There were several bugs here:
- We must always return an error when the peer tries to open a stream
from the wrong side.
- We must never return an error when GetOrOpenStream is called for a
stream that was already closed. GetOrOpenStream is called when a frame
on a stream is received, so this happens when we receive retransmissions
of STREAM frames for a closed stream. We only used to do that check for
peer-initiated streams, not for streams we opened ourselves.
This commit splits up handling of the crypto stream and the other streams in the framer, crypto setup, and the packer.
- Crypto stream data is handled separately and should never be sent unencrypted or FW-secure. Fixes#544.
- Non-crypto stream data is only sent with FW encryption on the server and only with non-FW or FW encryption on the client. Fixes#611.
The crypto stream is current excluded from flow control (#657), but that shouldn't be an issue in practice for now.
We used to reject duplicate and packets with packet numbers lower than
the LeastUnacked we received in a STOP_WAITING frame, because we didn't
accept overlapping stream data. For all other frames, duplicates never
were an issue. Now that we accept overlapping stream data, there's no
need to reject those packets, in fact, processing a delayed packet will
be beneficial for performance.
Pass the diversification nonce via a channel instead of setting it
directly. That way there is no need to protect the diversificationNonce
member by a mutex. Also prevents a possible deadlock that occurred when
SetDiversificationNonce was called before maybeUpgradeCrypto returned.
fixes#598
The mockReceivedPacketHandler returned the same ACK frame over and over
again, so that the loop in session.sendPacket() would send packets
containing this packet indefinitely.
The closeCallback was run when a session was closed, i.e. after the run
loop of the session stopped. Instead of explicitely calling this callback
from the session, the caller of session.run() can just execute the code
after session.run() returns.
