write_quota should only be decremented by 1 in datagram mode, otherwise we'll underflow and always allow writes through. This does not cause any existing tests to fail. (It will be useful once the bug in dtls1_do_write is fixed.) Change-Id: I42aa001d7264790a3726269890635f679497fb1c Reviewed-on: https://boringssl-review.googlesource.com/2831 Reviewed-by: Adam Langley <agl@google.com>kris/onging/CECPQ3_patch15