blockworks-foundation
/
solana
mirror of https://github.com/blockworks-foundation/solana.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

rebased

This commit is contained in:
Anatoly Yakovenko 2018-05-12 19:00:22 -07:00
parent a604dcb4c4
commit 2d635386af
10 changed files with 587 additions and 165 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
src/bank.rs
View File

@ -160,6 +160,7 @@ impl Bank {
/// Deduct tokens from the 'from' address the account has sufficient
/// funds and isn't a duplicate.
pub fn process_verified_transaction_debits(&self, tr: &Transaction) -> Result<()> {
info!("Transaction {}", tr.data.tokens);
let bals = self.balances
.read()
.expect("'balances' read lock in process_verified_transaction_debits");

180
src/crdt.rs
View File

@ -83,7 +83,7 @@ impl ReplicatedData {
/// * `listen` - listen for requests and responses
/// No attempt to keep track of timeouts or dropped requests is made, or should be.
pub struct Crdt {
table: HashMap<PublicKey, ReplicatedData>,
pub table: HashMap<PublicKey, ReplicatedData>,
/// Value of my update index when entry in table was updated.
/// Nodes will ask for updates since `update_index`, and this node
/// should respond with all the identities that are greater then the
@ -93,7 +93,7 @@ pub struct Crdt {
/// This Node will ask external nodes for updates since the value in this list
pub remote: HashMap<PublicKey, u64>,
pub update_index: u64,
me: PublicKey,
pub me: PublicKey,
timeout: Duration,
}
// TODO These messages should be signed, and go through the gpu pipeline for spam filtering
@ -106,6 +106,8 @@ enum Protocol {
//TODO might need a since?
/// from id, form's last update index, ReplicatedData
ReceiveUpdates(PublicKey, u64, Vec<ReplicatedData>),
/// ask for a missing index
RequestWindowIndex(ReplicatedData, u64),
}
impl Crdt {
@ -117,7 +119,7 @@ impl Crdt {
remote: HashMap::new(),
me: me.id,
update_index: 1,
timeout: Duration::new(0, 100_000),
timeout: Duration::from_millis(100),
};
g.local.insert(me.id, g.update_index);
g.table.insert(me.id, me);
@ -134,10 +136,10 @@ impl Crdt {
let mut me = self.my_data().clone();
me.current_leader_id = key;
me.version += 1;
self.insert(me);
self.insert(&me);
}
pub fn insert(&mut self, v: ReplicatedData) {
pub fn insert(&mut self, v: &ReplicatedData) {
// TODO check that last_verified types are always increasing
if self.table.get(&v.id).is_none() || (v.version > self.table[&v.id].version) {
//somehow we signed a message for our own identity with a higher version that
@ -169,11 +171,12 @@ impl Crdt {
let (me, table): (ReplicatedData, Vec<ReplicatedData>) = {
// copy to avoid locking durring IO
let robj = obj.read().expect("'obj' read lock in pub fn broadcast");
info!("broadcast table {}", robj.table.len());
let cloned_table: Vec<ReplicatedData> = robj.table.values().cloned().collect();
(robj.table[&robj.me].clone(), cloned_table)
};
let daddr = "0.0.0.0:0".parse().unwrap();
let items: Vec<(usize, &ReplicatedData)> = table
let nodes: Vec<&ReplicatedData> = table
.iter()
.filter(|v| {
if me.id == v.id {
@ -183,15 +186,30 @@ impl Crdt {
//filter nodes that are not listening
false
} else {
info!("broadcast node {}", v.replicate_addr);
true
}
})
.enumerate()
.collect();
let orders: Vec<_> = items.into_iter().cycle().zip(blobs.iter()).collect();
assert!(nodes.len() > 0);
info!("nodes table {}", nodes.len());
info!("blobs table {}", blobs.len());
// enumerate all the blobs, those are the indecies
// transmit them to nodes, starting from a different node
let orders: Vec<_> = blobs
.iter()
.enumerate()
.zip(
nodes
.iter()
.cycle()
.skip((*transmit_index as usize) % nodes.len()),
)
.collect();
info!("orders table {}", orders.len());
let errs: Vec<_> = orders
.into_par_iter()
.map(|((i, v), b)| {
.into_iter()
.map(|((i, b), v)| {
// only leader should be broadcasting
assert!(me.current_leader_id != v.id);
let mut blob = b.write().expect("'b' write lock in pub fn broadcast");
@ -199,13 +217,19 @@ impl Crdt {
blob.set_index(*transmit_index + i as u64)
.expect("set_index in pub fn broadcast");
//TODO profile this, may need multiple sockets for par_iter
s.send_to(&blob.data[..blob.meta.size], &v.replicate_addr)
info!("broadcast {} to {}", blob.meta.size, v.replicate_addr);
let e = s.send_to(&blob.data[..blob.meta.size], &v.replicate_addr);
info!("done broadcast {} to {}", blob.meta.size, v.replicate_addr);
e
})
.collect();
info!("broadcast results {}", errs.len());
for e in errs {
trace!("retransmit result {:?}", e);
match e {
Err(e) => return Err(Error::IO(e)),
Err(e) => {
error!("broadcast result {:?}", e);
return Err(Error::IO(e));
}
_ => (),
}
*transmit_index += 1;
@ -222,7 +246,11 @@ impl Crdt {
let s = obj.read().expect("'obj' read lock in pub fn retransmit");
(s.table[&s.me].clone(), s.table.values().cloned().collect())
};
let rblob = blob.read().expect("'blob' read lock in pub fn retransmit");
blob.write()
.unwrap()
.set_id(me.id)
.expect("set_id in pub fn retransmit");
let rblob = blob.read().unwrap();
let daddr = "0.0.0.0:0".parse().unwrap();
let orders: Vec<_> = table
.iter()
@ -243,15 +271,21 @@ impl Crdt {
let errs: Vec<_> = orders
.par_iter()
.map(|v| {
trace!("retransmit blob to {}", v.replicate_addr);
info!(
"retransmit blob {} to {}",
rblob.get_index().unwrap(),
v.replicate_addr
);
//TODO profile this, may need multiple sockets for par_iter
s.send_to(&rblob.data[..rblob.meta.size], &v.replicate_addr)
})
.collect();
for e in errs {
trace!("retransmit result {:?}", e);
match e {
Err(e) => return Err(Error::IO(e)),
Err(e) => {
info!("retransmit error {:?}", e);
return Err(Error::IO(e));
}
_ => (),
}
}
@ -278,29 +312,32 @@ impl Crdt {
(id, ups, data)
}
pub fn window_index_request(&self, ix: u64) -> Result<(SocketAddr, Vec<u8>)> {
if self.table.len() <= 1 {
return Err(Error::CrdtToSmall);
}
let mut n = (Self::random() as usize) % self.table.len();
while self.table.values().nth(n).unwrap().id == self.me {
n = (Self::random() as usize) % self.table.len();
}
let addr = self.table.values().nth(n).unwrap().gossip_addr.clone();
let req = Protocol::RequestWindowIndex(self.table[&self.me].clone(), ix);
let out = serialize(&req)?;
Ok((addr, out))
}
/// Create a random gossip request
/// # Returns
/// (A,B)
/// * A - Address to send to
/// * B - RequestUpdates protocol message
fn gossip_request(&self) -> Result<(SocketAddr, Protocol)> {
if self.table.len() <= 1 {
return Err(Error::GeneralError);
let options: Vec<_> = self.table.values().filter(|v| v.id != self.me).collect();
if options.len() < 1 {
return Err(Error::CrdtToSmall);
}
let mut n = (Self::random() as usize) % self.table.len();
while self.table
.values()
.nth(n)
.expect("'values().nth(n)' while loop in fn gossip_request")
.id == self.me
{
n = (Self::random() as usize) % self.table.len();
}
let v = self.table
.values()
.nth(n)
.expect("'values().nth(n)' in fn gossip_request")
.clone();
let n = (Self::random() as usize) % options.len();
let v = options[n].clone();
let remote_update_index = *self.remote.get(&v.id).unwrap_or(&0);
let req = Protocol::RequestUpdates(remote_update_index, self.table[&self.me].clone());
Ok((v.gossip_addr, req))
@ -334,7 +371,7 @@ impl Crdt {
// TODO we need to punish/spam resist here
// sig verify the whole update and slash anyone who sends a bad update
for v in data {
self.insert(v.clone());
self.insert(&v);
}
*self.remote.entry(from).or_insert(update_index) = update_index;
}
@ -354,9 +391,40 @@ impl Crdt {
);
})
}
fn run_window_request(
window: &Arc<RwLock<Vec<Option<SharedBlob>>>>,
sock: &UdpSocket,
from: &ReplicatedData,
ix: u64,
) -> Result<()> {
let pos = (ix as usize) % window.read().unwrap().len();
let mut outblob = vec![];
if let &Some(ref blob) = &window.read().unwrap()[pos] {
let rblob = blob.read().unwrap();
let blob_ix = rblob.get_index().expect("run_window_request get_index");
if blob_ix == ix {
// copy to avoid doing IO inside the lock
outblob.extend(&rblob.data[..rblob.meta.size]);
}
} else {
assert!(window.read().unwrap()[pos].is_none());
info!("failed RequestWindowIndex {} {}", ix, from.replicate_addr);
}
if outblob.len() > 0 {
info!(
"responding RequestWindowIndex {} {}",
ix, from.replicate_addr
);
sock.send_to(&outblob, from.replicate_addr)?;
}
Ok(())
}
/// Process messages from the network
fn run_listen(obj: &Arc<RwLock<Self>>, sock: &UdpSocket) -> Result<()> {
fn run_listen(
obj: &Arc<RwLock<Self>>,
window: &Arc<RwLock<Vec<Option<SharedBlob>>>>,
sock: &UdpSocket,
) -> Result<()> {
//TODO cache connections
let mut buf = vec![0u8; 1024 * 64];
let (amt, src) = sock.recv_from(&mut buf)?;
@ -378,7 +446,7 @@ impl Crdt {
//TODO verify reqdata belongs to sender
obj.write()
.expect("'obj' write lock in RequestUpdates")
.insert(reqdata);
.insert(&reqdata);
sock.send_to(&rsp, addr)
.expect("'sock.send_to' in RequestUpdates");
trace!("send_to done!");
@ -389,18 +457,30 @@ impl Crdt {
.expect("'obj' write lock in ReceiveUpdates")
.apply_updates(from, ups, &data);
}
Protocol::RequestWindowIndex(from, ix) => {
//TODO verify from is signed
obj.write().unwrap().insert(&from);
let me = obj.read().unwrap().my_data().clone();
info!(
"received RequestWindowIndex {} {} myaddr {}",
ix, from.replicate_addr, me.replicate_addr
);
assert_ne!(from.replicate_addr, me.replicate_addr);
let _ = Self::run_window_request(window, sock, &from, ix);
}
}
Ok(())
}
pub fn listen(
obj: Arc<RwLock<Self>>,
window: Arc<RwLock<Vec<Option<SharedBlob>>>>,
sock: UdpSocket,
exit: Arc<AtomicBool>,
) -> JoinHandle<()> {
sock.set_read_timeout(Some(Duration::new(2, 0)))
.expect("'sock.set_read_timeout' in crdt.rs");
spawn(move || loop {
let _ = Self::run_listen(&obj, &sock);
let _ = Self::run_listen(&obj, &window, &sock);
if exit.load(Ordering::Relaxed) {
return;
}
@ -458,7 +538,8 @@ mod test {
.map(|_| {
let (crdt, gossip, _, _) = test_node();
let c = Arc::new(RwLock::new(crdt));
let l = Crdt::listen(c.clone(), gossip, exit.clone());
let w = Arc::new(RwLock::new(vec![]));
let l = Crdt::listen(c.clone(), w, gossip, exit.clone());
(c, l)
})
.collect();
@ -514,7 +595,7 @@ mod test {
let yv = listen[y].0.read().unwrap();
let mut d = yv.table[&yv.me].clone();
d.version = 0;
xv.insert(d);
xv.insert(&d);
}
});
}
@ -531,7 +612,7 @@ mod test {
let yv = listen[y].0.read().unwrap();
let mut d = yv.table[&yv.me].clone();
d.version = 0;
xv.insert(d);
xv.insert(&d);
}
});
}
@ -549,10 +630,10 @@ mod test {
let mut crdt = Crdt::new(d.clone());
assert_eq!(crdt.table[&d.id].version, 0);
d.version = 2;
crdt.insert(d.clone());
crdt.insert(&d);
assert_eq!(crdt.table[&d.id].version, 2);
d.version = 1;
crdt.insert(d.clone());
crdt.insert(&d);
assert_eq!(crdt.table[&d.id].version, 2);
}
@ -568,8 +649,8 @@ mod test {
let c1_id = c1.my_data().id;
c1.set_leader(c1_id);
c2.insert(c1.my_data().clone());
c3.insert(c1.my_data().clone());
c2.insert(&c1.my_data());
c3.insert(&c1.my_data());
c2.set_leader(c1.my_data().id);
c3.set_leader(c1.my_data().id);
@ -577,14 +658,17 @@ mod test {
let exit = Arc::new(AtomicBool::new(false));
// Create listen threads
let win1 = Arc::new(RwLock::new(vec![]));
let a1 = Arc::new(RwLock::new(c1));
let t1 = Crdt::listen(a1.clone(), s1, exit.clone());
let t1 = Crdt::listen(a1.clone(), win1, s1, exit.clone());
let a2 = Arc::new(RwLock::new(c2));
let t2 = Crdt::listen(a2.clone(), s2, exit.clone());
let win2 = Arc::new(RwLock::new(vec![]));
let t2 = Crdt::listen(a2.clone(), win2, s2, exit.clone());
let a3 = Arc::new(RwLock::new(c3));
let t3 = Crdt::listen(a3.clone(), s3, exit.clone());
let win3 = Arc::new(RwLock::new(vec![]));
let t3 = Crdt::listen(a3.clone(), win3, s3, exit.clone());
// Create gossip threads
let t1_gossip = Crdt::gossip(a1.clone(), exit.clone());

3
src/entry_writer.rs
View File

@ -62,9 +62,10 @@ impl<'a> EntryWriter<'a> {
) -> Result<()> {
let mut q = VecDeque::new();
let list = self.write_entries(writer, entry_receiver)?;
trace!("New blobs? {}", list.len());
info!("New blobs? {}", list.len());
ledger::process_entry_list_into_blobs(&list, blob_recycler, &mut q);
if !q.is_empty() {
info!("broadcasting {}", q.len());
broadcast.send(q)?;
}
Ok(())

183
src/event_processor.rs Normal file
View File

@ -0,0 +1,183 @@
//! The `event_processor` module implements the accounting stage of the TPU.
use accountant::Accountant;
use entry::Entry;
use event::Event;
use hash::Hash;
use historian::Historian;
use recorder::Signal;
use result::Result;
use std::sync::mpsc::{channel, Sender};
use std::sync::{Arc, Mutex};
pub struct EventProcessor {
pub accountant: Arc<Accountant>,
historian_input: Mutex<Sender<Signal>>,
historian: Mutex<Historian>,
pub start_hash: Hash,
pub ms_per_tick: Option<u64>,
}
impl EventProcessor {
/// Create a new stage of the TPU for event and transaction processing
pub fn new(accountant: Accountant, start_hash: &Hash, ms_per_tick: Option<u64>) -> Self {
let (historian_input, event_receiver) = channel();
let historian = Historian::new(event_receiver, start_hash, ms_per_tick);
EventProcessor {
accountant: Arc::new(accountant),
historian_input: Mutex::new(historian_input),
historian: Mutex::new(historian),
start_hash: *start_hash,
ms_per_tick,
}
}
/// Process the transactions in parallel and then log the successful ones.
pub fn process_events(&self, events: Vec<Event>) -> Result<Entry> {
info!("start sending events {}", events.len());
let historian = self.historian.lock().unwrap();
let results = self.accountant.process_verified_events(events);
let events: Vec<_> = results
.into_iter()
.filter_map(|x| match x {
Ok(e) => Some(e),
Err(e) => {
info!("error {:?}", e);
None
}
})
.collect();
let sender = self.historian_input.lock().unwrap();
info!("sending events {}", events.len());
sender.send(Signal::Events(events))?;
// Wait for the historian to tag our Events with an ID and then register it.
let entry = historian.entry_receiver.recv()?;
info!("entry events {}", entry.events.len());
self.accountant.register_entry_id(&entry.id);
Ok(entry)
}
}
#[cfg(test)]
mod tests {
use accountant::Accountant;
use event::Event;
use event_processor::EventProcessor;
use mint::Mint;
use signature::{KeyPair, KeyPairUtil};
use transaction::Transaction;
#[test]
// TODO: Move this test accounting_stage. Calling process_events() directly
// defeats the purpose of this test.
fn test_accounting_sequential_consistency() {
// In this attack we'll demonstrate that a verifier can interpret the ledger
// differently if either the server doesn't signal the ledger to add an
// Entry OR if the verifier tries to parallelize across multiple Entries.
let mint = Mint::new(2);
let accountant = Accountant::new(&mint);
let event_processor = EventProcessor::new(accountant, &mint.last_id(), None);
// Process a batch that includes a transaction that receives two tokens.
let alice = KeyPair::new();
let tr = Transaction::new(&mint.keypair(), alice.pubkey(), 2, mint.last_id());
let events = vec![Event::Transaction(tr)];
let entry0 = event_processor.process_events(events).unwrap();
// Process a second batch that spends one of those tokens.
let tr = Transaction::new(&alice, mint.pubkey(), 1, mint.last_id());
let events = vec![Event::Transaction(tr)];
let entry1 = event_processor.process_events(events).unwrap();
// Collect the ledger and feed it to a new accountant.
let entries = vec![entry0, entry1];
// Assert the user holds one token, not two. If the server only output one
// entry, then the second transaction will be rejected, because it drives
// the account balance below zero before the credit is added.
let accountant = Accountant::new(&mint);
for entry in entries {
assert!(
accountant
.process_verified_events(entry.events)
.into_iter()
.all(|x| x.is_ok())
);
}
assert_eq!(accountant.get_balance(&alice.pubkey()), Some(1));
}
}
#[cfg(all(feature = "unstable", test))]
mod bench {
extern crate test;
use self::test::Bencher;
use accountant::{Accountant, MAX_ENTRY_IDS};
use bincode::serialize;
use event_processor::*;
use hash::hash;
use mint::Mint;
use rayon::prelude::*;
use signature::{KeyPair, KeyPairUtil};
use std::collections::HashSet;
use std::time::Instant;
use transaction::Transaction;
#[bench]
fn process_events_bench(_bencher: &mut Bencher) {
let mint = Mint::new(100_000_000);
let accountant = Accountant::new(&mint);
// Create transactions between unrelated parties.
let txs = 100_000;
let last_ids: Mutex<HashSet<Hash>> = Mutex::new(HashSet::new());
let transactions: Vec<_> = (0..txs)
.into_par_iter()
.map(|i| {
// Seed the 'to' account and a cell for its signature.
let dummy_id = i % (MAX_ENTRY_IDS as i32);
let last_id = hash(&serialize(&dummy_id).unwrap()); // Semi-unique hash
{
let mut last_ids = last_ids.lock().unwrap();
if !last_ids.contains(&last_id) {
last_ids.insert(last_id);
accountant.register_entry_id(&last_id);
}
}
// Seed the 'from' account.
let rando0 = KeyPair::new();
let tr = Transaction::new(&mint.keypair(), rando0.pubkey(), 1_000, last_id);
accountant.process_verified_transaction(&tr).unwrap();
let rando1 = KeyPair::new();
let tr = Transaction::new(&rando0, rando1.pubkey(), 2, last_id);
accountant.process_verified_transaction(&tr).unwrap();
// Finally, return a transaction that's unique
Transaction::new(&rando0, rando1.pubkey(), 1, last_id)
})
.collect();
let events: Vec<_> = transactions
.into_iter()
.map(|tr| Event::Transaction(tr))
.collect();
let event_processor = EventProcessor::new(accountant, &mint.last_id(), None);
let now = Instant::now();
assert!(event_processor.process_events(events).is_ok());
let duration = now.elapsed();
let sec = duration.as_secs() as f64 + duration.subsec_nanos() as f64 / 1_000_000_000.0;
let tps = txs as f64 / sec;
// Ensure that all transactions were successfully logged.
drop(event_processor.historian_input);
let entries: Vec<Entry> = event_processor.output.lock().unwrap().iter().collect();
assert_eq!(entries.len(), 1);
assert_eq!(entries[0].events.len(), txs as usize);
println!("{} tps", tps);
}
}

4
src/packet.rs
View File

@ -271,7 +271,9 @@ impl Blob {
break;
}
Err(e) => {
info!("recv_from err {:?}", e);
if e.kind() != io::ErrorKind::WouldBlock {
info!("recv_from err {:?}", e);
}
return Err(Error::IO(e));
}
Ok((nrecv, from)) => {

3
src/result.rs
View File

@ -18,7 +18,8 @@ pub enum Error {
BankError(bank::BankError),
SendError,
Services,
GeneralError,
CrdtToSmall,
GenericError,
}
pub type Result<T> = std::result::Result<T, Error>;

4
src/rpu.rs
View File

@ -73,7 +73,8 @@ impl Rpu {
) -> Result<Vec<JoinHandle<()>>> {
let crdt = Arc::new(RwLock::new(Crdt::new(me)));
let t_gossip = Crdt::gossip(crdt.clone(), exit.clone());
let t_listen = Crdt::listen(crdt.clone(), gossip, exit.clone());
let window = streamer::default_window();
let t_listen = Crdt::listen(crdt.clone(), window.clone(), gossip, exit.clone());
// make sure we are on the same interface
let mut local = requests_socket.local_addr()?;
@ -121,6 +122,7 @@ impl Rpu {
broadcast_socket,
exit.clone(),
crdt.clone(),
window,
blob_recycler.clone(),
broadcast_receiver,
);

158
src/streamer.rs
View File

@ -2,16 +2,17 @@
use crdt::Crdt;
#[cfg(feature = "erasure")]
use erasure;
use packet::{Blob, BlobRecycler, PacketRecycler, SharedBlob, SharedPackets, NUM_BLOBS};
use result::Result;
use packet::{Blob, BlobRecycler, PacketRecycler, SharedBlob, SharedPackets};
use result::{Error, Result};
use std::collections::VecDeque;
use std::net::UdpSocket;
use std::net::{SocketAddr, UdpSocket};
use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::mpsc;
use std::sync::{Arc, RwLock};
use std::thread::{spawn, JoinHandle};
use std::time::Duration;
pub const WINDOW_SIZE: usize = 2 * 1024;
pub type PacketReceiver = mpsc::Receiver<SharedPackets>;
pub type PacketSender = mpsc::Sender<SharedPackets>;
pub type BlobSender = mpsc::Sender<VecDeque<SharedBlob>>;
@ -70,7 +71,7 @@ fn recv_send(sock: &UdpSocket, recycler: &BlobRecycler, r: &BlobReceiver) -> Res
pub fn recv_batch(recvr: &PacketReceiver) -> Result<(Vec<SharedPackets>, usize)> {
let timer = Duration::new(1, 0);
let msgs = recvr.recv_timeout(timer)?;
debug!("got msgs");
trace!("got msgs");
let mut len = msgs.read().unwrap().packets.len();
let mut batch = vec![msgs];
while let Ok(more) = recvr.try_recv() {
@ -128,16 +129,58 @@ pub fn blob_receiver(
Ok(t)
}
fn find_next_missing(
locked_window: &Arc<RwLock<Vec<Option<SharedBlob>>>>,
crdt: &Arc<RwLock<Crdt>>,
consumed: &mut usize,
received: &mut usize,
) -> Result<Vec<(SocketAddr, Vec<u8>)>> {
if *received <= *consumed {
return Err(Error::GenericError);
}
let window = locked_window.read().unwrap();
let reqs: Vec<_> = (*consumed..*received)
.filter_map(|pix| {
let i = pix % WINDOW_SIZE;
if let &None = &window[i] {
let val = crdt.read().unwrap().window_index_request(pix as u64);
if let Ok((to, req)) = val {
return Some((to, req));
}
}
None
})
.collect();
Ok(reqs)
}
fn repair_window(
locked_window: &Arc<RwLock<Vec<Option<SharedBlob>>>>,
crdt: &Arc<RwLock<Crdt>>,
consumed: &mut usize,
received: &mut usize,
) -> Result<()> {
let reqs = find_next_missing(locked_window, crdt, consumed, received)?;
info!("repair_window {} {}", *consumed, *received);
let sock = UdpSocket::bind("0.0.0.0:0")?;
for (to, req) in reqs {
//todo cache socket
sock.send_to(&req, to)?;
}
Ok(())
}
fn recv_window(
window: &mut Vec<Option<SharedBlob>>,
locked_window: &Arc<RwLock<Vec<Option<SharedBlob>>>>,
crdt: &Arc<RwLock<Crdt>>,
recycler: &BlobRecycler,
consumed: &mut usize,
received: &mut usize,
r: &BlobReceiver,
s: &BlobSender,
retransmit: &BlobSender,
) -> Result<()> {
let timer = Duration::new(1, 0);
let timer = Duration::from_millis(200);
let mut dq = r.recv_timeout(timer)?;
let leader_id = crdt.read()
.expect("'crdt' read lock in fn recv_window")
@ -188,19 +231,27 @@ fn recv_window(
let b_ = b.clone();
let p = b.write().expect("'b' write lock in fn recv_window");
let pix = p.get_index()? as usize;
let w = pix % NUM_BLOBS;
if pix > *received {
*received = pix;
}
let w = pix % WINDOW_SIZE;
//TODO, after the block are authenticated
//if we get different blocks at the same index
//that is a network failure/attack
trace!("window w: {} size: {}", w, p.meta.size);
{
let mut window = locked_window.write().unwrap();
if window[w].is_none() {
window[w] = Some(b_);
} else {
debug!("duplicate blob at index {:}", w);
} else if let &Some(ref cblob) = &window[w] {
if cblob.read().unwrap().get_index().unwrap() != pix as u64 {
warn!("overrun blob at index {:}", w);
} else {
debug!("duplicate blob at index {:}", w);
}
}
loop {
let k = *consumed % NUM_BLOBS;
let k = *consumed % WINDOW_SIZE;
trace!("k: {} consumed: {}", k, *consumed);
if window[k].is_none() {
break;
@ -211,43 +262,71 @@ fn recv_window(
}
}
}
{
let buf: Vec<_> = locked_window
.read()
.unwrap()
.iter()
.enumerate()
.map(|(i, v)| {
if i == (*consumed % WINDOW_SIZE) {
assert!(v.is_none());
"_"
} else if v.is_none() {
"0"
} else {
"1"
}
})
.collect();
trace!("WINDOW: {}", buf.join(""));
}
trace!("sending contq.len: {}", contq.len());
if !contq.is_empty() {
trace!("sending contq.len: {}", contq.len());
s.send(contq)?;
}
Ok(())
}
pub fn default_window() -> Arc<RwLock<Vec<Option<SharedBlob>>>> {
Arc::new(RwLock::new(vec![None; WINDOW_SIZE]))
}
pub fn window(
exit: Arc<AtomicBool>,
crdt: Arc<RwLock<Crdt>>,
window: Arc<RwLock<Vec<Option<SharedBlob>>>>,
recycler: BlobRecycler,
r: BlobReceiver,
s: BlobSender,
retransmit: BlobSender,
) -> JoinHandle<()> {
spawn(move || {
let mut window = vec![None; NUM_BLOBS];
let mut consumed = 0;
let mut received = 0;
loop {
if exit.load(Ordering::Relaxed) {
break;
}
let _ = recv_window(
&mut window,
&window,
&crdt,
&recycler,
&mut consumed,
&mut received,
&r,
&s,
&retransmit,
);
let _ = repair_window(&window, &crdt, &mut consumed, &mut received);
}
})
}
fn broadcast(
crdt: &Arc<RwLock<Crdt>>,
window: &Arc<RwLock<Vec<Option<SharedBlob>>>>,
recycler: &BlobRecycler,
r: &BlobReceiver,
sock: &UdpSocket,
@ -263,8 +342,31 @@ fn broadcast(
#[cfg(feature = "erasure")]
erasure::generate_codes(blobs);
Crdt::broadcast(crdt, &blobs, &sock, transmit_index)?;
while let Some(b) = blobs.pop() {
recycler.recycle(b);
// keep the cache of blobs that are broadcast
{
let mut win = window.write().unwrap();
for b in &blobs {
let ix = b.read().unwrap().get_index().expect("blob index");
let pos = (ix as usize) % WINDOW_SIZE;
if let Some(x) = &win[pos] {
trace!(
"popped {} at {}",
x.read().unwrap().get_index().unwrap(),
pos
);
recycler.recycle(x.clone());
}
trace!("null {}", pos);
win[pos] = None;
assert!(win[pos].is_none());
}
while let Some(b) = blobs.pop() {
let ix = b.read().unwrap().get_index().expect("blob index");
let pos = (ix as usize) % WINDOW_SIZE;
trace!("caching {} at {}", ix, pos);
assert!(win[pos].is_none());
win[pos] = Some(b);
}
}
Ok(())
}
@ -275,12 +377,14 @@ fn broadcast(
/// * `sock` - Socket to send from.
/// * `exit` - Boolean to signal system exit.
/// * `crdt` - CRDT structure
/// * `window` - Cache of blobs that we have broadcast
/// * `recycler` - Blob recycler.
/// * `r` - Receive channel for blobs to be retransmitted to all the layer 1 nodes.
pub fn broadcaster(
sock: UdpSocket,
exit: Arc<AtomicBool>,
crdt: Arc<RwLock<Crdt>>,
window: Arc<RwLock<Vec<Option<SharedBlob>>>>,
recycler: BlobRecycler,
r: BlobReceiver,
) -> JoinHandle<()> {
@ -290,7 +394,7 @@ pub fn broadcaster(
if exit.load(Ordering::Relaxed) {
break;
}
let _ = broadcast(&crdt, &recycler, &r, &sock, &mut transmit_index);
let _ = broadcast(&crdt, &window, &recycler, &r, &sock, &mut transmit_index);
}
})
}
@ -463,7 +567,7 @@ mod test {
use std::sync::{Arc, RwLock};
use std::thread::sleep;
use std::time::Duration;
use streamer::{BlobReceiver, PacketReceiver};
use streamer::{default_window, BlobReceiver, PacketReceiver};
use streamer::{blob_receiver, receiver, responder, retransmitter, window};
fn get_msgs(r: PacketReceiver, num: &mut usize) {
@ -558,9 +662,11 @@ mod test {
blob_receiver(exit.clone(), resp_recycler.clone(), read, s_reader).unwrap();
let (s_window, r_window) = channel();
let (s_retransmit, r_retransmit) = channel();
let win = default_window();
let t_window = window(
exit.clone(),
subs,
win,
resp_recycler.clone(),
r_reader,
s_window,
@ -628,15 +734,27 @@ mod test {
let (crdt_leader, sock_gossip_leader, _, sock_leader) = test_node();
let (crdt_target, sock_gossip_target, sock_replicate_target, _) = test_node();
let leader_data = crdt_leader.read().unwrap().my_data().clone();
crdt_leader.write().unwrap().insert(leader_data.clone());
crdt_leader.write().unwrap().insert(&leader_data);
crdt_leader.write().unwrap().set_leader(leader_data.id);
let t_crdt_leader_g = Crdt::gossip(crdt_leader.clone(), exit.clone());
let t_crdt_leader_l = Crdt::listen(crdt_leader.clone(), sock_gossip_leader, exit.clone());
let window_leader = Arc::new(RwLock::new(vec![]));
let t_crdt_leader_l = Crdt::listen(
crdt_leader.clone(),
window_leader,
sock_gossip_leader,
exit.clone(),
);
crdt_target.write().unwrap().insert(leader_data.clone());
crdt_target.write().unwrap().insert(&leader_data);
crdt_target.write().unwrap().set_leader(leader_data.id);
let t_crdt_target_g = Crdt::gossip(crdt_target.clone(), exit.clone());
let t_crdt_target_l = Crdt::listen(crdt_target.clone(), sock_gossip_target, exit.clone());
let window_target = Arc::new(RwLock::new(vec![]));
let t_crdt_target_l = Crdt::listen(
crdt_target.clone(),
window_target,
sock_gossip_target,
exit.clone(),
);
//leader retransmitter
let (s_retransmit, r_retransmit) = channel();
let blob_recycler = BlobRecycler::default();

195
src/thin_client.rs
View File

@ -40,9 +40,9 @@ impl ThinClient {
pub fn recv_response(&self) -> io::Result<Response> {
let mut buf = vec![0u8; 1024];
info!("start recv_from");
trace!("start recv_from");
self.requests_socket.recv_from(&mut buf)?;
info!("end recv_from");
trace!("end recv_from");
let resp = deserialize(&buf).expect("deserialize balance in thin_client");
Ok(resp)
}
@ -50,7 +50,7 @@ impl ThinClient {
pub fn process_response(&mut self, resp: Response) {
match resp {
Response::Balance { key, val } => {
info!("Response balance {:?} {:?}", key, val);
trace!("Response balance {:?} {:?}", key, val);
self.balances.insert(key, val);
}
Response::LastId { id } => {
@ -89,7 +89,7 @@ impl ThinClient {
/// until the server sends a response. If the response packet is dropped
/// by the network, this method will hang indefinitely.
pub fn get_balance(&mut self, pubkey: &PublicKey) -> io::Result<i64> {
info!("get_balance");
trace!("get_balance");
let req = Request::GetBalance { key: *pubkey };
let data = serialize(&req).expect("serialize GetBalance in pub fn get_balance");
self.requests_socket
@ -98,7 +98,7 @@ impl ThinClient {
let mut done = false;
while !done {
let resp = self.recv_response()?;
info!("recv_response {:?}", resp);
trace!("recv_response {:?}", resp);
if let &Response::Balance { ref key, .. } = &resp {
done = key == pubkey;
}
@ -165,9 +165,11 @@ mod tests {
use std::io::sink;
use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::{Arc, RwLock};
use std::thread::JoinHandle;
use std::thread::sleep;
use std::time::Duration;
use std::time::Instant;
use streamer::default_window;
use tvu::{self, Tvu};
#[test]
@ -284,81 +286,108 @@ mod tests {
(leader, gossip, serve, replicate, events_socket)
}
fn replicant(
leader: &ReplicatedData,
exit: Arc<AtomicBool>,
alice: &Mint,
threads: &mut Vec<JoinHandle<()>>,
) {
let replicant = test_node();
let replicant_bank = {
let bank = Bank::new(&alice);
Arc::new(Tvu::new(
bank,
alice.last_id(),
None,
))
};
let mut ts = Tvu::serve(
&replicant_bank,
replicant.0.clone(),
replicant.1,
replicant.2,
replicant.3,
leader.clone(),
exit.clone(),
).unwrap();
threads.append(&mut ts);
}
fn converge(
leader: &ReplicatedData,
exit: Arc<AtomicBool>,
num_nodes: usize,
threads: &mut Vec<JoinHandle<()>>,
) -> Vec<SocketAddr> {
//lets spy on the network
let (mut spy, spy_gossip, _, _, _) = test_node();
let daddr = "0.0.0.0:0".parse().unwrap();
let me = spy.id.clone();
spy.replicate_addr = daddr;
spy.serve_addr = daddr;
let mut spy_crdt = Crdt::new(spy);
spy_crdt.insert(&leader);
spy_crdt.set_leader(leader.id);
let spy_ref = Arc::new(RwLock::new(spy_crdt));
let spy_window = default_window();
let t_spy_listen = Crdt::listen(spy_ref.clone(), spy_window, spy_gossip, exit.clone());
let t_spy_gossip = Crdt::gossip(spy_ref.clone(), exit.clone());
//wait for the network to converge
for _ in 0..30 {
let len = spy_ref.read().unwrap().table.values().len();
let mut min = num_nodes as u64;
for u in spy_ref.read().unwrap().remote.values() {
if min > *u {
min = *u;
}
}
info!("length {} {}", len, min);
if num_nodes == len && min >= (num_nodes as u64) {
warn!("converged! {} {}", len, min);
break;
}
sleep(Duration::new(1, 0));
}
threads.push(t_spy_listen);
threads.push(t_spy_gossip);
let v: Vec<SocketAddr> = spy_ref
.read()
.unwrap()
.table
.values()
.into_iter()
.filter(|x| x.id != me)
.map(|x| x.serve_addr)
.collect();
v.clone()
}
#[test]
#[ignore]
fn test_multi_node() {
logger::setup();
info!("test_multi_node");
const N: usize = 5;
trace!("test_multi_accountant_stub");
let leader = test_node();
let replicant = test_node();
let alice = Mint::new(10_000);
let bob_pubkey = KeyPair::new().pubkey();
let exit = Arc::new(AtomicBool::new(false));
let leader_bank = {
let bank = Bank::new(&alice);
Rpu::new(bank, alice.last_id(), Some(Duration::from_millis(30)))
Rpu::new(bank, alice.last_id(), None)
};
let replicant_bank = {
let bank = Bank::new(&alice);
Arc::new(Tvu::new(
bank,
alice.last_id(),
Some(Duration::from_millis(30)),
))
};
let leader_threads = leader_bank
let mut threads = leader_bank
.serve(leader.0.clone(), leader.2, leader.1, exit.clone(), sink())
.unwrap();
let replicant_threads = Tvu::serve(
&replicant_bank,
replicant.0.clone(),
replicant.1,
replicant.2,
replicant.3,
leader.0.clone(),
exit.clone(),
).unwrap();
//lets spy on the network
let (mut spy, spy_gossip, _, _, _) = test_node();
let daddr = "0.0.0.0:0".parse().unwrap();
spy.replicate_addr = daddr;
spy.serve_addr = daddr;
let mut spy_crdt = Crdt::new(spy);
spy_crdt.insert(leader.0.clone());
spy_crdt.set_leader(leader.0.id);
let spy_ref = Arc::new(RwLock::new(spy_crdt));
let t_spy_listen = Crdt::listen(spy_ref.clone(), spy_gossip, exit.clone());
let t_spy_gossip = Crdt::gossip(spy_ref.clone(), exit.clone());
//wait for the network to converge
for _ in 0..20 {
let ix = spy_ref.read().unwrap().update_index;
info!("my update index is {}", ix);
let len = spy_ref.read().unwrap().remote.values().len();
let mut done = false;
info!("remote len {}", len);
if len > 1 && ix > 2 {
done = true;
//check if everyones remote index is greater or equal to ours
let vs: Vec<u64> = spy_ref.read().unwrap().remote.values().cloned().collect();
for t in vs.into_iter() {
info!("remote update index is {} vs {}", t, ix);
if t < 3 {
done = false;
}
}
}
if done == true {
info!("converged!");
break;
}
sleep(Duration::new(1, 0));
for _ in 0..N {
replicant(&leader.0, exit.clone(), &alice, &mut threads);
}
let addrs = converge(&leader.0, exit.clone(), N + 2, &mut threads);
//contains the leader addr as well
assert_eq!(addrs.len(), N + 1);
//verify leader can do transfer
let leader_balance = {
let requests_socket = UdpSocket::bind("0.0.0.0:0").unwrap();
@ -368,47 +397,41 @@ mod tests {
let events_socket = UdpSocket::bind("0.0.0.0:0").unwrap();
let mut client = ThinClient::new(leader.0.serve_addr, requests_socket, events_socket);
info!("getting leader last_id");
trace!("getting leader last_id");
let last_id = client.get_last_id().wait().unwrap();
info!("executing leader transer");
let _sig = client
.transfer(500, &alice.keypair(), bob_pubkey, &last_id)
.unwrap();
info!("getting leader balance");
trace!("getting leader balance");
client.get_balance(&bob_pubkey).unwrap()
};
assert_eq!(leader_balance, 500);
//verify replicant has the same balance
let mut replicant_balance = 0;
for _ in 0..10 {
let mut success = 0usize;
for serve_addr in addrs.iter() {
let requests_socket = UdpSocket::bind("0.0.0.0:0").unwrap();
requests_socket
.set_read_timeout(Some(Duration::new(1, 0)))
.unwrap();
let events_socket = UdpSocket::bind("0.0.0.0:0").unwrap();
let mut client =
ThinClient::new(replicant.0.serve_addr, requests_socket, events_socket);
info!("getting replicant balance");
if let Ok(bal) = client.get_balance(&bob_pubkey) {
replicant_balance = bal;
let mut client = ThinClient::new(*serve_addr, requests_socket, events_socket);
for i in 0..10 {
trace!("getting replicant balance {} {}/10", *serve_addr, i);
if let Ok(bal) = client.get_balance(&bob_pubkey) {
trace!("replicant balance {}", bal);
if bal == leader_balance {
success += 1;
break;
}
}
sleep(Duration::new(1, 0));
}
info!("replicant balance {}", replicant_balance);
if replicant_balance == leader_balance {
break;
}
sleep(Duration::new(1, 0));
}
assert_eq!(replicant_balance, leader_balance);
assert_eq!(success, addrs.len());
exit.store(true, Ordering::Relaxed);
for t in leader_threads {
t.join().unwrap();
}
for t in replicant_threads {
t.join().unwrap();
}
for t in vec![t_spy_listen, t_spy_gossip] {
for t in threads {
t.join().unwrap();
}
}

21
src/tvu.rs
View File

@ -63,9 +63,12 @@ impl Tvu {
) -> Result<()> {
let timer = Duration::new(1, 0);
let blobs = verified_receiver.recv_timeout(timer)?;
trace!("replicating blobs {}", blobs.len());
let entries = ledger::reconstruct_entries_from_blobs(&blobs);
obj.bank.process_verified_entries(entries)?;
let res = obj.bank.process_verified_entries(entries);
if res.is_err() {
error!("process_verified_entries {} {:?}", blobs.len(), res);
}
res?;
for blob in blobs {
blob_recycler.recycle(blob);
}
@ -106,9 +109,10 @@ impl Tvu {
.set_leader(leader.id);
crdt.write()
.expect("'crdt' write lock before insert() in pub fn replicate")
.insert(leader);
.insert(&leader);
let t_gossip = Crdt::gossip(crdt.clone(), exit.clone());
let t_listen = Crdt::listen(crdt.clone(), gossip, exit.clone());
let window = streamer::default_window();
let t_listen = Crdt::listen(crdt.clone(), window.clone(), gossip, exit.clone());
// make sure we are on the same interface
let mut local = replicate.local_addr()?;
@ -140,6 +144,7 @@ impl Tvu {
let t_window = streamer::window(
exit.clone(),
crdt.clone(),
window,
blob_recycler.clone(),
blob_receiver,
window_sender,
@ -273,16 +278,18 @@ mod tests {
let cref_l = Arc::new(RwLock::new(crdt_l));
let t_l_gossip = Crdt::gossip(cref_l.clone(), exit.clone());
let t_l_listen = Crdt::listen(cref_l, leader_gossip, exit.clone());
let window1 = streamer::default_window();
let t_l_listen = Crdt::listen(cref_l, window1, leader_gossip, exit.clone());
//start crdt2
let mut crdt2 = Crdt::new(target2_data.clone());
crdt2.insert(leader_data.clone());
crdt2.insert(&leader_data);
crdt2.set_leader(leader_data.id);
let leader_id = leader_data.id;
let cref2 = Arc::new(RwLock::new(crdt2));
let t2_gossip = Crdt::gossip(cref2.clone(), exit.clone());
let t2_listen = Crdt::listen(cref2, target2_gossip, exit.clone());
let window2 = streamer::default_window();
let t2_listen = Crdt::listen(cref2, window2, target2_gossip, exit.clone());
// setup some blob services to send blobs into the socket
// to simulate the source peer and get blobs out of the socket to