Bump the library version, and add a function to retrieve the library
version, so that applications can verify that they are linked against a
fixed version of the library.
* Remove the curve25519_{sign, verify} functions which were unused and of
dubious quality.
Keeping track of the chain index is a useful thing to do, but is only required
if we've enabled diagnostics. Extend the session pickle format to make a space
for it, so that pickles can be transferred between the logging_enabled branch
and the master branch without loss of information.
Also add some tests for session pickling which explicitly check that we can
unpickle both formats of pickle.
This was introduced when I was experimenting with support for logging progress
in Olm. That is now relegated to the logging_enabled branch, so this should
probably be removed.
This also fixes the incompatibility of session pickles from the current master
branch with those from olm 0.1.0.
Putting the session_id inside the packed message body makes it hard to extract
so that we can decide which session to use. We don't think there is any
advantage to having thes sesion_id protected by the HMACs, so we're going to
move it to the JSON framing.
sha256.c and aes.c contain conflicting declarations, so we need to compile them
as separate units. This requires a bit more Makefile-shuffling; the build
directory now includes 'src' or 'lib' as appropriate, and we just mkdir -p
before each compilation.
There was some slightly overcomplex logic designed to save a couple of hash
operations when R(0) and R(1) were advanced, but the extra code was hard to
understand and didn't save much.
Factor the actual message encoding/decoding and encrypting/decrypting out to
separate functions from the top-level functions which do the base64-wrangling.
This is particularly helpful in the 'outbound' code-path where the offsets
required to allow room to base64-encode make the flow hard to see when it's all inline.
The behaviour when casting from a uint32_t which has overflowed (so has the top
bit set) to int32_t is implementation-defined, so let's avoid relying on it.
We don't need to have all of the top-level pickling functions in olm.cpp;
factor out the utilities to support it to pickle_encoding.cpp (and make sure
that they have plain-C bindings).
