Skip to content
This repository was archived by the owner on Nov 6, 2020. It is now read-only.
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
221 changes: 172 additions & 49 deletions util/network-devp2p/src/discovery.rs
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@

use ethcore_bytes::Bytes;
use std::net::SocketAddr;
use std::collections::{HashSet, HashMap, BTreeMap, VecDeque};
use std::collections::{HashSet, HashMap, VecDeque};
use std::mem;
use std::default::Default;
use std::time::{Duration, Instant, SystemTime, UNIX_EPOCH};
Expand All @@ -34,9 +34,8 @@ use network::IpFilter;

use PROTOCOL_VERSION;

const ADDRESS_BYTES_SIZE: u32 = 32; // Size of address type in bytes.
const ADDRESS_BITS: u32 = 8 * ADDRESS_BYTES_SIZE; // Denoted by n in [Kademlia].
const NODE_BINS: u32 = ADDRESS_BITS - 1; // Size of m_state (excludes root, which is us).
const ADDRESS_BYTES_SIZE: usize = 32; // Size of address type in bytes.
const ADDRESS_BITS: usize = 8 * ADDRESS_BYTES_SIZE; // Denoted by n in [Kademlia].
const DISCOVERY_MAX_STEPS: u16 = 8; // Max iterations of discovery. (discover)
const BUCKET_SIZE: usize = 16; // Denoted by k in [Kademlia]. Number of nodes stored in each bucket.
const ALPHA: usize = 3; // Denoted by \alpha in [Kademlia]. Number of concurrent FindNode requests.
Expand Down Expand Up @@ -119,7 +118,7 @@ impl Discovery {
discovery_round: 0,
discovery_id: NodeId::new(),
discovery_nodes: HashSet::new(),
node_buckets: (0..NODE_BINS).map(|_| NodeBucket::new()).collect(),
node_buckets: (0..ADDRESS_BITS).map(|_| NodeBucket::new()).collect(),
udp_socket: socket,
send_queue: VecDeque::new(),
check_timestamps: true,
Expand Down Expand Up @@ -155,8 +154,16 @@ impl Discovery {
fn update_node(&mut self, e: NodeEntry) {
trace!(target: "discovery", "Inserting {:?}", &e);
let id_hash = keccak(e.id);
let dist = match Discovery::distance(&self.id_hash, &id_hash) {
Some(dist) => dist,
None => {
warn!(target: "discovery", "Attempted to update own entry: {:?}", e);
return;
}
};

let ping = {
let bucket = &mut self.node_buckets[Discovery::distance(&self.id_hash, &id_hash) as usize];
let bucket = &mut self.node_buckets[dist];
let updated = if let Some(node) = bucket.nodes.iter_mut().find(|n| n.address.id == e.id) {
node.address = e.clone();
node.timeout = None;
Expand All @@ -181,7 +188,15 @@ impl Discovery {

/// Removes the timeout of a given NodeId if it can be found in one of the discovery buckets
fn clear_ping(&mut self, id: &NodeId) {
let bucket = &mut self.node_buckets[Discovery::distance(&self.id_hash, &keccak(id)) as usize];
let dist = match Discovery::distance(&self.id_hash, &keccak(id)) {
Some(dist) => dist,
None => {
warn!(target: "discovery", "Received ping from self");
return
}
};

let bucket = &mut self.node_buckets[dist];
if let Some(node) = bucket.nodes.iter_mut().find(|n| &n.address.id == id) {
node.timeout = None;
}
Expand Down Expand Up @@ -212,7 +227,7 @@ impl Discovery {
trace!(target: "discovery", "Starting round {:?}", self.discovery_round);
let mut tried_count = 0;
{
let nearest = Discovery::nearest_node_entries(&self.discovery_id, &self.node_buckets).into_iter();
let nearest = self.nearest_node_entries(&self.discovery_id).into_iter();
let nearest = nearest.filter(|x| !self.discovery_nodes.contains(&x.id)).take(ALPHA).collect::<Vec<_>>();
for r in nearest {
let rlp = encode_list(&(&[self.discovery_id.clone()][..]));
Expand All @@ -233,17 +248,17 @@ impl Discovery {
self.discovery_round += 1;
}

fn distance(a: &H256, b: &H256) -> u32 {
let d = *a ^ *b;
let mut ret:u32 = 0;
for i in 0..32 {
let mut v: u8 = d[i];
while v != 0 {
v >>= 1;
ret += 1;
/// The base 2 log of the distance between a and b using the XOR metric.
fn distance(a: &H256, b: &H256) -> Option<usize> {
for i in (0..ADDRESS_BYTES_SIZE).rev() {
let byte_index = ADDRESS_BYTES_SIZE - i - 1;
let d: u8 = a[byte_index] ^ b[byte_index];
if d != 0 {
Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we need to deal with timing attack here because I suppose this particular function is a crypto function? In the case where network connection is really stable, would it be possible for an attacker to conduct timing attack and leak some information through PING time? That function calls update_node and uses distance when inserting.

Don't know whether that's actually feasible, but if so, maybe we can also just deal with d != 0 case and return it after 32 loops.

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

To elaborate more on how this may be an issue (while a really subtle one), an attacker can use timing attack through PING on a stable network to guess the bucket index in a range of 8 (because of early return). Then, even if we applied eclipse attack countermeasure 3, the bucket index is still possible to be guessed with an offset of 8.

Public bucket index is an attack factor in the eclipse attack, so it may be possible to then use this information to conduct some sort of similar attacks.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, as long as the distance metric is public, I don't see the risk. If the distance is salted as recommended by the paper, this could be easily defended by sending the pong response packet before calling update_node (which would probably be a good idea regardless). Also, I generally think salting the distance is a bad idea...

let high_bit_index = 7 - d.leading_zeros() as usize;
return Some(i * 8 + high_bit_index);
}
}
ret
None // a and b are equal, so log distance is -inf
}

fn ping(&mut self, node: &NodeEndpoint) {
Expand Down Expand Up @@ -286,36 +301,53 @@ impl Discovery {
Ok(())
}

fn nearest_node_entries(target: &NodeId, buckets: &[NodeBucket]) -> Vec<NodeEntry> {
let mut found: BTreeMap<u32, Vec<&NodeEntry>> = BTreeMap::new();
let mut count = 0;
fn nearest_node_entries(&self, target: &NodeId) -> Vec<NodeEntry> {
let target_hash = keccak(target);
let target_distance = self.id_hash ^ target_hash;

let mut ret = Vec::<NodeEntry>::with_capacity(BUCKET_SIZE);

// Sort bucket entries by distance to target and append to end of result vector.
let append_bucket = |results: &mut Vec<NodeEntry>, bucket: &NodeBucket| -> bool {
let mut sorted_entries: Vec<&BucketEntry> = bucket.nodes.iter().collect();
sorted_entries.sort_unstable_by_key(|entry| entry.id_hash ^ target_hash);

let remaining_capacity = results.capacity() - results.len();
let to_append = if remaining_capacity < sorted_entries.len() {
&sorted_entries[0..remaining_capacity]
} else {
&sorted_entries
};
for entry in to_append.iter() {
results.push(entry.address.clone());
}
results.len() == results.capacity()
};

// Sort nodes by distance to target
for bucket in buckets {
for node in &bucket.nodes {
let distance = Discovery::distance(&target_hash, &node.id_hash);
found.entry(distance).or_insert_with(Vec::new).push(&node.address);
if count == BUCKET_SIZE {
// delete the most distant element
let remove = {
let (key, last) = found.iter_mut().next_back().expect("Last element is always Some when count > 0");
last.pop();
if last.is_empty() { Some(key.clone()) } else { None }
};
if let Some(remove) = remove {
found.remove(&remove);
}
}
else {
count += 1;
// This algorithm leverages the structure of the routing table to efficiently find the
// nearest entries to a target hash. First, we compute the XOR distance from this node to
// the target. On a first pass, we iterate from the MSB of the distance, stopping at any
// buckets where the distance bit is set, and skipping the buckets where it is unset. These
// must be in order the nearest to the target. On a second pass, we traverse from LSB to
// MSB, appending the buckets skipped on the first pass. The reason this works is that all
// entries in bucket i have a common prefix of length exactly 32 - i - 1 with the ID of this
// node.

for i in 0..ADDRESS_BITS {
if ((target_distance[i / 8] << (i % 8)) & 0x80) != 0 {
let bucket = &self.node_buckets[ADDRESS_BITS - i - 1];
if !bucket.nodes.is_empty() && append_bucket(&mut ret, bucket) {
return ret;
}
}
}

let mut ret:Vec<NodeEntry> = Vec::new();
for nodes in found.values() {
ret.extend(nodes.iter().map(|&n| n.clone()));
for i in (0..ADDRESS_BITS).rev() {
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't understand the point of the second pass.
If we didn't find any bucket in the first one, why would we find an entry in this one?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It partitions the buckets by whether the distance bit is set and searches one partition of buckets in the backwards pass and the other partition in the forward pass.

if ((target_distance[i / 8] << (i % 8)) & 0x80) == 0 {
let bucket = &self.node_buckets[ADDRESS_BITS - i - 1];
if !bucket.nodes.is_empty() && append_bucket(&mut ret, bucket) {
return ret;
}
}
}
ret
}
Expand Down Expand Up @@ -453,7 +485,7 @@ impl Discovery {
let target: NodeId = rlp.val_at(0)?;
let timestamp: u64 = rlp.val_at(1)?;
self.check_timestamp(timestamp)?;
let nearest = Discovery::nearest_node_entries(&target, &self.node_buckets);
let nearest = self.nearest_node_entries(&target);
if nearest.is_empty() {
return Ok(None);
}
Expand Down Expand Up @@ -614,7 +646,7 @@ mod tests {
}
discovery2.round();
}
assert_eq!(Discovery::nearest_node_entries(&NodeId::new(), &discovery2.node_buckets).len(), 3)
assert_eq!(discovery2.nearest_node_entries(&NodeId::new()).len(), 3)
}

#[test]
Expand All @@ -625,31 +657,122 @@ mod tests {
for _ in 0..1200 {
discovery.add_node(NodeEntry { id: NodeId::random(), endpoint: ep.clone() });
}
assert!(Discovery::nearest_node_entries(&NodeId::new(), &discovery.node_buckets).len() <= 16);
assert!(discovery.nearest_node_entries(&NodeId::new()).len() <= 16);
let removed = discovery.check_expired(true).len();
assert!(removed > 0);
}

#[test]
fn find_nearest_saturated() {
use super::*;
let mut buckets: Vec<_> = (0..256).map(|_| NodeBucket::new()).collect();

let key = Random.generate().unwrap();
let ep = NodeEndpoint { address: SocketAddr::from_str("127.0.0.1:40447").unwrap(), udp_port: 40447 };
let mut discovery = Discovery::new(&key, ep.address.clone(), ep.clone(), 0, IpFilter::default());

for _ in 0..(16 + 10) {
buckets[0].nodes.push_back(BucketEntry {
discovery.node_buckets[0].nodes.push_back(BucketEntry {
address: NodeEntry { id: NodeId::new(), endpoint: ep.clone() },
timeout: None,
id_hash: keccak(NodeId::new()),
});
}
let nearest = Discovery::nearest_node_entries(&NodeId::new(), &buckets);
let nearest = discovery.nearest_node_entries(&NodeId::new());
assert_eq!(nearest.len(), 16)
}

#[test]
fn routing_table_insertions_lookups() {
use super::*;
let ep = NodeEndpoint { address: SocketAddr::from_str("127.0.0.1:40448").unwrap(), udp_port: 40447 };
let node_ids_hex: [&str; 32] = [
"22536fa57acc12c4993295cbc26fef4550513496712b301ad2283d356c8108521244a362e64e6d907a0d0b4e65526699c5ae3cfebfc680505fe3b33d50672835",
"22c482f42401546f8dd7ed6b1c0cad976da6630730f1116614579ccb084791a528ff2676bfe94434de80e5d7e479f1ea1d7737077da3bd5e69a0f3e5bf596091",
"234c73e3a8f6835a7f9a9d2a896bff4908d66d21d5433a2c37d94f1fa9a6ca17d02388f31013ff87e3ad86506e76bd1006b9cac3815974a2b47c8d4f2124697e",
"2a5aaf4e2046c521e890dc82313c6151a55078f045a7e3d259f168238d029271cdd9a0943468d45c1e36a34a8a6d4de4b0262e48d3c8cfdd4c2aab5df42926b9",
"341d8c94d9670461186cfc1f66d4246cb12384940e9f621ec8d6c216b5d037cde5f7a41b70474ca36ced4a4f2fe91c9dc5a24a128414672661f78e8611d54bfd",
"3d9fd01851f3ae1bfd06b48e89738f29f9a2b4dce3ab7864df4fccca55d1ac88044956ba47d0c4cb44a19924626a3a3aa5a4de8958365cb7385111ce7b929200",
"406d5507a7fbc194a495800ae8cf408093336febc24d03d6c63756f522274ab02146ceb1b0213291a9a1544680503837519f88f1e8677d921de62c82935b4e6c",
"4c537f00805f320616ee49c7bc36e1d7e52a04a782b0cc00fd3d6b77200b027cef5f875ed38f1167fef4b02d7bd49a661812301d9d680bb62297131204c035f9",
"4fc8e3fdbdd7acad82b283ac52c121b805f3b15ffcaa6b2ca67b9e375aa88e978951ffa3d03ee13be99f0ee987db0bbfc6a7ca02b175e9123d79826025b4089d",
"55b5042a6910bc908a0520966e8cbcc92ac299bdb7efbfbcf703df1506fa0f9b09c5eeb930080de848d2864cca71f885942852c51233db0ee46fe0447306d61f",
"5d24f28b350c4c37fc4dad7f418e029992c9e4ac356bb3d9a1356ba1076339863c05044d7ceba233c65779401f8a3b38fe67b6a592c1be4834dc869f7bb932eb",
"5f6edaf2f2ae3003f4b4ff90b8e71a717c832c71a634d96e77fe046f9a88adc8de5718ff3c47659aea4cead5376df5b731e1b6530e6b0999f56ad75d4dabd3f6",
"6214c04211efe91abd23d65e2dc8e711b06d4fb13dcfd65b691dc51f58455b2145f9b38f523b72a45a12705a28d389308a34455720d774c9b805326df42b5a63",
"69df92573ddbbce88b72a930843dbb70728b2a020e0cc4e8ba805dcf7f19297bfc5def4ca447e9e6ec66971be1815b8f49042720431f698b6a87a185d94fa6c8",
"72ffc23de007cf8b6f4a117f7427b532d05861c314344ffa265175f57ee45dae041a710a4dc74124dba1dabdc0f52dfd21e3154d1d4285aab529810c6161d623",
"80b567f279a9512f3a66ebd8f87a93acd4d50bf66f5eff6d04039c1f5838e37021e981539659b33e0644b243fc9671209a80cbef40d1bcf7c7117d353cb45532",
"9009dc9e3bf50595f84271f46d4c7a5ad6971f7d2ffce1905bfc40a407d34fc5e2dcebd92746eadcd2c5fa4d5aaccb0e01b542d506b361851df3f19e6bc629a3",
"95264f56e091efeba911003fd01eeb2c81f6fc4bb7b10c92e4c7bfaf460b7246d232e61ad8a223d74870981a84e15b2d5134c25d931cb860c6912b20a2d3ac01",
"96013a472a9f7ff9c5c76b5ca958f14ee510d826703aa41d4c88eac51d30d14229b9f19f6e0469c37aaa6d2136a978a4aaa38ca766f48e53e569f84e44252962",
"a513c988cf8480ad2992caa64e3fa059ce07efda260dfeefed78e1d41ea3f97844603b8a9737eb633086fd9ac2f201200cb656cda8a91bf6cc500d6039db6f53",
"ab3311f38e3641c8b3b1fd36dd7f94b148166e267258e840d29d1859537c74f202bd3342359b3623f96c23fa662d1b65182a898bf20343744b37cb265182e500",
"ac8f41dbd637891a08c9cf715c23577bdd431ba40231682a5a9ba7fd6cb6d66c04f63d6d65c7d9f8737e641e05fdbeede57138a174f0d55e7835575dd6cddd98",
"accdad251888d53e4e18efee1e0d749d050216b14896efb657e9c7b1b78dab82a5b6fb3234017aa19a2f50475d73960f352d308b2e0e841cbebaf418362a4f21",
"b138622208f74d2b8e8fc10bcd4cf3302685cd77d339280a939474b92be8b93e441c50709e25c82cc88a2a4207e9f2938912d60600226efe322b43c6ef5e7aef",
"b4f64e1fa6a5cd6198b2515bde63fbdabaf7e7a31dbaf5369babbda4b8cd0bf5025ac4b7d2d6e6e3bc76c890df585d28d4815e464c8792ef677df9206864a12b",
"c1136e08a27c93812ae2dd47201d9e81c82d1995001b88dba9eec700e1d3385dfaf7ae834226c3c90a138f1808cd10b5502f49ee774a2bc707f34bd7d160b7bd",
"c203ae9b5d1953b0ac462e66338800ec26982e2af54bd444fc8978973191633d4f483e31b28233c07bb99f34d57c680fa5f8e093e64f13b235005b7ab6e2d594",
"c2e1067c58a9948e773e0a3637d946e26d95762f89ec9d35e2ad84f770309d94168d4e112c78d62b60efc6216bc5d31475f24307b1b8e0fa8dcbb18a10cb85f5",
"d60ecb1a89e0d5aeff14c9a95da9f5492eb15871c53563b86b7c5ddf0da74b4c29e682fdd22aae2290e0b16ef4b6d707ef55396ca98f755c95b689cf65ce5f80",
"df5ad4ea6242929df86f2162d1cc62b0e0a6f0a03428a39dea98f6a689335b5ceaf1f0696c17b717b141aeb45a29108d95c3a7d2d1d0bb3441219504ae672917",
"e1268f5dd9552a11989df9d4953bb388e7466711b2bd9882a3ed4d0767a21f046c53c20f9a18d66bae1d6a5544492857ddecb0b5b4818bd4557be252ddd66c71",
"e626019dc0b50b9e254461f19d29e69a4669c5256134a6352c6c30d3bc55d201a5b43fc2e006556cfaf29765b683e807e03093798942826244e4ee9e47c75d3f",
];
let node_entries = node_ids_hex.iter()
.map(|node_id_hex| NodeId::from_str(node_id_hex).unwrap())
.map(|node_id| NodeEntry { id: node_id, endpoint: ep.clone() })
.collect::<Vec<_>>();

let secret_hex = "6c71d1b8930d29e6371be1081f2c909c64b46440a1716314c3c9df995cb3aed1";
let key = Secret::from_str(secret_hex)
.and_then(|secret| KeyPair::from_secret(secret))
.unwrap();
let mut discovery = Discovery::new(&key, ep.address.clone(), ep.clone(), 0, IpFilter::default());

node_entries.iter().for_each(|entry| discovery.update_node(entry.clone()));

let expected_bucket_sizes = vec![
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 3, 7, 8, 12
];
let actual_bucket_sizes = discovery.node_buckets.iter()
.map(|ref bucket| bucket.nodes.len())
.collect::<Vec<_>>();
assert_eq!(actual_bucket_sizes, expected_bucket_sizes);

for entry in &node_entries {
let nearest = discovery.nearest_node_entries(&entry.id);
assert_eq!(nearest.len(), 16);
assert_eq!(nearest[0].id, entry.id);

let mut expected_ids: Vec<NodeId> = node_entries.iter().map(|entry| entry.id).collect();
expected_ids.sort_unstable_by_key(|id| keccak(id) ^ keccak(entry.id));
expected_ids.resize(BUCKET_SIZE, NodeId::default());

let actual_ids: Vec<NodeId> = nearest.iter().map(|entry| entry.id).collect();
assert_eq!(actual_ids, expected_ids);
}
}

#[test]
fn packets() {
let key = Random.generate().unwrap();
let ep = NodeEndpoint { address: SocketAddr::from_str("127.0.0.1:40447").unwrap(), udp_port: 40447 };
let ep = NodeEndpoint { address: SocketAddr::from_str("127.0.0.1:40449").unwrap(), udp_port: 40449 };
let mut discovery = Discovery::new(&key, ep.address.clone(), ep.clone(), 0, IpFilter::default());
discovery.check_timestamps = false;
let from = SocketAddr::from_str("99.99.99.99:40445").unwrap();
Expand Down