Prevent Requests/Caching of leader schedules for epochs beyond confirmed roots (#4033)

automerge
This commit is contained in:
carllin 2019-04-30 13:23:21 -07:00 committed by Grimes
parent c5f8b4960c
commit 1a259d4a3f
6 changed files with 99 additions and 44 deletions

View File

@ -266,7 +266,7 @@ impl BankingStage {
let poh = poh_recorder.lock().unwrap();
let leader_id = match poh.bank() {
Some(bank) => leader_schedule_cache
.slot_leader_at_else_compute(bank.slot() + 1, &bank)
.slot_leader_at(bank.slot() + 1, Some(&bank))
.unwrap_or_default(),
None => {
if poh.would_be_leader(DEFAULT_TICKS_PER_SLOT) {

View File

@ -131,7 +131,7 @@ pub fn process_blocktree(
vec![(slot, meta, bank, entry_height, last_entry_hash)]
};
let leader_schedule_cache = LeaderScheduleCache::new(*pending_slots[0].2.epoch_schedule());
let leader_schedule_cache = LeaderScheduleCache::new(*pending_slots[0].2.epoch_schedule(), 0);
let mut fork_info = vec![];
let mut last_status_report = Instant::now();
@ -188,6 +188,7 @@ pub fn process_blocktree(
bank.freeze(); // all banks handled by this routine are created from complete slots
if blocktree.is_root(slot) {
leader_schedule_cache.set_root(slot);
bank.squash();
pending_slots.clear();
fork_info.clear();
@ -219,7 +220,7 @@ pub fn process_blocktree(
let next_bank = Arc::new(Bank::new_from_parent(
&bank,
&leader_schedule_cache
.slot_leader_at_else_compute(next_slot, &bank)
.slot_leader_at(next_slot, Some(&bank))
.unwrap(),
next_slot,
));

View File

@ -15,41 +15,34 @@ pub struct LeaderScheduleCache {
// Map from an epoch to a leader schedule for that epoch
pub cached_schedules: RwLock<CachedSchedules>,
epoch_schedule: EpochSchedule,
max_epoch: RwLock<u64>,
}
impl LeaderScheduleCache {
pub fn new_from_bank(bank: &Bank) -> Self {
Self::new(*bank.epoch_schedule())
Self::new(*bank.epoch_schedule(), bank.slot())
}
pub fn new(epoch_schedule: EpochSchedule) -> Self {
Self {
pub fn new(epoch_schedule: EpochSchedule, root: u64) -> Self {
let cache = Self {
cached_schedules: RwLock::new((HashMap::new(), VecDeque::new())),
epoch_schedule,
}
max_epoch: RwLock::new(0),
};
cache.set_root(root);
cache
}
pub fn slot_leader_at(&self, slot: u64) -> Option<Pubkey> {
let (epoch, slot_index) = self.epoch_schedule.get_epoch_and_slot_index(slot);
self.cached_schedules
.read()
.unwrap()
.0
.get(&epoch)
.map(|schedule| schedule[slot_index])
pub fn set_root(&self, root: u64) {
*self.max_epoch.write().unwrap() = self.epoch_schedule.get_stakers_epoch(root);
}
pub fn slot_leader_at_else_compute(&self, slot: u64, bank: &Bank) -> Option<Pubkey> {
let cache_result = self.slot_leader_at(slot);
if cache_result.is_some() {
cache_result
pub fn slot_leader_at(&self, slot: u64, bank: Option<&Bank>) -> Option<Pubkey> {
if let Some(bank) = bank {
self.slot_leader_at_else_compute(slot, bank)
} else {
let (epoch, slot_index) = bank.get_epoch_and_slot_index(slot);
if let Some(epoch_schedule) = self.compute_epoch_schedule(epoch, bank) {
Some(epoch_schedule[slot_index])
} else {
None
}
self.slot_leader_at_no_compute(slot)
}
}
@ -94,6 +87,39 @@ impl LeaderScheduleCache {
None
}
fn slot_leader_at_no_compute(&self, slot: u64) -> Option<Pubkey> {
let (epoch, slot_index) = self.epoch_schedule.get_epoch_and_slot_index(slot);
self.cached_schedules
.read()
.unwrap()
.0
.get(&epoch)
.map(|schedule| schedule[slot_index])
}
fn slot_leader_at_else_compute(&self, slot: u64, bank: &Bank) -> Option<Pubkey> {
let cache_result = self.slot_leader_at_no_compute(slot);
// Forbid asking for slots in an unconfirmed epoch
let bank_epoch = self.epoch_schedule.get_epoch_and_slot_index(slot).0;
if bank_epoch > *self.max_epoch.read().unwrap() {
error!(
"Requested leader in slot: {} of unconfirmed epoch: {}",
slot, bank_epoch
);
return None;
}
if cache_result.is_some() {
cache_result
} else {
let (epoch, slot_index) = bank.get_epoch_and_slot_index(slot);
if let Some(epoch_schedule) = self.compute_epoch_schedule(epoch, bank) {
Some(epoch_schedule[slot_index])
} else {
None
}
}
}
fn get_epoch_schedule_else_compute(
&self,
epoch: u64,
@ -150,19 +176,17 @@ mod tests {
use crate::blocktree::get_tmp_ledger_path;
#[test]
fn test_slot_leader_at_else_compute() {
fn test_slot_leader_at() {
let (genesis_block, _mint_keypair) = GenesisBlock::new(2);
let bank = Bank::new(&genesis_block);
let cache = LeaderScheduleCache::new_from_bank(&bank);
// Nothing in the cache, should return None
assert!(cache.slot_leader_at(bank.slot()).is_none());
assert!(cache.slot_leader_at(bank.slot(), None).is_none());
// Add something to the cache
assert!(cache
.slot_leader_at_else_compute(bank.slot(), &bank)
.is_some());
assert!(cache.slot_leader_at(bank.slot()).is_some());
assert!(cache.slot_leader_at(bank.slot(), Some(&bank)).is_some());
assert!(cache.slot_leader_at(bank.slot(), None).is_some());
assert_eq!(cache.cached_schedules.read().unwrap().0.len(), 1);
}
@ -195,9 +219,9 @@ mod tests {
fn run_thread_race() {
let slots_per_epoch = MINIMUM_SLOT_LENGTH as u64;
let epoch_schedule = EpochSchedule::new(slots_per_epoch, slots_per_epoch / 2, true);
let cache = Arc::new(LeaderScheduleCache::new(epoch_schedule));
let (genesis_block, _mint_keypair) = GenesisBlock::new(2);
let bank = Arc::new(Bank::new(&genesis_block));
let cache = Arc::new(LeaderScheduleCache::new(epoch_schedule, bank.slot()));
let num_threads = 10;
let (threads, senders): (Vec<_>, Vec<_>) = (0..num_threads)
@ -210,7 +234,7 @@ mod tests {
.name("test_thread_race_leader_schedule_cache".to_string())
.spawn(move || {
let _ = receiver.recv();
cache.slot_leader_at_else_compute(bank.slot(), &bank);
cache.slot_leader_at(bank.slot(), Some(&bank));
})
.unwrap(),
sender,
@ -246,9 +270,7 @@ mod tests {
let cache = Arc::new(LeaderScheduleCache::new_from_bank(&bank));
assert_eq!(
cache
.slot_leader_at_else_compute(bank.slot(), &bank)
.unwrap(),
cache.slot_leader_at(bank.slot(), Some(&bank)).unwrap(),
pubkey
);
assert_eq!(cache.next_leader_slot(&pubkey, 0, &bank, None), Some(1));
@ -294,9 +316,7 @@ mod tests {
);
assert_eq!(
cache
.slot_leader_at_else_compute(bank.slot(), &bank)
.unwrap(),
cache.slot_leader_at(bank.slot(), Some(&bank)).unwrap(),
pubkey
);
// Check that the next leader slot after 0 is slot 1
@ -400,4 +420,35 @@ mod tests {
Some(expected_slot),
);
}
#[test]
fn test_schedule_for_unconfirmed_epoch() {
let (genesis_block, _mint_keypair) = GenesisBlock::new(2);
let bank = Arc::new(Bank::new(&genesis_block));
let cache = LeaderScheduleCache::new_from_bank(&bank);
assert_eq!(*cache.max_epoch.read().unwrap(), 1);
// Asking for the leader for the last slot in epoch 1 is ok b/c
// epoch 1 is confirmed
assert_eq!(bank.get_epoch_and_slot_index(95).0, 1);
assert!(cache.slot_leader_at(95, Some(&bank)).is_some());
// Asking for the lader for the first slot in epoch 2 is not ok
// b/c epoch 2 is unconfirmed
assert_eq!(bank.get_epoch_and_slot_index(96).0, 2);
assert!(cache.slot_leader_at(96, Some(&bank)).is_none());
let bank2 = Bank::new_from_parent(&bank, &Pubkey::new_rand(), 95);
assert!(bank2.epoch_vote_accounts(2).is_some());
// Set root for a slot in epoch 1, so that epoch 2 is now confirmed
cache.set_root(95);
assert_eq!(*cache.max_epoch.read().unwrap(), 2);
assert!(cache.slot_leader_at(96, Some(&bank2)).is_some());
assert_eq!(bank2.get_epoch_and_slot_index(223).0, 2);
assert!(cache.slot_leader_at(223, Some(&bank2)).is_some());
assert_eq!(bank2.get_epoch_and_slot_index(224).0, 3);
assert!(cache.slot_leader_at(224, Some(&bank2)).is_none());
}
}

View File

@ -157,6 +157,7 @@ impl ReplayStage {
&vote_account,
&cluster_info,
&blocktree,
&leader_schedule_cache,
)?;
Self::reset_poh_recorder(
@ -235,7 +236,7 @@ impl ReplayStage {
};
assert!(parent.is_frozen());
leader_schedule_cache.slot_leader_at_else_compute(poh_slot, &parent)
leader_schedule_cache.slot_leader_at(poh_slot, Some(&parent))
.map(|next_leader| {
debug!(
"me: {} leader {} at poh slot {}",
@ -308,12 +309,14 @@ impl ReplayStage {
vote_account_pubkey: &Pubkey,
cluster_info: &Arc<RwLock<ClusterInfo>>,
blocktree: &Arc<Blocktree>,
leader_schedule_cache: &Arc<LeaderScheduleCache>,
) -> Result<()>
where
T: 'static + KeypairUtil + Send + Sync,
{
if let Some(new_root) = locktower.record_vote(bank.slot()) {
bank_forks.write().unwrap().set_root(new_root);
leader_schedule_cache.set_root(new_root);
blocktree.set_root(new_root)?;
Self::handle_new_root(&bank_forks, progress);
}
@ -593,7 +596,7 @@ impl ReplayStage {
continue;
}
let leader = leader_schedule_cache
.slot_leader_at_else_compute(child_id, &parent_bank)
.slot_leader_at(child_id, Some(&parent_bank))
.unwrap();
info!("new fork:{} parent:{}", child_id, parent_id);
forks.insert(Bank::new_from_parent(&parent_bank, &leader, child_id));

View File

@ -51,7 +51,7 @@ fn retransmit(
);
for blob in &blobs {
let leader = leader_schedule_cache
.slot_leader_at_else_compute(blob.read().unwrap().slot(), r_bank.as_ref());
.slot_leader_at(blob.read().unwrap().slot(), Some(r_bank.as_ref()));
if blob.read().unwrap().meta.forward {
ClusterInfo::retransmit_to(&cluster_info, &neighbors, blob, leader, sock, true)?;
ClusterInfo::retransmit_to(&cluster_info, &children, blob, leader, sock, false)?;

View File

@ -85,10 +85,10 @@ fn should_retransmit_and_persist(
my_id: &Pubkey,
) -> bool {
let slot_leader_id = match bank {
None => leader_schedule_cache.and_then(|cache| cache.slot_leader_at(blob.slot())),
None => leader_schedule_cache.and_then(|cache| cache.slot_leader_at(blob.slot(), None)),
Some(bank) => match leader_schedule_cache {
None => slot_leader_at(blob.slot(), &bank),
Some(cache) => cache.slot_leader_at_else_compute(blob.slot(), bank),
Some(cache) => cache.slot_leader_at(blob.slot(), Some(bank)),
},
};