This is a big PR merged many small commits together.
We add the message validation function in libp2p layer.
In the validation function, we check the following conditions
1) the p2p message is a valid consensus message
2) the p2p message sender has a valid public key
3) the sender's public key is in the current committee
4) log the number of invalid/valid messages
After the validation, the valid messages will be forward to the network,
while the invalid messages will be filtered out.
The messages intended for the validator will be handled in the consensus layer.
Signed-off-by: Leo Chen <leo@harmony.one>
* [node] Unused package level big.Int
* [p2p] Throttle receives to a max of 100
* [project] Remove msgq
* [node][p2p] Only allow 100 max message handling at time
* [node][p2p] Throttle max message handler with Weighted Semaphore - init w/ uniform weight
* [node] Use go routine for passing handled block
* [p2p] Leftover commented out
* [node] Add log for when cannot pay semaphore cost
* [node] Bump max message handlers
* [internal] Remove file_embed
* [project] Remove unused file, unused function
* [consensus] Unused values
* [config] Unused field & function
* [internal] Unused address & hash
* [project] More unused tests, values, types
* [project] Move conn logger just to bootnode
* [rpc] Remove dead functions
* [internal] More unused functions
* [project] Remove dead keystore dir package
* [p2p] Need lock to grab all Topics
* [node] Force side effect of topic join, give each topic own weighted semaphore
* [p2p] Remove unused Method
* [p2p] More dead function
* [node] Unused BCPeers field
* [node] Need to exit early if have ping
* [p2p] Remove unnecessary & confusing abstraction
* [p2p] Expose topics
* [project] Dead test
* [node] If fields always same, then just use the constant
* [project] Remove overcomplicated ctxerror
* [project] Remove more dead tests, adjust error replacing ctxerror
* [project] More dead tests, fix travis complaint on Errorf
* [node.sh] Remove is-genesis
this is solve the problem of validators in different network connected
with each others.
* mainet is still using the original harmony prefix to keep backward
compatibility
* pangaea uses "pangaea" as network prefix
* testnet uses "testnet" as network prefix
All nodes in Pangaea and Testnet need to restart to re-connect with each
other. Mainnet nodes have no changes.
Signed-off-by: Leo Chen <leo@harmony.one>
This is because we are introducing more and more parallel P2P concepts
(libp2p peer versus our own peer, for example). Explicit tagging makes
it easier for untrained eyes to understand which side the code refers
to: If it's prefixed with libp2p_, it's libp2p; if not, it's our p2p.
While we are at this, apply the std → 3rd-party → local import grouping,
and also rename multiaddr to ma (the upstream canonical name of
go-multiaddr) in the affected files.
This is a move-only refactor, intended to bubble the Host interface
(which is the main p2p interface) to the top. It also has the effect of
reducing name stutter (host.Host → p2p.Host).
Leo Chen
Edgar Aroutiounian
Jian Zhen
Richard Liu
Eugene Kim